ANALYSES OF TAU AND RADIXiN: TWO ACCESSORY
PROTEINS OF THE CYTOSKELETON IN MAMMALIAN
CELLS
BY
MICHAEL D. HENRY
B.S., University of Georgia
1989
Submitted to the Department of Biology in partial fulfillment of the
requirements for the degree of
Doctor of Philosophy in Biology
at the
Massachusetts Institute of Technology
February 1996
0 1995 Massachusetts Institute of Technology
All rights reserved
Signature of Author
Certified by
Tlhesis'Advisor: Protessor .vrank Solomon
Accepted by
Chairman of the~Graduate Committee: ProfessorFrank~Solomon
OF TE CHNOLOGY
NOV 2 2 1995
1
LUBR ARIES
ANALYSES OF TAU AND RADIXJN: TWO ACCESSORY PROTEINS OF
THE CYTOSKELETON IN MAMMALIAN CELLS
by
Michael D. Henry
Submitted to the department of Biology on October 10, 1995 in partial
fulfillment of the requirements for the degree of Doctor of Philosophy in
Biology
ABSTRACT
Tau and radixin are two proteins that have been implicated in organizing
the microtubule- and microfilament-based cytoskeletons respectively. Here
we have examined the structure and function of these proteins in cultured
mammalian cells. To test tau function, we inhibited its expression in
embryonal carcinoma P19 cells with antisense RNA. The results show that
inhibition of tau expression does not inhibit the ability of P19 cells to
undergo morphological differentiation into neurons- a process that is
known to depend on microtubule function.
Radixin is a member of the ezrin-radixin-moesin (ERM) protein family and
is known to localize in cells to cortical structures in which there is a close
apposition between the plasma membrane and underlying microfilamentrich cytoskeleton. To define the regions of the radixin molecule that specify
its subcellular localization, we expressed full-length and truncated
versions of the molecule in cultured mammalian cells and determined
their localization. Exogenous full-length radixin localized in a manner
similar to endogenous ERM proteins. Moreover, expression of full-length
radixin was correlated with the disappearance of endogenous moesin from
cortical structures suggesting that these two ERM proteins compete for
localization in cortical structures. Localization of the full-length molecule
depended on distinct determinants in both the carboxy- and amino-terminal
domains of the protein. High level expression of the carboxy-terminal
domain of radixin had deleterious effects on cells including the induction of
abnormal cortical processes. Neither the full-length molecule nor its
amino-terminal domain had these effects on cells. These results suggested
that the amino-terminal domain of radixin modulated the function of the
carboxy-terminal domain in the context of the full-length molecule. This
hypothesis was tested in vitro. We found that the amino- and carboxyterminal domains interact with one another with high affinity in solution.
This inter-domain interaction could inhibit intermolecular binding of other
proteins from cell extracts. Taken together, these studies have suggested
an overall model for the molecular organization of radixin which might
explain its localization and function in dynamic cortical cytoskeletal
structures involved in cell motility.
Thesis Advisor:
Title:
Dr. Frank Solomon
Professor of Biology
2
ACKNOWLEDGEMENTS
I would like to thank the following people for specific contributions to the
work presented in each chapter:
CHAPTER TWO and APPENDIX TWO: Drs. Jon Dinsmore and Arthur
Lander, Jill Hahn, Karl Yen, and Sean Walsh for experimental guidance
and contributions.
CHAPTER THREE: Dr. Charo Gonzalez Agosti for experimental
contributions. The work presented in this chapter is published: Henry et al.
1995.
CHAPTER FIVE: Most of the experimental work in this chapter was
performed by Dr. Margaret Magendantz. Dr. Arthur Lander also
contributed to the experiments presented in this chapter. My contribution
to this work was in aid of providing an intellectual framework for some of
the experiments. The result of our collaboration is published: Magendantz
et al., 1995.
APPENDIX THREE: Most of the experimental work presented in this
appendix was performed by Nancy-Lorena Torres, an undergraduate
whose work in the lab I had the great privilege of supervising.
I owe a tremendous debt of gratitude to the following people:
To past and present members of the Solomon lab: Julie Archer, Letty Vega,
John Dinsmore, Adam Grancell, Vida Praitis, Bettina Winckler, Adelle
Smith, Jill Hahn, Jim Fleming, Etchell Cordero, Karl Yen, Nancy Torres,
Charo Gonzalez, and especially Margaret Magendantz for copious
instruction, insightful conversation, rewarding collaboration, and, of
course, indecent amounts of chocolate.
To classmates: Mike Brodsky, John Crispino, Juli Klemm, Sumati Murli,
Eric Schmidt, Tracy Smith, and especially Brian and Brenda Kennedy
without whose friendship these years would have lacked a significant
amount of vitality, croquet, and flyfishing.
To the members of my thesis committee: Richard Hynes for guidance and
support in my career both during and after MIT, Tyler Jacks for teaching
me to keep a shrewd Yankee eye toward the bottom line of any avenue of
inquiry, and Arthur Lander for witty banter and for encouragement to be
clever in the design of experiments and parsimonious in the interpretation
of such.
To mi professor6 Frank Solomon: who instilled in me, among other first
principles, an appreciation of the beauty and power of a controlled
experiment, a taste for the kickin' barbeque to be had at J&E's, and most of
all, a sense of the profound humanity of the scientific enterprise.
To Sloane Henry: whose limitless patience, forbearance and love made
these years possible and to whom I dedicate this work.
3
TABLE OF CONTENTS
Title Page
Abstract
Ackowledgements
Table of Contents
List of Figures and Tables
1
2
3
4
7
CHAPTER ONE:
Modulation of cytoskeletal form and function by accessory proteins.
Summary
Cytoskeletal Form and Function in Animal Cells
Molecular Mechanisms for Organization of the Cytoskeleton
Identification and Characterization of Tau and Radixin: Two
Accessory Proteins of the Cytoskeleton in Cultured Mammalian
Cells
10
11
17
25
CHAPTER TWO:
Antisense inhibition of tau protein expression in embryonal
carcinoma P19 cells.
31
32
35
39
58
Summary
Introduction
Materials and Methods
Results
Discussion
CHAPTER THREE:
Molecular dissection of radixin: Distinct and interdependent
functions of the amino- and carboxy-terminal domains.
64
65
68
74
110
Summary
Introduction
Materials and Methods
Results
Discussion
CHAPTER FOUR:
Deletion analysis of radixin's carboxy-terminal domain.
115
116
120
137
Summary
Materials and Methods
Results
Discussion
CHAPTER FIVE:
Inter-domain interactions of radixin in vitro.
4
Summary
Introduction
Materials and Methods
Results
Discussion
141
142
143
146
156
CHAPTER SIX:
Intermolecular interactions of radixin.
Summary
Introduction
Materials and Methods
Results
Discussion
159
160
162
165
177
CHAPTER SEVEN:
A model for the molecular organization of radixin.
Summary
The Model
Unresolved Issues
Testing the Model
183
184
189
193
APPENDIX ONE:
Isolation of a euploid embryonal carcinoma P19 cell line.
Summary
Materials and Methods
Results and Discussion
196
197
198
APPENDIX TWO:
Characterization of a cell-substratum adhesion deficient embryonal
carcinoma P19 cell line.
Summary
Materials and Methods
Results
Discussion
205
206
208
228
APPENDIX THREE:
Expression and localization of HA-radixin constructs in embryonal
carcinoma P19 cells.
Summary
Introduction
Materials and Methods
Results
Discussion
233
234
236
237
244
5
LITERATURE CITED
246
6
LIST OF FIGURES AND TABLES
FIGURES:
1-1
2-1
2-2
2-3
2-4
2-5
2-6
3-1
3-2
3-3
3-4
3-5
3-6
3-7
3-8
3-9
3-10
3-11
3-12
3-13
3-14
3-15
4-1
4-2
4-3
4-4
Protein sequence comparison of ERM proteins and band 4.1 family
members. p. 29
Tau mRNA expression in EC P19 cells. p. 41
Antisense constructs. p. 43
RA differentiation of pGKBA-derived cell lines. p. 47
Tau protein expression in RA-induced pGKBA-derived cell lines.
p. 5 1
RA differentiation of pCXN2-derived cell lines. p. 55
Tau protein expression in RA-induced pCXN2-derived cell lines. p.
57
Localization of endogenous ERM proteins in NIH-3T3 cells. p. 77
Stable expression of HA-radixin constructs in NIH-3T3 cells. p. 79
Localization of full-length HA-radixin polypeptides in NIH-3T3 cell
lines. p. 83
Localization of truncated HA-radixin polypeptides in NIH-3T3 cell
lines. p. 85
Displacement of endogenous moesin from cortical structures by HAradixin polypeptides in NIH-3T3 cells. p. 87
HAC-RAD, but not HAC-RADC, displaces endogenous moesin from
cleavage furrows in NIH-3T3 cells. p. 89
Moesin protein expression in NIH-3T3 cell lines expressing HAradixin proteins. p. 89
Endogenous ERM protein expression and localization in HtTA-1
cells. p. 93
Transient expression of HA-radixin constructs in HtTA-1 cells. p. 95
Localization of HA-radixin polypeptides in transiently transfected
HtTA-1 cells. p. 97
Optical sectioning of HtTA-1 cells transiently expressing HANRADC. p. 99
Localization of HA-radixin in cleavage furrows of transiently
transfected HtTA-1 cells. p. 103
Localization of HA-radixin constructs in the ventral cytoplasm of
transiently transfected HtTA-1 cells. p. 105
Expression of the carboxy-terminal HA-radixin constructs results in
an increased number of multinucleated cells. p. 107
Transient expression of HA-radixin proteins in NIH-3T3 cells. p. 109
Expression of radixin carboxy-terminal domain deletion constructs
in HtTA-1 cells. p. 123
Effects of carboxy-terminal domain deletion constructs on cortical
structures. p. 125
Subcellular localization of radixin carboxy-terminal domain deletion
constructs. p. 129
Radixin carboxy-terminal domain deletion constructs do not displace
moesin from cortical structures. p. 133
7
4-5
5-1
5-2
5-3
6-1
6-2
6-3
6-4
6-5
7-1
Al-1
A1-2
A2-1
A2-2
A2-3
A2-4
A2-5
A2-6
A2-7
A2-8
A3-1
A3-2
The carboxy-terminal domain of moesin induces abnormal cortical
structures in HtTA-1 cells. p. 135
Immunoblot analyses of column eluates. p. 149
Affinity co-electrophoresis of the N- and C- domains of radixin. p. 153
Co-expression of the amino-terminal domain does not suppress the
effects of expression of the carboxy-terminla domain in HtTA-1 cells.
p. 155
Detergent fractionation of HA-radixin polypeptides. p. 167
Cellular fractionation of HA-radixin polypeptides. p. 167
The carboxy-terminal domain of radixin binds F-actin in solution. p.
171
Scatchard analysis of data from C-6 actin binding experiment. p. 173
Co-immunoprecipitation of cellular proteins with radixin and its
domains. p. 175
A model for the molecular organization of radixin. p. 185
Karyotype analysis of embryonal carcinoma P19 lab stock. p. 201
Characterization of P19-A4- a euploid embryonal carcinoma P19 cell
line. p. 203
RA-induced differentiation of P19 cell line TA3A. p. 211
Extended culture of TA3A. p. 213
Loss of cell-substratum adhesion in TA3A cells is a specific effect of
exposure to RA. p. 215
Adherence properties of TA3A cells on a variety of culture substrata.
p. 217
Development of TA3A phenotype after exposure to retinoic acid. p.
221
TA3A phenotype is independent of culture density. p. 223
Expression of neuron-specific markers by retinoic acid-induced
TA3A cells. p. 225
Neurite extension in aggregates. p. 227
Expression of HA-radixin constructs in uninduced and RA-induced
P19 cell lines. p. 239
Co-localization of HA-radixin proteins and ezrin in uninduced P19
cell lines. p. 241
TABLES:
2-1
4-1
Growth of tau antisense lines. p. 49
Effects of radixin carboxy-terminal domain deletion constructs on
cytokinesis. p. 131
8
CHAPTER ONE:
Modulation of cytoskeletal form and function by accessory
proteins.
9
SUMMARY
The importance of the cytoskeleton in cell shape and motility has
been recognized for many years. Recent advances in molecular biology
have begun to define the molecular components of the cytoskeleton, yet a
detailed understanding of how these molecules affect cytoskeletal form and
function remains elusive. One of the central mysteries is how a relatively
simple set of cytoskeletal polymers can assume the varied roles required in
distinct cell types. One working hypothesis is that non-covalent
associations of other proteins with cytoskeletal polymers play crucial roles
in organizing the cytoskeleton for diverse functions. These cytoskeletal
modulating proteins are collectively referred to as accessory proteins. In
this chapter, we will illustrate the range of cytoskeletal diversity, provide
some examples of the roles that accessory proteins play in the cytoskeleton,
and introduce two accessory proteins -tau and radixin- which are the focus
of this work.
10
CYTOSKELETAL FORM AND FUNCTION IN ANIMAL CELLS
Metazoa are characterized by a variety of differentiated cell types.
Often the most readily observable hallmark of the differentiated state is an
asymetric morphology. Because lipid membranes are fluid, other intraand extra-cellular determinants are required for the establishment and
maintenance of these asymetric morphologies. Extraction of membranes
with non-ionic detergents reveals an insoluble matrix that retains the preextracted cell shape (Webster et al., 1978; Ben-Ze'ev et al., 1979). Among the
components of this matrix are the cytoskeletal polymers- microfilaments,
microtubules, and intermediate filaments. The cytoskeleton is not,
however, simply a collection of rigid elements passively filling space within
the amorphous membrane. Indeed, forces generated by motor-driven
sliding of polymers against one another or by polymerization of the
polymers themselves underlie many motile cell behaviors.
Pharmacological agents that specifically disrupt microfilament and
microtubule arrays demonstrate that these two cytoskeletal polymers are
required for cell morphogenesis and motility. Below, we briefly describe the
cytoskeletal polymers and a few of the specialized organelles formed by
these polymers in living cells.
The cytoskeletal polymers- The cytoskeleton is composed of three
distinct fiber systems: microtubules, microfilaments, and intermediate
filaments. Since the focus of this study is on the microtubule and
microfilament components of the cytoskeleton, we have largely restricted
further comment to these two systems. Initially, microtubules and
microfilaments were described microscopically and thereby differentiated
according to the diameter of the fiber. Later, molecular analysis revealed
that each of these fibers is a polymer composed of identical, repeating
protein subunits. The 24 nm-diameter microtubules are polymers
composed of the heterodimeric unit of oc- and B- tubulin (Dustin, 1978). As
the name implies, microtubules are tubes. The heterodimeric tubulin
subunits are arrayed into thirteen longitudinal rows (the protofilaments)
which form the walls of the tube. With few exceptions, this basic structure
of microtubules is conserved in all eukaryotic cells. It is thought that the
tubulin heterodimers join in a head-to-tail fashion within the
11
protofilament. This arrangement confers a structural polarity on the
microtubule lattice which is evident in the properties of microtubule ends.
One end of the microtubule (the plus end) grows faster than the other end
(the minus end) during microtubule polymerization. Minus ends of
microtubules also tend to be associated with proteinaceous structures in
cells, called microtubule organizing centers, from which many
microtubules might radiate.
The 7 nm-diameter microfilaments are composed of a "string" of
actin "beads". Monomeric actin is conventionally referred to as G-actin
while the polymeric form is called F-actin. The structure of G-actin is
known at atomic resolution (Kabsch et al., 1990). Like microtubules,
microfilament ends exhibit a functional polarity. In this case, the faster
growing end is called the barbed end and the slower growing end is called
the pointed end based on the way that the S1 myosin fragment decorates Factin. Again, as for microtubules, the basic structure of microfilaments is
remarkably conserved among all eukaryotes.
Cytoskeletal organelles- Microtubules and microfilaments are
typically arrayed into organelles that carry out specialized functions.
Often, the activities of one or more cytoskeletal organelles are coordinated to
achieve complex cellular behavior. During mitosis, the microtubule-based
spindle is charged with segregating the chromosomes. Following
chromosome segregation, the contractile ring of microfilaments serves to
separate the mitotic cell into two daughter cells. Little is known about how
these two organelles communicate with one another, but the fidelity with
which chromosome segregation precedes cell division at each mitosis
indicates that such communication must exist. Before one can dissect the
higher order relationships among cytoskeletal organelles, it is important to
understand the structure and function of individual organelles. Below we
will describe in more detail a few of the cytoskeletal organelles that are
relevant to these studies.
Microtubule organelles- A particularly striking microtubule
organelle is found within the neuritic processes of nerve cells. Bundles of
microtubules course through the neurite parallel to its long axis. In axons,
the microtubules exhibit a uniform polarity with the plus ends of the
12
microtubule all towards the distal tip of the neurite (Heidemann and
McIntosh, 1981). At least one function that these axonal microtubules serve
is to transport macromolecules from the cell body through the neurite to its
tip. This function is vital since axons can be quite long and are devoid of
protein synthetic machinery. Early evidence that axons served a transport
function came from Weiss and Hiscoe (1948) who observed "damming" on
the proximal side of constricted nerve fibers. Later, work defined distinct
classes of transported molecules using radioactive tracers (Willard et al.,
1974). One of these classes, the fast component, moves at the rate of 2-4 pm
per second and represents vesicles and small granules transported along
microtubule tracks. At the other extreme, the slow component moves at
about 0.002-0.01 gm per second and among its constituents is tubulin
(Hoffman and Lasek, 1975). More recent experiments have demonstrated
that the tubulin moves through the neurite in a coherent phase consistent
with the view that polymer is the transported form of tubulin (Reinsch et
al., 1991). In addition to this transport function, the neuritic arrays of
microtubules also support the structure of neuronal processes. This is
demonstrated by treatment of neurons in culture with microtubule
depolymerizing drugs which results in the rapid retraction of neurites
(Yamada et al., 1970; Solomon and Magendantz, 1981).
Microfilament organelles- The cortical actin cytoskeleton is marked
by a close apposition or a physical union of microfilaments with the plasma
membrane. Among other roles, this organelle lends shape and structural
integrity to the pliable plasma membrane. Clear examples of this function
are found in the membrane skeleton of mammalian erythrocytes and in the
brush border microvilli of intestinal epithelial cells (Marchesi, 1985;
Mooseker, 1985). In these two cell types, respectively, the cortical
cytoskeleton maintains the form of plasma membrane to withstand
tremendous shear forces present in circulatory system or to effect a greater
surface to volume ratio for nutrient absorption in the gut. The cortical
cytoskeleton in these two cell types are particularly well characterized
because of their comparative simplicity and availability in large quantities
for biochemical analysis.
In other cell types, the cortical cytoskeleton plays important roles in
dynamic membrane structures, including those associated with cell
13
motility. For instance, cultured animal cells exhibit a number of motile
surface protrusions (Abercrombie, 1970). Filopodia are fingerlike
projections that extend up to several microns from the cell body and are
often in contact with the culture substratum. When observed over time,
these structures may alternately grow, shrink until they are resorbed by the
cell body, detach from the substratum, and make new attachments
elsewhere. This sort of behavior is suggestive of a role for filopodia in
exploring the extracellular environment. Another, motile cortical
structure is the lamellipodium. This structure is characterized by a
flattened sheet of membrane that spreads out over the culture substratum.
While this structure appears to move away from the cell body, there are also
extensions -membrane ruffles- that appear to move back toward the cell
body. Often, moving cells have a single, well defined lamellipodium that is
oriented at the front of the cell in the direction of movement. This type of
lamellipodium is referred to as the leading edge.
Both filopodia and lamellipodia are rich in microfilaments, but these
microfilaments are less stereotyped in organization compared to the stable
membrane skeletons of mammalian erythrocytes or intestinal brush border
microvilli. Microfilaments in filopodia tend to be parallel bundles whereas
those in lamellipodia form an orthogonal meshwork (Small et al., 1982).
Electron microscopic inspection of fibroblast leading edges reveals that
filopodia and lammellipodia are a continuum of microfilaments rather
than distinct organelles- the lattice of microfilaments in the lamellipodium
is consolidated into the bundled array present in the filopodium. In both
structures, the barbed ends of the cortical microfilaments are oriented
toward the plasma membrane (Small, 1978).
That filopodial and lamellipodial activity is often polarized in the
form of a leading edge initially suggested that these structures played a key
role in the translocation of cells across surfaces. Indeed, treatments that
disrupt microfilaments cause these type of cortical cytoskeletal structures
to collapse and cell crawling to cease (Albrecht-Buehler and Lancaster,
1976). The molecular mechanisms underlying the protrusion of filopodia
and lamellipodia are not yet clear. Among the current models for actin
based motility in animal cells is one that posits actin polymerization as a
driving force for membrane protrusion (reviewed in Condeelis, 1993). The
essence of this model is that thermal vibration in the plasma membrane
14
allows for the interposition and subsequent addition of actin monomers onto
the barbed ends of anchored microfilaments. This newly polymerized actin
results in displacement of the membrane. This model is consistent with
the force generating capacity of actin polymerization as calculated from
thermodynamic considerations and thermal oscillation properties of
biological membranes. Recently, support for this model of actin based
motility was garnered from an unexpected source. Studies on the invasive
pathogenic bacteria Listeria and Shigella reveal that these organisms move
about the cytoplasm of host cells by polymerizing host actin into a structure
called a comet tail (Tilney et al., 1992). Observation of the actin dynamics in
comet tails indicated that propulsion is driven by polymerization proximal
to the bacterium at a rate that is consistent with the rate of movement of the
bacterium and, interestingly, the rates of membrane protrusive activity
seen in cultured cells (Theriot et al., 1992).
Composite cytoskeletal organelles- The classification of the
cytoskeletal organelles described above as strictly microtubule and
microfilament organelles is somewhat arbitrary since often more than one
type of polymer occupies any cellular compartment. However, in some
cytoskeletal organelles, the functions of more than one polymer type may be
integrated to accomplish a particular task. These we have designated
composite cytoskeletal organelles. One example of a composite organelle is
the neuronal growth cone. At the distal tip of growing neurite there exists a
motile domain called the growth cone. Growth cones possess many of the
same motile actin-containing cortical cytoskeletal features, like the
filopodia and lamellipodia described above. Evidence indicates that the
growth cone filopodia play a role in transducing signals from the local
environment of the growth cone (Davenport et al., 1993). In fact, in
preparations of grasshopper limbs, where resolution of growth cones is
possible during neurite outgrowth, growth cone filopodia are observed to
reach out and touch guidepost cells that direct growth cone steering events.
Disruption of filopodial microfilaments with cytochalasin results in
disoriented axonal outgrowth (Bentley and Toroian-Raymond, 1986). As
mentioned above, the neurite shaft trailing the growth cone is filled with
microtubules that play a role in macromolecular transport to the growth
cone. High resolution imaging of microtubules revealed that they also
15
extend well into the growth cone and occupy the same compartments with
microfilaments (Tanaka and Kirschner, 1991; Sabry et al., 1991). These
growth cone microtubules displayed several interconvertible arrangements
and could sometimes be detected deep in lamellipodia where they closely
apposed the membrane. Interestingly, in the same study, just prior to
growth cone turning, microtubules were often consolidated in the future
direction of growth. This provocative result suggests that microtubles and
microfilaments may cooperate at sites where growth cone steering
decisions are being made.
Another example of a composite cytoskeletal organelle is the
marginal band of avian erythrocytes. In mature erythrocytes, a band
consisting of 10-14 microtubule profiles trace an elliptical orbit around the
cell periphery (Miller and Solomon, 1984). Microfilaments are also
concentrated in the region of the marginal band microtubules (Kim et al.,
1987). One role for the marginal band is in the morphogenesis of these
cells. During development chicken erythrocytes change from an initial
spherical form to a lentil-shaped mature form. Concomitant with this
shape change, initially diffuse microtubule arrays are first bundled then
the bundles are consolidated into the final position of the marginal band.
Treatment of developing erythrocytes with cytochalasin to disrupt
microfilaments results in abnormal process formation in the immature
spherical cells (Winckler and Solomon, 1991). Taken together, these
observations have suggested a mechanism for the morphological transition
in chicken erythrocytes in which the relatively rigid microtubules pushing
from within are constrained by the plasma membrane and cortical
cytoskeleton. The net effect of these opposing forces is that microtubules
take the path of least resistance which is to circle the equator of the ellipse
in a bundle.
16
MOLECULAR MECHANISMS FOR ORGANIZATION OF THE
CYTOSKELETON
In the preceding section, we described a few of the many known
cytoskeletal organelles. From these examples, it is clear that microtubules
and microfilaments may adopt a wide variety of structural and functional
arrangements. What are the molecular determinants of cytoskeletal
organization? Over the years, several models have emerged that explain, in
part, how the cell might achieve the complex molecular relationships
present in the cytoskeleton. In this section, we will briefly describe and
illustrate the salient points of each of the predominant paradigms. At the
outset however, we would like to stress that these models should not be
considered as absolutes. Indeed, there are examples to support each of the
models presented. An extant challenge in molecular cell biology is to
understand how these mechanisms are integrated within the cell.
Modulation of polymer dynamics- In vitro assembly reactions of
purified tubulin and actin demonstrate that monomers are in an
equilibrium with polymer. Under defined conditions in solution, there
exists a concentration of tubulin and actin such that a steady state is
reached between the monomeric and polymeric pools- polymerization is
balanced by depolymerization. This is termed the critical concentration.
Above the critical concentration, there is net polymer assembly and below
this value there is net disassembly. The critical concentration for the fastgrowing barbed ends of microfilaments is around 1gM while for the slower
growing pointed ends this value is near 8 gM under physiological
conditions. For microtubules, the critical concentration is around 14 gM.
However, Mitchison and Kirschner (1984a) found that when the minus
ends of microtubules are sequestered by organizing centers (centrosomes),
microtubule growth is supported at lower concentrations. Direct
observation of individual microtubules emanating from organizing centers
after dilution below the critical concentration led to a surprising finding.
Some microtubules continued to grow while others disappeared. To explain
this unusual behavior, Mitchison and Kirschner (1984b) proposed the
dynamic instability model for microtubule growth. Microtubule dynamics
are governed by four parameters- polymerization rate, depolymerization
17
rate, transition from growing to shrinking (catastrophe), and transition
from shrinking to growing (rescue). The essence of the dynamic instability
model is that newly polymerized tubulin subunits have a guanosine
triphosphate molecule bound- the so-called GTP cap. After addition to the
polymer, the GTP is hydrolyzed to guanosine diphosphate. Since the off
rate of GDP-tubulin is much higher than that for GTP-tubulin, when a
microtubule loses its GTP cap, catastrophe occurs and depolymerization is
rapid.
These in vitro studies have relevance for microfilaments and
microtubules in cells. Since estimates of the cellular concentration of actin
range from 150 to 900 gM are well above the critical concentration of actin
assembly, one would expect most of the actin to be polymerized. Moreover,
under these conditions, there would seem to be little opportunity for the sort
of dynamic actin polymerization evident in the motile cortical structures
described above. However, several non-actin protein factors influence
microfilament polymerization in the cell. Due to the activities of monomer
sequestering proteins, like thymosin 14, and barbed-end capping proteins,
such as gelsolin, blocking actin polymerization, a considerable portion of
the actin in the cell is maintained as G-actin (Carlier and Pantaloni, 1994).
This pool of G-actin, which exhibits a focal distribution in cultured cells
(Cao, et al., 1993) may be available for localized F-actin assembly
(Cassimeris et al., 1992). Profilin plays a more complex role in actin
dynamics. Genetic experiments support an essential role for profilin in
actin-based functions in yeast and Drosophila and in intracellular motility
of Listeria (Haarer, et al. 1990; Cooley, et. al., 1992; Theriot, et al., 1994).
Biochemical data indicates that profilin may facilitate transfer of thymosin
14 sequestered G-actin to uncapped barbed ends, but the precise role of
profilin in animal cells is unclear. For instance, overexpression of profilin
in CHO cells results in an increase in F-actin, whereas microinjection of
profilin in fibroblasts leads to microfilament disassembly (Finkel et al.,
1993; Cao et al., 1992)
In cells, as in the test tube, microtubules are highly dynamic
structures. Co-existing with in the same cell are both growing and
shrinking microtubules (Schultze and Kirschner, 1988). While the
dynamic instability model for microtubule assembly is widely accepted as a
plausible mechanism to explain, in part, microtubule dynamics in cells,
18
certain aspects of this model remain to be firmly demonstrated. For
instance, although there is some direct evidence for a GTP-cap on
microtubules in vitro, there is no direct evidence for this structure
occurring in vivo (Mitchison, 1993). Analogous to the microfilament barbed
end capping proteins, other proteins such as pericentrin and y- tubulin
occur in association with the minus ends of microtubules at organizing
centers effectively blocking subunit addition at this end. In contrast to the
monomer sequestering activities regulating the actin cytoskeleton, there
are as yet no proteins that have been shown to sequester assembly
competent tubulin dimer. However, recent evidence indicates that Rbl2p, a
protein that binds to B-tubulin in yeast cells, could play an important role in
microtubule assembly by regulating dimer formation (Archer et al., 1995).
In conclusion, factors that modulate the polymer dynamics of cytoskeletal
proteins most certainly influence when and where cytoskeletal polymers
form in cells.
Polymer molecular heterogeneity- Another potential mechanism for
regulating the cytoskeleton exists in the form of molecular heterogeneity of
polymer proteins. Such heterogeneity comes in two basic forms- protein
isotypes encoded by distinct genes and post-translational modifications.
There are at least six actin isoforms that are encoded by separate genes
(Vandekerckhove and Weber, 1978). These include smooth muscle a- and yactin, cardiac and skeletal muscle a- actin and non-muscle B- and y- actin.
In certain non-muscle cells, there may be differential sorting of actin
isoforms. Using isoform specific antisera, DeNofrio et al. (1989) showed
that in microvascular pericytes B- and y- actin is enriched in motile cellular
domains whereas a- actin tended to be distributed in stress fibers. One
possible mechanism for this sorting could be the translation of localized Bactin mRNA (Lawrence and Singer, 1986). However, the functional basis
for F-actin sorting is unclear at present.
Higher eukaryotic organisms have around 6 a-tubulin and 6 Btubulin genes each. These have been relatively well conserved throughout
evolution (Sullivan, 1988). In general, different tubulin isotypes, even
highly evolutionarily divergent isotypes, may coassemble in vivo arguing
that isotypes are not sufficient to specify particular microtubule organelles
(Bond et al., 1986). However, certain isotypes may confer special properties
19
to microtubules. For instance, Antarctic fishes, which require cold-stable
microtubules, also have unique a-tubulin genes (Detrich et al., 1987).
Tubulins are also the targets for extensive post-translational
modifications. Among these are de-tyrosination, acetylation, and
polyglutamylation (reviewed in Joshi and Cleveland, 1990). Again though,
the functional significance of these modifications is not certain. It is
known that axonal arrays of microtubules are stable in the presence of
depolymerizing drugs and also tend to be de-tyrosinated. However,
enhanced stability of these microtubules does not seem to be a direct
consequence of de-tyrosination (Khawaja, et al., 1988).
Lower eukaryotic organisms present a much simpler set of tubulin
isotypes and post-translational modifications and also afford the possibility
of testing function by gene replacement. Saccharomyces cerevisiae has two
a-tubulin genes and one B-tubulin gene. One of the two a- tubulin genes is
essential, although overexpression of the non-essential a-tubulin can
compensate for the loss of the essential gene (Schatz et al. 1986). Indeed,
yeast are quite sensitive to the relative levels of a- and B-tubulin (Katz et al.,
1990; Weinstein and Solomon, 1990). The role of post-translational
modifications of tubulin has also been investigated by genetic analysis.
Replacement of the single Tetrahymena themophila a-tubulin gene with a
mutant form incapable of being modified at an evolutionarily conserved
acetylation site had no effect on these cells (Gaertig et al., 1995). These
rigorous tests of tubulin function in these simple organisms argue against
an essential role for both specific isotypes and post-translational
modifications of tubulin. While not definitive, the accumulated evidence
thus far points toward molecular heterogeneity playing an auxiliary role in
specifying cytoskeletal organization.
Polymer accessory proteins- Another mechanism by which the
cytoskeleton can be modulated is by the action of non-covalently associated
proteins- the accessory proteins. Both microtubules and microfilaments
have characteristic sets of accessory proteins. We have already mentioned
several such proteins in the discussion of the regulation of actin polymer
dynamics- monomer sequestering proteins and barbed end capping
proteins. For the microfilament cytoskeleton, there exists a wide range of
accessory proteins. A full accounting of this long list is well beyond the
20
scope of this introduction. Therefore, we will just briefly describe the
classes of microfilament accessory proteins in non-muscle cells to illustrate
the range of their activities.
Microfilament accessory proteins can be broadly separated into motor
and non-motor proteins. Motor proteins might play important roles in cell
motility by mediating the sliding of microfilaments relative to one another
and/or to a substratum over which a cell is walking. A relative to myosin
II, the mechanoenzyme responsible for muscular contraction, myosin I
was initially identified as an actin-stimulated ATPase activity from
Acanthamoeba (Pollard and Korn, 1973). Subsequently, this enzyme was
shown to support ATP-dependent movement of F-actin in in vitro motility
assays (Albanesi et al., 1985). Myosin I's from a host of non-muscle cell
types now comprise a large family of molecules. The function of myosin I
has been investigated in cells. Disruption of myosin heavy chain
expression in Dictyostelium by antisense inhibition or gene disruption leads
to mutant phenotypes in these cells (Knecht and Loomis, 1987; De Lozanne
and Spudich, 1987). These mutants are defective in cytokinesis and a
developmentally regulated morphogenetic event. These phenotypes are
consistent with myosin I playing a key role in motile cell behavior.
However, the cells still move. Individual cells walk across substrata
exhibiting the classic (wild-type) motile behaviors such a extension of
membrane ruffles, pseudopods, and other cortical cytoskeletal structures.
There may be other myosin I-like molecules in these cells that can
compensate for the reduction or loss of myosin I. Alternatively, myosin I
might be required for some motile behaviors and not others.
Non-motor microfilament accessory proteins come in a variety of
flavors. These include crosslinking proteins such as filamin which can
link actin filaments into an orthogonal network forming an actin gel in
solution (Wang and Singer, 1987). Other crosslinking proteins arrange
microfilaments into parallel bundles. Still other classes of microfilament
accessory proteins can nucleate F-actin polymerization and sever preexisting microfilaments. Examples of the last three classes can be found in
a single molecule. Villin is a sort of molecular jack-of-all-trades which
takes its name from the source of its initial purification in intestinal brush
border microvilli. In biochemical studies with purified villin, this molecule
can sever, crosslink into bundles, and cap the barbed ends of actin
21
filaments or nucleate the assembly of new filaments (Matsudaira and
Janmey, 1988). The particular activities that villin exhibits at any one time
are regulated by calcium and phosphoinositides. Transfection of CV-1 cells
with a full-length villin cDNA construct induces the presence of long,
intestinal microvilli-like processes suggesting that villin plays an
important role in microvilli formation (Friederich et al., 1989). One final
class of microfilament accessory proteins are those which stabilize F-actin
by binding along the sides of microfilaments. An example of this type is
tropomyosin. Tropomyosins occur in a wide variety of muscle and nonmuscle tissues. Biochemical studies have demonstrated that tropomyosin
can protect actin filaments from severing (Fattoum et al., 1983).
Endogenous and microinjected tropomyosins localize to stress fibers
(Pittenger and Helfman, 1992). However the role of these proteins in the
establishment or maintenance of these relatively stable F-actin structures
remains to be elucidated.
By way of comparison to microfilaments, microtubules may seem
rather bereft accessory protein functions. Microtubules do have motor
proteins- dynein and kinesin. Cytoplasmic dynein was first identified as an
ATPase activity from sea urchin eggs (Weisenberg and Taylor, 1968).
Later, this protein was found associated with microtubules from bovine
brain preparations and this form was shown to translocate to the minus
ends of microtubules in an ATP-dependent manner in vitro (Paschal et al.,
1987). The presence of a minus-end directed motor in neurons has led to
speculation that this motor plays a role in retrograde axonal transport
(Vallee et al. 1989). More is known about kinesin. This protein was first
purified as an ATP-binding protein from squid giant axons (Vale et al.,
1985). In vitro motility assays of the purified protein revealed an ATPdependent plus end-directed motor activity. Like myosins, kinesins
represent a large family of related proteins. Mutant alleles of kinesin heavy
chain homologs in Aspergillus nidulans, Saccharomyces cerevisiae, and
Drosophila melanogaster support a role for kinesin in mitotic or meiotic
spindle function (reviewed in Endow and Titus, 1992). Other experiments
support a role for kinesin in fast axonal transport. Treatment of cultured
neurons with kinesin antisense oligonucleotides resulted in a reduction of
neurite length and aberrant distribution of growth cone localized proteins
(Ferreira et al., 1992). A more pertinent finding is that antibody
22
perturbation studies show that kinesins play essential roles in anterograde
vesicular transport in squid giant axons (Brady et al., 1990).
Microtubules also have non-motor accessory proteins. These
activities include proteins or protein complexes that nucleate microtubule
assembly at microtubule organizing centers. As mentioned in the section
on polymer dynamics, the centrosome is slowly yielding to molecular and
cellular analysis. Microinjection of anti- y-tubulin antibodies inhibits
reassembly of microtubules after drug-induced depolymerization (Joshi et
al., 1990. Immunodepletion of cell extracts with anti-y-tubulin antibodies
blocks the ability of sperm centrsomes to nucleate microtubule growth in a
cell free system (Felix et al., 1994; Stearns and Kirschner, 1994). Likewise,
depletion of pericentrin leads to the same effect, although y-tubulin is
recruited from the cell extract to the nucleating centrosome indicating that
y-tubulin is not sufficient for nucleation (Doxsey et al., 1994). There is
recent evidence for an activity, katainin, that severs microtubules in vitro
(McNally and Vale, 1993). Such a protein could be of tremendous
importance for organizing the cytoskeleton; however, little is known about
the in vivo function of this protein at present. By far the most extensively
studied non-motor microtubule associated proteins are the so-called fibrous
microtubule-associated proteins (Wiche et al., 1991). Weisenberg (1972)
demonstrated that microtubules could be taken through successive rounds
of polymerization and depolymerization in vitro. Since then, others have
used this technique as a means for enrichment of proteins that specifically
associate with microtubules. These studies initially defined a class of
proteins called microtubule-associated proteins or MAPs (for example
Weingarten et al., 1975). Over time, this group of proteins came to include
MAPlA, MAP1B, MAP2, MAP4 and tau. These proteins all shared the
features of high affinity binding to microtubules. Other, perhaps weaker
microtubule binding MAPs-the chartins- were identified as proteins that
co-align with microtubules in cells and are extracted along with tubulin by
calcium (Duerr et al., 1981). Instead of reviewing the current evidence for
the in vivo function for all of these MAPs, we will restrict comments below
to two MAPs- tau and MAP2- that are germane to the present study.
The examples of accessory proteins above demonstrate the many
ways in which these molecules can modulate and specify the organization
of the cytoskeleton. In the work presented here, we have endeavored to
23
understand further the cellular functions of accessory proteins. Below, we
will introduce the particular accessory proteins that are the focus of the
present study.
24
IDENTIFICATION AND CHARACTERIZATION OF TAU AND
RADIXiN: TWO ACCESSORY PROTEINS THE CYTOSKELETON IN
CULTURED MAMMALIAN CELLS
Tau- As mentioned above, MAP2 and tau were identified by virtue of
their ability to co-assemble with microtubules. Among the other
demonstrated biochemical properties of tau and MAP2 is stimulation of
microtubule polymerization and microtubule bundling (Cleveland et al.
1977). In neuronal cells tau and MAP2 occur in the neuritic processes with
enrichment of tau in axons and MAP2 in dendrites (Matus et al., 1981;
Peng et al., 1986). Taken together these observations suggested that tau and
MAP2 might mediate the bundling of the microtubule arrays found in those
organelles. Transfection of tau and MAP2 into cultured cells induces the
presence of bundles of microtubules (Kanai et al., 1989; Lewis et al., 1989).
Molecular cloning revealed that these two proteins have related amino-acid
sequences in their carboxy-terminal domains (Lee et al., 1988; Lewis et al.,
1988). These conserved motifs were shown to be capable of promoting
microtubule assembly in vitro (Joly et al., 1989). Ultrastructural analysis of
tau and MAP2 by immunoelectron microscopy indicated that these
molecules project from the surface of the microtubule lattice (Hirokawa et
al., 1988 a,b). These results along with the biochemical data fueled
speculation that these projection domains mediated lateral interactions
between microtubules. Indeed, transfection studies in cultured cells using
MAP2 deletion constructs identified a short hydrophobic segment
responsible for the observed bundling properties of MAP2 in cells and
suggested that MAP2 molecules dimerized via this domain (Lewis et al.,
1989). However, subsequent to this study, these authors provided evidence
that cast doubt on their earlier data (Lewis and Cowan, 1990).
Furthermore, more recent studies also argue against MAP2-mediated
microtubule bundling by dimerization of MAP2 molecules (Burgin et al.,
1994). What mechanisms might account for microtubule bundling? One
hypothesis put forth by Chapin et al. (1991), is based on the observation that
taxol- a microtubule stabilizing drug- induces microtubule bundling. This
theory holds that bundles are an energetically favorable state for stable
microtubules. If MAP2 and tau do not directly mediate microtubule
25
bundling in cells, what are their functions? We have addressed this
question for tau in Chapter Two.
Radixin- For a number of years, the Solomon laboratory has used the
chicken erythrocyte as a model system to study the molecular specification
of the cytoskeleton. As described above, these cells are available in bulk, the
cytoskeleton -the marginal band- in this cell is rather simple, and
microtubules and microfilaments are though to play a critical role in the
morphogenesis of this cell type (Winckler and Solomon, 1991). That
accessory proteins are critical for the specification of the marginal band is
demonstrated by the following experiment. In detergent extracted cells, the
marginal band microtubules can be selectively depolymerized by a
temperature shift so that endogenous tubulin can be separated from the
remaining cytoskeleton. When exogenous calf brain tubulin is added to
these erythrocyte cytoskeletons in a buffer that supports polymerization, the
marginal band microtubules re-form in detail- the correct number of
microtubules course around the original equator of the extracted cell (Swan
and Solomon, 1984). This experiment argues that polymer-extrinsic factors
specify the structure of the band. Birgbauer and Solomon (1989) conducted
an immunological screen for non-tubulin marginal band proteins. They
reasoned that MAPs could play a role in specification of the band and so
used cycled chicken brain microtubule proteins as an immunogen for the
production of monoclonal antibodies. Antibodies were screened for specific
localization to the marginal band. One antibody, 13H9, emerged that
fulfilled the criteria set forth in the screen. This antibody recognized an 80
kDa protein on Western blots of chicken cytoskeletal proteins and also
stained cortical cytoskeletal structures in fibroblasts and neurons (Goslin et
al., 1989; Birgbauer, 1991). Through the considerable efforts of a number of
members of the Solomon laboratory, the identity of the 13H9 antigen was
established. mAb 13H9 recognized an epitope conserved in all members of
the ezrin-radixin-moesin (ERM) protein family.
The ERM proteins were being isolated and characterized
concomitantly with the work on the marginal band from the Solomon
laboratory. At that point in time, the ERM proteins were thought to be
involved in mediating interactions between the cytoskeleton and plasma
membrane. This hypothesis stemmed from two findings: 1) the proteins
26
were localized to cortical cytoskeletal sites, and 2) molecular cloning
revealed that the amino-terminal domains of ERM proteins are similar to
the amino-terminal domain of band 4.1 (Fig. 1-1). All of the members of the
band 4.1 superfamily share the property of being closely associated with the
plasma membrane. Other studies implicated the amino-terminal domain
of band 4.1 in binding to the integral membrane protein glycophorin (Leto et
al., 1986). Sequencing of the distinct genes that encode ERM proteins in
different laboratories revealed that these proteins are highly similar to one
another and created the unfortunate situation where the proteins are
named individually rather than as isoforms of one another.
The similarity of the amino acid sequences of ERM proteins
hampered immunological efforts to determine the cellular and tissue
distribution of ERM proteins. Winckler et al. (1994) developed mono-specific
anti-peptide antisera that have shed light on this issue. The results from
this study indicated that radixin and ezrin localize to the position of the
marginal band, with radixin being the predominant species. In a
developmental series, radixin localized to the position of the marginal band
only after the microtubules had arrived there, making it unlikely that this
protein is involved in the establishment of marginal band microtubules. In
fact, during development, radixin is more closely co-localized with F-actin.
Data indicate that the arrival of radixin in the band is correlated with the
resistance of marginal band microfilaments to drug-induced
depolymerization, supporting a possible role for radixin in stabilization of
the marginal band. In Chapters Three through Seven, we have extended
the studies on radixin with an emphasis on understanding its function in
cultured mammalian cells.
27
Figure 1-1. Protein sequence comparison of ERM proteins and band 4.1
family members. Ezrin and moesin are compared to full-length radixin,
and the amino-terminal domains of band 4.1 family members are
compared to the amino-terminal domain of radixin using the BestFit
sequence comparison program. Percent similarity is indicated at the right
of the figure. Actual or predicted structural features of radixin are
indicated above the schematic diagram of radixin.
28
Radixin: A member of the ERM protein family
and Band 4.1 superfamily
The ERM proteins
band 4.1 homology
alpha helix
F-actin
PPP binding
% similarity to
radixin
100
Radixin
87
Ezrin
I
ceN I
Moesin
The Band 4.1 superfamily
89
% similarity to
radixin N-domain
Band 4.1
52
Merlin
78
Talin
iI
rn-rn
H-
47
EM10
73
PTPH1
55
PTPaseMEG
51
m pq
pq
29
CHAPTER TWO:
Antisense inhibition of tau protein expression in embryonal
carcinoma P19 cells.
30
SUMMARY
Utilizing the properties of embryonal carcinoma P19 cells, we have
studied the role of the microtubule-associated protein tau during neuronal
differentiation. P19 cells can be induced to form either neural cultures
(including neurons and glia) in response to retinoic acid (RA) or muscle
cultures in response to dimethyl sulfoxide (DMSO). Because P19 cells can
assume a number of different cell fates, like the early embryonic cells from
which they were derived, these cells may provide a system to assess tau
function during early stages of neuronal differentiation. To test tau
function in P19 cells, we have expressed an antisense RNA element to
inhibit specifically tau protein expression in these cells. Our results
indicate that inhibition of tau expression, unlike inhibition of another
microtubule-associated protein, MAP2, in these same cells, has no gross
effect on the expression of a differentiated neuronal morphology.
31
INTRODUCTION
The dramatic changes in cell shape that occur during cellular
differentiation are often manifested in changes in cytoskeletal organization.
Considerable effort has focused on defining the molecular basis of the
regulation of cytoskeletal organization during cellular differentiation.
Neurons are a particularly attractive model system for the study of cellular
specification of cytoskeletal organization because their striking
morphologies play an explicit role in nervous system function. The axons
and dendrites, the most prominent morphological features of neurons,
contain dense, longitudinal bundles of microtubules. Drug interference
experiments indicate that these microtubules are necessary for both the
establishment and maintenance of neuritic processes. Since microtubule
associated proteins (MAPs) are known to bind to and stabilize microtubules
in vitro and in vivo, and since the expression patterns of some MAPs are
modulated during neuronal differentiation, MAPs may be significant determinants of neuronal cell shape (Matus, 1990; Wiche et al., 1991).
We have investigated the function of the neuronal MAP tau. Tau was
initially identified on the basis of its ability to co-assemble with tubulin in
vitro, and to promote that assembly reaction (Weingarten et al., 1975;
Cleveland et al., 1977). Tau stabilizes microtubules by modulating dynamic
instability (Dreschel et al., 1992)Immunolocalization revealed that tau is
concentrated in axons of mature neurons (Binder et al., 1985), in the
somatodendritic domain of developing neurons (Kosik and Finch, 1987) and
even in glial cells (Papasozomenos and Binder, 1987). Tau is encoded by a
single gene, conserved among vertebrates, but the protein exists as a set of
polypeptides generated by both alternative splicing (Himmler, 1989) and at
least one type of covalent modification (Lindwall and Cole, 1984). During
normal development there is a shift from expression of lower molecular
weight tau isoforms to higher molecular weight isoforms (Kosik et al.,
1989). In Alzheimer's disease patients, tau in brain tissue is abnormally
phosphorylated (Grundke-Iqbal et al., 1986). It is also a principal
constituent of the paired helical filaments associated with neurofibrillary
tangles- a pathological hallmark of the disease (Lee et al., 1991). However,
the role of tau in this disease is unclear.
32
Several groups have assayed tau function in model cell culture
systems. Introduction of tau into non-neuronal cells by either
microinjection of the protein (Drubin and Kirschner, 1986) or transfection
with tau cDNA (Kanai et al., 1989; Lewis et al., 1989) shows that exogenous
tau decorates microtubules and promotes tubulin polymerization.
Expression of tau in moth ovary cells induces the formation of long,
microtubule-filled processes (Knops et al., 1991). In two neuronal cell
systems, disruption of tau expression with antisense oligonucleotides
affects process extension. In primary cerebellar neurons, the treated cells
initiated neurites but failed to elaborate axons (Caceres and Kosik, 1990;
Caceres et al., 1991). The processes of nerve growth factor-induced PC12
cells eventually retracted their neurites in response to antisense
oligonucleotides (Hanemaaijer and Ginzburg, 1991). In another study in
PC12 cells, nerve growth factor-induced transfectants expressing tau
antisense RNA showed a decrease in mean neurite length compared to
controls (Esmaeli-Azad et al., 1994). Taken together, the results are
consistent with tau playing a central role in stabilizing neuritic
microtubule arrays.
We have investigated tau protein function in embryonal carcinoma
P19 cells by inhibiting its synthesis through the constitutive expression of a
tau antisense transcript. P19 cells can be induced to differentiate into
neural cells with RA or into muscle cells with DMSO (McBurney and
Rogers, 1982; McBurney et al., 1982). RA-induced P19 cultures include cells
that possess many characteristics of neurons, such as the extension of
neuritic processes and the expression of neuron-specific biochemical
markers (Jones-Villeneuve, 1983; McBurney et al., 1988); and also contain
cells resembling several non-neuronal cell types (Jones-Villeneuve et al.,
1982).
P19 cells are an attractive model system for the study of tau protein
function. P19 cells are relatively uncommitted precursor cells since they
can be driven down either neural or muscle differentiation pathways.
Therefore, the possibility exists to examine tau's contribution to early stages
of differentiation. Since the expression of many genes are induced de novo
by RA or DMSO in P19 cells, antisense RNA to these genes expressed in the
uninduced state would lack a target and so not be expected to have effects on
uninduced cells. This feature would allow for the selection of cell lines
33
without antisense RNA expression imposing an additional selective
pressure. Previously, others in this lab used an antisense strategy in P19
cells to examine the role of another microtubule component, MAP2, in
neuronal differentiation. The results demonstrated that MAP2 expression
is required for two of the phenotypes associated with the RA-induced
differentiation of these cells: the extension of neurites and decreased cell
proliferation (Dinsmore and Solomon, 1991). Here we show that inhibition
of tau expression has consequences different from those associated with
MAP2 inhibition. The reduction in tau levels do not affect neuronal
morphology or cell growth as a response to RA in P19 cells.
34
MATERIALS AND METHODS
Cell culture. The EC cell line P19 (McBurney, 1982) was kindly
provided by Dr. M. McBurney (University of Ottawa, Ottawa, Canada). P19A4, a euploid P19 subclone isolated during these studies and described in
Appendix One, was used for some experiments where indicated. Routine
culture and induction of P19 cells were done as previously described
(Rudnicki and McBurney, 1987). The RA and DMSO inductions were
performed as modified by Dinsmore and Solomon (1991). Briefly, 2.4 X 106
cells were plated in 100 mm bacteriological grade dishes in MEMa + 2%
fetal calf serum supplemented with 0.5 gM all-trans retinoic acid (Sigma).
Under these conditions, cells do not adhere to the culture dish and instead
form aggregates suspended in the medium. Medium described above was
replaced after 2 days incubation. After 4 days incubation, cell aggregates
were harvested and trypsinized to form a single cell suspension. From this
suspension, 3 X 106 cells were plated in tissue culture grade dishes in
MEM(a) + 7.5% calf serum/2.5% fetal calf serum. Normally, cells treated
in this manner have fully elaborated neurites and expressed neuronspecific biochemical markers 3 d after plating on tissue culture plastic.
Cells were photographed and harvested for biochemical analysis at this
time unless otherwise noted. Cells were examined with a Nikon Diaphot
inverted phase microscope. All phase micrographs were recorded on
Kodak Plus-X-Pan film, ASA125, developed in Microdol-X (Kodak).
DNA constructs and transfection. The vector pGEM7-KJ1-Sal
(described in Dinsmore and Solomon, 1991) was modified by partial
digestion with BamHI, blunting the ends with the KIenow fragment of E.
coli DNA polymerase and ligating with T4 DNA ligase to eliminate a
BamHI site 3' of the phosphoglycerate kinase (PGK) polyadenylation signal
sequence. This construct, named pGKBA, was used in these studies. The
pCXN2 vector was provided by Dr. Jun-ichi Miyazaki and is described in
(Niwa et al., 1991).
The F13 murine tau cDNA clone was kindly provided by Dr. G. Lee
(Brigham and Women's Hospital, Boston, MA). The F13 clone contains 186
bases of 5' untranslated sequence and 562 bases (54%) of coding sequence of
a tau cDNA from 6-day-old mouse (Lee et al., 1988). F13 was digested with
35
PstI or SailI and HinDIII to yield a 710 bp fragment (spanning nucleotides
-186 to +524) and a 765 bp fragment (spanning nucleotides -186 to +562)
respectively. These fragments were isolated from a 1% low-melt agarose
gel, phenol extracted and ethanol precipitated. The 3'-overhang ends of the
Pst I fragment were blunted with the 3'-5' exonuclease activity of T4 DNA
polymerase and the 5'-overhang ends of the SalI/HinDIII fragment were
blunted with Kienow enzyme.
The pGKBA vector was prepared for insertion of the PstI-blunted tau
fragment by cutting at the BamHI site 3' of the neo gene and blunting the
ends with Klenow enzyme. The pCXN2 vector was prepared for insertion of
the SalI/HinDIII fragment into the XhoI cloning site downstream of the Bactin promoter by XhoI digestion and blunting the ends with Klenow
enzyme. The linearized, blunt-ended vectors were treated with calf
intestinal phosphatase and isolated from a 1% low-melt agarose gel. The
blunt-ended inserts and vectors were ligated using T4 DNA ligase, and the
ligation mixture was used to transform E. coli. Plasmids were isolated
from transformants, analyzed by restriction enzyme digestion to determine
insert orientation. The resultant plasmids are named and diagrammed in
figure 2-2. All enzymes used for these procedures were obtained from New
England Biolabs and used according to the manufacturers instructions.
For transfection, plasmids were linearized at the NsiI site (for
pGKBA-based plasmids) or the ScaI site (for pCXN2-based plasmids) and
transfected using the calcium phosphate coprecipitate method (Cullen,
1987; Graham and van der Eb, 1973) as described for EC cells (Rudnicki et
al., 1988). Stably transfected P19 cells were selected in the presence of 400
jg/ml Geneticin (Gibco-BRL) (or 600pg/ml for pCXN2 vector transfectants).
The resultant colonies were picked, expanded in the continued presence of
the selecting agent, and stored in liquid nitrogen. Cell lines derived in this
manner were periodically passaged through medium containing 400 gg/ml
Geneticin.
Analysis of mRNA. RNA was isolated from P19 cells by phenol
extraction (Wallace, 1987). For Northern blots, RNA was separated on a 1%
agarose gel containing formaldehyde by the method of Ogden and Adams
(1987). After electrophoresis, gels were soaked for 30 min in 0.05 N NaOH,
neutralized in 0.1M Tris-HCi plus 0.15 M NaCl for 30 min and transferred
36
to a GeneScreen nylon membrane in a vacuum blotting apparatus
(Stratagene) for one hour with lOX SSC as transfer buffer. After transfer,
RNA was crosslinked to the membrane in a Stratalinker (Stratagene) at an
energy setting of 240,000 microjoules. The blots were hybridized at 65*C
according to the method of Church and Gilbert (1984) with the PstI insert
from tau clone F13 (the same sequence used to make the sense and
antisense constructs) that was labeled with 3 2 P by the random primer
method (Feinberg and Vogelstein, 1983) with a kit from United States
Biochemical Corporation.
Analysis of protein. For preparation of whole-cell extracts, cells were
washed once with 10 ml PBS per 100 mm plate at 37 0 C and then harvested
in two ways: i) medium containing non-adherent, aggregated cells was
collected in a centrifuge tube and ii) cells adherent to the culture
substratum were harvested by adding 1 ml of PBS at 37*C, scraping the
cells from the dish with a rubber policeman, and collecting the cell
suspension in a centrifuge tube. Cells were pelleted by centrifugation at
1,000 x g for 5 min at 37*C in a tabletop centrifuge and chilled on ice. Cell
pellets were resuspended in 5 times the pellet volume with ice cold PBS plus
0.5% NP-40 and the following protease inhibitors: aprotinin (1 pg/ml),
leupeptin (1 g/ml), pepstatin (1 jig/ml), 0.2 mM PMSF, tosyl-L-arginine
methyl ester (10 gg/ml), and soybean trypsin inhibitor (5 gg/ml); and
incubated on ice for 10 min. The NP-40 lysates were then centrifuged at
7,000 rpm for 5 min at 40 C to sediment the nuclei. The supernatant was
removed, an aliquot was saved for protein quantitation, and the remaining
cell extract was reduced as described (Dinsmore and Sloboda, 1988).
Protein concentrations in cell extracts were determined by the method of
Lowry (1951) as modified by Schacterle and Pollack (1973).
For preparation of heat-stable proteins, NP-40 lysate was brought to
0.75 M NaCl and 10 mM DTT, boiled 5 min, chilled on ice, and spun at
15,000 rpm for 15 min at 40 C in a microfuge (Vallee, 1985). The supernatant
was removed and reduced as described above. The amounts of heat-stable
extracts used during electrophoresis were normalized according to the
protein concentrations of the cell extracts from which they were derived.
Proteins were separated on 4-10% acrylamide, 2-8 M urea linear
gradient gels (0.8 mm) according to the buffer formulations of Laemmli
37
(1970). The following molecular mass markers (Sigma Chemical
Company) were present on all gels: rabbit muscle myosin (205 kD), 1galactosidase (116 kD), phosphorylase b (97.4 kD), bovine serum albumin
(66kD), ovalbumin (45kD), and carbonic anhydrase (29 kD).
Proteins separated on polyacrylamide gels were electrophoretically
transferred to a 0.2 gm nitrocellulose membrane (Schleicher and Schuell)
as described by King et al. (1985) with modifications described by Dinsmore
and Solomon (1991). Nitrocellulose blots were blocked in PBS plus 0.05%
Tween-20 for 15 min and reacted with the 5E2 mAb specific for tau
(described in Galloway et al., 1987), a gift of Dr. K. S. Kosik (Harvard
Medical School, Boston, MA.) diluted 1:100 in PBS. For detection of mouse
immune complexes, blots were reacted with alkaline phosphatase
conjugated goat anti-mouse second antibodies (BioRad) or 1 2 5 1-conjugated
sheep anti-mouse second antibodies (Amersham). Blots were processed
colorimetrically for the detection of alkaline phosphatase conjugates with
the ImmunoSelect development substrate (Gibco-BRL), or
autoradiographically for the detection of 1251 conjugates by exposure to preflashed Kodak X-Omat AR film at -70'C using a Dupont Cronex Lightning
Plus intensifying screen. Quantitation of radioactive bands was performed
using a Phosphorimager (Molecular Dynamics).
38
RESULTS
Expression of tau mRNA in P19 cells. Previous work indicated that
there were a number of MAPs in P19 cells, but-that at least two of them MAP2 and tau - were expressed specifically in RA-induced cells (Dinsmore
and Solomon, 1991). It is important for the antisense strategy used here
that tau expression be induced at the mRNA level only after exposure to
RA. If this criterion is met, there should be no target for the constitutively
expressed antisense transcript in the undifferentiated state. We used a
portion of a murine tau cDNA to probe Northern blots of total RNA from
uninduced, DMSO-induced, and RA-induced P19 cells and found that a tau
message (approximately 6kb) was detectable only in the RA-induced cells
(Fig. 2-1).
Generation of P19 cell lines expressing tau antisense sequence from
the pGKBA vector. We used the constructs shown in Figure 2-2 A-C to
transfect P19 cells. When expressed in this configuration, the 700bp tau
antisense element is fused to the neomycin resistance transcript (Fig. 2-2
A) driven by the PGK promoter. This type of antisense element is
advantageous for two reasons: 1) selection for neomycin resistance is tightly
linked to antisense expression and 2) a similar vector was used previously
to inhibit MAP2 expression in P19 cells (Dinsmore and Solomon, 1991). The
700bp of tau sequence, spanning nucleotides -186 to +564, includes the ATG
start codon of both known tau primary transcripts and represents only a
portion of the 1040 bases of the tau coding region. Two control constructs
contain either the same tau sequence in the sense orientation (Fig. 2-2 B) or
the neomycin resistance selectable marker alone (Fig. 2-2 C). We selected
stable transfectants in the presence of 400 gg/ml G-418. We isolated, in
total, 20 tau antisense lines, 8 tau sense lines, and 9 vector lines from two
separate transfections with the pGKBA-derived constructs.
Expression of the tau antisense construct in two independently
isolated tau antisense lines, TA2B and TA3A, was confirmed by an RNase
protection assay for the tau antisense transcript. This transcript was
present in these lines, but in none of the four control lines examined. TA3A
expressed more of the tau antisense transcript than TA2B (data not shown).
39
Figure 2-1. Tau mRNA expression in EC P19 cells. RNA was harvested
and 10 pg of total RNA was separated on a 1% agarose gel as described in
Materials and Methods. Blot was probed with tau cDNA PstI fragment
from clone F13. Lane 1 contains RNA from uninduced P19 cells; lane 2,
RNA from DMSO-induced P19 cells; and lane 3, RNA from RA-induced
cells. The 6 kb tau transcript is present only in the RA-induced cells. The
high MW species in lane 3 has been identified previously on Northern blots
probed with tau sequences different from the one used here (Kosik et al.,
1989).
40
.6kb
1 23
Figure 2-2. Antisense constructs. Plasmids were constructed as described
in Materials and Methods. (A) The pGKBA-derived plasmids express the
antisense element as a fusion to the 3' end of the neomycin resistance
transcript. (B) A control plasmid expressing the same tau region in (A) in
the sense orientation. (C) pGKBA vector control. (D) The pCXN2-derived
plasmids express separate antisense and neomycin resistance transcripts.
(E) pCXN2 vector control.
42
A) pGKBA-TA1
prom.
neo
B) pGKBA-TS1
prom.
neo
C) pGKBA
D) pCXN2-TA1 IF-actin prom.
E) pCXN2
neo
PGK prom.L
-actin prom.
I
T
tk-prom.
43
PGK pA
tau sense
PGK pA
IPGK pA
tau antisense
I
tau antisense
I
I
tk-prom.
neo
t~o~
I
All uninduced cell lines exhibited similar morphology and growth
properties during routine culture (Table 2-1 "uninduced").
Retinoic acid-induced differentiation of pGKBA-derived antisense cell
lines. We carried pGKBA-derived tau antisense and control cell lines
through the standard RA-differentiation protocol described in Materials
and Methods. RA-induced cultures 3 days after plating onto tissue culture
plastic are shown in Figure 2-3. Many of the antisense lines (8/20)
presented a "clumpy" appearance that suggested cells were more adherent
to one another than to the tissue culture plastic substratum (examples in
Fig. 2-3 A and B). This effect does not result from variations in cell density
since this parameter was held constant at the time of plating. Upon
extended culture, the clustered cells would spread out and neurites would
become evident (see Fig. A2-2 for example). Other tau antisense lines
(10/20) showed typical culture morphology after RA induction.
Two of the tau antisense lines, TAlA and TA3A, showed a much
more dramatic phenotype. After plating the 4 day RA-induced cell
suspension into tissue culture dishes, the cells re-formed aggregates that
were completely non-adherent to the substratum (Figs. 2-3 C and A2-1).
Unlike the other clumpy RA-induced antisense lines described above, TA1A
and TA3A aggregates became only weakly adherent to the culture
substratum after extended periods, but no neurites were evident (see Figure
A2-2). Southern blot analysis of genomic DNA demonstrated that TAlA
and TA3A have identically integrated antisense expression constructs (data
not shown). Therefore, these two lines most probably were not derived from
two independent integration events. Rather, it is more likely that these two
lines arose from cells separated after integration of the tau antisense
construct during the selection or clonal expansion processes.
Some of the control lines (2/3 sense and 2/3 vector) also exhibited the
clumpy phenotype. Figure 2-3 D shows this in a tau sense expressing cell
line TS2A. Figure 2-3 E shows an unaffected vector transfected cell line,
N2B, that is comparable to untransfected RA-induced P19 control cells (Fig.
2-3 F). None of the control lines showed the markedly different phenotype
present in TA1A and TA3A. These features of RA-differentiated P19 cell
lines were reproducible over many independent trials.
44
We compared cell growth in tau antisense lines with control lines
after retinoic acid exposure (Table 2-1). The results indicate that there are
no remarkable differences in the overall growth rate during RA induction
between tau antisense and control lines. Control lines have doubling times
around 30 hours. Tau antisense lines vary within 10% of this value. As
expected, the growth rates of all lines are slower during RA induction.
We assayed tau protein levels in RA-induced tau antisense and
control lines. Cells were induced to differentiate with RA and protein
extracts were harvested from these cells after 3 d of culture on tissue
culture plastic as described in Materials and Methods. We subjected the
extracts to Western blot analysis with a mAb specific for tau (Fig. 2-4A). In
P19 cells, there are two predominant tau isoforms expressed at 50 kDa and
58 kDa. Upon extended exposure of autoradiograms, two minor isoforms at
48 kDa and 52 kDa can be detected (not shown). Tau protein heterogeneity
has been previously described (see Introduction). Interestingly, the Mr of
the predominant tau species in P19 cells is more characteristic of the Mr of
tau proteins found in embryonic neural tissue (48-50 kDa) than of the
higher molecular weight tau isoforms (55-62kDa) present in mature tissues
(Kosik et al., 1989). Figure 2-4 B is a graphical representation of the results
from several experiments. Among the tau antisense lines assayed, TA3A
(and TAlA) exhibited the most dramatic reduction of protein compared to
controls (20 ± 3 % control value). Other independent tau antisense lines
showed more modest reductions of tau protein (TA2B: 76 ± 4 % and 2TA6: 69
+ 10% control value). Tau sense lines showed similar levels of tau protein to
vector lines (data not shown).
Retinoic acid-induced differentiation of pCXN2-derived antisense cell
lines. The results presented above did not conclusively establish the effects
of tau antisense inhibition in P19 cells. The clumpy RA-induced phenotype
was shared by both tau antisense and control lines. The more severe loss of
cell-substratum adhesion seen in TA3A cultures was probably a singularity
and so alone could not substantiate the role of tau in this phenotype.
However, since TA3A also showed the most dramatic reduction of tau, it
was possible that this line was the only one with sufficient antisense
inhibition to produce the observed behavior. Therefore, we used two
approaches in an attempt to refine our experiments. First, we considered
45
Figure 2-3. RA differentiation of pGKBA-derived cell lines. pGKBA-derived
cell lines were taken through the RA differentiation protocol as described in
Materials and Methods. Cultures were photographed 3d after plating onto
tissue culture plastic. A-C tau antisense lines: (A)-TA2B (B)-2TA6 (C)TA3A; (D) tau sense line TS2A; (E) vector control line N2B; (F)
untransfected P19.
46
B
Tool
Table 2-1. Growth of tau antisense lines. Cell growth rates were
determined by taking cell counts with a hemacytometer during routine
growth "uninduced" between splitings (-48 hour interval) or during RAinduction "RA-induced" (-96 hour interval). Growth constant k was
determined by the following formula for log-phase growth: k= ln(xt/xo)/t
where xt=final cell number; xo=initial cell number and t=48 for uninduced
and 96 for RA-induced. Doubling time (td)= 1n2/k is expressed in hours ±
standard deviation. Number of independent determinations (n) for each
value is indicated. Tau antisense lines: TA2B, TA3A, 2TA6; tau sense line
TS2A; vector control line N2B; and untransfected P19 line are shown. n.d.not determined.
48
Table 2-1:
Doubling Time (hours)
RA-induced
Uninduced
Cell Line
U ~E
(n=4)
TA2B
20.2
+
1.9
(n=17)
29.1 ± 1.2
TA3A
18.3
+
2.5
(n=29)
26.4 ±1.7 (n=7)
n.d.
2TA6
33.2 ± 2.3
(n=3)
TS2A
17.3
±
1.9
(n=30)
30.0 ± 1.3
(n=6)
N2B
20.0
+
2.9
(n=30)
30.8 ± 2.8
(n=7)
29.5 ± 0.9
(n=3)
P19
n.d.
49
Figure 2-4. Tau protein expression in RA-induced pGKBA-derived cell
lines. pGKBA-derived cell lines were taken through the RA differentiation
protocol as described in Materials and Methods. Protein was harvested 3d
after plating onto tissue culture plastic and analyzed for tau expression as
described in Materials and Methods. (A) Tau western blot. Lane 1-TA2B;
lane 2-TA3A; lane 3-2TA6; lane 4-N2B (B) Graphical representation of tau
western blotting data. Band intensities were determined by
Phosphorimager analysis as described in Materials and Methods. The data
was compiled from three separate induction experiments.
50
(A)
-97
-45
1
2
3
4
(B)
120-
100-
80-
S6040
CI)
20
0
--
N2B
-
TA2B
2TA6
P19 cell line
TA3A
the possibility that the clumpy phenotype arose due to variation within the
parental P19 stock used for the transfection. In fact, we sometimes noted
small clumps of cells in the P19 parental line, but not to the same degree
found in the transfected lines. As a possible remedy for this, we recloned
our P19 line from single cells. We isolated a euploid clone that exhibited
model differentiation properties to use as the background for subsequent
experiments. Isolation and characterization of this clone, P19-A4, is
described in detail in Appendix One. Second, in an attempt to boost the
expression of the antisense element, possibly resulting in more tau protein
inhibition, we chose another expression vector, pCXN2, which was reported
to give high level protein expression in P19 cells (Fukuchi et al., 1992).
The pCXN2-derived constructs are shown in Figure 2-2 D and E. In
this configuration, the anti-sense element and the neomycin resistance
gene are under the control of separate promoters. The antisense element is
driven by the strong 1-actin promoter and the weaker thymidine kinase
promoter drives the expression of a weakened neomycin resistance allele.
In principle, these features require a high copy number of integrated
pCXN2 to achieve neomycin resistance which could lead to higher
antisense RNA expression. We cloned the tau antisense element, similar
to that used for the pGKBA vector, into the pCXN2 vector as described in
Materials and Methods. We transfected the P19-A4 parental line and
selected for stable transfectants in 600 g/ml G-418. We isolated and
characterized 8 tau antisense and 7 vector lines from this transfection.
We carried these lines through the RA differentiation protocol as
described in Materials and Methods. Representative lines, 3 days after
plating to tissue culture plastic, are shown in Figure 2-5. Again, some of
the tau antisense lines (4/8) showed varying degrees of clumpiness. Figure
2-5 A-C shows a range of tau antisense line behavior. However, none of
these lines were impaired in cell-substratum adhesion to the degree of
TA3A (Fig. 2-5 D). Control vector lines also showed a clumpy culture
appearance (5/7) that could be clearly distinguished from untransfected
cells (compare Fig. 2-5 E and F). As with the previous set of lines, the
clumpy phenotype seemed to be a property of having gone through the
clonal selection process. We obtained similar results from a transfection of
pGKBA-derived tau antisense and control plasmids in the new P19-A4
background. Some of both tau antisense (2/5) and vector lines (3/5) exhibited
52
a clumpy RA-induced phenotype and no lines resembling TA3A were
isolated (data not shown).
We examined the levels of tau protein expression in pCXN2-derived
lines as described above for pGKBA-derived lines. Figure 2-6 A shows a tau
western blot of protein extracts probed for tau expression taken from RAinduced cell lines shown in Figure 2-5. All three antisense lines show
reduced levels of tau protein compared to the vector control.
Phoshporimager analysis of this blot demonstrated that the levels of tau
protein in antisense lines TAB12, TAB14, and TAB15 were comparable to or
lower than TA3A (Fig. 2-6B). Tau protein level in vector control line NB14
was similar to untransfected P19-A4.
Taken together, several points emerge from these experiments.
First, the clumpy phenotype does not segregate with expression of a
particular transgene, and is not correlated with reduced levels of tau. Both
TAB12 and TAB15 show an approximately five-fold reduction of tau, but
only TAB12 shows the clumpy phenotype. Second, the dramatic loss of cellsubstratum adhesion in TA3A is not strictly correlated with reduction of
tau expression. TAB12, TAB14, and TAB15 have similarly reduced levels of
tau protein but none display the dramatic loss of cell-substratum adhesion
seen in TA3A. Finally, inhibition of tau protein expression, at least at the
levels achieved here, does not grossly affect the expression of a
differentiated neuronal morphology in RA-induced P19 cells. For example,
TAB15 cultures, although only expressing a fraction of the wild-type level of
tau protein, are morphologically indistinguishable from P19-A4 cultures
and display numerous neurite bearing cells.
53
Figure 2-5. RA differentiation of pCXN2-derived cell lines. pCXN2-derived
cell lines were taken through the RA differentiation protocol as described in
Materials and Methods. Cultures were photographed 3d after plating onto
tissue culture plastic. A-C tau antisense lines: (A)-TAB12 (B)-TAB14 (C)TAB15 (D)- TA3A; (E) vector control line NB14; (F) untransfected P19A4.
54
Figure 2-6. Tau protein expression in RA-induced pCXN2-derived cell
lines. pCXN2-derived cell lined were taken through the RA differentiation
protocol as described in Materials and Methods. Protein was harvested 3d
after plating onto tissue culture plastic and analyzed for tau expression as
described in Materials and Methods. (A) Tau western blot. Lane 1-NB14;
lane 2-TAB12; lane 3-TAB14; lane 4-TAB15; lane 5-P19-A4; lane 6-TA3A. (B)
Graphical representation of band intensities in (A) were determined by
Phosphorimager analysis as described in Materials and Methods.
56
(A)
.50
. 33
1
2
3
4
5
6
(B)
120-
100
80
60
40
20m
NB14
TAB12
TAB14
TAB15
P19 cell line
P19-A4
TA3A
DISCUSSION
Reduced levels of tau protein do not affect expression of a differentiated
neuronal morphology in RA-induced embryonal carcinoma P19 cells. We
have studied the function of the tau protein, taking advantage of the
properties of P19 cells. These cells, after exposure to RA, characteristics of
neuronal differentiation, such as extension of neurites and expression of
neuron-specific genes, including the tau gene itself. We have shown that
P19 cells constitutively expressing tau antisense RNA have reduced levels
of tau protein - at most a five-fold reduction- following RA-induced
differentiation. In the face of this reduced level of tau protein, P19 cells are
able to elaborate neurites and appear indistinguishable from control
cultures.
We have not ruled out more subtle effects of diminution of tau
expression on the establishment of a differentiated neuronal phenotype.
Although tau antisense and control cultures are similar in appearance,
there could be small differences in the number of neurite-bearing cells in
these cultures. Quantitative analysis would require immunocytochemical
staining to rigorously identify and count neuronal cell bodies. There might
also be qualitative differences in the neurons present in differentiated tau
antisense cultures. Possible differences include: length or caliber of
neurites, rate of growth of neurites, response of neurites to cytoskeletaldepolymerizing drugs, and electrophysiological properties of the neurons.
Answers to these questions await higher resolution analyses.
Perhaps the levels of tau inhibition achieved here are inadequate to
significantly impair tau function in P19 cells. One might reasonably expect
that a cell would be sensitive to quantitative changes in proteins that are
involved in cell structure (Katz et al., 1990). However, the modest decrease
in tau expression in these studies does not seem to affect neuronal
differentiation. More drastic inhibition of tau could produce a detectable
phenotype. Two different types of antisense expression vectors yielded a
similar five-fold maximum reduction of tau protein. The antisense element
itself was similar in design to that used previously to target MAP2
(Dinsmore and Solomon, 1991). In that study, ten-fold reductions in MAP2
protein levels were achieved. It is unclear how to increase the efficacy of
the tau antisense RNA inhibition strategy used here. Some genes might be
refractory to this type of inhibition. Another study utilizing stable
58
expression of antisense RNA to interfere with tau expression in PC12 cells
only managed a 23% reduction of tau protein relative to controls (EsmaeliAzad et al., 1994).
Tau function assayed by other experiments. Tau function in vivo
previously has been investigated by two sorts of experiments. Ectopic
expression in various non-neuronal cell types indicates that tau can
stabilize microtubules (Drubin and Kirschner, 1986; Kanai et al., 1989;
Lewis et al., 1989; Knops et al., 1991). More pertinent is the observation that
application of tau antisense oligonucleotides to cells committed to neuronal
differentiation interferes with the rate and extent of neurite extension
(Caceres and Kosik, 1990; Caceres et al., 1991; Hanemaaijer and Ginzburg,
1991). These inhibition experiments have been interpreted as consistent
with a role for tau in stabilizing the microtubules which underlie neurite
extension.
The differences between the results presented here and those
obtained from tau antisense oligonucleotide experiments with primary
neurons or PC12 cells might arise from differences in the experimental
systems. P19 cells are not committed to a single pathway of differentiation
and therefore may pass through earlier developmental states than
committed primary neurons or developmentally restricted PC12 cells. In
fact, both primary neurons and PC12 cells constituitively express tau,
whereas RA induces de novo synthesis of tau protein in P19 cells. Perhaps
the more primitive P19 cells possess a mechanism that can compensate for
decreased levels of tau. Alternatively, antisense oligonucleotides could be a
more effective agent at inhibiting tau expression. Unfortunately, a
comparison of the levels of tau inhibition among all the various
experimental paradigms cannot be made because this information was not
determined for the antisense oligonucleotide studies in primary neurons
and PC12 cells.
During these studies, Hirokawa and colleagues published their
results with the tau knockout mouse (Harada et al., 1994). Contrary to
expectations, these animals developed nervous systems and thrived as
adults. The only outstanding phenotype in these mice was a slight
reduction in the microtubule density in certain small-caliber axons.
Furthermore, hippocampal neurons derived from these mice had similar
59
neurite extension properties, microtuble stability, and microtubule
dynamics to control neurons when assayed in vitro. Although one is
tempted to conclude that tau function is largely dispensible at the
organismal level, the lack of a significant phenotype could reflect
compensation for a loss of tau by a redundant gene function. In support of
this notion, this same study documented an increase in MAPlA
expression.
Still, it is difficult to reconcile the results from genetic ablation of tau
in animals (and in cells derived from those animals) and inhibition of tau
expression in primary neurons and PC12 cells. The in vitro experiments
suggest a critical role for tau in neuriteogenesis that is not borne out in
vivo. The results presented here are in better agreement with the tau
knockout mouse. One possibility is that P19 cells better mimic neuronal
development as it occurs in vivo. That is, RA-induced P19 cells may pass
through similar developmental states in the culture dish that neuronal
precursors pass through in the embryo. If this is so, then the mechanism
that compensates for the loss of tau in mice may be the same one postulated
above to be operating in P19 cells.
Tau and MAP2: Divergent functions in P19 cells? Both tau and
MAP2 were originally identified on the basis of their ability to co-assemble
with microtubules cycled through successive rounds of polymerization and
depolymerization. Studies in cells argue for parallel functions for tau and
MAP2. Both proteins induce the formation of microtubule bundles in
transfected non-neuronal cells and process formation in moth ovary cells
(Knops et al., 1991; LeClerc et al., 1993). However, other work suggests
different roles for tau and MAP2. Antisense oligonucleotide inhibition of
MAP2 in primary cerebellar macroneurons blocks initial neurite
extension, whereas tau inhibition in these same cells blocks polarization of
neurites into axons and dendrites (Caceres and Kosik, 1990; Caceres et al.,
1992).
In RA-induced P19 cells, inhibition of MAP2 by antisense RNA has
two consequences: it blocks the expression of a differentiated neuronal
morphology and it blocks the typical decrease in cell proliferation that
accompanies RA-induction. One question that arises from this study is
whether this effect was specific to MAP2 or would inhibition of other MAPs
60
yield the same results. To examine this question, we inhibited tau
expression in P19 cells. The results presented above indicate that inhibition
of tau and MAP2 in P19 cells have distinct effects. Unlike MAP2, inhibition
of tau does not grossly affect expression of a differentiated neuronal
morphology and does not significantly affect the cell proliferative response
to RA.
There is a problem, though, in firmly establishing divergent roles for
these MAPs that is inherent in the antisense inhibition approach. As
mentioned above, the levels of inhibition achieved for tau in this study were
lower than those achieved for MAP2 previously. Therefore, it is formally ,
possible that the apparent differences in the consequences of tau and MAP2
inhibition may simply result from quantitative differences in the two
experiments. It will be interesting to see if a MAP2 knockout mouse will
shed light on the issue of functional divergence between tau and MAP2.
Unusual cell-substratum adhesion behavior in P19 cell lines. During
this work, we found that nearly half of all the stably transfected P19 cell
lines we isolated exhibited what we call a clumpy phenotype. After plating
into tissue culture dishes after RA induction, cells initially appear to
adhere to one another better than they adhere to the substratum. With time
(3-6 days), these clumps spread out and neurite-bearing cells become
readily apparent. The tendency to clump after RA-induction was a stable
property of P19 cell lines- similar behavior for individual lines was
reproducible over many RA inductions. Parental lines and other
transfected lines did not show the same degree of clumping behavior,
although some small clumps of cells could be observed in these lines. The
basis for this behavior is not understood, but it was not correlated with the
level of tau protein expression. Speculation as to the nature of the clumpy
behavior is further complicated by the fact that it was present at the same
frequency in transfectants derived from the original P19 parental stock and
a euploid clone from those cells. Finally, others, both in our laboratory and
in published work have not previously reported this tendency of stably
transfected P19 cells. One possible explanation for this artifact is that some
P19 cells in the lab parental stock, including euploid cells in that stock,
carried a genetic element that conferred a modest selective advantage over
other cells when they were challenged with neomycin. This same element
61
might then confer the clumpy phenotype when the stably transfected were
induced to differentiate with RA. Obviously, we can draw no firm
conclusions here.
The tau antisense cell line TA3A at first appeared to be an extreme
case of clumpiness. Following plating to tissue culture plastic after RAinduction, this line did not adhere at all to the solid substratum and instead
the cells adhered to one another to form aggregates. Although this line
showed reduced levels of tau protein, other tau antisense lines showed
similarly reduced tau protein levels but not the striking behavior after RA
induction. Out of more than 40 P19 cell lines isolated and characterized in
this study, plus many more characterized previously in the lab, only two
lines have shown this profound loss of cell-substratum adhesion- TAlA and
TA3A. However, Southern blot analysis revealed that these two lines had
identically integrated the tau antisense vector. The most probable
interpretation of this result is that these lines are derived from the same
genomic integration of the tau antisense vector. Therefore we really only
have one example of this phenotype to date. Although it is still possible that
a reduced tau level is a co-factor in the RA-dependent loss of cellsubstratum adhesion, we conclude that this phenotype resulted from a
spontaneous mutation in the TA3A line. We have characterized this
mutant phenotype further in Appendix Two.
62
CHAPTER THREE:
Molecular dissection of radixin Distinct and interdependent
functions of the amino- and carboxy-terminal domains.
63
SUMMARY
The ERM proteins - ezrin, radixin, and moesin - occur in particular
cortical cytoskeletal structures. Several lines of evidence suggest that they
interact with both cytoskeletal elements and plasma membrane
components. Here we describe the properties of full-length and truncated
radixin polypeptides expressed in transfected cells. In stable transfectants,
exogenous full-length radixin behaves much like endogenous ERM
proteins, localizing to the same cortical structures. However, the presence
of full-length radixin or its carboxy-terminal domain in cortical structures
correlates with greatly diminished staining of endogenous moesin in those
structures, suggesting that radixin and moesin compete for a limiting
factor required for normal associations in the cell. The results also reveal
distinct roles for the amino- and carboxy-terminal domains. At low levels
relative to endogenous radixin, the carboxy-terminal polypeptide is
associated with most of the correct cortical targets except cleavage furrows.
In contrast, the amino-terminal polypeptide is diffusely localized
throughout the cell. Low level expression of full-length radixin or either of
the truncated polypeptides has no detectable effect on cell physiology.
However, high-level expression of the carboxy-terminal domain
dramatically disrupts normal cytoskeletal structures and functions. At
these high levels, the amino-terminal polypeptide does localize to cortical
structures, but does not affect the cells. We conclude that the behavior of
radixin in cells depends upon activities contributed by separate domains of
the protein, but also requires modulating interactions between those
domains.
64
iNTRODUCTION
Many experiments demonstrate that the cytoskeleton influences the
topography, behavior and organization of the plasma membrane. However,
the detailed molecular basis for this regulation is not clearly understood. A
crucial part of solving this problem is the identification of proteins that
mediate interactions between components of the plasma membrane and the
cytoskeletal polymers. The highly related ezrin, radixin and moesin
proteins - the ERM proteins - are good candidates for molecules that play
such a role. Each of these proteins was first isolated from distinct tissues
(Bretscher, 1983; Tsukita et al., 1989; Lankes, et al., 1988). Antisera against
these proteins show that they occur in cellular domains marked by a close
juxtaposition of the plasma membrane and underlying cytoskeleton. Such
domains include microvilli in cultured cells and intestinal epithelia
(Bretscher, 1983), the marginal band of nucleated erythrocytes (Birgbauer
and Solomon, 1989), filopodia and lamellipodia in migrating cells
(Birgbauer, 1991), neuronal growth cones (Goslin et al., 1989; Birgbauer et
al., 1991), and cleavage furrows in dividing cells (Sato et al., 1991).
Molecular cloning of the genes encoding the ERM proteins (Gould et
al., 1989; Funayama et al., 1991; Lankes et al., 1991) demonstrated that they
share 70% overall amino acid identity, and that their amino-termini are
about 35% identical to the amino-terminus of Band 4.1 (Conboy et al., 1986;
see Fig. 1-1). The inclusion of the ERM proteins and related proteins, like
merlin which is thought to be involved in neurofibromatosis type II
(Trofatter et al., 1993; Rouleau et al., 1993), in the band 4.1 superfamily
(Rees et al., 1990) is suggestive of a membrane-cytoskeletal linker role.
Some evidence indicates that Band 4.1 links the plasma membrane and
underlying cytoskeleton in human erythrocytes (Anderson and Lovrien,
1984), and, that its amino-terminal domain may be important for
interaction with the integral membrane protein glycophorin (Leto et al.,
1986).
In addition to their cortical localization, there is also direct
experimental evidence indicating involvement of ERM proteins in plasma
membrane-cytoskeleton interactions. First, a drug interference
experiment shows that depolymerization of microtubules disrupts the
localization of ERM proteins to cortical structures of growth cones (Goslin
65
et al., 1989; Ch. Gonzalez Agosti, unpublished observations). Second,
transiently expressed polypeptides representing the amino- and carboxyterminal portions of ezrin localize to cell surface microvilli (Algrain et al.,
1993). Third, a functional test of ERM proteins suggests a role consistent
with such bivalent interactions: diminution of ERM levels by application of
anti-sense oligonucleotides affects both cell-cell and cell-substratum
adhesion (Takeuchi et al., 1994). Other studies have identified putative
ERM protein binding partners in the plasma membrane and cytoskeleton.
All three ERM proteins co-immunoprecipitate with the integral membrane
protein CD44 (Tsukita et al., 1994) in several cell types, and the carboxyterminal domain of ezrin interacts with F-actin, but not G-actin, in vitro
(Turunen et al. 1994).
Although the evidence above is consistent with ERM proteins acting
as molecular links between the cytoskeleton and plasma membrane, what
purposes these links serve, how they are established and maintained, or
whether ERM proteins are involved in processes other than physically
connecting these two organelles are outstanding questions. Their presence
in both dynamic structures such as the protrusive lamellipodia and
comparatively stable structures such as the marginal band, suggests that
the interactions between ERM proteins and cortical binding partners are
under tight spatial and temporal control. For instance, Bretscher (1989)
noted that ezrin was rapidly phosphorylated and recruited to membrane
ruffles in epidermal growth factor-stimulated A431 cells, but the
mechanisms that regulate this sort ERM protein behavior are still
unknown.
It has been suggested that different ERM proteins might occupy
structurally or functionally distinct subcellular domains (Sato et al., 1992).
However, other studies, including the present one, indicate that ERM
proteins localize to identical subcellular domains (Franck et al., 1993;
Winckler et al., 1993). That such highly similar proteins occupy the same
subcellular locales raises the possibility the these proteins may have at least
partially overlapping functions. Indeed, in the anti-sense inhibition
experiment mentioned above, the cell adhesion phenotypes are manifest
only when levels of all three ERM proteins are reduced in concert; although
there is also some evidence in this study that the roles for the ERM proteins
may not be entirely similar (Takeuchi et al., 1994).
66
With these issues in mind, we describe an analysis of exogenously
expressed radixin polypeptides in cultured cells. The results demonstrate
that exogenous radixin can behave like endogenous ERM proteins,
localizing to the same cortical structures. We also show evidence that
suggests that radixin and moesin share an important common binding
partner, and that these proteins may be functionally interchangeable.
These properties of the full-length radixin molecule are mimicked by low
levels of the carboxy-terminal domain of the protein, but not the aminoterminal domain. At higher levels, the carboxy-terminal domain in the
absence of the amino-terminal domain profoundly alters cell morphology
and division; phenotypes that may reflect functions for ERM proteins. Our
results demonstrate the distinct activities associated with domains of
radixin as well as crucial interactions between those domains.
67
MATERIALS AND METHODS
Cell culture. NIH-3T3 cells were maintained in DME supplemented
with 10% bovine calf serum (HyClone, Logan, UT). HtTA-1 cells (a HeLa
cell line derivative stably expressing the tetracycline-repressible
transactivating element (Gossen and Bujard, 1992) were obtained with
permission from Dr. Hermann Bujard, (University of Heidelberg;
Heidelberg, Germany) from the laboratory of Dr. Hans-Martin Jack (Loyola
University; Chicago, USA). Isolation and characterization of HtTA-1 cells
is described in Damke et al. (1995). HtTA-1 cells were maintained in DME
supplemented with 10% fetal bovine serum (HyClone, Logan, UT), and
400 g/ml G-418 sulfate (Geneticin; Gibco/BRL; Gaithersburg, MD). All
cells were incubated at 37*C under 5% C02 in a humidified chamber and
routinely subcultured.
DNA constructs and transfection. We introduced the influenza
hemagglutinin (HA) epitope tag (YPYDVPDYA; Field et al., 1988) onto the
amino-and carboxy-terminus of both full-length and truncated forms of the
murine radixin coding sequence in the following manner. First, we
synthesized two double-stranded oligonucleotides containing the following
features, in a 5' to 3' orientation, in the same translational reading frame:
oligonucleotide #1- an initiation codon (which is part of a 5' NcoI-site in
both oligonucleotides #1 and #2), followed by the nucleic acid sequence
encoding the HA-epitope, the 6-base XhoI cloning site, and finally, a stop
codon; oligonucleotide #2- an initiation codon , followed by the 6-base XhoI
cloning site, the nucleic acid sequence encoding the HA-epitope, and
finally, a stop codon. Also, to maintain the proper reading frame in these
oligonucleotide sequences with respect to the initiation codon, a codon for
alanine was inserted at what becomes residue #2 in the translated protein.
All oligonucleotides used in this study were synthesized in the MIT
Biopolymers Facility. Using T4 DNA ligase, we inserted each of these
oligonucleotides into the NcoI-BamHI-digested pMFG retroviral vector
(Dranoff et al., 1993) obtained from Dr. Richard Mulligan, MIT. The
resulting plasmids were named pMFG-HAN (derived from oligonucleotide
#1) and pMFG-HAC (derived from oligonucleotide #2). All enzymes
utilized for recombinant DNA work were obtained from New England
68
Biolabs (Beverly, MA) and used according to the manufacturers
specifications. Next, we prepared murine radixin coding sequences for
insertion into pMFG-HAN and pMFG-HAC by polymerase chain reaction
mediated amplification (GeneAmp PCR Kit; Perkin-Elmer, Branchburg,
NJ), with a Perkin-Elmer DNA thermal cycler, from the pR2ESS plasmid
(containing a cDNA clone of murine radixin; obtained from Dr. Akira
Nagafuchi; National Institute for Physiological Sciences; Okazaki, Japan
(Funayama et al., 1991)). The following oligonucleotide primer pairs were
used to direct the synthesis of full-length and truncated forms of radixin
while simultaneously adding XhoI cloning sites in-frame with, and directly
apposed to, the 5'- and 3'- ends of the radixin coding sequence: full-length
(residues 2-583)- Primer #15'CCGCTCGAGCCGAAGCCAATCAATGTAAG and Primer #25'CCGCTCGAGCATGGCTTCCAACTCATCG; amino-terminus (residues
2-317)- Primer #1 and Primer #3-5'CCGCTCGAGCTGCTTCTGATGCAAAACC; carboxy-terminus (residues
318-583)- Primer #2 and Primer #4-5'CGGCTCGAGCTAGAAAGGGCACAATTAG. After treatment with T4
polynucleotide kinase, the resulting polymerase chain reaction products
were ligated into the EcoRI site of pUC19 made blunt by treatment with
Klenow enzyme and dephosphorylated by treatment with calf intestinal
phosphatase. The resulting plasmids pUC-RAD (full-length), pUC-RADN
(amino-terminus), and pUC-RADC (carboxy-terminus) were cut with XhoI
and the radixin inserts were separated from the pUC vector on an agarose
gel and purified with a Qiaex gel extraction kit (Qiagen; Chatsworth, CA).
These three radixin fragments with XhoI sites on their 5'- and 3'- ends
were inserted into the unique XhoI site adjoining the HA-epitope sequence
in pMFG-HAN and pMFG-HAC and those ligants containing inserts in the
proper orientation were identified by diagnostic restriction enzyme digests.
Note: Due to the extra sequence added by the XhoI site, insertion of the
radixin fragments into pMFG-HAN and pMFG-HAC results in the addition
of a leucine and a glutamate residue immediately flanking both the aminoand carboxy- termini of the radixin coding sequence. The resulting six
plasmids, representing the full-length, and amino- and carboxy- terminal
portions of murine radixin bearing the HA-epitope tag at either their
amino- or carboxy- termini, were designated pMFG-HAN-RAD; pMFG69
HAN-RADN; pMFG-HAN-RADC; pMFG-HAC-RAD; pMFG-HAC-RADN;
pMFG-HAC-RADC (the names of these plasmids correspond to the
constructs shown in Fig. 3-2A). The integrity of the junctions between the
radixin coding sequence and the flanking sequence containing the HAepitope was confirmed by DNA sequencing (Sequenase Version 2.0 DNA
Sequencing Kit; United States Biochemical Co.; Cleveland, OH).
For expression of the HA-radixin constructs in HtTA-1 cells, we
digested pMFG-HAN-RAD; pMFG-HAN-RADN; pMFG-HAN-RADC;
pMFG-HAC-RAD; pMFG-HAC-RADN; and pMFG-HAC-RADC separately
with NcoI, treated the linearized plasmids with Klenow enzyme to blunt the
NcoI ends, digested with BamHI, and isolated those fragments, containing
the radixin coding sequence plus the HA-epitope tag, by purification from
an agarose gel. These fragments were ligated into the multiple cloning site
of the tetracycline-regulatable expression plasmid pUHD 10-3 (kindly
provided by Dr. Hermann Bujard) that we prepared by digestion with
EcoRI, treatment of the linearized plasmid with Klenow enzyme to blunt
the EcoRI ends, digestion with BamHI, and purification on an agarose gel.
Insertion of HA-radixin constructs into pUHD10-3 was confirmed by
diagnostic restriction enzyme digests. The resulting plasmids were named
pUHD-HAN-RAD; pUHD-HAN-RADN; pUHD-HAN-RADC; pUHD-HACRAD; pUHD-HAC-RADN; and pUHD-HAC-RADC. The polypeptides
encoded by these plasmids are identical to those expressed from the pMFGderived plasmids.
Both stable and transient transfection protocols were modifications of
the calcium phosphate transfection protocol of Graham and Van der Eb
(1973). Plasmid DNAs were prepared from Qiagen columns. For stable
transfection of NIH-3T3 cells, we co-transfected pMFG-HA radixin
constructs with pSV2-neo (Southern and Berg, 1982) at a 10 to 1 molar ratio
respectively. We selected stably transfected NIH-3T3 cells in the presence of
600 gg/ml G-418 sulfate. The resultant colonies were picked, expanded in
the continuous presence of the selecting agent, and stored in liquid
nitrogen. Cell lines derived in this manner were periodically passaged
through medium containing 600 pg/ml G-418 sulfate. HtTA-1 cells were
transfected as described by Damke et al. (1995).
70
Antibodies. Polyclonal antibodies #464, #457, and #454 raised against
unique peptides from murine ezrin, radixin, and moesin, respectively,
were described previously (Winckler et al., 1994). They were affinity
purified as described in Winckler et al. (1994) except that anti-radixin
antibody #457 was affinity-eluted from the radixin-immunoreactive band of
chicken erythrocytes. Monoclonal antibody 12CA5 (Niman et al., 1983) was
obtained from Berkeley Antibody Co. (Richmond, CA)
Protein extracts and immunoblotting. We prepared whole cell
extracts from cultured cells by first lysing the subconfluent monolayers in
PBS containing 2% SDS and a protease inhibitor cocktail consisting of:
0.04U/ml Aprotinin, lg/ml PMSF, lg/ml leupeptin, and 1gg/ml pepstatin
(Sigma Chemical Co.; St. Louis, MO). An aliquot of this lysate was
reserved for determination of protein concentration by the method of Lowry
(1951) using a detergent compatible analysis system from Bio-Rad (Melville,
NY). To the remaining lysate, we added Laemlli sample buffer (Laemlli,
1970) and boiled this mixture for 5 minutes.
For immunoblotting, we separated protein samples on a 7.5%
polyacrylamide gel with a 5% stacker according to the method of Laemlli
(1970). Proteins were electrophoretically transferred to 0.2 gm
nitrocellulose filters (Schleicher and Schuell; Keene, NH) essentially as
described by Tobin et al. (1979). Transfer of protein to filter and equivalency
of loads was confirmed by staining with 0.2% Ponceau S (Sigma) in 3%
TCA. For blotting with antibodies #464, #457, and #454, the nitrocellulose
filters were first blocked with 5% BSA (Sigma) in TBS supplemented with
0.1% Tween-20 and 0.05% sodium azide for two hours. Next, the affinitypurified antibodies were diluted 1:100 into blocking buffer and incubated
overnight at room temperature, washed extensively in TBS supplemented
with 0.1% Tween-20, incubated one hour in blocking buffer containing a
1:1,000 dilution of 1 2 5 1-labeled Protein A (DuPont-New England Nuclear;
Boston, MA), and washed in TBS supplemented with 0.1% Tween-20 again.
For blotting with mAb 12CA5, the nitrocellulose filters were first blocked
with 5% Nonfat Dry Milk (Nestle Food Company; Glendale CA) in TBS
supplemented with 0.1% Tween-20 and 0.05% sodium azide overnight.
Next, the mAb 12CA5 was diluted 1:1,000 into blocking buffer and incubated
one hour at room temperature, washed extensively in TBS supplemented
71
with 0.1% Tween-20, incubated one hour in blocking buffer containing a
1:10,000 dilution of horseradish peroxidase-conjugated sheep anti-mouse
IgG F(ab')2 (Amersham; Arlington Heights, IL) and washed in TBS
supplemented with 0.1% Tween-20 again. 1251 signal was detected
autoradiographically on DuPont-NEN Reflection Film at -70*C using a
DuPont Reflection intensifying screen and quantified using a
PhosphorImager (Molecular Dynamics; Sunnyvale, CA). Horseradish
peroxidase signal was detected by enhanced chemiluminescence with the
Lumiglo reagent kit (Kirkegaard and Perry Laboratories; Gaithersburg,
MD) on DuPont-NEN Reflection Film at room temperature.
Indirect immunofluorescence microscopy. Cells grown on glass
coverslips were fixed for 30 minutes in PBS containing 4%
paraformaldehyde, permeablized for 10 minutes in PBS containing 0.5%
NP-40, and washed three times in PBS. For immunostaining, fixed,
permeablized cells were first blocked in PBS containing 10% Normal Goat
Serum (Vector Laboratories Inc.; Burlingame, CA). Then, the cells were
incubated with the appropriate primary antibody diluted in PBS containing
1% BSA for 30 minutes at 37 0 C, washed three times in PBS, incubated with
the appropriate secondary antibody - FITC-conjugated goat anti-rabbit IgG
F(ab')2 (Tago; Burlingame, CA) for antibodies #464, #457, and #464 or
rhodamine or FITC-conjugated goat anti-mouse IgG F(ab')2 (Organon
Teknika; Durham, NC) for mAb 12CA5 - diluted in PBS containing 1% BSA
for 30 minutes at 37*C, and washed three times in PBS. In some
experiments, during secondary antibody incubation, cells were stained
with rhodamine-conjugated phalloidin (Molecular Probes; Eugene, OR)
and/or DAPI (4,6-Diamidino-2-phenylindole; Sigma) to reveal F-actin and
DNA respectively. The coverslips were mounted onto glass slides using
Gelvatol mounting medium containing an anti-fade agent - 15 mg/ml 1,4diazabicyclo[2,2,2] octane (Aldrich Chemical Co.; Milwaukee, WI).
Cells were examined by conventional microscopy on a Zeiss Axioplan
microscope (Carl Zeiss Inc.; Thornwood, NY) using 63X 1.4 N.A. and 100X
1.3 N.A. objectives. Images were recorded on Kodak Tri-X-Pan 400 film
(Eastman Kodak Co.; Rochester, NY). For confocal microscopy cells were
examined with a Bio Rad MRC 600 scanning laser confocal microscope.
FITC- labeled second antibody and rhodamine- labeled phalloidin were
72
detected with a yellow high sensitivity filter and blue high sensitivity filter
respectively. Images taken from a video monitor were recorded on Kodak
Plus-X-Pan 125 film.
73
RESULTS
Expression and localization of ERM proteins in NIH-3T3 cells.
Winckler et al. (1994) previously showed that NIH-3T3 cells express each of
the ERM proteins. Antisera generated against unique peptides from the
predicted sequence of ezrin (antibody #464), radixin (antibody #457) and
moesin (antibody #454) each recognize a distinct band in extracts of NIH3T3 cells. By immunofluorescence, the antibodies against radixin and
moesin brightly stain cortical structures such as microvilli and filopodia as
well as the cleavage furrows of dividing cells (Fig. 3-1 B-H). The antibodies
also stain lamellipodia and ruffling edges, although less intensely (data not
shown). These localizations are consistent with those reported in other cell
types (Bretscher, 1983; Franck et al.; 1993; Sato et al., 1992). Anti-ezrin did
not give detectable staining in these cells (Fig. 3-1 A). This result may
represent different localization or lower levels of ezrin, but it is not a
property of the anti-ezrin antiserum, which gives definitive staining in
another cell type (Winckler et al., 1994). Thus, NIH-3T3 cells express each
of the ERM proteins, and two of them - radixin and moesin - localize in
patterns indistinguishable from one another.
Stable expression of epitope-tagged forms of radixin in NIH-3T3 cell
lines. We generated constructs encoding the full-length protein and its
amino- and carboxy-terminal portions, fused to the HA-epitope. The
division into amino- and carboxy-terminal polypeptides occurs at codon 318,
so the amino-terminal polypeptide contains all of the sequence homologous
to band 4.1. The inclusion of the epitope tag allows us to distinguish
immunologically the exogenous from the endogenous gene products.
Versions of the constructs bearing the HA-epitope sequence at either their
amino- or carboxy-termini (see Materials and Methods) allow us to control
for effects of the extra sequence on the conformation or function of the
proteins. Figure 3-2 A shows a schematic diagram of the constructs.
To generate stable NIH-3T3 lines expressing these proteins, we cotransfected the pMFG retroviral vector, containing each of the radixin
constructs, and pSV2-neo and selected for resistance to G-418 sulfate (see
Materials and Methods). We recovered tens to hundreds of drug-resistant
colonies for each of the constructs in each of two independent transfections.
74
We picked five to ten drug-resistant colonies for each construct from each
transfection, expanded, and screened for expression of HA-radixin
constructs by immunoblotting with mAb 12CA5. As expected, the
expression levels varied somewhat among lines derived from each of the
constructs - as much as ten-fold - as assessed by western blots. Figure 3-2 B
shows such blots on extracts from lines expressing relatively high levels of
each of the six HA-radixin constructs. The apparent sizes of both the fulllength and truncated epitope-tagged constructs were as expected
considering the addition of the HA sequence. Previous reports documented
the anomalously high apparent molecular weight of both full-length ezrin
(Gould et al., 1989) and the carboxy-terminal domain of ezrin (Algrain et
al., 1993) on SDS-PAGE compared to that calculated from the predicted
amino acid sequence. Like ezrin, the apparent molecular weights of the
HA-tagged forms of both the full-length and the carboxy-terminal
polypeptide of radixin are higher (81 and 46 kD respectively) than their
calculated molecular weights (69 and 35 kD respectively).
The levels of expression of the two full-length constructs - HAC-RAD
and HAN-RAD, tagged at their carboxy- and amino-termini respectively
(Fig. 3-2 B, lanes 1, 2) - are comparable to one another. In contrast, the
levels of the truncated constructs were significantly lower than those of the
full-length constructs; for comparison, in Figure 3-2 B, lanes 3-7 were
exposed 20-times longer than lanes 1 and 2. At these longer exposures,
bands recognized by mAb 12CA5 but not dependent upon transfection
become apparent (Fig. 3-2 B, lane 7). We compared the relative levels of
exogenous radixin to the endogenous molecule in these stable lines by
blotting with antibody #457 and measuring the signal by PhosphorImager.
By this assay in the line with the highest expression level, the steady state
level of the full-length tagged proteins was approximately four-fold higher
than that of endogenous radixin (Fig. 3-2 C). Because expression levels of
the truncated HA-radixin constructs are much lower (more than 20-fold)
than the full-length HA-radixin constructs, we estimate that the expression
levels of the truncated HA-radixin polypeptides are at least five-fold lower
than the level of endogenous radixin.
In general, the cells expressing HA-radixin constructs were
indistinguishable from the parental NIH 3T3 line with respect to cell
morphology (see Figs. 3-3 and 3-4) and growth rate (data not shown). The
75
Figure 3-1. Localization of endogenous ERM proteins in NIH-3T3 cells.
NIH-3T3 cells were processed for indirect immunofluorescence with antiezrin antibody #464 (A), anti-radixin antibody #457 (B-D), and anti-moesin
antibody #454 (E-H) as described in Materials and Methods. Anti-radixin
and anti-moesin specifically label cortical structures such as filopodia
(arrows in B and E), microvilli (arrow in F) and cleavage furrows (C,G).
Phase-contrast images of dividing cells in C and G are shown in D and H
respectively. Anti-ezrin is not detectable in any discrete structure (A). Bar:
20 gim.
76
Figure 3-2. Stable expression of HA-radixin constructs in NIH-3T3 cells.
(A) Diagram of HA-radixin constructs. We added the influenza
hemagglutinin (HA) epitope [1 to the amino- (HAN) and carboxy- (HAG)
terminus of the coding sequence of full-length murine radixin -amino acids
1-583- (RAD); the amino-terminus [0111-amino acids 1-318 corresponding to
the Band 4.1 homology- (RADN); and the carboxy-terminus [M -amino
acids 319-583 containing the putative F-actin binding domain- (RADC) as
described in Materials and Methods. (B) Western blot analysis of lysates
taken from stably transfected NIH-3T3 cell lines. 20 micrograms of total
protein taken from lines stably expressing HA-radixin constructs was
analyzed by immunoblotting with mAb 12CA5 directed to the HA-epitope as
described in Materials and Methods. Lane 1- HAC-RAD; Lane-2 HANRAD; Lane-3 HAN-RADN; Lane 4-HAC-RADN; Lane 5-HAN-RADC; Lane
6-HAC-RADC; Lane 7-Untransfected NIH-3T3. The film used for Lanes 3-7
was exposed 20-fold longer than the film for Lanes 1-2. Positions of the fulllength (FL), amino-terminal (N), and carboxy-terminal (C) polypeptides are
indicated to the right of the blot. Several species dependent on mAb 12CA5,
but unrelated to the HA-radixin polypeptides, appeared at the longer
exposure time used for Lanes 3-7. These species are also detectable in
lysates from untransfected NIH-3T3 (Lane 7). Despite this background
immunoreactivity detected by Western blotting, immunostaining
untransfected NIH-3T3 cells with mAb 12CA5 gave a low background and
no discrete structures were stained (see Figure 3H). Positions of molecular
weight standards are indicated to the left of the blot. (C) Comparison of
endogenous vs. exogenous radixin expression in HACRAD expressing cell
line. Protein extracts were probed for radixin expression with antibody
#457 and the signal was quantitated as described in Materials and
Methods. Lane 1-untransfected NIH-3T3 cells. Lane 2- NIH-3T3 cells
expressing HAC-RAD.
78
A.
HAN-RAD
HAC-RAD
HAN-RADN
HAC-RADN
HAN-RADC
le
HAC-RADC
B.
135
81
i
44v-.
RD
FL
"
12
3
c.
1
2
45
-4N
67
one exception is that cultures expressing the carboxy-terminal polypeptides
contained multi-nucleated cells at a relatively high frequency. The
frequency does vary, however, among the individual lines, and in any case
occurs more robustly in cells expressing high levels of the carboxy-terminal
constructs (see below).
Subcellular localization of HA-radixin polypeptides in stably
transfected NIH-3T3 cell lines. We examined the localization of each of the
six HA-radixin polypeptides by immunofluorescence microscopy. The
HAC-RAD protein - the full-length molecule tagged on its carboxy-terminus
- is in the same cellular domains as endogenous radixin shown in Figure 31: in cortical structures such as cell surface microvilli, filopodia and
cleavage furrows and, less intensely, in ruffling edges and lamellipodia
(Fig. 3-3 A-D). However, the same full-length radixin polypeptide with the
epitope tag at its amino-terminus - HAN-RAD - displayed no distinct
localization, and instead was present diffusely in the cytoplasm (Fig. 3-3 EG). In particular, the HAN-RAD protein was not concentrated at the cell
cortex. Neither of the amino-terminal polypeptides, HAN-RADN and HACRADN, localized to discrete structures and, like HAN-RAD, appeared to be
diffuse throughout the cytoplasm (Fig. 3-4 A-C). Both of the carboxyterminal polypeptides, HAN-RADC and HAC-RADC, showed localization
patterns that were similar, but not identical, to that of HAC-RAD. Like the
full-length protein, the carboxy-terminal polypeptides were present in
cortical structures. Although, compared to either HAC-RAD or
endogenous radixin, they were more enriched in ruffling edges and
lamellipodia (Fig. 3-4 F). In two ways, however, the localization of the
carboxy-terminal polypeptides differed strikingly from HAC-RAD. First, in
some cells, these polypeptides co-align with linear, phalloidin positive
structures near the ventral surface of the cell (large arrows, Fig. 3-4 G and
H) that likely represent stress fibers. This result is consistent with the
localization of the carboxy-terminal polypeptide of ezrin in transient
transfection experiments (Algrain et al., 1993). Second, we could not detect
HAN-RADC and HAC-RADC in cleavage furrows (compare Fig. 3-4 D and
E with Fig. 3-3 C and D). We examined at least six lines that showed
expression by Western blot for the full-length and truncated HA-radixin
constructs. The localization characteristics described above for each
80
polypeptide were the same in all independently isolated lines examined.
We also stably expressed HA-radixin constructs in P19 embryonal
carcinoma cells. In these cells the localization properties of the
polypeptides were the same as in NIH-3T3 cells (see Appendix Three).
Displacement of endogenous moesin by epitope-tagged radixin
polypeptides. Since moesin was localized to the same cortical structures in
untransfected cells that the HA-radixin proteins were localizing to in the
stably transfected cell lines wished to know the disposition of moesin in
these lines. Taking advantage of the fact that the endogenous moesin and
the products of the transfected genes are distinguishable by their reactivity
with different antibodies, we performed double label immunofluorescence
on the stably transfected cell lines. The results demonstrate that those
radixin polypeptides that do localize to cortical structures affect the
behavior of endogenous moesin. As shown in Figure 3-5, cells expressing
HAC-RAD and HAC-RADC, both of which show typical ERM localization,
exhibit a substantial diminution of anti-moesin staining in microvilli and
filopodia (Fig. 3-5 A, B, G, and H). A similar, but less dramatic, reduction
was observed in cells expressing HAN-RADC (data not shown). Moesin
localization in HAN-RAD, HAN-RADN, and HAC-RADN lines is
quantitatively indistinguishable from untransfected cells (Fig. 3-5 C-F). In
addition, moesin localization in cleavage furrows was also diminished in
cells expressing the full-length polypeptide HAC-RAD (Fig. 3-6 A).
However, cells expressing HAN-RADC or HAC-RADC, which themselves
do not localize to cleavage furrows, showed normal moesin staining in that
structure (Fig. 3-6 C). These apparent qualitative differences among the
different transfected constructs are not due to a quantitative difference in
moesin levels in those lines. Western blotting with antibody #454
demonstrated that all lines expressing HA-radixin constructs maintain
similar amounts of moesin, and showed no evidence of its down-regulation
(Fig. 3-7).
Consequences of transient expression of HA-radixin constructs in
HeLa cells. Some of the results reported above for the amino- and carboxyterminal domains of radixin in stably transfected NIH-3T3 cells differ from
those reported for the corresponding domains of ezrin in transiently
81
Figure 3-3. Localization of full-length HA-radixin polypeptides in NIH-3T3
cell lines. NIH-3T3 cell lines stably expressing HAC-RAD (A-D) or HANRAD (E-G) and untransfected NIH-3T3 cells (H) were processed for indirect
immunofluorescence with mAb 12CA5 as indicated in Materials and
Methods. A and B are, respectively, ventral and dorsal focal planes of the
same cell. HAC-RAD is localized to cortical structures including: filopodia
(small arrow in A), ruffling edges (large arrow in A), microvilli (arrow in
B) and to cleavage furrows (C). HAN-RAD does not localize to discrete
structures in either non-dividing cells (E) or dividing cells (F). D and G are
phase contrast images of dividing cells shown in C and F. No specific
staining is detectable in untransfected cells (H). Bar: 20gm.
82
r
i
'
%
Figure 3-4. Localization of truncated HA-radixin polypeptides in NIH-3T3
cell lines. NIH-3T3 cell lines stably expressing HAN-RADN (A-C) or HACRADC (D-H) were processed for indirect immunofluorescence with mAb
12CA5 and phalloidin as indicated in Materials and Methods. HAN-RADN
does not localize to discrete structures in either non-dividing cells (A) or
dividing cells (B). F and G are, respectively, dorsal and ventral focal planes
of the same cell. Results similar to those shown for HAN-RADN were
obtained for lines expressing HAC-RADN (data not shown). HAC-RADC is
localized to cortical structures including: microvilli (small arrow in F),
ruffling edges (large arrow in F), filopodia (small arrow in G). Staining in
ruffling edges is much more intense in lines expressing carboxy-terminal
constructs. HAC-RADC also associates with linear elements in the ventral
cytoplasm (large arrow in G) which co-align with F-actin containing
microfilaments revealed by phalloidin staining (large arrow in H). HACRADC does not localize to cleavage furrows in dividing cells (D). H is the
rhodamine channel showing phalloidin staining in the same focal plane as
G. C and E are the phase contrast images of the dividing cells shown in B
and D. Results similar to those shown for HAC-RADC were obtained for
lines expressing HAN-RADC (data not shown). Bar: 20gm.
84
-........
.F
Figure 3-5. Displacement of endogenous moesin from cortical structures by
HA-radixin polypeptides in NIH-3T3 cells. Lines expressing HA-radixin
constructs were processed for indirect immunofluorescence double-labeled
with anti-moesin Ab#454 (A,C,E,G) and anti-HA mAb 12CA5 (B,D,F,H).
Anti-moesin staining is dramatically reduced in cortical structures in cells
expressing HAC-RAD (A,B) and HAC-RADC (G,H). Moesin localization in
cortical structures was normal in cells expressing HAN-RAD (C,D) and
HAN-RADN (EF).
86
Figure 3-6. HAC-RAD, but not HAC-RADC, displaces endogenous moesin
from cleavage furrows in NIH-3T3 cells. Lines expressing HA-radixin
constructs were processed for indirect immunofluorescence double-labeled
with anti-moesin Ab#454 (A,C) and anti-HA mAb 12CA5 (B,D). Dividing
cells expressing HAC-RAD (A,B) and HAC-RADC (C,D) are shown. Antimoesin staining is dramatically reduced in cleavage furrows of cells
expressing HAC-RAD, but not in cells expressing HAC-RADC.
Figure 3-7. Moesin protein expression in NIH-3T3 cell lines expressing
HA-radixin proteins. The level of moesin expression in NIH-3T3 cell lines
was determined by Western blot analysis with antibody #454 as described in
Materials and Methods. The identities of the samples are indicated in the
figure. Samples from HA-radixin expressing cell lines were taken from
two independently isolated lines.
88
I
cGNH[
t
N~rVU-NV
DaVU-NV4
CJW-HIN
transfected CV-1 cells (Algrain et al., 1993). These differences might arise
from differences in expression levels of the radixin and ezrin polypeptides
in the two experiments. We noted that there were some indications that
stable expression of high levels of the amino- and/or carboxy-terminal
polypeptides might not be tolerated by the cells. First, the highest
expression level of each of these polypeptides in stable lines was
significantly less than 20% that of the full-length proteins. Second, as noted
above, we detected an abnormally high incidence of multi-nucleated cells in
HAN-RADC and HAC-RADC cultures. For these reasons, we examined
the effects of high-level, transient expression of full-length and truncated
forms of radixin.
We placed the various HA-radixin constructs under the control of the
tetracycline-repressible promoter developed by H. Bujard and co-workers
(Gossen and Bujard, 1992), and transfected them into HtTA-1 cells, a HeLa
cell line derivative stably expressing the essential tetracycline-sensitive
transactivator element (see Materials and Methods). Expression of radixin
and moesin is detectable by Western blotting extracts from HtTA-1 cells
with #457 and #454 antisera respectively (Fig. 3-8 A). There is less moesin
and radixin signal per gram of total protein in HtTA-1 cells as compared to
NIH-3T3 cells. This could be a reflection of lower expression levels of ERM
proteins in HtTA-1 cells. Ezrin expression cannot be ascertained in HtTA-1
cells because the human ezrin sequence differs from the mouse epitope that
antibody #465 was raised against. The localization patterns of moesin and
radixin in HtTA-1 cells were similar to those in NIH-3T3 cells (Fig. 3-8 B-E).
Cortical structures including filopodia, microvilli, and cleavage furrows
(see insets in Fig. 3-8 B-E) all showed immunoreactivity. In general, the
staining of these structures in HtTA-1 cells was less robust -fewer cells
with clearly positive staining structures- with the anti-moesin antibody
#454 than with the anti-radixin antibody #457.
We evaluated the expression of the HA-radixin constructs 48 hours
after transfection. Western blot analysis showed that these constructs were
expressed at approximately equal levels (Fig. 3-9 A). This result suggests
that the lower steady-state expression levels of the truncated HA-radixin
constructs, relative to the full-length versions, that we observed in stably
transfected NIH-3T3 cells are not due to inherent instability of the truncated
polypeptides. The transient expression level of HA-radixin constructs is
90
approximately 20-fold higher than the level of endogenous radixin in HtTA1 cells, as assessed by Western blotting and Phosphorimager analysis of
extracts from transfected cultures with antibody #457 (Fig. 3-9 B). Because
only 10-40% of cells in these transiently transfected cultures expressed the
HA-radixin constructs, the relative expression levels in these cells must be
even higher.
By immunofluorescence, HAC-RAD localizes to cortical structures,
like endogenous radixin (Fig. 3-10 A-C). HAN-RAD is diffuse in the
cytoplasm, even in brightly staining cells (Fig. 3-10 D-F). Thus, the results
in transiently transfected HtTA-cells expressing these full-length
constructs are consistent with those obtained from the stable transfectants
of NIH-3T3 cells expressing much lower levels of the full-length constructs.
However, in contrast to the results in the stable transfectants, both aminoterminal polypeptides (HAN-RADN and HAC-RADN) are concentrated in
cortical structures (Fig. 3-10 G-I). Transient, high-level expression of
either the full-length or amino-terminal HA-radixin constructs did not
result in any gross morphological alterations. There were, however,
striking consequences of expressing either of the carboxy-terminal
polypeptides at high levels. Many cells exhibited multiple long, tapered
processes covering the dorsal surface of the cells that were clearly
distinguishable by phase contrast microscopy. Other extensions that made
contact with the substratum like filopodia were much larger than similar
structures in normal cells (Fig. 3-10 J-L). The confocal micrographs shown
in Figure 3-11 better illustrate the three dimensional nature of these
processes. The processes stain intensely with both mAb 12CA5 and with
phalloidin, suggesting that they contain F-actin (Fig. 3-11 A-F, left hand
panel in each). On the contrary, microtubules were not enriched in these
processes and we only occasionally observed them in the wide bases of the
processes proximal to the cell soma. In fact, the microtubule arrays in
cells expressing HAN-RADC and HAC-RADC appeared similar to cells
expressing other HA-radixin constructs and non-expressing cells (data not
shown). In the same population, cells exhibiting weaker immunoreactivity
with mAb 12CA5 showed either modest versions of the processes or none at
all. We never observed these processes in mAb 12CA5 negative cells or in
cells expressing the full-length or amino-terminal HA-radixin constructs.
91
Figure 3-8. Endogenous ERM protein expression and localization in HtTA1 cells. (A) Western blot analysis of HtTA-1 and NIH-3T3 cell extracts with
anti-moesin antiserum #454 (blot on left) or anti-radixin antiserum #457
(blot on right). 25gg total protein was examined for each sample. The
radixin-staining doublet in the NIH-3T3 lane has been detected before with
antibody #457 (M. Magendantz, unpublished result) however, the nature of
the higher molecular weight species is unclear. Immunolocalization of
radixin with antibody #457 (B) and moesin with antibody #454 (D) in HtTA-1
cells as described in Materials and Methods. Both moesin and radixin
localize to cortical cytoskeletal structures (including cleavage furrows in
dividing cells (see insets)) in HtTA-1 cells. C and E are phase contrast
images of B and D respectively.
92
0
4
I
NIH-3T3
HtTA-1
NIH-3T3
HtTA-1
Figure 3-9. Transient expression of HA-radixin constructsin HtTA-1 cells.
(A) Lysates were harvested from HtTA-1 cells 48 hours after transfection
with HA-radixin constructs. At this time, 10-40% of the cells expressed
mAb 12CA5 immunoreactivity as judged by indirect immunofluorescence
microscopy. 20 micrograms of total protein was examined by Western blot
analysis with mAb 12CA5. Lane 1- HAN-RAD; Lane 2- HAC-RAD; Lane 3HAN-RADN; Lane 4- HAN-RADC; Lane 5- Untransfected HtTA-1 cells.
Positions of the full-length (FL), amino-terminal (N), and carboxy-terminal
(C) polypeptides are indicated to the right of the blot. The species migrating
just above the full-length protein is unrelated to expression of HA-radixin
constructs and is present in untransfected HtTA-1 cells (lane 5). Positions
of molecular weight standards are indicated to the left of the blot. (B)
Comparison of endogenous vs. exogenous radixin expression in HtTA-1
cells transfected with HACRAD. Protein extracts were probed for radixin
expression with antibody #457 and the signal was quantitated as described
in Materials and Methods. Lane 1-untransfected HtTA-1 cells. Lane 2HtTA-1 cells expressing HAC-RAD.
94
135
1
44
2
B.eM
1
2
3 4
Figure 3-10. Localization of HA-radixin polypeptides in transiently
transfected HtTA-1 cells. HtTA-1 cells transfected with HAC-RAD (A-C),
HAN-RAD (D-F), HAN-RADN (G-I), and HAC-RADC (J-L) were fixed 48
hours after transfection and examined by indirect immunofluorescence
microscopy with mAb 12CA5 as described in Materials and Methods. mAb
12CA5 immunoreactivity is shown for the same cell in a dorsal (A,D,G,J)
and ventral (B,E,H,K) focal plane and in phase contrast (C,F,I,L). HACRAD localizes to microvilli (arrow in A), filopodia (arrow in B). HAN-RAD
does not localize to discrete structures (D,E). IAN-RADN localizes to
dorsal microvilli (arrow in G) and to filopodia in contact with the
substratum (arrow in H). Results similar to those shown for cells
transfected with HAN-RADN were obtained for cells transfected with HACRADN (data not shown). Expression of HAC-RADC induces the presence of
long, tapered processes that can be seen in dorsal focal planes (arrow in J)
and in contact with the substratum (arrow in K). These processes are
clearly visible by phase contrast microscopy (arrow in L). Results similar to
those shown for cells transfected with HAC-RADC were obtained for cells
transfected with HAN-RADC. Bar: 30m.
96
Figure 3-11. Optical sectioning of a HtTA-1 cell transiently expressing
HAN-RADC. HtTA-1 cells expressing HAN-RADC were fixed 48 hours
after transfection and processed for indirect immunofluorescence by double
labeling with mAb 12CA5 and phalloidin. Cells were examined by confocal
microscopy as described in Materials and Methods. Each panel A-F shows
the phalloidin channel on the left side and mAb 12CA5 channel on the right
side. A is a compilation of all focal planes taken in the z-dimension. B-F
are single focal planes taken in 1sm steps in a dorsal to ventral direction.
Processes emanating from the entire dorsal surface of the cell are long,
tapered, and filled with F-actin. Bar: 10m.
98
I
Because mAb 12CA5 signal intensity made the imaging of cleavage
furrows and ventral focal planes by conventional microscopy difficult, we
took optical sections with a confocal microscope to examine HA-radixin
localization in these areas. Figure 3-12 shows cleavage furrows in dividing
cells expressing HA-radixin constructs stained with mAB 12CA5 and
phalloidin. These images demonstrate that the HAC-RAD polypeptide colocalizes with F-actin throughout the contractile ring, but that HAN-RAD
cannot be detected in this structure. However, unlike the situation in the
NIH-3T3 cell lines, both of the amino-terminal HA-radixin constructs
localized to cleavage furrows in the transiently transfected HtTA-1 cells
(HAN-RADN is shown in Fig. 3-12 C). This difference is considered in the
Discussion section below. However, HAN-RADC, although it co-localized
with F-actin in other regions of the cortex, is not detectable in the
contractile ring (Fig. 3-12 D); HAC-RAD-C shows similar behavior (data not
shown). Thus, in HtTA-1 cells expressing high levels of HA-radixin
constructs, both the full-length molecule and amino-terminal domain are
capable of localizing to the F-actin rich cleavage furrow while the same
high levels of the carboxy-terminal polypeptides are not detectable in this
structure.
At the higher levels of expression facilitated by the transient
transfection system, none of the HA-radixin constructs, particularly the
carboxy-terminal polypeptides, co-localized precisely with ventral F-actin
filaments in HtTA-1 cells (Fig. 3-13 A and B). Instead, expression of the
carboxy-terminal polypeptides often disrupted normal phalloidin staining
in the ventral cytoplasm. Figure 3-13 C shows a cell expressing HANRADC, stained both with anti-HA antibody and phalloidin. The latter
staining is diffuse or punctate, in contrast with the linear elements present
in control cells (compare left hand panels in Fig. 3-13 A and C).
Expression of the carboxy-terminal polypeptides apparently also
interferes with normal cytokinesis. Many cells brightly staining with mAb
12CA5 had two or more nuclei 48 hours after transfection (Fig. 3-14 A-C).
We scored the number of nuclei per cell in cultures transfected with each of
the six HA-radixin constructs co-stained with mAb 12CA5 and DAPI. The
results showed that the incidence of multi-nucleated cells is much higher
in cells expressing the carboxy-terminal polypeptides than in cells
100
expressing either the full-length or the amino-terminal polypeptides (Fig 314 D). This result was also confirmed by flow cytometry (data not shown).
Transient expression of HA-radixin constructs in NIH-3T3 cells.
Unlike their behavior when stably expressed in NIH-3T3 cells, the
transiently expressed amino-terminal polypeptides localize to cortical
structures and the carboxy-terminal polypeptides induce the presence of
abnormal cortical structures in HtTA-1 cells. To test the possibility that the
differences in the results of these experiments are due to differences
between NIH-3T3 cells and HtTA-1 cells, we transiently transfected NIH3T3 cells with the HA-radixin constructs. We expressed these proteins
under the control of a strong B-actin promoter (see Appendix Three for a
description of these constructs). NIH-3T3 cells, 48 hours after transfection,
are shown in Figure 3-15. The full-length constructs behaved in these cells
as they did at lower expression levels, that is, HAC-RAD localized to the
appropriate cortical structures whereas HAN-RAD remained diffuse in the
cytoplasm (Fig. 3-15 A and B). High-level expression of these constructs
had no apparent effects on cell morphology. In contrast, brightly mAb
12CA5 staining NIH-3T3 cells transfected with the carboxy-terminal
constructs showed morphological abnormalities resembling those found in
HtTA-1 cells transiently transfected with these constructs, including long,
tapered processes. (Fig. 3-15 D). Many of these NIH-3T3 cells were
multinucleate like the carboxy-terminal construct transfected HtTA-1 cells.
These results indicate that the morphological abnormalities and
cytokinesis defects described above for HtTA-1 cells transfected with the
carboxy-terminal radixin construct are a consequence of high level
expression of this construct and not due to the particular cell type assayed.
NIH-3T3 cells transfected with the amino-terminal constructs did not show
the abnormal processes present in the carboxy-terminal transfectants, but
these cells appeared to be in distress. Very few positive staining cells were
evident 48 hours after transfection. Those that were present were
unusually rounded with many membrane blebs (Fig. 3-15 C).
101
Figure 3-12. Localization of HA-radixin polypeptides in cleavage furrows of
transiently transfected HtTA-1 cells. HtTA-1 cells expressing HA-radixin
constructs were fixed 48 hours after transfection and processed for indirect
immunofluorescence by double labeling with mAb 12CA5 and phalloidin.
Cells were examined by confocal microscopy as described in Materials and
Methods. Each panel A-D shows the phalloidin channel on the left side and
mAb 12CA5 channel on the right side. HAC-RAD (A) and HAN-RADN (C)
localize to the F-actin containing contractile ring, but HAN-RAD (B) and
HAN-RADC (D) do not. Bar: 10gm.
102
Figure 3-13. Localization of HA-radixin constructs in the ventral cytoplasm
of transiently transfected HtTA-1 cells. HtTA-1 cells expressing HAC-RAD
(A) HAN-RADC (B,C) were fixed 48 hours after transfection and processed
for indirect immunofluorescence by double labeling with mAb 12CA5 and
phalloidin. Cells were examined by confocal microscopy as described in
Materials and Methods. Each panel A-C shows the phalloidin channel on
the left side and mAb 12CA5 channel on the right side. Cytoplasmic
microfilaments were evident in ventral focal planes of cells expressing
HAC-RAD (arrow in left panel of A) and in some cells expressing HANRADC (arrow in left panel of B) but neither of these HA-radixin constructs
co-localized with these structures (compare left and right panels in A and
B). In other cells expressing HAN-RADC (C), these linear F-actin
structures were disrupted and phalloidin staining in this region of the cell
was diffuse or punctate (arrow in left panel of C). Bar: 10m.
104
I
Figure 3-14 Expression of the carboxy-terminal HA-radixin constructs
results in an increased number of multi-nucleated cells. HtTA-1 cells
expressing HA-radixin constructs were fixed 48 hours after transfection
and processed for indirect immunofluorescence by double labeling with
mAb 12CA5 and DAPI. A cell expressing HAN-RADC is shown in A-C.
This mAb 12CA5-positive cell (A) has two nuclei which can be seen by DAPI
staining (B) and by phase contrast (C). Cultures of HtTA-1 cells transfected
with each of the HA-radixin constructs were scored for mAb 12CA5
immunoreactivity and nuclei number (D). 200-400 well-spread, nondividing cells were counted for each construct. The results shown are
pooled from three independent transfections. There is a dramatic increase
in cultures expressing carboxy-terminal HA-radixin constructs compared
to non-expressing cells in the same culture and to cells expressing either
the full-length or the amino-terminal HA-radixin constructs. Bar: 30gm.
106
PER CENT CELLS WITH
TWO (FOUR) NUCLEl
EXPRESSED
CONSTRUCT
HAC-RAD
12CA5(+)
<1
12CA5(-)
<1
HAN-RAD
2
0
N-TERMINUS
1
2
C-TERMINUS
24 (5)
<1
Figure 3-15. Transient expression of HA-radixin proteins in NIH-3T3 cells.
NIH-3T3 cells were transfected with HA-radixin constructs as described in
Materials and Methods. 48 hours after transfection, cells were processed
for immunofluoresence with mAb 12CA5 as described in Materials and
Methods. Shown are cells expressing HAC-RAD (A), HAN-RAD (B),
HANRADN (C), and HANRADC (D).
108
DISCUSSION
NIH-3T3 cells express each of the ERM proteins, and two of them radixin and moesin - show a typical localization to cortical cytoskeletal
structures. To identify the elements of ERM sequence that specify cellular
interactions, we expressed full-length radixin and its amino- and carboxyterminal domains in cultured cells, at both low and high levels relative to
endogenous radixin. At low levels, the full-length protein localized to the
appropriate cortical structures: ruffling edges, filopodia and lamellipodia,
microvilli and the cleavage furrow of dividing cells. The carboxy-terminal
polypeptide behaved similarly, with two notable exceptions: it did not
localize to cleavage furrows, and it localized abnormally, to some stress
fibers. In cells expressing either the full-length or carboxy-terminal
polypeptides, the staining of endogenous moesin is substantially
diminished. In contrast, the amino-terminal domain of radixin was only
diffuse throughout the cell, and its expression had no effect on moesin
staining. Only when expressed at levels significantly higher than the
endogenous pool did the amino-terminal polypeptide show cortical
localization. At those levels, the carboxy-terminal domain induced
significant disruptions of normal cytoskeletal structures and functions.
These results demonstrate that determinants of radixin localization reside
in both the amino- and carboxy-terminal domains of the protein, and that
these domains may interact in cis to achieve localization. The results also
suggest that all ERM proteins may have a common intracellular ligand
that is both essential and limiting for localization.
Functional interactions between the amino- and carboxy-terminal
domains of radixin. Our results suggest that the cellular associations of
radixin are not the simple sum of the activities of the parts of the molecule.
A previous study found that both the amino- and carboxy-terminal domains
of ezrin associated with cell surface microvilli, but suggested that correct
targeting of that protein depends primarily on the amino-terminal
sequence (Algrain et al., 1993). The discrepancy between that conclusion
and those in this paper may be explained by the effects of high level
expression in transient systems versus more modest levels in stable
transfectants examined here. Both amino- and carboxy-terminal domains
110
of radixin do contain determinants of localization, but those determinants
apparently can interact with one another in cis to account for the properties
of the full-length molecule. For example, the carboxy-terminal polypeptide,
unlike the full-length molecule, fails to localize to cleavage furrows, binds
to stress fibers, and at high levels disrupts normal cytoskeletal structures
and functions. All of these activities are modified in the context of the fulllength molecule - that is, when the amino-terminal domain is present in
cis. Perhaps, then, the amino-terminus of radixin exerts a regulatory
influence over the activities of the carboxy-terminus. Such an interaction
might explain why extensive efforts to detect a direct interaction between Factin and ezrin in vitro were negative (Bretscher, 1983), while the isolated
carboxy-terminal domain can bind (Turunen et al. 1994). In cells, too, the
carboxy terminal polypeptide co-localizes with microfilaments (Algrain et
al., 1993; and Fig. 3-11). There is precedent for such intramolecular
regulation between domains of a cytoskeletal protein. Johnson and Craig
(1994, 1995) demonstrate in vitro that an such an interaction modulates the
binding of talin to the amino-terminal domain of vinculin, and of F-actin
binding to the carboxy-terminal domain. Perhaps intramolecular
interactions between the amino- and carboxy-terminal domains of radixin
may explain why the ability of the full length molecule to localize to cortical
cytoskeletal structures is dependent on the placement of the epitope tag.
Interactions of radixin with cellular factors necessary for
localization. Stable expression of the carboxy-terminal tagged full-length
radixin largely abolishes moesin staining in cortical structures. The
simplest interpretation of this result is that radixin and moesin compete for
some common binding partner that is limiting for localization to cortical
structures. This interacting element may be present at the cortex itself. A
candidate molecule is CD44, known to interact with ERM proteins (Tsukita
et al., 1994). Alternatively, it could be a cytoplasmic protein with which
ERM proteins interact before localization. For example, ERM proteins
might localize as homo- or hetero-oligomeric complexes (Gary and
Bretscher, 1993; Andreoli et al., 1994). Over-expression of radixin could
compete with other ERM proteins for participation in such complexes.
However, since the full-length protein both localizes to the cleavage furrow
and competes for moesin localization in that structure, while the carboxy111
terminal fragment has neither of these activities, it is likely that
competition takes place at the cortical localization site itself.
The carboxy-terminal domain of radixin localizes to cortical
structures and, even at levels less than 20% those of endogenous radixin,
substantially diminishes moesin staining in cortical structures. This
behavior, too, can be explained by competition for a common and necessary
binding partner, but also requires that the relative stoichiometry of that
partner compared to total ERM proteins is quite small. In addition, either
the affinity of the carboxy-terminal polypeptide for the partner is
considerably greater than that of full-length ERM proteins - again,
suggestive of intramolecular regulation of the interactions of isolated
domains - or the pool of endogenous ERM proteins competent to interact
with the partner is only a small fraction of the total complement.
Interchangeability of ERM proteins in NIH-3T3 cells? The
displacement of moesin by exogenously expressed radixin occurs without
any apparent phenotype. This indicates that the tagged version of radixin
can substitute for endogenous moesin without seriously disrupting cellular
functions such as spreading or cytokinesis. That two ERM proteins, 75%
identical, are interchangeable at cortical localization sites supports the
notion that these proteins may be functionally redundant. Takeuchi et al.
(1994) report that anti-sense inhibition of either ezrin or radixin expression
can have different phenotypes than inhibition of moesin expression. Such a
phenotypic difference can arise from qualitative differences among the
three proteins, but functionally interchangeable elements can give different
null phenotypes for quantitative reasons as well (Schatz et al. , 1986).
Consequences of high-level expression of truncated forms of radixin.
The interest in assessing the effects of high level expression of the aminoand carboxy- terminal domains of radixin stemmed from the initial
observation that the only stably transfected lines recovered expressed these
constructs at a low stoichiometry to the endogenous radixin pool. This
suggested that higher expression levels of truncated radixin were not
tolerated by the cell. It is not yet clear why high level expression of the
amino-terminus should be toxic. Indeed, HtTA-1 cells expressing this
radixin domain do not appear to be affected in transient assays. In
112
contrast, transient expression of the amino-terminal domain perturbs
NIH-3T3 cells. These cells are rounded with many membrane blebs and
appear to be having difficulty adhering to the culture substratum. More
work is necessary to determine if this phenotype is a specific defect in cell
adhesion or whether these cells are dying and rising off the dish. In the
antisense inhibition experiments, a cocktail of oligonucleotides targeting all
three ERM proteins inhibited cell-substratum adhesion (Takeuchi et al.,
1994). If a cell adhesion phenotype is established for expression of the
amino-terminal domain, this could be a dominant negative effect.
It is more obvious why we did not recover cell lines expressing high
levels of the carboxy-terminal fragment. High level expression of the
carboxy-terminal domain has several negative consequences for cells: it
induces the formation of long processes all over the surface of the cell,
processes unlike normal cortical extensions such as filopodia or microvilli;
it disrupts ventral F-actin filaments; and it interferes with cytokinesis even
though the polypeptide itself is not found in cleavage furrows. These
phenotypes are not the simple consequence of over-expressing any actinassociated protein. For example, the carboxy-terminal domain of
caldesmon, transiently expressed in CHO cells, associates with and
stabilizes F-actin, but does not induce cortical processes (Warren et al.,
1994). It is possible that the long processes induced by high levels of the this
radixin domain sequester a substantial fraction of components required to
form F-actin, and so indirectly affect cytokinesis. Indeed, the phenotype
observed here in NIH-3T3 cells may reflect generalizable interactions of
ERM protein carboxy-terminal domains. During these studies, Martin et
al. (1995) reported that baculovirus mediated high-level expression of the
carboxy terminus of ezrin induced process formation in Sf9 moth ovary
cells. Furthermore, over-expression of the carboxy-terminal domain of the
D. melanogaster moesin homolog in fission yeast produced irregularly
shaped and multi-nucleated cells (Edwards et al., 1994). If the carboxytermini of ERM proteins are centrally involved in the organization of the
cytoskeleton at particular cortical sites, perhaps modulating membrane
protrusive activity at those sites, then the over-expression of this domain
could lead to the observed morphological phenotypes.
113
CHAPTER FOUR:
Deletion analysis of radixin's carboxy-terminal donain.
114
SUMMARY
The results presented in Chapter Three indicate that the carboxyterminal domain of radixin (amino acids 318-583) has several properties
when expressed in cells: 1) it induces the formation of unusual cortical
structures, 2) it disrupts cytokinesis, and 3) it competes with moesin for
localization in cortical structures. In this chapter, we have further
dissected the carboxy-terminus by deletion analysis. We found that both the
morphogenic and cytokinesis disruption activities map to amino-acids 509583. However, the same minimal fragment capable of producing these
phenotypes did not displace cortical moesin. In fact, none of the deletion
constructs tested displaced cortical moesin as did the entire carboxyterminal domain. A fragment consisting of amino acids 318-400 is capable
of localization to cortical cytoskeletal structures in the transiently
transfected cells. Additionally, we show that high-level expression of the
carboxy-terminal domain of moesin (amino acids 318-577) has effects on
cells similar to those described for the carboxy-terminal domain of radixin,
providing more evidence that these effects are generalizable to the carboxyterminal domains of all ERM proteins.
115
MATERIALS AND METHODS
Construction of radixin carboxy-terminus deletion mutants. We
constructed a set of deletion constructs representing the carboxy-terminal
domain of radixin by PCR amplification using the primers shown below.
PCR primers:
Primer
Sequence
I
RADFWD2
RADFWD3
RADFWD4
RADFWD5
RADREV1
RADREV3
RADREV4
RADREV5
MOEFWD1
MOEFWD2
MOEREV1
5'CGGCTCGAGCTAGAAAGGGCACAATTAG
5' CGGCTCGAGTCTGCAATCGCCAAGCAAG
5'*CGGCTCGAGCACAAAGCTTTTGCAGCTC
5'eCGGCTCGAGCGAAGCGAGGAGGAGTG
5'*CCGCTCGAGCATGGCTTCAAACTCATCG
5'CCGCTCGAGCTTGGCTTCTTCTGCAGC
5'eCCGCTCGAGCTGCCACTCAGTAGCTTC
5'*CCGCTCGAGGTGGTTCATCACCCCCTC
5'*CTCGAGCCGAAGACGATCAGTGTG
5'.CTCGAGATGGAGCGTGCTCTCCTG
5'eCTCGAGCATGGACTCAAACTCATC
In addition to the appropriate complementary radixin sequence, each
primer carried an XhoI site in the same reading frame as the radixin
coding sequence. The template used for each of these reactions was the
plasmid pR2ESS containing the murine radixin cDNA clone described in
Chapter Three. We used the following primer pairs for amplification of the
truncated radixin constructs: RADC-1 (residues 318-400)-RADFWD2 and
RADREV3; RADC-2 (residues 318-446)-RADFWD2 and RADREV4; RADC-3
(residues 318-508) RADFWD2 and RADREV5; RADC-4 (residues 401-583)
RADFWD3 and RADREV1; RADC-5 (residues 447-583) RADFWD4 and
RADREV1; RADC-6 (residues 509-583) RADFWD5 and RADREV1.
Amplification reactions were carried out in a Perkin-Elmer DNA thermal
cycler using a Gene-Amp PCR kit (Perkin-Elmer) according to
manufacturers instructions. The PCR products were run on a 1% agarose
gel, the band containing the appropriate size fragment was excised from
116
the gel, and DNA from this gel slice was purified using Qiaex resin
(Qiagen). The gel purified PCR products were subcloned directly into the
pT7-blue PCR product cloning vector (Novagen) according to
manufacturers instruction. Diagnostic digest were performed on these
plasmids to verify that the appropriate radixin fragment was present in the
plasmid. These plasmids were digested with XhoI and inserted into the
XhoI site of pUHD-HAC. The set of carboxy-terminus deletion constructs is
diagrammed in Figure 4-1 A.
pUHD-HAC was constructed from pUHD10-3 which allows for
tetracycline-regulable expression (described in Chapter Three). First, an
endogenous XhoI site was eliminated from pUHD 10-3 by digestion with
XhoI followed by blunting with Kienow enzyme and religation. This site
occurred in the backbone plasmid sequence and therefore was not located in
a critical regulatory region. This step was necessary because in the
following step we introduced an XhoI site that serves as a unique cloning
site. An oligonucleotide was inserted into this modified pUHD 10-3 vector
between the EcoRI and XbaI sites. This double-stranded oligonucleotide
contained the following features in a contiguous reading frame: an ATG
initiation codon, the six base XhoI site, and the HA epitope sequence
followed by a stop codon. Insertion of the XhoI-flanked radixin sequences
into the XhoI site of pUHD-HAC results in fusion of the radixin coding
sequences to the HA epitope tag. Due to the presence of the XhoI cloning
sites that flank the inserted fragments, leucine and glutamate residues
were also present flanking the radixin polypeptides. The plasmids
constructed in this manner were prepared on Qiagen columns and
transfected into cells as described in Chapter Three.
RT-PCR cloning of moesin carboxy-terminal domain. We cloned the
carboxy-terminal domain of murine moesin (amino acids 318-577) in the
following manner. First, we extracted total RNA extracted from NIH-3T3
by the method of Wallace (1987). Then, using a first-strand cDNA synthesis
kit (Pharmacia), we synthesized a single stranded cDNA from the RNA
template using a moesin specific primer. This primer MOEREV1 (see
Table 4-1) is complementary to the 3' end of the coding region on the moesin
transcript. Next, we used the reaction product above as a template for PCR
amplification of full-length moesin using MOEFWD1 and MOEREV1. The
117
product from this reaction was gel purified as described above and cloned
into the pT7-blue vector. The identity of the insert as murine moesin was
confirmed by restriction digestion with XhoI, BamHI, and PstI. The
appropriate sized fragments were liberated from the plasmid according to
the murine moesin cDNA sequence published by Sato et al. (1992). This
plasmid was named pT7-MOE.
We constructed the carboxy-terminal domain by PCR amplification
using pT7-MOE as a template and MOEFWD2 and MOEREV1 as a primer
set. These primers amplify the moesin sequence between codons 318 and
577 and contain XhoI sites in frame with the moesin coding sequence at
their 5' ends. The PCR product was gel purified as described above and
cloned into the pT7-blue vector. The resultant plasmid pT7-MOEC was
verified by restriction digests. The moesin carboxy-terminal fragment was
released from pT7-MOEC by digestion with XhoI and inserted into the XhoI
site of pUHD-HAC described above. The resultant plasmid, pUHD-HACMOEC allowed for tetracycline regulable expression of the carboxy-terminal
domain of moesin fused to the HA epitope. This plasmid was prepared on a
Qiagen column and transfected into cells as described in Chapter Three.
Co-localization of HA-radixin polypeptides and moesin. We colocalized HA-radixin and moesin in transiently transfected NIH-3T3 cells
as described in Chapter Three for stably transfected NIH-3T3 cells with the
following modifications. For transient expression in NIH-3T3 cells of the
pUHD plasmids described above carrying the carboxy-terminal deletion
constructs we co-transfected these plasmids with plasmid pUHD 15-1
carrying the tetracycline transactivating element. 48 hours after
transfection, cells were processed for immunofluorescence with mAb
12CA5. In control experiments, we found that a number of commercial
preparations (Cappell, Tago) of the fluorescein-conjugated goat anti-rabbit
F(ab') 2 antisera used for detection of anti-moesin antibody #454 weakly
cross-reacted with mAb 12CA5. This signal confounded interpretation of
experiments in transiently transfected cells expressing high levels of HAtagged proteins. Therefore, for these experiments we used a mAb 12CA5 at
1:10,000 dilution. This measure reduced the signal from the cross reactivity
to the point where results were interpretable. All other conditions,
118
including the use of affinity eluted antiserum #454, were similar to those
described in Chapter Three.
119
RESULTS
Expression of radixin carboxy-terminal domain deletion constructs
in HtTA-1 cells. The set of deletion constructs used to dissect the activities
of the carboxy-terminal domain and a summary of the activities of these
deletion constructs are diagrammed in Figure 4-1 A. We harvested protein
from cells transfected with these constructs 48 hours after transfection and
examined expression of the HA-tagged radixin constructs by Western blot
analysis with mAb 12CA5 (Fig. 4-1 B). Each of the constructs is expressed,
although RADC-2-5 show several lower molecular weight bands that
probably represent proteolytic breakdown products. Assuming that the
highest molecular weight species represents the full-length translation
product, the observed molecular weights for these polypeptides are the
following: RADC-1 (11.5 kDa); RADC-2 (16.7 kDa); RADC-3 (24.0 kDa);
RADC-4 (25.7 kDa); RADC-5 (21.6 kDa); RADC-6 (10.5 kDa). The predicted
molecular weights for these radixin fragments plus the extra sequence
added by the epitope tag are the following: RADC-1 (12.5 kDa); RADC-2 (17.6
kDa); RADC-3 (22.3 kDa); RADC-4 (22.8 kDa); RADC-5 (17.7 kDa); RADC-6
(11.0 kDa). The observed molecular weights for RADC-1, 2, and 6 are
within 1 kDa of the predicted values. In contrast, RADC-3, 4, and 5 are
running 1.5, 2.9, and 3.9 kDa, respectively, larger than predicted. The
anomalous migration of the carboxy-terminal domain of radixin in SDSpolyacrylamide gel electrophoresis was reported in Chapter Three. These
results map the element that leads to the unexpectedly slow migration
between amino acids 447-508. In this strech of amino acids there are 8
contiguous proline residues (amino-acids 469-477). Others have noted
effects of polyprolyl tracts on the migration properties of other proteins.
Although each protein is clearly expressed, there are some
differences in the apparent expression levels. RADC-1 and RADC-6 are
present in much lower abundance than the other constructs. This result
was reproduced in another independent transfection. Some of this
difference can be attributed to the fact that fewer cells in these transiently
transfected cultures are expressing the RADC-1 and RADC-6 polypeptides.
This was evident after examining the number of mAb 12CA5 stained cells
in RADC-1 and RADC-6 transfected cultures compared to the others (data
not shown). However, other mechanisms can also account for the lower
120
steady state expression levels of these polypeptides. Unfortunately, it is
difficult to make accurate determinations of the amount of heterologous
protein expressed in individual cells in transient expression experiments.
Induction of abnormal cortical processes by radixin carboxy-terminal
domain deletion constructs. We examined the localization of the carboxyterminal domain deletion constructs 48 hours after transfection into HtTA1 cells. Figure 4-2 shows ventral focal planes of mAb 12CA5 stained HtTA-1
cells transfected with each of the HA-tagged carboxy-terminal deletion
constructs. For comparison, HtTA-1 cells transfected with full-length
radixin and the entire carboxy-terminal domain are shown in Figure 4-2 A
and E respectively. RADC-1, 2, and 3 transfectants (Fig. 4-2 B-D) all look
like the full-length transfectants. There are occasional extended processes
in contact with the substratum which may be large filopodia or retraction
fibers. These structures are also present in full-length transfectants. In
contrast, like cells transfected with the entire carboxy-terminal domain,
cells expressing RADC-4, 5, and 6 show large numbers of long, tapered
processes in the ventral focal plane (Fig. 4-2 F-H). There are also abnormal
extensions emanating from the dorsal surfaces of cells expressing these
three constructs that resemble those shown in Figure 3-11 for cells
transfected with the entire carboxy terminal domain (see arrow in Fig. 4-3
E for an example). These abnormal dorsal structures do not appear in cells
transfected with RADC-1, 2, or 3 see Fig. 4-3 A for an example).
Although the cellular morphology induced by RADC-6 is clearly
distinguishable from RADC-1, -2, and -3, and full-length radixin, the
character of the processes induced by RADC-6 differ somewhat from those
in RADC-4, -5, and the entire carboxy-terminal domain. The processes on
RADC-6 transfectants are generally neither as long nor as thick as the
other transfectants (Fig. 4-2 H). However, there were rare examples in the
RADC-6 transfected cultures that did have long processes. The less robust
processes in RADC-6 transfectants could be due to lower expression levels
of this polypeptide in the transfected cells. That this could be the case is
suggested by the Western blotting data shown in Figure 4-1 B. Taken
together, these data demonstrate that the element responsible for the
morphogenic properties radixin's carboxy-terminal domain lie between
amino acids 509 and 583.
121
Figure 4-1. Expression of radixin carboxy-terminal domain deletion
constructs in HtTA-1 cells. (A) Schematic diagram of radixin carboxyterminal domain deletion constructs and summary of their properties. A
set of contiguous, non-overlapping constructs covering the carboxyterminal domain of was generated as described in Materials and Methods.
The amino acid breakpoints are indicated in the diagram. All constructs
were expressed in cells as fusion proteins bearing the HA epitope tag. To
the right of the constructs is a summary of their properties when expressed
in cells. The data supporting this summary is presented in Figures 4-2, -3
and -4 and Table 4-2. (B) Western blot analysis of proteins taken from HtTA1 cells transfected with radixin carboxy-terminal domain deletion
constructs RADC-1 through RADC-6. Proteins were harvested and
analyzed for expression of these constructs with mAb 12CA5 48 hours after
transfection as described in Materials and Methods. Samples are identified
beneath the blot and molecular weight standards (in kilodaltons) are at the
right of the blot.
122
A
Dissection of radixin's carboxy-terminal domain.
Abnormal
Construct
Morphology
U
RADC
318
513
I
Cytokinexis
Defects
S
Moesin
Displace.
S
+
+
+
+
+
+
+
+
+
RADC-1
Rjji-
6
RADC-3
411
518
RADC-4
RAD-5II
7
ff
IM"M
5JR;AM;
DC
B
-
45.0
31.0
.
21.5
.
14.4
-
6.5
.
Cq1
Cv)
LO
Figure 4-2 Effects of carboxy-terminal domain deletion constructs on
cortical structures. HtTA-1 cells were transfected with radixin carboxyterminal domain deletion constructs RADC-1 through RADC-6 plus control
constructs and processed for immunofluorescence with mAb 12CA5 as
described in Materials and Methods. Ventral focal planes of brightly
staining cells are shown. (A) Full-length radixin; (B) RADC-1; (C) RADC2; (D) RADC-3; (E) entire radixin carboxy-terminal domain; (F) RADC-4;
(G) RADC-5 (H) RADC-6.
124
Localiation of radixin carboxy-terminal domain deletion constructs
in transiently transfected HtTA-1 cells. We examined the subcellular
localization of the radixin carboxy-terminal domain deletion constructs in
more detail. Figure 4-3 shows these results for RADC-1 and RADC-4. At
least some of the RADC-1 polypeptide appears to be cortically localized in
these transiently transfected cells. In dorsal focal planes, enriched
staining is evident in microvilli (Fig. 4-3 A). In ventral focal planes, RADC1 immunoreactivity is detectable in filopodial processes (Fig. 4-3 B). This
sort of staining pattern was present even in some of the less intensely
staining cells, suggesting that RADC-1 localized to cortical structures in
cells expressing lower amounts of this polypeptide. Phalloidin staining of
these cells shows that stress fibers in the expressing cells are
indistinguishable from those in the surrounding non-expressing cells (Fig.
4-3 C). We obtained similar results for cells transfected with RADC-2 and
RADC-3.
RADC-4 also appears to be cortically localized, although as
mentioned above its expression induces the presence of abnormal cortical
processes. These brightly mAb 12CA5 staining structures can be seen in
both dorsal (Fig. 4-3 E) and ventral (Fig. 4-3 F) focal planes. The phalloidin
staining in cells expressing RADC-4 is noticeably different from the
surrounding non-expressing cells. Often, but not always, instead of bright,
linear stress fibers, the phalloidin staining in the ventral cytoplasm
appeared diffuse (see arrow in Fig. 4-3 G). Finally, cells expressing RADC4 tended to be more spread than non-expressing cells. This is evident by the
increased area occupied these cells and the lower density of the edges of
cells in phase contrast (Fig. 4-3 H). Similar results were obtained for
RADC-5 and 6. Like the entire carboxy-terminal domain, none of the
carboxy-terminal deletion constructs localized to cleavage furrows (data not
shown).
Cytokinesis phenotypes of cells transfected with radixin carboxyterminal domain deletion constructs. We scored the number of
multinucleated cells expressing the carboxy-terminal domain deletion
polypeptides and compared to the entire carboxy-terminal domain. Table 41 shows the results expressed as percent of multinucleated cells. Results
reported in Figure 3-14 D established that the background of multinucleated
126
cells in HtTA-1 cultures was 2%. In contrast, in this assay the carboxyterminal domain of radixin showed a nearly ten-fold increase in the
numbers of expressing cells with a multinucleate phenotype. Expression of
RADC-1, -2, and -3 did not seem to have an effect on cytokinesis as judged by
the frequency of multinucleated cells. All of the values for these constructs
were within the background range. However, RADC-4, -5, and -6 did have a
demonstrable effect on cytokinesis. Although the values for these
constructs were not as high as for the entire carboxy-terminal domain, they
were from two- to five- fold higher than background. Again, like the results
reported above on induction of abnormal processes, RADC-6 had the least
dramatic effect of the RADC-4, -5, and -6 polypeptides. Perhaps the
differences in this assay between the entire carboxy-terminal domain and
its truncated versions -RADC-4, -5, and -6- are due to differences in
expression levels as mentioned previously. Alternatively, the region of the
carboxy-terminal domain covered by RADC-1 could make a contribution to
disruption of cytokinesis when in the context of the rest of the carboxyterminal domain but not alone. In sum, these results indicate that the
same region that is responsible for the cell morphogenetic effects when the
carboxy-terminal domain of radixin is expressed at high levels in cells,
amino acids 509-583, is also responsible for the disruption of cytokinesis.
Moesin displacement properties of radixin carboxy-terminal domain
deletion constructs. We examined moesin displacement by the carboxyterminal domain deletion constructs in transiently transfected NIH-3T3
cells as described in Materials and Methods. Figure 4-4 shows the results
from this analysis. Because a high mAb 12CA5 dilution was used for these
experiments (see Materials and Methods), an arrow indicates the
mAb12CA5 positive cell(s) in Figure 4-4 B, D, and F for reference. As in the
stably transfected NIH-3T3 cells expressing the entire carboxy-terminal
domain, moesin immunoreactivity in cortical structures is diminished in
NIH-3T3 cells transiently transfected with this domain (Fig. 4-4 A and B).
In contrast, none of the deletion constructs derived from this domain
showed this activity. For simplicity, only C-3 (Fig. 4-4 C and D) and C-6
(Fig. 4-4 E and F) are shown.
127
Figure 4-3. Subcellular localization of radixin carboxy-terminal domain
deletion constructs. HtTA-1 cells were transfected with radixin carboxyterminal domain deletion constructs RADC-1 through RADC-6 and
processed for double-label immunofluorescence with mAb 12CA5 and
phalloidin as described in Materials and Methods. For simplicity, only
RADC-1 (A-D) and RADC-4 (E-H) are shown. Dorsal (A and E) and ventral
(B and F) mAb 12CA5 staining and phalloidin staining in the ventral
cytoplasm (C and G) are of the same cells shown in phase contrast (D and
H). Arrow in E indicates abnormal dorsal cortical processes in a cell
expressing RAC-4. Arrow in G indicates a disrupted pattern of phalloidin
staining in the ventral cytoplasm of a cell expressing RAC-4.
128
Table 4-1 Effects of radixin carboxy-terminal domain deletion constructs on
cytokinesis. HtTA-1 cells were transfected with radixin carboxy-terminal
domain deletion constructs RADC-1 through RADC-6 and processed for
double-label immunofluorescence with mAb 12CA5 and DAPI to visualize
nuclei as described in Materials and Methods. Cells were examined by
immunofluorescence microscopy. The number of nuclei in mAb12CA5
positive-staining cells was counted. Cells with more than one nucleus were
designated as multinucleate. This number divided by the total number of
mAb12CA5 positive-staining cells equals % multinucleate. The values
RADC-1, 2, and 3 are within the background range of multinucleated cells
for the HtTA-1 line (see Fig. 3-14). n indicates the number of individual
cells counted for each construct.
130
Table 4-1:
Construct
%Multinucleate
I
RADC
RADC-1
RADC-2
RADC-3
RADC-4
RADC-5
RADC-6
n
ii.
14.7
1.1
1.3
1.6
9.2
6.6
4.1
131
365
538
672
495
636
442
387
Figure 4-4. Radixin carboxy-terminal domain deletion constructs do not
displace moesin from cortical structures. HtTA-1 cells were transfected
with radixin carboxy-terminal domain deletion constructs RADC-1 through
RADC-6 and processed for double-label immunofluorescence with antimoesin antibody #454 (A, C, E) and mAb 12CA5 (B, D, F) as described in
Materials and Methods. Cells transfected with the entire radixin carboxyterminal domain (A and B), RADC-3 (C and D), and RADC-6 (E and F) are
shown. For clarity, arrows indicate positive mAb 12CA5-staining cells.
132
Figure 4-5. The carboxy-terminal domain of moesin induces almormal
cortical structures in HtTA-1 cells. (A) Expression of moesin carboxyterminal domain in HtTA-1 cells. Protein from cells transfected with a HAtagged versions of the carboxy-terminal domain of murine moesin (lane1)
(see Materials and Methods for details of this construct) the carboxyterminal domain of radixin (lane 2) and mock transfected cells (lane 3) was
harvested 48 hours after transfection and subjected to Western blot analysis
with mAb12CA5 as described in Materials and Methods. Positions of
molecular weight standards (in kilodaltons) are indicated to the left of the
blot. (B-E) Subcellular localization of an HA-tagged version of the carboxyterminal domain of murine moesin in HtTA-1 cells. Cells were processed
for immunofluorescence with mAb 12CA5 and phalloidin 48 hours after
transfection with an HA-tagged version of the carboxy-terminal domain of
murine moesin as described in Materials and Methods. (B-E) are images of
the same cell. B and C show abnormal cortical structures induced by the
expression of this construct in dorsal and ventral focal planes respectively.
Phalloidin staining shows that abnormal structures are rich in F-actin (D)
and these structures are easily visualized by phase contrast microscopy (E).
134
A
97.4
66.2
45.0
31.0
1
2
3
E
Expression and localization of moesin carboxy-terminal domnain. To
explore further the notion that the effects of high level expression of the
carboxy-terminal domain of radixin might also be properties of high level
expression of the carboxy-terminal domains of other members of the ERM
family, we expressed the carboxy-terminal domain of moesin in HtTA-1
cells. We cloned and prepared an HA-tagged version of the carboxyterminus of murine moesin (amino acids 318-577) for expression in HtTA-1
cells as described in Materials and Methods. Figure 4-5 A shows a Western
blot of extracts taken from cells 48 hours after transfection with the carboxyterminal domain of moesin (Fig. 4-5 A lane 1) and the corresponding
carboxy-terminal domain of radixin (Fig. 4-5 A lane 2). The apparent Mr of
the moesin and radixin fragments are 42.8 kDa and 45.7 kDa respectively.
This difference of approximately 3 kDa could be attributed to the fact that
moesin lacks the polyproline tract present in radixin from residues 469-477.
This interpretation is consistent with the migration properties in SDS gels
of the radixin carboxy-terminus deletion constructs reported in Figure 4-1
B. However, the Mr's of both the moesin and radixin carboxy-terminal
fragments are much larger than their predicted sizes of 32.6 kDa and 33.1
kDa. The property that accounts for this increased apparent molecular
weight in SDS gels does not seem to be present in the truncated polypeptides
derived from the carboxy-terminal domain of radixin.
High level expression of the carboxy-terminal domain of moesin has
deleterious effects on HtTA-1 cells like the carboxy-terminal domain of
radixin. It induces the formation of cortical cytoskeletal structures that
can be seen clearly in both dorsal (Fig. 4-5 B) and ventral (Fig. 4-5 C) focal
planes. These processes are filled with F-actin (Fig. 4-5 D) and are clearly
resolved by phase contrast microscopy (Fig. 4-5 E). These peculiar cortical
cytoskeletal structures are indistinguishable from those present on cells
transfected with the carboxy-terminal domain of radixin.
Expression of the moesin carboxy-terminal domain also disrupts
cytokinesis. When we examined cells transfected with the HA-tagged
moesin carboxy-terminal domain construct and scored the number of
nuclei in HA-positive cells, we found that 5.4% of HA-positive cells also had
more than one nucleus. This value is approximately five-fold higher than
the background percentage of multinucleate cells found in HtTA-1 cultures.
136
DISCUSSION
Cell morphogenic and cytokinesis disruption properties map to
amino acids 509-583 of the carboxy-terminus of rWain. The data presented
above indicate that the element(s) responsible for the deleterious
consequences of high-level expression of the carboxy-terminal domain of
radixin lie between amino acids 509 and 583. These consequencesinduction of abnormal cortical processes and disruption of cytokinesismight both be attributed to a direct or indirect effect of the last 75 aminoacids of radixin on the actin cytoskeleton. For instance, the abnormal
processes are filled with F-actin and microfilaments are known to play a
crucial role in contractile ring function during cytokinesis. Partly because
the carboxy-terminal domain of radixin itself does not localize to the
cleavage furrow, we argued in Chapter Three that the effect of high level
expression of this domain on cytokinesis is likely secondary to its effects
which lead to the abnormal cell morphology. This hypothesis holds
through deletion analysis of radixin's carboxy terminal domain. In
principle, there could have been separate elements in the 266 amino acid
carboxy terminal domain that caused the morphological defects and
cytokinesis disruption respectively. Instead, these two phenotypes
segregated with the same smaller piece of the carboxy-terminal domain.
There is in vitro evidence for an F-actin binding site in the isolated
carboxy-terminus of other ERM proteins. An F-actin binding activity
detected by affinity chromatography mapped to the last 34 amino acids of
ezrin (Turunen et al. , 1994). Interestingly, these authors reported that
deletion of the final six amino acids of ezrin, which are identical to those in
radixin, abolished the F-actin binding activity. In our experiments, the
natural carboxy terminus of radixin is followed by the HA-epitope.
Therefore, it seems that if the effects of the carboxy terminus of radixin are
due to a direct interaction with F-actin, small deletions, but not extensions
from the carboxy-terminus affect F-actin binding. An F-actin binding site
has also been mapped to the carboxy-terminal 48 amino-acids of moesin by
blot overlay (Pestonjamasp et al. , 1995). Since the final 25 amino-acids are
nearly identical among radixin, ezrin, and moesin, it is highly likely that
radixin also contains an F-actin binding site at its extreme carboxy
terminus that would be detected in similar assays. F-actin binding by the
137
carboxy-terminal domain of radixin is examined in more detail in Chapter
Six.
We expected to be able to identify a smaller fragment of the carboxyterminal domain capable of displacing moesin from cortical structures.
However, none of the deletion constructs tested here possessed the
displacement activity of the entire carboxy-terminal domain. There are
many possible reasons for this negative result. One is that the
displacement domain extends beyond amino acids 400-509. Perhaps this
region is necessary for proper folding of a putative displacement domain or
multiple elements present in this larger span of amino acids are required
for displacement. In this case, none of the deletion constructs would be
capable of displacement. Alternatively, one trivial possibility is that the
epitope tag disrupts displacement activity in the context of the deleted
fragments but not the entire carboxy-terminal domain. At least we can
draw one conclusion from these studies on moesin displacement. The
same minimal fragment that is capable of disrupting cell morphology and
cytokinesis (amino acids 509-583) does not displace moesin from cortical
structures. This indicates that moesin displacement is a separable activity
from disruption of cell morphology and cytokinesis.
A cortical localization determinant present in amino acids 318-400?
The polypeptide between amino acids 318 to 400 appears to localize to
cortical structures including filopodia and microvilli. It is sometimes
difficult to make definitive judgments based on localization in transfected
cells expressing high levels of the protein of interest. However, other
evidence indicates that the cortical localization of this radixin fragment at
high levels reflects true targeting to this locale and not diffusion of an
abundant protein throughout the cell. Full-length radixin bearing an HAepitope tag at its amino terminus does not localize to cortical structures
when transiently expressed at high levels (see Fig. 3-10 D and E). Still,
other experiments will be necessary to firmly conclude that this portion of
radixin is capable of targeting to cortical structures including expression of
this fragment at lower levels in stable transfectants.
If amino acids 318-400 do target to the cortex, what is the nature of its
associations there? In Chapter Three, we found that the amino-terminal
domain of radixin localized to cortical structures when expressed at high
138
levels, but not at low levels. We interpreted this as a reflection of a low
affinity interaction with a cortical binding partner. Indeed, radixin may
have a number of contacts spread through several regions of the protein
that specify its localization to cortical cytoskeletal structures. These
contacts may occur with one or more other proteins. Such a multi-faceted
binding arrangement is not unprecedented at membrane-cytoskeletal
interfaces. Microinjected talin head and tail domains are each
independently capable of targeting to focal contacts (Nuckolls et al., 1990).
Focal contacts are known to have a large collection of proteins and many of
these proteins have been shown to interact with one another in vitro. In
principle, these large assemblies of proteins, associating with low
affinities, might allow for highly regulated, dynamic membranecytoskeletal interactions.
Deleterious effects of high-level expression of the moesin carboxyterminal domain. We found that high level expression of the carboxyterminus of moesin in mammalian cells had effects similar to high level
expression of the carboxy-terminus of radixin, namely induction of cortical
processes and disruption of cytokinesis. This result suggests that these
effects are generalizable properties of the carboxy-terminal domain of ERM
proteins. In fact, this result was not entirely unexpected. High level
expression of the carboxy-terminus of moesin disrupts cell shape in
Schizosaccharomyces pombe (Edwards et al., 1994). Furthermore,
overexpression of the carboxy-terminus of ezrin induces cortical processes
in Sf9 insect cells and deletion of the last 26 amino acids of this domain
alleviates these consequences (Martin et al., 1995). Taken together, the
results indicate that the cellular consequences of high level expression of
the carboxy-terminal domain of radixin are conserved among the carboxyterminal domains of the ERM protein family.
139
CHAPTER FIVE:
Inter-domain interactions of radixin in vitro.
140
SUMMARY
We have assayed the domains of radixin for binding activities in
vitro. Affinity columns bearing the amino-terminal domain of radixin
selectively bound a small subset of the proteins of the chicken erythrocyte
cytoskeleton. Two of those proteins were identified as radixin itself and
band 4.1. In contrast, the carboxy-terminal domain of the molecule bound
neither protein, and full-length radixin did not bind band 4.1 (binding of
full-length radixin to itself was not evaluated). Columns bearing a mixture
of the amino- and carboxy- terminal domains of radixin also failed to bind
radixin and band 4.1. These results suggested that the amino- and carboxyterminal sequences can interact with one another either in cis or in trans,
and so interfere with radixin's interactions with other ligands. Using
affinity co-electrophoresis, we confirmed a direct interaction in solution
between the two radixin domains; the data are consistent with the
formation of a 1:1 complex with a dissociation constant of -5 X 10-8 M.
Competition between intramolecular and intermolecular interactions may
help to explain the provocative and dynamic localization of ERM proteins
within cells.
141
INTRODUCTION
The results presented in Chapter Three demonstrate that in living
cells, separable domains of radixin contribute information that specify
appropriate localization in the cell. For example, at low levels of
expression, the carboxy-terminal domain of radixin can localize to all of the
structures in which ERM proteins are normally found, save the cleavage
furrow, and it associates quite clearly with one cellular element - stress
fibers - where ERM proteins are not typically found. The information
necessary to target radixin to the cleavage furrow is in the amino-terminal
domain of the protein.
These cellular assays also reveal evidence for regulatory interactions
between the domains of radixin. Expressed at high levels, the carboxyterminal domain causes dramatic disruption of normal cell morphology
and interferes with cell division, while the amino-terminal domain has
neither phenotype. However, both consequences of high level expression of
the carboxy-terminal domain are suppressed by the presence of the aminoterminal domain in the context of the full-length molecule. Perhaps, then,
the deleterious interactions of one domain are prevented by an interaction
with the other domain. We have tested this model using in vitro methods.
Here, we present evidence that the amino- and carboxy-terminal domains
of radixin can bind each other in vitro. We also show that this binding
event blocks the binding of other ligands. However, unlike co-expression of
the domains of ezrin in insect cells, expression in HeLa cells of the aminoterminal domain of radixin in trans does not suppress the deleterious
consequences of expression of the carboxy-terminal domain of radixin.
Possible explanations for this latter result are discussed.
142
MATERIALS AND METHODS
Isolation of bacterially expressed chicken radixin polypeptides. We
cloned chicken radixin cDNAs by PCR (First-Strand cDNA Synthesis Kit;
Pharmacia) from chicken embryo fibroblast mRNA. The clones
representing the full-length, amino-terminal (codons 1-318) and carboxyterminal (codons 319-585) sequences were ligated into PQE-70 (Qiagen)
using the SphI and BglII sites, so that the last amino acid of each sequence
is followed by arginine, serine and then 6 histidines and a stop codon. E.
coli DH5aF'IQ transformants with the correct radixin insert were
identified by restriction digests, and by DNA sequencing (Sequenase
Version 2.0 DNA Sequencing Kit; United States Biochemical Corp.) at the
junction of radixin insert and vector. Expression of the three His6-tagged
radixin polypeptides and His6-tagged dihydrofolate reductase was induced
by growth in 1mM isopropyl-1-thio-B-D-galactopyranoside for three hours.
Cell pellets were lysed (50mM NaH2PO4/NaHPO4, pH 8.0, 300mM NaCl,
20mM imidazole, 1mM Pefabloc, 1mM leupeptin, ImM pepstatin, 0.007
TIU/ml aprotinin and 1mg/ml lysozyme), sonicated and clarified by
centrifugation, then stored as aliquots at -80 0 C. High-speed supernatants
containing the His6-tagged polypeptides were incubated with 0.5ml Ni-NTA
Sepharose CL-6B resin (Qiagen) in lysis buffer supplemented with 10mM Bmercaptoethanol. The material was washed batch-wise twice with 15ml
wash buffer (50mM NaH2PO4/NaHPO4 , pH 8.0, 300mM NaCl, 40mM
imidazole) plus 10% glycerol, and once in wash buffer alone.
Affinity chromatography. For affinity adsorption experiments, His6tagged proteins were left bound to Ni-NTA matrices and incubated with a
chicken erythrocyte cytoskeletal fraction prepared as follows: Freshly
washed erythrocytes were extracted with 0.1% NP40 in PM2G (100mM
Pipes, 1mM MgSO4, 2mM EGTA, pH 6.9) containing .007 TIU/ml
aprotinin, 1mM PMSF, and 1mM leupeptin. After a wash in the same
buffer the pellet was extracted in 8M urea in phosphate-buffered saline with
protease inhibitors. The urea extract was spun at 11,000 x g and the
supernatant collected, dialyzed in PBS and frozen as aliquots. The
erythrocyte proteins were incubated with the affinity matrices batchwise for
20 minutes 4'C, then repeatedly washed and centrifuged to remove
143
unbound proteins. The matrices were loaded into columns and eluted with
two 0.5ml aliquots of elution buffer (50mM NaH2PO4/NaHPO4, pH8.0,
300mM NaCl and 125mM imidazole). Fractions were boiled in sample
buffer, and analyzed by 7.5% SDS-PAGE. Proteins were detected by silver
stain or by immunoblotting as described in Chapter Three, using antibodies
that detect epitopes in the amino-domain of ERM proteins (#220), the
carboxy-domain of radixin (#457), and previously characterized antibodies
again chicken erythrocyte band 4.1 (Conboy et al., 1988).
Affinity co-electrophoresis (ACE). For affinity co-electrophoresis
(ACE), high-speed supernatants of His6-tagged radixin amino- and
carboxy-domains were purified on Ni-NTA resin columns. Glutathione Stransferase was prepared by expression of pGEX-3X (Pharmacia) in E. coli
strain HB101. Protein concentrations were determined by the Bradford
assay (Bio Rad). ACE gels were cast as described (Lee and Lander, 1991;
Lim et al., 1991) using 1% low gelling temperature agarose in 125mM
potassium acetate, 50mM Hepes adjusted to pH 7.5 with NaOH. Gels were
run at 60 volts for 4 hours, and the proteins then transferred to
nitrocellulose by capillary action and analyzed by immunoblotting.
Retardation coefficients were calculated as previously described, including
the application of a correction for "overrunning" (electrophoresis of the
detected species beyond the end of the zones containing the retarding
species (Lim et al., 1991)). Dissociation constants were calculated from
non-linear least squares fitting of plots of corrected retardation coefficient
versus concentration of retarding protein (San Antonio et al., 1991). Data
were fit to a general form of the binding equation that is appropriate even
when the concentration of the detected species is not << Kd (Lim et al.,
1991).
Co-expression of amino- and carboxy-terminal domains of radixin in
HeLa cells. For co-expression of the amino-and carboxy-terminal domains
in HtTA-1 cells, we transfected 2 plates of cells with aliquots of a calciumphosphate precipitate containing equimolar concentrations of pUHDHACRADN (described in Chapter Three) and pCXN2-HACRADC
(described in Appendix Three). From these plasmids combined, expression
of the carboxy-terminal domain is constitutive and expression of the amino144
terminal domain is tetracycline-regulable. Immediately after transfection,
one of the two dishes was supplemented with 2gg/ml tetracycline to repress
expression of the amino-terminal domain. 48 hours after transfection, the
cells were processed for immunofluorescence and protein was harvested
and analyzed as described in Chapter Three.
145
RESULTS
Distinct binding domains of radixin in vitro. To prepare radixin
affinity columns, the high speed supernatants of bacterial extracts
expressing His6 versions of full-length radixin (FL), or its amino-terminal
domain (N-domain) or carboxy-terminal domain (C-domain), were applied
to Nickel-NTA Agarose as described above. In each case, the radixin
polypeptide encoded by the plasmid is by far the most abundant polypeptide
bound, although several bacterial bands are apparent by silver staining
(data not shown).
As a source of potential binding partners for radixin, we used the
proteins of the cytoskeletal fraction of chicken erythrocytes. All of the
cytoskeletal radixin in these cells is in a single structure, the marginal
band, from which it can only be extracted by strong chaotropic agents
(Birgbauer and Solomon, 1989). The cells also are available in large
quantities. We prepared detergent-extracted cytoskeletons from
suspensions of chicken erythrocytes, solubilized these in 8M urea, removed
the urea by dialysis, and applied the extracts to the three radixin affinity
columns described above. After extensive washing, columns were eluted
with 125mM imidazole and the eluted proteins analyzed by SDS-PAGE.
Several erythrocyte proteins that bind to each of the columns, were detected
by silver stain, but we have identified by immunoblotting two known
proteins that bind either exclusively or preferentially to the N-domain.
-A -80kd erythrocyte protein is detected by silver stain in the eluant
from the N-domain column. That protein is identified as radixin by
immunoblotting with antibody #457, specific for the carboxy-terminus of
radixin, and antibody #220, which binds to an epitope present in the aminotermini of all ERM proteins (Fig. 5-1 A, AMINO both C and N lanes). This
band is not detectable in the eluates from the C-domain column, by either
immunoblotting, (Fig. 5-1 A, CARBOXY) or silver stain. These data are
consistent with the finding of Andreoli et al. (1994), who demonstrated that
full-length ezrin bound to its own amino-terminus substantially more
tightly than to its own carboxy-terminus. We do not detect ezrin or moesin
as proteins bound to the N-domain column, perhaps because they are much
less abundant in chicken erythrocytes than radixin (B. Winckler,
146
unpublished). We can not determine if radixin is bound to the FL-column,
since the two proteins should co-migrate.
-A -110kd polypeptide detected by silver stain binds preferentially to
the N-domain column. That protein is identified as band 4.1 in
immunoblots, using antibodies against chicken erythrocyte band 4.1 (Fig. 51A, AMINO, 4.1 lane). Granger and Lazarides (1985) demonstrated that in
chicken erythrocytes band 4.1 occurs in multiple isoforms with a wide
range of molecular weights, but that a species of -115kd (in their gel
system) is the major one. Our -110kd band co-migrates with the major
band 4.1 element in our hands. In some experiments, the antibodies
identify a much less intense band in the eluants from the C-domain column
and the FL column.
Several properties of the observed protein binding suggest that it is
specific. First, the binding is highly selective. Although some erythrocyte
proteins appear to bind to all three columns, in fact they and the specific
polypeptides named above account only for a small subset of the total
complement of proteins in the extract. Second, we can detect both tubulin
and vinculin by immunoblotting of the column flowthrough, but we detect
no signal above background among the proteins bound to the column (Fig.
5-1B). Third, neither band 4.1 nor radixin bind to column bearing an
irrelevant His-tagged protein of comparable size, dihydrofolate reductase.
In contrast, those erythrocyte proteins that bind to all three radixin
columns also bind to the DHFR control (data not shown).
We do not know if band 4.1 and radixin are ligands of radixin in the
cell. However, these experiments do suggest that, under the conditions of
this assay, specific associations do occur between a small subset of chicken
erythrocyte proteins and discrete domains of radixin.
Inhibition of interactions between the N-domain and in vitro ligands
by the C-domain. The observation that columns containing the radixin Ndomain retain a protein-band 4.1-that is not retained by FL columns raises
the possibility that, in full-length radixin, the carboxy-terminal domain
inhibits the binding properties of the amino-terminal domain. This effect
could occur because the presence of the domains as contiguous sequences
affects their conformation and therefore their binding properties.
Alternatively, the two domains could physically interact with each other in
147
Figure 5-1. Immunoblot analyses of column eluates. (A) High imidazole
eluates from columns bearing His-tagged polypeptides (AFFINITY
COLUMN) plus bound chicken erythrocyte proteins were collected. The
His-tagged polypeptides included full-length radixin (FULL-LENGTH), the
NH2- (AMINO) and COOH- (CARBOXY) terminal domains of radixin, and
dihydrofolate reductase (DHFR). The eluates were separated on 7.5% SDSPAGE, transferred to nitrocellulose, and probed with antibodies
recognizing the carboxy terminus (C) or amino terminus (N) of radixin, or
band 4.1 (4.1). The arrowhead indicates the position of radixin, the arrow
indicates the position of band 4.1 the numbers (kilodaltons) indicate
mobilities of four molecular weight markers. (B) Partitioning of vinculin,
radixin, and tubulin between flow through and column bound material,
detected by Western blots of fractions loaded to represent equal starting
material.
148
A
200-
111S9
66-*
44
4
*
45-1
C N 4.
PULLNTH
C N 4.C
AMINO
C N 4.
AINO +
CARUOY
CN 4.
AMINO +
DHMR
C N 4.1
CARUOXY
B
VINCULIN
RADIXIN
TUBULIN
FLOW
THROUGH
BOUND
C N 4.1
DHPR
a way that excludes other ligands. To distinguish between these
possibilities, we assayed the ability of the C-domain to interfere with the
binding properties of the N-domain: As shown in Figure 5-1 A, columns
bearing either the N-domain alone, or the N-domain mixed with a control
protein, DHFR, both bind radixin and band 4.1. In contrast, columns
bearing a mixture of the N-domain and C-domain bind neither radixin nor
band 4.1 (Fig. 5-1A, AMINO+CARBOXY). The results suggest that the Cdomain can inhibit the binding of ligands to the N-domain.
Direct detection of binding between the amino- and carboxy-terminal
domains of radixin in solution. The observation that the C-domain can
inhibit the binding properties of the N-domain strongly suggests that these
two domains bind to one another. To demonstrate that such direct binding
does occur, in solution, and to estimate the strength of binding, we used the
technique of affinity co-electrophoresis ("ACE"). Briefly, the two His6tagged polypeptides were subjected to electrophoresis within a single 1%
agarose gel in a physiological buffer (125mM KOAc, 50mM Hepes, pH 7.65).
We loaded the N-domain (at 10- 7 M) into a long transverse slot. We cast (in
agarose) the C-domain into 9 rectangular wells at concentrations from 0 to
750 nM. The anode was placed so that the faster migrating N-domain pass
through the zones containing the C-domain during most of the
electrophoresis. The mobility of the N-domain was then detected by
transferring the proteins out of the gel and probing with an amino
terminal-specific antibody. Figure 5-2 A demonstrates that, where
migrating N-domain encountered the C-domain, the migration of the
former was retarded in a manner that varied directly with the
concentration of the latter. In contrast, the migration of purified
glutathione-S-transferase--a control protein--was not affected by the Cdomain over the same range of concentrations (Fig. 5-2 B).
From measurements of mobility retardation in Figure 5-2 A, we can
estimate the dissociation constant for the interaction of the amino- and
carboxy-terminal polypeptides of radixin. Figure 5-2 C shows the analysis
of one such experiment. As described by San Antonio et al. (1993), to avoid
problems arising from saturation of the film, we used a Phosphorimager to
determine the true midpoint of each of the bands. The data have been fit by
an equation that assumes 1:1, non-cooperative binding, and a concentration
150
of N-domain of 5 X 10-8 M. The curve indicates an apparent value of Kd of
4.5 X 10- 8 M. A second experiment, analyzed in the same way, gave a value
for Kd of 4.2 X 10- 8 M (data not shown). Equations that assume higher order
or cooperative binding fit the data significantly less well (not shown).
The amino-terminal domain does not suppress in trans the
morphogenic properties of the carboxy-terminal domain in HeLa cells.
Based on the data presented above, one might predict that co-expression of
the amino- terminal domain would supress the deleterious effects of
expression of the carboxy-terminal domain in cells. In fact, there is
precedent for this effect. Simultaneous expression of the amino-terminal
domain of ezrin suppresses the morphogenic properties of ezrin carboxyterminal expression in insect cells (Martin et al., 1995). To test this in
mammalian cells, we co-expressed the domains of radixin in HtTA-1 cells.
To avoid problems that might arise from inconsistensies in co-precipitate
preparation, we split a single calcium phosphate precipitate-containing the
carboxy-terminal domain under the control of a constitutive promoter and
the amino-terminal domain under the control of the tetracyclinerepressible promoter-between two dishes of cells. One of the two dishes was
supplemented with tetracycline to repress expression of the aminoterminal domain. Proteins from the transfected cells were harvested 48
hours after transfection and analyzed by western blot with mAb 12CA5.
Figure 5-3 A shows that in the absence of tetracycline both the amino-and
carboxy-terminal polypeptides are expressed at roughly similar quantities.
In contrast, in the presence of tetracycline, only the carboxy-terminal
domain is expressed. We then scored cells stained with mAb 12CA5 for the
presence of the abnormal processes characteristic of cells expressing high
levels of the carboxy-terminal domain. We found no significant difference
in the percentage of 12CA5-positive cells with abnormal cortical processes
between cells expressing the carboxy-terminal domain alone (47.8%; n=596)
and those expressing both domains (51.9%; n=810). The processes were also
qualitatively similar between cells expressing the carboxy-terminal domain
alone (Fig. 5-3 B) and those expressing both domains ( Fig. 5-3 C).
151
Figure 5-2. Affinity co-electrophoresis of the N- and C- domains of radixin.
(A) Column-purified His6-N-domain (10- 7 M) was loaded into a long gel slot
perpendicular to the direction of electrophoresis. Column-purified His6 -Cdomain was loaded, at the concentrations given, into multiple lanes
parallel to the direction of electrophoresis. After electrophoresis, during
which time the migrating front of N-domain traversed the zones containing
the C-domain, the contents of the gels were transferred to nitrocellulose,
and visualized by immunoblotting with antibody against the N-domain. (B)
Electrophoresis was carried out as in panel A, except that glutathione-Stransferase was loaded into the slot and was detected using antibodies
specific for that protein. (C) Analysis of the binding of the N- and Cdomains of radixin as revealed by affinity co-electrophoresis. Reorr, the
corrected value of the retardation coefficient, quantifies the electrophoretic
retardation of the N-domain. C-Term gives the nominal concentration of
the C-domain in the lanes. Typically, the concentration of the detected
species can be ignored in ACE experiments because it is <<Kd. Here, that
assumption does not apply because detection of the N-domain required
relatively high concentrations. The nominal initial concentration of the Ndomain was 10- 7 M, establishing an upper limit. Because the band
broadens and diffuses during electrophoresis, the actual concentration of
N-domain was likely lower. Varying the assumed concentration of the Ndomain in the analysis from its highest possible value (10-7 M) to 1.25 x 108 M yielded values for Kd that varied about 4-fold range. The
data have been
fit to an equation derived from the definition of Kd, using the assumption
that bound fraction is equal to Rcorr/R., where R., represents the limiting
value of RcOrr when the concentration of the carboxy-terminal domain is
arbitrarily large. The curve was obtained using non-linear least squares
methods, in which Kd and R. were taken as variables to be fit
simultaneously.
152
I
9
z
9
LIO-
0
1i
0
[HUI
106)
[HUI
0n
'I;
0
0
0
0
Figure 5-3. Co-expression of the amino-terminal domain does not suppress
the effects of expression of the carboxy-terminla domain in HtTA-1 cells.
(A) Co-expression of amino- and carboxy-terminal polypeptides in HtTA-1
cells. HtTA-1 cells were co-transfected with a vector constitutively
expressing the carboxy-terminal domain (HACRADC) and one expressing
the amino-terminal domain (HACRADN) under the control of a
tetracycline repressible promoter (see Materials and Methods). After
transfection, cells were split equally into two dishes. We supplemented one
of these dishes with tetracycline to repress expression of the aminoterminal domain. Protein extracts 48 hours after transfection were
subjected to western blot analysis with mAb 12CA5. Lane 1- mock
transfected; lane 2- transfected plus tetracycline lane 3- transfected minus
tetracycline. The positions of the amino-(N) and carboxy-(C) terminal
polypeptides are indicated to the right of the blot. (B and C) Cells
transfected as described above were processed 48 hours after transfection
for immunofluorescence with mAb12CA5 as described in Materials and
Methods. B-plus tetracycline C-minus tetracycline.
154
A
-
1
-
2
3
~C
DISCUSSION
In summary, the data indicate that radixin has separable domains
capable of specific binding interactions in vitro. Binding partners for
radixin's domains include band 4.1 and radixin itself. Furthermore, the Nand C-domains of radixin can bind each other, in solution and with high
affinity. Like the intact full-length protein, the reconstituted complex of the
N- and C-domains fails to interact with band 4.1. Since the estimated
concentration of the N-domain and C-domain polypeptides in bacterial
extracts is -2.5 X 10-6 M, it is likely that complexes between those
polypeptides formed before adsorption to the column (Fig. 5-1 A). Taken
together with the results of transfection experiments presented in Chapter
Three, in which high level expression of the carboxy-terminal domain of
radixin had deleterious consequences that the full length protein did not
have, these data strongly suggest that direct interactions between the
amino- and carboxy-terminal domains of the radixin molecule in vivo
inhibit the interaction of radixin with other molecules. Presumably, such
inhibition is overcome under appropriate circumstances, either because
ligands with higher affinity or effective local concentration successfully
compete with the interaction between the amino- and carboxy terminal
domains, or because regulatory modifications (e.g. phosphorylation)
modulate the affinity of that interaction.
Discrepancies in results from co-expression of ERM protein domains
in cultured cells. Contrary to the findings of Martin et al. (1995) in which
the amino-terminal domain of ezrin suppressed in trans the morphogenic
properties of the carboxy-terminal domain of ezrin in moth ovary cells, we
found that co-expression of the two domains of radixin did not attenuate the
effects of expression of radixin's carboxy-terminal domain in HeLa cells.
While amino-terminal domain's suppression of the carboxy-terminal
domain's phenotype is a prediction of the results presented above, there are
several possible explanations for the negative result that we report here.
First, the baculovirus expression system used in Martin et al. (1995) most
likely produces several orders of magnitude higher expression levels of the
domains of ezrin. Perhaps these quantities are necessary for complex
formation in the cell. Alternatively, intermolecular interactions in
156
mammalian cells could compete for inter-domain interactions. For
example, the amino-terminal domain probably interacts with at least one
other molecule to allow for its cortical localization at high expression levels.
This binding reaction could out compete amino- to carboxy-terminal
domain binding if it was of higher affinity. Again, higher expression levels
in the moth ovary cells could saturate this binding reaction or these cells
might lack appropriate binding partners for the domains of mammalian
ERM proteins altogether. Other possibilities include intrinsic differences
between ezrin and radixin and possible interference of the HA-epitope tags.
The latter possibility seems unlikely because we measured the direct
association of radixin domains bearing His6 tags at their carboxy termini.
Intra- or inter-molecular interaction of the domains of radixin? It is
reasonable to propose that the interaction between amino- and carboxyterminal domains of radixin is intramolecular. Such a situation would be
strikingly similar to what has been observed in studies of vinculin, in
which intramolecular interaction between head and tail domains has been
shown to compete with the binding of presumptive ligands (Johnson and
Craig, 1994; 1995). Since ERM proteins can self associate, however, we
cannot rule out the possibility that interactions of the amino- and carboxy
terminal domains of radixin may also be intermolecular (Gary and
Bretscher, 1993; 1995). In this regard it is noteworthy that, in the present
study, affinity columns of the amino-terminal domain of radixin bound full
length radixin, but columns of the carboxy-terminal domain did not.
Similarly, Andreoli et al. (1994) reported that full-length ezrin binds its own
amino-terminus substantially more tightly than its own carboxy-terminus.
Although negative results must be interpreted with caution (protein
fragments, and proteins immobilized on solid supports, may have
artifactually altered binding properties), it is possible that interactions
between ERM proteins are not mediated by the same amino-domain to
carboxy-domain interaction demonstrated in the present study. A detailed
analysis of binding domains and their activities, both in vivo and in vitro,
will resolve these issues.
157
CHAPTER SIX:
Intermolecular interactions of radixin.
158
SUMMARY
The experiments described in Chapters Two through Five indicate
the following: 1) that appropriate subcellular localization of radixin occurs
through discrete determinants in the amino- and carboxy-terminal
domains, 2) that radixin competes with other ERM proteins for a factor
necessary for localization to cortical-cytoskeletal sites, and 3) that the
intermolecular associations of radixin are likely to be modulated by
intramolecular interactions between the amino-and carboxy-terminal
domains. Each of these findings suggest that radixin binding partners
exist in cells and make certain predictions about the relationship between
the domains of radixin and these putative binding partners. The
experiments described in this chapter are aimed at further elucidating the
nature of radixin's intermolecular associates. We found, contrary to
expectations for a cytoskeletal accessory protein, that the carboxy-terminal
domain of radixin was entirely released into the soluble fraction after
treatment of cells with non-ionic detergents. Also, the carboxy-terminal
polypeptide in this detergent soluble fraction did not show evidence for
association with detergent labile F-actin. These assays argue against a
tight association between insoluble and soluble pools of F-actin. However,
in direct binding assays, we did detect a specific, though relatively weak,
association between the carboxy-terminal domain and F-actin. An attempt
to identify cellular binding partners by co-immunoprecipitation in HeLa
cells yielded a 160 kDa candidate binding protein to the amino-terminal
domain.
159
JNTRODUCTION
ERM proteins are localized to cortical cytoskeletal structures. One
way that ERM proteins might achieve this particular cellular distribution
is if they bind to one or more other proteins present in the plasma
membrane and submembranous cytoskeleton. There is evidence for such
binding partners. All three ERM co-immunoprecipitate with the integral
membrane protein CD44 (Tsukita et al., 1994). Although initial efforts to
detect direct binding between intact ERM proteins and actin indicated that
these proteins did not bind (Bretscher, 1983), more recent evidence suggests
that the carboxy-terminal domain has an F-actin binding site that is
masked in the full-length molecule. Beads coated with a bacteriallyexpressed version of the carboxy-terminus of moesin are capable of
precipitating F-actin (Turunen et al., 1994). In blot overlay experiments, an
F-actin probe identifies the carboxy-terminal domain of moesin
(Pestonjamasp et al., 1995). It is important to stress that in both of these
experiments, binding was specific for F-actin- no binding was detected with
monomeric actin. Although the preceding experiments are suggestive of a
direct interaction between ERM proteins and F-actin, Shuster and Herman
(1995) provide evidence that ezrin from microvascular pericytes binds
indirectly and selectively to 1-actin. Other studies indicate that ERM
proteins may bind to each other in homo- and hetero-oligomeric complexes
(Gary and Bretscher, 1993). Finally, the data presented in Chapter Five
indicate that band 4.1 might be a radixin binding partner in the chicken
erythrocyte marginal band.
The work presented in Chapters Three, Four, and Five suggests that
as measured in some assays, the binding interactions of the parts of
radixin do not equal the binding interactions of the whole molecule. While
this could reflect exposure of binding sites in the isolated domains that are
never accessible for binding in the full-length molecule, another proposal is
that the exposure of these masked binding sites is biologically regulated in
the full-length molecule. In this chapter, we have explored the binding
interactions of radixin and its domains in cultured mammalian cells.
Here, we have employed subcellular fractionation techniques to address
broadly the nature of radixin's binding partners. The data are consistent
with a weak F-actin binding activity in the carboxy-terminal domain. Also,
160
we have attempted to identify proteins associated with radixin by coimmunoprecipitation. We found a 160 kDa protein that coimmunoprecipitated specifically with the amino-terminal domain in HeLa
cells.
161
MATERIALS AND METHODS
Subcellular fractionation experiments. For detergent extraction
experiments, 5x10 5 cells were grown overnight in 35mm tissue culture
dishes. The following day, these cells were washed with PBS at room
temperature, then lysed with 500 g1 extraction buffer consisting of : PM2G[100 mM PIPES; 2mM EGTA; 2M glycerol (pH 6.9)] (Solomon et al., 1979)
-plus 0.1% NP-40 and supplemented with the following protease inhibitor
cocktail- 0.04 TIU/ml aprotinin; 1 gg/ml PMSF; 1 gg/ml pepstatin A; and 1
gg/ml leupeptin (Sigma)- for 10 minutes at room temperature. After
incubation, 200 g1 lysate was removed and SDS sample buffer was added to
1X concentration. This sample was designated soluble or "S" fraction. We
washed the material remaining on the plate with 1 ml extraction buffer,
then dissolved it in 250 pl 1X SDS sample buffer. This sample was removed
from the plate and designated insoluble or "I" fraction. Both fractions were
boiled for 5' min and stored at -20'C until use. To detect HA-radxiin
polypeptides in these fractions, equal volumes were run on 7.5% SDS
polyacrylamide gels, transferred to nitrocellulose filters, and probed with
mAb 12CA5 as described in Chapter Three. Since the I fraction was
brought up in roughly half the volume used for the S fraction, the I fraction
is present roughly 2X concentrated on the SDS gels relative to the S fraction.
For actin fractionation experiments, cells were lysed and treated
essentially as described in Watts and Howard (1992). Cells on 100mm
dishes were lysed for 15 minutes at room temperature in 1 ml of the
following buffer: [10mM imidazole; 40 mM KCl; 10 mM EGTA (pH 7.15)]
plus 1% Triton X-100 supplemented with the following protease inhibitor
cocktail- 0.04 TIU/ml aprotinin; 1 gg/ml PMSF; 1 gg/ml pepstatin A; and 1
gg/ml leupeptin (Sigma). Lysed cells were scraped from the tissue culture
dish with a rubber policeman, decanted into an eppendorf tube, and spun at
12,00OXg in a microfuge for 2 minutes to pellet insoluble material. An
aliquot of the supernatant was removed and SDS sample buffer was added
to this aliquot for a final volume of 120 [d. This sample was designated low
speed supernatant or "LSS". The remaining supernatant was decanted to
an ultra-centrifuge and used in the subsequent steps. The pellet remaining
in the eppendorf tube was washed 1X with lml lysis buffer and120 p1 lysis
buffer plus 1X SDS sample buffer was added. This sample was designated
162
low speed pellet or "LSP". The remaining supernatant from the low speed
spin was placed in a TLA 100.3 rotor and centrifuged in a Beckman TL100
ultracentrifuge for 5 minutes at 82,000 rpm (360,00OXg). An aliquot of the
supernatant was removed and SDS sample buffer was added to this aliquot
for a final volume of 120 p1. This sample was designated high speed
supernatant or "HSS". The remaining supernatant was aspirated and 120
R1 lysis buffer plus 1X SDS sample buffer was added. This sample was
designated high speed pellet or "HSP". To detect HA-radixin polypeptides
in these fractions, equal volumes were run on 7.5% SDS polyacrylamide
gels, transferred to nitrocellulose filters, and probed with mAb 12CA5 as
described in Chapter Three.
Actin binding assays. We performed actin co-pelleting assays as
described by Matsudaira (1992) with modifications. His6 -tagged versions of
the entire carboxy-terminal domain (amino acids 318-583), RADC-3 (amino
acids 318-508), and RADC-6 (amino acids 509-583) -see figure 4-1 for a
diagram- were constructed and prepared from murine radixin cDNA
essentially as described for preparation of chicken His 6 -tagged proteins as
described in Chapter Five and dialyzed in actin binding assay buffer
(ABAB) [50-200 mM NaCl; 1mM MgC 2; 10mM PIPES; 0.1mM EGTA (pH
8.0)]-ATP (0.2mM final concentration) and DTT (0.2mM final
concentration) were added fresh to this stock solution before use. Rabbit
muscle F-actin was prepared from an acetone powder preparation (Sigma)
by resuspension in the polymerization buffer described by Spudich and Watt
(1971) [10mM Tris-Cl; 100mM NaCl; 1mM MgCI 2 ; 1mM ATP; 0.3mM NaN3
(pH 8.0)]. Protein concentrations were determined by Lowry assay. We
mixed the radixin polypeptides with F-actin together with ABAB in a final
reaction volume of 50p1 and incubated them at 4'C for 2 hours. After
incubation, the reaction mixture was transferred to a 5X20mm Beckman
airfuge tube. The fractions were spun in an airfuge at 23p.s.i. (-105,00OXg)
at room temperature for 15 minutes. The supernatant fraction was
removed (-50pl) and SDS sample buffer was added. This fraction was
designated supernatant or "S". The sucrose cushion was removed and the
pellet was resuspended in ABAB plus SDS sample buffer to the same final
volume as the S sample. This sample was designated pellet or "P". The
samples were boiled for 5 minutes then equal volumes of S and P samples
163
were run on 10% or 15% SDS polyacrylamide gels. The radixin proteins
were visualized by Coomassie staining and quantitated by densitometry.
Co-immunoprecipitation experiments. First, we metabolically
labeled stably transfected NIH-3T3 or transiently transfected HtTA-1 cells
in the following manner. 106-107 cells in 60mm culture dishes were
washed 2X in DME- medium [DME without cysteine, glutamine, or
methionine supplemented with 4mM glutamine (Sigma)] and then
incubated in this medium for 15 minutes to deplete intracellular stores of
cysteine and methionine. Then this medium was aspirated and cells were
incubated in 2ml of DME- supplemented with 200gCi
3 5 S-cysteine
and 35S-
methionine (Express Label, NEN) for 2 hours. Following this short-term
metabolic labeling, the cells were washed in PBS and then lysed in 0.5 ml
[50mM Tris-Cl (pH 8.0); 150mM NaCl; 0.1% NP-40 supplemented with the
following protease inhibitor cocktail- 0.04 TIU/ml aprotinin; 1 .g/ml PMSF;
1 pg/ml pepstatin A; and 1 pg/ml leupeptin (Sigma)] for 30 min at 4*C on a
rocking platform. The lysate was then transferred to an eppendorf tube and
spun at 12,00OXg for 5 minutes to remove insoluble material. Pilot
experiments showed that all of the radixin was released into the soluble
pool under these conditions. Next, the lysates were pre-cleared by
incubation on a rotator with 3gg irrelevant IgG2b for 1 hour at 40 C.
Immune complexes were removed by incubation for 1 hour at 40 C with
100gl of 10% protein A-sepharose (Sigma) and pelleting for 15 seconds at
maximum speed in a microfuge. The pre-cleared lysate was then
incubated with 3gg mAb 12CA5 for 1 hour at 4*C. Immune complexes were
removed by incubation for 1 hour at 4'C with 100pl of 10% protein Asepharose and pelleting for 15 seconds at maximum speed in a microfuge.
The beads were washed 2 times in 1 ml lysis buffer, then dissolved in 100gl
SDS sample buffer. To separate and visualize proteins in immune
complexes, samples were run on a 5-10% gradient gel. The gel was fixed in
7.5% glacial acetic acid/50% methanol then incubated in ENHANCE
(DuPont-NEN) according to manufacturers instructions for fluorography of
radiolabelled proteins. The gel was then dried and exposed to film.
164
RESULTS
Detergent fractionation of HA-radixin constructs in NIH-3T3 cells.
Previous reports from several groups demonstrate that the ERM proteins
are largely solubilized when cells are extracted with non-ionic detergents
(Gould et al., 1986; Bretscher, 1989; Birgbauer et al., 1991). We performed
these extractions on NIH-3T3 cell lines stably expressing HA-radixin
constructs. Figure 6-1 shows the results of extracting cells with 0.1% NP-40
in the cytoskeletal stabilization buffer PM2G. HAN-RAD and HAC-RAD
showed similar behavior: there is approximately two-fold more of each
protein in the detergent-soluble fraction than remains associated with the
extracted preparations. This result contrasts with their differential
localization properties as assessed by immunoflourescence microscopy (see
Fig. 3-3). Microscopic examination of extracted cells stained with mAb
12CA5 showed that the pre-extraction staining pattern of HAC-RAD was
largely disrupted, although some faint staining was detectable in microvilli
(data not shown). In contrast, the microtubule and microfilament
cytoskeletons were intact in these extracted cells as previously reported
(Solomon et al., 1979). Similar to the full-length proteins, we found the
amino-terminal domain in both fractions. Although compared to the fulllength proteins, more of the amino-terminal polypeptide is associated with
the insoluble material.
The behavior of the carboxy-terminal domain in this assay is wholly
different from the full-length molecule and amino-terminal domain. The
carboxy-terminal fragment is not detectable in the detergent-extracted
cytoskeletons, despite the fact that it clearly localized to cytoskeletal
structures in unextracted cells. Instead, all of the HA-tagged carboxyterminal domain was present in the detergent soluble fraction. As a
control, we probed these fractions taken from cells expressing the carboxyterminal domain with anti-vimentin antibodies (data not shown). All of
this material was in the detergent insoluble fraction indicating that other
proteins in these cells behave as expected in this assay. We reproduced
these results over several independent trials.
The carboxy-terminal domain does not co-pellet with detergent labile
F-actin. Some evidence supports the notion that a population of detergent
165
Figure 6-1. Detergent fractionation of HA-radixin polypeptides. Cultures of
NIH-3T3 cells stably expressing HA-radixin constructs were extracted with
0.1% NP-40 and separated into soluble (S) and insoluble (I) fractions as
described in Materials and Methods. 40gl of each of these fractions were
subjected to SDS-PAGE and Western blot analysis with mAb 12CA5.
Insoluble loads are two-fold more concentrated compared to soluble loads.
For simplicity, only HAN-RADN and HAN-RADC are shown, but similar
results were obtained with HAC-RADN and HAC-RADC.
Figure 6-2. Cellular fractionation of HA-radixin polypeptides. Cultures of
NIH-3T3 cells expressing HA-radixin constructs and untransfected NIH3T3 cells were lysed in 1% Triton X-100, transferred to an eppendorf tube
and spun at low speed as described in Materials and Methods. The pellet
from this step is designated LSP. An aliquot of the supernatant, LSS, was
taken and the remaining supernatant was subjected to high-speed
centrifugation to pellet F-actin as described in Materials and Methods. The
supernatant and pellet from this step were designated HSS and HSP,
respectively. Equivalent amounts of protein, based on the original lysis
volume, were separated by SDS-PAGE. Fractions were subjected to
Western blot analysis with mAb 12CA5 (HACRAD, HANRADN,
HANRADC) for transfected cells or pan-ERM antiserum #220 (ERM) for
untransfected cells. Only the relevant bands are shown.
166
NP-40 FRACTION
CONSTRUCT
s
I
HAC-RAD
HAN-RAD
-
HAN-RADN
HAN-RADC
Protein
__
_
_
_
HACRAD
HANRADN
4m
HANRADC
ERM
M
-
..
am
labile F-actin exists in cells (Cassimeris et al., 1990; Watts and Howard,
1992). It is thought that this population consists of short F-actin oligomers
not crosslinked into the higher order insoluble cytoskeletal matrix. We
wished to know if detergent soluble HA-radixin proteins associated with
this population of F-actin. We lysed NIH-3T3 cell lines expressing HAradixin constructs with non-ionic detergent and isolated F-actin from the
lysates by high-speed centrifugation. The results are shown in Figure 6-2.
Under these lysis conditions (which differ from those in the previous
experiment - see Materials and Methods) , all of the full-length protein and
carboxy-terminal domain and most of the amino-terminal domain are
released into the soluble fraction (compare LSP with LSS). Interestingly,
with this treatment, all of the full-length HA-radixin is released into the
soluble fraction. This contrasts with the behavior of this protein in the
previous experiment (compare HAC-RAD in Figs. 6-1 and 6-2). The basis
for these differential extraction properties are not clear. After
centrifugation of the lysate at >300,000g for 5 minutes, a small quantity of
the full-length protein is found in the pellet. A substantial portion of the
amino-terminal domain pellets under these conditions. However, there is
no carboxy-terminal polypeptide detectable in the high-speed pellet fraction.
We obtained similar results when HA-radixin polypeptides were isolated
and fractionated from transiently transfected HtTA-1 cells (data not
shown). Furthermore, when we took untransfected NIH-3T3 cells through
this assay, we found that endogenous ERM proteins, detected by pan-ERM
antiserum #220, behaved like full length HA-radixin (Fig. 6-2, ERM). As a
control to determine if actin was detectable in the high speed pellet fraction,
we probed Western blots with an anti-actin mAb and found that a
substantial amount of total cellular actin was present in the high speed
pellet (data not shown).
Direct in vitro F-actin binding assays. The preceding experiment
argue against a tight association of radixin or its domains with F-actin in
cells. However, these assays might not be sensitive to a weak binding
interaction. To test directly the proposition that the carboxy-terminal
domain interacts with F-actin, we performed in vitro binding experiments.
We mixed purified, bacterially expressed full-length and truncated forms of
radixin with a purified F-actin solution. When this mixture is exposed to
168
high centrifugal force, the F-actin pellets along with any associated
proteins leaving monomeric actin and unassociated proteins in the
supernatant. Pilot experiments indicated that at physiological salt
concentrations, very little of the full-length molecule pellets in an actindependent manner (data not shown). However, we found that actindependent pelleting of the full-length protein was influenced by the salt
concentration. At lower levels of NaCl, more of the full-length molecule
pelleted, but this fraction was still a minor component of the total protein in
the reaction (data not shown). Interestingly, Hanzel et al., (1991) reported
that the association of ezrin with detergent extracted cytoskeletons from
gastric parietal cells was sensitive to salt in the same manner reported
here.
At physiological salt concentrations, a substantial amount of the
carboxy-terminal domain was found to pellet in an actin-dependent
manner. To investigate this binding further, we titrated actin against a
constant concentration (5pM) of radixin polypeptides. As shown in Figure
6-3 A the C-3 portion of the carboxy-terminl domain lacking the final 75
amino acids did not show any actin-dependent enrichment in the pellet
fractions. The small amount of protein that does show up in the pellet
fraction is probably aggregated material. In contrast, the C-6 fragment of
murine radixin showed a clear enrichment in the pellet fraction that
depended on the concentration of actin (Fig. 6-3B). It was difficult to test the
entire carboxy-terminal domain in this sort of assay because it migrates
close to actin and is therefore obscured at the higher actin concentrations.
The result presented above indicated that the final 75 amino acids of
radixin are capable of binding F-actin in solution. To measure the strength
of this interaction, we titrated concentrations of C-6 ranging from 1 to 50
pM against 2 gM actin. A Scatchard analysis of the data from this
experiment is presented in Figure 6-4. The plot indicates that there are two
non-identical, independent binding sites for F-actin. A higher affinity
binding site on C-6 has a Kd- 3 pM and is predicted to have a binding
stoichiometry of one to one with actin. A lower affinity site also exists and
was not saturable over the concentration range tested.
169
Figure 6-3. The carboxy-terminal domain of radixin binds F-actin in
solution. Actin binding assays were performed with the C-3 (A) and C-6 (B)
regions of the carboxy-terminal domain of radixin (see Fig. 4-1 for a
schematic of these polypeptides) as described in Materials and Methods. In
these experiments, actin was titrated against a constant amount (5 RM) of
radixin polypeptide. Concentrations of actin are indicated at the top of blot.
Binding reactions were subjected to high-speed centrifugation to separate
F-actin in the pellet fraction (P) and G-actin in the supernatant fraction (S).
The positions of actin and the radixin polypeptides are indicated at the right
of the gel. The results show that the presence of C-6, but not C-3, in the
pellet fraction is influenced by the concentration of F-actin.
170
A
C-3
[actin] pM
Fraction:
25
12.5
P S P
6.25
3.13
S P S P
0
1.56
S
P S
3.13
1.56
P
s
actin.,
C-3
B
C-6
[actin]
sM:
Fraction:
actin__
C-6
25
12.5
6.25
0
P S P S P S P S P S P S
Figure 6-4. Scatchard analysis of data from C-6 actin binding experiment.
Concentrations of C-6 ranging from 1 to 50 gM were titrated against 2 kM
actin as described in Materials and Methods. Intensities of Coomassie blue
stained bands in S and P fractions were quantified by densitometry. Values
were corrected for amount of C-6 which pellets independent of actin (about
3% of total in these experiments.
172
C-6 actin binding
data
0.4Bmax= -. 95 mol C-6/ mol actin
Kd= -3 gM
0.3-
0.2-
0.1 0
1
2
[C-6eactin]
3
(g M)
4
5
Figure 6-5. Co-immunoprecipitation of cellular proteins with radiin and
its domains. HA-radixin polypeptides were immunoprecipitated with
mAb12CA5 from extracts taken from 3 5 S-methionine-labelled NIH-3T3
cells (left blot) or HtTA-1 cells (right blot) expressing HA-radixin constructs
as described in Materials and Methods. After separation of the
immunoprecipitates by SDS-PAGE, the radioactive bands were visualized
by fluorography. The identity of the particular HA-radixin polypeptide
being expressed in each extract is indicated under each lane.
Untransfected NIH-3T3 and HtTA-1 cells were included as background
controls. The positions of the immunoprecipitating HA-radixin
polypeptides are indicated to the right of the blots and the positions of the
molecular weight standards and a 160kDa protein co-immunoprecipitating
with the amino -terminal domain of radixin in HtTA-1 cell extracts are
indicated to the left of the blots.
174
alI
S
4
4
z
HtTA-1
HANRADC
HANRADN
HACRAD
HANRAD
NIH3T3
RANRADC
HANRADN
HACRAD
HANRAD
A 160 kDa protein co-immunoprecipitates with the amino-terminal
domain in HeLa cells. One approach to identifying interacting proteins is
co-immunoprecipitation. Addition of the epitope tag sequence allows for the
detection of proteins that co-immunoprecipitate with the domains of radixin
as compared to the full-length molecule. We immunoprecipitated HAradixin polypeptides from 35S-labeled cells expressing HA-radixin
constructs with mAb 12CA5. The results are shown in Figure 6-5. Both
versions of the full-length protein and amino-terminal domain are
precipitated by mAb 12CA5 in transfected NIH-3T3 and HtTA-1 cells. Other
experiments demonstrated that roughly half of the total amount of HAradixin polypeptides are precipitated from cell extracts under the
conditions used (data not shown). In contrast, there was no detectable
enrichment of the carboxy-terminal domain in these radiolabelled
preparations. This result was not due to poor labeling of the carboxyterminal domain relative to the other proteins because Western blots with
mAb12CA5 showed that the carboxy-terminal domain was not detectable in
the immunoprecipitated material (data not shown). One possibility to
explain this result is that the HA-epitope in the carboxy-terminal domain is
occluded in these cell extracts. However, both HAN-RADC and HACRADC failed to immunoprecipitate (data not shown). Although the fulllength and amino-terminal polypeptides were immunoprecipitated by
mAb12C5, no other species co-enriched with these proteins in NIH-3T3
cells. However, in HtTA-1 cells there was specific co-enrichment of a 160
kDa protein in amino-terminal domain precipitates. This band was not
enriched in either full-length immunoprecipitate or in mock transfected
cells. Curiously, as mentioned above, this species was not detected in NIH3T3 cells stably expressing the amino-terminal domain.
176
DISCUSSION
Insights into the intermolecular associations of radixn through
subeellular fractionation studies. Resistance to extraction by non-ionic
detergent is a hallmark of cytoskeletal accessory proteins. The behavior of
ERM proteins in this sort of assay is complex and depends on the cell type
examined. For instance, in at least two cell types -chicken erythrocytes and
gastric parietal cells- ERM proteins are highly enriched in the detergent
insoluble fraction (Birgbauer and Solomon, 1989; Hanzel et al., 1991). In
some cultured mammalian cell lines however, there seems to be
substantially more ERM protein present in the detergent soluble fraction
(Gould et al., 1986; Bretscher, 1989). In P19 cells, RA induced
differentiation results in a dramatic increase in the detergent insoluble
fraction of ERM proteins (Birgbauer et al., 1991). This might be interpreted
as consistent with the notion that more ERM protein is being recruited to
the cortical cytoskeleton in the differentiated state, but another experiment
suggests that ERM proteins can be recruited to the cortex without a
detectable change in detergent fractionation properties. Epidermal growth
factor induces rapid phosphorylation and concomitant recruitment of ezrin
to the cortical cytoskeleton in A-431 cells. However, no change is detected in
the detergent fractionation properties of this protein after EGF stimulation
(Bretscher, 1989). Some of the variability in these results might be
attributable to different extraction conditions used in each experiment, but
it is still not clear how the presence of ERM proteins in the submembranous
cytoskeleton correlates with their detergent fractionation properties.
With this background in mind, we have examined the subcellular
fractionation of the domains of radixin. The original membranecytoskeletal linker model for the ERM proteins (see Chapter One) suggested
that the amino-terminal portion of the molecule should interact with an
integral membrane component which left the carboxy-terminal domain to
interact with the cytoskeleton. This model makes the simple prediction that
the amino-terminal domain should be detergent soluble and the carboxyterminal domain should be insoluble. Quite to the contrary, we found that
in stably transfected cells, the amino-terminal domain was present in both
the soluble and insoluble fractions, in roughly equal abundance, and the
carboxy-terminal domain partitioned exclusively with the detergent soluble
177
fraction. Our results do not agree with those from another study. Algrain
et al., (1993) found that in transiently transfected CV-1 cells the aminoterminal domain of ezrin was present almost entirely in the detergent
soluble fraction whereas the carboxy-terminal domain of ezrin cleanly
partitioned with the detergent insoluble fraction. We cannot explain why
our experiments with the domains of radixin should differ so drastically
from their results with the domains of ezrin. We do not think it is the result
of transient versus stable expression in the cells because we find similar
detergent fractionation results for the domains of radixin in both stably
transfected NIH-3T3 cells and transiently transfected HtTA-1 cells.
Our results are puzzling. The carboxy-terminal domain in the stably
transfected cells clearly localizes to the cortical cytoskeleton and stress
fibers. Yet, it is completely soluble. This result suggests that localization of
ERM proteins to cortical cytoskeletal structures is not strictly correlated
with resistance to detergent extraction. Further evidence for this
proposition comes from a comparison of HAC-RAD and HAN-RAD. Both of
these proteins show similar detergent fractionation properties, but only
HAC-RAD localizes to cortical cytoskeletal structures. Based on our
results, it is reasonable to suggest that the element that confers detergent
insolubility on the full-length molecule resides in the amino-terminal
domain, since this domain shows some resistance to detergent extraction.
However, the biological relevance of the detergent soluble versus detergent
insoluble fractions of the total pool of ERM proteins remains at issue.
One idea that we explored here was that the detergent soluble
fraction of radixin or its domains associated with detergent labile F-actin.
Detergent labile F-actin was first characterized by Watts and Howard (1992)
in polymorphonuclear leukocytes. This population of F-actin is thought to
be comprised of short oligomers that are not cross-linked into the detergent
insoluble matrix. The function of this type of F-actin is unknown, but some
studies indicate that this pool of F-actin is involved in the rapid extension of
processes when polymorphonuclear leukocytes are stimulated with
chemotactic peptides (Cassimeris et al., 1990; Watts and Howard, 1993).
Other studies indicate that detergent labile F-actin is present in
lamellipodia of motile cells (Small et al., 1995). We found that a small
amount of the full-length and amino-terminal polypeptides co-pelleted with
the detergent labile F-actin, but none of the carboxy-terminal domain was
178
present in this fraction. This result argues against the association of the
carboxy-terminal domain with detergent labile F-actin although the copelleting assay used here might not detect low affinity interactions.
An F-actin binding site in the carboxy-terminal domnin of radixin.
Previous work by others suggested the existence of an F-actin binding site
present in the carboxy-terminal domain of ezrin and moesin that was
masked in the full-length molecule. Tsukita et al. (1989) originally
identified radixin as a microfilament barbed-end capping protein from rat
liver adherens junctions, but there have been no subsequent reports to
substantiate this function. Moreover, barbed-end capping activities have
not been reported for the other ERM proteins. The work presented in
Chapter Four identified a domain at the extreme carboxy-terminus that
had effects on cells consistent with its interaction with the microfilament
cytoskeleton. We tested the possibility that an F-actin binding site existed in
the carboxy-terminus of radixin. Our results indicate final 75 amino acids
of radixin can bind to F-actin in solution. This interaction is specific in that
it is a property of this fragment in isolation or in the larger context of the
carboxy-terminal domain, but not of other proteins- for example full-length
radixin or the C-3 fragment of the carboxy-terminal domain. The C-6
fragment of radixin binds stoichiometrically with actin with Kd- 3 RM.
This result is comparable to the report that the isolated carboxy-terminal
domain of another membrane-cytoskeleton linking protein, vinculin, binds
F-actin with Kd~ 1 gM (Johnson and Craig, 1995). Our findings are also in
good agreement with others that have used solid phase assays to
demonstrate association of F-actin with the extreme carboxy terminus of
other ERM proteins (see Introduction). We did not detect a quantitative
difference of actin in the pellet fractions as a function of the concentration of
radixin carboxy-terminal polypeptides. This negative result argues against
a barbed-end capping activity for radixin. An inversely proportional
relationship should exist between radixin protein concentration and the
amount of pelletable F-actin if it were a bona fide barbed-end capping
protein. As mentioned, we did not detect F-actin binding by the carboxyterminus in cell fractionation experiments. This could be explained by the
fact that a weak binding interaction could be lost due to the high dilution
factors after cell disruption. Could this weak binding between the carboxy179
terminus of radixin and F-actin be meaningful? Given the high
concentration of actin in the cell (150-600 gM) and the possibility that the
local concentrations of actin in cortical membrane structures such as
filopodia where ERM proteins reside could be even higher, a low affinity
interaction might be expected. Perhaps, then, this low affinity interaction
as measured in solution is reflective of a direct interaction between radixin
and F-actin in the cell.
Another possibility is that the putative interaction between cortically
localized ERM proteins and F-actin is indirect. Evidence for an indirect
interaction with F-actin comes from Shuster and Herman (1995). They
show evidence that the association of full-length ezrin with 1-actin is
indirect. A factor that supports ezrin binding to 1-actin might be
fractionated from vascular pericytes (C. Shuster, personal
communication). Elucidation of the nature of the association between ERM
proteins and F-actin awaits further exploration.
Association of the amino-terminal domain of radixin with a 160kDa
protein in HeLa cells. We found that a protein with an apparent molecular
weight of 160 kDa co-immunoprecipitates from HeLa cell extracts
specifically with the amino-terminal domain of radixin. Although we have
not demonstrated that this is a direct association between these two
polypeptides, the 160 kDa species could be a binding partner for this domain
of radixin in cells. However, we do not detect this protein coimmunoprecipitating with the amino-terminal domain in extracts taken
from stably transfected NIH-3T3 cells. We do not know the basis for this
result. However, that the 160 kDa protein associates with the aminoterminal domain, but not the full-length molecule, is further evidence that
specific binding sites may be accessible in the isolated domains of radixin,
but masked in the full-length molecule.
It is difficult to speculate on the identity of the 160kDa species simply
from its molecular weight. Tsukita et al. (1994) reported an interaction
between ERM proteins and a 140 kDa isoform of CD44. CD44 is a
heterogeneous group of polypeptides (Lesley et al., 1993). Therefore we
cannot exclude this protein as the one interacting with the amino-terminal
domain of radixin in our studies solely on the basis of molecular weight.
However, the 160 kDa co-immunoprecipitating protein is probably too large
180
to be band 4.1 which was implicated as a binding partner for radixin in the
chicken erythrocyte marginal band in Chapter Five. Although band 4.1
shows considerable molecular heterogeneity, no species has yet been
reported at or near 160 kDa in mammals (Conboy et al., 1991). Future
experiments with specific antibodies will test the possibility that the 160 kDa
protein is either CD44 or band 4.1.
181
CHAPTER SEVEN:
A model for the molecular organization of radixin.
182
SUMMARY
The work presented in Chapters Two through Six was directed
toward understanding the function of radixin. Initially, we asked what
pieces of the molecule were responsible for its targeting to cortical
cytoskeletal structures in cells. We found that the localization of radixin is
dependent on distinct determinants resident in the amino- and carboxyterminal domains of the protein. These structural domains of radixin have
quite different functional consequences when expressed in cells at high
levels. In particular, expression of the carboxy-terminal domain, but not
the amino-terminal domain or the full-length protein, has deleterious
consequences including induction of abnormal cortical processes. It is
possible that this reflects an amplification of the normal function of
radixin. These results suggested that the amino-terminal domain might be
regulating the activities of carboxy-terminal domain in the context of the
complete molecule. We tested this idea in vitro. Briefly, we showed that the
amino- and carboxy- terminal domains interact directly in solution with
high affinity. This interaction inhibits the binding of the amino-terminal
domain to other proteins consistent with the notion that certain specific
binding sites are masked in the whole molecule. Taken together, the
results have suggested an overall organization for the radixin molecule. In
this chapter we outline this model, discuss unresolved issues about ERM
protein structure and function, and consider future prospects for testing
the hypothesis presented here.
183
ThE MODEL
The following model for the structure of the radixin molecule is
based, in part, on our analyses of the in vivo and in vitro functions of its
domains. Three findings in particular sketched the outline for this rather
simple model: 1) when separated from the full length molecule, the
carboxy-terminal domain exerts dramatic effects on the structure and
function of the actin-based cytoskeleton in cells 2) the amino- and carboxyterminal domains can form a high affinity complex in solution 3) formation
of this complex inhibits binding of other molecules to the amino-terminal
domain. Figure 7-1 shows our current conception of how the
structure/function relationship of the radixin molecule might be organized.
The essential features of this model are that radixin can exist in at least two
stable conformational states in the cell and that a biological regulatory
mechanism controls the inter conversion between these two states. In the
"closed" state, the amino-and carboxy-terminal domains are tightly
associated with one another within the molecule. In this state,
intermolecular binding sites are masked. Presumably this conformation of
the molecule is incapable of localization to cortical structures. The binding
between the amino- and carboxy-terminal domains in the closed state is
probably very tight. We measured a dissociation constant in the nanomolar
range for the complex between the isolated domains in Chapter Five.
Although it is also possible that steric constraints within the intact
molecule lead to a higher actual dissociation constant that that measured
here for the isolated domains, entropic considerations suggest that if
binding between these domains occurs in the intact molecule then the
dissociation constant for this interaction could be much lower than
nanomolar. For this reason, it seems necessary to invoke a mechanism
that can regulate the very tight intramolecular association. Due to the
action of an unknown regulatory mechanism (represented by the arrow),
the intramolecular association is weakened and the molecule assumes an
"open" state. In this state, intermolecular binding sites are unmasked so
that radixin is free to associate with other molecules (X and Y for
simplicity). This conformational state of the molecule does have the ability
to localize to cortical structures. We envision that the modifications that
promote the open state are reversible (hence, the bidirectional arrow) so
184
Figure 7-1: A model for the molecular organization of radixin
CONFORMATION:
OPEN
CLOSED
Mr
X
N
N
C
C
LOCALIZATION:
CYTOPLASMIC
CORTICAL
that the total pool of radixin in the cell could be divided between these two
conformational states. Depending on the particular needs of the cell, the
ratio between "open" and "closed" molecules might be rapidly adjusted.
This mechanism could explain the complex and dynamic localization
patterns of radixin in motile membrane-cytoskeletal structures such as, for
example, growth cone filopodia.
Other structural information about radixin. Although no atomic
resolution structural information for radixin exists at present, it is worth
considering computer modeling of secondary and tertiary structure of
radixin in light of the model presented above. The higher-order structural
features predicted from the primary sequence are a globular aminoterminal domain, comprised of the first 300 amino acids, which bear
sequence similarity to the amino-terminal domains of the other band 4.1
superfamily members, followed by an extended carboxy-terminal portion of
the molecule. From roughly residues 300 to 470, there is a strong
propensity for alpha helix formation. The model proposes a linker region
that is flexible to allow for conformational changes in the molecule. This
alpha helical region could serve this role. Alpha helices are typically short
-10 to 15 residues- and thought to be rather rigid, but long alpha helices can
allow for flexibility (see discussion of the alpha helical region of caldesmon
below). Instead of a long continuous alpha helix, this region might consist
of several shorter helices packed together. This sort of arrangement has
also been shown to be capable of dramatic, regulable changes in
conformation (Carr and Kim, 1993). In radixin, a strech of 8 proline
residues follows the helical region which could break the helical domain
due to the constraints that proline residues impose on the peptide backbone.
Polyproline tracts can also assume dynamic helical conformations (Lin and
Brandts, 1990). It is worth noting, however, that radixin and ezrin have
this polyproline region, but moesin does not. The final 106 amino acids are
largely hydrophilic with a net charge of -8. As discussed and demonstrated
in Chapter Six, the F-actin binding domain resides in this region. Taken
together the putative structure of radixin is consistent with the proposed
model: a globular amino-terminal domain binding to the extended carboxyterminal domain with some provision for a flexible linker between the two
domains.
186
Others have mapped the interacting regions in the amino- and
carboxy-terminal domains of ezrin. Using a bioassay, Martin et al. (1995)
showed that amino acids 1-233 suppressed the cell extension properties of
amino acids 310-585 of ezrin in moth ovary cells implying that these two
domains physically interact. In blot overlay assays, Gary and Bretscher
(1995) showed that a minimal polypeptide of amino acids 1-296 bind to
carboxy-terminal amino acids 479-585. Given the high degree of sequence
conservation between radixin and ezrin, it will not be surprising to see that
these same smaller domains of radixin are capable of interacting.
Intramolecular masking of F-actin binding domains: a common
motif in actin binding proteins? As mentioned in Chapter One, many actinbinding proteins have a chimeric nature-they are composed of several
domains each with different tasks. Several of these proteins appear to be
organized in a manner similar to that proposed above for radixin and are
considered in turn below.
Vinculin- The following picture of vinculin, another membranecytoskeleton linking protein, was emerging during our studies and has
obviously influenced our thinking about radixin. Vinculin is a protein that
is enriched at cell-cell and cell-substratum junctions where there is close
apposition of the cytoskeleton and cell-membrane. For many years, it was
thought that vinculin served as a linker between talin, which bound the
membrane spanning integrins, and a-actinin, which contacted F-actin. At
least one group claimed that native vinculin bound F-actin directly, albeit
very weakly, but direct binding of vinculin to F-actin remained
controversial (Gilmore and Burridge, 1995). This issue took an interesting
turn when Johnson and Craig (1994) first showed that proteolytic
fragments of vinculin- the 95 kDa head domain and the 30 kDa tail- bound
tightly to one another (Kd -50 nM) and that talin and the tail domain
compete for binding to the head domain. Subsequently, this same group
demonstrated that unlike the intact, full-length molecule, the tail domain
associates reasonably well (Kd -1gM) with F-actin (Johnson and Craig,
1995). Furthermore, the head domain could inhibit the binding of the tail
domain to F-actin. These results led the authors to propose that an interdomain interactions must be regulated in the whole molecule to allow for
engagement of intermolecular binding partners. It is interesting that
187
radixin and vinculin, which bear little primary sequence similarity to one
another, may be organized similarly in tertiary structure. Perhaps this is
an evolutionarily conserved design for proteins that are involved in
dynamic associations between the membrane and cytoskeleton.
Caldesmon- Caldesmon is an actin-binding protein that plays a role
in calcium-dependent regulation of smooth muscle cell contraction. The
amino-terminal domain of caldesmon binds myosin and the carboxyterminal domain binds actin and calmodulin. These two domains are
connected by long alpha helical region in muscle isoforms of caldesmon.
There is evidence that this strech of 146 amino acids exists as a single,
continuous alpha helix (Albert Wang et al., 1991). Physico-chemical
studies have demonstrated two conformations of the caldesmon molecule in
solution- one extended and one in a hairpin configuration in which the
amino- and carboxy-termini in close proximity with one another (Martin et
al., 1991). This arrangement attests to the flexibility of the central alpha
helical region. Finally, calmodulin binding to the carboxy-terminus
inhibits myosin binding to the amino-terminal domain indicating that
intramolecular communication may be important for the function of this
molecule (Crosbie et'al., 1994).
In sum, the sort of mechanism proposed here for radixin - regulation
of intermolecular associations by modulation of intramolecular
associations- is probably a relatively common theme in nature. For
instance, src family kinases are thought to exist in an inactive
conformation due to intramolecular binding of their SH2 domains to a
phosphotyrosine residue. Engagement of SH3 domains with ligands may
activate the molecule by releasing the SH2 domain for intermolecular
binding (Superti-Furga et al., 1993). For these signaling molecules, as well
as for radixin and the other proteins mentioned, this design could allow for
rapidly registering biological events in a spatially and temporally restricted
manner.
188
UNRESOLVED ISSUES
Our understanding of ERM proteins is in many ways still
rudimentary. I would like to comment here on what I believe to be some of
the most important issues that need to be addressed to more fully appreciate
the biology of these molecules.
Cellular regulatory mechanisms- A variety of information points to
the fact that function of ERM proteins is highly regulated. For example, the
localization of ERM proteins in growth cone filopodia changes rapidly in
response to changes in motility of that organelle (C. Gonzalez Agosti,
unpublished results). In the model proposed above, we speculate that some
regulatory mechanism(s) modulates the conformational state of the
molecule. The open state is competent for localization in cortical
cytoskeletal structures, such as growth cone filopodia, while the closed
state is not able to localize. What mechanism(s) might account for the
opening of the molecule? Generally, these mechanisms might involve
either covalent or non-covalent modifications to the molecule. For instance,
radixin might bind non-covalently to a particular ligand which causes the
opening of the molecule. In the case of growth cone filopodia, this ligand
might be sensitive to extracellular cues. One covalent modification,
phosphorylation, is already correlated with ezrin localization to cortical
structures. Bretscher (1989) found that upon epidermal growth factor
stimulation of A-431 cells, ezrin is rapidly (within seconds) recruited to cell
surface structures and this event was correlated with the kinetics of
tyrosine phosphorylation of this molecule. One of the residues that is the
site of epidermal growth factor induced phosphorylation in ezrin, tyrosine145, is conserved among the ERM protein family members (Kreig and
Hunter, 1992). However, it is not yet clear what role, if any,
phosphorylation plays in cortical localization of ERM proteins. If
phosphorylation is not involved in the establishment of the open state, then
perhaps it is involved with its maintenance. Because the localization
properties of ERM proteins are dynamic, we envision that the processes
which lead to the opening of the molecule are rapidly reversible. In this
regard, dephosphorylation could be a candidate mechanism for the reverse
189
reaction. It is interesting that several phosphatases share the band 4.1
homology in their amino-terminal domains (see Fig. 1-1).
Intriguing possibilities for the regulation of ERM proteins are raised
by recent experiments with small GTPases. The activity of these regulatory
molecules are tightly regulated by guanine nucleotide binding and
hydrolysis. The GTPases rho, rac, and cdc42 have been implicated in the
modulation of cortical cytoskeletal structures (Nobes and Hall, 1995). These
molecules might regulate ERM protein conformation directly, or they could
lead to the generation of small-molecule second messengers that might
modulate ERM protein structure and function (Abo et al., 1992; Takaishi et
al., 1995; Hartwig et al., 1995).
Binding partners- The cellular binding partners for ERM proteins
are not yet firmly established, in part because these molecules have not
been particularly amenable to approaches for identifying binding partners
that rely on chemical affinity such as co-immunoprecipitation. Perhaps
this is a reflection of the transient nature of ERM protein associations in the
cell. However, some of the suspected binding partners can be considered in
terms of the model. That the amino- and carboxy- termini of radixin can
bind to one another suggests that this molecule might form dimers or
oligomers. Although first principles suggest that the intramolecular
binding should be highly favored over the multimeric configuration due to
entropic considerations, these domains might be arranged in the intact
molecule in such a way that inter-ERM protein interactions predominate
over inter-domain interactions. Physico-chemical data on ezrin purified
from brush border microvilli indicated that it behaved as a monomer
(Bretscher, 1983). Subsequently, Gary and Bretscher (1993) found that a
small amount of total ezrin and moesin co-immunoprecipitated from A431
cell extracts. Similarly, Andreoli et al. (1994) reported that ezrin coimmunoprecipitated with two proteins presumed to be radixin and moesin.
However, it was not entirely clear from these studies whether the ezrin and
other ERM proteins in these complexes are directly associated. The
immune complexes isolated by Gary and Bretscher were also unusual in
that they were extremely stable: exogenously applied ezrin did not exchange
into the complex. Later, using blot overlay assays, these authors showed
that the amino- and carboxy-terminal domains of ezrin bind one another
190
(Gary and Bretscher, 1995). This led to the notion that dimer formation was
a result of "head-to-tail" interactions. If the same interaction between the
isolated domains of radixin that we measured in Chapter Five mediates the
formation of multimeric forms of radixin, then each radixin molecule
should have two high affinity binding sites for another. This bivalent
association could explain the remarkable stability of the apparent homoand hetero-dimeric complexes. Still, the physiological relevance of ERM
protein remains to be determined. Perhaps this low abundance
configuration of ERM proteins is a byproduct of the open state, postulated in
the model, that is not functionally important. Although dimeric forms of
radixin could have properties that monomeric forms of radixin do not have,
our work suggests that certain intermolecular association could be
disfavored. Head-to-tail dimers could mask intermolecular binding sites in
the same way that we observed these sites masked in the complex between
the amino- and carboxy-terminal domains of radixin.
The nature of the association between ERM proteins and actin
remains enigmatic. It is an attractive proposition for ERM proteins to bind
directly to F-actin. This putative interaction could, in part, explain their
localization in cortical cytoskeletal structures in a wide variety of cell types.
We have shown here evidence for a specific, albeit weak, interaction
between the extreme carboxy terminus of radixin. A weak interaction
between ERM proteins and F-actin could be biologically relevant. As
mentioned previously, the local concentrations of F-actin in cortical
cytoskeletal structures could be quite high. Moreover, abundant evidence
indicates that ERM proteins are involved in dynamic, transient interactions
at cortical sites so weak binding might be expected. Focal contacts serve as
a model for this latter argument. Many components of this multi-protein
complex have transient associations in the cell that, when measured in
biochemical assays, are relatively weak (Horwitz et al., 1986). However,
caution is generally warranted when interpreting the meaning of weak
binding interactions. Our experiments are further complicated by the fact
that we have only detected binding of a portion of radixin, not the full-length
molecule. Of course, one can easily reconcile this fact with the model
presented above.
191
Redundancy of ERM protein function- Another outstanding issue is
the functional interchangeability of ERM proteins. We have not directly
addressed that issue in this work, but some of our results suggest that ERM
proteins are functionally redundant in cellular assays. For instance,
radixin can apparently replace moesin (and ezrin- see Appendix Three) in
cortical cytoskeletal structures without any observable effects on cells.
Perhaps the different cellular localization patterns and tissue expression
profiles of ERM proteins only reflect the relative abundance of these three
proteins in different tissues, and do not indicate specialized functions.
Genetic analysis might resolve this issue.
ERM proteins and the band 4.1 superfamily- What are the
relationships between ERM proteins and the rest of the band 4.1
superfamily? We showed in Chapter Five that radixin can associate with
band 4.1, but more work is necessary to determine the relevance of this
association. It is possible that the amino-terminal domains of other band
4.1 family members (see Fig. 1-1) act like the amino-terminal domain of
radixin by regulating carboxy-terminal functions. In fact, there is
preliminary support for this notion. In a manner analogous to radixin,
expression of the carboxy-domain of merlin has effects on cells unlike those
of the full-length protein or its amino-terminal domain (R. Shaw,
unpublished results).
ERM protein function? - It seems that the function of these molecules
is probably more complex than simply acting as a physical link between the
plasma membrane and the cytoskeleton. Our results point toward an active
role for radixin in organizing the submembranous cytoskeleton, although
the details of this activity are far from certain. Perhaps ERM proteins
mediate lateral interactions between microfilaments and the plasma
membrane. Alternatively, ERM proteins might uncap cortical
microfilaments to allow for polymerization. These sorts of functions could
play critical roles in organizing cell shape and motility. Along these lines,
it is interesting that ERM proteins are among the proteins recruited to the
"comet tail" formed by the bacterium Listeria monocytogenes as it moves
about within the cytosol of eukaryotic cells (Temm-Grove et al., 1994).
192
TESTING THE MODEL
In this work, we have focused primarily on the properties of the
isolated domains of radixin. These studies have led to the model proposed
above. Obviously, this model must be tested and understood for the whole
radixin molecule, as this is its naturally occurring form. Below we suggest
several lines of experimentation designed to explore the structure and
function of the native radixin molecule in more depth.
Molecular structure- The essential feature of the proposed model for
radixin involves a conformational change of the molecule. In this lies the
difficulty of probing its 3-dimensional structure in a cellular context, an
exceedingly difficult task. As a start, we need more structural information
about purified ERM proteins. For instance, our model indicates that the
complex formed by the separate amino- and carboxy-terminal domains of
radixin is an intramolecular complex in the intact molecule. This
proposition needs to be tested. One approach to this question would be a
combination of chemical crosslinking and protease digestion. Using point
crosslinkers followed by protease digestion, one might be able to establish
that residues close to one another in the inter-domain complex are close to
each other in the whole molecule. Other information could come from low
resolution techniques such as electron microscopy. For example, this kind
of data indicated a "balloon-on-a-string" shape for vinculin which is
consistent with the notions about the organization of that molecule (Milan,
1985). More refined structural information could come from X-ray
crystallography or NMR spectroscopy of radixin. If the general features of
the model are correct, then the results from these studies are likely to
depend on the source of radixin analyzed. One possibility is that the open
conformation of the molecule is not stable outside of the cell, in which case,
this conformation would be difficult to study. A possible source for large
quantities of radixin in the open form is the chicken erythrocyte. The bulk
of radixin in the chicken erythrocyte is localized to the marginal band
where it is resistant to extraction by non-ionic detergents (Birgbauer and
Solomon, 1989; Winckler, 1994). The confusing detergent extraction
properties of ERM proteins notwithstanding (see Discussion in Chapter
Six), perhaps radixin in the marginal band is in a stable, open state
193
participating in intermolecular associations. If the stability of the open
state is due to a covalent modification, then it is possible that after urea
extraction of radixin from the marginal band, an open form of the molecule
can be isolated for study.
Regulatory mechanisms- Of course, for full understanding of the
model, the mechanism regulating the conformational transition must be
understood. Drug interference experiments might suggest the involvement
of one or more signal transduction pathways. Also, identification of
radixin binding partners might lead to in vitro assays for ERM protein
function. For instance, one could determine if the engagement of another
protein ligand by radixin enabled F-actin binding by the full-length protein.
Results from structural and functional assays of radixin like those
described above might lead to more refined mutagenesis of the radixin
molecule. One might predict that point mutations in critical residues could
create a constituitively open conformation of the full-length molecule. This
mutant form of full-length radixin should have properties similar to those
of the isolated carboxy-terminal domain when expressed at high levels in
cultured cells.
194
APPENDIX ONE:
Isolation of a euploid embryonal carcinoma P19 cell line.
195
SUMMARY
As described in Chapter Two, we isolated a number of P19 cell lines
that exhibited a clumpy phenotype upon RA-induction. There was a small
amount of clumping detectable in the P19 parental stock. Therefore, we
considered the possibility that isolation of clumpy P19-transfectants was a
reflection of variation within the parental P19 stock. We conducted a
karyotype analysis on these cells to determine chromosome number. P19
cells had been reported to be euploid (McBurney and Rogers, 1982). We
found that the P19 stock in the lab was not uniformly euploid. Instead, they
exhibited a bimodal distribution with part of the population centered around
the euploid chromosome number of 40 and another part of the population
centered around 72. To re-isolate a euploid P19 line from the mixed
parental stock, we subjected the parental cells to limiting dilution cloning
and screened karyotypes of these clones. We isolated a euploid P19 cell line,
P19-A4, that exhibited model differentiation properties. We subsequently
used P19-A4 cells for transfection experiments. However, we continued to
isolate clumpy stable lines from this background.
196
MATERIALS AND METHODS
Karyotype analysis. To examine karyotypes, we used the procedure
described by Rudnicki and McBurney (1987). Briefly, 1 X 106 cells grown
overnight in a 60 mm dish were arrested in metaphase by treatment with
0.06 jig/ml colcemid (Sigma) for 3 hours. Then, we washed the cells in PBS
and dissociated them from the culture substratum with 1 ml trypsin-EDTA.
Next, we swelled the cells in a hypertonic buffer containing 0.56% KO for 8
min at room temperature. After a spin to collect the osmotically swelled
cells, we fixed the cells in 10 ml fresh Carnoy's fixative (3 methanol: 1
glacial acetic acid) by dropwise addition to pelleted cells. After a 10 min
incubation, we pelleted the fixed cells, resuspended them in 10 ml fresh
fixative, re-pelleted them, and resuspended them in 1 ml fixative. We
applied drops of this cell suspension to a clean microscope slide and dried
them under a lightbulb. Finally, we stained them for 5 min in Giemsa
stain (Sigma), washed them in water and after drying, we mounted a glass
coverslip over the slide. We examined metaphase chromosome spreads
prepared in this way under the 100x phase objective on a Zeiss Axioplan
microscope.
Cloning P19 cells. We cloned the P19 parental stock by limiting
dilution. Briefly, we made serial dilutions of a P19 stock cell suspension
calculated to yield 10, 1, and 0.1 cells in a 0.2 ml volume of MEMa + 7.5%
calf serum/ 2.5% fetal calf serum. We plated 0.2 ml of the cell suspensions
at the three densities indicated above in each well of three separate 96-well
microtiter dishes. After plating, we noted wells containing single cells and
then incubated the plates for 10 days in the tissue culture incubator under
standard conditions. After growth of single cells to small colonies, we
picked those colonies and expanded them using standard tissue culture
technique.
197
RESULTS AND DISCUSSION
P19 cells are an embryonal carcinoma line derived from teratomas
formed by grafting egg cylinders from C3H/He mice to testes in male
animals. The original P19 isolate was reported by McBurney and Rogers
(1982) to be a normal male karyotype 40:XY. This karyotype was stable in
culture during the course of a cloning and selection procedure. We (the
Solomon laboratory) received P19 cells in 1989 from the McBurney
laboratory where they were isolated in 1982. The history of the cells during
this period is unknown. Once in the Solomon lab, P19 cells were expanded
for several doublings, aliquoted, and frozen in liquid nitrogen for long term
preservation.
Results described in Chapter Two suggested the presence of variants
within the P19 parental stock. Nearly half of the stably transfected cell lines
derived from these cells displayed a clumpy culture appearance after RAinduction. The P19 parental stock also showed a small amount of
clumping. Therefore, we decided to re-clone the parental P19 stock in an
attempt to avoid this undesirable outcome in future experiments. Prior to
this cloning step, we performed a karyotype analysis to see if the parental
stock was still euploid as previously reported.
We performed this analysis on the original P19 cell stock received in
the lab which had been stored under liquid nitrogen for approximately four
years prior to this analysis. We counted chromosome number for 116
metaphases from these cells. The results are shown in Figure A1-1. There
is clearly karyotypic heterogeneity in this cell line. In fact, the population
presents a bimodal distribution with centers around 40 and 72
chromosomes. Aneuploidy is a common feature of most established cell
lines. However, as previously mentioned, the original P19 isolates were
reported to be relatively stable euploid cells.
We isolated a euploid clone from this mixed parental stock for use in
subsequent transfection experiments. Since the clumpy RA-induced
behavior was not reported for the original P19 isolates, we hoped that a recloned euploid P19 line would not give rise to this behavior in stable
transfectants. We cloned P19 cells by limiting dilution as described in
Materials and Methods. We expanded these clones and performed
karyotype analyses to screen for euploid lines. Out of 44 lines screened, 2
198
were euploid as judged by having >90% metaphases with 40 chromosomes.
A chromosome spread from one of these two euploid lines, P19-A4, is
shown in Figure A1-2 A. It is not clear why the frequency of recovering
euploid clones did not reflect the frequency of euploid cells in the original
parental population. Perhaps the cloning efficiency of aneuploid P19 cells
is higher than that of euploid P19 cells.
This euploid P19 cell line -P19-A4- was carried through the RA and
DMSO induction protocols as described in Chapter Two. This cell line
displayed model culture morphology with both of these protocols (Fig. A1-2
C and D). At the level of gross morphology, P19-A4 cultures were
indistinguishable from the parental line (compare Figs. A1-2D and 2-5F for
P19-A4 with Fig. 2-3F for the parental stock.
As mentioned in Chapter Two, we used P19-A4 as a background for
production of stably transfected lines. Unfortunately, about half of the cell
lines derived from P19-A4 exhibited a clumpy phenotype like those derived
from the parental stock. Therefore, the cloning procedure described did not
remedy that problem.
We examined the karyotypes from some of the transfected P19-A4 cell
lines. Out of 6 tau antisense transfectants, 4 had >90% euploid metaphases,
and out of 6 vector control transfectants, 5 had 90% euploid metaphases.
Therefore, it seems that the chromosome number of P19-A4 is reasonably
stable through a transfection and selection process. Finer resolution
analysis would be necessary to determine the degree of other types of
genomic instability, such as translocations or deletions, in P19-A4.
Cultured cell lines typically have unstable genomes. One notable
exception is mouse embryonic stem cells. In fact, their genomic stability
makes targeted gene disruption in mice possible. Euploid P19 cells might
also be attractive candidates for gene targeting experiments. Though these
transformed cells would not be useful for production of transgenic mice,
one could study the differentiation properties of these cells bearing specific
genetic ablations in vitro. Just such an approach has been taken to study
1-integrin function in embryonal carcinoma F9 cells (Stephens et al.,
1993).
199
Figure Al-1. Karyotype analysis of embryonal carcinoma P19 lab stock. A
karyotype analysis was performed on P19 cells as described in Materials
and Methods. Chromosome number was counted only in well-spread
metaphases. The number of metaphases counted with a particular
number of chromosomes were plotted in this graph.
200
P19 KARYOTYPE ANALYSIS 4/2/93
20
1918171615141312-
w 11CL
0.
E
10
10
987-
65
43 21-
# chromosomes
Figure AI-2. Characterization of P19-A4- a euploid embryonal carcinoma
P19 cell line. P19-A4 was isolated as described in the text. (A) shows a
euploid metaphase spread from this cell line. (B-D) phase contrast images
of uninduced (B); DMSO-induced (C); and RA-induced (D) cultures.
202
04
41j
w
4b
,
APPENDIX TWO:
Characterization of a cell-substratum adhesion deficient
embryonal carcinoma P19 cell line.
204
SUMMARY
During the course of studies presented in Chapter Two, we isolated a
P19 cell line with unique properties, TA3A. This line was derived from a
transfection with a tau antisense element. When induced to differentiate
with RA, this line lost the ability to adhere to solid substrata. Instead,
TA3A cells adhered to one another to form aggregates. This behavior did
not strictly depend on depletion of tau in these cells and was distinct from
the clumpy behavior described for some lines in Chapter Two. Here we
show that the TA3A phenotype is a RA-specific cell-autonomous response.
Although these cells cannot attach to solid substrata, and therefore cannot
extend neurites on these surfaces, the cells do retain the ability to extend
neurites within the aggregates they form. We conclude that TA3A is a
novel variant of P19 cells.
205
MATERIALS AND METHODS
RA induction in monolayer culture. Uninduced cells were plated in
a 100 mm bacteriological grade petri dish at a density of 1x10 5 cells/ml in 12
ml MEMa plus 7.5% bovine calf serum, 2.5% fetal bovine serum and
allowed to aggregate. After 4 h, aggregates (10-20 cells in size) were
harvested and replated on a 100 mm tissue culture grade dish in MEMx
plus 7.5% bovine calf serum, 2.5% fetal bovine serum and allowed to adhere
to the dish for 16 h. During this period, the cells spread out of the
aggregates so that cells were growing in small monolayer patches on the
tissue culture dish. After aggregates adhered to the dish, medium was
aspirated, cells were washed once with 1X PBS and induced to differentiate
with 0.6 pM RA in MEMx plus 2% fetal bovine serum. Cells were cultured
in this medium for 3 d, with a medium change after 2 d, and then
photographed. We noted that using this method of RA induction we did not
observe morphological evidence of neural differentiation by as large a
proportion of the cells as observed using the aggregate method described in
Chapter Two.
Labeling and imaging P19 cells in aggregates. For studies in which
P19 cells were tracked over time within aggregates, the cells were
trypsinized to form a single cell suspension, plated in 5 ml alpha-modified
MEM plus 7.5% bovine calf serum, 2.5% fetal bovine serum in 60 mm
bacteriological grade dishes, and 5-chloromethylfluorescein diacetate
(CMFDA) and 5-(and-6)-(((4-chloromethyl)benzoyl)amino) tetramethylrhodamine (CMTMR) (Molecular Probes) were added separately to one dish
each of TA3A and N2B to a final concentration of 25 tm. Cells were
incubated at 37'C for 10 min, then pelleted at 500 rpm for 5 min at 37'C.
Cells were resuspended in 5 ml fresh medium, re-pelleted, and
resuspended in 5 ml fresh medium again. Other cells were carried
through the same procedure without the addition of fluorescent dyes. No
significant loss of viability was noted after treatment of cells in this
manner. Both labeled and unlabeled cells were counted with a
hemacytometer, and then were mixed together at the following
labeled:unlabeled ratios: 1:10 (which was useful for examining large
numbers of neurite bearing cells within the same aggregate) and 1:100
206
(which allowed for comparison of individual neurites). The background of
unlabeled cells was varied in some experiments from all TA3A to all N2B.
The mixed cells were plated on bacteriological grade dishes so that the
normally adherent N2B cells would stay in the aggregates and incubated
for various times with medium changes every 2 d. At 1 d intervals,
aggregates were removed from the culture dishes, placed in a well of 96well plate (Costar), washed once with 1X PBS, and fixed with 3.7%
paraformaldehyde for 30 min at 37*C. After fixation, aggregates were
washed 3 times with 1X PBS, placed on glass microscope slides, covered
with a drop of mounting medium containing Gelvatol (Monsanto) and
25mg/ml 1,4-diazabicyclo [2.2.21-octane (Aldrich) to retard photobleaching,
and a glass coverslip was placed over the aggregates. Although this caused
deformation (flattening) of the aggregates, it did not significantly alter the
morphology of the processes as compared to fixed aggregates that were
examined in a drop of PBS without being affixed to a slide.
For immunostaining of aggregates, 4d RA-induced TA3A and N2B
cells were plated onto bacteriological grade dishes without mixing and
cultures were incubated for 3 d. After 3 d, aggregates were removed from
the dishes and fixed as described above. The fixed aggregates were then
treated with 0.5% NP-40 in PBS for 30 min at 37'C. Aggregates were
reacted with a 1:400 dilution of polyclonal 1-tubulin antiserum 429
(described in (Kim et al., 1987)in PBS for 3 h at 37*C, washed 3 times with
1X PBS and then reacted with a 1:40 dilution of Texas Red-conjugated goat
anti-rabbit second antibody for 1 h at 37'C and washed 3 times with IX PBS.
Some aggregates were treated with the second antibody alone. The
aggregates were mounted on microscope slides as described above.
Aggregates were imaged with a Bio Rad MRC 600 scanning laser
confocal microscope. Images of differentially labeled cells were collected
with a yellow high sensitivity filter to detect CMFDA labeled cells and with
a blue high sensitivity filter to detect CMTMR labeled cells and merged with
Bio Rad software two create pseudocolor micrographs. Immunostained
aggregates were imaged with a blue high sensitivity filter to detect Texas
Red second antibody. Hard copy prints of the video images were made with
a Sony UP-5000 Color Video Printer.
207
RESULTS
P19 cell line TA3A shows a RA-dependent aggregation. TA3A was
isolated from P19 cells transfected with the tau antisense vector as
described in Chapter Two. Figure A2-1 illustrates the distinct steps in the
standard protocol for differentiation of P19 cells to produce neural cultures.
TA3A cells are indistinguishable from control cells at the light microscope
level in the uninduced state (Fig. A2-1 A and B), in the aggregates formed
during RA-induction in bacteriological grade dishes (Fig. A2-1 C and D), or
as single cells released from these aggregates by trypsinization in
preparation for re-plating onto tissue culture dishes (Fig. A2-1 E and F).
However, striking differences become apparent upon plating cells after
treatment with RA. RA-induced control P19 cells attach to tissue culture
plastic within 4 hours, spread and extend processes within 24 hours, and
have fully elaborated the neural phenotype by 72 hours (Fig. A2-1H). In
contrast, when RA-induced TA3A cells are plated, they do not attach even
transiently to the tissue culture substratum. Instead, in as little as four
hours they have adhered to one another to form aggregates (Fig. A2-1G).
Another cell line that showed an identical pattern of tau antisense vector
integration, TAlA, also showed this behavior.
The TA3A phenotype was distinct from the clumpy phenotype of
transfected cell lines described in Chapter Two. As mentioned in Chapter
Two, nearly half of the transfected P19 cell lines isolated displayed a clumpy
culture morphology after exposure to RA. Although this phenomenon did
not depend on the type of construct used for transfection, we wondered
whether the TA3A phenotype was related to the clumpy behavior. Upon
extended culture (6 days), initially clumpy lines spread out (Fig. A2-2). In
contrast, TA3A aggregates do not spread out and most of these aggregates
are still completely non-adherent after this time. After ten days, more
TA3A aggregates were loosely adherent, but no neurites were evident (data
not shown).
The failure of the TA3A cells to adhere properly was strictly
dependent on prior exposure to RA. TA3A cells treated with DMSO adhere
normally to tissue culture plastic and form aligned syncytia, as do control
lines (Figs. A2-3A and B). Also, TA3A cells carried through the neural
208
differentiation protocol, but in the absence of RA, retain their ability to
adhere to tissue culture plastic (Fig. A2-3 C and D).
The adhesion defect was manifest whether the cells were initially
exposed to RA in suspension or in monolayer. P19 cells were also RAinduced in monolayer as described in Materials and Methods. TA3A cells
treated with RA in monolayer culture differed markedly from control cells:
after 3 d RA treatment, they formed aggregates that loosely adhered to the
tissue culture dish, whereas control cells remained in monolayer (Fig. A2-3
E-H).
We examined whether plating on substrata other than tissue culture
plastic would enable RA-induced TA3A cells to adhere properly. We plated
1x10 6 4 d RA-induced TA3A cells and control cells in 6-well dishes
containing wells coated with either collagen I (rat tail tendon), fibronectin
(human), laminin (mouse), or poly-L-lysine (Collaborative Biomedical
Products). TA3A cultures were clearly distinct from control cultures and
formed aggregates on each of these substrata, demonstrating that the
aggregation phenotype is not specific to tissue culture plastic (Fig. A2-4).
However, on collagen I and laminin, many of the aggregates did attach and
some neurites were evident.
Aggregation of RA-induced TA3A cells results from a cellautonomous defect in attachment to solid substrata. Formally, two models
could explain the aggregation of RA-induced TA3A cells. One possibility is
that RA-induced TA3A cells gain the ability to aggregate in such a way that
cell-cell interactions are exclusive of cell-substratum interactions.
Alternatively, TA3A cells might lose the ability to adhere to solid substrata
in response to RA in a cell-autonomous manner. Aggregate formation
would then be a default state resulting from cells that are able to adhere to
one another, but are not able to attach to the tissue culture substratum. If
the latter hypothesis is correct, TA3A cells must lose the ability to attach to
solid substrata some time between the uninduced state, when they adhere
normally, and 4 d of RA treatment. Although, as previously shown, there
were no apparent differences (as judged by phase microscopy) between the
TA3A and control cell aggregates during the 4 d RA-induction period, we
wished to know when the adherent, uninduced TA3A cells made the
transition to the non-adherent state. We removed aggregates from
209
Figure A2-1. RA-induced differentiation of P19 cell line TA3A. TA3A
(A,C,E, and G) and vector control cell line N2B (B,D,F,and H) were induced
to differentiate with RA as described in Materials and Methods and were
photographed at points along the induction process. (A and B) Uninduced
cells; (C and D) aggregated cells in bacteriological grade dishes, after 2 d
treatment with RA; (E and F) 4 d RA-treated cells plated for 10 min on
tissue-culture plastic after trypsinization of aggregates; (G and H) 4 d RAinduced cells plated for 3 d on tissue culture plastic. Bar, 100tm.
210
Figure A2-2. Extended culture of TAA. Tau antisense cell lines TA2B (A
and B) and TA3A (C and D) were exposed to RA for 4 d in aggregates,
trypsinized to form a single-cell suspension, and replated in tissue culture
plastic dishes. The medium, without RA, was replaced every 2 d and the
cultures were photographed 3 d (A and C) and 6 d (B and D) after being
plated on tissue culture plastic. Arrowheads indicate aggregates that are
adherent to the dish when the plate is rocked gently. Bar, 100 gm.
212
Days Plated on
Tissue Culture
Plastic
TA2B
TA3A
3-
6
U
Figure A2-3. Loss of cell-substratum adhesion in TASA cells is a specific
effect of exposure to RA. TA3A (A,C,E, and G) and vector control cell line
N2B (B,D,F, and H) were induced to differentiate into muscle cells with
DMSO (A and B), aggregated in bacteriological grade plastic without the
addition of RA (C and D), or induced to differentiate with RA in monolayer
culture (E and G) as described in Materials and Methods. Cells were plated
in tissue culture dishes and photographed 3 d after plating (A and D), just
prior to addition of RA (E and F), or 3d after addition of RA (G and H).
Arrows in A and B denote regions of aligned syncitia, a characteristic of
muscle differentiation, in both tau antisense cells and vector control cells.
Bar, 100 Rm.
214
Figure A2-4. Adherence properties of TASA cells on a variety of culture
substrata. TA3A (A-E) and vector control cell line N2B (F-J) were induced
to differentiate with RA as described in Materials and Methods. Equal
numbers of 4d RA-induced cells were plated into uncoated tissue culture
dishes (A and F) or dishes coated with collagen I (B and G), fibronectin (C
and H). laminin (D and I), or poly-L-lysine (E and J). Cells were
photographed 3 days after plating.
216
r~~u
bacteriological grade dishes after 1, 2, and 3 d of exposure to RA,
trypsinized them and replated the single cell suspension on tissue culture
grade dishes. After 1 d of RA exposure, TA3A cells were able to adhere to
the tissue culture dish as single cells, but most of the cells failed to spread
and remained rounded and phase bright (Fig. A2-5A). In contrast, control
cells attached to and spread over the culture surface, although after this
short exposure very few neurite bearing cells were observed (Fig. A2-5B).
Over a period from 1 to 3 d of RA exposure, TA3A cells progressively lost the
ability to attach to the tissue culture surface and formed aggregates while
control cells attached and more neurite-bearing cells appeared in cultures
over time (Fig. A2-5 C-F).
If the loss of cell-substratum adhesion does not depend on cell-cell
contact then it should be manifest at low cell densities where this contact is
rare. To test this, we plated RA-induced TA3A and control cells at densities
between 2.7 X 10 4 and 2.7 X 10 5 cells/cm 2 . TA3A cells did not attach and
spread on the tissue culture plastic at any density. Instead, at the lowest
cell densities, we observed single, rounded non-adherent cells (Fig. A2-6).
The number and size of aggregates increased with cell density. In
contrast, control cells were adherent to the culture substratum at all
densities demonstrating that at low density plating P19 cells are normally
capable of adhering to substrata (Fig. A2-6). Taken together, the results
presented in this section argue that the aggregation seen in RA-induced
TA3A cultures is the result of a cell-autonomous defect in cell-substratum
adhesion.
Expression of neuron-specific biochemical markers in RA-induced
TAA. We assayed the levels of several neuron specific proteins in TA3A
and control cultures. We prepared cells and harvested protein as described
in Chapter Two. As mentioned previously in Chapter Two, TA3A
expresses about five-fold less tau protein in response to RA as compared to
control lines (Fig. A2-7A).
TA3A was competent to express other biochemical markers of
neuronal differentiation. Western blot analysis revealed that MAP2 and
GAP-43, previously shown to be induced by RA in P19 cells (Dinsmore and
Solomon, 1991), were also induced by RA in TA3A (Fig. A2-7 B and C),
although at slightly reduced levels compared to the vector control cells.
218
This result may indicate that the levels of these proteins are regulated, in
part, by adherence to solid substrata.
Neurite extension in aggregates. Because attachment is a
prerequisite for neurite extension over a planar culture surface, and TA3A
cells did not adhere well to any of the substrata tested above, we could not
assay the ability of these cells to extend neurites over the culture dish.
However, we investigated whether TA3A cells were able to extend neurites
on the surface of, or within, the aggregates they formed after RA induction.
TA3A and vector control cells were exposed to RA, differentially labeled
with cell-permeant dyes, mixed together so that the two cell types could be
visualized within the same aggregate, and plated in bacteriological dishes
so that the normally adherent vector control cells would aggregate. Figure
A2-8A shows that the fluorescein-labeled TA3A cells, like the rhodaminelabeled vector control cells, exhibit highly polarized morphologies, with
long, branching processes extending many times the length of the cell
bodies. Processes were present on a large proportion of the labeled cells and
coursed throughout the entire aggregate. Both cell types displayed these
features beginning 2-3 d after exposure to RA regardless of the composition
of the background of unlabeled cells, which was varied from all TA3A cells
to all vector control cells. The processes were apparent in both cell types for
6d after labeling, after which time the fluorescent markers did not clearly
label individual cells. There were no conspicuous differences in either the
number of processes emanating from cell bodies or the length and
complexity of processes between the TA3A cells and vector control cells. In
addition to the morphological similarities of these processes to other
neurites, two lines of evidence indicate that the processes within the
aggregates are indeed neurites. First, the same type of processes are not
present in either TA3A or vector control cells that have not been treated
with RA (a small proportion of cells do have cytoplasmic extensions, but
these rarely exceed one cell body diameter in length; data not shown).
Second, immunostaining of aggregates with an anti-tubulin antibody
revealed that microtubules in both the TA3A cells and vector control cells
are arranged in brightly staining bundles similar to the staining patterns
seen in neurites extended over planar surfaces (Fig. A2-8 B and C).
219
Figure A2-5. Development of TA3A phenotype after exposure to retinoic
acid. Tau antisense cell line TA3A (A, C, and E) and vector control cell line
N2B (B, D, and F) were induced to differentiate with RA as described in
Materials and Methods. During the 4 day RA-induction, aggregates from
the bacteriological grade dishes were harvested at the times indicated,
trypsinized, and plated in tissue culture plastic dishes with no RA in the
medium. Cells were photographed 2 d after plating in tissue culture plastic
dishes. (A and B) 1 d RA-treatment; (C and D) 2 d RA-treatment; (E and F)
3 d RA-treatment. Bar, 100pm.
220
DAYS IN
RA
Tau Anti-Sense
Neo Alone
2El
3
I'
Figure A2-6. TASA phenotype is independent of culture density. TA3A (AE) and vector control cell line N2B (F-J) were induced to differentiate with
RA as described in Materials and Methods. After 4 days of RA-induction,
cells were plated onto tissue culture dishes at the following densities (cells/
cm 2 ): 2.7 X 10 4 (A and F); 5.3 X104 (B and G); 1.1X 10 5 (C and H); 2.0 X 10 5
(D and I) 2.7 X 10 5 (E and J). Cells were photographed 3 days after plating.
222
D
-
Figure A2-7. Expression of neuron-specific markers by retinoic acidinduced TA3A cells. Proteins were harvested from 4 d RA-induced TA3A
(lane 1) and vector control line N2B (lane 2) 3 d after plating induced cells
onto tissue culture plastic as described in Materials and Methods. Equal
amounts of each protein sample were separated on 4-10 %linear gradient
polyacrylamide gels, and then electrophoretically transferred to a 0.2 Rm
nitrocellulose membrane as described in Materials and Methods. The blots
were immunostained with the primary antibodies described below, reacted
with appropriate second antibody and the second antibody was detected
either colorimetrically (A) or isotopically (B,C) as described in Materials
and Methods. (A) Tau expression. 200 gg of heat-stable protein was loaded
on gel. Blot was immunostained with a 1:100 dilution of mAb 5E2 specific
for tau This antibody recognizes two predominant species in the molecular
weight range for tau isoforms. The entire blot is shown and the positions of
66 and 45 kd molecular weight standards are indicated. (B) MAP2
expression. 100 g of whole-cell extract was loaded on gel. Blot was
immunostained with a 1:1000 dilution of mAb AP-14 specific for MAP2.
Only the relevant sections of the blot are shown. (C) GAP-43 expression.
100 g of whole-cell extract was loaded on gel. Blot was immunostained
with a 1:2000 dilution of mAb 9-1E12 specific for GAP-43. Only the relevant
sections of the blot are shown.
224
(A)
066 kd
'445 kd
(B) ..
(C)
a
12
Figure A2-8. Neurite extension in aggregates. (A) TA3A cells labeled with
CMFDA (green) and N2B cells labeled with CMTMR (red) were mixed,
cultured, and imaged as described in Materials and Methods. This
confocal micrograph shows an example of the processes extended by TA3A
and N2B cells. Aggregate shown is composed of 1 labeled: 100 unlabeled
cells and all unlabeled cells are TA3A. (B and C) Aggregates composed of
all TA3A cells (B) or all N2B cells (C) were immunostained with an
antibody specific for 13-tubulin and imaged by confocal microscopy as
described in Materials and Methods. Microtubules stain as dense bundles
that course through the aggregates. Bars: (B and C) 50 gm.
226
jr- .-
DISCUSSION
Here we have characterized a P19 cell line variant that, when
exposed to RA produces cultures that are no longer able to adhere to solid
substrata. RA-induced TA3A cells exist as aggregates of cells and are
capable of extending neurites within these aggregates. The loss of cellsubstratum adhesion is cell-autonomous and is dependent on exposure to
RA. The molecular basis for this phenotype is unknown, but it is not
strictly dependent on reduced levels of tau protein.
Cell adhesion phenotypes of TASA. The data presented here indicate
that undifferentiated TA3A cells are adherent, but they lose their ability to
adhere to solid substrata after even a 24 hour exposure to RA. TA3A cells
are deficient in both the establishment of cell-substratum adhesion (a
monodispersed suspension of RA-induced TA3A cells fail to attach to all
substrata tested); and the maintenance of that adhesion (TA3A cells in
monolayer lose normal adhesion when treated with RA). These
observations suggest that the undifferentiated precursors, and the DMSOinduced muscle cells derived from them, rely upon a different mechanism
for cell-substratum adhesion than do the neural cells. In animals, both
cell-cell and cell-substratum adhesion events are important for various
aspects of neuronal differentiation (for review see Hynes and Lander, 1992).
Two lines of evidence demonstrate that the failure to adhere to solid
substrata is a cell-autonomous phenomenon and not a direct consequence
of cell-cell interactions among different populations of cells within the
TA3A cultures. First, a time course of RA-induction reveals single,
unaggregated TA3A cells that are apparently impaired in cell substratum
adhesion (rounded appearance on the culture dish) after a short exposure
to RA. Second, when plated at low densities, where cell-cell contact is
infrequent, TA3A cells still fail to adhere to the solid substrata unlike
controls. These data also argue strongly against the existence of a
subpopulation of cells in TA3A cultures that are able to sequester otherwise
adherent TA3A cells into aggregates.
All TA3A cells exposed to RA express the defect in cell-substratum
adhesion. RA-induced P19 cultures are known to contain several nonneuronal cell types including glial and fibroblast-like cells (Jones228
Villeneuve et al., 1982). The results suggest that both of these cell types rely
on a similar mechanism for cell-substratum adhesion.
The failure to attach means these cells cannot extend neurites over
solid substrata. However, the cells do retain the ability to extend neurites
within the aggregates they form. This suggests that both the ability to
adhere to other cells and the mechanisms for neurite extension are not
significantly impaired in TA3A.
It was somewhat surprising to see the degree and rapidity of neurite
extension in aggregates of both TA3A and control cells. Unambiguous
neurites were detectable in aggregates by immunostaining with antitubulin antibodies by 3d of RA-induction. Several reports indicate that a
number of neuron-specific markers are expressed at this time. However,
McBurney et al. (1988) did not detect robust neuronal ultrastructure by
electron microscopy of sectioned aggregates until 9 days post RA-treatment.
Perhaps the discrepancy in these results lies in the methods for visualizing
the neurites. Confocal immunofluorescence microscopy used here may be
sensitive to nascent neuronal processes whereas electron microscopy only
detects mature processes with densely packed microtubule and
neurofilament arrays.
Speculation on the nature of the defect in cell-substratumn adhesion in
TA3A cells. Several lines of evidence indicate that, except for the
catastrophic loss of cell-substratum adhesion, RA-induced TA3A cells are
otherwise fairly normal. They retain the ability to adhere to one another
and form densely packed aggregates. The cells in these aggregates are
capable of extending neurites and expressing neuron-specific markers like
control cells. What is, then, the nature of the defect in these cells that leads
to the RA-dependent loss of cell-substratum adhesion?
A host of cell-intrinsic and extracellular factors are known to
influence neuronal cell-substratum adhesion. Briefly, those may be
characterized as cytoskeletal, transmembrane, and extracellular. Proper
cell-substratum adhesion depends on the integrated function of many
molecules that play structural and regulatory roles. Several examples
follow. Both microfilaments and microtubules participate in the
establishment and maintenance of cell-substratum adhesion (LeTourneau,
1975). Drug interference studies indicate that these two cytoskeletal
229
elements are also necessary for neurite extension. Therefore, at some level,
these two systems are functioning in TA3A cells. Transmembrane cell
surface molecules such as the integrins are known to mediate cell
attachment to substrata (reviewed in Hynes, 1992). Expression and
function of these molecules is also known to be modulated during neuronal
differentiation (Neugebauer and Reichardt, 1991). Interestingly, targeted
disruption of 131 integrin in F9 embryonal carcinoma cells, which can be
induced to differentiate into parietal endoderm, results in constitutitively
impaired cell substratum adhesion, but does not affect differentiationspecific gene expression in these cells (Stephens et al., 1993). Preliminary
evidence suggests that B1 integrin expression levels in RA-differentiated
TA3A cells are normal. Finally, extracellular matrix molecules like
laminin play a role in cell adhesion. However, it does not seem likely that
the defect in TA3A cell adhesion lies in an extracellular factor for two
reasons. TA3A cells did not adhere to a variety of extracellular matrix
substrates. Additionally, when differentially labeled and mixed with
control cells, most TA3A cells still form aggregates on tissue culture dishes
(data not shown). These results argue that TA3A cells are not deficient in
production of an extracellular component necessary for adhesion.
It will be a tall order to determine the molecular nature of the TA3A
phenotype. Our work indicates that tau and GAP-43 are significantly
reduced in TA3A cells. From the work presented in Chapter Two, it is
clear that the TA3A phenotype does not solely result from decreased levels
of tau protein. However, we cannot rule out the possibility that this
phenotype results from a combination of decreased tau and another lesion.
To learn more about the molecules involved in the TA3A phenotype, one
could use probes to molecules implicated in cell-substratum adhesion in
other systems to survey the TA3A cells. Molecules in RA-induced TA3A
cells whose expression is decreased compared to RA-induced control cells
would be interesting candidates for further study. These candidates could
then be transfected into TA3A cells tested for their ability to rescue cellsubstratum adhesion after RA-induction. This general approach was used
to identify vinculin as a molecule that rescued cell-substratum adhesion in
an embryonal carcinoma F9 mutant cell line (Samuels et al., 1993).
However, this sort of experiment does not definitively identify to original
molecular defect in the mutant cell line. Overexpression of a number of
230
components could promote cell-substratum adhesion. If the mutation that
caused the TA3A phenotype resulted from the chromosomal integration of
the tau antisense vector, in principle, one could clone and sequence
flanking DNA sequences to see if a known gene had been disrupted. This
approach would be daunting too because of the size and complexity of
mammalian genes, the possibility of multiple vector integration events, and
the instability of cultured cell line genomes.
231
APPENDIX THREE:
Expression and localization of HA-radixin constructs in embryonal
carcinoma P19 cells.
232
SUMMARY
ERM proteins localize to cortical structures in neuronal growth
cones. To determine the localization properties of full-length and truncated
HA-radixin constructs in neurons, we have stably expressed these
constructs in P19 cells. The results indicate that in uninduced P19 cell
lines, the localization patterns of both full-length radixin and its amino-and
carboxy-terminal domains resemble those described in Chapter Three for
these constructs in stably transfected NIH-3T3 cells. Preliminary evidence
suggests that radixin could compete with ezrin for cortical localization in
P19 cells as it does with moesin in NIH-3T3 cells. We then differentiated
P19 cells into neurons with retinoic acid treatment. RA-induced cell lines
expressing full-length and truncated HA-radixin constructs showed
evidence of appropriate neural differentiation including the presence of
neurite-bearing cells. However, definitive growth cone localization for the
HA-radixin constructs was not obtained.
233
INTRODUCTION
The exquisite cell-cell connectivity that underlies the functioning of
the nervous system is established, in part, due to an extraordinary cellular
structure- the growth cone. Growth cones project outward from the
neuronal cell body and often travel over enormous distances relative to their
size. They also accomplish this migration with remarkable accuracy targeting appropriate tissues and, in some cases, perhaps specific cells.
Growth cones utilize diffusible, cell surface bound, and extracellular
matrix signals to guide their journey. A variety of evidence indicates that
growth cone filopodia actively explore the extracellular environment for
guidance cues during development (Bastiani and Goodman, 1984; Sretavan
and Reichardt, 1993). The provocative localization of ERM proteins in the
distal tips of growth cone filopodia raises the question of what role these
molecules play in growth cone function.
Goslin et al. (1989) found that ERM proteins localized to growth cone
filopodia in hippocampal neurons actively extending processes over a
culture substratum. The mAb13H9 staining in these growth cones
partially overlapped the distribution of F-actin and its presence in the
growth cone depended on intact microtubules. Recent work in the Solomon
laboratory suggests the distribution of ERM proteins in growth cones
suggests that ERM proteins are sensitive to guidance cues (C. Gonzalez
Agosti unpublished results). For instance, ERM protein immunoreactivity
rapidly decreases when chicken sypathetic neurons are induced to collapse
by nerve growth factor withdrawal. Also, they rapidly redistribute to the
side of the growth cone that will become the new direction of migration
when growth cones are exposed to a galvanotactic field.
ERM proteins are also found in growth cones present in RA-induced
cultures of P19 cells (Birgbauer et al., 1991). Prior to RA-induction, ERM
proteins are detected in cortical cytoskeletal structures such as filopodia
and microvilli. RA-induces upregulation of ERM proteins so that they are
expressed two-to four-fold higher in the RA-induced cells. For the studies
above in hippocampal neurons and P19 cells, the identity of the ERM
proteins involved was not known, since pan-ERM immunological reagents
were used. Winckler et al. (1994) developed specific probes for each of the
ERM proteins. By Western blot analysis, they found that all three ERM
234
proteins were expressed in both uninduced and RA-induced P19 cells (C.
Gonzalez Agosti and B. Winckler, unpublished results). In contrast, only
ezrin was detected in cortical structures in uninduced cells and in growth
cone filopodia in RA-induced cells (C. Gonzalez Agosti and B. Winckler,
unpublished results). This finding is an interesting complement to the
results from 3T3 cells where all three ERM proteins are expressed, but only
moesin and radixin are detectable in cortical structures.
To explore the role of ERM proteins in growth cone function, we have
expressed full-length and truncated forms of radixin in P19 cells. The
localization properties of these proteins in P19 cells are similar to their
localization properties in NIH-3T3 cells. Preliminary evidence suggests
that exogenous radixin competes with ezrin for cortical localization. We
did not find any obvious effects on RA-induced neuronal differentiation
associated with the heterologous expression of radixin or its fragments.
235
MATERIALS AND METHODS
DNA constructs and transfection. We used the pCXN2 vector for
expression of HA-radixin constructs in P19 cells. These were constructed
with cloning techniques similar to those described in more detail in
Chapters Two and Three. Briefly, we prepared the HA-radixin inserts by
digestion of pMFG-HAN-RAD; -HAC-RAD; -HAN-RADN; -HAC-RADN;
-HAN-RADC; and -HAC-RADC (all described Chapter Three) with NcoI
and BamHI and then blunting the ends with Klenow enzyme. This
produced blunt-ended fragments carrying the entire coding sequence for
the HA-radixin polypeptides. Using T4 DNA ligase, we inserted these
fragments into the pCXN2 vector (described in Chapter Two) that was first
digested with XhoI and treated with Kienow enzyme to blunt the ends. The
resultant plasmids were named pCXN2-HAN-RAD; -HAC-RAD; -HANRADN; -HAC-RADN; -HAN-RADC; and -HAC-RADC.
After purification of plasmid DNA on Qiagen columns, we
introduced these plasmids into P19-A4 cells (described in Appendix One)
using the calcium phosphate method of Graham and van der Eb (1973) and
selected for stable transfectants in the presence of 600 pg/ml G-418.
Immunofluorescence detection of HA-radixin constructs and ezrin
in P19 cells. We used the same immunostaining protocol described in
Chapter Three for NIH-3T3 cells. Ezrin was detected with antiserum #465
affinity purified as described in Winckler et al. (1995). We prepared RAinduced P19 cells for immunofluorescence in the following manner: P19
cells were RA-induced as described in Chapter Two. After 4 days of RAinduction, aggregates were trypsinized and dispersed to form a single cell
suspension. These cells were plated at low density onto 12-mm round
coverslips previously coated with 1Og laminin (Collaborative Research).
Cells were allowed to adhere, spread, and extend neurites for 24 hours
before fixation and processing for immunofluorescence as described for
uninduced cells above.
236
RESULTS
Expression of HA-radixin constructs in P19 cells. We expanded G418 resistant colonies and screened them for expression of HA-radixin
constructs. Many of these cell lines showed detectable expression of the
appropriate HA-radixin protein: HAN-RAD (4/4); HAC-RAD (8/10); HANRADN (10/10); HAC-RADN (3/10); HAN-RADC (8/9); HAC-RADC (1/11).
Examples of lines expressing the highest levels of HA-tagged proteins are
shown in Figure A3-1. In general, the highest expression levels for HAradixin constructs in P19 cells were 2-10-fold higher than highest
expression levels of the same HA-radixin constructs in NIH-3T3 cells (data
not shown). Some or all of this difference might be attributable to the
different expression vectors used in these two cell types.
Another difference between P19 and NIH-3T3 cell lines stably
expressing HA-radixin constructs is that unlike the case in NIH-3T3 cells,
the amino- and carboxy-terminal fragments of radixin are expressed at
comparable levels to the full length protein. Curiously though, while this is
the case for the amino-terminal HA-tagged HAN-RADN and HAN-RADC
constructs, the carboxy-terminal HA-tagged HAC-RADN and HAC-RADC
P19 transfectants screened expressed much lower levels of these constructs
compared to the full-length molecules. We do not know the basis for these
differences in the behavior in the truncated HA-radixin constructs. Figure
A3-1 also shows HA-radixin expression after RA-induction. The
expression levels are comparable for each HA-radixin construct before and
after RA-induction.
Co-localization of HA-radixin polypeptides and ezrin in unmduced
P19 cell lines. All of the lines expressing full length and truncated HAradixin constructs showed normal P19 cell morphology. We stained these
cells with mAb 12CA5 to reveal the localization of HA-radixin proteins. As
mentioned above, ezrin is the predominant ERM protein detected in cortical
structures in P19 cells. So, we also stained the cells with antibody #465 to
detect ezrin. As in NIH-3T3 cells, HAC-RAD localizes to cortical
cytoskeletal structures including filopodia, lammellipodia, and microvilli.
Figure A3-2 B shows some of these features. Also, P19 cells often have
prominent membrane blebs on their dorsal surfaces. HAC-RAD is also
237
Figure A3-1. Expression of HA-radixin constructs in uninduced and RAinduced P19 cell lines. P19 cell lines expressing HA-radixin constructs
were isolated as described in the text. Proteins from these lines in the
uninduced state (UN) or after RA-induction (RA) were harvested and
subjected to Western blot analysis with mAb 12CA5 as described in
Materials and Methods. 30 gg protein was analyzed for each sample.
Samples are identified beneath the blot and positions of molecular weight
markers (in kilodaltons) and full-length (FL), amino-terminal (N), and
carboxy-terminal (C) polypeptides are indicated to the left of the blot.
238
221.
115.
vomw
67.
C
N
.o
-46
3. II
Figure A3-2. Co-localization of HA-radixin proteins and ezrin in
uninduced P19 cell lines. Uninduced P19 cell lines expressing HACRAD
(A and B), HANRAD (C and D), HANRADN (E and F), and HANRADC (G
and H) were subjected to double-label immunofluorescence analysis with
anti-ezrin antibody #465 (A, C, E, and G) and mAb 12CA5 (B, D, F, and H)
as described in Materials and Methods. In A and B, arrowheads indicate
ezrin staining cortical structures where mAb 12CA5 staining is excluded
and, conversely, straight arrows indicate mAb 12CA5 staining cortical
structures where ezrin staining is excluded. In B, curved arrow points to a
brightly mAb 12CA5 staining membrane bleb. In G and H curved arrow
points to a membrane bleb positively staining with both mAb 12CA5 and
anti-ezrin antibody #465. In H, straight arrow points to mAb 12CA5
staining that co-aligns with a stress fiber in the ventral cytoplasm.
240
found in this type of cortical cytoskeletal structure (see curved arrow in Fig.
A3-2 B). Figure A3-2 A shows ezrin localization in the same HAC-RAD
expressing cells. Although there is clearly still some ezrin
immunoreactivity in cortical structures in these cells, it does seem
somewhat diminished compared to lines expressing other HA-radixin
constructs (see Fig. A3-2 C and E). Also, there are examples of mutually
exclusive staining patterns (see markers in Fig. A3-2 A and B). In this
micrograph it is difficult to determine if the mutually exclusive staining
patterns occurred within a single cell or between different cells due to the
tendency for P19 cells to grow as clusters. However, we did find other
examples in which single isolated cells showed either HAC-RAD or ezrin
staining in cortical structures (data not shown).
Both HAN-RAD and the amino-terminal HA-radixin constructs
behaved in stably transfected P19 lines as they did in NIH-3T3 cell lines.
HAN-RAD did not localize to cortical structures and remained diffuse
throughout the cytoplasm (Fig. A3-2D). In HAN-RAD expressing cells,
ezrin robustly localized to cortical structures (Fig. A3-2 C). This type of
staining pattern was typical for ezrin in P19 cells. The amino-terminal
HA-radixin constructs did not localize to cortical structures either. Like
HAN-RAD, they were diffuse in the cytoplasm. Figure A3-2F shows the
localization of HAN-RADN. Ezrin staining in these cells appears normal
(Fig. A3-2 E). We obtained similar results for HAC-RADN (data not
shown).
The carboxy-terminal HA-radixin polypeptides stably expressed in
P19 cells localized in a manner similar to their localization in NIH-3T3 cell
lines. HAN-RADC expressing cells are shown in Figure A3-2J. This
radixin fragment localizes to cortical structures including the prominent
membrane blebs. Also, like the findings in NIH-3T3 cells, this carboxyterminal portion of radixin co-aligns with stress fibers in the ventral
cytoplasm. This is not a characteristic feature of ERM protein localization
in P19 cells. Ezrin staining in P19 cells expressing the carboxy-terminal
domain was not typical of normal ezrin staining in P19 cells. Overall, the
staining was less intense or not detectable in the same cortical structures
where the HAN-RADC was detected (Fig. A3-2 G). However, there was still
ezrin immunoreactivity in some cortical structures. Particularly,
membrane blebs showed co-localization of the carboxy-terminal fragment
242
and ezrin (curved arrows in Fig. A3-2 G and H). We obtained similar
results for HAC-RADC.
RA-induction of P19 cell lines expressing full length and truncated
HA-radiin constructs. We assessed the effects of full-length and truncated
HA-radixin constructs on neuronal differentiation in P19 cell lines. We
differentiated P19 cell lines expressing these constructs with RA as
described in Chapter Two. All of the RA-induced cell lines gave rise to
neurite bearing cells. The number of neurite bearing cells present in
cultures of P19 cell lines expressing full-length and truncated HA-radixin
constructs were similar to untransfected cultures (data not shown). We
note that some of the transfected lines showed the same clumpy behavior
that was described in Chapter Two.
We attempted to detect localization of HA-radixin constructs in
growth cones emanating from these cells. However, we were unsuccessful
at preparing isolated growth cones for immunofluorescence. For definitive
localization studies, one must stain growth cones that are in isolation on
the culture substratum. We could not preserve the small, isolated growth
cones that were adherent and spread on laminin-coated coverslips
throughout the fixation and staining protocols. We tried several measures
to preserve these delicate structures, including layering the fixative over
the cells in a sucrose solution to avoid shear forces associated with
aspiration of the culture medium. More effort will be required to stabilize
these structures for immunoflourescence studies.
243
DISCUSSION
We have examined expression of HA-radixin constructs in stably
transfected P19 cell lines. The results show that the localization patterns of
full length and truncated HA-radixin proteins in P19 cell lines are similar
those described for NIH-3T3 cell lines in Chapter Two. This provides
confirmation that the localization patterns are a property of the radixin
molecules and not dependent on the particular cellular context in which
they are expressed.
We also found that expression of either full-length or truncated HAradixin constructs did not grossly interfere with RA-induced neuronal
differentiation in these cells. The presence of neurite-bearing cells in the
cultures suggests that growth cone function is unimpaired in these cell
lines. Unfortunately, due to technical difficulties, we were not able to
examine the localization properties of HA-radixin proteins in growth cones.
One of the most interesting findings presented here is that
heterologous expression of radixin in P19 cells is correlated with reduced
staining of ezrin from cortical structures. This result is similar to those
described in Chapter Three in which expression of radixin is correlated
with decreased moesin staining in cortical structures in NIH-3T3 cells.
Thus, the displacement phenomenon may extend to other ERM proteins
and other cell types. However, some caution is warranted with these
results. Ezrin immunoreactivity is still detectable in cortical structures in
P19 cell lines expressing either HAC-RAD or the carboxy-terminal
fragments. Although, ezrin staining in these cell lines is clearly
diminished compared to lines expressing HAN-RAD and the aminoterminal fragments. The moesin staining in HA-radixin expressing NIH3T3 cells seemed to be much more uniformly diminished. There are
several possibilities for the apparent differences in the effectiveness of the
displacement. We did not rigorously clone the P19 cell lines. Perhaps the
lines examined are a mixture of cells expressing HA-radixin to varying
degrees including some cells that do not express HA-radixin at all.
Another more interesting possibility is that the relative pools of HA-radixin,
endogenous ERM proteins, and the element necessary for cortical
localization could differ between cellular contexts. For example, there
could be more of the crucial localization element available in P19 cells as
244
compared to NIH-3T3 cells. Alternatively, in P19 cells the ratio of HAradixin to ezrin might be lower than the ratio of HA-radixin to moesin in
NIH-3T3 cells. Furthermore, since endogenous ERM proteins are
upregulated in P19 cells by RA treatment, the ratio of HA-radixin to ezrin
could be different in P19 cells between the uninduced and RA-induced
states. This could also be reflected in the cortical localization patterns of
HA-radixin and ezrin in RA-induced P19 cells as compared to uninduced
cells. In fact, during our attempts to detect HA-radixin in growth cones, we
noticed that RA-induced HAC-RAD and carboxy-terminal fragment
expressing cells did not show the crisp cortical localization found in the
uninduced cells. One possibility for this result is that higher levels of
endogenous ERM protein in the RA-induced cells out-competes HA-radixin
for cortical localization.
245
LITERATURE CITED
Abercrombie, M., J. E. M. Heaysman, S. M. Pegrun. 1970. The
locomotion of fibroblasts in culture. I. Movements of the leading edge.
Exp. Cell Res. 59:393-398.
Abo, A., A. Boyhan, I. West, A. J. Thrasher, and A. W. Segal. 1992.
Reconstitution of neutrophil NADPH oxidase activity in the cell-free
system by four components: p67-phox, p47-phox, p2lracl, and
cytochrome b-245. J. Biol. Chem. 267: 16767-16770.
Albanesi, J. P., H. Fujisaki, J. A. Hammer, E. D. Korn, R. Jones, and T. D.
Pollard. Mononeric Acanthamoeba myosins support movement in
vitro. J. Biol. Chem. 260:8649-8652.
Albert Wang, C.-L., J. M. Chalovich, P. Graceffa, R. C. Lu, K. Mabuchi,
and W. F. Stafford. 1991. A long helix from the central region of
smooth muscle caldesmon. J. Biol. Chem. 266:13958-13963.
Albrecht-Buehler, G., and R. M. Lancaster. 1976. A quantitative
description of the extension and retraction of surface protrusions in
spreading 3T3 mouse fibroblasts. J. Cell Biol. 71:370-382.
Algrain, M., 0. Turunen, A. Vaheri, D. Louvard, and M. Arpin. 1993.
Ezrin contains cytoskeleton and membrane binding domains
accounting for its proposed role as a membrane-cytoskeletal linker.
J. Cell Biol. 120:129-139.
Anderson, R. A. and R. E. Lovrien. 1984. Glycophorin is linked by Band 4.1
protein to the human erythrocyte membrane skeleton. Nature
(Lond.). 307:655-658.
Andreoli, C., M. Martin, R. Le Borgne, H. Reggio, and P. Mangeat. 1994.
Ezrin has properties to self-associate at the plasma membrane. J.
Cell Sci. 107:2509-2521.
Archer, J. E., L. R. Vega, and F. Solomon. 1995. Rbl2p, a yest protein that
binds to 1-tubulin and participates in microtubule function in vivo.
Cell 82: 425-434.
Bastiani, M. J., and C. S. Goodman. Neuronal growth cones: specific
nteractions mediated by filopodial insertion and induction of coated
vesicles. Proc. Natl. Acad. Sci. USA 81:1849-1853.
Bentley, D., and A. Toroian-Raymond. 1986. Disoriented pathfinding by
pioneer neurone growth cones deprived of filopodia by cytochalasin
treatment. Nature 323: 712-715.
246
Ben-Ze'ev, A., A. Duerr, F. Solomon, and S. Penman. 1979. The outer
boundary of the cytoskeleton: a lamina derived from plasma
membrane proteins. Cell 17:859-865.
Binder, L. I., A. Frankfurter, and L. I. Rebhun. 1985. The distribution of t
au in the mammalian central nervous system. J. Cell Biol. 101: 13711378.
Birgbauer, E. and F. Solomon. 1989. A marginal band-associated protein
has properties of both microtubule- and microfilament- associated
proteins. J. Cell Biol. 109:1609-1620.
Birgbauer, E., J. H. Dinsmore, B. Winckler, A. D. Lander, and F. Solomon.
1991. Association of ezrin isoforms with the neuronal cytoskeleton.
J. Neurosci. Res. 30: 232-241.
Birgbauer, E. 1991. Cytoskeletal interactions of ezrin in differentiated cells.
PhD thesis, MIT.
Bond, J. F., J. L. Fridovich-Keil, L. Pillus, R. C. Mulligan, and F. Solomon.
1986. A chicken-yeast chimeric B-tubulin protein is incorporated into
mouse mictotubules in vivo. Cell 44: 461-468.
Brady, S. T., K. K. Pfister, and G. S. Bloom. 1990. A monoclonal antibody
against kinesin inhibits both anterograde and retrograde fast axonal
transport in squid axoplasm. Proc. Natl. Acad. Sci. USA 87:10611065.
Bretscher, A. 1983. Purification of an 80,000-dalton protein that is a
component of the isolated microvillus cytoskeleton, and its
localization in non-muscle cells. J. Cell Biol. 97:425-432.
Bretscher, A. 1986. Purification of the intestinal microvillus cytoskeletal
proteins villin, fimbrin and ezrin. Methods Enzymol. 134:24-37.
Bretscher, A. 1989. Rapid phosphorylation and reorganization of ezrin and
spectrin accompany morphological changes induced in A-431 cells by
epidermal growth factor. J. Cell Biol. 108:921-930.
Burgin, K. E., B. Ludin, J. Ferralli, and A. Matus. 1994. Bundling of
microtubules in transfected cells does not involve an autonomous
dimerization site on the MAP2 molecule. Mol. Biol. Cell 5:511-517.
Caceres, A., and K. S. Kosik. 1990. Inhibition of neurite polarity by tau
antisense oligonucleotides in primary cerebellar neurons. Nature
343:461-463.
247
Caceres, A., S. Potrebic, and K. S. Kosik. 1991. The effect of tau antisense
oligonucleotides on neurite formation of cultured cerebellar
macroneurons. J Neurosci 11: 1515-23.
Caceres, A. J. Mautino, and K. S. Kosik. 1992. Suppression of MAP2 in
cultured cerebellar macroneurons inhibits minor neurite formation.
Neuron 9:607-618.
Cao, L. G., G. G. Babcock, P. A. Rubenstein and Y.-L Wang. 1992. Effects
of profilin and profilactin on actin structure and function in living
cells. J. Cell Biol. 117: 1023-1031.
Cao, L. G., D. J. Fishkind, and Y.-L. Wang. 1993. Localization and
dynamics of non-filamentous actin in cultured cells. J. Cell Biol.
123:173-181.
Carlier, M.-F., and D. Pantaloni. 1994. Actin assembly in response to
extracellular signals: role of capping proteins, thymosin B4 and
profilin. Sem. in Cell Biol. 5:183-191.
Carr, C. M., and P. S. Kim. 1993. A spring-loaded mechanism for the
conformational change of influenza hemagglutinin. Cell 17:823-832.
Cassemeris, L., McNeill, H., and S. H. Zigmond. 1990. Chemoattractant
stimulated polymorphonuclear leukocytes contain two populations of
actin filaments that differ in their spatial distributions and relative
stabilities. J. Cell Biol. 110:1067-1075.
Cassemeris, L. D. Safer, V. T. Nachmias, and S. Zigmond. 1992.
Thymosin B4 sequesters the majority of G-actin in resting human
polymorphonuclear leukocytes. J. Cell Biol. 119:1261-1270.
Chapin, S. J., J. C. Bulinski, and G. G. Gundersen. 1991. Microtubule
bundling in cells. Nature 349:24.
Church, G. M., and W. Gilbert. 1984. Genomic Sequencing. Proc. Natl.
Acad. Sci. USA 81: 1991-1995.
Cleveland, D. W., S. Hwo, and M. W. Kirschner. 1977. Physical and
Chemical Properties of Purified Tau Factor and the Role of Tau in
Microtubule Assembly. J. Mol. Biol. 116: 227-247.
Conboy, J., Y. K. Kan, S. B. Shohet, and N. Mohandas. 1986. Molecular
cloning of protein 4.1, a major structural element of the human
erythrocyte membrane skeleton. Proc. Natl. Acad. Sci. USA.
83:9512-9516.
Conboy, J., J. Y. Chan, J. A. Chasis, Y. W. Kan, N. Mohandas. 1991.
Tissue- and development-specific alternative RNA splicing regulates
248
expression of multiple isoforms of erythroid membrane protein 4.1.
J. Biol. Chem. 266:8273-8280.
Condeelis, J. 1993. Life at the leading edge: the formation of cell
protrusions. Ann. Rev Cell Biol. 9:411-444.
Cooley, L., E. Verheyen, and K. Ayers. 1992. chickadee encodes a profilin
required for intracellular cytoplasm transport during Drosophila
oogenesis. Cell 69: 173-184.
Crosbie, R. H., J. M. Chalovich, and E. Reisler. 1994. Flexation of
caldesmon: effect of conformation on the properties of caldesmon.
Manuscript submitted.
Cullen, B. R. 1987. Use of eukaryotic expression technology in the
functional analysis of cloned genes. Meth. Enzymol. 152: 684-704.
Damke, H., M. Gossen, S. Freundlieb, H. Bujard, and S. L. Schmid. 1995.
Tightly regulated and inducible expression of a dominant interfering
dynamin mutant in stably transformed HeLa cells. Methods.
Enzymol. In press.
Davenport, R. W., P. Dou, V. Rehder, and S. B. Kater. 1993. A sensory role
for growth cone filopodia. Nature 361: 721-723.
DeLozanne, A., and J. A. Spudich. 1987. Disruption of the Dictyostelium
myosin heavy chain gene by homologous recombination. Science
236:1086.
DeNofrio, D., T. C. Hoock, and I. M. Herman. 1989. Functional sorting of
actin isoforms in microvascular pericytes. J. Cell Biol. 109:191-202.
Detrich, H. W., V. Prasad, and R. F. Luduena. 1987. Cold-stable
microtubules from antarctic fishes contain unique alpha tubulins. J.
Biol. Chem. 262:8360-8366.
Dinsmore, J. H., and R. D. Sloboda. 1988. Calcium and calmodulindependent phosphorylation of a 62 kD protein induces microtubule
depolymerization in sea urchin mitotic apparatuses. Cell 53: 769-780.
Dinsmore, J. H., and F. Solomon. 1991. Inhibition of MAP2 expression
affects both morphological and cell division phenotypes of neuronal
differentiation. Cell 64: 817-826.
Doxsey, S. J., P. Stein, L. Evans, P. D. Calarco, and M. Kirschner. 1994.
Pericentrin, a highly conserved centrosome protein involved in
microtubule organization. Cell 76:639-650.
249
Dranoff, G., E. Jaffe, A. Lazenby, P. Golumbek, H. Levitsky, K. Brose, V.
Jackson, H. Hamada, D. Pardoll and R. C. Mulligan. 1993.
Vaccination with irradiated tumor cells engineered to secrete
murine granulocyte-macrophage colony-stimulating factor
stimulates potent, specific, and long-lasting anti-tumor immunity.
Proc. Natl. Acad. Sci. USA. 90:3539-3543.
Dreschel, D. N., A. A. Hyman, M. H. Cobb, and M. W. Kirschner. 1992.
Modulation of the dynamic instability of tubulin assembly by the
microtubule-associated protein tau. Mol. Biol. Cell 3:1141-1154.
Drubin, D. G., and M. W. Kirschner. 1986. Tau protein function in living
cells. J. Cell Biol. 103: 2739-2746.
Duerr, A., D. Pallas, and F. Solomon. 1981. Molecular analysis of
cytoplasmic microtubules in situ: identification of both widespread
and specific proteins. Cell 24:203.
Dustin, P. 1978. Microtubules. Springer-Verlag. New York.
Edwards, K. A., R. A. Montague, S. Shepard, B. A. Edgar, R. L. Erikson,
and D. P. Kiehart. 1994. Identification of Drosophila cytoskeletal
proteins by induction of abnormal cell shape in fission yeast. Proc.
Natl. Acad. Sci. USA. 91:4589-4593.
Endow, S. A., and M. A. Titus. 1992. Genetic approaches to molecular
motors. Ann. Rev. Cell Biol. 8: 29-66.
Esmaeli-Azad, B., J. H. McCarty, and S. C. Feinstein. 1994. Sense and
antisense transfection analysis of tau function: tau influences net
microtubule assembly, neurite outgrowth and neuritic stability. J.
Cell Sci. 107:869-879.
Fattoum, A., J. H. Hartwig, and T. P. Stossel. 1983. Isolation and some
structural and functional properties of macrophage tropomyosin.
Biochemistry 22:1187-1193.
Feinberg, A., and B. Vogelstein. 1983. A technique for radiolabeling DNA
restriction endonuclease fragments to high specific activity.
Analytical Biochem. 132: 6-13.
Felix, M. A., C. Antony, M. Wright, and B. Maro. 1994. Centrosome
assembly in vitro: role of gamma tubulin recruitment in Xenopus
sperm aster formation. J. Biol. Chem. 124: 19-31.
Ferreira, A., J. Niclas, R. D. Vale, G. Banker, and K. S. Kosik. 1992
Suppression of kinesin expression in cultured hippocampal neurons
using antisense oligonucleotides. J. Cell Biol. 117: 595-606.
250
Field, J., J.-I. Nikawa, D. Broek, B. MacDonald, L. Rodgers, I. A. Wilson,
R. A. Lerner, and M. Wigler. 1988. Purification of a RAS-responsive
adenylyl cyclase complex from Saccharomyces cerevisiae by use of an
epitope addition method. Mol. Cell. Biol. 8:2159-2165.
Finkel, T., J. A. Theriot, K. R. Dise, G. F. Tomaselli, P. J. GoldschmidtClermont. 1993. Dynamic actin structures stabilized by profilin.
Proc. Natl. Acad. Sci. USA 91:1510-1514.
Franck, Z., R. Gary, and A. Bretscher. 1993. Moesin, like ezrin, colocalizes
with actin in the cortical cytoskeleton in cultured cells, but its
expression is more variable. J. Cell Sci. 105:219-231.
Friederich, E., C. Huet, M. Arpin, and D. Louvard. 1989. Villin induces
microvilli growth and actin redistribution in transfected fibroblasts.
Cell. 59:461-475.
Fukuchi, K.-I. K. Kamino, S. S. Deeb, A. C. Smith, T. Dang, and G. M.
Martin. Overexpression of amyloid precursor protein alters its
normal processing and is associated with neurotoxicity. Biochem.
Biophys. Res. Comm. 182:165-173.
Funayama, N., A. Nagafuchi, N. Sato, Sa. Tsukita, and Sh. Tsukita. 1991.
Radixin is a novel member of the band 4.1 family. J. Cell Biol.
115:1039-1048.
Gaertig, J. M. A. Cruz, J. Bowen, L. Gu, D. G. Pennock, and M. A.
Gorovsky. 1995. Acetylation of lysine-40 in alpha-tubulin is not
essential in Tetrahymena thermophila. J. Cell Biol. 129: 1301-1310.
Galloway, P. G., G. Perry, K. S. Kosik, and G. Pierluigi. 1987. Hirano
bodies contain tau protein. Brain Res. 403: 337-340.
Gary, R. and A. Bretscher. 1993. Heterotypic and homotypic associations
between ezrin and moesin, two putative membrane-cytoskeletal
linking proteins. Proc. Natl. Acad. Sci. USA. 90:10846-10850.
Gary, R., and A. Bretscher. 1995. Ezrin self association involves binding of
an N-terminal domain to a normally masked C-terminal domain
that includes the F-actin binding site. Mol. Biol Cell 6:1061-1075.
Gilmore, A. P., and K. R. Burridge. 1995. Cryptic sites in vinculin. Nature
373:197.
Goslin, K., E. Birgbauer, G. Banker, and F. Solomon. 1989. The role of
cytoskeleton in organizing growth cones: a microfilament-associated
growth cone component depends upon microtubules for its
localization. J. Cell Biol. 109:1621-1631.
251
Gossen, M. and H. Bujard. 1992. Tight control of gene expression in
mammalian cells by tetracycline-responsive promoters. Proc. Natl.
Acad. Sci. USA. 89:5547-5551.
Gould, K. L., J. A. Cooper, A. Bretscher, and T. Hunter. 1986. The proteintyrosine substrate, p81, is homologous to a chicken microvillar core
protein. J. Cell Biol. 102:660-669.
Gould, K. L., A. Bretscher, F. S. Esch, and T. Hunter. 1989. cDNA cloning
and sequencing of the protein-tyrosine kinase substrate, ezrin,
reveals homology to band 4.1. EMBO (Eur. Mol. Biol. Organ.) J.
8:4133-4142.
Graham, F. L. and A.J. van der Eb. 1973. A new technique for the assay of
infectivity of human adenovirus DNA. Virology. 52:456-467.
Granger, B., and E. Lazarides. 1984. Cell 37:597-607.
Granger, B., and E. Lazarides. 1985. Nature 313:238-241.
Grundke-Iqbal, I., K. Iqbal, Y.-C. Tung, M. Quinlan, H. M. Wisniewski,
and L. I. Binder. 1986. Abnormal phosphorylation of the
microtubule-associated protein tau in Alzheimer cytoskeletal
pathology. Proc. Natl. Acad. Sci. USA 83:4913-4917.
Haarer, B. K., S. H. Lillie, A. E. M. Adams, V. Magoden, W. Bandlow, and
S. S. Brown. 1990. Purification of profilin from Saccharomyces
cerevisiae and analysis of profilin-deficient cells. J. Cell Biol.
110:105-114.
Hanemaaijer, R., and I. Ginzburg. 1991. Involvement of mature tau
isoforms in the stabilization of neurites in PC12 cells. J. Neurosci.
Res. 30: 163-171.
Hanzel, D., H. Reggio, A. Bretscher, J. G. Forte, and P. Mangeat. 1991.
The secretion-stimulated 80K phosphoprotein of parietal cells is
ezrin, and has properties of a membrane cytoskeletal linker in the
induced apical microvilli. EMBO J. 10:2363-2373.
Harada, A., K. Oguchi, S. Okabe, J. Kuno, S. Terada, T. Oshima, R. SatoYoshitake, Y. Takei, T. Noda, and N. Hirokawa. 1994. Altered
microtubule organization in small-calibre axons of mice lacking tau
protein. Nature 369:488-491.
Hartwig, J. H., G. M. Bokoch, C. L. Carpenter, P. A. Janmey, L. A. Taylor,
A. Toker, and T. P. Stossel. 1995. Thrombin receptor ligation and
activated rac uncap actin filament barbed ends through
phosphoinositide synthesis in permeablized human platelets. Cell
82:643-653.
252
Heidemann, S. R., and J. R. McIntosh. 1981. Visualization of the
structural polarity of microtubules. Nature 286:517.
Henry, M. D., C. Gonzalez Agosti. 1995. Molecular dissection of radixin:
distinct and interdependent functions of the amino- and carboxyterminal domains. J. Cell Biol. 129:1007-1022.
Himmler, A. 1989. Structure of the Bovine Tau Gene: Alternatively Spliced
Transcripts Generate a Protein Family. Mol. Cell Biol. 9: 1389-1396.
Hirokawa, N., Y. Shomura, and S. Okabe. 1988a. Tau proteins: the
molecular structure and mode of binding on microtubules. J. Cell
Biol. 107:1449-1460.
Hirokawa, N., S.-I. Nisanaga, Y. Shomura. 1988b. MAP2 is a component
of cross-bridges between microtubules and neurofilaments in the
neuronal cytoskeleton: Quick-freeze, deep etch immunoelectron
microscopy and reconstitution studies. J. Neurosci. 8:2769-2779.
Hoffman, P. N., and R. J. Lasek. 1975. The slow component of axonal
transport. J. Cell Biol. 66:351-366.
Horwitz, A., K. Duggan, C. Buck, M. C. Beckerle, and K. Burridge. 1986.
Interaction of plasma membrane fibronectin receptor with talin--a
transmembrane linkage. Nature 320: 531-533.
Hynes, R. 0. 1992. Integrins: versatility, modulation, and signalling in
cell adhesion. Cell 69:11-25.
Hynes, R. 0., and A. D. Lander. 1992. Contact and adhesive specificities in
the associations, migrations, and targeting of cells and axons. Cell
68:303-322.
Johnson, R. P., and S. W. Craig. 1994. An intramolecular association
between the head and tail domains of vinculin modulates talin
binding. J. Biol. Chem. 269: 12611-12619.
Johnson, R. P., and S. W. Craig. 1995. F-actin binding site masked by the
intramolecular association of vinculin head and tail domains.
Nature 373:261-264.
Joly, J. C., G. Flynn, and D. L. Purich. 1989. The microtubule-binding
fragment of MAP2:Location of the protease accessible site and
identification of an assembly promoting peptide. J. Cell Biol.
109:2289-2294.
253
Jones-Villeneuve, E. M. V., M. W. McBurney, K. A. Rogers, and V. I.
Kalnins. 1982. Retinoic acid induces embryonal carcinoma cells to
differentiate into neurons and glial cells. J. Cell Biol. 94: 253-262.
Jones-Villeneuve, E. M. V., Rudnicki, M. A., Harris, J. F., McBurney, M.
F. 1983. Retinoic Acid Induced Differentiation of Embryonal
Carcinoma Cells. Mol. Cell Biol. 3: 2271-2279.
Joshi, H. C., anf D. W. Cleveland. 1990. Diversity among tubulin subunits:
toward what functional end? Cell Motil. Cytoskel. 16:159-163.
Joshi, H. C. M. J. Palacios, L. McNamara, and D. W. Cleveland. 1992.
Nature 356:80-53.
Kabsch, W., H. G. Mannherz, D. Suck, E. F. Pai, and K. C. Holmes. 1990.
Atomic structure of the actin:DNase I complex. Nature 347:37-49.
Kanai, Y., R. Takemura, T. Oshima, H. Mori, Y. Ihara, M. Yanagisawa,
T. Masaki, and N. Hirokawa. 1989. Expression of Multiple Tau
Isoforms and Microtubule Bundle Formation in Fibroblasts
Transfected with a Single Tau cDNA. J. Cell Biol. 109: 1173-1184.
Katz, W., B. Weinstein, and F. Solomon. 1990. Regulation of tubulin levels
and microtubule assembly in Saccharomyces cerevisiae:
consequences of altered tubulin gene copy number. Mol. Cell Biol.
10:5294-5304.
Khawaja, S., G. G. Gunderson, and J. C. Bulinski. 1988. Enhanced
stability of microtubules enriched in de-tyrosinated tubulin is not a
direct function of detyrosination level. J. Cell Biol. 106: 141-149.
Knecht, D. A., and W. F. Loomis. 1987. Developmental consequences of the
lack of myosin heavy chain in Dictyostelium discoideum. Science 236:
1081.
Knops, J., K. S. Kosik, G. Lee, J. D. Pardee, L. Cohen-Gould, and L.
McConlogue. 1991. Overexpression of Tau in a nonneuronal cell
induces long cellular processes. J. Cell Biol. 114: 725-733.
Kim, S., M. Magendantz, W. Katz, and F. Solomon. 1987. Development of a
differentiated microtubule structure: formation of the chicken
erythrocyte marginal band in vivo. J. Cell Biol. 104:51-59.
King, S. M., T. Otter, and G. B. Witman. 1985. Characterization of
monoclonal antibodies against Chlamydamonas flagellar dyneins by
high-resolution protein blotting. Proc. Natl. Acad. Sci. 82: 4717-4721.
Kosik, K. S., and E. A. Finch. 1987. MAP2 and tau segregate into dendritic
and axonal domains after the elaboration of morphologically distinct
254
neurites: an immunocytochemical study of cultured rat cerebrum.
J. Neurosci. 7: 3142-3153.
Kosik, K. S., L. D. Orrechio, S. Bakalis, and R. L. Neve. 1989.
Developmentally regulated expression of specific tau sequences.
Neuron 2: 1389-1397.
Kreig, J., and T. Hunter. Identification of the two major epidermal growth
factor-induced tyrosine phosphorylation sites in the microvillar core
protein ezrin. J. Biol. Chem. 267: 19258-19265.
Laemlli, U. K. 1970. Cleavage of structural proteins during the assembly of
the head of bacteriophage T4. Nature (Lond.). 227:680-685.
Lankes, W., A. Greismacher, J. Gruenwald, R. Schwartz-Albeiz, and R.
Keller. 1988. A heparin binding protein involved in inhibition of
smooth-muscle cell proliferation. Biochem. J. 251:831-842.
Lankes, W. T. and H. Furthmayr. 1991. Moesin: a member of the protein
4.1-talin-exrin family of proteins. Proc. Natl. Acad. Sci. USA.
88:8297-8301.
Lawrence, J. B., and R. H. Singer. 1986. Intercellular localization of
messenger RNAs for cytoskeletal proteins. Cell 45:407-415.
Leclerc, N., K. S. Kosik, N. Cowan, T.P. Pienkowski and P. W. Baas. 1993.
Process formation in Sf9 cells induced by the expression of a MAP2Clike construct. Proc. Natl. Acad. Sci. USA 70:6223-6227.
Lee, G., N. Cowan, and M. Kirschner. 1988. The Primary Structure and
Heterogeneity of Tau Protein from Mouse Brain. Science 239: 285-288.
Lee, M., and A. D. Lander. 1991. Proc. Natl. Acad. Sci. USA 88:2768-2772.
Lee, V.M.-Y., B. J. Balin, L. Otovos, J. Q. Trojanowski. 1990. A68: a major
subunit of paired helical filaments and derivatized forms of normal
tau. Science 251:675-678.
Lesley, J., R. Hyman, and P. W. Kincaid. 1993. CD44 and its interaction
with the extracellular matrix. Advavces Immunol. 54:271-335.
Leto, T. L., I. Correas, T. Tobe, R. A. Anderson, and W. C. Horne. 1986.
Structure and function of human erythrocyte cytoskeletal protein 4.1.
In Membrane Skeletons and Cytoskeletal-Membrane Associations .
V. Bennet, C. M. Cohen, S. Lux, and J. Palek, editors. Alan R. Liss,
New York. 201-209.
Letourneau, P. C. 1975. Cell-to-substratum adhesion and guidance of
axonal elongation. Dev. Biol. 44: 92-101.
255
Lewis, S. A., D. Wang, and N. J. Cowan. 1988. Microtubule-associated
protein MAP2 shares a microtubule-binding motif with tau protein.
Science 242:936-939.
Lewis, S. A., I. E. Ivanov, G.-H. Lee, and N. J. Cowan. 1989. Organization
of microtubules in dendrites and axons is determined by a short
hydrophobic zipper in microtubule-associated proteins MAP2 and
tau. Nature 342: 498-505.
Lewis, S. A., and N. J. Cowan. 1990. Microtubule bundling. Nature
345:674.
Lim, W. A., R. T. Sauer, and A. D. Lander. 1991. Analysis of DNA-Protein
interactions by affinity coelectrophoresis. Methods Enzymol. 208:196210.
Lin, L.-N., and J. F. Brandts. 1980. Kinetic mechanism for the
conformational transitions between poly-L-prolines I and II: a study
utilizing the cis-trans specificity of a proline-specific protease.
Biochemistry 19:3055-3059.
Lindwall, G., and R. D. Cole. 1984. The Purification of Tau Protein and the
Occurence of Two Phosphorylation States of Tau in Brain. J. Biol.
Chem. 259: 12241-12245.
Lowry, 0. H., J. Rosebrough, A.S. Farr, and R. J. Randall. 1951. Protein
measurement with the folin phenol reagent. J. Biol. Chem. 193:265275.
Magendantz, M., M. D. Henry, A. Lander, and F. Solomon. 1995.
Interdomain interactions of radixin in vitro. J. Biol. Chem. In press.
Marchesi, V. T. 1985. Stabilizing infrastructure of cell membranes. Ann.
Rev. Cell Biol. 1:531-561.
Martin, F., M.-C. Harricane, E. Audemard, F. Pons, and D. Mornet. 1991.
Conformational change of turkey-gizzard caldesmon induced by
specific chemical modification with carbodiimide. Eur. J. Biochem.
195:335-342.
Martin, M. C. Andreoli, A. Shaquet, P. Montcourrier, M. Algrain, and P.
Mangeat. 1995. Ezrin NH2-terminal domain inhibits the cell
extension activity of the COOH-terminal domain. J. Cell Biol. 128:
1081-1093.
Matsudaira, P. 1992. Mapping structural and functional domains in actin
binding proteins. In The cytoskeleton a practical approach. K. L.
256
Carraway and C. A. C. Carraway eds. Oxford University Press: New
York pp.88-90.
Matsudaira, P., and P. Janmey. 1988. Pieces in the actin severing puzzle.
Cell 54:139-140.
Matus, A., R. Bernhardt, T. Hugh-Jones. 1981. Proc. Natl. Acad. Sci. USA
78:3010-3014.
Matus, A. 1990. Microtubule-associated proteins. Curr. Opin. Cell Biol. 2:
10-14.
McBurney, M. W., E. M. V. Jones-Villeneuve, M. K. S. Edwards, and P. J.
Anderson. 1982. Control of muscle and neuronal differentiation in a
cultured embryonal carcinoma cell line. Nature 299: 165-167.
McBurney, M. W. and Rogers, B. J. 1982. Isolation of Male Embryonal
Carcinoma Cells and Their Chromosome Replication Patterns. Dev.
Biol. 89: 503-508.
McBurney, M. W., Reuhl, K. R., Ally, A. I., Nasipuri, S., Bell, J. C., Craig,
J. 1988. Differentiation and Maturation of Embryonal CarcinomaDerived Neurons in Culture. J. Neurosci. 8: 1063-1073.
McNally, F. J., and R. D. Vale. 1993. Identification of katainin, an ATPase
that severs and disassembles stable microtubules. Cell 75:419-429.
Milam, L. M. 1985. Electrom microscopy of rotary shadowed vinculin and
vinculin complexes. J. Mol. Biol. 184:543-545.
Miller, M. and F. Solomon. 1984. Kinetics and intermediates of marginal
band reformation: evidence for peripheral determinants of marginal
band organization. J. Cell Biol. 99:2108.
Mitchison, T. J. 1993. Localization of an exchangeable GTP-binding site at
the plus ends of microtubules. Science 261:1044-1047.
Mitchison, T., and M. Kirschner. 1984a. Microtubule assembly nucleated
by isolated centrosomes. Nature 312: 232-237.
Mitchison, T., and M. Kirschner. 1984b. Dynamic instability of
microtubule growth. Nature 312: 237-242.
Mooseker, M. S. 1985. Organization, chemistry, and assembly of the
cytoskeletal apparatus of the intestinal brush border. Ann. Rev. Cell
Biol. 1:209-241.
Neugebauer, K. M., and L. Reichardt. 1991. Cell-surface regulation of biintegrin activity on developing retinal neurons. Nature.350: 68-71.
257
Niman, H. L., R. A. Houghten, L. E. Walker, R. A. Reisfeld, I. A. Wilson,
J. A. Hogle, and R. A. Lerner. 1983. Generation of protein-reactive
antibodies by short peptides is an event of high frequency:
implications for the structural basis of immune recognition. Proc.
Natl. Acad. Sci. USA. 80:4949-4953.
Niwa, H., K. Yammamura, and J. Miyazaki. 1991. Efficient selection for
high-expression transfectants with a novel eukaryotic vector. Gene
108:192-199.
Nobes, C. D., and A. Hall. 1995. Rho, rac, and cdc42 GTPases regulate the
assembly of multimolecular focal complexes associated with actin
stress fibers, lamellipodia, and filopodia. Cell 81: 53-62.
Nuckolls, G. H., C. E. Turner, and K. Burridge. 1990. Functional studies
of the domains of talin. J. Cell Biol. 110:1635-1644.
Ogden, R. C., and D. A. Adams. 1987. Elactrophoresis in agarose and
acrylamide gels. Meth. Enzymol. 152: 61-87.
Papasozomenos, S. C., and L. I. Binder. 1987. Phosphorylation Determines
Two Distinct Species of Tau in the Central Nervous System. Cell
Motil. Cytoskeleton 8: 210-226.
Paschal, B. M., H. S. Sheptner, R. Vallee. 1987. MAP1C is a microtubuleactivated ATPase which translocates microtubules in vitro and has
dynein-like properties. J. Cell Biol. 105: 1273-1282.
Peng, I., L. I. Binder, and M. M. Black. 1986. J. Cell Biol. 102:252-262.
Pestonjamasp, K., M. R. Amieva, C. P. Strassel, W. M. Nauseef, H.
Furthmayr, and E. J. Luna. 1995. Moesin, Ezrin and p205 are actinbinding proteins associated with neutrophil plasma membranes.
Mol. Biol. Cell 6: 247-259.
Pittenger, M. F., and D. M. Helfman. 1992. In vitro and in vivo
characterization of four fibroblast tropomyosins produced in bacteria:
TM-2, TM-3, TM-5a, and TM-5b are co-localized in interphase
fibroblasts. J. Cell Biol. 118: 841-858.
Pollard, T. D., and E. D. Korn. 1973. Acanthamoeba myosin I. Isolation
from Acanthamoeba castellanii of an enzyme similar to muscle
myosin. J. Biol. Chem. 248:4682-4690.
Rees, D. J. G., S. E. Ades, S. J. Singer, and R. 0. Hynes. Sequence and
domain structure of talin. Nature (Lond.). 347:685-689.
258
Reinsch, S. S., T. J. Mitchison, and M. Kirschner. 1991. Microtubule
polymer assembly and transport during axonal elongation. J. Cell.
Biol. 115:365-379.
Rouleau, G. A., P. Merel, M. Lutchman, M. Sanson, J. Zucman, C.
Marineau, K. Hoang-Xuan, S. Demezuk, C. Desmaze, B. Plougastel,
S. M. Pulst, G. Lenoir, E. Bijlsma, R. Fashold, J. Dumanski, P.
Jong, D. Parry, R. Eldridge, A. Aurias, 0. Delattre, and G. Thomas.
1993. Alteration in a new gene encoding a putative membraneorganizing protein causes neuro-fibromatosis type 2. Nature (Lond.).
363:515-521.
Rudnicki, M. A., and M. W. McBurney. 1987. Cell culture methods and
induction of differentiation of embryonal carcinoma cell lines. In
Teratocarcinomasand embryonic stem cells: a practical approach.
Edited by E. J. Robertson. 19-49. Oxford: IRL Press Limited.
Rudnicki, M. A., M. Ruben, and M. W. McBurney. 1988. Regulated
expression of a human cardiac actin gene during differentiation of
multipotential embryonal carcinoma cells. Mol. Cell. Biol. 8: 406-417.
Sabry, J. H., T. P. O'Connor, L. Evans, A. Toroian-Raymond, M. Kirschner
and D. Bentley. 1991. Microtubule begavior during guidance of
pioneer neuron growth cones in situ. J. Cell Biol. 115: 381-395.
Samuels, M., R. M. Ezzell, T. J. Cardozo, D. R. Critchley, J.-L. Coll, and E.
D. Adamson. 1993. Expression of chicken vinculin complements the
adhesion-defective phenotype of a mutant mouse F9 embryonal
carcinoma line. J. Cell Biol. 121:909-921.
San Antonio. J. D., J. Slover, J. Lawler, M. J. Karnovsky, and A. D.
Lander. 1993. Biochemistry 32:4746-4755.
Sato, N., S. Yonemura, T. Obinata, Sa. Tsukita, and Sh. Tsukita. 1991.
Radixin, a barbed end-capping actin-modulating protein, is
concentrated at the cleavage furrow during cytokinesis. J. Cell Biol.
113:321-330.
Sato, N., N. Funayama, A. Nagafuchi, S. Yonemura, Sa. Tsukita, and Sh.
Tsukita. 1992. A gene family consisting of ezrin, radixin, and
moesin. Its specific localization at actin filament/plasma membrane
association sites. J. Cell Sci. 103:131-143.
Schacterle, G. R., and R. L. Pollack. 1973. A simplified method for the
quantitative assay of small amounts of protein in biologic material.
Anal. Biochem. 51: 654-655.
259
Schatz, P. J., F. Solomon, and D. Botstein. 1986. Genetically essential and
non-essential alpha tubulin genes specify functionally
interchangeable proteins. Mol. Cell Biol. 6:3722.
Schulze, E. S., and M. W. Kirschner. 1988. Direct observation of
microtubule dynamics in living cells. Nature 334:356-359.
Shuster, C. B., and I. M. Herman. Indirect association of ezrin with Factin: Isoform specificty and calcium sensitivity. J. Cell Biol.
128:837-848.
Small, J. V., G. Isenberg, J. E. Celis. 1978. Polarity of actin at the leading
edge of cultured cells. Nature 272: 638-639.
Small, J. V., G. Rinnerthaler, H. Hinssen. 1982. Organization of actin
meshworks in cultured cells: the leading edge. Cold Spring Harb
Symp Quant Biol 46:599-611.
Small, J. V., M. Herzog, and K. Anderson. 1995. Actin filament
organization in the fish keratocyte lamellipodium. J. Cell Biol. 129:
1275-1286.
Stearns, T., and M. Kirschner. 1994. In vitro reconstitution of centrosome
assembly and function: the central role of gamma-tubulin. Cell:
76:623-637.
Stephens, L. E., J. E. Sonne, M. L. Fitzgerald, and C. H. Damsky. 1993.
Targeted deletion of beta 1 integrins in F9 embryonal carcinoma cells
affects morphological differentiation but not tissue specific gene
expression. J. Cell Biol. 123: 1607-1620.
Solomon, F., and M. Magendantz. 1981. Cytochalasin separates
microtubule disassembly from loss of asymmetric morphology. J.
Cell Biol. 89: 157-161.
Solomon, F., M. Magendantz, and A. Salzman. 1979. Identification with
cellular microtubules one of the coassembling microtubuleassociated proteins. Cell. 18:431-438.
Southern, P. J., and P. Berg. 1982. Transformation of mammalian cells to
antibiotic resistance with a bacterial gene under control of the SV40
early region promoter. J. Mol. Appl. Gen. 1:327-341.
Spudich, J. A., and Watt, S. 1971. J. Biol. Chem. 246:4866.
Sretavan, D. W., and L. F. Reichardt. 1993. Time-lapse video analysis of
retinal ganglion cell axon pathfinding at the mammalian optic
chiasm: growth cone guidance using intrinsic chiasm cues. Neuron
10:761-777.
260
Sullivan, K. F. 1988. Structure and utilization of tubulin isotypes. Ann.
Rev. Cell Biol. 4:687-716.
Superti-Furga, G., S. Fumagalli, M. Koegl, S. A. Courtneidge, and G.
Draetta. 1993. Csk inhibition of c-Src activity requires both the SH2
and SH3 domains of Src. EMBO J. 12:2625-2634.
Swan, J. A.,and F. Solomon. 1984. Reformation of the marginal band of
avian erythrocytes in vitro using calf-brain tubulin: preipheral
determinants of microtubule form. J. Cell Biol. 99: 2108-2113.
Takaishi, K., T. Sasaki, T. Kameyama, S. Tsukita, S. Tsukita, and Y.
Takai. 1995. Translocation of activated rho from the cytoplasm to
membrane ruffling area, cell-cell adhesion sites and cleavage
furrows. Oncogene 11: 39-48.
Takeuchi, K., N. Sato, H. Kasahara, N. Funayama, A. Nagafuchi, S.
Yonemura, Sa. Tsukita, and Sh. Tsukita. 1994. Perturbation of cell
adhesion and microvilli formation by antisense oligonucleotides to
ERM family members. J. Cell Biol. 125:1371-1384.
Tanaka, E. M., and M. W. Kirschner. 1991. Microtubule behavior in the
growth cones of living neurons during axon elongation. J. Cell Biol.
115:345-363.
Temm-Grove, C. J., B. M. Jockush, M. Rohde, K. Niebuhr, T. Chakraborty,
and J. Wehland. 1994. Exploitation of microfilament proteins by
Listeria monocytogenes: microvillus-like composition of the comet
tails and vectorial spreading in epithelial sheets. J. Cell Sci. 107:
2951-2960.
Theriot, J. A., T. J. Mitchison, L. G. Tilney, and D. A. Portnoy. 1992. The
rate of actin-based motility of intracellular Listeria monocytogenes
equal the rate of actin polymerization. Nature 357:257-260.
Theriot, J. A., J. Rosenblatt, D. A. Portnoy, P. J. Goldschmidt-Clermont
and T. J. Mitchison. 1994. Involvement of profilin in the actin-based
mobility of Listeria monocytogenes in cells and cell-free extracts.
Cell 76:505-517.
Tilney, L. G., D. J. DeRosier, and M. S. Tilney. 1992. How Listeria exploits
host cell actin to form its own cytoskeleton. I. Formation of a tail and
how that tail might be involved in movement. J. Cell Biol. 118:71-81.
Tobin, H., T. Staehelin, and J. Jordon. 1979. Electrophoretic transfer of
proteins from polyacrylamide gels to nitrocellulose sheets; procedure
and some applications. Proc. Natl. Acad. Sci. USA. 76:4350-4354.
261
Trofatter, J. A., M. M. MacCollin, J. L. Rutter, J. R. Murrell, M. P. Duyao,
D. M. Parry, R. Eldridge, N. Kley, A. G. Menon, K. Pulaski, V. H.
Hasse, C. M. Ambrose, D. Munroe, C. Bove, J. L. Haines, R. L.
Martuza, M. E. MacDonald, B. R. Seizinger, M. P. Short, A. J.
Buckler, and J. F. Gusella. 1993. A novel moesin-, ezrin-, radixinlike gene is a candidate for the neurofibromatosis 2 tumor
suppressor. Cell . 72:791-800.
Tsukita, Sa., Y. Hieda, and Sh. Tsukita. 1989. A new 82-kD barbed endcapping protein (radixin) localized in the cell-to-cell adherens
junction: purification and characterization. J. Cell Biol. 108:23692382.
Tsukita, Sa., K. Oishi, N. Sato, J. Sagara, A. Kawai, and Sh. Tsukita. 1994.
ERM family members as molecular linkers between the cell surface
glycoprotein CD44 and actin-based cytoskeletons. J. Cell Biol.
126:391-401.
Turunen, 0., T. Wahlstrom, and A. Vaheri. 1994. Ezrin has a COOHterminal actin binding site that is conserved in the ezrin protein
family. J. Cell Biol. 126: 1445-1453.
Vale, R. D., T. S. Reese, M. P. Sheetz. 1985. Identification of a novel forcegenerating protein, kinesin, involved in microtubule-based motility.
Cell 42:39-50.
Vallee. 1985. On the Use of Heat Stability as a Criterion for the
Identification of Microtubule Associated Proteins (MAPS). Biochem.
Biophys. Res. Comm. 133: 128-133.
Vallee, R. B., H. S. Sheptner, and B. M. Paschal. 1989. The role of dynein
in retrograde axonal transport. Trends Neurosci. 12: 66-70.
Vanderkerckhove, J., and K. Weber. 1978. At lest six different actins are
expressed in a higher mammal: an analysis based on the amino acid
sequence of the amino-terminal tryptic peptide. J. Mol. Biol. 126: 783802.
Wallace, D. M. 1987. Large and small scale phenol extractions. Meth.
Enzymol. 152: 33-41.
Wang, K., and S. J. Singer. 1977. Interaction of filamin with F-actin in
solution. Proc. Natl. Acad. Sci. 74:2021-2025.
Warren, K. S., J. L-C. Lin, D. D. Wamboldt, and J. J.-C. Lin. 1994.
Overexpression of human fibroblast caldesmon fragment containing
actin-, Ca++/calmodulin-, and tropomyosin-binding domains
stabilizes endogenous tropomyosin and microfilaments. J. Cell Biol.
125:359-368.
262
Watts, R. G., and T. H. Howard. 1992. Evidence for a gelsolin-rich, labile
F-actin pool in human polymorphonuclear leukocytes. Cell Motil.
Cytoskel. 21:25-37.
Watts, R. G., and T. H. Howard. 1993. Mechanisms for actin
reorganization in chemotactic factor-activated polymorphonuclear
leukocytes. Blood 81:2750-2757.
Webster, R. E., D. Henderson, M. Osborn, and K. Weber. 1978. Threedimensional electron microscopal visualization of the skeleton of
animal cells: immunoferritin identification of actin- and tubulincontaining structures. Proc. Natl. Acad. Sci. USA 75:5511-5515.
Weingarten, M., A. H. Lockwood, S. Y. Hwo, and M. W. Kirschner. 1975.
A protein factor essential for microtubule assembly. Proc. Natl.
Acad. Sci. USA 72: 1858-1862.
Weinstein, B. and F. Solomon. 1990. Consequences of overexpression of
tubulin in Saccharomyces cerevisiae: differences between alphatubulin and beta-tubulin. Mol. Cell Biol. 10:5294-5304.
Weisenberg, R. 1972. Microtubule formation in vitro in solutions
containing low calcium concentration. Science 177:1104.
Weisenberg, R., and E. W. Taylor. 1968. Studies on ATPase activity of sea
urchin eggs and the isolated mitotic apparatus. Exp. Cell Res.
53:372-384.
Weiss, P., and H. B. Hiscoe. 1948. Experiments on the mechanism of nerve
growth. J. Exp. Zoo. 107:315-395.
Wiche, G., C. Oberkanins, and A. Himmler. 1991. Molecular structure
and function of microtubule-associated proteins. Int. Rev. Cytol. 124:
217-273.
Willard, M., W. M. Cowan, and P. R. Vagelos. 1974. The polypettide
composition of intra-axonally transported proteins: evidence for four
transport velocities. Proc. Natl. Acad. Sci. USA 71: 2183-2187.
Winckler, B. 1994. The role of the cytoskeleton in the morphogenesis of the
avian erythrocyte. PhD thesis. MIT.
Winckler, B. and F. Solomon. 1991. A role for microtubule bundles in the
morphogenesis of chicken erythrocytes. Proc. Natl. Acad. Sci. USA
88:6033-6037.
Winckler, B., Ch. Gonzalez Agosti, M. Magendantz, and F. Solomon. 1994.
Analysis of a cortical cytoskeletal structure: a role for ezrin-radixin-
263
moesin (ERM proteins) in the marginal band of chicken erythrocytes.
J. Cell Sci. 107:2523-2534.
Yamada, K. M., B. S. Spooner, and N. K. Wessels. 1970. Axon growth:
foles of microfilaments and microtubules. Proc. Natl. Acad. Sci.
USA 66:1206-1212.
264