ANALYSIS OF INTERACTIONS BETWEEN RADIXIN AND ITS
LIGANDS NHE-RF AND LAYILIN
BY
ETCHELL ANN CORDERO
B.S., University of California, Berkeley
1991
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 1999
@1999 Massachusetts Institute of Technology
All rights reserved
Signature of Author
Department of Biology
Certified by
--
-
-- - --
--
-- - --
-
-- -
-
-
-
-
-
-
-
-
-
--
-
-
-
Thesis Ad isor Professor Frank Solomon
Accepted by
Chairman of the Graduate Comn tte
SSACHUSETTS INSTITUTE
LIBRARIES
ANALYSIS OF INTERACTIONS BETWEEN RADIXIN AND ITS
LIGANDS NHE-RF AND LAYILIN
BY
ETCHELL ANN CORDERO
Submitted to the Department of Biology on February 1, 1999
in partial fulfillment of the requirements for the degree of
Doctor of Philosophy in Biology
ABSTRACT
ERM proteins (ezrin, radixin, moesin) are cytoskeletal-membrane linkers. The
regulation of the interactions between ERM proteins and their ligands may
modulate the organization of cortical structures in the cell. Previously we identified
NHE-RF as a ligand for merlin and ERM proteins (Murthy 1998). In this paper, we
characterize the interaction between NHE-RF and radixin as well as identify a novel
ligand for radixin - layilin. Using affinity chromatography binding assays, we show
that layilin and NHE-RF bind directly to radixin at its amino-terminal domain. Our
results show that an interdomain interaction between the amino- and carboxyterminal domains of radixin inhibits the ability of the ligands, NHE-RF and layilin,
to bind to the amino-terminal domain of radixin. Furthermore, we show that the
conformation of the amino-terminal domain changes in the presence of PIP. In the
presence of PIP, this interdomain interaction is inhibited and ligand binding to fulllength radixin is enhanced. Taken together, our results give a mechanistic rationale
for the regulation of the interaction between ERM proteins and their ligands by
phospholipids.
Thesis Advisor: Dr. Frank Solomon
Title: Professor of Biology
2
Acknowledgements
To Shau Neen - I couldn't have done this without you. You had dreams and
visions for me that I never considered for myself. You knew I had the abilities to get
a Ph.D. before I knew it was an option. Thank you for being my colleague and best
friend.
To Frank - You are my mentor, my therapist and my advisor for the rest of my life.
Thank you for being confident in my abilities and teaching me what is important in
life and in science.
To my collaborators Charo Gonzalez-Agosti, Vijaya Ramesh, and Mark Borowsky.
Thank you for sharing your reagents and your ideas with me. You helped me bring
my work to a higher level and it was a joy working with such nice people.
To my thesis committee members Tyler Jacks, Richard Hynes, Frank Gertler and
Vijaya Ramesh - I appreciate the guidance you have given me to make my thesis
stronger. I also thank you for making my thesis defense a stimulating and exciting
experience and I regret not taking advantage of these meetings earlier in my career.
To Jim - You are "a ray of sunshine in my life" and my favorite baymate. Thanks
for putting up with all my moody days and for knowing when and when not to be a
bully. Now you don't have to worry about me taking up more than my half of the
bay - enjoy!
To Letty - You are my girlfriend. Thank you for filling my days in lab with
(neverending) stories, hugs, love and smiles.
To Alice Rushforth and Sylvia Sanders - Thank you for showing me that it's
possible to be a good mother and scientist at the same time.
To the rest of the Solomon Lab - Margaret Magendantz, Kate Compton, Adelle
Smith, and Will Chen - thank you for your helpful discussions and advice at group
meetings. I wish the best for you all.
To the rest of the people in my life who gave me something to do besides
benchwork: Thanks to all my friends who made my life outside of lab memorable
and fun. Thanks to my ultimate frisbee buddies who gave me a sport to play and
memories of success that I'll keep with me for the rest of my life. Finally, thanks to
my family who unconditionally supported everything I did.
yeah!
3
Table of Contents
Title Page
Abstract
Acknowledgements
Table of Contents
List of Figures
1
2
3
4
6
CHAPTER 1:
Introduction
Thesis overview
Bibliography
7
8
29
30
CHAPTER 2:
Intra- and inter-domain interactions
Introduction
Materials and Methods
Results
Discussion
Bibliography
36
37
40
47
72
78
CHAPTER 3:
Radixin and NHE-RF interactions
Introduction
Materials and Methods
Results
Discussion
Bibliography
80
81
85
88
108
114
CHAPTER 4:
Radixin and layilin interactions
Introduction
Materials and Methods
Results
Discussion
Bibliography
117
118
120
123
134
138
4
CHAPTER 5:
Analysis of Merlin
Introduction
Materials and Methods
Results
Discussion
Bibliography
139
140
143
145
149
153
CHAPTER 6:
Model and conclusions
Bibliography
156
157
167
CHAPTER 7:
Future Prospects
170
171
CHAPTER 8:
Appendix - Publications
175
176
5
List of Figures
Figure 1-1
ERM family of proteins
Figure 1-2
Current model of ERM regulation
Figure 2-1
Detecting an amino- and carboxy-terminal domain
interaction
Figure 2-2
Phospholipids bind to radixin at its amino-terminal
domain
Figure 2-3
Phospholipids regulate the interdomain interaction
Figure 2-4
Cross-linkers capture an intra-molecular interaction
Figure 2-5
Formation of the faster band is dependent on an
intact cross-linker and the conformation of the
protein
Figure 2-6
Cross-linker kinetics
Figure 2-7
Cross-linking domains of ERM proteins
Figure 2-8
Phospholipids disrupt the formation of the
intramolecular interaction
Figure 2-9
Specific phospholipids disrupt cross-linking
Figure 2-10
Intra- and inter-domain interactions of radixin are
regulated by phospholipids
Figure 3-1
Immunofluorescence of NHE-RF and ERM proteins
Figure 3-2
NHE-RF is a ligand for radixin
Figure 3-3
NHE-RF binds to radixin in a concentration
dependent manner
Figure 3-4
Binding between NHE-RF and the amino-terminal
domain of radixin is saturable
Figure 3-5
Radixin binds to NHE-RF on glutathione beads
Figure 3-6
An interdomain interaction between the amino- and
carboxy-terminal domains of radixin block NHE-RF
binding to radixin
6
Figure 3-7
Phospholipids enhance NHE-RF binding to full
length radixin
Figure 3-8
Cross-linking radixin does not affect ligand binding or
domain competition
Figure 3-9
Model for phospholipid regulation of radixin binding
to NHE-RF
Figure 4-1
Layilin and ERM proteins partially co-localize at the
ruffling edges
Figure 4-2
Layilin is a ligand for radixin
Figure 4-3
An interdomain interaction blocks layilin binding
Figure 4-4
Phospholipids enhance layilin binding to full length
radixin
Figure 4-5
Model for phospholipid regulation of radixin binding
to layilin
Figure 5-1
Merlin forms an intramolecular interaction which
can be captured by cross-linkers
Figure 6-1
Model for ERM regulation
7
CHAPTER 1:
INTRODUCTION
The cortical cytoskeleton participates in crucial aspects of cell motility,
morphological differentiation and organization of the plasma membrane.
Understanding the molecular basis for those functions remains a central question
for cell biologists. It seems likely that specific molecules are required to localize
differentially in the membrane, where they mediate specific organizations of
cytoskeletal and membrane components.
Among the cortical cytoskeletal components, the localization patterns of ERM
proteins (ezrin, radixin, and moesin) are particularly intriguing. Ezrin, radixin and
moesin - the ERM proteins, closely related by primary sequence - localize to a wide
variety of cortical structures associated with microfilaments. This pattern suggests
that they may participate in the establishment and maintenance of differentiated
morphology. The discovery of a fourth, somewhat less closely related polypeptide,
merlin, as the product of a tumor suppressor gene greatly increased interest in this
family of proteins.
IDENTIFICATION OF EZRIN, RADIXIN AND MOESIN
Ezrin was first identified by Bretscher (Bretscher, 1983) in the course of a
systematic analysis of the intestinal microvillar cytoskeleton, an organelle that has
been a valuable tool for understanding actin-membrane interactions. A
collaboration between groups demonstrated that ezrin was identical to a previously
identified endogenous substrate for tyrosine kinase activity (Gould et al., 1986).
Biochemical searches for actin- and membrane-associated proteins led to the
identification of moesin (Lankes et al., 1988) and radixin (Tsao et al., 1990). The
Chapter 1
8
Introduction
three proteins were identified as a family on the basis, first, of their sequence
similarities (between 72% and 80% identity for the pairwise comparisons of the
three proteins). Subsequent experiments revealed that they share many other
properties. The ERM proteins are most related in their amino-terminal halves
(between 85 and 89% identity), which in turn have significant sequence homology to
the amino-terminal domain of band 4.1 (Figure 1-1). The carboxy-terminal domain
varies among the three proteins, although they may all specify binding sites for
ligands including actin (see below). Whether this diversity has any functional
consequences is not established, but it has permitted generation of antibodies specific
for each of the polypeptides.
Other proteins have also been identified as members of the Band 4.1
superfamily by sequence homology, and so are related to ERM proteins. EM10, a
protein identified as a surface antigen on the tapeworm that causes echinocosis, is
55% identical to ezrin in their amino-terminal domains (Frosch et al., 1991). A
series of protein tyrosine phosphatases have also been identified because of their
homology to the N-terminal domain of Band 4.1 and ERM proteins (Yang and
Tonks, 1991; L'Abbe et al., 1994; Sawada et al., 1994; Higashitsuji et al., 1995). The
most provocative sequence homology is to merlin, the NF2 gene product (see
below). What is not clear is what common elements of structure and function this
family represents.
Chapter 1
9
Introduction
ERM PROTEINS LOCALIZE TO SPECIALIZATIONS OF THE CELL CORTEX
Although the immunofluorescence localizations and relative abundance of
each of the proteins may vary among cell types (Berryman, 1993), at least two of the
proteins are detectable in nearly every cell type tested. There also is no compelling
evidence demonstrating segregation of individual ERM proteins within the same
cell. Instead, there are several examples of two or more proteins co-localizing
precisely. These descriptive experiments support the notion that different ERM
proteins may have overlapping, if not identical, functions. They also permit the
discussion of ERM localization without distinguishing among the family members.
The localized fraction of cellular ERM proteins is cortical, in particular to
domains of the cortex juxtaposed to F-actin structures: ruffling edges, blebs,
microspikes, and neuronal growth cones (Bretscher, 1983; Gould et al., 1986;
Pakkanen et al., 1987; Pakkanen, 1988; Goslin et al., 1989; Birgbauer, 1991; Berryman,
1993; Franck, 1993; Gonzalez et al., 1996). ERM proteins are also prominent in nonmotile regions such as intestinal microvilli (Bretscher, 1983; Gould et al., 1986;
Pakkanen et al., 1987; Pakkanen, 1988; Berryman, 1993). In mitotic cells, ERM
proteins are greatly enriched at the cleavage furrow (Sato et al., 1991; Franck, 1993;
Henry et al., 1995).), adherens junctions (Tucker et al., 1989) and the marginal band
of nucleated erythrocytes (Birgbauer and Solomon, 1989). The localization of ERM
proteins, then may represent a common structural feature - for example, specialized
interactions between cortex and cytoskeleton - rather than a function dedicated to
motility. And while ERM proteins can co-localize precisely with certain cellular
actin structures, they are conspicuously absent from others, in particular the
Chapter 1
10
Introduction
prominent cytoplasmic actin fibers, such as stress fibers. Even in organelles where
both ERM and F-actin are enriched, such as the growth cone, their patterns overlap
rather than coincide (Goslin et al., 1989). In this sense, ERM proteins do not behave
as do simple F-actin associated proteins. These restrictions on ERM localization - to
a subset of both cortical structures and F-actin - are evidence of in vivo mechanisms
that regulate crucial interactions. Analysis of possible regulatory mechanisms is a
major focus of the field.
INTERACTIONS IN CIS AND IN TRANS:
ERM PROTEINS AS BIFUNCTIONAL CONNECTORS
A series of in vitro and in vivo assays detect two sorts of non-covalent
interactions for ERM proteins: intermolecular binding of other species; and
interdomain binding that could either mediate self-association to form oligomers or
act in an intramolecular fashion to stabilize conformations of individual molecules.
Results from several laboratories suggest that these two types of interactions may be
mutually exclusive. Perhaps the sites necessary for binding of other species are
occluded when the domains of ERM proteins are able to interact with one another.
This reciprocal relationship could provide the molecular mechanism for regulation
of ERM interactions and localization. As discussed below, this model explains some
but not all of the relevant experimental data.
Heterologous ligands: Given their localization in actin-rich structures, it is
possible that ERM proteins bind directly to F-actin. Different assays, using different
isoforms of actin and different sources of ERM proteins, produce conflicting binding
Chapter 1
11
Introduction
results (Turunen et al., 1994; Pestonjamasp et al., 1995; Shuster and Herman, 1995;
Yao et al., 1995; Yao et al., 1996). Attempts using standard actin pelleting assays to
detect an interaction between actin and full-length ERM proteins have been
unsuccessful (Bretscher, 1986), Cordero and Solomon, unpublished observations).
In the experiments which succeeded in detecting an interaction between actin and
ERM proteins, the ERM proteins are denatured (Pestonjamasp et al., 1995) or ERM
fragments are used rather than the full-length polypeptide (Turunen et al., 1994).
Those assays delineate the sequences in the carboxy-terminus necessary for this
interaction (Turunen et al., 1994). Others have argued that native ezrin interacts
specifically with the
P-actin isoform and not with skeletal actin (Shuster and
Herman, 1995; Yao et al., 1995; Roy et al., 1997). In at least one of those cases (Shuster
and Herman, 1995), this unusual specificity appears to depend upon the presence of
other proteins, which could be considered as consistent with the requirement that
the association with F-actin is not a simple binding reaction but rather one that
must be regulated. Thus, it is unclear whether the specificity they detect is a
property of ERM proteins or other proteins associating with ERM proteins. Using a
microtiter binding assay similar to ELISA tests, Roy et al (1997) have identified an
actin binding site present in the amino-terminus of ezrin which has not been
detected in any previous assay. One explanation for this discrepancy may be that the
state of the protein, specifically the presence or absence of a protein tag, affects the
conformation of the amino-terminal domain. The authors suggest that a free
amino-terminus is crucial for unmasking the actin-binding properties at the amino-
Chapter 1
12
Introduction
terminus. The details of the ERM protein interaction with actin are yet to be
resolved.
The full-length molecule must have ligands other than F-actin. For example,
nearly wild type amounts of radixin expressed in stable transfectants not only
localize normally, but displace endogenous moesin from all its normal loci (Henry
et al., 1995). This result suggests that the cell contains a saturable element required
for proper localization of ERM proteins. To identify such ligands, Tsukita and
colleagues (Tsukita et al., 1994) analyzed the proteins that co-immunoprecipitate
with anti-ERM antibodies from cultured cells. They showed that a fraction of the
cellular CD44, a cell surface glycoprotein expressed in a wide variety of cells,
associates with ERM proteins both by co-precipitation and by co-localization at the
level of light microscopy. The domains of co-localization include surface microvilli,
cell-cell adhesion sites, and the cleavage furrow. Significantly, there appear to be
significant domains of at least CD44 staining that do not co-localize with ERM,
suggesting again that the presence of this ligand is not sufficient to specify ERM
localization. Several other membrane-associated proteins, such as CD43 , ICAM-1
and ICAM-2 (Heiska et al., 1998; Serrador et al., 1998; Yonemura et al., 1998) have
also been identified as binding partners for ERM proteins. Ezrin was identified as a
ligand for ICAM-1 after placenta lysate was incubated with a matrix containing
ICAM-1, and one of the most prominent bands eluting from the affinity column
was ezrin (Heiska et al., 1998). In addition, a peptide encompassing the cytoplasmic
region of ICAM-2 coupled to Sepharose beads interacted with ezrin (Heiska et al.,
1998). Three proteins, ICAM-2, CD43 and CD44, share a positively charged amino
Chapter 1
13
Introduction
acid motif which was found to be necessary and sufficient for ERM protein binding
(Yonemura et al., 1998). Using deletion constructs in in vitro binding studies, these
authors mapped the ERM binding site to the first 20 amino acids of the cytoplasmic
domain of CD43 and CD44. These amino acids were also important for the correct
co-localization and immunoprecipitation of CD44 with ERM proteins. Lastly, by
site-directed mutagenesis the authors identified the positively charged amino acid
clusters in these three integral membrane proteins as necessary for their interaction
with ERM proteins (Yonemura et al., 1998).
Several ligands integrate ERM proteins into the Rho signal transduction
pathway. The Rho-GDP dissociation inhibitor, Rho-GDI, a regulator of the GTPase
activity for Rho, immunoprecipitates with anti-ERM protein antibodies from
cultured cells (Hirao et al., 1996). Another protein in the Rho signal transduction
pathway, the myosin binding subunit (MBS) of myosin phosphatase, co-localizes
with moesin to Rho induced ruffling edges in MDCK cells. By coimmunoprecipitation experiments, moesin and MBS interact with one another and
binding studies showed a direct interaction between MBS and the N-terminal
domains of moesin and ezrin (Fukata et al., 1998). Furthermore, the interaction
between ERM proteins and the pelletable fraction from cell culture lysates increases
in the presence of GTPyS and decreases after the addition of C3 toxin, a specific
inhibitor of Rho (Heiska et al., 1996). These data suggest that ERM proteins may be
recruited to the plasma membrane and its interactions may be regulated in a Rho
dependent manner.
Chapter 1
14
Introduction
The effect of phosphorylating ERM proteins and other possible methods of
regulating intermolecular interactions will be discussed later in this chapter.
Interdomain interactions:
The amino- and carboxy-terminal halves of the
molecule not only bind to other ligands, but to one another as well. Such an
interaction was first suggested by the demonstration that a significant fraction of
ezrin and moesin from cultured cells could be detected in complexes that contain
both proteins (Gary and Bretscher, 1993). By gel overlay assays, purified ezrin and
moesin bind to one another, suggesting that the interaction may be direct and
binary. Mapping the sequences involved in these associations identifies two
domains, in the two extremes of the molecule (Gary and Bretscher, 1995). A direct
interaction between the domains is detectable by affinity co-electrophoresis; the data
are consistent with a 1:1 complex of the amino- and carboxy-terminal halves of the
protein, with a dissociation constant Kd = -4x1O- 8 M (Magendantz et al., 1995).
Significantly, many of the binding studies described above between ERM
proteins and their ligands suggest that the relevant binding site which is embedded
in the separate domains is not functional in full-length molecules except upon
denaturation. Several ligands such as actin, CD44, band 4.1 and NHE-RF/EBP50
bind to a separate domain of ERM proteins more efficiently than to the full-length
proteins (Tsukita et al., 1994; Magendantz et al., 1995; Pestonjamasp et al., 1995;
Reczek et al., 1997; Murthy et al., 1998). These data suggest that interdomain
interactions may compete with binding of other ligands. Some predictions of this
Chapter 1
15
Introduction
view are borne out by the properties of ERM polypeptides ectopically expressed in
cultured cells (see below).
These results support a model in which the amino-terminus interacts with
some membrane-associated protein at the same time as the carboxy-terminus acts
with a cytoskeletal protein (Figure 1-2). One could imagine that various regulators
could interact with ERM proteins to open up the molecule and make both sites
accessible; or that the appropriate juxtaposition of the two ligands competes
successfully with the intramolecular binding reaction. Either method would limit
the interaction of ERM proteins to only a specific region of the cell. The following
section discusses the consequence of misregulated interactions.
DISSECTING ERM FUNCTIONS
Disruption of protein expression. Studying function of ERM proteins in tissue
culture cells by interference experiments is complicated by the likelihood that the
three distinct gene products have overlapping roles. Takeuchi et al. (1994)
performed a series of systematic antisense oligonucleotide experiments which
demonstrated a function of the three proteins together. Oligonucleotides designed
to interfere specifically with expression of each of the ERM transcripts dramatically
decreased the level of that protein. For cells attached to substrata, depletion of one
ERM protein, or any pair of ERM proteins, had no effect on cell morphology. Loss of
all three ERM proteins did have an effect: after 2 days, the cells rounded up and
eventually detached. The result suggests that ERM proteins have a direct or indirect
effect on maintaining adhesion to solid substrata. That the phenotype is manifested
Chapter 1
16
Introduction
only after a long time may mean that the structures containing ERM proteins in
these cells must be relatively stable. Another assay may distinguish among the
three proteins. Cells treated with the antisense oligonucleotide directed to moesin
can re-attach after trypsinization, but those depleted of ezrin or radixin, respectively,
can do so only partially or not at all. Similarly, the effect of antisense
oligonucleotides on the surface morphology of thymoma cells in suspension is
obvious when all three proteins are depleted; disrupted singly, only loss of moesin
has a detectable effect.
These experiments may signify that each of the proteins has
a different function. But the data fit equally well with a quantitative rather than
qualitative difference among the three proteins: that the severity of the phenotype
may increase with the relative abundance of the particular polypeptide.
Consequences of over-expressing ERM domains. As discussed in the previous
section, in vitro experiments detect an actin binding site in the carboxy-terminal
domain conserved in the ERM proteins, binding sites for membrane-associated
proteins at the amino-terminal domain and an interaction between the carboxy- and
the amino-terminal domains. Expression of the domains in transfected cells
partially confirm and extend these results.
At moderate levels of expression comparable to that of the endogenous
protein, the carboxy-terminal domain localizes to the same cortical structures as
transfected full-length protein as well as endogenous ERM proteins, except one - the
cleavage furrow (Algrain et al., 1993; Henry et al., 1995). Like the full-length protein,
the carboxy-terminal domain displaces endogenous ERM proteins. Unlike the full-
Chapter 1
17
Introduction
length protein, the carboxy-terminal domain localizes to cytoplasmic actin structures
in the cell such as stress fibers. At high levels, the carboxy-terminal domain induces
the formation of aberrant cortical projections filled with F-actin (Henry et al., 1995;
Martin et al., 1995). These cells also fail to undergo cytokinesis; therefore, high
levels of the carboxy-terminus can disrupt this function without localizing to the
cleavage furrow, perhaps by sequestering some other essential component (Henry et
al., 1995).
These results suggest that the C-terminal domain contains enough
information to localize it to many cortical structures, by a mechanism that competes
with endogenous ERM proteins. The properties of the transfected cells demonstrate
that the carboxy-terminal sequence specifies a binding site sufficient to localize it to
the cortex and apparently supplant the endogenous protein. These properties are
not readily explicable if the carboxy-terminus can bind only to the amino-terminus.
It is important to note that ectopic expression of other actin binding proteins
typically has quite different phenotypes than those observed for this ERM fragment.
Among many examples, gelsolin at high levels causes disruption of the major Factin filaments (Finidori et al., 1992). In contrast, transfected full-length villin
induces projections like those seen with the carboxy-terminus, but not full-length,
ERM proteins (Friederich et al., 1989).
The consequences of expressing the full-length protein or the amino-terminal
domain are quite different from one another. For example, the amino-terminal half
of the protein at moderate levels of expression does not show appropriate
localization. When significantly over-expressed, it does localize to all the structures
Chapter 1
18
Introduction
that normally contain ERM proteins - including the cleavage furrow (Henry et al.,
1995). These results suggest that there are cortical binding sites in this domain perhaps even one specific for the cleavage furrow. And the full-length protein
behaves indistinguishably from the endogenous protein, even when substantially
over-expressed. Therefore, the deleterious interactions that mediate the phenotypes
of carboxy-terminal over-expression are suppressed in cis (Henry et al., 1995; Martin
et al., 1995) - and, in one study, even in trans, when the two fragments are cooverexpressed (Martin et al., 1995). These results strengthen the notion that the
intermolecular interactions of ERM proteins with other proteins are regulated by
virtue of being mutually exclusive with intramolecular interactions.
POTENTIAL REGULATORS OF ERM INTERACTIONS
The mechanism by which ERM proteins are restricted to domains of the
cortical cytoskeleton is likely to be a crucial aspect of function. Perhaps binding is
determined by the physical proximity of two ligands, juxtaposed so that they can
compete effectively with postulated interdomain interactions. Alternatively, the
intermolecular interactions could be regulated by covalent modification, or by
binding of some small molecular ligand.
ERM proteins are phosphorylated in many cells types on tyrosine as well as
threonine and serine residues (Gould et al., 1986; Bretscher, 1989; Egerton, 1992;
Tsukita and Yonemura, 1997). After EGF stimulation of A431 cells, ERM proteins
are phosphorylated within 30 seconds and dephosphorylated after 20 minutes
(Bretscher, 1989). This fast and transient time course parallels the formation and
Chapter 1
19
Introduction
disappearances of cortical structures after EGF stimulation (Bretscher, 1989).
Recently, several experiments suggest that ERM proteins may be part of the Rho
signal transduction pathway. In LPA-stimulated cells, ERM proteins are
phosphorylated in a Rho-kinase dependent manner and ERM proteins relocalize
into cortical protrusions (Shaw et al., 1998). In vitro binding studies showed that the
phosphorylation state of ERM proteins affects their binding properties. In a gel
overlay assay, the N-terminal domain of radixin- bound to the unphosphorylated
form of the C-terminal domain of radixin, but not to the phosphorylated form
(Matsui et al., 1998). The phosphorylation state of the C-terminal domain did not
affect its ability to bind actin in a pelleting assay (Matsui et al., 1998). These
experiments suggest that the phosphorylation of ERM proteins may regulate their
intermolecular interactions by decreasing the competing interdomain interactions.
But the fact that the full-length protein is inefficiently phosphorylated and the ERM
proteins are phosphorylated and dephosphorylated within two minutes (Matsui et
al., 1998) suggests that this method of regulation may be only one of many factors
necessary to regulate ERM protein binding to their ligands.
The interaction between ERM proteins and their ligands may also be
regulated by binding to small molecules.
ERM proteins interact specifically with
phosphoinositides as assayed by pelleting with phospholipid micelles (Niggli et al.,
1995) and by their elution profile in the presence of PL through a column containing
Superose 12R gel (Hirao et al., 1996). Furthermore, the presence of phospholipids
affects the binding of several ERM proteins to their ligands in vitro. At
physiological conditions, CD44 binds only weakly to ezrin, radixin and moesin. The
Chapter 1
20
Introduction
presence of phosphatidyl inositol (PI), phosphatidyl inositol-4-phosphate (PIP) and
phosphatidyl inositol 4, 5 bisphosphate (PIP 2) enhanced the binding of CD44 and
ERM proteins (Hirao et al., 1996). Phosphoinositides also enhanced ezrin binding to
ICAM-2 and induced the interaction between ezrin and ICAM-1 (Heiska et al., 1998).
That phospholipids have a role in regulating protein-protein interactions has been
previously documented with other cytoskeletal proteins, but the exact mechanism
by which phospholipids work is not yet known. Understanding how ERM proteins
are recruited to specific regions of the cell will increase our understanding of their
role in the cell.
IDENTIFICATION OF MERLIN
A detailed function for any member of the ERM family still remains
unknown. However the identification of merlin, raises the possibility that aspects
of ERM function may be involved in cell growth control. Mutations in merlin are
implicated for neurofibromatosis type 2, a human disease causing the formation of
multiple nervous system tumors, especially schwannomas. A first crucial question
is whether or not merlin is indeed a member of the ERM family. That is, does
merlin behave like an ERM protein in those assays that are available? Transfected
merlin in COS cells gives a punctate pattern, and general staining of the membrane
(den Bakker et al., 1995). More apposite is the identification of endogenous merlin
in both fibroblast and meningioma cells in motile domains of the cortex (Gonzalez
et al., 1996). Despite this similarity to ERM proteins, the anti-merlin staining in
meningioma cells is clearly distinguishable from that of anti-ezrin or anti-moesin.
Chapter 1
21
Introduction
In Drosophila, the merlin homolog is in quite different localizations than is the
moesin homolog, providing a system for genetic tests of functions of both proteins
(McCartney and Fehon, 1996). These striking differences between merlin and ERM
proteins may mean that properties of merlin overlap but are not coincident with
those of other ERM proteins. Obviously, there are further assays that could assess
parallels of properties: does ectopic expression of merlin domains have phenotypes?
do those domains bind to one another, or to other ERM ligands? Finally, the
intriguing finding that transfection of merlin can suppress some phenotypes of
transformed cells (Tikoo et al., 1994; Sherman et al., 1997) suggests an assay for ERM
function as well, and the possibility of identifying ERM and merlin sequences that
are essential for certain properties.
ERM PROTEIN HOMOLOGS IN OTHER ORGANISMS
The possibilities for understanding the relationship between ERM structure
and function are enhanced by the identification of homologs in several invertebrate
species. D. melanogaster expresses both moesin and merlin proteins (Edwards, 1994;
McCartney and Fehon, 1996). These two proteins are differentially distributed in the
developing fly; moesin distributes to the plasma membrane, while merlin was seen
in punctate structures at the plasma membrane and cytoplasm. Analysis of mutant
phenotypes suggested that merlin is involved in endocytic processes by studies in
cultured cells. Sea urchins express a moesin homolog and both localization data
and drug studies suggested that it interacts with actin (Bachman and McClay, 1995).
There is also evidence for homologs in Schistosorna japonicum (Kurtis et al., 1997)
Chapter 1
22
Introduction
and in C. elegans (Genbank, accession: U10414). An ERM-homolog, called EM10,
was identified in the tapeworm Echinococcus multilocularis (Frosch et al., 1991).
EM10 is a highly immunogenic protein, and is the dominant antigen associated
with the parasitic disease alveolar echinococcosis. The sequence of EM10 is 46.9%
identical to that of human ezrin. As for other homologs, the EM10 homology is
more pronounced in the amino-terminal half of the protein. To learn more about
EM10 and ERM function, we have compared the properties of these proteins when
they are ectopically expressed in cultured cells. In particular, following on studies of
ERM proteins, we analyzed the localization and phenotypic consequences of
expressing full-length EM10 and its amino- and carboxy-terminal halves. The
results demonstrate that EM10 sequences behave similarly but not identically to
those of ERM-proteins (Hubert, Cordero et al., in press). By studying the role of
ERM protein homologs in other organisms, which may be genetically tractable, we
may uncover the role for ERM proteins in animal cells.
OTHER CYTOSKELETAL PROTEINS USE SIMILAR MODES FOR REGULATING
THEIR INTERACTIONS WITH LIGAND
The generally-accepted model of regulation for ERM proteins (Figure 1-2) in
which a reciprocal relationship between interdomain binding and intermolecular
binding is not novel for ERM proteins. A similar model was proposed earlier for
vinculin and its binding to talin and actin (Johnson and Craig, 1995). In that case,
the interaction between the carboxy-terminal domain and F-actin is blocked by the
presence of the amino-terminal domain polypeptide. The actin binding activity is
Chapter 1
23
Introduction
not detected with full-length vinculin. This same binding property is detected
between vinculin and talin - the vinculin head binds to talin whereas the fulllength protein does not. An interdomain interaction has been proposed for this
effect (Johnson and Craig, 1995). This interdomain interaction identified for
vinculin is disrupted by phospholipids which allows vinculin to bind to talin
(Johnson and Craig, 1994; Gilmore and Burridge, 1996).
Another cytoskeletal protein which uses these same method for regulation is
band 4.1 and its interaction with glycophorin (Anderson and Marchesi, 1985). The
binding between these two proteins increases in the presence of a
polyphosphoinositide cofactor. Thus the regulation of intermolecular interactions
by phospholipids occurs with other cytoskeletal proteins and their ligands.
Chapter 1
24
Introduction
Figure 1-1: ERM family of proteins
Radixin is a member of the ERM (ezrin, radixin, moesin) family of proteins.
The amino-terminal domains are highly homologous. The ERM family is
part of the Band 4.1 superfamily which includes the proteins Band 4.1, talin, a
homologous protein found in echinococcus - EM10 and the product of a
tumor suppressor gene - merlin.
Chapter 1
25
Introduction
.. ...............
The Band 4.1 Superfamily
ERM Family
N
C
Ezrin
Radixin
84%
Moesin
U
83%/
EM-10
58%
Talin
Band 4.1
35%
Merlin
62%
............................
Figure 1-2: Current model of ERM regulation
The current model in the field is that the full-length ERM protein forms an
interdomain interaction which blocks the binding sites present on the
individual domains. It isn't until a modification occurs that the ERM protein
changes to the "open" conformation and makes the binding sites available to
their ligands.
Chapter 1
27
Introduction
-
Cortical
Proteins
-Actin
ERM proteins
.........
........
......................................
....
...
THESIS OVERVIEW
In my thesis, I use two proteins, NHE-RF and layilin, to study the regulation
of ERM protein binding to its ligands. NHE-RF interacts with ezrin, radixin and
moesin in affinity chromatography binding assays (Murthy et al., 1998; Reczek and
Bretscher, 1998). NHE-RF also binds to a protein homologous to ERM proteins,
merlin which is the product of the tumor suppresser gene, NF2 (Murthy et al., 1998).
In HeLa cells, NHE-RF co-aligns with moesin at the microvilli, filopodia and
ruffling edges (Murthy et al., 1998). NHE-RF was originally identified as a
regulatory co-factor necessary for protein kinase A inhibition of a Na+/H+
exchanger (NHE) (Weinman et al., 1995). A second ligand I study for my thesis is
layilin. Layilin is a novel transmembrane protein which binds talin and has
homology to C-type lectins. It localizes to ruffling edges in NIL8 and CHO cells
(Borowsky and Hynes, 1998).
My work shows that layilin and NHE-RF are direct ligands for radixin and
their binding sites map to the amino-terminal domain of radixin. Ligand binding to
the amino-terminal domain is decreased by the presence of the C-terminal domain
of radixin in cis as well as in trans. The interaction between radixin and its ligands
can be enhanced by phosphatidyl-inositol 4 phosphate (PIP). We show that PIP
changes the conformation of radixin at the amino-terminal domain and decreases
the amino- and carboxy- interdomain interaction. By releasing this interdomain
interaction, phosphoinositides can regulate the interaction between radixin and its
ligands.
Chapter 1
29
Introduction
BIBLIOGRAPHY
Algrain, M., Turunen, 0., Vaheri, A., Louvard, D., and Arpin, M. (1993). Ezrin
contains cytoskeleton and membrane binding domains accounting for its proposed
role as a membran-cytoskeletal linker. J. Cell Biol. 120, 129-139.
Anderson, R. A., and Marchesi, V. T. (1985). Regulation of the association of
membrane skeletal protein 4.1 with glycophorin by a polyphosphoinositide. Nature
318, 295-298.
Bachman, E. S., and McClay, D. R. (1995). Characterization of moesin in the sea
urchin Lytechinus variegatus: redistribution to the plasma membrane following
fertilization is inhibited by cytochalasin B. J Cell Sci 108, 161-71.
Berryman, M., Zsofia Franck and Anthony Bretscher (1993). Ezrin is concentrated in
the apical microvilli of a wide variety of epithelial cells whereas moesin is found
primarily in endothelial cells. Journal of Cell Science 105, 1025-1043.
Birgbauer, E., and Solomon, F. (1989). A marginal band-associated protein has
properties of both microtubule- and microfilament-associated proteins. Journal of
Cell Biology 109, 1609-1620.
Birgbauer, E. C. (1991). Cytoskeletal interactions of ezrin in differentiated cells:
M.I.T.).
Borowsky, M. L., and Hynes, R. 0. (1998). Layilin, a novel talin-binding
transmembrane protein homologous with C- type lectins, is localized in membrane
ruffles. J Cell Biol 143, 429-42.
Bretscher, A. (1983). Purification of an 80,000-dalton protein that is a component of
the isolated microvillus cytoskeleton, and its localization in nonmuscle cells. J. of
Cell Biology 97, 425-532.
Bretscher, A. (1986). Purification of the intestinal microvillus cytoskeletal proteins
villin, fimbrin, and ezrin. Meth. Enzymology 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. Journal of Cell Biology 108, 921-930.
den Bakker, M., Riegman, P., Hekman, R., Boersma, W., Janssen, P., van der Kwast,
T., and Zwarthoff, E. (1995). The product of the NF2 tumour suppressor gene
localizes near the plasma membrane and is highly expressed in muscle cells.
Oncogene 10, 756-763.
Chapter 1
30
Introduction
Edwards, K. A., Ruth A. Montague, Scott Shepard, Bruce A. Edgar, Raymond L.
Erikson and Daniel 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.
Egerton, M., Wilson H. Burgess, Douglas Chen, Brian J. Druker, Anthony Bretscher
and Lawrence E. Samelson (1992). Identification of Ezrin as an 81-kDa TyrosinePhosphorylated Protein in T Cells. Journal of Immunology 149, 1847-1852.
Finidori, J., Friederich, E., Kwiatkowki, D., and Louvard, D. (1992). In vivo analysis
of functional domains from villin and gelsolin. Jour. Cell Biol. 116, 1145-1155.
Franck, Z., Ronald Gary and Anthony Bretscher (1993). Moesin, like ezrin,
colocalizes with actin in the cortical cytoskeleton in cultured cells, but its expression
is more variable. Journal of Cell Science 105, 219-231.
Friederich, Huet, Arpin, and Louvard (1989). Villin induces microvilli growth and
actin redistribution in transfected fibroblasts. cell 59, 461-475.
Frosch, P. M., Frosch, M., Pfister, T., Schaad, V., and Bitter-Suermann, D. (1991).
Cloning and characterisation of an immunodominant major surface antigen of
Echinococcus multilocularis. Mol Biochem Parasitol 48, 121-130.
Fukata, Y., Kimura, K., Oshiro, N., Saya, H., Matsuura, Y., and Kaibuchi, K. (1998).
Association of the myosin-binding subunit of myosin phosphatase and moesin:
dual regulation of moesin phosphorylation by Rho-associated kinase and myosin
phosphatase. J Cell Biol 141, 409-18.
Gary, R., and Bretscher, A. (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 Bretscher, A. (1995). Ezrin self-association involves binding of an Nterminal 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 Burridge, K. (1996). Regulation of vinculin binding to talin and
actin by phosphatidyl-inositiol-4-5- bisphosphate. Nature 381, 531-535.
Gonzalez, -. A., C., Xu, L., Pinney, D., Beauchamp, R., Hobbs, W., Gusella, J., and
Ramesh, V. (1996). The merlin tumor suppressor localizes preferentially in
membrane ruffles. Oncogene 13, 1239-1247.
Goslin, K., Birgbauer, E., Banker, G., and Solomon, F. (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.
Chapter 1
31
Introduction
Gould, K. L., Bretcher, A., Cooper, J. A., and Hunter, T. (1986). The protein-tyrosine
kinase substrate, p81, is homologous to a chicken microvillar core protein. JCB 102,
660-669.
Heiska, L., Alfthan, K., Gronholm, M., Vilja, P., Vaheri, A., and Carpen, 0. (1998).
Association of ezrin with intercellular adhesion molecule-1 and -2 (ICAM-1 and
ICAM-2). Regulation by phosphatidylinositol 4, 5- bisphosphate. J Biol Chem 273,
21893-900.
Heiska, L., Kantor, C., Parr, T., Critchley, D. R., Vilja, P., Gahmberg, C. G., and
Carpen, 0. (1996). Binding of the cytoplasmic domain of intercellular adhesion
molecule-2 (ICAM-2) to alpha-actinin. J Biol Chem 271, 26214-9.
Henry, M., Gonzalez-Agosti, C., and Solomon, F. (1995). Molecular dissection of
radixin: Distinct and interdependent functions of the amino- and carboxy-terminal
domains. Journal of Cell Biology 129, 1007-1022.
Higashitsuji, H., Arii, S., Furutani, M., Imamura, M., Kaneko, Y., Takenawa, J.,
Nakayama, H., and Fujita, J. (1995). Enhanced expression of multiple protein
tyrosine phosphatases in the
regenerating mouse liver: isolation of PTP-RL10, a novel cytoplasmic-type
phosphatase with sequence homology to cytoskeletal protein 4.1. Oncogene 10, 407414.
Hirao, M., Sato, H., Kondo, T., Yonemura, S., Monden, M., Sasaki, T., Takai, Y.,
Tsukita, S., and Stukit, S. (1996). Regulation mechanism of ERM
(ezrin/radixin/moesin) protein/plasma membrane association: possible
involvement of phosphatidylinositol turnover and Rho-dependent signaling
pathway. Jour. Cell Biol. 135, 37-51.
Johnson, R. P., and Craig, S. W. (1994). An intramolecular association between the
head and tail domains of vinculin modulates talin binding. Journal of Biological
Chemistry 269, 12611-12619.
Johnson, R. P., and Craig, S. W. (1995). F-actin binding site masked by the
intramolecular association of vinculin head and tail domains. Nature 373, 261-264.
Kurtis, J. D., Ramirez, B. L., Wiest, P. M., Dong, K. L., El-Meanawy, A., Petzke, M. M.,
Johnson, J. H., Edmison, J., Maier, R. A., Jr., and Olds, G. R. (1997). Identification and
molecular cloning of a 67-kilodalton protein in Schistosoma japonicum
homologous to a family of actin-binding proteins. Infect Immun 65, 344-7.
L'Abbe, D., Banville, D., Tong, Y., Stocco, R., Masson, S., Ma, S., Fantus, G., and
Shen, S.-H. (1994). Identification of a novel protein tyrosine phosphatase with
Chapter 1
32
Introduction
sequence homology to the cytoskeleltal proteins of the band 4.1 family. FEBS Letters
356, 351-356.
Lankes, W., Griesmacher, A., Grunwald, J., Schwartz-Albietz, R., and Keller, R.
(1988). A heparin-binding protein involved in inhibition of smooth-muscle cell
proliferation. Biochem. J. 251, 831-842.
Magendantz, M., Henry, M., Lander, A., and Solomon, F. (1995). Inter-domain
interactions of radixin in vitro. Jour. Biol. Chem. 270, 25324-25327.
Martin, M., Andreoli, C., Sahuquet, A., Montcourrier, P., Algrain, M., and Mangeat,
P. (1995). Ezrin NH2-terminal domain inhibits the cell extension activity of the
COOH-terminal domain. Jour. Cell Biol. 128, 1081-1093.
Matsui, T., Maeda, M., Doi, Y., Yonemura, S., Amano, M., Kaibuchi, K., and Tsukita,
S. (1998). Rho-kinase phosphorylates COOH-terminal threonines of
ezrin/radixin/moesin (ERM) proteins and regulates their head-to-tail association. J
Cell Biol 140, 647-57.
McCartney, B., and Fehon, R. (1996). Distinct cellular and subcellular patterns of
expression imply distinct functions for the Drosophila homologues of moesin and
the neurofibromatosis 2 tumor supressor, merlin. Jour. Cell Biol. 133, 843-852.
Murthy, A., Gonzalez-Agosti, C., Cordero, E., Pinney, D., Candia, C., Solomon, F.,
Gusella, J., and Ramesh, V. (1998). NHE-RF, a regulatory cofactor for Na(+)-H+
exchange, is a common interactor for merlin and ERM (MERM) proteins. J Biol
Chem 273, 1273-6.
Niggli, V., Andreoli, C., Roy, C., and Mangeat, P. (1995). Identification of a
phosphatidylinositiol-4,5-bisphosphate-binding domainin the N-terminal region of
ezrin. FEBS Letters 376, 172-176.
Pakkanen, R. (1988). Immunofluorescent and immunochemical evidence for the
expression of cytovillin in the microvilli of a wide range of cultured human cells. J.
Cell. Biochem. 38, 65-75.
Pakkanen, R., Hedman, K., Turunen, 0., Wahlstrom, T., and Vaheri, A. (1987).
Microvillus-specific Mr 75,000 plasma membrane protein of human
choriocarcinoma cells. J. of Histochem. Cytochem. 35, 809-816.
Pestonjamasp, K., Amieva, M., Strassel, C., Nauseef, W., Furthmayr, H., and Luna,
E. (1995). Moesin, ezrin and p205 are actin-binding proteins associated with
neutrophil plasma membranes. Mol. Biol. Cell 6, 247-259.
Chapter 1
33
Introduction
Reczek, D., Berryman, M., and Bretscher, A. (1997). Identification of EBP50: A PDZcontaining phosphoprotein that associates with members of the ezrin-radixinmoesin family. J Cell Biol 139, 169-79.
Reczek, D., and Bretscher, A. (1998). The carboxyl-terminal region of EBP50 binds to a
site in the amino- terminal domain of ezrin that is masked in the dormant
molecule. J Biol Chem 273, 18452-8.
Roy, C., Martin, M., and Mangeat, P. (1997). A dual involvement of the aminoterminal domain of ezrin in F- and G- actin binding. J Biol Chem 272, 20088-95.
Sato, N., Yonemura, S., Obinata, T., Tsukita, S., and Tsukita, S. (1991). Radixin, a
barbed-end-capping actin-modulating protein, is concnetrated at the cleavage furrow
during cytokinesis. Journal of Cell Biology 113, 321-330.
Sawada, M., Ogata, M., Fujino, Y., and Hamaoka, T. (1994). cDNA cloning of a novel
protein tyrosine phosphatase with homology to cytoskeletal protein 4.1 and its
expression in T-lineage cells. BBRC 203, 479-484.
Serrador, J. M., Nieto, M., Alonso-Lebrero, J. L., del Pozo, M. A., Calvo, J.,
Furthmayr, H., Schwartz-Albiez, R., Lozano, F., Gonzalez-Amaro, R., SanchezMateos, P., and Sanchez-Madrid, F. (1998). CD43 interacts with moesin and ezrin and
regulates its redistribution to the uropods of T lymphocytes at the cell-cell contacts.
Blood 91, 4632-44.
Shaw, R. J., Henry, M., Solomon, F., and Jacks, T. (1998). RhoA-dependent
phosphorylation and relocalization of ERM proteins into apical membrane/actin
protrusions in fibroblasts. Mol Biol Cell 9, 403-19.
Sherman, L., Xu, H. M., Geist, R. T., Saporito-Irwin, S., Howells, N., Ponta, H.,
Herrlich, P., and Gutmann, D. H. (1997). Interdomain binding mediates tumor
growth suppression by the NF2 gene product. Oncogene 15, 2505-9.
Shuster, C., and Herman, I. (1995). Indirect association of ezrin with F-actin: isoform
specificity and calcium sensitivity. Jour. Cell. Biol. 128, 837-848.
Tikoo, A., Varga, M., Ramesh, V., Gusella, J., and Maruta, H. (1994). An anti-ras
function of neurofibromatosis type 2 gene product (NF2/Merlin). Jour. Biol. Chem.
269, 23387-23390.
Tsao, H., Aletta, J. M., and Greene, L. A. (1990). Nerve growth factor and fibroblast
growth factor selectively activate a protein kinase that phosphorylates high
molecular weight microtubule-associated proteins. Detection, partial purification,
and characterization in PC12 cells. J Biol Chem 265, 15471-80.
Chapter 1
34
Introduction
Tsukita, S., Oishi, K., Sato, N., Sagara, J., Kawai, A., and Tsukita, S. (1994). ERM
family members as molecular linkers between the cell surface glycoprotein CD44
and actin-based cytoskeletons. Journal of Cell Biology 126, 391-401.
Tsukita, S., and Yonemura, S. (1997). ERM proteins: head-to-tail regulation of actinplasma membrane interaction. Trends Biochem Sci 22, 53-8.
Tucker, R. P., Garner, C. C., and Matus, A. (1989). In Situ localization of microtubuleassociated protein mRNA in the developing and adult rat brain. Neuron 2, 12451256.
Turunen, 0., Wahlstrom, T., and Vaheri, A. (1994). Ezrin has a COOH-terminal
actin-binding site that is conserved in the ezrin protein family. Journal of Cell
Biology 126, 1445-1453.
Weinman, E. J., Steplock, D., Wang, Y., and Shenolikar, S. (1995). Characterization of
a protein cofactor that mediates protein kinase A regulation of the renal brush
border membrane Na(+)-H+ exchanger. J Clin Invest 95, 2143-9.
Yang, Q., and Tonks, N. K. (1991). Isolation of a cDNA clone encoding a human
protein-tyrosine phosphatase with homology to the cytoskeletal-associated proteins
band 4.1, ezrin and talin. Proc. Natl. Acad. Sci. USA 88, 5949-5953.
Yao, X., Chaponnier, C., Gabbiani, G., and Forte, J. G. (1995). Polarized distribution of
actin isoforms in gastric parietal cells. Mol Biol Cell 6, 541-57.
Yao, X., Cheng, L., and Forte, J. (1996). Biochemical characterization of ezrin-actin
interaction. Jour. Biol. Chem. 271, 7224-7229.
Yonemura, S., Hirao, M., Doi, Y., Takahashi, N., Kondo, T., and Tsukita, S. (1998).
Ezrin/radixin/moesin (ERM) proteins bind to a positively charged amino acid
cluster in the juxta-membrane cytoplasmic domain of CD44, CD43, and ICAM-2. J
Cell Biol 140, 885-95.
Chapter 1
35
Introduction
CHAPTER 2:
INTRA- AND INTER-DOMAIN INTERACTIONS
The interaction between the amino and carboxy-terminal domains of ERM
proteins is well documented. By gel overlay assays, purified ezrin and moesin bind
to one another, and the sequence involved in the association was mapped to the
two domains in the extreme domains of the molecule (Gary and Bretscher, 1995).
Furthermore, for many assays, the activities which are seen with fragments of ERM
proteins are diminished when the assays are done with full-length proteins. In
vitro experiments show that several ligands such as actin, CD44, band 4.1 and NHERF/EBP50 bind to a separate domain of ERM proteins more efficiently than to the
full-length proteins suggesting an inhibitory effect of the complementary domain
(Tsukita et al., 1994; Magendantz et al., 1995; Pestonjamasp et al., 1995; Murthy et al.,
1998; Reczek and Bretscher, 1998).
This difference between full-length and fragment
activities is also seen in in vivo experiments. Unlike the full-length ERM protein,
the exogenously expressed carboxy-terminus localizes to stress fibers (Algrain et al.,
1993), and upon overexpression induces the formation of aberrant actin-rich cortical
extensions in animal cells (Algrain et al., 1993; Henry et al., 1995). These data could
be explained by a direct interaction between the amino- and carboxy-terminal
domains. In fact an interaction between the domains is detectable by affinity coelectrophoresis; the data are consistent with a 1:1 complex of the amino- and
carboxy-terminal halves of the protein with a Kd of ~4x1O-8 M (Magendantz et al.,
1995). In the full-length proteins, this interdomain interaction could mask the
binding sites and inhibit the binding activities seen with the domains of the protein.
Chapter 2
37
Domain interactions
Since the interdomain interaction may inhibit the binding between ERM
proteins and their ligands, the regulation of this interdomain interaction can
modulate the interaction between radixin and its ligands.
The effect of
phosphorylating the carboxy-terminal domain of radixin was shown to decrease its
ability to bind to the amino-terminal domain of radixin, but it did not affect its
ability to bind to actin (Matsui et al., 1998). In this case the release of the
interdomain interaction by a phosphorylation event may increase the interaction
between radixin and actin. Another possible method of regulating these
interactions is by binding to small molecule regulators. The presence of
phospholipids affects the binding of several ERM proteins to their ligands in vitro.
CD44 does not bind to full-length ezrin at physiological conditions except in the
presence of PIP2 (Hirao et al., 1996). PIP2 also increases the affinity between ezrin
and ICAM-1 and ICAM-2 (Heiska et al., 1998). Although binding sites for specific
phospholipids have been identified at the amino-terminal domain of ezrin (Niggli
et al., 1995), the exact mechanism for how these phospholipids affect the proteinprotein interactions is not known.
In this chapter, I describe the interdomain interactions between the amino
and carboxy-terminal domains of radixin. I show that this interdomain interaction
can be disrupted by the presence of PIP. The interdomain interaction and its
modulation by phospholipids may represent a regulated change in the tertiary
structure of the protein. A specific conformation of radixin can be captured by crosslinkers. Cross-linking reagents capture an intramolecular interaction at the aminoterminal domain of radixin and this intramolecular interaction is disrupted by the
Chapter 2
38
Domain interactions
presence of specific phospholipids. PIP changes the conformation of the aminoterminal domain and decreases the interdomain interaction. As I will describe later,
the loss of the interdomain interaction increases ligand binding to full-length
radixin.
Chapter 2
39
Domain interactions
MATERIALS AND METHODS
Antibodies
To detect radixin proteins by immunoblotting, we used polyclonal antibodies
specific for the amino-terminal domain of ERM proteins (220) or a carboxy-terminal
sequence specific for radixin (457-3) (Winckler, 1994).
Recombinant Proteins
His6-tagged versions of murine radixin constructs were expressed and
purified as previously described for chicken radixin (Magendantz et al., 1995). The
His6-tagged radixin constructs we used were His6-tagged full-length radixin
(HisRadFL, residues 1-583, strain EC74), the amino-terminal domain of radixin
(HisRadN, residues 1-318, strain EC75), and the carboxy-terminal domain of radixin
(HisRadC, residues 319-583, strain EC76). To purify the protein, a saturated culture
was diluted 1:10 in 500 ml of LB with 100gg/ml ampicillin and grown for 2 hours at
37*C. We induced protein expression by adding isopropyl-p-Dthiogalactopyranoside (IPTG) to a final concentration of 1mM and grew for an
additional 3-4 hours. The cells were collected by centrifugation in a Beckman JA-14
rotor at 5000rpm for 20 minutes. We froze the cell pellets in a dry ice/ethanol
solution and stored them at -85'C and quickly thawed the pellets in 10 ml lysis
buffer (50mM NaH 2PO 4 /NaHPO 4, pH 8.0, 300mM NaCl; 20mM imidazole, pH 8.0;
1mM pefabloc, 1mM leupeptin, 1mM pepstatin, 0.007 TIU/ml aprotinin), lysed the
Chapter 2
40
Domain interactions
cells by sonication, and collected the high speed supernatant by centrifugation at
15000 rpm in a Beckman JA-20 rotor for 20 minutes. The lysate was supplemented
with 10mM f-mercaptoethanol then incubated with 3.0 ml Ni-NTA Sepharose CL6B resin (Qiagen) which had been pre-equilibrated with Buffer I (300mM NaCl,
50mM NaH 2PO 4/NaHPO 4, 20mM imidazole, pH 8.0) for 20 minutes at 4'C. The
beads were washed twice with Buffer V (40mM imidazole buffer, 300mM NaCl,
50mM NaH 2PO 4/NaHPO 4, pH 8.0, 10% glycerol) and once with Buffer VI (40mM
imidazole buffer, 300mM NaCl, 50mM NaH 2PO 4/NaHPO 4, pH 8.0). We loaded the
beads into a 5ml column and eluted the protein fractions with Buffer VII (250mp.
imidazole, 300mM NaCl, 50mM NaPhosphate pH8.0). The fractions containing the
highest concentrations of proteins were pooled.
All purified proteins were dialyzed against PBS using rotating Pierce Slide-alyzers for four hours changing the buffer every hour. We added a protease inhibitor
cocktail (1mM pefabloc, 1mM leupeptin, 1mM pepstatin, 0.007 TIU/ml aprotinin
and 1mM phenylmethylsulfonyl fluoride). The protein preparations were separated
into small aliquots, quickly frozen in liquid nitrogen and stored at -40'C.
Immediately before each experiment, protein aliquots were quickly thawed and held
on ice until used.
We created a GST tagged version of the carboxy-terminal domain of radixin
(GSTRadC) using a previously described plasmid containing the carboxy-terminal
domain of radixin behind an inducible promoter - pUHD-HAC-RADC (Henry et al.,
1995). The plasmid pUHD-HAC-RADC, was digested at its XhoI sites (New England
Biolabs Inc. Beverly, MA) which were engineered to flank the coding sequence of
Chapter 2
41
Domain interactions
the protein (Henry et al., 1995). The radixin insert was separated from the pUHD
vector on an agarose gel and purified using a Qiaex gel extraction kit (Qiagen,
Chatsworth, CA). This insert was ligated into a unique XhoI site within the
polylinker of pGEX-5X-1 (Pharmacia) which keeps the coding sequence in frame.
Escherichia coli DH5aF'IQ containing the correct insert orientation was identified by
diagnostic restriction enzyme digests (EC88).
GST fusions of the carboxy-terminal domain of radixin (GSTRadC, residues
319-583) were expressed in Escherichia coli DH5cL and purified using glutathione
beads. A saturated culture was diluted 1:30 into 450ml LB medium containing
50 jg/ml ampicillin. We cultured the cells at 37'C until an OD600 reading of a
sample of the culture reached between 0.45-0.60. Protein expression was induced by
incubating the cells with 0.05mM isopropyl-1--D-thiogalactopyranoside (IPTG) for
two hours. The cells were harvested by centrifugation 5000 rpm in a Beckman JA-14
rotor at 4'C for 20 min. We lysed the cells by freeze/thawing and removed the
insoluble fraction by centrifugation (16000 rpm in a Beckman JA-20 rotor at 4'C).
The high speed supernatant was incubated with glutathione Sepharose 4B beads
(Pharmacia) with 0.2% Tween-20 for 1 hour at 4'C. The beads were washed with
PBS/0.2% Tween-20 and eluted the proteins with glutathione elution buffer (5mM
glutathione, 50mM Tris pH 8.0, 0.2% Tween-20). We dialyzed and stored these
proteins as stated for the His 6 tagged radixin proteins.
Chapter 2
42
Domain interactions
Phospholipids
The phospholipids used in this study were: PIP2 (Phosphatidylinositol 4,5bisphosphate disodium salt); PS (O-(3-sn-Phosphatidyl)-L-serine sodium salt); IP3 (Dmyo-inositol-1,4,5-triphosphate penta-potassium salt); PIP (Phosphatidylinositol 4phosphate sodium salt); OAG (1-oleolyl-2-acetyl-rac-glycerol); and PC (Phosphatidyl
choline). All phospholipids were purchased from Sigma. The phospholipids were
resuspended in water to a final concentration of 1mg/ml, then sonicated for 5
minutes; the stock solutions were frozen in small aliquots using liquid nitrogen and
stored at -40'C. The solutions of PC and OAG did not completely clarify. The
phospholipids were sonicated for 30s - 1min immediately before use in the crosslinking assays.
Cross-linking reagents
The cross-linking reagents and abbreviations used in this study were: BS 3
(Bis[sulfosuccinimidyl]suberate), DTSSP (3,3'-Dithiobis[sulfosuccinimidylpropionate]); DSP (Dithiobis[succinimidyl-propionatel); BSOCOES (Bis[2(succinimidoxycarbonyloxy)ethylsulfone); EGS (ethylene glycolbis[succinimidylsuccinate]); DST (Disuccinimidyl tartarate); DTBP (Dimethyl 3,3'dithiobispropionimidate 2HCl); and EDC (1-Ethyl-3[3-dimethylaminopropyl]carbodiimide Hydrochloride). All were purchased from Pierce. We prepared fresh
10x stock solutions in DMSO immediately before use.
Chapter 2
43
Domain interactions
Phospholipid Pelleting Assays
Phospholipid stock aliquots were thawed and sonicated for 30 seconds - 1
minute immediately before use. Beckman T-100 tubes were blocked with 0.3mg/ml
BSA for 30 minutes at room temperature. The proteins were pre-airfuged in a
Beckman Air-driven ultracentrifuge for 20 minutes at 23psi to remove any
pelletable aggregates. We incubated the precleared proteins with phospholipid and
added PBS to 150g1. The mixture was airfuged for 20 minutes at 23 psi in a Beckman
airfuge. We collected 50 p1 of the supernatant and resuspended the pellet in the
same volume of 1x GSD. The samples were run on an SDS-PAGE and the proteins
were detected by immunoblotting.
Binding Assays
To detect an interaction between the amino- and carboxy-terminal domains,
GSTRadC was bound to the matrix. 40gl of glutathione-agarose beads (Pharmacia)
were swelled in 1ml PBS for 1 hour at 4*C and washed with PBS. We incubated
150pl of GSTRadC (0.2nmoles) with the glutathione beads agarose beads for 20
minutes. The beads were washed with 1 ml PBS and incubated 150gl HisRadN for
20 minutes at 4*C turning end over end. The beads were again washed with PBS. If
phospholipids were used, they were added at the same times as HisRadN and are
added in the wash buffers as well. The bound proteins were eluted with 150
glutathione elution buffer (5mM glutathione, 50mM Tris, pH8.0, 0.2% Tween-20) in
a batchwise manner and the eluant was boiled in GSD. The samples were then
Chapter 2
44
Domain interactions
analyzed by SDS-PAGE and the proteins were detected by coomassie or
immunoblotting.
Cross-linking Assay
In the cross-linking assays, purified radixin was pre-incubated for 15 minutes
at 25'C in PBS. In some experiments, a denaturant is added during this preincubation step prior to the addition of cross-linking reagents but the final reaction
volume remained constant. For experiments requiring phospholipids, the
appropriate phospholipids were present throughout the reaction including the preincubation step. The cross-linker was added to a final concentration of 50-fold the
protein concentration and typically was allowed to react for 10 min at 25'C. The
reaction was stopped by adding a quenching solution which contained excess amines
(125 mM Tris, 125 mM glycine, pH 8.0) for an additional 15 min at 25'C. The
samples were then run on a 7.5% SDS-PAGE (HisRadFL) or 10% SDS-PAGE
(HisRadN and HisRadC) and analyzed by immunoblotting as described below.
Western Blot Analysis
For immunoblotting, we separated protein samples on a 7.5% or 10%
polyacrylamide gel with a 5% stacker according to the method of Laemlli (1970).
Proteins were electrophoretically transferred to 0.2gm nitrocellulose filters
(Schleicher and Schuell; Keene, NH) essentially as described by Tobin et al (1979).
Total protein transferred to the nitrocellulose filter was detected by Ponceau S
staining (0.2% Ponceau S (Sigma) in 3% TCA). For all the antibodies used, we
Chapter 2
45
Domain interactions
blocked the blots with TBST (50 mM Tris pH 8.0, 150 mM NaCl, Tween-20 0.1 %) for
1-2 hours at room temperature and probed the blots overnight with the primary
antibody. To detect HisRadC, HisRadN and NHE-RF, we used 457-3 (1:500), 220
(1:1000) and NP1 (1:1000) diluted into TBST, respectively. We washed the blots
extensively with TBST and detected the primary antibodies with 11 2 5-labeled Protein
A (DuPont - New England Nuclear; Boston, MA) (1:1000) diluted in TBST. After
extensive washing with TBST, the signal was detected using xray film at -70'C.
Chapter 2
46
Domain interactions
RESULTS
Phospholipids regulate an interdomain interaction of radixin
We previously identified an interdomain interaction between the amino-and
carboxy-terminal domains of radixin by affinity co-electrophoresis (ACE) and
determined the Kd of this interaction to be about 50 nM (Magendantz et al., 1995).
To determine if these protein fragments also interact by affinity chromatography I
used domains of radixin which were differentially tagged. We made a GST-tagged
carboxy-terminal domain of radixin (GSTRadC) and a His6-tagged amino-terminal
domain of radixin (HisRadN). A constant amount of GSTRadC (0.2nmoles) was
bound to glutathione beads and incubated with increasing amounts of HisRadN.
We washed the beads to remove nonspecific protein interactions and eluted the
specifically bound proteins with excess glutathione. HisRadN bound specifically to
beads containing GSTRadC (Figure 2-1A); it did not bind to beads containing GST
(data not shown). We estimated the amount of HisRadN binding to GSTRadC by
quantifying the intensity of the autorad bands using an IS1000 Digital imaging
system. The binding of HisRadN to GSTRadC appears to saturate at a ratio of about
1mole of HisRadN per 2 moles of GSTRadC (Figure 2-1B). By Coomassie, GSTRadC
is present as a doublet which may be an indication of protein degradation. If only a
portion of the protein preparation was capable of binding to HisRadN, this would
account for the 1:2 binding ratio of HisRadN to GSTRadC. In fact, in a converse
binding assay, with HisRadN bound to Ni-NTA beads and GSTRadC in solution, the
binding of the higher molecular weight band of GSTRadC was enriched as compared
Chapter 2
47
Domain interactions
to the lower molecular weight band. Suggesting that the lower molecular weight
band is actually a degradation product and may not have the binding properties of
the full-length carboxy-terminal domain.
The regulation of the interdomain interaction within ERM proteins may be
one way of modulating the binding between ERM proteins and their ligands. As
mentioned in the introduction, binding to small molecule regulators, such as
phospholipids, may regulate ERM protein binding to their ligands (Hirao et al., 1996;
Heiska et al., 1998). I have confirmed a phosphatidyl inositol 4-phosphate, PIP,
binding site present on the amino-terminal domain of radixin using pelleting
assays. I incubated purified His-tagged full-length radixin (HisRadFL), carboxyterminal domain (HisRadC) or amino-terminal domain of radixin (HisRadN) with
PIP or buffer alone. I centrifuged these samples and collected the pelleted protein.
In the presence of PIP, HisRadFL and HisRadN pelleted (Figure 2-2A and C), but
HisRadC did not (Figure 2-2B). In the presence of phosphatidyl choline (PC),
HisRadFL does not pellet (data not shown). Thus as was shown for ezrin (Niggli et
al., 1995), radixin has a binding site for specific phospholipids such as PIP, but not PC.
The binding of phospholipids to ERM proteins may increase ligand binding
by eliminating the interdomain interaction shown above. To test this hypothesis,
we asked if the presence of PIP affected the interaction between the domains of
radixin and used PC as a control. We performed affinity chromatography binding
experiments with GSTRadC and HisRadN in the absence or presence of
phospholipids. We incubated 0.2nmoles GSTRadC bound to glutathione beads with
equimolar concentrations of HisRadN in the absence of phospholipids or in the
Chapter 2
48
Domain interactions
presence of either 100gM PIP or 100pM PC. HisRadN bound to beads bearing
GSTRadC (Figure 2-3, lane 1), but not to beads bearing GST alone (data not shown).
In the presence of PIP, the amount of bound HisRadN was significantly diminished
(Figure 2-3, lane 2). The presence of the control phospholipid, PC, has no effect on
the interdomain interaction (Figure 2-3, lane 3). Thus specific phospholipids which
bind to radixin regulate an interdomain interaction and this may affect the ability of
the full-length radixin to interact with its ligands.
Assay for the conformation of radixin
The interdomain interaction and its modulation by phospholipids may
represent a regulated change in the tertiary structure of the protein. We used
covalent cross-linkers to identify such conformations.
We incubated purified His 6 -
tagged, bacterially-expressed radixin (HisRadFL) with 8 different cross-linkers of
different lengths but all reacting with nucleophilic side chains (see Materials and
Methods). Five of the eight cross-linkers produced a second, faster migrating band
of -74kD (Figure 2-4, arrow), clearly distinct from the unreacted protein of molecular
weight 80kD (Figure 2-4, arrowhead). The formation of the faster migrating band
through SDS-PAGE suggests that the covalent cross-linkers captured a more
compact conformation of the protein.
The altered mobility is dependent upon an intact covalent cross-linker and is
not simply a consequence of covalent modifications of nucleophilic side chains on
radixin. Two of the reagents shown in Figure 2-4, DTSSP (lane 2) and DSP (lane 3),
contain internal disulfide bonds. When the products of these cross-linking
Chapter 2
49
Domain interactions
reactions were incubated with a reducing agent before SDS-PAGE analysis, the 74kD
band is no longer detected (Figure 2-5A, lanes 3, 4 (DTSSP) and lanes 5, 6 (DSP)). In
addition, pre-incubation of radixin with protein denaturants such as 0.02% SDS
(Figure 2-5B, lanes 3 and 4) or 6M urea (Figure 2-5B, lanes 5 and 6) prior to adding
the cross-linker completely inhibits the appearance of the faster band. These data
suggest that the formation of the faster mobility band depends upon the intact crosslinker capturing a specific secondary and tertiary structure of the protein.
Although different preparations of radixin differ modestly in the final extent
of formation of the faster band (we tested 9 independent preparations), none react to
convert all of the radixin to the 74kD form We estimated the percent of the protein
converted by measuring the intensity of the bands from an autoradiogram. No
more than 55% of the radixin ran as a 74kD band in any of the preparations. This is
not a consequence of the kinetics of the reaction, which shows no apparent lag
3
(Figure 2-6A) and which plateaus within 10 minutes (BS , Figure 2-6A and B; similar
results for DTSSP and DSP, data not shown). Longer reaction times do not increase
the formation of the faster mobility band, but cause a general decrease in total
protein level (data not shown). Since the formation of the faster mobility band
depends upon the covalent cross-linking of two nucleophilic groups in the
appropriate positions, the efficiency of cross-linking will be affected if the reagent,
having reacted with one such group, hydrolyzes or reacts with an irrelevant second
side chain. Alternatively, the proportion that is not cross-linked may represent a
conformation of the protein unsuited for cross-linking. The presence of a covalent
cross-linker creates an alteration in molecular weight in an SDS-PAGE.
Chapter 2
50
The
Domain interactions
formation of this faster mobility band is fast, dependent on an intact cross-linker and
dependent on the conformation of the protein. This assay has given us a tool to
study the conformation of radixin.
Cross-linking the domains of radixin
To localize the structural elements within radixin that participate in the
formation of cross-linked species, we reacted the amino- and carboxy-terminal
domains, either individually or together, with cross-linking reagent. A faster
mobility cross-linked band similar to that seen with the full-length protein was
produced only after cross-linking the amino-terminal fragment (Figure 2-7, lane 4,
arrow). As shown with HisRadFL, the extent of formation of this band with
HisRadN was about 40% of the total protein. Significantly, we found no evidence
for a faster mobility band from HisRadC (Figure 2-7, lane 6). Furthermore, when
both HisRadN and HisRadC were simultaneously incubated with cross-linkers a
cross-linked product containing both domains is not produced (data not shown).
The results suggest that the cross-linking behavior of the full-length protein can be
explained by the cross-linking of an intra-molecular interaction within the isolated
amino-terminal domain.
The cross-linking reaction also captured higher molecular weight bands,
indicative of a dimer or oligomer interaction. HisRadFL and HisRadN both form
high molecular weight bands, although their formation varies depending on the
specific protein preparation. In addition, the high molecular weight bands are more
prevalent with the HisRadN samples than with HisRadFL samples (Figure 2-8, lanes
Chapter 2
51
Domain interactions
2 and 4, open arrows). The higher molecular weight forms for HisRadFL and
HisRadN may be the consequence of general protein aggregation. In fact very high
concentrations of HisRadN will precipitate out of solution. On the other hand,
HisRadC consistently formed a single higher molecular weight band at
approximately the size of a dimer (Figure 2-7, lane 6, open arrow). The significance
of HisRadC forming a dimer is still unclear.
Phospholipids affect radixin cross-linking
Since phospholipids affect the affinity between the amino- and carboxyterminal domains of radixin, we tested if phospholipids function by changing the
conformation of radixin as detected by cross-linkers. I incubated either HisRadFL or
HisRadN with phospholipids before introducing the cross-linking reagent BS 3 . The
presence of PIP inhibits the formation of the lower molecular weight band for both
HisRadFL and HisRadN (Figure 2-8, lanes 3 and 7 respectively). PIP does not
interfere with the reaction of the cross-linking agent with nucleophilic side chains,
since cross-linker dependent high molecular weight bands are still formed in its
presence (Figure 2-8, open arrowheads). The inhibition by phospholipids of crosslinking reactions shows the same specificity as in the binding reactions: PIP has an
effect on the formation of the intramolecular interaction by cross-linking and it
disrupted the amino- and carboxy- interdomain interaction while PC has no effect
on the intramolecular interaction or the interaction between the radixin domains
(Figure 2-8, lanes 4 and 8). Therefore, an intramolecular interaction at the amino-
Chapter 2
52
Domain interactions
terminal domain of radixin can be captured by cross-linking reagents and this
intramolecular interaction is disrupted by the presence of specific phospholipids.
I also tested if other phospholipids would have an effect on the formation of
the 74kD band. I performed cross-linking reactions in the presence of various
concentrations of PIP, PIP2 and PS. These phospholipids inhibit the formation of
the faster mobility band in a concentration dependent manner (Figure 2-9). PIP and
PIP2 effectively inhibit the formation of the lower molecular weight band at a
concentration of 15gM (Figure 2-9). Phosphatidyl serine, required more than 15gM
to have an effect (Figure 2-9). Other lipids such as PC, IP3 and OAG do not bind to
ezrin (Niggli) and they had no effect on the formation of the lower molecular
weight band after cross-linking (data not shown). These data suggest that the
intramolecular interaction within the amino-terminal domain of radixin can be
disrupted by specific phospholipids.
Chapter 2
53
Domain interactions
Figure 2-1: Detecting an amino- and carboxy-terminal domain interaction
A. Glutathione beads containing GSTRadC (0.2nmoles) was incubated with
increasing amount of HisRadN. We washed the beads with PBS and eluted
the specifically bound proteins with excess glutathione. We analyzed the
proteins using SDS-PAGE and stained the proteins with coomassie blue.
B. We quantified the amount of HisRadN binding to GSTRadC by measuring
the band densities using the IS-1000 Digital Imaging System. The binding of
HisRadN to GSTRadC shows saturation behavior consistent with the
formation of a 1:2 complex.
Chapter 2
54
Domain interactions
A)
[GSTRadC] on Beads:
[HisRadN] Added:
0.2 nmoles
N~
U3
0
.
0
S
GSTRadC
---
+ HisRadN
1
2
3
4
B)
0.1
'
0.08-
E
c
= 0.06sRadN binding
S-a-Hi
g 0.04-
Iam 0.02
01
0
0.1
0.2
0.3
0.4
HisRmdN Added (nmoles)
0.5
Figure 2-2: Phospholipids bind to radixin at the amino-terminal domain
Purified radixin protein was precleared of pelletable aggregates by centrifugation
in a Beckman air driven ultracentrifuge. We incubated radixin with sonicated
PIP (+) (lanes 2 and 4) or buffer alone (-) (lanes 1 and 3) at room temperature and
centrifuged the mixture at 23 psi for 20 minutes in a Beckman airfuge. We
collected the supernatent (SN) and resuspended the pellet with 1X GSD. The
samples were run on an SDS-PAGE and the proteins were detected by
immunoblotting.
A, Pelleting assay done with 1.5gM amino-terminal domain of radixin
(HisRadN) in the presence of PIP (+) or the absence of PL (-).
B, Pelleting assay done with 1.5gM carboxy-terminal domain of radixin
(HisRadC) in the presence of PIP (+) or the absence of PL (-).
C, Pelleting assay done with 1.5gM full-length radixin (HisRadFL) in the
presence of PIP (+) or the absence of PL (-).
Chapter 2
56
Domain interactions
A)
HisRadN
Pellet
SN
+
m
3
2
1
4
HisRadC
B)
SN
Pellet
2
1
4
3
HisRadFL
C)
Pellet
SN
-+
S0
1
2
3
4
Figure 2-3: Phospholipids regulate the interdomain interaction
GSTRadC (0.2 nmoles) was bound to glutathione beads and incubated with
HisRadN (0.2 nmoles) in the absence (lane 1) of phospholipids or in the
presence of 100gM PIP (lane 2) or 100gM PC (lane 3). All reaction buffers and
washes included the appropriate phospholipids if necessary. The eluted
proteins were run on 10% SDS-PAGE. HisRadN (upper panel) and GSTRadC
(lower panel) were detected by western blot analysis.
Chapter 2
58
Domain interactions
Phospholipid:
0
z
0
0
- HisRadN Binding
4- GSTRadC on beads
1
2
3
Figure 2-4: Cross-linkers capture an intra-molecular interaction
A. Various cross-linkers capture an intra-molecular interaction. Purified
radixin (HisRadFL, 250nM) was incubated with various cross-linking reagents
(12.5gM). The cross-linkers used were - BS 3, DTSSP, DSP, BSOCOES, EGS, or
buffer alone (lanes 1-6 respectively). The cross-linking reactions were
quenched and the products were resolved on a 7.5% SDS-PAGE. Radixin is
detected by western blot analysis. In this figure and in subsequent figures, the
arrowhead indicates the position of purified radixin; the arrow indicates the
position of the 74kD band; and the numbers along the left-hand side show the
mobility of molecular weight markers in kilodaltons.
Chapter 2
60
Domain interactions
-&
BS
3
DTSSP
DSP
BSOCOES
EGS
None
Figure 2-5: Formation of the faster band is dependent on an intact cross-linker and the
conformation of the protein
A, The formation of the 74kD band is reversed by reduction of cross-linkers
containing disulfide bonds. Radixin was incubated with the cross-linker DTSSP
(lanes 3 and 4), DSP (lanes 5 and 6) or without cross-linker (lanes 1 and 2), then
incubated with the reducing agent DTT (lanes 1, 3, and 5) or buffer alone (lanes
2, 4, and 6) before analysis as in figure 2-2A.
B, The formation of the 74kD band is inhibited by pre-incubation with
protein denaturants. Radixin was preincubated with 0.2% SDS (lanes 3 and 4)
or 6M urea (lanes 5 and 6) prior to the addition of the cross-linker BS 3 . The
reaction products were analyzed as described in figure 2-2A. Neither
denaturant had an effect on the protein in the absence of BS 3 (lanes 3 and 5).
Chapter 2
62
Domain interactions
A)
Cross-linker:
None
DTSSP
1
3
DSP
DTT:
B)
Denaturant:
Cross-linker (BS):
2
4
NoneNoe0.02%
SDS
-
+
-
+
5
6
6M
Urea
-
+
mgs.0
1
2
3
4 5
6
Figure 2-6: Cross-linker kinetics
A, The formation of the 74kD band occurs within 15 seconds. Purified
radixin was incubated with BS 3 and the reaction was stopped by the addition
of quench solution at the times indicated . The reaction products were
analyzed as described in figure 2-2A.
B, The percent of radixin in the 74kD band was quantified by measuring the
density of the bands from the autoradiograms using an IS1000 digital imaging
system. The data show that the extent of reaction reaches a plateau after 7-10
minutes. Other cross-linking reagents give similar kinetics (data not shown).
Chapter 2
64
Domain interactions
A)
Cross-linking
Reaction time: 0
5
1545s 2
7 10 min
B)
30ca)
CO
CCO
C
20
-
10
-
02cu
*0
0I
4
~U~
8
J
8
I
I
1'0
1
Cross-linking Reaction Time
(minutes)
Figure 2-7: Cross-linking domains of ERM proteins
Purified full-length radixin (HisRadFL, lanes 1 and 2) or domains of radixin,
HisRadN (lanes 3 and 4) and HisRadC (lanes 5 and 6), were incubated in the
presence (lanes 2, 4, 6) or absence (lanes 1, 3, 5) of cross-linker BS 3 . The
products of these reactions were run on 7.5% (lanes 1 and 2) or 10% (lanes 3-6)
SDS-PAGE and analyzed by immunoblotting using anti-radixin antibodies. In
this figure, the arrowhead indicates the molecular weight of the purified
protein; the arrow indicates the position of a possible intramolecular
interaction band; and the open arrow indicates the positions of possible
intermolecular interaction bands.
Chapter 2
66
Domain interactions
HisRadN
HisRadFL
HisRadC
Cross-linker:
2009786
130-
130-
70 45-
40
79-
30
3143-
1
2
3 4
5
6
Figure 2-8: Phospholipids disrupt the formation of the intramolecular interaction
Radixin constructs were incubated with BS 3 in the presence of PIP (lanes 3, 7,
and 11), PC (lanes 4, 8 and 12), or buffer alone (lanes 1,2, 5, 6, 9 and 10). PIP
affects the formation of the intramolecular interaction bands (filled arrow),
but not the formation of the intermolecular interaction bands (open arrow).
PC has no effects on the formation of any cross-linked products (lanes 4 and
8).
Chapter 2
68
Domain interactions
HisRadFL
Croselinker:
+
+
HisRadC
HisRadN
+
Croselinker:
+ +
+
-
++
10
11
go1U-=
60645-i
3031
64-
1
2
3
4
5 6
7
8
9
12
Figure 2-9: Specific phospholipids disrupt cross-linking
We incubated HisRadFL with various concentrations of PIP2, PIP or PS before
the addition of BS 3. The reaction was quenched and analyzed as described in
Figure 2-2A. PIP2 and PIP inhibited the formation of the faster mobility band
most efficiently, whereas PS inhibited the formation only at the highest
concentration tested.
Chapter 2
70
Domain interactions
[Phospholipid]:
BS3:
116-
PIP:
97
66 -
120116-
PIP2 :
97 66 -
120 -
Phosphatidyl
Serine:
11697
66-
-
5
1
+
50
+
pM
DISCUSSION
The general model in the field is that an interdomain interaction blocks the
binding sites present on the individual domains of ERM proteins. We have shown
in our lab that the two domains of radixin interact with one another with a Kd of
about 50 nM using affinity co-electrophoresis (ACE). In this chapter, I describe
experiments which confirm this interdomain interaction using affinity
chromatography. But unlike the ACE experiments, the affinity data show a 1:2
HisRadN:GSTRadC protein binding ratio (Figure 2-1B) This may be due to the
quality of the GSTRadC protein preparation. By coomassie and immunoblotting, we
detect GSTRadC as a doublet. The lower molecular weight band may be a
degradation product of GSTRadC since an antibody specific for the carboxy-terminal
domain of radixin still recognized this band. In a converse binding assay, with
HisRadN bound to Ni-NTA beads and GSTRadC in solution, the binding of the
higher molecular weight band of GSTRadC was enriched as compared to the lower
molecular weight band. Taking this into account, we detect an interaction between
the carboxy- and amino-terminal domains which has a binding ratio closer to 1:1.
Phospholipids play a role in the regulation of several protein-ligand
interactions (Johnson and Craig, 1995; Hirao et al., 1996; Heiska et al., 1998). One
possible way phospholipids can have this effect is by decreasing an inhibitory
activity such as an interdomain interaction which blocks the binding sites present
within the domains of the protein. In this chapter I have also shown phospholipids
bind to radixin at the amino-terminal domain and this binding regulates the
Chapter 2
72
Domain interactions
interaction between the amino- and carboxy-terminal domains. Since PIP has an
effect on the protein-protein interaction in trans this data suggest that the PIP effect
occurs within one domain of the protein and is not a global effect which requires an
intact protein.
That phospholipids affect the interaction between the two domains of radixin
conflicts with conclusions made from previous work done by Niggli et al. (1995). In
their experiment, PIP2 binding did not have an effect on the interaction between
ezrin fragments which were created by protein cleaving (Niggli et al., 1995).
To
identify and interaction between the amino- and carboxy- fragments the authors
incubated the fragments together and ran the samples through a sephadex column.
The fractions from the column were analyzed by gel electrophoresis and
immunoblotting. The carboxy- and amino-terminal protein fragments co-eluted in
the same fraction from the column, thus the authors concluded that they identified
an interaction between the two domains of ezrin. When they performed the same
assay in the presence of PIP2 the elution patterns of these fragments did not change.
From this result, the authors concluded that phospholipids do not affect the
interdomain interaction. My results present an alternative explanation of their
data. Cross-linking the fragments alone showed that the individual molecules of
these domains interact with one another to form higher molecular weight bands on
an SDS-PAGE. On a sephadex column, these proteins may elute as higher
molecular weight complex as well. In my cross-linking assay, I show that the
presence of phospholipids does not affect the formation of this high molecular
weight product and in fact it may enhance its formation (Figure 2-8). So the higher
Chapter 2
73
Domain interactions
molecular weight forms which they detect and conclude are not affected by
phospholipids may be homo-dimerization or homo-oligomerization and not
actually an interaction between the amino- and carboxy-terminal domains. Thus
although my data may seem to conflict with those identified by Niggli et al, upon
closer examination, my results suggest an alternative model of their data that they
might not have considered.
We have uncovered a mechanism for phospholipid regulation by developing
an assay for the conformation of radixin. Cross-linkers capture an intramolecular
interaction within radixin (Figure 2-4) which is detected as a faster mobility band on
SDS-PAGE. The formation of this intramolecular interaction is fast, reversible and
dependent on the conformation of the protein (Figure 2-5). The amino-terminal
domain alone forms an intramolecular interaction which is captured by crosslinkers (Figure 2-7). Therefore, we have localized the region of the protein
responsible for the shift in molecular weight in full-length radixin to the aminoterminal domain of radixin. The conformation of the amino-terminal domain
which can be captured by cross-linkers to form the faster mobility band, can interact
with the carboxy-terminal domain. This conformation changes in the presence of
phospholipids such that the amino-terminal domain does not form the faster
mobility band (Figure 2-8) and does not bind to the carboxy-terminal domain in the
presence of phospholipids (Figure 2-3). Therefore, PIP changes the conformation of
the radixin at its amino-terminal domain and in this conformation, an interdomain
interaction is inhibited. The regulation of this interdomain interaction may
modulate the interaction between radixin and its ligands.
Chapter 2
74
Domain interactions
In addition to capturing an intramolecular interaction, cross-linkers capture
an intermolecular interaction which creates higher molecular weight bands (Figure
2-7). Unlike the formation of the faster mobility band, these high molecular weight
bands formed by with purified HisRadN or HisRadFL are not affected by the
presence of phospholipids.
Moreover, the band intensities and specific band
patterns that are formed varies with each protein preparation. We suggest that
these bands are non-specific artifacts of the cross-linking reactions. HisRadC
consistently formed a single high molecular weight band. The formation of this
band was enhanced by the presence of PIP. It is unclear what the significance of
these bands are.
In this chapter, I have described the interaction of radixin with one ligand itself. In the following chapters, I will show data which describe the interaction of
radixin with two other ligands - layilin and NHE-RF. I will show how the
interdomain interaction described in this chapter interferes with ligand binding in
cis and in trans. But phospholipids enhance the interaction between radixin and its
ligands - presumably by decreasing the interdomain interaction as described in this
chapter. By understanding the inter- and intra-domain interactions of radixin, we
can understand the interactions between radixin and heterologous ligands such as
NHE-RF and layilin.
Chapter 2
75
Domain interactions
Figure 2-10: Intra- and interdomain interactions of radixin are regulated by
phospholipids
In this chapter, I have presented evidence that radixin forms an intra- and
interdomain interaction.
A, An intramolecular interaction was detected after reacting purified radixin
with covalent protein cross-linkers. The intramolecular interaction was
detected in the full-length protein (left column) as well as with the aminoterminal domain (right column).
B, Phospholipids change the conformation of radixin. In the presence of PIP,
but not PC, the intramolecular interaction is disrupted. This affects the
interdomain interaction such that the two domains do not bind to one
another.
Chapter 2
76
Domain interactions
.................
..
..
...................
A)
N-terminal domain
B)
Full length protein
in cis
in trans
N-terminal domain
Full length protein
C-terminal domain
.4a
PIP
f
*41m 4IP
PIP
BIBLIOGRAPHY
Algrain, M., Turunen, 0., Vaheri, A., Louvard, D., and Arpin, M. (1993). Ezrin
contains cytoskeleton and membrane binding domains accounting for its proposed
role as a membran-cytoskeletal linker. J. Cell Biol. 120, 129-139.
Gary, R., and Bretscher, A. (1995). Ezrin self-association involves binding of an Nterminal domain to a normally masked C-terminal domain that includes the F-actin
binding site. Mol. Biol. Cell 6, 1061-1075.
Heiska, L., Alfthan, K., Gronholm, M., Vilja, P., Vaheri, A., and Carpen, 0. (1998).
Association of ezrin with intercellular adhesion molecule-1 and -2 (ICAM-1 and
ICAM-2). Regulation by phosphatidylinositol 4, 5- bisphosphate. J Biol Chem 273,
21893-900.
Henry, M., Gonzalez-Agosti, C., and Solomon, F. (1995). Molecular dissection of
radixin: Distinct and interdependent functions of the amino- and carboxy-terminal
domains. Journal of Cell Biology 129, 1007-1022.
Hirao, M., Sato, H., Kondo, T., Yonemura, S., Monden, M., Sasaki, T., Takai, Y.,
Tsukita, S., and Stukit, S. (1996). Regulation mechanism of ERM
(ezrin/radixin/moesin) protein/plasma membrane association: possible
involvement of phosphatidylinositol turnover and Rho-dependent signaling
pathway. Jour. Cell Biol. 135, 37-51.
Johnson, R. P., and Craig, S. W. (1995). F-actin binding site masked by the
intramolecular association of vinculin head and tail domains. Nature 373, 261-264.
Magendantz, M., Henry, M., Lander, A., and Solomon, F. (1995). Inter-domain
interactions of radixin in vitro. Jour. Biol. Chem. 270, 25324-25327.
Matsui, T., Maeda, M., Doi, Y., Yonemura, S., Amano, M., Kaibuchi, K., and Tsukita,
S. (1998). Rho-kinase phosphorylates COOH-terminal threonines of
ezrin/radixin/moesin (ERM) proteins and regulates their head-to-tail association. J
Cell Biol 140, 647-57.
Murthy, A., Gonzalez-Agosti, C., Cordero, E., Pinney, D., Candia, C., Solomon, F.,
Gusella, J., and Ramesh, V. (1998). NHE-RF, a regulatory cofactor for Na(+)-H+
exchange, is a common interactor for merlin and ERM (MERM) proteins. J Biol
Chem 273, 1273-6.
Niggli, V., Andreoli, C., Roy, C., and Mangeat, P. (1995). Identification of a
phosphatidylinositiol-4,5-bisphosphate-binding domainin the N-terminal region of
ezrin. FEBS Letters 376, 172-176.
Chapter 2
78
Domain interactions
Pestonjamasp, K., Amieva, M., Strassel, C., Nauseef, W., Furthmayr, H., and Luna,
E. (1995). Moesin, ezrin and p205 are actin-binding proteins associated with
neutrophil plasma membranes. Mol. Biol. Cell 6, 247-259.
Reczek, D., and Bretscher, A. (1998). The carboxyl-terminal region of EBP50 binds to a
site in the amino- terminal domain of ezrin that is masked in the dormant
molecule. J Biol Chem 273, 18452-8.
Tsukita, S., Oishi, K., Sato, N., Sagara, J., Kawai, A., and Tsukita, S. (1994). ERM
family members as molecular linkers between the cell surface glycoprotein CD44
and actin-based cytoskeletons. Journal of Cell Biology 126, 391-401.
Winckler, B., Charo Gonzalez Agosti, Margaret Magendantz and Frank Solomon
(1994). Analysis of a cortical cytoskeletal structure: a role for ezrin-radixin-moesin
(ERM proteins) in the marginal band of chicken erythrocytes. Journal of Cell Science
107,2523-2534.
Chapter 2
79
Domain interactions
CHAPTER 3:
RADIXIN AND NHE-RF INTERACTIONS
A catalog of ligands has been established for ERM proteins. This list can be
classified into several types of proteins: transmembrane proteins such as CD44,
ICAM-1, ICAM-2 and ICAM-3 (Tsukita et al., 1994; Heiska et al., 1996; Serrador et al.,
1998; Yonemura et al., 1998); cytoskeletal proteins such as actin (Turunen et al., 1994;
Roy et al., 1997); and proteins associated with a signal transduction pathway such as
Rho-GDI and MBP(Hirao et al., 1996; Fukata et al., 1998).
The identification of these
ligands and the localization of these proteins suggested that ERM proteins are
cytoskeleton-membrane linkers. The challenge that remains is to decipher how
ERM proteins bind to these ligands and what role they play in reorganizing the cell.
In this chapter, I describe the identification of a Na+/H+ exchanger regulatory factor
(NHE-RF) as a ligand for ERM proteins and use this ligand to study the regulation of
radixin binding.
NHE-RF was originally identified as a regulatory co-factor necessary for
protein kinase A inhibition of NHE, Na+/H+ exchanger (Weinman et al., 1993).
NHEs are the most significant and widespread mechanism for maintaining the pH
in the cell. It has been shown that pH plays an important role in several cell
activities such as cell growth, response to growth factors, cell differentiation and
oncogenesis. To date there are 5-6 different isoforms of these antiporters. They
contain 10-12 membrane spanning regions at the amino-terminus and a cytoplasmic
region at the carboxy-terminus. Of the different isoforms, NHE1 is by far the most
ubiquitous form and is the Na+/H+ exchanger necessary for pH homeostasis and
regulating cellular volume. The other NHE isoforms are distributed in various
Chapter 3
81
Radixin and NHE-RF
tissues (Orlowski and Grinstein, 1997). Within individual cells, NHE localizes to
the border of lamellipodia and in areas that are enriched with vinculin, talin and Factin. This gave some indication that NHE may interact with the cytoskeleton
(Orlowski and Grinstein, 1997).
By using cell lines with modified NHE expression, the Barber lab (Vexler et
al., 1996; Tominaga and Barber, 1998) was able to show that NHE is a necessary factor
for Rho- induced formation of stress fibers and integrin mediated cell spreading.
They used four cell models: CCL39 fibroblasts expressed only the NHE1 isoform;
CCL39 + EIPA which is a pharmacological reagent which specifically inhibits NHE1;
PS120 is a derivative of CCL39 which is deficient in NHE1 expression; and
PS120/NHE1 is PS120 with NHE1 stably expressed. Using these cell lines, they were
able to show that LPA stimulation required NHE1 activity. They also showed that
Rho- stimulated formation of stress fibers were only seen in cells that had active
NHE1 (Vexler et al., 1996). Furthermore other cell activities such as cell attachment
to fibronectin and cell spreading were inhibited without NHE1 activity. Thus NHE
activity appears to be a significant factor in cell activities which change the
organization of cell shape and the organization of the cell.
There are several possible methods for regulating NHE activity. NHE has
several possible phosphorylation sites and the protein is constitutively
phosphorylated in resting cells. It is also further phosphorylated in the presence of
growth factors or phosphatase inhibitors. But phosphorylation cannot account for
all of the regulation since a mutant protein missing all its possible phosphorylation
sites is still functional. Furthermore there isn't a correlation between the
Chapter 3
82
Radixin and NHE-RF
phosphorylation of the protein and its activity (Orlowsky et al., 1997). Another
method of regulating NHE activity is directly binding to other proteins. Calcineurin
B Homolog Protein (CHP) associates with the antiporter at a site near the
membrane. CHP binding inhibits NHE activity and the phosphorylation state of
CHP binding determines its ability to bind to NHE. Another protein which binds to
NHE is NHE-RF, NHE regulatory factor. Using fractionation and reconstitution
experiments, NHE-RF was identified as a co-factor necessary for protein kinase A,
PKA, inhibition of NHE. NHE-RF is a co-factor distinct and dissociable from the
Na+/H+ exchanger (Weinman et al., 1995). NHE-RF is a 44kD protein and a target
for PKA phosphorylation. Thus other methods than direct phosphorylation of NHE
may be regulating its activities.
NHE-RF was identified as a binding partner for ERM proteins. Using a twohybrid screen, NHE-RF was identified as a ligand for merlin, a protein homologous
to ERM proteins and the product of a tumor suppressor gene, neurofibromatosis
type 2, NF2 (Murthy et al., 1998). It was also identified from placenta and brain
extracts as a ligand for ezrin using affinity chromatography and named EBP50, ERM
binding protein (molecular weight) 50 kD (Reczek et al., 1997). In this chapter, I
identify NHE-RF as a direct ligand for radixin and map the domains of their
interaction to the amino-terminal domain of radixin and the carboxy-terminal
domain of NHE-RF. I also use this ligand to determine the mode of regulation for
binding to radixin. In conjunction with the results from Chapter 2, I have formed a
model of regulation for radixin binding to NHE-RF. My data show that
phospholipids regulate the interaction between radixin and its ligands by changing
Chapter 3
83
Radixin and NHE-RF
the conformation of the amino-terminal domain, disrupting the interdomain
interaction and allowing the interaction between radixin and NHE-RF.
Chapter 3
84
Radixin and NHE-RF
MATERIALS AND METHODS
Antibodies
To detect radixin proteins by western, we used polyclonal antibodies against
the amino-terminal domain of ERM proteins (220) or a carboxy-terminal sequence
specific for radixin (457-3) (Winckler et al., 1994). We used NP1 to detect the fulllength and the carboxy-terminal domain of NHE-RF (Murthy et al., 1998).
Recombinant Proteins
In this chapter, we used His6 -tagged full-length radixin (HisRadFL), the
amino-terminal domain of radixin (HisRadN), and the carboxy-terminal domain of
radixin (HisRadC) which was purified as described in Chapter 2.
GST fusion proteins of full-length NHE-RF (GST-IFL) (Ramesh, unpublished)
and the carboxy-terminal domain of NHE-RF (GST-IC270) (Ramesh, unpublished)
were expressed in Escherichia coli DH5a and purified using glutathione beads as
described for GSTRadC in chapter 2.
Binding Assays
For the binding assays, we used Ni-NTA Sepharose CL-6B resin (Qiagen) to
bind His6 -tagged proteins or glutathione-agarose beads to bind to GST-tagged
proteins. In experiments using Ni-NTA beads, the beads were equilibrated with
Buffer I (40mM imidazole buffer, 300mM NaCl, 50mM NaPhosphate, pH 8.0) before
adding protein. The His 6-tagged radixin constructs were brought up to 150 t1 with
Chapter 3
85
Radixin and NHE-RF
PBS and incubated with the Ni-NTA resin for 20 min turning end over end at 4'C.
After washing the beads with buffer I, we added 150gl of the stated concentration of
GST-IFL or GST-IC270 in PBS to the radixin beads. The beads with the bound
proteins were washed with 500gl of Buffer V (40mM imidazole buffer, 300mM
NaCl, 50mM NaPhosphate, pH 8.0, 10% glycerol) to remove non-specific protein
binding. When phospholipids were used in the experiment, they were present in
the binding buffers as well as all the wash buffers. The bound proteins were eluted
with 150g1 Buffer VII (250mg imidazole, 300mM NaCl, 50mM NaPhosphate pH8.0).
The eluate was boiled in GSD and stored at -40'C .
For glutathione bead binding assays, we followed the same protocol as
described in Chapter 2. GST-IFL or GST-IC270 was bound to pre-swelled glutathione
beads. The beads were washed with PBS and His tagged radixin was incubated with
the NHE-RF on the beads for 20 minutes at 4'C. In some cases, we reacted radixin
with cross-linkers before incubating it with the beads. Purified radixin was
incubated with cross-linkers as described in Chapter 2, I quenched the reaction with
excess amines. I incubated this to glutathione beads containing GST-NHE-RF for 20
minutes at 4'C. The beads were washed with PBS and the bound proteins eluted
with 150gl of glutathione elution buffer (5mM glutathione, 50mM Tris pH 8.0, 0.2%
Tween-20). The eluants were boiled in GSD and the samples were analyzed by
immunoblotting.
Chapter 3
86
Radixin and NHE-RF
Affinity Preparation of hNHE-RF from Cell Lysates
RIPA (50mM Tris, pH 7.5, 150mM NaCl, 1% Nonidet P-40, 0.5% deoxycholate,
0.1% SDS containing a 1X protease inhibitor mixture) lysate from COS-7 cells
overexpressing hNHE-RF was incubated with 600pmol of the His 6-tagged full-length
(HisRadFL), amino-terminal domain (HisRadN), or carboxyl-terminal domain of
radixin (HisRadC). The complexes were separated from the reaction mixture using
Ni-NTA beads. The beads were washed (20mM imidazole, 50 mM NaPhosphate,
300mM NaCl, pH 8.0), and the specifically bound complexes were eluted with the
same buffer containing 400mM imidazole and the samples were boiled in SDS.
Immunofluorescence
The immunofluorescence staining of moesin and NHE-RF was performed as
described previously (Gonzalez et al., 1996). Anti-HA monoclonal antibody(1:100)
was used as a primary antibody to detect the localization of hNHE-RF. The affinitypurified moesin antibody was used as previously described (Henry et al., 1995). Cells
were examined on a Nikon microscope using a 10 X 1.2 N.A. and 60 X 1.4 N.A.
objectives. For confocal microscopy, cells were examined with Leica TCS-NT 4D
scanning laser confocal microscope.
Chapter 3
87
Radixin and NHE-RF
RESULTS
Evidence for an interaction between ERM proteins and NHE-RF
NHE-RF, a Na+/H+ exchanger regulatory factor, was identified in a yeast twohybrid screen for interactors with merlin, a homolog of ERM proteins (Murthy et
al., 1998).
NHE-RF localizes to the ruffling edges, microvilli and filopodia, in HeLa
cells exogenously expressing the protein. Immunofluorescence studies detected
partial co-localization between NHE-RF and endogenous moesin at the microvilli
on the dorsal surface of the cell (Figure 3-1, panels a-c) and at the leading edges and
filopodia on the ventral surface (Figure 3-1, panels d-f) (Murthy et al., 1998). This
result suggests that these proteins may interact in vivo.
To determine if an interaction between ERM proteins and NHE-RF exists, we
performed affinity chromatography using lysates from cells expressing HA-tagged
NHE-RF. We incubated cell lysates from these cells with Ni beads containing Histagged constructs of radixin. NHE-RF showed a strong association with the aminoterminal domain of radixin (HisRadN) (Figure 3-2A, lane 1), but not with the
carboxy-terminal domain of radixin (HisRadC) (lane2) or beads alone (lane 4). The
interaction with the full-length protein (HisRadFL) is reduced compared with the
amino-terminal domain alone (Figure 3-2A, lane 3). These results suggest that
NHE-RF interacts with ERM proteins at the amino-terminal domain but this
binding is inhibited in the full-length protein.
Chapter 3
88
Radixin and NHE-RF
NHE-RF interacts directly with ERM proteins
Since the previous experiments were performed using extracts, it is possible
that other cellular proteins mediate the observed interaction. To determine if NHERF can bind directly to radixin, we generated GST-tagged full-length NHE-RF (GSTIFL) (Figure 3-2B, lane 2) and GST-tagged carboxy-terminal domain of NHE-RF
(GST-IC270, residues 270-358, Figure 3-2B, lane 1); and purified the proteins by
affinity chromatography. These NHE-RF protein preparations were incubated with
His6 -tagged radixin polypeptides purified on Ni-NTA agarose beads. Both GST-IFL
and GST-IC270 bound to radixin (Figure 3-2C). GST-IFL and GST-IC270 bound more
strongly to the amino-terminal domain of radixin (HisRadN) (Figure 3-2C, lane 1
and 5, respectively) than to His6-tagged full-length protein (HisRadFL) (lane 3 and 7,
respectively). However, there is no detectable GST-IFL or GST-IC270 bound to the Cterminal domain of radixin (HisRadC) (lane 2 and 6, respectively). This mapped the
interaction between radixin and NHE-RF to the carboxy-terminal domain of NHERF and the amino-terminal domain of radixin.
The binding sites for radixin on the beads are not saturated. We performed
the same binding assays as described above but varied the bead volume. As we
decreased the volume of beads, the amount of radixin and thus the amount of GSTIFL (data not shown) and GST-IC270 (Fig 3-3A) binding to the beads remained
constant. The amount of NHE-RF binding is dependent on the amount of radixin
bound to beads. As the concentration of HisRadFL (Fig 3-3B, lanes 7-9) and HisRadN
(lanes 1-3) decreased, the amount of GST-IC270 and GST-IFL bound to the beads
Chapter 3
89
Radixin and NHE-RF
correspondingly decreased. Even at the highest concentrations of HisRadC tested
(0.6nmoles), we detected no GST-NHE-RF bound to the beads above background
levels (Figure 3-3B, lanes 4-6). Thus NHE-RF binding is dependent on the
concentration of HisRadN and HisRadFL bound to the beads.
To determine if the binding between radixin and NHE-RF is saturable, I
determined the binding curves for GST-IFL and GST-IC270 on beads containing
HisRadN or HisRadFL. A fixed concentration (0.2nmoles) of HisRadN (Figure 34A) or HisRadFL (data not shown) was bound to Ni-NTA beads. An increasing
amount of GST-IFL or GST-IC270 was incubated with the radixin beads. The bound
proteins were separated by SDS-PAGE and the proteins detected by immunoblotting.
I measured the intensity of the bands from the autoradiograms using an IS-1000
Digital Imaging System and calculated the amount of NHE-RF bound to beads. Both
NHE-RF constructs bound to HisRadN with a saturation behavior consistent with
the formation of about 1:1 complex (Figure 3-4B, open and closed circles).
Conversely, the binding of neither GST-NHE-RF constructs to HisRadFL reached
saturation levels at the concentrations we tested (Figure 3-4B, open and closed
squares).
An interaction between radixin and NHE-RF is also detected if NHE-RF is
fixed on a matrix. GST-IFL and GST-IC270 were bound to glutathione beads and
incubated with His-tagged radixin constructs. The beads were washed with
phosphate buffered saline, and the proteins were eluted with glutathione elution
buffer (5mM glutathione, 50mM Tris pH 8.0, 0.2% Tween-20). The results are
similar, but not identical to the results from the Ni-matrix binding assays. HisRadN
Chapter 3
90
Radixin and NHE-RF
bound to both GST-IFL and GST-IC270 constructs (Figure 3-5A and B, lanes 1, 4) but
not to beads alone (lane 7). We did not detect HisRadC binding to either NHE-RF
construct (lanes 2 and 5) but HisRadC and HisRadFL seems to bind nonspecifically to
glutathione beads (lane 8 and 9 respectively). Contrary to the results from the Ni
binding assays, when NHE-RF is on the matrix, I do not detect an interaction
between HisRadFL and NHE-RF by coomassie staining (Figure 3-5A) or western
blotting (Figure 3-5B). This difference in binding may be due to a difference between
the conformation of radixin on a fixed matrix versus in solution.
Thus there is a direct interaction between the amino-terminal domain of
radixin and NHE-RF. We mapped the interaction to the carboxy-terminal domain
of NHE-RF and the amino-terminal domain of radixin. If I use a protein fragment
containing the amino-terminal domain of radixin, detecting its interaction with
NHE-RF is not dependent on which of the two proteins is bound to the matrix.
Conversely, the ability of full-length radixin to bind NHE-RF is different depending
on whether full-length radixin is bound to the matrix or in solution.
Interdomain interactions inhibit ligand binding
In the binding experiments between NHE-RF and radixin, full-length radixin
bound less efficiently to NHE-RF than did the amino-terminal domain. Several in
vivo and in vitro experiments from our lab and other labs suggest that the binding
activities of ERM domains are suppressed in the full-length protein (Algrain et al.,
1993; Turunen et al., 1994; Henry et al., 1995; Hirao et al., 1996; Reczek et al., 1997;
Takahashi et al., 1997; Murthy et al., 1998). This property may be explained by
Chapter 3
91
Radixin and NHE-RF
interactions between the domains of ERM proteins that compete with their
interactions with ligands (Gary and Bretscher, 1995; Magendantz et al., 1995). This
model could account for the difference in binding abilities between the aminoterminal domain and full-length radixin which I reported in the section above.
Accordingly, we tested HisRadC for its ability to compete with binding of both GSTIFL and GST-IC270 to HisRadN. We prepared Ni-NTA beads containing a fixed
amount of HisRadN and varying amounts of HisRadC. NHE-RF was incubated
with the beads as previously described. For both GST-IFL (Figure 3-6A) and GSTIC270 (data not shown), the observed binding decreases with increasing amounts of
HisRadC (Figure 3-6A). These data suggest that HisRadC binding to HisRadN
competes with binding of either GST-IFL or GST-IC270.
This interdomain interaction can also occur in solution to block HisRadN
binding to its ligands. In this experiment, GST-IFL and GST-IC270 were bound to
glutathione beads. I incubated these beads with a single concentration of HisRadN
and varying amounts of HisRadC in a PBS solution. Increasing amounts of
HisRadC inhibited HisRadN binding to both GST-IFL and GST-IC270 bound to
glutathione beads (Figure 3-6B). Little or no HisRadC itself bound to NHE-RF on
glutathione beads. The amount seen with the highest concentration of HisRadC
may be due to non-specific binding as seen in Figure 3-5A, lane 8. Thus the radixin
interdomain interaction can block the binding between radixin and NHE-RF if both
protein fragments containing the amino- and carboxy-terminal domains of radixin
are fixed on a matrix or if the protein fragments are in solution.
Chapter 3
92
Radixin and NHE-RF
Phospholipids affect the NHE-RF binding to radixin
The presence of phosphoinositides enhances the interactions between ERM
proteins and ICAM-1, ICAM-2 and CD44 (Hirao et al., 1996; Heiska et al., 1998). To
determine if phospholipids also have an effect on the interaction between radixin
and NHE-RF, we incubated radixin with GST-IFL in the presence of phospholipids.
Since I showed that HisRadFL interacts with PIP in a pelleting assay, we chose PIP
for these experiments as well and used PC as a negative control. We observed an
increase in binding of GST-IFL to full-length radixin in the presence of PIP (100 jIM)
(Figure 3-7A, lane 2) but the same concentration of PC has no detectable effect on the
ligand binding to radixin (Figure 3-7A, lane 3). The effect of PIP on binding is
dependent on the concentration of the phospholipid in the reaction. If increasing
amounts of PIP is added to the binding reaction, an increased amount of GST-IFL
binds to full-length radixin (Figure 3-7B).
Thus the binding of radixin to NHE-RF
can be regulated by the presence of phospholipids. As mentioned in Chapter 2 an
interdomain interaction between the amino and carboxy-terminal domains is
disrupted by the presence of phospholipids and phospholipids change the
conformation of radixin at the amino-terminal domain and this decreases the
affinity of the amino-terminal domain for the carboxy-terminal domain. The
decrease in affinity between the domains can account for the increase in ligand
binding.
Chapter 3
93
Radixin and NHE-RF
Figure 3-1: Immunofluorescence of NHE-RF and ERM proteins
Co-localization of hNHE-RF with moesin by double label
immunocytochemistry.
Fluorescein isothiocyanate-conjugated secondary
anti-rabbit antibody was used to detect hNHE-RF (a, d) and rhodamineconjugated secondary anti-mouse antibody was used to detect endogenous
moesin (b, e) in HeLa cells. On the dorsal surface of the cells (a-c), the regions
containing these two proteins strongly coincide (c). On the ventral surface
close to the substrate (d-f), the two protein co-localize very clearly at ruffling
membrane and filopodia.
Chapter 3
Bar, 20m. (Murthy et al 1998)
94
Radixin and NHE-RF
NHE-RF
Moesin
Combined
Murthy et al. 1998
Figure 3-2: NHE-RF is a ligand for radixin
A, Affinity precipitation of NHE-RF from cell lysates. RIPA lysate from COS-7 cells
overexpressing hNHE-RF was incubated with His 6-tagged radixin constructs bound
to Ni-NTA agarose beads. The beads were extensively washed, the eluants
separated on 10% SDS-PAGE and immunoblotted with affinity purified antihNHE-RF antibody (NP1). The arrow shows the hNHE-RF at 50kD.
B, GST-tagged and purified NHE-RF constructs. Full-length NHE-RF (GST-IFL)
and the carboxy-terminal domain of NHE-RF (residues 270-358, GST-IC270) were
expressed in DH5a cells. The GST-tagged proteins purified on glutathione beads as
described in the material and methods section. The purified proteins were run on
10% SDS-PAGE and visualized by coomassie blue staining.
C, NHE-RF interacts directly with radixin at its amino-terminal domain. 0.2
nmoles of the amino-terminal domain of radixin (HisRadN) (lanes 1 and 5),
the carboxy-terminal domain of radixin (HisRadC) (lanes 2 and 6) or full-length
radixin (HisRadFL) (lanes 3 and 7) was bound to Ni-NTA beads. GST-IFL (lanes
1-4) or GST-IC270 (lanes 5-8) was incubated with Ni-NTA beads alone (lanes 4
and 8) or with beads containing radixin protein. The beads were washed with a
low concentration imidazole buffer and the proteins specifically bound to the
beads were eluted with a high concentration imidazole buffer. The eluted
proteins were run on a SDS-PAGE and the western blots were probed for NHERF. GST does not bind to any of the radixin constructs.
Chapter 3
96
Radixin and NHE-RF
I
0
Aci'
I
I
II
0
-A)
I
M,
I
m~
-4
I
S
I.
-HisRadN
-HisRadC
-HisRadFL
-Beads alone
G)
S
_
--
G)
-HisRadN
,A
-HisRadC
0
-HisRadFL
-Beads alone 0
HisRadN
HisRadC
HisRadFL
Beads alone
4
z
x
m
I
I
I
I
GST-IC270
f
fowS
GST-IFL
Figure 3-3: NHE-RF binds to radixin in a concentration dependent manner
A, Binding sites for radixin on beads is not saturated. 0.2nmoles of radixin
was incubated with various volumes of Ni-NTA bead slurry. I incubated the
radixin bead slurry with GST-IC270, washed the beads and eluted the bound
proteins. The volume of each step remained constant throughout the
experiments. The bound proteins were run on an SDS-PAGE and analyzed by
immunoblotting. Similar results were received if radixin were incubated
with GST-IFL (data not shown).
B, NHE-RF binding to radixin is dependent on radixin concentration on the
beads. Various concentrations (0.06, 0.2 and 0.6nmoles) of His6-tagged radixin
protein was bound to Ni-NTA beads. I incubated these beads with 0.6nmoles
of GST-IFL (data not shown) and GST-IC270. The beads were extensively
washed and the bound proteins eluted and analyzed on an SDS-PAGE and
stained with coomassie blue.
Chapter 3
98
Radixin and NHE-RF
A)
GST-IC270
Bead Volume:
Radixin on beads:
1Op1
25pl
50p
N C R B N C R B N C R B
97 66 41 31
ae
HisRadC
HisRadN
GST-IC270
wOw,
"Now
HisRadFL
-
B)
Radixin Construct:
HisRadN
0
[Radixin]: ' *4 6 C;o
HisRadC
HisRadFL
(0
O
coo
600
I
f
;*
'
'
-
--
HisRadFL
HisRadC
HisRadN
GST-IC270
Figure 3-4: NHE-RF binding to the amino-terminal domain of radixin is saturable
A, NHE-RF binding to HisRadN is saturable. 0.2nmoles of HisRadN was
bound to Ni-NTA beads and incubated with increasing amounts of NHE-RF GST-IFL (lanes 1-5) or GST-IC270 (lanes 5-10) (0.01, 0.02, 0.06, 0.2, 0.4 nmoles).
The beads were washed with a low concentration imidazole buffer and the
proteins were eluted with high imidazole buffer and analyzed by
immunoblotting for NHE-RF as described in the materials and methods.
NHE-RF binding to HisRadFL does not saturate at the concentrations we
tested. 0.2nmoles of HisRadFL was bound to Ni-NTA beads and incubated
with NHE-RF constructs as described above (data not shown).
B, Binding curves for NHE-RF and radixin interaction. I estimated the
amount of NHE-RF binding to the beads by quantifying the intensities of the
band from autoradiograms using the IS1000 digital imaging system. The
binding of both GST-NHE-RF constructs to HisRadN saturates with a
concentration ratio consistent with about 1:1 stoichiometry (GST-IFL, closed
circles; GST-IC270, open circles). GST-NHE-RF binding to HisRadFL does not
saturate at the concentrations of NHE-RF we tested (GST-IFL, closed squares;
GST-IC270, open squares).
Chapter 3
100
Radixin and NHE-RF
A)
GST-IC270
GST-IFL
[NHE-RF] Added
(nmoles)
v m 0 0 0 V
o
o oc; o
m C
0 0
c; C5o
GST-IFL -
GST-IC270 -
1
2
3 4
5
6 7 8
9 10
B)
0.35
0.3
-.
HisRadFL + IFL
-o-HisRadFL + IC270
-- HisRadN + IFL
-o-HisRadN + IC270
-U-
0.2
g 0.15
0.
-
-
*0.05
0
0
0.2
0.6
0.4
NHE-RF Added (nmoles)
0.8
Figure 3-5: Glutathione bead binding assays - NHE-RF on the matrix
A, Radixin binds to NHE-RF on glutathione beads. GST-IFL and GST-IC270
were bound to glutathione beads. I incubated NHE-RF beads (lanes 1-6) or
blank (lanes 7-9) with HisRadN, HisRadC and HisRadFL. I washed the beads
with phosphate buffered saline and eluted the bound proteins with excess
glutathione. The eluant was boiled in GSD, separated on a 10% SDS-PAGE
and stained with coomassie blue.
B, Immunoblot analysis of the same experiment as described in 3-5A. The
radixin proteins were detected by a mixture of antibodies which recognize
both the amino- (220) and carboxy-terminal (457-3) domains of radixin. I used
protein A conjugated with 1125 to visualize the protein on xray film.
Chapter 3
102
Radixin and NHE-RF
OD
-41
.06
G
l
IC1
CL"
z 0o
wDc K
mm
Ch,
HisRadFL
HisRadC
HisRadN
His RadC
HisRadFL
HisRadN
His RadCO
HIsRadFL
HisRadN
0
4
w
c
0
(3o
-
Cn1
N)
-.1
r2zo
I
I-
G)
(I)
7'
I
I
I
II
HisRadFL
HisRadC
HisRadC
H=fld
HisRadFL
HisRadNL
HisRadCN
HisfladFL
HisRadFL
00
0
Figure 3-6: An interdomain interaction between the amino- and carboxy-terminal
domains of radixin block NHE-RF binding to radixin
A, Simultaneously binding HisRadC and HisRadN to the matrix inhibits
NHE-RF binding to HisRadN. GST-IFL binds to beads containing HisRadN
(lane 1). We incubated GST-NHE-RF (0.1 nmoles) with Ni-NTA beads
containing HisRadN (0.1 nmoles) and increasing amounts of HisRadC (lanes
1-4 and 5-8 respectively - 0, 0.1, 0.3, 1.0 nmoles). The binding assays were
performed as described above. The proteins were run on SDS-PAGE and the
western blots probed for NHE-RF (lanes 5-8) or HisRadC (lanes 1-4).
B, HisRadC inhibits HisRadN binding to NHE-RF fixed on a matrix.
HisRadN (0.1nmoles) was mixed with HisRadC (0.02 or 0.6 nmoles) or buffer
alone and then incubated with glutathione beads containing GST-IFL (lanes 13) or GST-IC270 (lanes 4-6). We washed the beads with phosphate buffered
saline and eluted the proteins were eluted with excess glutathione. The
amount of HisRadN, HisRadC, GST-IFL or GST-IC270 was detected by
immunoblotting for each protein.
Chapter 3
104
Radixin and NHE-RF
A)
[HisRadN] :
0.1 nmoles
[HisRadC]
o
(nmoles)
:
-V--0M
,.
0
+GST-IFL
+HisRadC
1
B)
2
3
4
IC270
IFL
[HisRadN]:
0.1 nmoles
[HisRadC]:
o
6 o 6 ci
-
HisRadN
+ HisRadC
+-GST-IFL
GST-IC270
1
2
3
4
5
6
Figure 3-7: Phospholipids enhance NHE-RF binding to full-length radixin
A, HisRadFL (0.2 nmoles) was bound to Ni-NTA beads and incubated with
GST-NHE-RF (0.2 nmoles). In addition, the reaction buffers and washes
contained 100 M PIP (lane 2) or 100 M PC (lane 3). The eluted proteins were
detected by western analysis: NHE-RF (upper panel) and HisRadFL (lower
panel).
B, Increasing amounts of PIP increases NHE-RF binding. We incubated
HisRadFL and GST-IFL in the presence of 0gM, 15gM or 50gM PIP in all the
reaction buffers. As the concentration of PIP increased, the amount of NHERF binding increased.
Chapter 3
106
Radixin and NHE-RF
A)
NHE-RF
Phospholipid:
Ligand binding
+
0
z
a. U
C.
-a qw4
Radixin on beads 1
B)
2
3
NHE-RF
[PIP] pM
0 15 50
9766-
45-
31-
-4-
GST-IFL
DISCUSSION
In this study, we have characterized the binding between radixin and NHERF, a protein which was originally identified as a regulatory co-factor for a Na+/H+
exchanger (Weinman et al., 1993).
Using NHE-RF, we studied the regulation of
ligand binding to ERM proteins. We mapped the domains involved in binding to
the amino-terminal domain of radixin and the carboxy-terminal domain of the
NHE-RF. The binding properties of NHE-RF did not depend on whether I used a
fragment of the protein containing the carboxy-terminal domain of NHE-RF or fulllength NHE-RF. On the other hand, a protein fragment containing the aminoterminal domain of radixin bound more efficiently to NHE-RF than did full-length
radixin. Our data show that the differences between binding properties of fulllength radixin versus its fragment may be due to an interdomain interaction within
the full-length protein between its amino- and carboxy-terminal domains. We
report that NHE-RF binding to the amino-terminal domain is blocked by the
presence of the carboxy-terminal domain in trans. Furthermore, we showed that
PIP enhances the NHE-RF binding to full-length radixin. In chapter 2, we showed
that PIP changed the conformation of the amino-terminal domain of radixin and
disrupted the interaction between the amino- and carboxy-terminal domains. The
release of this interdomain interaction may eliminate the competition of the
carboxy-terminal domain binding to the amino-terminal domain and makes the
binding sites in radixin more available to its ligands. Thus PIP regulates the
Chapter 3
108
Radixin and NHE-RF
intermolecular interaction between radixin and NHE-RF by changing the
conformation of radixin at the amino-terminal domain which disrupts the
interdomain interaction between the amino- and carboxy-terminal domains of
radixin and increases radixin binding to its ligands (Model Figure 3-8). By
controlling intermolecular interactions in this way, the cell can control the
recruitment of proteins to specific regions of the cell.
Several cytoskeletal proteins show a discrepancy between the binding
activities of the full-length protein and its fragments. For example, the head
domain of vinculin has an increased affinity for talin compared to the intact
vinculin molecule (Johnson and Craig, 1994). Individual domains of ERM proteins
bind to actin, Rho-GDI and CD44 more efficiently than do the full-length ERM
proteins (Johnson and Craig, 1994; Turunen et al., 1994; Hirao et al., 1996; Takahashi
et al., 1997). In this paper, we show that the binding of NHE-RF to the aminoterminal domain of radixin is stronger than to full-length radixin (Figure 3-2C). In
addition, the interaction between NHE-RF and the N-terminal domain of radixin
saturates at a stoichiometry of about 1:1 whereas the interaction between NHE-RF
and full-length radixin does not reach saturation at any of the concentrations of
NHE-RF we tested (Figure 3-4). We have shown that an interdomain interaction
may account for the decreased binding between full-length radixin and its ligands
(Figure 3-6). The carboxy-terminal domain of ezrin also inhibits NHE-RF/EBP50
binding to the amino-terminal domain of ezrin (Reczek et al., 1997). Thus the
interdomain interaction masks the binding sites for other proteins. The regulation
of this interdomain interaction may modulate ligands binding to radixin.
Chapter 3
109
Radixin and NHE-RF
Phospholipids play a role in the regulation of several protein-protein
interactions (Johnson and Craig, 1994; Hirao et al., 1996; Heiska et al., 1998). In this
chapter, I show that the presence of PIP enhances ligand binding to full-length
radixin (Figure 3-7). Our data from Chapter 2 show that the interdomain interaction
between the carboxy- and amino-terminal domain of radixin in trans is disrupted in
the presence of PIP (Figure 2-2). This suggests that the effect of PIP occurs within
one domain of the protein and is not a global effect which requires an intact protein.
I have also shown that radixin contains a binding site for PIP at its amino-terminal
domain (Figure 2-2). Therefore, PIP may change the affinity between the two
domains of radixin by causing a conformational change within the amino-terminal
domain.
Binding to phospholipids may not be the exclusive method of regulating
ligand binding to radixin. Even in the presence of PIP, ligand binding to full-length
radixin is not saturated and does not reach the same levels of ligand binding as the
amino-terminal domain (data not shown). One possible reason for these results is
that a population of radixin is denatured and is intractable to PIP regulation.
Another possibility is that PIP binding to radixin is not saturated. A titration of PIP
concentrations up to 100gM showed increasing NHE-RF or layilin binding to fulllength radixin (Figure 3-7B). Yet another possibility is that a second signal enhances
an intermolecular interaction. The phosphorylation of ERM proteins may also play
a role in regulating intermolecular interactions (Takahashi et al., 1997; Fukata et al.,
1998; Matsui et al., 1998; Shaw et al., 1998). It may be a concerted effort between the
Chapter 3
110
Radixin and NHE-RF
phosphorylation event and binding to phospholipids that stabilizes ERM protein
binding to its ligands at specific regions of the cell.
Whether radixin binding recruits NHE-RF to a specific region of the cell or
radixin binds to NHE-RF already bound at the membrane through its interaction
with NHE is unknown. One possibility is that the cell responds to a signal and
increases the level of PIP and PIP 2 in a specific location of the membrane. In these
regions, ERM proteins are activated such that the binding sites for its ligands are
exposed. This allows ERM proteins to bind to their ligands, such as NHE-RF near
the cell membrane. This interaction would allow cytoskeletal rearrangements in
this specific region of the cell. Conversely, if ERM proteins are responsible for
relocalizing NHE-RF, recruiting NHE-RF to this region would allow PKA
inhibition of NHE in this area. The biological relevance of this protein-protein
interaction has yet to be elucidated.
Chapter 3
ill
Radixin and NHE-RF
Figure 3-8: Model for phospholipid regulation of radixin binding to NHE-RF
An interdomain interaction between the amino- and carboxy-terminal
domains blocks the binding site for NHE-RF present within the aminoterminal domain. In the presence of PIP, there is a conformational change
within the amino-terminal domain and this decreases the affinity between
the amino and carboxy-terminal domains of radixin. Without the
competition of the carboxy-terminal domain, NHE-RF can bind to a
previously occluded binding site within the amino-terminal domain of
radixin.
Chapter 3
112
Radixin and NHE-RF
..........
11..............
Radixin
,Aff
0
a PIP
AZ
NHE
NHE-RF
Actin
BIBLIOGRAPHY
Algrain, M., Arpin, M., and Louvard, D. (1993). Wizardry at the cell cortex. Current
Biology 3, 451-454.
Fukata, Y., Kimura, K., Oshiro, N., Saya, H., Matsuura, Y., and Kaibuchi, K. (1998).
Association of the myosin-binding subunit of myosin phosphatase and moesin:
dual regulation of moesin phosphorylation by Rho-associated kinase and myosin
phosphatase. J Cell Biol 141, 409-18.
Gary, R., and Bretscher, A. (1995). Ezrin self-association involves binding of an Nterminal domain to a normally masked C-terminal domain that includes the F-actin
binding site. Mol. Biol. Cell 6, 1061-1075.
Gonzalez, -. A., C., Xu, L., Pinney, D., Beauchamp, R., Hobbs, W., Gusella, J., and
Ramesh, V. (1996). The merlin tumor suppressor localizes preferentially in
membrane ruffles. Oncogene 13, 1239-1247.
Heiska, L., Alfthan, K., Gronholm, M., Vilja, P., Vaheri, A., and Carpen, 0. (1998).
Association of ezrin with intercellular adhesion molecule-1 and -2 (ICAM-1 and
ICAM-2). Regulation by phosphatidylinositol 4, 5- bisphosphate. J Biol Chem 273,
21893-900.
Heiska, L., Kantor, C., Parr, T., Critchley, D. R., Vilja, P., Gahmberg, C. G., and
Carpen, 0. (1996). Binding of the cytoplasmic domain of intercellular adhesion
molecule-2 (ICAM-2) to alpha-actinin. J Biol Chem 271, 26214-9.
Henry, M., Gonzalez-Agosti, C., and Solomon, F. (1995). Molecular dissection of
radixin: Distinct and interdependent functions of the amino- and carboxy-terminal
domains. Journal of Cell Biology 129, 1007-1022.
Hirao, M., Sato, H., Kondo, T., Yonemura, S., Monden, M., Sasaki, T., Takai, Y.,
Tsukita, S., and Stukit, S. (1996). Regulation mechanism of ERM
(ezrin/radixin/moesin) protein/plasma membrane association: possible
involvement of phosphatidylinositol turnover and Rho-dependent signaling
pathway. Jour. Cell Biol. 135, 37-51.
Johnson, R. P., and Craig, S. W. (1994). An intramolecular association between the
head and tail domains of vinculin modulates talin binding. Journal of Biological
Chemistry 269, 12611-12619.
Magendantz, M., Henry, M., Lander, A., and Solomon, F. (1995). Inter-domain
interactions of radixin in vitro. Jour. Biol. Chem. 270, 25324-25327.
Matsui, T., Maeda, M., Doi, Y., Yonemura, S., Amano, M., Kaibuchi, K., and Tsukita,
S. (1998). Rho-kinase phosphorylates COOH-terminal threonines of
Chapter 3
114
Radixin and NHE-RF
ezrin/radixin/moesin (ERM) proteins and regulates their head-to-tail association.
Cell Biol 140, 647-57.
J
Murthy, A., Gonzalez-Agosti, C., Cordero, E., Pinney, D., Candia, C., Solomon, F.,
Gusella, J., and Ramesh, V. (1998). NHE-RF, a regulatory cofactor for Na(+)-H+
exchange, is a common interactor for merlin and ERM (MERM) proteins. J Biol
Chem 273, 1273-6.
Orlowski, J., and Grinstein, S. (1997). Na+/H+ exchangers of mammalian cells. J Biol
Chem 272, 22373-6.
Reczek, D., Berryman, M., and Bretscher, A. (1997). Identification of EBP50: A PDZcontaining phosphoprotein that associates with members of the ezrin-radixinmoesin family. J Cell Biol 139, 169-79.
Roy, C., Martin, M., and Mangeat, P. (1997). A dual involvement of the aminoterminal domain of ezrin in F- and G- actin binding. J Biol Chem 272, 20088-95.
Serrador, J. M., Nieto, M., Alonso-Lebrero, J. L., del Pozo, M. A., Calvo, J.,
Furthmayr, H., Schwartz-Albiez, R., Lozano, F., Gonzalez-Amaro, R., SanchezMateos, P., and Sanchez-Madrid, F. (1998). CD43 interacts with moesin and ezrin and
regulates its redistribution to the uropods of T lymphocytes at the cell-cell contacts.
Blood 91, 4632-44.
Shaw, R. J., Henry, M., Solomon, F., and Jacks, T. (1998). RhoA-dependent
phosphorylation and relocalization of ERM proteins into apical membrane/actin
protrusions in fibroblasts. Mol Biol Cell 9, 403-19.
Takahashi, K., Sasaki, T., Mammoto, A., Takaishi, K., Kameyama, T., Tsukita, S., and
Takai, Y. (1997). Direct interaction of the Rho GDP dissociation inhibitor with
ezrin/radixin/moesin initiates the activation of the Rho small G protein. J Biol
Chem 272, 23371-5.
Tominaga, T., and Barber, D. L. (1998). Na-H exchange acts downstream of RhoA to
regulate integrin-induced cell adhesion and spreading. Mol Biol Cell 9, 2287-303.
Tsukita, S., Oishi, K., Sato, N., Sagara, J., Kawai, A., and Tsukita, S. (1994). ERM
family members as molecular linkers between the cell surface glycoprotein CD44
and actin-based cytoskeletons. Journal of Cell Biology 126, 391-401.
Turunen, 0., Wahlstrom, T., and Vaheri, A. (1994). Ezrin has a COOH-terminal
actin-binding site that is conserved in the ezrin protein family. Journal of Cell
Biology 126, 1445-1453.
Vexler, Z. S., Symons, M., and Barber, D. L. (1996). Activation of Na+-H+ exchange is
necessary for RhoA-induced stress fiber formation. J Biol Chem 271, 22281-4.
Chapter 3
115
Radixin and NHE-RF
Weinman, E. J., Steplock, D., Corry, D., and Shenolikar, S. (1993). Identification of
the human NHE-1 form of Na(+)-H+ exchanger in rabbit renal brush border
membranes. J Clin Invest 91, 2097-102.
Weinman, E. J., Steplock, D., Wang, Y., and Shenolikar, S. (1995). Characterization of
a protein cofactor that mediates protein kinase A regulation of the renal brush
border membrane Na(+)-H+ exchanger. J Clin Invest 95, 2143-9.
Winckler, B., Gonzalez Agosti, C., Magendantz, M., and Solomon, F. (1994). Analysis
of a cortical cytoskeletal structure: a role for ezrin-radixin-moesin (ERM proteins) in
the marginal band of chicken erythrocytes. Journal of Cell Science 107, 2523-2534.
Yonemura, S., Hirao, M., Doi, Y., Takahashi, N., Kondo, T., and Tsukita, S. (1998).
Ezrin/radixin/moesin (ERM) proteins bind to a positively charged amino acid
cluster in the juxta-membrane cytoplasmic domain of CD44, CD43, and ICAM-2. J
Cell Biol 140, 885-95.
Chapter 3
116
Radixin and NHE-RF
CHAPTER 4:
RADIXIN AND LAYILIN INTERACTIONS
The localization patterns of ERM proteins to the plasma membrane rich in
actin was the first indication that these proteins play a role in the interaction
between the cytoskeleton and the membrane. The identification of membrane
associated proteins as ligands gave further evidence for ERM proteins playing this
role in the cell. An integral membrane protein CD44 was one of the first ligands
identified for ERM proteins (Tsukita et al., 1994). It was identified by coimmunoprecipitation studies using an anti-moesin antibody. It was later shown
that CD44 interacts directly with ezrin and radixin as well. Using similar methods,
another membrane component, ICAM-1 was identified as a ligand for ezrin. Ezrin
from placental lysate bound to the cytoplasmic domain of ICAM-1 fixed onto a
matrix (Heiska et al., 1998). Synthetic peptides were used to demonstrate a direct
interaction between the cytoplasmic domain of ICAM-2 and ezrin (Heiska et al.,
1998). In this chapter we identify another integral membrane protein, layilin, as a
ligand for ERM proteins.
Layilin was identified as a ligand for talin, a protein homologous to ERM
proteins, in a yeast two-hybrid screen. Layilin is a novel transmembrane protein
and has homology to C-type lectins. The cellular distribution of layilin is distinct
from integrin, another talin ligand. Integrins are found in more stable cell-matrix
connections such as focal contacts whereas layilin localizes to the ruffling edges of
cells. This may be a method of segregating talin to distinct structures of the cell that
require the cytoskeleton-membrane linkages talin provides. Borowsky et al. (1998)
Chapter 4
118
Radixin and layilin
suggest that layilin and talin form a transient connection at ruffling edges and these
sites ultimately form focal contacts after recruiting integrins.
Since talin head is 58% homologous to ERM proteins, we asked if layilin also
interacted with radixin. In this chapter, we have identified layilin as a direct ligand
for radixin. We also show that like NHE-RF, the interaction between radixin and
layilin can be regulated by the presence of specific phospholipids.
Chapter 4
119
Radixin and layilin
MATERIALS AND METHODS
Antibodies
A monoclonal pan-ERM antibody (13H9) was used for immunofluorescence
experiments (Birgbauer and Solomon, 1989). We detected radixin protein by
immunoblotting using antibodies against the amino-terminal domain of ERM
proteins (220) and against the carboxy-terminal domain (457-3) as described in the
previous chapters. Affinity eluted polyclonal antibodies were used for layilin
detection (619G2); serum depleted of anti-layilin antibodies was used as a negative
control (619FT) (Borowsky and Hynes, 1998).
Immunoblotting
Protein samples were run on SDS-PAGE and transferred to nitrocellulose as
described in Chapter 2. For detecting radixin, we blocked the nitrocellulose filters for
1-2 hours with TBST at room temperature, changing the buffer several times within
this period. I added the primary antibody 220 (1:1000) diluted in TBST to detect the
amino-terminal domain of radixin or 457-3 (1:500) to detect the carboxy-terminal
domain of radixin. I incubated the primary antibody with the filters overnight at
room temperature. I extensively washed the filters with TBST. To detect the bound
antibodies, I used protein A conjugated with 1125 diluted in TBST. This solution
incubated with the filter for 1 hour at room temperature. I washed these blots with
excess TBST and exposed them to film at -70 0 C.
Chapter 4
120
Radixin and layilin
To detect layilin, we blocked the blots with 5% milk in TBST for 1-2 hours at
room temperature. I added the primary antibody, 619G2 (1:1000) in TBST with 5%
milk and incubated the blot overnight at 4*C. I washed the blots with TBST and
detected the protein using HRP conjugated to goat anti-mouse antibody. We used
Enhanced chemiluminescence to detect this secondary antibody as described by the
manufacturer.
Recombinant Proteins
In this chapter, we used His6 -tagged full-length radixin (HisRadFL), the
amino-terminal domain of radixin (HisRadN), and the carboxy-terminal domain of
radixin (HisRadC) which was purified as described in Chapter 2.
A GST fusion protein of the carboxy-terminal domain of layilin (GST-layilin
261-374, (Borowsky and Hynes, 1998)) was expressed in Escherichia coli DH5a and
purified using glutathione beads as described for GSTRadC in chapter 2.
Indirect Immunofluorescence
NIH-3T3 cells were grown on fibronectin-coated (5gg/ml) glass coverslips
overnight and fixed with 4% paraformaldehyde in PBS then permeabilized with
0.1% NP40 in PBS for 15 minutes at room temperature. I blocked the coverslips
with 10% Normal Goat Serum in PBS (Vector Laboratories, Inc., Burlingame, CA)
for 15 min at 37*C then rinsed the coverslips with PBS.
A monoclonal pan-ERM
antibody (13H9) was diluted to 1:50 in 1% BSA / PBS. A polyclonal affinity purified
antibody against layilin (619G2, (Borowsky and Hynes, 1998)) was diluted to 1:50 in
Chapter 4
121
Radixin and layilin
1% BSA / PBS. As a control, I used the same serum after preabsorption of the
antibody with the antigen (619FT) diluted to 1:50 in 1% BSA / PBS (Borowsky and
Hynes, 1998). We used fluorescein isothiocyanate-conjugated goat anti-mouse
(1:1000) to detect 13H9 and tetramethylrhodamine isothiocyanate-conjugated goat
F(ab')2 anti-rabbit IgG (1:500) to detect 619G2 and 619FT. The coverslips were fixed
onto slides using gelvatol and DABCO to decrease fading.
Chapter 4
122
Radixin and layilin
RESULTS
Colocalization of layilin and ERM proteins
Both layilin and ERM proteins localize to cortical structures in various cell
types (Bretscher, 1983; Birgbauer and Solomon, 1989; Goslin et al., 1989; Sato, 1991;
Franck, 1993; Borowsky and Hynes, 1998). To determine if they co-localize in the
same cell, we double stained NIH-3T3 cells for endogenous layilin and all three
ERM proteins. In these cells, ERM proteins show the typical localization to ruffling
edges (arrows), filopodia (arrowheads) and microvilli (data not shown) (Figure 4-1A,
panels a, d, and g). Layilin is conspicuous in ruffling edges (Figure 4-1A, panels b
and e, arrows). We show that this anti-layilin antiserum also gives artifactual
staining of the nucleus by using serum which was preincubated with the antipeptide antigen (Figure 4-1B, panel h, open arrowheads, (Borowsky and Hynes,
1998). The distributions of the two proteins coalign crisply in the ruffling edges
(Figure 4-1A, panels c and f, arrows), but not in the filopodia (Figure 4-1A, panes c
and f, arrowheads). This localization pattern suggests that an interaction may exist
between these proteins.
Layilin is a ligand for radixin
To verify the interaction between layilin and radixin, we incubated COS7 cell
lysates with purified radixin constructs immobilized on Ni-NTA beads. The beads
were washed and the bound proteins eluted from the beads using a high
concentration imidazole buffer. We analyzed the eluants by resolving the 'proteins
Chapter 4
123
Radixin and layilin
on an SDS-PAGE and immunoblotting for layilin protein. Layilin from the cell
extract bound to beads containing the amino-terminal domain of radixin (HisRadN)
(Figure 4-2A, lane 1). Layilin bound to beads containing the carboxy-terminal
domain of radixin (HisRadC) to the same degree as to beads alone (Figure 4-2A, lane
2) There is only a slight enhancements of layilin binding to full-length radixin
(HisRadFL) bound to beads as compared to background levels (Figure 4-2A, lane 3).
These data confirm an interaction between layilin and radixin.
Since the above binding experiment is done with lysates, it is possible that
other cellular factors mediate this interaction. To determine if the proteins directly
interact with one another, we used a purified GST-tagged cytoplasmic domain of
layilin (GSTLayCyt, (Borowsky and Hynes, 1998)). These preparations were then
incubated with His6 -tagged radixin polypeptides purified and bound to Ni-NTA
agarose beads. As we saw with GST-NHE-RF, GSTLayCyt bound to HisRadN (Figure
4-2B, lane 1). We detected no GSTLayCyt bound to HisRadFL (Figure 4-2B, lane 3) or
HisRadC (lane 2). Layilin does not bind to beads alone (Figure 4-2B, lane 4) and GST
polypeptide itself does not bind to any of the radixin constructs in this assay (data
not shown).
Thus layilin is a direct ligand of radixin and its binding site maps to
the amino-terminal domain of radixin.
Interdomain interactions of radixin inhibit layilin binding
In chapter 2, I examined the interdomain interaction between the amino and
carboxy-terminal domains of radixin and in chapter 3 we report that this
interdomain interaction blocks the binding site for NHE-RF binding. Accordingly,
Chapter 4
124
Radixin and layilin
we tested if this interdomain interaction also competed with layilin binding to the
amino-terminal domain of radixin. We prepared Ni-NTA beads containing a fixed
amount of HisRadN and varying amounts of HisRadC. GSTLayCyt were incubated
with the beads and the binding assay performed as described above. The observed
binding of layilin decreased with increasing amounts of HisRadC bound to the beads
(Figure 4-3). This suggests that HisRadC binding to HisRadN competes with binding
of GSTLayCyt. This binding property is similar to that seen with NHE-RF.
Phospholipids induce layilin binding to full-length radixin
We detected no or little layilin binding to HisRadFL (Chapter 4-2A and B,
lanes 3). Upon longer burns, we detect very low levels of layilin interacting with
HisRadFL (data not shown). NHE-RF binding to full-length radixin is enhanced by
the presence of phospholipids (Chapter 3). We tested if specific phospholipids
enhance layilin binding to full-length radixin as well. A fixed amount of HisRadFL
was bound to Ni-NTA beads. GSTLayCyt was incubated with radixin in the absence
of phospholipids or in the presence of 100gM PIP or 100gM PC. In the presence of
PIP, we see an increase in layilin binding to radixin (Figure 4-4B, lane 2). The
control phospholipid, PC, has no effect on layilin binding (Figure 4-4B, lane 3). Thus
Layilin binding to radixin can be enhanced by specific phospholipids. Again, these
results are similar to those seen with NHE-RF (Figure 3-7).
Chapter 4
125
Radixin and layilin
Figure 4-1 Layilin and ERM proteins partially co-localize at the ruffling edges
A, Cellular distribution of ERM proteins and layilin. Using indirect
immunofluorescence, NIH-3T3 cells were stained with an antibody against
ERM proteins (ezrin, radixin and moesin), followed by fluoresceinconjugated goat anti-mouse antibody (panels a and d). Layilin was localized
in these same cells using affinity purified antibody followed by rhodamineconjugated anti-rabbit antibody (panels b and e). The yellow staining in
panels c and f indicates co-localization of layilin and ERM proteins at the
ruffling edges (arrows) but not at the filopodia (arrowheads).
B, Staining with anti-layilin depleted control serum. ERM proteins localized
to filopodia (arrowheads) and ruffling edges (arrows) as described above
(panel g). Anti-layilin serum which has been depleted over an antigen
peptide column detects perinuclear staining (panel h, open arrowheads).
There is no overlap between ERM staining and layilin control serum at the
ruffling edges (panel i, arrows).
Chapter 4
126
Radixin and layilin
ERM
Layilin
Combined
A)
B)
ERM
Control Serum
Combined
A
Figure 4-2 Layilin is a ligand for radixin
A, Layilin in COS-7 cells binds to the amino-terminal domain of radixin. 0.2
nmoles of HisRadN (lane 1), HisRadC (lanes 2) or HisRadFL (lane 3) were
bound to Ni-NTA beads. RIPA (50mM Tris, pH 7.5, 150mM NaCl, 1%
Nonidet P-40, 0.5% deoxycholate, 0.1% SDS containing a 1X protease inhibitor
mixture) lysates from COS-7 were incubated with radixin beads (lanes 1-3) or
beads alone (lane 4). The beads were washed and the specifically bound
proteins eluted from the beads. The samples were boiled in GSD and
separated by SDS-PAGE. We detected layilin binding by immunoblotting as
described in the material and methods.
B, Layilin is a direct ligand for radixin. Radixin constructs were bound to
beads as described in Figure 4-2A. GSTLayCyt (0.2nmoles) was incubated with
beads containing radixin protein (lanes 1-3) or with Ni-NTA beads alone (lane
4). The beads were washed with a low imidazole concentration buffer and the
proteins specifically bound to the beads were eluted with a high imidazole
buffer. The eluted proteins were run on a SDS-PAGE, transferred to
nitrocellulose filters and probed for layilin.
Chapter 4
128
Radixin and layilin
C,,
I
I
I-
C,,
7'
C)
t
S
~.
c~)
I
I
(0
- Beads alone
- HisRadN
- HisRadC
- HisRadFL
00
CO
a
-.-
I
Beads alone
HisRadFL
HisRadC
HisRadN
Figure 4-3 An interdomain interaction blocks layilin binding
GSTLayCyt binds to beads containing HisRadN (lane 1). To determine if
HisRadC could compete with layilin binding to HisRadN, GSTLayCyt
(0.2nmoles) was incubated with beads containing HisRadN (0.2nmoles) and
increasing amounts of HisRadC (lanes 1-4 respectively - 0, 0.2, 0.4 and 2.0
nmoles of HisRadC). The binding assay was performed as previously
described and the bound proteins were run on SDS-PAGE and the filters were
probed for layilin (upper panel) or HisRadC (lower panel).
Chapter 4
130
Radixin and layilin
0.2 nmoles
[HisRadN]
[HisRadC]
(nmoles)
Layilin binding
c'J 19t
0o6 c
(0
0
3
4
-
CRad on beads -+
1
2
Figure 4-4 Phospholipids enhance layilin binding to full-length radixin
Phospholipid enhancement of binding is specific for PIP. HisRadFL (0.2
nmoles) was bound to Ni-NTA beads and incubated with GSTLayCyt (0.2
nmoles). In addition, the reaction buffers and washes contained 100gM PIP
(lane2) or 100gM PC (lane 3). The eluted proteins were detected by western
analysis for layilin (upper panel) and HisRadFL (lower panel).
Chapter 4
132
Radixin and layilin
Phospholipid:
Ligand binding
+
Radixin on beads
-+
0
z
0:
0
a.
Aw
1
2
3
DISCUSSION
ERM proteins localize to cortical structures rich in actin. Several
transmembrane proteins have been identified as ligands for ERM proteins. In this
chapter I describe a novel transmembrane protein layilin, which I have identified as
a ligand for radixin. Layilin binds to the amino-terminal domain of radixin. Like
NHE-RF, this binding site is blocked by the presence of the carboxy-terminal domain
in cis, as seen with the full-length protein, or in trans using domains expressed as
individual polypeptides.
This interdomain interaction can be weakened by the
presence of specific phospholipids and in the absence of the interdomain
interaction, layilin binding to full-length radixin increases.
A variety of other transmembrane proteins are reported to be ERM protein
ligands such as CD44, CD43, ICAM-1 and ICAM-2. A basic amino acid stretch is
shared among these proteins and was identified as the amino acids necessary for
ERM protein binding (Yonemura et al., 1998). Layilin does not contain an amino
acid sequence homologous to these regions, thus the interaction between radixin
and layilin may be due to another motif. Yet it will be interesting to determine if
these ERM protein ligands share other properties.
Ezrin, moesin and radixin are about 80% homologous to one another. It is
not known whether layilin interacts with these other ERM proteins. Layilin was
originally identified from a two hybrid screen for interactors with the talin head (the
amino-terminal domain). The percent homology between the ERM proteins at
their amino-terminal domain is higher than the homology between radixin and
Chapter 4
134
Radixin and layilin
talin. Therefore it seems likely that layilin interacts with the other ERM proteins as
well.
I have now shown that although layilin and NHE-RF are two very different
proteins, the former a transmembrane protein and the latter a cytoplasmic protein,
they both bind to radixin with similar binding properties. Their binding sites are
blocked by an interdomain interaction and their binding is regulated by
phospholipids. That these properties are seen with such different types of proteins
strengthens our model (Figure 4-5). With layilin as with NHE-RF, the binding site
at the amino-terminal domain of radixin is blocked in the full-length protein. In
the presence of phospholipids, there is a conformational change at the aminoterminal domain. This conformation of the amino-terminus decreases its affinity
for the carboxy-terminal domain. Without the competition of the carboxy-terminal
domain, the layilin binding site at the amino-terminal domain is available and
layilin can bind to radixin.
Chapter 4
135
Radixin and layilin
Figure 4-5: Model for phospholipid regulation of radixin binding to layilin
We mapped the layilin binding site to the amino-terminal domain of radixin.
An interdomain interaction blocks this binding site in cis as well as in trans.
As we have previously shown, the presence of PIP eliminates the
interdomain interaction by changing the conformation of the aminoterminal domain. In the absence of the interdomain competition, the
binding site is available for layilin and this increases the layilin binding to
full-length radixin.
Chapter 4
136
Radixin and layilin
...
......
-..........
--.1. 1 ...............
.......
......
-..........................................
Radixin
4PIP
Layilin
Radixin
Actin
...
.................
BIBLIOGRAPHY
Birgbauer, E., and Solomon, F. (1989). A marginal band-associated protein has
properties of both microtubule- and microfilament-associated proteins. Journal of
Cell Biology 109, 1609-1620.
Borowsky, M. L., and Hynes, R. 0. (1998). Layilin, a novel talin-binding
transmembrane protein homologous with C- type lectins, is localized in membrane
ruffles. J Cell Biol 143, 429-42.
Bretscher, A. (1983). Purification of an 80,000-dalton protein that is a component of
the isolated microvillus cytoskeleton, and its localization in nonmuscle cells. J. of
Cell Biology 97, 425-532.
Franck, Z., Ronald Gary and Anthony Bretscher (1993). Moesin, like ezrin,
colocalizes with actin in the cortical cytoskeleton in cultured cells, but its expression
is more variable. Journal of Cell Science 105, 219-231.
Goslin, K., Birgbauer, E., Banker, G., and Solomon, F. (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.
Heiska, L., Alfthan, K., Gronholm, M., Vilja, P., Vaheri, A., and Carpen, 0. (1998).
Association of ezrin with intercellular adhesion molecule-1 and -2 (ICAM-1 and
ICAM-2). Regulation by phosphatidylinositol 4, 5- bisphosphate. J Biol Chem 273,
21893-900.
Sato, N., Shigenobu Yonemura, Takashi Obinata, SAchiko Tsukita and Shoichiro
Tsukita (1991). Radixin, a Barbed-End-capping Actin-modulating Protein, Is
Concnetrated at the Cleavage Furrow during Cytokinesis. Journal of Cell Biology
113, 321-330.
Tsukita, S., Oishi, K., Sato, N., Sagara, J., Kawai, A., and Tsukita, S. (1994). ERM
family members as molecular linkers between the cell surface glycoprotein CD44
and actin-based cytoskeletons. Journal of Cell Biology 126, 391-401.
Yonemura, S., Hirao, M., Doi, Y., Takahashi, N., Kondo, T., and Tsukita, S. (1998).
Ezrin/radixin/moesin (ERM) proteins bind to a positively charged amino acid
cluster in the juxta-membrane cytoplasmic domain of CD44, CD43, and ICAM-2. J
Cell Biol 140, 885-95.
Chapter 4
138
Radixin and layilin
CHAPTER 5:
ANALYSIS OF MERLIN
Neurofibromatosis 2 (NF2) is a human disease which causes bilateral
formation of schwannomas of the eighth cranial nerve. Analysis of these
schwannomas showed the loss of heterozygosity or mutations of both NF2 alleles
(Martuza and Eldridge, 1988; Mautner et al., 1996).
These characteristics identified
NF2 as a tumor suppressor gene. The gene responsible for NF2 was mapped to
chromosome 22 and named schwannomin (Rouleau et al., 1993) or merlin
(Trofatter et al., 1993). The sequence of merlin showed that it is highly homologous
to the ERM family of proteins especially within the amino-terminal domain (63%).
The gene encodes a protein with a predicted molecular weight of 66 kD, and unlike
other ERM proteins, merlin runs at its expected size on an SDS-PAGE. By studying
merlin, we may be able to learn more about the functions of ERM proteins and the
role of cytoskeletal proteins in tumorigenesis.
The severity of neurofibromatosis 2 varies. Some cases of the disease are
severe and occur early in life. Other cases are mild and are late onset. Ruttledge
(1996) analyzed 111 NF2 cases and correlated the mutation to the severity of the
disease. The most severe phenotypes had mutations which caused a protein
truncation. On the other hand, cases with a single amino acid change had mild
phenotypes. A recent review by Turunen et al (1998) discussed the possible protein
conformation changes due to the mutations in merlin found in NF2 patients. The
authors predict that several of the mutations are located in regions of the protein
that would cause drastic conformational changes and affect ligand binding. These
types of studies may aid our understanding of which domains of merlin, and ERM
Chapter 5
140
Merlin
proteins are important for their functions in the cell and possibly which ligand
interactions are necessary to prevent tumorigenesis.
The localization pattern of ERM proteins gave the first indication of their
function as a cytoskeletal-membrane linking proteins. The localization patterns
from some studies show a similarity between merlin and ERM localization, others
note a distinction. Some of the first immunofluorescence studies were done in COS
cells transfected with NF2 cDNA. In these cells, merlin localized to punctate areas
and to stress fibers (den Bakker et al., 1995). Merlin staining was not detected in
untransfected cells. In cultured human fibroblasts and primary meningioma cells,
merlin localized to the ruffling edges with some association with F-actin filaments
(Gonzalez et al., 1996). This staining pattern is more reminiscent of ERM
localization patterns. The differences may lie in the cell types used or the quality of
the antibodies. But there is some evidence that merlin may be acting in similar
regions of the cells as ERM proteins.
The localization patterns indicate that merlin may be able to associate with
actin. The carboxy-terminal actin binding site is conserved (94% homology) within
the ERM family of proteins. The site is less conserved in merlin. Merlin isoforms 1
and 2 are 30-40% identical and 43-57% similar (Huang 1998). The original actin
binding studies done with ERM proteins produced conflicting results (Turunen et
al., 1994; Pestonjamasp et al., 1995; Shuster and Herman, 1995; Yao et al., 1995; Yao et
al., 1996). Studies done with merlin also show some discrepancies. By blot overlay,
Huang et al (1998) detected F-actin binding to moesin, but not to merlin isoform 1.
Using co-polymerization assays Xu et al (1998) reported different results - in vitro
Chapter 5
141
Merlin
transcribed merlin associated with F-actin. In vivo, green fluorescent protein (GFP)
conjugated to the carboxy-terminal domain does not localize to stress fibers or
cortical actin in the cell (Huang et al., 1998). Therefore, it is still unknown if merlin
has similar actin binding activities as ERM proteins.
The conflicting data from early ERM protein binding studies may have been
due to a interdomain interaction which blocks the binding sites for actin. A similar
interaction may occur with merlin. Using affinity chromatography assays, there is
some suggestion of an interdomain interaction between the amino- and carboxyterminal domains of merlin isoform 1 as well as an interaction between the fulllength proteins (Huang et al., 1998). How these interactions affect ligand binding is
still known.
To further assess the similarities and differences between radixin and merlin,
I have done cross-linking assays with wild type merlin protein and a mutant form
of merlin. The mutation is a substitution of a tyrosine for an asparagine at position
220. This mutation was identified as a shift on single-strand conformational
polymorphism (SSCP) in affected individuals, but not in unaffected individuals
(MacCollin et al., 1993). In this assay, the proteins behave in similar fashion to ERM
proteins.
Chapter 5
142
Merlin
MATERIALS AND METHODS
Recombinant Proteins
The merlin proteins were provided by the Ramesh lab. Briefly, GST-tagged merlin
was purified from glutathione beads and the GST tag was removed by cleaving the
protein with thrombin. The purified merlin constructs were aliquotted to small
volumes and frozen until used.
Cross-linking reagent
BS 3 (Bis[sulfosuccinimidyl]suberate) was diluted in DMSO to a 1OX stock solution
immediately before use.
Phospholipids
The reagents (and abbreviations) used in this study were: PIP2 (Phosphatidylinositol
4,5-bisphosphate disodium salt); PS (O-(3-sn-Phosphatidyl)-L-serine sodium salt); PIP
(Phosphatidylinositol 4-phosphate sodium salt); and PC (Phosphatidyl choline). All
phospholipids were purchased from Sigma. The phospholipids were dissolved in
water to a final concentration of 1mg/ml, then sonicated; the stock solutions were
frozen in liquid nitrogen and stored at -30'C. The solutions of PC and OAG did not
completely clarify. The phospholipids were sonicated for 30s - 1min immediately
before use in the crosslinking assays.
Chapter 5
143
Merlin
Western Blot Analysis and Antibodies
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 nitrocellulose filters (Schleicher and Schuell;
Keene, NH) essentially as described by Tobin et al (1979). Total protein transferred to
the nitrocellulose filter was detected by Ponceau S staining (0.2% Ponceau S (Sigma)
in 3% TCA). We blocked the blots with TBST for 1-2 hours at room temperature
and probed the blots overnight with the primary antibody. To detect merlin by
immunoblotting, we used MP4 (1:100) or MP8 (1:200) in TBST. We washed the blots
with TBST and detected the antibodies with 112 5-labeled Protein A (DuPont - New
England Nuclear; Boston, MA). After extensive washing with TBST, the signal was
detected using at -70'C.
Cross-linking Assay
In the cross-linking assays, merlin was pre-incubated for 15 minutes at 25'C in PBS.
For experiments requiring phospholipids, the appropriate phospholipids were
present throughout the reaction including the pre-incubation step. The cross-linker
was added to a final concentration of 50-fold the protein concentration and typically
was allowed to react for 10 min at 25*C. The reaction was stopped by adding a
quenching solution (125 mM Tris, 125 mM glycine, pH 8.0) for an additional 15 min
at 25'C. The samples were then run on a 7.5% SDS-PAGE and analyzed by
immunoblotting as described above.
Chapter 5
144
Merlin
RESULTS
Merlin is 62% homologous to ERM proteins within the amino-terminal
domain. Since these proteins are so similar within the amino-terminal domain, we
asked if we could capture a conformation of merlin which is similar to the
conformation we detected with radixin. Purified merlin was used in the crosslinker assays as described for radixin in the previous chapters. After preincubating
the protein in PBS at room temperature, I added cross-linker, BS 3 , for 10 minutes
and quenched the reaction. I resolved the protein on a 7.5% SDS-PAGE and probed
the blots with antibodies against merlin. Similar to what we saw with radixin,
covalent cross-linkers capture an intra-molecular interaction within merlin (Figure
5-1A, lanes 1 and 2). We tested if phospholipids could disrupt the formation of the
faster mobility band. This would indicate a change in the conformation of the
protein that is regulated by phospholipids. In the presence of PIP, we do not detect a
shift in molecular weight of merlin after its reaction with cross-linkers (Figure 5-1A,
lane 3). Thus there is a protein conformation of merlin and radixin such that the
same cross-linker captures a faster mobility band that can be regulated by
phospholipids.
A missense mutation in the merlin protein at the amino acid position 220
was identified in NF2 patients (MacCollin et al., 1993). We tested if this mutant
protein had the same cross-linking properties as radixin and wild type merlin. We
incubated bacterially expressed and purified mutant protein (Mer220) with a
covalent cross-linker, BS 3 . I quenched the reaction with excess amines and ran the
Chapter 5
145
Merlin
samples on a 7.5% SDS PAGE. Our data show a similar shift in the molecular
weight in the presence of cross-linkers (Figure 5-1B, lanes 1 and 2). Furthermore,
the formation of this faster mobility band is disrupted in the presence of PIP (lane 3),
but PC has no effect (lane 4). Thus the mutation at the carboxy-terminal domain of
merlin has no effect on the conformation of merlin which is captured by the crosslinkers or on the effect of PIP on the proteins conformation as assayed by crosslinking.
Another similarity between ERM proteins and merlin which we have shown
is that merlin interacts with NHE-RF (see Murthy et al., 1998, in the appendix). The
binding properties are not identical to those seen with ERM proteins. NHE-RF
interacts with the amino-terminal domain of radixin more efficiently than with
full-length radixin (Chapter 3). Unlike radixin, both the amino-terminal domain of
merlin and full-length merlin interact equally well with NHE-RF. The exact
consequences of this discrepancy is not known, but is something that should be
explored (See Future Prospects, Chapter 7).
Chapter 5
146
Merlin
Figure 5-1 Merlin forms an intramolecular interaction which can be captured by
cross-linkers
A, Merlin forms an intramolecular interaction. GST-merlin was purified
and thrombin cleaved to remove the GST tag. The protein was incubated
with cross-linker, BS 3 in the presence or absence of PIP for 10 minutes. The
reaction was quenched with excess amines and the samples boiled in GSD. I
analyzed the proteins on a 7.5% SDS-PAGE, and detected merlin by
immunoblotting. The arrowhead indicates the molecular weight of purified
protein. The arrow indicates the molecular weight of a cross-linker
dependent faster mobility band.
B, A mutant isoform of merlin which has a missense mutation at position
220 (Merlin 220) also forms an intramolecular interaction. GST-merlin 220
was purified and thrombin cleaved to remove the GST tag. The purified
protein was incubated with a covalent cross-linker in the absence or presence
of phospholipids, PIP or PC, as described in figure 5-1A. The samples were
boiled in GSD and separated on a 7.5% SDS-PAGE. The arrowhead indicates
the molecular weight of purified protein and the arrow indicates the
molecular weight of a captured intramolecular interaction.
Chapter 5
147
Merlin
A)
Cross-linker :
- + +
PIP:
209
-
-
+
-
137-
84-
2 3
1
B)
Phospholipid:
Cross-linker:
-
+ + +
.4
1234
DISCUSSION
Understanding merlin may help us understand ERM proteins and the roles
of these proteins in growth regulation. Our data suggest that a conformation of
merlin exists which is similar to radixin. This conformation is captured by covalent
protein cross-linkers. Furthermore, this intramolecular interaction can be disrupted
by the presence of specific phospholipids. This intramolecular interaction and its
regulation by phospholipids are not affected by a missense mutation at amino acid
220. Given the high identity between these proteins at the amino-terminal domain,
these results are not surprising. I think the key to understanding why merlin is a
tumor suppressor gene and ERM proteins are not is to compare the similarities and
differences between these proteins.
Several experiments show similar but distinct results between ERM proteins
and merlin. Expressing full-length merlin in fetal lung fibroblast cell line VA13
cells caused the formation of elongated cell processes (Koga 1998). This result is
similar to those with ERM proteins, although this phenotype is seen only after the
expression of the C-terminal domain alone and not the full-length protein (Algrain
et al., 1993; Henry et al., 1995). Thus the amino-terminal domains of ERM proteins
suppress the effect of the carboxy-terminal domains but the expression of full-length
merlin produces similar results. The difference in phenotypes between full-length
merlin and full-length ERM proteins may represent different mechanisms
converging to create similar phenotypes. The differences may also indicate that the
interdomain interaction within ERM proteins which blocks binding sites does not
Chapter 5
149
Merlin
occur in merlin. Resolving the reason for such similarities and differences between
these proteins may suggest which domains are important for organizing the
cytoskeleton in a regulated fashion.
McCartney et al (1996) showed that although both moesin and merlin are
expressed in Drosophila, they show distinct subcellular localizations. Merlin
localized in cells to punctate structures associated with the plasma membrane and
the cytoplasm. Moesin showed typical ERM protein localization - continuous
staining along the plasma membrane and cortical structures. This indicates that
these proteins may play different roles in Drosophila.
Disrupting the expression of the proteins also give similar but distinct results.
Use of antisense oligodeoxynucleotides in schwann-like STS26T cells suggested a
function for merlin. The expression of anti-sense pODNs inhibited cell spreading
and attachment (Huynh and Pulst, 1996). This phenotype was reversible - after the
removal of the antisense pODNs the cells attach. Furthermore, the phenotype seen
with the antisense pODNs is not due to cell viability (Huynh and Pulst, 1996). These
results are similar to antisense experiments done with ERM proteins although, the
appearance of a phenotype was detected only after disrupting the expression of two
or more of the ERM proteins simultaneously (Takeuchi, 1994). It may be that the
redundancy of ezrin radixin and moesin prevent the appearance of these proteins as
tumor suppressors. Yet it's interesting that although merlin is very similar to ERM
proteins, they cannot compensate for the loss of merlin.
The most intriguing aspect of merlin is its ability to act as a tumor suppressor.
Initially, merlin was classified as a tumor suppressor gene by genetics. Results from
Chapter 5
150
Merlin
in vitro studies also showed merlin acting as a tumor suppressor. Merlin
expression can cause the reversal of malignant phenotypes. A malignant phenotype
can be mimicked in NIH-3T3 cells by expressing v-HA-Ras. The growth of these
cells is anchorage independent and they have lost contact inhibition.
Overexpressing full-length merlin reversed these malignant phenotypes caused by
v-HA-Ras. In addition, the expression of the amino-terminal domain of merlin
also suppressed the malignant phenotype, but is much less effective (Tikoo et al.,
1994). The specificity of merlin to suppress growth in schwannoma cells was shown
by Sherman, L (1997). They reported that merlin, but not radixin can suppress the
growth of schwannoma cells in vitro and in vivo. Furthermore, this effect is
specific for the NF2 isoform which lacks exon 16 (NF2-17). NF2-16, which contains
exon 16, but not exon 17, has no effect on the growth of schwannoma cells. Thus it
is interesting that in the first assay the amino-terminal domain of merlin (which is
60% homologous to ERM proteins) can suppress a malignant phenotype, but in the
second assay, the crucial area of merlin seems to be in the last exon.
It is not surprising that merlin, a cytoskeletal protein, can act as a tumor
suppressor. Several other actin cytoskeleton associated proteins have tumor
suppressor capabilities. The overexpression of vinculin and x-actinin reverses SV40
induced malignant transformation (Fernandez et al., 1992; Gluck et al., 1993).
Tropomyosin 1 and gelsolin reverses the malignant transformation caused by Ras
mutants such as v-HA-Ras and v-ki-Ras (Mullauer et al., 1993; Prasad et al., 1993).
Catenin is another cytoskeletal associated protein which is involved in
tumorigenesis. Catenin links cadherin to actin filaments (Tsukita et al., 1992) and
Chapter 5
151
Merlin
the disruption of the cadherin-catenin complex is involved in the pathogenesis of
several human tumors. Therefore there is precedent for a cytoskeletal protein, such
as merlin, playing a role in tumorigenesis.
Chapter 5
152
Merlin
BIBLIOGRAPHY
Algrain, M., Turunen, 0., Vaheri, A., Louvard, D., and Arpin, M. (1993). Ezrin
contains cytoskeleton and membrane binding domains accounting for its proposed
role as a membran-cytoskeletal linker. J. Cell Biol. 120, 129-139.
den Bakker, M., Riegman, P., Hekman, R., Boersma, W., Janssen, P., van der Kwast,
T., and Zwarthoff, E. (1995). The product of the NF2 tumour suppressor gene
localizes near the plasma membrane and is highly expressed in muscle cells.
Oncogene 10, 756-763.
Fernandez, L. A., MacSween, J. M., You, C. K., and Gorelick, M. (1992). Immunologic
changes after blood transfusion in patients undergoing vascular surgery. Am J Surg
163, 263-9.
Gluck, U., Kwiatkowski, D. J., and Ben-Ze'ev, A. (1993). Suppression of
tumorigenicity in simian virus 40-transformed 3T3 cells transfected with alphaactinin cDNA. Proc Natl Acad Sci U S A 90, 383-7.
Gonzalez, -. A., C., Xu, L., Pinney, D., Beauchamp, R., Hobbs, W., Gusella, J., and
Ramesh, V. (1996). The merlin tumor suppressor localizes preferentially in
membrane ruffles. Oncogene 13, 1239-1247.
Henry, M., Gonzalez-Agosti, C., and Solomon, F. (1995). Molecular dissection of
radixin: Distinct and interdependent functions of the amino- and carboxy-terminal
domains. Journal of Cell Biology 129, 1007-1022.
Huang, L., Ichimaru, E., Pestonjamasp, K., Cui, X., Nakamura, H., Lo, G. Y., Lin, F. I.,
Luna, E. J., and Furthmayr, H. (1998). Merlin differs from moesin in binding to Factin and in its intra- and intermolecular interactions. Biochem Biophys Res
Commun 248, 548-53.
Huynh, D. P., and Pulst, S. M. (1996). Neurofibromatosis 2 antisense
oligodeoxynucleotides induce reversible inhibition of schwannomin synthesis and
cell adhesion in STS26T and T98G cells. Oncogene 13, 73-84.
MacCollin, M., Mohney, T., Trofatter, J., Wertelecki, W., Ramesh, V., and Gusella,
(1993). DNA diagnosis of neurofibromatosis 2. Altered coding sequence of the
merlin tumor suppressor in an extended pedigree [published erratum appears in
JAMA 1994 Oct 12;272(14):1104]. Jama 270, 2316-20.
J.
Martuza, R. L., and Eldridge, R. (1988). Neurofibromatosis 2 (bilateral acoustic
neurofibromatosis). N Engl J Med 318, 684-8.
Mautner, V. F., Baser, M. E., and Kluwe, L. (1996). Phenotypic variability in two
families with novel splice-site and frameshift NF2 mutations. Hum Genet 98, 203-6.
Chapter 5
153
Merlin
Mullauer, L., Fujita, H., Ishizaki, A., and Kuzumaki, N. (1993). Tumor-suppressive
function of mutated gelsolin in ras-transformed cells. Oncogene 8, 2531-6.
Pestonjamasp, K., Amieva, M., Strassel, C., Nauseef, W., Furthmayr, H., and Luna,
E. (1995). Moesin, ezrin and p205 are actin-binding proteins associated with
neutrophil plasma membranes. Mol. Biol. Cell 6, 247-259.
Prasad, G. L., Fuldner, R. A., and Cooper, H. L. (1993). Expression of transduced
tropomyosin 1 cDNA suppresses neoplastic growth of cells transformed by the ras
oncogene. Proc Natl Acad Sci U S A 90, 7039-43.
Rouleau, G. A., Merel, P., Lutchman, M., Sanson, M., Zucman, J., Marineau, C.,
Hoang-Xuan, K., Demczuk, S., Desmaze, C., and Plougastel, B. e. a. (1993). Alteration
in a new gene encoding a putative membrane-organizing protein causes
neurofibromatosis type 2. Nature 363, 515-521.
Shuster, C., and Herman, I. (1995). Indirect association of ezrin with F-actin: isoform
specificity and calcium sensitivity. Jour. Cell. Biol. 128, 837-848.
Takeuchi, K., Naruki Sato, Hideko Kasahara, Noriko Funayama, Akira Nagafuchi,
Shigenobu Yonemura, Sachiko Tsukita and Shoichiro Tsukita (1994). Perturbation
of Cell Adhesion and Microville Formation by Antisense Oligonucleotides to ERM
Family Members. Journal of Cell Biology 125, 1371-1384.
Tikoo, A., Varga, M., Ramesh, V., Gusella, J., and Maruta, H. (1994). An anti-ras
function of neurofibromatosis type 2 gene product (NF2/Merlin). Jour. Biol. Chem.
269, 23387-23390.
Trofatter, J. A., MacCollin, M. M., Rutter, J. L., Murrell, J. R., Duyao, M. P., Parry, D.
M., Eldridge, R., Kley, N., Menon, A. G., Pulaski, K., Haase, V. H., Ambrose, C. M.,
Munroe, D., Bove, C., Haines, J. L., Martuza, R. L., MacDonald, M. E., Seizinger, B. R.,
Short, M. P., Buckler, A. J., and Gusella, J. F. (1993). A novel moesin-, ezrin-, radixinlike gene is a candidate for the neurofibromatosis 2 tumor suppressor. Cell 72, 791800.
Tsukita, S., Nagafuchi, A., and Yonemura, S. (1992). Molecular linkage between
cadherins and actin filaments in cell-cell adherens junctions. Curr Opin Cell Biol 4,
834-9.
Turunen, 0., Wahlstrom, T., and Vaheri, A. (1994). Ezrin has a COOH-terminal
actin-binding site that is conserved in the ezrin protein family. Journal of Cell
Biology 126, 1445-1453.
Yao, X., Chaponnier, C., Gabbiani, G., and Forte, J. G. (1995). Polarized distribution of
actin isoforms in gastric parietal cells. Mol Biol Cell 6, 541-57.
Chapter 5
154
Merlin
Yao, X., Cheng, L., and Forte, J. (1996). Biochemical characterization of ezrin-actin
interaction. Jour. Biol. Chem. 271, 7224-7229.
Chapter 5
155
Merlin
CHAPTER 6:
MODEL AND CONCLUSIONS
In this study, we have characterized the binding between radixin and two
ligands: layilin, a novel transmembrane protein (Borowsky and Hynes, 1998); and
NHE-RF, a protein which was originally identified as a regulatory co-factor for a
Na+/H+ exchanger (Weinman et al., 1993). Using these two proteins, we studied
the regulation of ligand binding to ERM proteins. The binding properties for layilin
and NHE-RF were very similar in a variety of binding experiments with radixin.
The binding sites for both ligands mapped to the amino-terminal domain of radixin
and both ligands bound more strongly to a protein fragment containing only the
amino-terminal domain of radixin than to the full-length protein. Our data show
that this difference in binding may be due to an interdomain interaction since the
ligand binding to the amino-terminal domain is blocked by the presence of carboxyterminal domain in trans. Furthermore, we show that PIP changes the
conformation of the amino-terminal domain and disrupts the interaction between
the amino- and carboxy-terminal domains. The elimination of this interdomain
interaction makes the binding sites in radixin more available to its ligands. By
controlling intermolecular interactions, the cell can control the recruitment of
proteins to specific regions of the cell (Figure 6-1).
Several other cytoskeletal proteins show a discrepancy between binding
activities between the full-length protein and fragments of the protein. Talin has an
increased affinity for the head domain of vinculin compared to the intact vinculin
molecule (Johnson and Craig, 1994). Individual domains of ERM proteins bind to
Chapter 6
157
Model & Conclusions
actin, Rho-GDI and CD44 more efficiently than to the full-length ERM proteins
(Turunen et al., 1994; Hirao et al., 1996; Takahashi et al., 1997). In this paper, we
show that the binding of NHE-RF and layilin to the N-terminal domain of radixin is
stronger than to full-length radixin (Figure 3-2 and 4-2). This is not due to the
concentration of the proteins since the interaction between NHE-RF and the Nterminal domain of radixin saturates at a stoichiometry of about 1:1 whereas the
interaction between NHE-RF and full-length radixin does not reach saturation at
any of the concentrations of NHE-RF we tested (Figure 3-4). We have shown that
the decrease in binding between full-length radixin and its ligands may be due to an
interdomain interaction (Figure 3-6 and 4-3). An interdomain interaction also
blocks the binding of ezrin to NHE-RF/EBP50 (Reczek and Bretscher, 1998) and
radixin binding to protein 4.1 (Magendantz et al., 1995). The regulation of this
interdomain interaction may modulate ligands binding to radixin.
Phospholipids play a role in the regulation of several protein-protein
interactions (Johnson and Craig, 1995; Hirao et al., 1996; Heiska et al., 1998). In this
paper, we show that the presence of PIP enhances ligand binding to full-length
radixin (Figure 3-7 and 4-4).
One possible way phospholipids can have this effect is
by decreasing an inhibitory activity such as an interdomain interaction which blocks
the binding sites present within the domains of the protein. Our data show that the
interdomain interaction between the carboxy- and amino-terminal domain of
radixin in trans is significantly disrupted in the presence of PIP (Figure 2-3).
This
suggests that the effect of PIP is within one domain of the protein and is not a global
effect which requires an intact protein. Since PIP binds to radixin at its amino-
Chapter 6
158
Model & Conclusions
terminal domain (Figure 2-2) it may change the affinity between the two domains of
radixin by causing a conformational change within the amino-terminal domain.
We have developed an assay for the conformation of radixin. An
intramolecular interaction within radixin can be captured by covalent cross-linkers
(Figure 2-4). This cross-linked form is detected as a faster mobility band on SDSPAGE. The formation of this intramolecular interaction is fast, reversible and
dependent on the conformation of the protein (Figures 2-5 and 2-6). We have
localized the region of the protein responsible for the shift in molecular weight to
the amino-terminal domain of radixin (Figure 2-7). The cross-linked form of the
amino-terminal domain has the ability to bind ligands as well as the carboxyterminal domain of radixin. The form of the amino-terminal domain which does
not form the faster mobility band - that is in the presence of PIP - does not bind to
the carboxy-terminal domain (Figure 2-3). Thus using a cross-linking assay, we can
detect two conformations of radixin, one which can be cross-linked to form a faster
mobility band on an SDS-PAGE and another which cannot.
From these results, we propose a model for the mechanism by which PIP
affects ERM protein-ligand interactions (Figure 6-1). The conformation of the
amino-terminal domain which forms an intra-molecular interaction is conducive
to ligand binding as well as binding to the carboxy-terminal domain. Thus the
ligand must compete with the carboxy-terminal domain to bind to radixin. In the
presence of phospholipids, the conformation of the amino-terminal domain
changes and it does not form an intra-molecular interaction. In this conformation,
the affinity between the amino-terminal domain and the carboxy-terminal domain
Chapter 6
159
Model & Conclusions
decreases but the interaction between the ligand and radixin is not affected. Once
this interdomain interaction is disrupted by phospholipids, there is a shift in
equilibrium which increases ligand binding to the full-length protein.
In conclusion, we characterized radixin binding to two different ligands,
layilin and NHE-RF, and the regulation of these interactions by PIP. This regulation
is manifested by a change in the conformation of the amino-terminal domain due
to PIP binding. The conformational change decreases an interdomain interaction
which makes the binding sites for the ligands available. Thus the amino-terminal
domain appears to interpret PIP signals to control radixin interaction with its
ligands. The interaction between ERM proteins and membrane associated proteins
may be promiscuous - that is it binds to several different proteins at the membrane
and not to a single ligand. A few of the ligands for ERM proteins share an ERM
protein binding motif, but this motif is not conserved in all of the ligands. There
may be a reason for this promiscuous binding. By not having to search for a specific
ligand at the membrane, ERM proteins can quickly attach since it can bind to several
different proteins. This may be a crucial aspect of cytoskeletal rearrangement in
active regions of the cell such as ruffling edges.
From signal to cytoskeletal reorganization
The work presented in this thesis gives us a hint of how the cytoskeleton may
be reorganized in response to an extracellular signal. As I stated above, the
intermolecular interaction between ERM proteins and their ligands is enhanced by
the presence of phospholipids. Thus phospholipids may be a very important
Chapter 6
160
Model & Conclusions
regulator for the formation of cortical structures. Several labs have shown this to be
true. When hepatocytes are treated with wortmannin, a specific inhibitor for PI 3kinase, the levels of phosphoinositides is decreased and the formation of microvilli
is inhibited (Lange et al., 1998). This suggests a connection between the formation of
cortical structures and phosphoinositides. Stimulation of cells by serum derived
factors such as lysophosphatidic acid (LPA) affects the organization of actin
structures in the cell. LPA stimulation also increases the PIP levels in human
platelets without changing the PI or PIP2 levels (Mani et al., 1996). Some of the most
intriguing data is the discovery that enzymes such as PI 4-kinase and PI 4-phosphate
5-kinase activity are elevated in carcinoma cells (Weber et al., 1996). These data
suggest that the phospholipid levels are crucial in organizing the cytoskeletal
structures and may have a role in carcinogenesis.
Although my thesis focuses on the affect of phospholipids on ERM proteins,
several other cytoskeletal proteins are affected by phosphoinositides. For example,
ERM proteins not only share homology to band 4.1, but these proteins also share
binding properties. Band 4.1 was one of the first proteins to be identified as a crucial
protein for the organization of the membrane cytoskeleton. Band 4.1 increases the
association between actin and spectrin (Fowler and Taylor, 1980) and it binds to the
membrane through its association with glycophorin (Anderson and Lovrien, 1984)
and band 3 (Pasternack et al., 1985). Through these interactions, band 4.1 is thought
to be involved in the maintenance of cell shape. The interaction of band 4.1 and
glycophorin is regulated by specific phospholipids. Phosphatidyl inositol 4,5
bisphosphate (P14,5P), phosphatidyl inositol 4-phosphate (PI4P) and to a lesser
Chapter 6
161
Model & Conclusions
degree, phosphatidic acid and phosphatidyl serine (PS) increase the interaction
between band 4.1 and glycophorin (Anderson and Marchesi, 1985). Thus the
presence of phosphoinositides regulates protein-protein interactions between band
4.1 and its ligands in a similar manner as with ERM proteins and their ligands.
Another cytoskeletal protein which is affected by phosphoinositides is
vinculin. Vinculin is also thought to act as a linking protein between the actin
cytoskeletal network and the cytoplasmic face of the cell membrane. Vinculin is
very similar to ERM proteins in that binding sites for its ligands, such as talin and
actin, are masked in the full length protein but are seen with fragments of vinculin
(Johnson and Craig, 1994; Johnson and Craig, 1995). Furthermore, vinculin binds to
acidic phospholipids (Goldmann et al., 1995) and the binding of these phospholipids
decreases the head - tail interaction within vinculin (Weekes et al., 1996) and may
increase the binding of vinculin with its ligands by disrupted the head-tail
interaction. Although vinculin is found in cell-cell contacts and cell-adhesion sites
whereas ERM proteins are found in microvilli, filopodia and ruffling edges,
vinculin and ERM proteins use similar strategies to regulate their interactions with
their ligands. Thus phospholipid metabolism may be important for many different
interactions near the membrane.
In Balb/c 3T3 cells, PIP2 was found to bind alpha-actinin as well as vinculin.
Using antibodies specific for PIP2 in western blot analysis, alpha-actinin and
vinculin were identified as PIP2 abundant proteins (Fukami et al., 1994). The
authors also showed that alpha-actinin in the cytoskeletal fraction was bound to
PIP2, but alpha-actinin in the cytosolic fraction was not (Fukami et al., 1994). Alpha-
Chapter 6
162
Model & Conclusions
actinin is known to bind and cross-link actin, increasing its viscosity in solution.
The ability of alpha-actinin to cross-link actin increases in the presence of PIP2
(Fukami et al., 1994). Thus phosphoinositides not only affect the interaction
between linker proteins and their ligands, but also cytoskeletal proteins to actin.
Lastly, another important cytoskeletal protein known to be affected by
phospholipids is profilin. Profilin binds to both actin monomers and
polyphosphoinositides (Chaudhary et al., 1998). Profilin is a small (12-15 KDa),
ubiquitous protein which sequesters monomeric actin in a 1:1 complex. Profilin
also decreases the critical concentration at the barbed end of actin filaments and
promotes actin polymerization when the barbed ends are free (Carlier and
Pantaloni, 1997; Schluter et al., 1997). Phospholipids also have an effect on profilin
interactions. Profilin binds to PIP2 at physiological conditions (Machesky et al., 1990)
and the binding of PIP2 and actin to profilin are mutually exclusive (Lassing and
Lindberg, 1985). Furthermore, the presence of phosphoinositides inhibit the effects
of profilin on actin polymerization (Goldschmidt-Clermont et al., 1991) - that is
profilin binding to actin decreases and actin polymerization increases.
Taking all these data together, we can speculate how an extracellular signal
coordinates the reorganization of the actin cytoskeleton. One can imagine that an
extracellular signal such as LPA stimulates the cell and increases the local
concentration of phosphoinositides by increasing the activities of the signal
transduction enzymes - PI kinase and PIP kinase. In these specific regions,
phosphoinositides bind to the cytoskeletal proteins. Profilin binds to the
phosphoinositides and releases the actin monomers. ERM proteins bind to
Chapter 6
163
Model & Conclusions
phosphoinositides which changes it to the "open" conformation and links
cytoskeletal proteins to the membrane. This causes the local rearrangement of the
actin cytoskeleton including possibly actin cross-linking by alpha-actinin. The same
process may occur with vinculin or band 4.1 instead of ERM proteins, but this may
occur in different regions of the cell or in different cell types. At some point, PIP2 is
hydrolyzed by phospholipase Cy and produces secondary messengers IP3 and DAG.
At the same time without the phospholipids bound to the proteins the linker
proteins may release their ligands. This may be one method for creating a
temporary connection between the cytoskeleton and the membrane in regions of the
cell such as ruffling edges where the actin structures are changing and not
permanent.
This is just one possible scenario. I have not included what effect
phosphorylation would have on ERM proteins or other proteins as the Rho-signal
transduction pathway is triggered. This is discussed in the introduction. Needless
to say, the process gets more complicated as the activities of different proteins are
inhibited or stimulated upon phosphorylation. The effect of phosphorylation in
conjunction with phospholipid binding is just one aspect of ERM protein - ligand
binding that could be addressed in the future. In the next chapter, I discuss other
avenues that can be explored and continued from this work.
Chapter 6
164
Model & Conclusions
Figure 6-1 Model for ERM regulation
PIP regulates ligand binding to radixin by changing the conformation of the
amino-terminal domain. In the closed conformation, an amino-carboxy
interdomain interaction blocks the ligand binding sites available within the
fragments of radixin. The presence of PIR changes the conformation of the Nterminal domain by inhibiting an intra-molecular interaction and this
decreases the interdomain interaction. The previously occluded binding sites
are then available to its ligands and this allows full-length ERM proteins to
bind to other proteins such as NHE-RF and layilin.
Chapter 6
165
Model & Conclusions
model slide
Chapter 6
166
Model & Conclusions
Radixin
A4,44 PIP
NHE
Layilin
NHE-RF
Radixin
Actin
Actin
BIBLIOGRAPHY
Anderson, R. A., and Lovrien, R. E. (1984). Glycophorin is linked by band 4.1 protein
to the human erythrocyte membrane skeleton. Nature 307, 655-658.
Anderson, R. A., and Marchesi, V. T. (1985). Regulation of the association of
membrane skeletal protein 4.1 with glycophorin by a polyphosphoinositide. Nature
318, 295-298.
Borowsky, M. L., and Hynes, R. 0. (1998). Layilin, a novel talin-binding
transmembrane protein homologous with C- type lectins, is localized in membrane
ruffles. J Cell Biol 143, 429-42.
Carlier, M. F., and Pantaloni, D. (1997). Control of actin dynamics in cell motility. J
Mol Biol 269, 459-67.
Chaudhary, A., Chen, J., Gu, Q. M., Witke, W., Kwiatkowski, D. J., and Prestwich, G.
D. (1998). Probing the phosphoinositide 4,5-bisphosphate binding site of human
profilin I. Chem Biol 5, 273-81.
Fowler, V., and Taylor, D. L. (1980). Spectrin plus band 4.1 cross-link actin.
Regulation by micromolar calcium. J Cell Biol 85, 361-76.
Fukami, K., Endo, T., Imamura, M., and Takenawa, T. (1994). alpha-Actinin and
vinculin are PIP2-binding proteins involved in signaling by tyrosine kinase. J Biol
Chem 269, 1518-22.
Goldmann, W. H., Senger, R., Kaufmann, S., and Isenberg, G. (1995). Determination
of the affinity of talin and vinculin to charged lipid vesicles: a light scatter study.
FEBS Lett 368, 516-8.
Goldschmidt-Clermont, P. J., Machesky, L. M., Doberstein, S. K., and Pollard, T. D.
(1991). Mechanism of the interaction of human platelet profilin with actin. J Cell
Biol 113, 1081-9.
Heiska, L., Alfthan, K., Gronholm, M., Vilja, P., Vaheri, A., and Carpen, 0. (1998).
Association of ezrin with intercellular adhesion molecule-1 and -2 (ICAM-1 and
ICAM-2). Regulation by phosphatidylinositol 4, 5- bisphosphate. J Biol Chem 273,
21893-900.
Hirao, M., Sato, H., Kondo, T., Yonemura, S., Monden, M., Sasaki, T., Takai, Y.,
Tsukita, S., and Stukit, S. (1996). Regulation mechanism of ERM
(ezrin/radixin/moesin) protein/plasma membrane association: possible
involvement of phosphatidylinositol turnover and Rho-dependent signaling
pathway. Jour. Cell Biol. 135, 37-51.
Chapter 6
167
Model & Conclusions
Johnson, R. P., and Craig, S. W. (1994). An intramolecular association between the
head and tail domains of vinculin modulates talin binding. Journal of Biological
Chemistry 269, 12611-12619.
Johnson, R. P., and Craig, S. W. (1995). F-actin binding site masked by the
intramolecular association of vinculin head and tail domains. Nature 373, 261-264.
Lange, K., Brandt, U., Gartzke, J., and Bergmann, J. (1998). Action of insulin on the
surface morphology of hepatocytes: role of phosphatidylinositol 3-kinase in insulininduced shape change of microvilli. Exp Cell Res 239, 139-51.
Lassing, I., and Lindberg, U. (1985). Specific interaction between phosphatidylinositol
4,5-bisphosphate and profilactin. Nature 314, 472-475.
Machesky, L. M., Goldschmidt-Clermont, P. J., and Pollard, T. D. (1990). The affinities
of human platelet and Acanthamoeba profilin isoforms for polyphosphoinositides
account for their relative abilities to inhibit phospholipase C. Cell Regul 1, 937-50.
Magendantz, M., Henry, M., Lander, A., and Solomon, F. (1995). Inter-domain
interactions of radixin in vitro. Jour. Biol. Chem. 270, 25324-25327.
Mani, I., Gaudette, D. C., and Holub, B. J. (1996). Increased formation of
phosphatidylinositol-4-phosphate in human platelets stimulated with
lysophosphatidic acid. Lipids 31, 1265-8.
Pasternack, G. R., Anderson, R. A., Leto, T. L., and Marchesi, V. T. (1985).
Interactions between protein 4.1 and band 3. J. Biol. Chem. 260, 3676-3683.
Reczek, D., and Bretscher, A. (1998). The carboxyl-terminal region of EBP50 binds to a
site in the amino- terminal domain of ezrin that is masked in the dormant
molecule. J Biol Chem 273, 18452-8.
Schluter, K., Jockusch, B. M., and Rothkegel, M. (1997). Profilins as regulators of
actin dynamics. Biochim Biophys Acta 1359, 97-109.
Takahashi, K., Sasaki, T., Mammoto, A., Takaishi, K., Kameyama, T., Tsukita, S., and
Takai, Y. (1997). Direct interaction of the Rho GDP dissociation inhibitor with
ezrin/radixin/moesin initiates the activation of the Rho small G protein. J Biol
Chem 272, 23371-5.
Turunen, 0., Wahlstrom, T., and Vaheri, A. (1994). Ezrin has a COOH-terminal
actin-binding site that is conserved in the ezrin protein family. Journal of Cell
Biology 126, 1445-1453.
Chapter 6
168
Model & Conclusions
Weber, G., Shen, F., Prajda, N., Yeh, Y. A., Yang, H., Herenyiova, M., and Look, K. Y.
(1996). Increased signal transduction activity and down-regulation in human cancer
cells. Anticancer Res 16, 3271-82.
Weekes, J., Barry, S. T., and Critchley, D. R. (1996). Acidic phospholipids inhibit the
intramolecular association between the N- and C-terminal regions of vinculin,
exposing actin-binding and protein kinase C phosphorylation sites. Biochem J 314,
827-32.
Weinman, E. J., Steplock, D., Corry, D., and Shenolikar, S. (1993). Identification of
the human NHE-1 form of Na(+)-H+ exchanger in rabbit renal brush border
membranes. J Clin Invest 91, 2097-102.
Chapter 6
169
Model & Conclusions
CHAPTER 7:
FUTURE PROSPECTS
There are several experiments that can be done to increase our understanding
of the interactions between radixin and its ligands and their roles in cell motility
and morphology.
Layilin studies
Due to the scarcity of the protein, I was not able to do several experiments
with layilin that I was able to do with NHE-RF. By a rough estimation, the binding
affinity between radixin and layilin seems lower than NHE-RF and radixin. To
confirm this, one can do a titration study between radixin and layilin. The
GSTLayCyt construct was a generous gift from Mark Borowsky. With the construct
in hand, we can make large protein preparations of GSTLayCyt. From these studies
we might be able to determine an approximate Kd and use this to compare the
binding affinity of each of these ligands to radixin. It would be interesting to know if
one protein had a stronger affinity for radixin. This may be one way ERM proteins
determine which ligand it binds to at the cell membrane.
Layilin and NHE-RF combination studies
After the dissociation constant is identified for layilin, you can test if there are
differences between layilin and NHE-RF in affinity chromatography binding assays.
Can the two ligands, NHE-RF and layilin bind radixin at the same time or is the
binding mutually exclusive? Can one ligand displace the other? Does this
correspond with the calculated binding constants?
Chapter 7
171
Future Prospects
Map the binding sites for the ligands
If the binding of NHE-RF and layilin are mutually exclusive, is this due to
overlapping binding sites or a conformational change in radixin? The binding sites
can be mapped by using deletion constructs of the proteins in the binding assays.
Once the ligand binding sites are identified in radixin, we can determine if the
sequence is conserved in the other ERM proteins. Once the radixin binding sites on
the ligands are identified, we can determine if these are similar to one another and
if there are other proteins with similar sequences.
Testing the effect of pH on ligand interactions
The interaction between NHE-RF and radixin suggests an intriguing aspect of
regulation of protein-protein interactions. Since NHE-RF is a regulatory factor for a
Na+/H+ exchanger, it may recruit radixin to regions of the cell which have a
fluctuating pH as compared to other regions of the cell. Studies have shown that pH
is important for several cellular activities (see chapter 3). Thus one can check the
binding between NHE-RF and radixin using buffers with varying pH. Is the NHERF/radixin interaction more sensitive than the layilin/radixin interaction to a
change in pH?
Testing other modes of regulation
Binding to phospholipids may not be the exclusive method of regulating
ligand binding to radixin. Even in the presence of PIP, ligand binding to radixin is
not saturated and does not reach the same levels of ligand binding as the aminoChapter 7
172
Future Prospects
terminal domain. One possibility is that a population of radixin is denatured and is
intractable to PIP regulation. Another possibility is that PIP binding to radixin is not
saturated. A titration of PIP concentrations up to 50gM showed increasing NHE-RF
or layilin binding to full-length radixin (Chapters 3 and 4). Yet another possibility is
that a second signal enhances an intermolecular interaction. The phosphorylation
of ERM proteins may also play a role in regulating intermolecular interactions
(Takahashi 1997, Fukata 1998, Shaw 1998, Matsui 1998). It may be a concerted effort
between the phosphorylation event and binding to phospholipids that stabilizes
ERM protein binding to its ligands at specific regions of the cell. We may test this
hypothesis by phosphorylating the protein and performing the binding studies in
the presence of phospholipids. It would not be surprising if the two signals worked
simultaneously on ERM proteins since the Rho signalling pathway not only
phosphorylates proteins but also triggers the production of phospholipids.
In vivo experiments
One of the main questions in the field is how ERM proteins function in vivo.
The identification of NHE-RF and layilin presents a way to approach this question.
As I have shown in my thesis, the interaction between radixin and these ligands can
be detected in in vitro binding assays. These interactions may be identified in vivo
by immunoprecipitating the complex. Once immunoprecipitation conditions are
worked out, the formation of the complex can be compared in several different cell
types and under several different conditions. For example, the amount of NHERF/radixin complex immunoprecipitated from a population of wounded and
Chapter 7
173
Future Prospects
unwounded cells can be compared. The ratio of layilin/radixin complex or NHERF/radixin complex may change in P19 cells which have been stimulated to form
neuronal or fibroblast cells. One can ask if the phosphorylation of radixin changes
and if this affects the formation of these complexes after serum starvation and LPA
stimulation of the cells.
Merlin studies
It is still unclear how many properties merlin shares with other ERM
proteins. The high homology at the amino-terminal domain and the divergence of
the carboxy-terminal domain presents a variety of intriguing possibilities of shared
and distinct properties. By understanding the differences and similarities, we may
understand how merlin acts as a tumor suppressor but ERM proteins do not. One of
the questions you may ask is if the amino- and carboxy-terminal domains of ERM
proteins and merlin are interchangeable. That is, can the carboxy-terminal domain
of merlin compete with NHE-RF and layilin binding to the amino-terminal domain
of radixin as was shown for the carboxy-terminal domain of radixin? Do the
different isoforms of merlin act differently in the binding assays? You can also
compare the binding affinities between merlin and NHE-RF or layilin. Is it much
different from the binding affinities of these ligands to radixin?
By identifying the
differences and similarities between radixin and merlin, we may be able to
determine why merlin is a tumor suppressor and ERM proteins are not.
Chapter 7
174
Future Prospects
CHAPTER 8:
APPENDIX - PUBLICATIONS
PUBLICATIONS
Murthy, A., Gonzalez-Agosti, C., Cordero, E., Pinney, D., Candia, C., Solomon,
F., Gusella, J., and Ramesh, V. (1998). NHE-RF, a regulatory cofactor for Na+H+ Exchange, is a common interactor for Merlin and ERM (MERM) proteins.
J. Biol. Chem. 273, 1273-1276.
Hubert, K., Cordero, E., Frosch, M., and Solomon, F. (1998). Activities of the
EM10 protein from Echinococcus multilocularis in cultured mammalian cells
demonstrate functional relationships to ERM family members. (in press, Cell
Motility and the Cytoskeleton)
Cordero, E., Borowsky, M., Gonzalez-Agosti, C., Ramesh, V., Hynes, R., and
Solomon, F. Regulation and binding analysis of radixin: a cytoskeletonmembrane linking protein to NHE-RF and layilin. (submitted)
Chapter 8
176
Publications
THE JouRNAL OF BIOLOGICAL cHEMISTRY
Communication
Vol. 273, No. 3, Issue of January 16, pp. 1273-1276, 1998
C 1998 by The American Society for Biochemistry and Molecular Biology, Inc.
Printed in U.S.A.
NHE-RF, a Regulatory Cofactor
for Na*-H+ Exchange, Is a
Common Interactor for Merlin
and ERM (MERM) Proteins*
(Received for publication, October 22, 1997)
Anita Murthyt, Charo Gonzalez-Agostit§,
Etchell Corderol, Denise Pinney, Cecilia Candia,
Frank Solomon1, James Gusella, and
Vijaya Rameshil
From the MolecularNeurogenetics Unit, Massachusetts
GeneralHospital, Charlestown, Massachusetts 02129
and the Center for Cancer Research, Massachusetts
Institute of Technology, Cambridge,Massachusetts
02139
We have identified the human homologue of a regulatory cofactor of Na*-H* exchanger (NHE-RF) as a novel
interactor for merlin, the neurofibromatosis 2 tumor
suppressor protein. NHE-RF mediates protein kinase A
regulation of Na'-H' exchanger NHE3 to which it is
thought to bind via one of its two PDZ domains. The
carboxyl-terminal region of NHE-RF, downstream of the
PDZ domains, interacts with the amino-terminal protein
4.1 domain-containing segment of merlin in yeast twohybrid assays. This interaction also occurs in affinity
binding assays with full-length NIIE-RF expressed in
COS-7 cells. NHE-RF binds to the related ERM proteins,
moesin and radixin. We have localized human NHE-RF
to actin-rich structures such as membrane ruffles, microvilli, and filopodia in HeLa and COS-7 cells, where it
co-localizes with merlin and moesin. These findings suggest that hNHE-RF and its binding partners may participate in a larger complex (one component of which
might be a Na'-H+ exchanger) that could be crucial for
the actin filament assembly activated by the ERM proteins and for the tumor suppressor function of merlin.
Neurofibromatosis 2 (NF2), 1 is a dominantly inherited disorder characterized by bilateral occurrence of vestibular schwannomas and other brain tumors, especially meningiomas, and
schwannomas of other cranial nerves and spinal nerve roots
(1). The NF2 gene isolated by positional cloning encodes merlin, named for its striking similarity with moesin, ezrin, and
radixin, three closely related proteins commonly referred to as
* This work was supported in part by National Institutes of Health
Grants NS24279 and NIDCD 03354 (to A. M.), the Deafness Foundation, and Neurofibromatosis Inc. (Massachusetts Chapter). The costs of
publication of this article were defrayed in part by the payment of page
charges. This article must therefore be hereby marked "advertisement"
in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
t The first two authors contributed equally to this work.
§ Supported in part by an young investigator award from the National Neurofibromatosis Foundation.
tiTo whom correspondence should be addressed: Molecular Neurogenetics Unit, Massachusetts General Hospital, Bldg. 149, 13th St.,
Charlestown, MA 02129. Tel.: 617-724-9733; Fax: 617-726-5736.
' The abbreviations used are: NF2, neurofibromatosis 2; MERM, merlin, ezrin, radixin, moesin; NHE-RF, regulatory cofactor of Na+-H+
exchanger; NHE1-5, Na'-H' exchanger isoforms; aa, amino acid(s);
GST, glutathione S-transferase; HA, hemagglutinin; GFP, green fluorescent protein; PAGE, polyacrylamide gel electrophoresis.
This paper is available on line at http://www.jbc.org
the ERM family, a subclass of the protein 4.1 superfamily
thought to link cytoskeletal components with proteins in the
cell membrane (2, 3). The ERM proteins share -78% amino
acid identity with each other, and all three are 45-47% identical to merlin (4).
In cultured cells, ERM proteins are highly concentrated in
regions of contact between actin filaments and the plasma
membrane, acting as possible linkers between integral membrane and cytoskeletal proteins (5-9). The carboxyl termini of
both ezrin and moesin bind directly to actin in vitro (10, 11) via
a conserved actin binding site present in ezrin, moesin, and
radixin but not in merlin. These findings suggest that the
carboxyl terminus of the ERM proteins is responsible for their
association with the actin-based cytoskeleton. Recently, however, another actin binding site in the amino-terminal domain
of ezrin has been characterized in vitro and shown to be conserved in moesin, radixin, and merlin (12). The highly conserved amino-terminal half of the ERM proteins also contains
the domain responsible for interaction with membrane proteins, particularly the glycoprotein CD44 (13). ERM-CD44 complexes are also associated with RhoGDI (RhoGDP dissociation
inhibitor) (14), and the ERM proteins have been directly implicated in Rho- and Rac-dependent cytoskeletal reorganization in
permeabilized cells (15).
We have reported that endogenous merlin localizes to the
actin-rich motile regions (i.e. leading and ruffling edges) in
human fibroblast and meningioma cells where it co-localizes
with actin but is not associated with stress fibers (16). Merlin
when overexpressed in cells, however, localizes to membrane
ruffles as well as to other actin-rich structures such as microvilli and filopodia thus resembling the ERM proteins (17).
Given the similarity between merlin and the ERM proteins, it
is likely that merlin's normal function also involves interactions with membrane and cytoskeletal components. We have
used the yeast two-hybrid interaction trap strategy to search
for proteins that may associate with merlin. Here, we describe
a human cDNA (merintc) that interacts not only with merlin,
but also with the ERM proteins via their conserved aminoterminal domain. This cDNA encoded the carboxyl terminus of
the human homologue of NHE-RF, which was originally identified as a cofactor that mediates protein kinase A inhibition of
a renal brush-border membrane Na'-H' exchanger in rabbit
(18). Exogenously expressed human NHE-RF also co-localizes
with merlin and moesin. Our findings suggest that NHE-RF is
a biologically significant interactor for the MERM (merlin and
ERM) family of proteins and may be a participant in the activation of the Na'-H' exchange required for actin cytoskeleton
reorganization. During the preparation of this manuscript a
report appeared showing that NHE-RF binds to ezrin (19).
EXPERIMENTAL PROCEDURES
MolecularCloning of Merintc Using the Yeast Two-hybrid SystemDNA encoding the merlin sequence (aa 8-595, isoformi) was polymerase chain reaction-amplified and cloned into the yeast LexA DNA binding vector pEG202 (20), and the plasmid was designated as pmerbait.
To identify proteins interacting with merlin isoform 1, yeast strain
EGY88 was sequentially transformed with the pmerbait and a human
fetal frontal cortex interaction library fused to the activation domain of
GAL4 in the plasmid pJG4-5 and selected on ura~, his-, trp~, leuplates. A partial merinte clone was obtained (see "Results") and subsequently used as a probe to screen a human fetal brain library in AZAP
(Stratagene) by standard techniques to obtain a full-length clone. These
clones were sequenced, and sequences were analyzed using software
1273
1274
NHE-RF Interacts with MERM Proteins
A
PDZl
-i
aa
FIG. 1. A, schematic representation of
hNHE-RF. Lines represent the 5'- and 3'untranslated region. The coding region is
indicated by the box. Shaded areas represent the two PDZ domains. The black area
(aa 290-358) represents the MERM binding domain. B, sequence comparison of
human NHE-RF with rabbit and mouse
NHE-RF and E3KARP. The amino acid
identities between these sequences are
boxed. The hNHE-RF sequence has been
deposited in GenBankTM, and the accession number is A5036241.
aa 149-236
11-97
B
MERM
PDZ2
\M
mm
1,01M
10
aa 290-358
20
30
hNHE-RF MSADAAAGAPI -'PRLCCLEKGPNYGF LGEKG
rNHE-RF M4SADAAAGAP -PRLCCLEKGPNGYGFWLHGEK
IRLCCLEKGPNGYGFHLHGEKG
MNHE-RF NSADAAA
LGEKGR
RP_;RL0
b
E3KARP
EQGYGFH
so
hGNHE-RF
90
40
50
60
70
QYIRLVEPGSPA KAGLLAGDRLVVNGENVEK
YIRLVEPGSPAEKAGLLAGDRLVEVNGENVEK
LLAGDRLVEVNGENVEK
IRLVEPGSPAEC
iL* AGDRLVEVNGVNVE
I 8VEPGSPAR
110
100
L
E
rNHE-RF ETHQQVVSRIRAALNAVRLLVVDPETDE
MNHE-RF ETHQQVVSRIRAALNAVRLLVVDPETD
EPWTQV.R IQVQqLEV4NQRETLPD
E3KA
ISO
160
|DKSNPE ELRP0LCTMKXG
hNNE-RF
rNHE-RF VEKSH -.
mNHE-RF NEKSH-KKDVSGP
E3KARP
E-LNPRLC
KXG
ELRPRLCTMKKG
.ERPRTL K
220
hNHE-RF KQHGDVVSAI
rNHE-RF KQHGDVVjA
mNHE-RF KQHGDVVSAIC
E3KARP LRHAV
- 230
GODKLL
KLL
D
L
D
L
300
290
hNKE-RF SPRPALIF SASSDTSEELNSQDSP
rNHE-RF SPRPALAIRSASSDTSEEIASQDS
mNHE-RF SPRPALSASSDTSEELNSDS
- -SDIEDGSAWK
E3KARP
RKLO
G
IREEL
14C
130
120
PPAA]VQGAGNENEPRE
PRPPAAPGEQGPGENEPRF
GP
DSEPPAADTEAGDQLN- -
TATDPWEPKPDWAHTGSHSSEAG
180
170
190
200
210
FNLIS4DKSKPGQFIRSVDPDSPAEASGLRAQDRIVEVNGVCMEG
IVDPDSPAEASGLWQDRIVEVNGVCNEG
GFNLSDK GMFI
VDPDsPAEASGLRAQDRIVEVGVCMEG
YGFNLHSDKSKPGQFI
GQ~IRsVDPiSAiAUiSGLRAQDRifEVNGQfNJVG
YGFNLHSD
280
270
260
250
240
FTNOIQKENSREALAE
TDEFFKKCRVIPSQEHLNGPL
SE
FTNGEIQK ~A
TDEFFKK VMPSS|HLYGPL
E
,
GE
PL
PS
TDEFFKK
PL
E
FT
SPAQLNGGSACSSRSD
R
330
320
-- SSSDPILDFNIS
v1
-SSSSDPILDFQIS
310
QDSTAPSS
STAPSS
SSSSDPILD
PSS
LHL
QESG --------------
ISL
350
340
ERAilQKRSSKRAPQM.D
ERAHQKRSSJRAPQMD
ERAHQKRSSKRAPQMD
RAPQMD
360
hNHE-RF W5KKNELF9NI
rNHE-RF WSgK@ELFS
MNHE-RF WSKKNELFSNI
E3XARP gUIff|dFgr
provided by the Genetics Computer Group (GCG) and the BLAST
network service at the National Center for Biotechnology Information
(NCBI) (21).
Protein Fusion Constructs-Full-length(aa 1-595) as well as the
amino (aa 1-332) and carboxyl (aa 308-595) portions of merlin were
expressed as GST fusion proteins using the vector pGEX2T (16). Anencoding a naturally occurring NF2
other full-length merlin construct
20
-> Tyr (22) was generated by site-directed
missense mutation, Asn
mutagenesis (Stratagene) and expressed as a GST fusion protein. Similarly full-length (aa 1-577), amino (aa 1-332), and carboxyl (aa 307577) segments of human moesin were expressed as GST fusion proteins
using the vector pGEX4T1. Expression and purification of the GST
fusion proteins were performed as described previously for merlin (16)
and using the standard method of Smith and Johnson (23) for moesin.
For radixin, His,-tagged constructs were employed, and these have
been described previously (24).
Mammalian Expression-The entire coding sequence of hNHE-RF(1-358) engineered to have the influenza hemagglutinin (HA) epitope
tag at the 5' end was cloned into the mammalian expression vector
pcDNA3. The entire coding sequence of merlin was expressed as a GFP
(green fluorescent protein) fusion protein employing the vector
pEGFP-N1 (CLONTECH). These expression constructs were transfected in COS-7 cells and HeLa cells by the calcium phosphate method
using the Cell Phect kit (Pharmacia Biotech Inc.) employing 10-20 sig
of the plasmid DNA. Transient transfections were performed, and cells
were harvested after 72 h.
Antibodies-An antipeptide rabbit polyclonal antibody was raised
against the carboxyl-terminal amino acids SLAMAKERAHQKR (aa
328-340) specific to human NHE-RF by Research Genetics, Inc.,
Huntsville, AL. The antiserum obtained (NP1) was further affinitypurified. The anti-merlin antibodies and the antibodies for moesin have
been described previously (16, 25). The hybridoma supernatant for the
HA-tag was kindly provided by Dr. Ed Harlow (Massachusetts General
Hospital, Boston, MA).
Affinity Precipitationof ANHE-RF from Cell Lysates-COS-7 cell
lysates overexpressing hNHE-RF were incubated with 300 pmol of
GST-merlin or GST-moesin immobilized on glutathione-Sepharose 4B
(GSH) beads. The beads were washed with phosphate-buffered saline
containing Pefabloc, resuspended in Laemmli loading buffer, subjected
to 10% SDS-PAGE, and immunoblotted with anti-hNHE-RF (affinitypurified NP1). For radixin, COS-7 cell lysates overexpressing the interactor were incubated with 600 pmol of the Hiss-tagged full-length,
amino-terminal, or carboxyl-terminal radixin polypeptides. The complexes were separated from the reaction mixture using Ni-NTA beads.
The beads were washed (20 mm imidazole, 50 mm sodium phosphate,
300 mm NaCl, pH 8.0), and the specifically bound complexes were eluted
with the same buffer containing 400 mM imidazole separated on SDSPAGE gel and detected with NP1 antibody as described above.
Immunofluorescence-The immunofluorescence staining was performed as described previously (16). Anti-HA monoclonal antibody (1:
100) was used as a primary antibody to detect the localization of
hNHE-RF. The affinity-purified moesin antibody was used as described
earlier (9). Cells were examined on a Nikon microscope using a 40 x 1.3
NA. and 60 x 1.4 N.A. objectives. For confocal microscopy, cells were
examined with Leica TCS-NT 4D scanning laser confocal microscope.
RESULTS AND DISCUSSION
To identify proteins that associate with merlin, we screened
a human fetal frontal cortex interaction library with pmerbait.
About 106 primary transformants were pooled and replated (at
a multiplicity of 20) onto galactose leu- selection plates. Thirtythree colonies, which showed galactose-dependent growth and
blue color on leu~ plates and on 5-bromo-4-chloro-3-indolyl
-D-galactopyranoside medium, respectively, were identified,
and plasmids containing the cDNA clones were isolated. Restriction mapping and hybridization experiments revealed that
these cDNAs were clustered into three different groups of overlapping clones, representing three novel cDNAs. In this study
we report the characterization of one of the cDNAs, an 880-base
pair fragment which we refer to as merintc. We tested merintc
for associations in yeast strains containing either the amino(aa 1-341) or the carboxyl- (aa 342-595) terminal portion of
merlin as bait. Results of these tests indicated that merintc
specifically associates with full-length and amino-terminal
merlin constructs (data not shown).
The largest merintc fusion clone was sequenced, and a
BLASTX search of GenBank revealed strong similarity with
the carboxyl-terminal segment of rabbit NHE-RF, a protein
that mediates protein kinase A inhibition of the renal brushborder membrane Na'-H' exchanger (18). A full-length cDNA
clone was isolated from a human fetal brain library and was
sequenced, revealing 88% identity (across the open reading
frame) with rabbit NHE-RF. The predicted 358 amino acid
protein shares 86%identity with both the rabbit protein and its
mouse homologue and contains 2 PDZ repeats between amino
NHE-RF Interacts with MERM Proteins
A
A
B
ti
1275
isI" 01
4-.
FIG. 2. Affinity precipitation of hNHE-RF from cell lysates.
RIPA (50 mi Tris, pH 7.5, 150 mm NaCl, 1% Nonidet P-40, 0.5%
deoxycholate, 0.1% SDS containing a 1x protease inhibitor mixture)
lysates from COS-7 cells overexpressing hNHE-RF were incubated with
merlin or moesin expressed as GST fusion proteins and immobilized on
GSH beads (A), and radixin expressed with a His-tag and eluted from
Ni-NTA beads (B). The beads were extensively washed, separated on
10% SDS-PAGE, and immunoblotted with affinity-purified anti
hNHE-RF antibody (NP1). N and C represent the amino and carboxyl
regions of merlin, moesin, and radixin. FL represents full-length merlin
and radixin, and Mut-Mer is an NF2-associated missense mutation of
merlin. GST protein expressed alone is shown as a control (GST). Total
cell extracts from either COS-7 cells overexpressing full-length
hNHE-RF or wild type COS-7 cells are also shown as controls. The
supernatants that were not bound to the beads are shown to point out
that the amino portion of moesin is completely bound by hNHE-RF (A).
The arrow shows the hNHE-RF at 50 kDa.
acids 11-97 and 149-236, with an expected molecular mass of
39 kDa (Fig. 1). Consequently, merinte appears to derive from
the human homologue of NHE-RF which we refer to as hNHERF. Northern blot analysis of human tissues revealed that
hNHE-RF is ubiquitously expressed with highest levels in kidney, liver, and pancreas (data not shown).
To confirm the interaction of full-length hNHE-RF with merlin, we expressed hNHE-RF with a 5'HA tag in COS-7 cells and
tested its ability to bind with different segments of merlin
expressed as bacterial GST fusion proteins and immobilized on
GSH beads. We detected the bound protein in immunoblots,
using the anti-hNHE-RF antibody (NP1). As shown in Fig. 2,
hNHE-RF bound specifically to the amino-terminal and fulllength merlin fusion proteins, but did not bind to either the
carboxyl terminus of merlin or to GST alone. Binding to a
mutant merlin, encoding an NF2-associated missense muta2
tion Asn o -* Tyr, is reduced compared with the wild type (Fig.
2, lane 4). Thus, as was predicted by the two-hybrid analysis,
hNHE-RF associates with the amino-terminal but not with the
carboxyl-terminal domain of merlin.
Because the amino-terminal domain of merlin shows significant similarity with ERM proteins, we also examined the
interaction of NHE-RF with moesin and radixin. We separated
GST-moesin and His-tagged radixin complexes from cell lysates expressing hNHE-RF by affinity chromatography. As
shown in Fig. 2, hNHE-RF showed a strong association with
the amino termini of both moesin and radixin but not with the
carboxyl terminus of either protein. The interaction with fulllength radixin (Fig. 2B) is reduced compared with the aminoterminal domain alone. A similar result has been obtained with
full-length moesin (data not shown) although no such difference was noted for merlin (Fig. 2A). The immobilized amino
termini of moesin and radixin completely bound and removed
hNHE-RF from the lysate, whereas immobilized merlin, incubated with an equivalent amount of hNHE-RF, bound only a
fraction of the protein (Fig. 2).
We have also determined the subcellular localization of exogenously expressed hNHE-RF in HeLa cells using immunocytochemistry. hNHE-RF localizes to the ruffling membrane, mi-
B
I.,
FIG. 3. Co-localization of hNHE-RF with merlin and moesin by
double immunocytochemistry. A, merlin is visualized as GFP-
tagged protein (a, d). Rhodamine-conjugated secondary anti-mouse an-
tibody was used to detect hNHE-RF (b, e). The partial co-localization of
these two proteins can be visualized in COS-7 cells at the ruffling
membrane (small arrows), microvilli and filopodia (large arrow,c), and
similar co-localization in microvilli is observed in HeLa cells (t). Bar, 10
gm. B, fluorescein isothiocyanate-conjugated secondary anti-mouse antibody was used to detect hNHE-RFh (a,d), and rhodamine-conjugated
secondary anti-mouse antibody was used to detect endogenous moesin
(b, e) in HeLa cells. At the top of the cells (a-c), there is strong overlap
of these two proteins in the microvilli (c). In the lower part of the cells
close to substrate (d-f), the two proteins co-localize very clearly at
ruffling membrane and filopodia. Bar, 20 pan.
crovilli, and filopodia reminiscent of the ERM proteins and of
overexpressed merlin (Fig. 3). When we co-expressed both merlin and hNHE-RF in COS-7 and HeLa cells, we also observed
the co-localization of these two proteins in membrane ruffles,
microvilli, and filopodia (Fig. 3A). Double immunostaining of
exogenous hNHE-RF and endogenous moesin in HeLa cells
revealed co-localization in microvilli, ruffling membrane, and
filopodia (Fig. 3B). However, hNHE-RF does not overlap with
merlin and moesin in all microvilli and filopodia, and similarly
merlin and moesin are not seen throughout the membrane
ruffles where we observe hNHE-RF. The frequent co-localization of hNHE-RF with merlin and moesin is, however, consistent with the interaction of these proteins.
We demonstrate here that hNHE-RF, a widely expressed
protein whose rabbit homologue was first identified by its involvement in mediating protein kinase A regulation of the
renal brush-border Na*-H' exchanger (18) interacts both with
merlin and with its close relatives, moesin and radixin from the
ERM family. Interestingly, while this manuscript was in preparation, Reczek et al. (19) identified hNHE-RF as a physiologically relevant interactor (which they named EBP50) for ezrin,
the third ERM protein.
hNHE-RF possesses two PDZ domains, motifs thought to
mediate protein-protein interactions, particularly at the
plasma membrane during assembly of components involved in
NHE-RF Interacts with MERM Proteins
1276
might abrogate regulation of a Na+-H+ exchanger in NF2
target cells and lead to formation of meningiomas and
schwannomas.
Acknowledgments-We thank Dr. Yimin Ge of the Cutaneous Biology
Research Center at Massachusetts General Hospital for the excellent
help with the confocal microscopic images, Dr. N. Ramesh of Childrens'
Hospital, Boston for valuable suggestions, and members of our laboratory for helpful comments on the manuscript.
MERM
REFERENCES
1. Evans, D. G. R., Huson, S. M., Donnai, D., Neary, W., Blair, V., Teare, D.,
Newton, V., Strachan, T., Ramsden, R., and Harris, R. (1992) J.Med. Genet.
29, 841-846
2. Gusella, J. F., Ramesh, V., McCollin, M., and Jacoby, L. B. (1996) Curr. Opin.
Genet. Dev. 6, 87-92
FIG. 4. A model connecting the Na+-H+ exchangers of the
plasma membrane to the actin cytoskeleton via the interaction
of hNHE-RF to the NF2 tumor suppressor merlin and related
ERM (MERM) proteins.
cellular signaling (26, 27). E3KARP (also called TKA-1), a
protein recently isolated as an interactor for the small intestinal and renal proximal tubule brush-border Na+-H' exchanger
NHE3, shares 51% identity with hNHE-RF primarily in the
PDZ domains (28). However, the interaction with the aminoterminal domain of merlin and the ERM proteins occurs outside of the hNHE-RF PDZ domains, via the carboxyl-terminal
region spanning amino acids 290-358. This suggests that the
hNHE-RF and its MERM binding partners might participate in
a larger complex, one component of which might be a Na+-H+
exchanger.
Multiple forms of Na+-H+ exchangers exist in mammalian
cells, and at least five of these, designated as NHE1-5, sharing
an overall 34-60% amino acid identity have been cloned and
characterized (29). Whereas NHE3 expression is limited to
kidney and intestine, NHE1 is a housekeeping protein that is
ubiquitously expressed (30). NHE1 has been shown to accumulate in membrane ruffles where vinculin, talin, and F-actin are
concentrated suggesting that Na'-H+ exchangers are sequestered to these regions by interacting with the cytoskeletal
network (31). NHEI is necessary for Rho-induced stress fiber
formation (32).
Our demonstration of NHE-RF as a binding partner for
MERM family members suggests a potential functional connection with the involvement of Na'-H' exchange in cytoskeletal
rearrangement (Fig. 4). Indeed, a recent study demonstrates
that ERM proteins are also essential for Rho- and Rac-induced
cytoskeletal changes such as stress fiber assembly, formation of
focal adhesions, elaboration of lamellipodia, and membrane
ruffling (15). The prospect that MERM-NHE-RF interaction
plays a role in these changes and the fact that NHE-RF is
expressed in tissues not containing NHE3 raise the question of
whether NHE-RF is also capable of regulating other Na+-H+
exchangers, particularly NHE1. Interestingly, increased
Na'-H' exchange has long been known to be associated with
cell proliferation, differentiation, and neoplastic transformation (33). Thus, an intriguing possibility that merits investigation is that merlin-NHE-RF interaction is the basis of merlin's
tumor suppressor function since failure of the interaction
3. Tsukita, S., Yonemura, S., and Tsukita, S. (1997) Curr. Opin. Cell Biol. 9,
70-75
4. Trofatter, J. A., MacCollin, M. M., Rutter, J. L., Murrell, J. R., Duyao, M. P.,
Parry, D. M., Eldridge, R., Kley, N., Menon, A. G., Pulaski, K., Haase, V. H.,
Ambrose, C. M., Munroe, D., Bove, C., Haines, J. L., Martuza, R. L.,
MacDonald, M. E., Seizinger, B. R., Short, M. P., Buckler, A. J., and
Gusella, J. F. (1993) Cell 72, 791-800
5. Bretcher, A. (1983) J. Cell Biol. 97, 425-432
6. Birgbauer, E. (1991) Cytoskeletal Interactions of Ezrin in Differential Cells.
Ph.D. thesis, Massachusetts Institute of Technology, Cambridge, MA
7. Sato, N., Yonemura, S., Obinata, T., Tsukita, S., and Tsukita, S. (1991) J.Cell
Biol. 113, 321-330
8. Sato, N., Funaysma, N., Nagafuchi, A., Yonemura, S., Tsukita, S., and
Tsukita, S. (1992) J.Cell Sci. 103, 131-143
9. Henry, M. D., Gonzalez-Agosti, C., and Solomon, F. (1995) J. Cell Biol. 129,
1007-1022
10. Turunen, 0., Wahlstrom, T., and Vaheri, A. (1994) J.Cell Biol. 126, 1445-1453
11. Pestonjamasp, K, Amieva, M. R., Strassel, C. P., Nauseef, W. M., Furthmayr,
H., and Luna, E. J. (1995) Mol. Biol. Cell 6, 247-259
12. Martin, M., Roy, C., Montcourrier, P., Sahuquet, A., and Mangeat, P. (1997)
Mol. Biol. Cell 8, 1543-1557
13. Tsukita, S., Oishi, K, Sato, N., Sagara, J., Kawai, A., and Tsukita, S. (1994)
J. Cell Biol. 126, 391-401
14. Hirao, M., Sato, N., Kondo, T., Yonemura, S., Monden, M., Sasaki, T., Takai,
Y., Tsukita, S., and Tsukita, S. (1996) J. Cell Biol. 135, 37-51
15. Mackay, D., Each, F., Furthmayr, H., and Hall, A. (1997) J. Cell Biol. 138,
927-938
16. Gonzalez-Agosti, C., Xu, L., Pinney, D., Beauchamp, R., Hobbs, W., Gusella, J.,
and Ramesh, V. (1996) Oncogene 13, 1239-1247
17. Xu, L., Gonzalez-Agosti, C., Beauchamp, R., Pinney, D., Sterner, C., and
Ramesh, V. (1998) Exp. Cell Res., in press
18. Weinman, E. J., Dubinsky, W., and Shenolikar,
S. (1995) J. Clin. Invest. 95,
2143-2149
19. Reczek, D., Berryman, M., and Bretcher, A. (1997) J.Cell Biol. 139, 169-179
20. Gyuris, J., Golemis, E. A., Chertkov, H., and Brent, R. (1993) Cell 75, 791-803
21. Altschul, S. F., Thomas, M. L., Alejandro, S. A., Zhang, J., Zhang, Z., Miller,
W., and Lipman, D. J. (1997) Nucleic Acids Res. 25, 3389-3482
22. MacCollin, M., Mohney, T., Trofatter, J., Wertelecki, W., Ramesh, V., and
Gusella, J. F. (1993) J. Am. Med. Assoc. 270, 2316-2320
23. Smith, D. B., and Johnson, K S. (1988) Gene 67, 31-40
24. Magendantz, M., Henry, M. D., Lander, A., and Solomon, F. (1995) J. Biol.
Chem. 270, 25324-25327
25. Winckler, B., Gonzalez-Agosti, C., Magendantz, M., and Solomon, F. (1994)
J. Cell Sci. 107, 2523-2534
26. Tsunoda, S., Sierralta, J., Sun, Y., Bodner, R., Suzuki, E., Becker, A., Socolich,
M., and Zuker, C. S. (1997) Nature 388, 243-249
27. Fanning, A. S., and Anderson, J. M. (1996) Curr. Biol. 6, 1385-1388
28. Yun, C. H. C., Oh, S., Zizak, M., Steplock, D., Tsao, S., Tse, C.-M., Weinman,
E. J., and Donowitz, M. (1997) Proc.Natl.Acad. Sci. U. S. A. 94, 3010-3015
29. Orlowski, J., and Grinstein, S. (1997) J.Biol. Chem. 272, 22373-22376
30. Yun, C. H. C., Tse, C.-M., Nath, S. K., Levine, S. A., Brant, S. R., and Donowitz,
M. (1995) Am. J.Physiol. Gi-G11
31. Grinstein, S., Woodside, M., Waddell, T. K., Downey, G. P., Orlowski, J.,
Pouyssegur, J., Wong, D. C. P., and Foskett, J. K. (1993) EMBO J. 12,
5209-5218
32. Vexler, Z. S., Simons, M., and Barber, D. L. (1996) J. Biol. Chem. 271,
22281-22284
33. Harguindey, S., Pedraz, J. L., Canero, R. G., Diego, J. P. D., and Cragoe, E. J.
(1995) Crit. Rev. Oncogene. 6, 1-33