Vol. 264, No. 27,Issue of September 25, pp. 16083-16092,1989
Printed in U.S.A.
THEJOURNAL OF BIOLOGICAL
CHEMISTRY
0 1989 by The American Society for Biochemistry and Molecular Biology, Inc
A Novel 58-,kDaProtein Associates with the GolgiApparatus and
Microtubules*
(Received for publication, April 27, 1989)
George 23. Bloom$ and TeresaA. Brashear
From the Department of Cell Biology and Neuroscience, University of Texas Southwestern Medical Center, Dallas, Texas 75235
With the aim of identifying proteins involved in linkdination of Golgi and microtubule functions. To facilitate this
ing microtubules to other cytoplasmic structures, mi- cooperativity, the Golgi apparatus is usually located in close
crotubule-binding proteins were isolated from
rat liver proximity to themicrotubule-organizing center (MTOC),’ the
extracts by a taxol-dependent procedure. The major site from which most microtubules emanate and where they
non-tubulin component, a 58-kDa protein (designated are maximally concentrated in the cell. This intimate affilia58K),was purified to homogeneity by gel filtration tion between the MTOC region and Golgi is maintained even
chromatography. To aid further characterization of while the MTOC migrates, as occurs during myoblast differ58K,purified prepiarationsof the protein were
used as entiation into myotubes (Tassin et al., 1985), chemotaxis of
immunogen for the production of monoclonal antibod- macrophages (Nemere et al., 1985), binding of natural killer
ies. Fivedifferent monoclonals wereobtained,and
cells to target cells (Kupfer et al., 1983), and the movement
each of these reacted on immunoblots of liver homogenates with a single band that comigrated with 58K. of fibroblaststowardexperimental wounds (Kupfer et al.,
Based onthe resultsof immunochemical, peptide map- 1982). When microtubule depolymerizing drugs, such as col58K was cemid, nocodazole, or vinblastine, are administered to cells,
ping, and microsequencingexperiments,
found to be unrelated structurally to similarly sized Golgi components redistribute into fragments that are scatcytoskeleton-associated proteins, such as tubulin, tau, tered throughout the cytoplasm (Rogalski and Singer, 1984;
vimentin, or keratin, and to represent a new protein Wehland et al., 1983). Other agents that alter thedistribution
species. Several i n ivitro properties of 58K were found of microtubules in cultured cells have also led to fragmentato be characteristic of microtubule-associated proteins. tion of the Golgi. Included among these, for example, are
externally administered taxol or microinjected anti-tubulin
For instance, 58K cosedimented quantitatively with
microtubules out of liver extracts, stimulated polym- (Wehland et al., 1983). Taken together, these studies have
erization of tubulin, and bound to microtubules in a indicated that the structuralintegrity of the Golgi apparatus
saturable manner. Cn contrast to traditional microtu- requires the presence of normally organized microtubules
bule-associated proteins, however, 58K was not found within the cytoplasm.
to be distributed uniformly along
microtubules in cells.
To account for this, it is likely that specific proteins are
Immunofluorescence microscopy of cultured hepatoma involved in cross-bridging the Golgi to microtubules in the
cells revealed, instead, that 58K is associated princi- vicinity of the MTOC. A protein that serves this function
pally with theGolgi apparatus. Moreover, Golgi mem- would presumably possess binding sites located on the exterior
branes isolated frolm rat liver were observed by im- or cytoplasmic surfaces of both microtubules and the Golgi
munoblotting tocontainsignificantlevels
of 58K, apparatus. Based on the possibility that such proteins might
which, upon subfraetionation of the membranes, partitioned as if it were a peripheral membrane protein be recovered from cytosol in conjunction with conventional
exposed to the cytoplasmic side of the Golgi. These microtubule-associated proteins (MAPs), we applied a taxolcollective results havebeen evaluated in termsof ear- dependent method for isolating MAPs (Bloom et al., 1985b;
lier evidence that the intracellularposition and struc- Vallee, 1982) to liver extracts, a rich source of Golgi memtural integrity of the Golgi relies on the presence and branes, and searched among the products for proteins that
organization of microtubules. In that context, the ob- also display binding to theGolgi. As a result of having taken
servations reported heresuggest that thein vivo func- this approach, we have identified, purified, and characterized
tion of 58K is to provide an anchorage site for micro- a novel 58-kDa protein (designated 58K) which exhibits properties that arecharacteristic inpart of both MAPs and Golgitubules on the outer surface
of the Golgi.
associated proteins.
Like many MAPS whose in vitro properties have been
characterized in depth, 58K was found to cosediment quanThe Golgi apparatus consists of an interconnected, titatively with microtubules out of tissue extracts, stimulate
branched network of membrane-bounded stacks and tubules, tubulin assembly, and exhibit saturation binding to microtuand is the site from which secretory vesicles originate. Follow- bules. Unlike most MAPs, though, the primary intracellular
ing their emergence from the Golgi, these vesicles move along location of 58K was found to be the Golgi apparatus, rather
microtubules to the cell perimeter, necessitating close coor- than microtubules, as judged by immunofluorescence microscopy of cultured hepatoma cells. Consistent with this obser* This work was suppolted by National Institutes of Health Grant vation was our finding of significant levels of 58K in Golgi
GM-35364 and National Science Foundation Grant DMB-8701164
(to G. S. B.). The costs of publication of this article were defrayed in
part by the payment of p;ige charges. This article must therefore be
hereby marked “aduertkement” in accordance with 18 U.S.C. Section
1734 solely to indicate this fact.
$ To whom correspondence should be addressed.
’
The abbreviations used are: MTOC, microtubule-organizing center; MAP, microtubule-associated protein; HPLC, high pressure liquid chromatography; SDS-PAGE, sodium dodecyl sulfate-polyacrylamide gel electrophoresis; EGTA, (ethylenebis(oxyethylenenitrilo)1tetraacetic acid.
16083
16084
A Protein
That
Associates with Microtubules and
the
Golgi
quantitatively recovered in thesupernatant, along with a small
amount of soluble brain tubulin.
Purification of 58K-The microtubule-binding proteins isolated
from liver were concentrated 9-fold by dialysis uersus solid sucrose
and fractionated by,gel filtration chromatography usingeither a 4000
SW Spherogel TSK column on aWaters HPLCsystem or a Toyopearl
HW-55F column in conjunction with a low pressure peristaltic pump.
Fractions containing dilute 58K at 90-99% purity could be obtained
by this step, and
were pooled and concentrated. Forsome applications,
the sucrose present following the final concentration step was reEXPERIMENTALPROCEDURES
moved using a desaltingcolumn. Aliquots of purified 58K were frozen
Materials-Taxol was generously provided by Dr. Matthew Suffin liquid nitrogen and stored at -80 "C.
ness of the NationalCancer Institute. Tissueculture media and
Purification of Tau-Bovine brain microtubules were purified by a
serum were obtained from Sigma and Hy-Clone (Logan, UT), respec- taxol-dependent procedure (Vallee, 1982). Heat-stable MAPs (Kim
tively. Freund's adjuvant was from GIBCO. Rabbit antisera specific et al., 1979) were then isolated from the microtubules by adding 0.1
for mouse antibody isotypes were acquired from ICN. Antibodies volume of 100 mM dithiothreitol and 0.25 volume of PEM suppledirected against mouse or goat IgG, and labeled with fluorescein mented with 4 M NaCl to 1 volume of pelleted microtubules, boiling
isothiocyanate, tetramethyl rhodamineisothiocyanate,horseradish
the mixture for 5 min and centrifuging for 30 min a t 4 "C in a Sorvall
peroxidase, or alkaline phosphatase were purchased from the Cappel SA-600 rotor a t 17,000 rpm (41,800 X gmaX).
Tau was then purified
Division of Organon Teknika N.V. (Malvern, PA), Fisher, or Jackson
from the supernatantby gelfiltration chromatography using column
a
ImmunoResearch (West Grove, PA). Nitro blue tetrazolium and 5- of Toyopearl HW-55F.
bromo-4-chloro-3-indoy1phosphate were obtained from Promega BioIsolation and Subfractionation of Rat Liver Golgi Membranes-A
tec (Madison, WI). Wheat germ agglutinin labeled with fluorescein previously described method for isolating Golgi-enriched membrane
isothiocyanate was from Vector Laboratories (Burlingame, CA). Two fractions from rat liver (Leelavathi et al., 1970) was used with two
murine monoclonal antibodies to bovine brain tau (Binder et al., 1985; adjustments. First, protease inhibitors were added to the tissue hoPapasozomenos and Binder, 1987) were kindly provided by Dr. Lester
mogenization buffer. Second, in some preparations theGolgi-enriched
Binder. Polyclonal rabbit antibodies to BHK-21 cell vimentin and material loaded onto the final discontinuous sucrose gradient
was
mouse skin keratin were generously supplied by Dr. Robert Goldman. brought to 1.25 M, instead of 1.1 M sucrose, underlaid with 1.4 M
Reagents for the bicinchoninic acid (Smith etal., 1985) and Coomassie sucrose, and overlaid with 1.1 and 0.5 M sucrose. As described in the
Blue G-250 (Bradford, 1976) proteinassays were purchased from original procedure (Leelavathi et al., 1970), the Golgi membranes
Pierce. All other chemicals and reagents were purchased from Sigma collected at the0.5/1.1 M interface but were separated more effectively
or Polysciences (Warrington, PA). DEAE-Sephadex A-50m (Phar- from cytosolic proteins as a result of the second modification.
Prior to further processing, Golgi membranes were diluted by the
macia LKB Biotechnology Inc.) was used for anion exchange chromatography. A Beckman (San Ramon, CA) HPLC column (4000 SW addition of several volumes of sucrose-free buffer, and thesuspensions
Spherogel TSK) and low pressure columns of Toyopearl HW-55F were centrifuged for 2 h at 4 "C in a TLA 100.3 rotor at 50,000 rpm
(Supelco, Bellefonte, PA) were used for gel filtration chromatography. (100,000 X g,,,) in a Beckman TL-100 tabletop ultracentrifuge. The
Nitrocellulose paper (BA85; 0.45 em) was acquired from Schleicher pelleted Golgi membranes were then resuspended in buffer to a final
& Schuell (Keene, NH). Disposable, 115-ml filters (0.45 ern) were protein concentration of0.4-0.8 mg/ml. Extraction of Golgi memobtained from the Nalge Division of the Sybron Corporation (Roch- branes with Triton X-114 was performed as described (Bordier, 1981).
Salt extraction was accomplished by the addition of solid KC1 to a
ester, NY).
Purification of Tubulin-Microtubules were purified from bovine final concentration of0.5 M in suspensions of Golgi membranes,
brain by one to three cycles of GTP-stimulated assembly at 37 "C in centrifugation at 34,000 rpm (50,000 X gm,) for 30 min at 4 'C in a
the presence of glycerol and cold-induced disassembly according to TLA 100.2 rotor, and resuspension of the pellet to original volume.
Other Biochemical Methods-Amino acid sequence data for 58K
the method of Murphy (1982). Anion exchange chromatography using
a column of DEAE-Sephadex A-50m (Vallee and Borisy, 1978) was were obtained by afour-step process. First, 58K was purified as
described here. Second, -400 pmol of 58K were subjected to electrothen used to separate tubulin from the MAPs.
Isolation of Microtubule-binding Proteins-Livers were dissected phoresis on a 7.5% polyacrylamide-SDS gel, after which the protein
from adult rats thathad been allowed unrestricted access to food and was transferred to nitrocellulose and digested with trypsin (Aebersold
were killed by exposure to lethal doses of ether. Seven livers weighing et al., 1987). Third, tryptic peptides were resolved by two successive
-105 g total were used for most preparations. Immediately following reverse phase chromatographic steps, using an Applied Biosystems
(Foster City, CA) Model 130A HPLC and a Brownlee (Santa Clara,
dissection, the livers were minced into small pieces, and a Polytron
CA) Model RP300 C8 column (2.1 X 100 mm). The first separation
tissue grinder was used to homogenize the tissue in 1.5 volumes of
was performed in 0.1% trifluoroacetic acid using a gradientof 0-50%
ice-cold PEM buffer (0.1 M piperazine-N,N'-bis(2-ethanesulfonic
acetonitrile. Selected peaks were then loaded onto anidentical column
acid), pH 6.8-7.0, 1 mM MgS04, 1 mM EGTA) supplemented with 1 in 0.1% ammonium acetate, and an elution gradient of 0-50% acetomM dithiothreitol and protease inhibitors (1 mM phenylmethylsulfo- nitrile in 0.1% ammonium acetate was applied. The fourth and final
nyl fluoride; 10 pg/ml each leupeptinand pepstatinA). All subsequent
step involved collection of second dimension HPLC peaks on Whatsteps were performed at 0-4 "C unless specified otherwise. The ho- man GF/C paper, reduction and alkylation of cysteine residues (Anmogenate was centrifuged for 20 min at 40,000 rpm (186,000 X g,,,)
drews and Dixon, 1987), and sequencing of the peptides onan Applied
in a Beckman 45Ti rotor, and theresulting supernatant was spun for Biosystems Model 470A Sequencer.
90 min in the same rotor a t 37,000 rpm (159,000 X gmax). Thefinal
SDS-polyacrylamide gel electrophoresis (SDS-PAGE) was persupernatant was then clarified further by passage through 0.45-pm formed as described earlier (Bloom et al., 1988) using 4-16% polyfilters.
acrylamide and 0-6 M urea gradients in the separatinggel. Coomassie
Purified brain tubulin was polymerized at 37 "Cwith a 1.5 M excess Blue R-250 was used to stain gels of this type. Molecular weight
of taxol and added to the ice-cold liver extract to a final exogenous standards included bovine erythrocyte carbonic anhydrase (29,000),
tubulin concentration of 0.5 mg/ml. All succeeding centrifugation
chicken ovalbumin (45,000), bovine serum albumin (66,000), rabbit
steps were performed in a Beckman 60 Ti rotor at 23,000 rpm (53,000 muscle phosphorylaseb (97,400), Escherichia coli 8-galactosidase
X gmax)
for 30 min at 4 "C. Following a 10-min incubation, the solution
(116,000), and rabbit muscle myosin (205,000). Quantitative densiwas centrifuged through 0.2 volume cushions of 10% sucrose in PEM tometry of gels stained with Coomassie Blue R-250 was accomplished
buffer. The pellets, which contained polymerized brain tubulin deco- using an LKB (Piscataway, NJ) Model 2222-010 Ultroscan XL scanrated with liver microtubule-binding proteins, were resuspended to ning laser densitometer.Purified tubulin or 58K was used as a
10% the original extract volume in PEM buffer supplemented with
calibration standard and to determine the linear range of protein
dithiothreitol and protease inhibitors, and frequently 0.5% Triton X- amount Versus absorbance at 633 nm. All measurements were within
100. Following resuspension, the microtubules were centrifuged again. this range. Peptide mapping by limited proteolysis (Cleveland et al.,
The washed pellets were then resuspended in PEM supplemented
1977) involved the use of 15% acrylamide gels and the Laemmli
with 0.5 M KC1 to one-eighth the original extract volume and centri- (1970) buffers and were stained by an ammoniacal silver method
(Wray et al., 1981). Either of two colorimetric protein assays (Bra(!fuged once more. The pellets consisted of polymerized brain tubulin
and were discarded. Microtubule-binding Droteins from liver were ford, 1976; Smith et al., 1985) was used in conjunction with hovirw
membranes isolated from rat liver. Analysis of Golgi subfractions by immunoblotting providedevidence that 58K is a
peripheral membrane proteinexposed on thecytoplasmic side
of the Golgi. The presence of 58K at that site might enable
the protein t o form cross-bridges to closely apposed microtubules, emphasizing the possibility that the i n vivo function of
58K is to anchor theGolgi apparatus to microtubules.
A Protein ThatAssociates with Microtubules and the Golgi
16085
serum albumin or y-globulin as standards.
B
Immunological Methods-A BALB/c mouse was immunized sub7
cutaneously on fouroccasions a t I-3-week intervals with purified
58K emulsified in Freund's complete adjuvant. One week after the
final injection, the mouse was bled from the tail vein, and antibody
content of the serum was evaluated by immunoblotting of microtu"Dynein
bule-bindingproteins isolated fromrat liver. A strong response
against 58K was noted a t a 1:500 dilution of serum, and the mouse
205was immunized a final time by intraperitoneal injection of purified
58K in the absence
of adjuvant. Three days later, splenic lymphocytes
from the immunized mouse were fused with NS1 murine myeloma
cells using our standardprotocol (Bloom et al.,1984a). Five clonesof
116hybridomaswere obtainedaftertwosteps
of cloning by limiting
dilution in 96-well dishes.
97.4Coverslips containing cultured hepatoma cells (rat Faza, mouse
66 Hepa-l-6-J, or human Hep-GB) or adsorbed, 58K-decorated micro58K
tubules were immersed into -20 "C methanol for 5-10 min as a
D
]Tubulin
fixation step. Monoclonal anti-58K and goat anti-mouseIgG labeled
with tetramethyl rhodamine isothiocyanate
were used as primary and
45secondary antibodies, respectively.Fluorescein
isothiocyanate-labeled wheat germ agglutinin wasincluded at the second antibody step
for some experiments. In some cases, cells were incubated with a
29third antibody, tetramethyl rhodamine isothiocyanate-labeled rabbit
anti-goat IgG.
Our standard procedure for transferring proteins from SDS-polyacrylamide gels to nitrocellulose (Bloom et al., 1984a) was used for
FIG. 1. Isolation of microtubule-binding proteins from rat
this study. In most experiments,
a peroxidase-labeled goat anti-mouse
IgG was usedas a second antibody, and 4-chloro-1-naphthol
was used liver and purification of 58K. A, purified bovine brain tubulin
as a reagent for visualizing immunoreactive protein. The experiment was assembled using taxol and added to rat liver cytosol to a final
documented in Fig. 8, though, involved the use of alkaline phospha- concentration of 0.5 mg/ml. Centrifugation of this solution ( l a n e 1 )
yielded a supernatant ( l a n e 2) depleted of liver microtubule-binding
tase-labeled rabbit anti-mouse IgG as a second antibody, and nitro
phosphate (Blake et proteins, which were enriched in themicrotubule pellet ( l a n e 3). The
blue tetrazolium and5-bromo-4-chloro-3-indoyl
pellet was resuspendedin PEMbuffer supplemented with0.5% Triton
a/.,1984) as chromogens.
Electron Microscopy-Microtubules were polymerized at 37 "Cby X-100 and centrifuged. The supernatant ( l a n e 4 ) was discarded, and
addition of either taxol to 10 pM or G T P t o 10 mM to a mixture of the washed microtubule pellet (lane 5 ) was resuspended in PEM
purified preparations of tubulin and58K. Negative staining of micro- containing 0.5 M KCI. A final centrifugation step yielded a supernaof the microtubule-binding proteins
tubules adsorbed to Formvar-coated grids was performed exactly as tant ( l a n e 6) containing nearly all
described by Langford (1983). For thin sectioning, microtubules
were from liver plus some brain tubulin and a microtubule pellet ( l a n e 7)
for 30 min at 4 "C, and composed mainly of tubulin. A molecular mass scale in kilodaltons
pelleted by centrifugation a t 50,000 X gmaX
(see "Experimental Procedures") is shown to the left. The positions
fixed for 2 h with 2% glutaraldehyde, 1%tannic acid in PEM buffer.
The pellet was then post-fixed in OsO,, embedded in Spurr's resin, of 58K, tubulin, and dynein are indicatedto the right. B, the microA, lone 6) were concentratedand
sectioned, and counterstained with uranyl acetate and lead citrate. tubule-bindingproteins(asin
Specimens were viewed and photographed on a Phillips EM 300 a t fractionated further on a 7.8 X 300 mm HPLC column packed with
58K were
4000 SW Spherogel TSK. Several fractions containing pure
80 kV.
obtained, the peak fractionbeing shown here.
-
RESULTS
A high molecular weight band (MI > 300,OO)was found to
Purification of58K-A
taxol-based procedure originally
designed for the isolation of MAPS revealed a protein whose be the second most prominent microtubule-binding protein
electrophoretic mobility is slightly less than thatof a-tubulin in rat liver. This band was identified as the heavy chain of
to be the major microtubule-binding protein in rat liver. A cytoplasmic dynein (MAPIC) based on its -20-s sedimentamolecular mass of 58 kDa was estimated for this protein by tion coefficient, failure to react with broadly cross-reactive
SDS-PAGE, and we refer to the protein accordingly as 58K. monoclonal antibodies to MAPlA(Bloom et dl., 1984a, 1984b)
As illustrated in Fig. L4 by SDS-PAGE, the taxol procedure and MAPlB (Bloom et al., 1985a),comigration in SDS-PAGE
led to partial purification of 58K. Purified bovine brain tu- with brain dynein, and photosensitive cleavage into two lower
bulin was assembled with taxol and added to liver cytosol molecularweight fragments in the presence of ATP and
(lane 1). The microtubule-supplemented liver extract was vanadate (Paschal et al., 1987). Numerous additional minor
then centrifuged, yielding a supernatant depleted of microtu- polypeptides were observed in the MAP fraction, including
bule-binding proteins (lane 2) anda pellet (lane 3) that material of M, -200,000, which probably represents the liver
contained brain tubulin and microtubule-binding proteins version (Kotani et al., 1988) of MAP4 (Parysek et al., 1984a,
from liver. The pellet was resuspended in buffer and spun 1984b).
again. Proteins that were nonspecifically entrapped in the
Monoclonal Antid8K Antibodies-To facilitate further
pellet, as well as some microtubule-binding proteins and tu- characterization of 58K we generated five mouse monoclonal
bulin, were solubilized at this step (lane 4 ) , and the washed antibodies that react specifically with the protein (see Figs. 2,
microtubule pellet (lane5)was resuspended in buffer contain- 4, and 6-8). All of the antibodies belong to theIgGl class, as
ing 0.5 M KCl. A final centrifugation step produced a super- judged by double immunodiffusion uersus rabbit antisera spenatant (lane 6 ) containing all of the microtubule-binding cific for mouse antibody isotypes. As can be seen in Fig. 2,
proteins and a small amount of tubulin and a pellet ( l a n e 7) when a mixture of all five antibodies was used to stain ablot
composed of nearly pure tubulin. 58K could then be purified of rat liver homogenate, a single immunoreactive band with
to homogeneity (Fig. 1B) by gel filtration chromatography of the electrophoretic mobitity of58Kwas detected. Identical
the isolated microtubule-binding proteins (as in Fig. LA, lane results were obtained when each antibody was used individ6 ) .Seven rat livers weighing a total of about 100 g were used ually. Each antibody was immunoreactive with a unique specas starting material for typical preparations and routinely trum of CNBr fragments of 58K on immunoblots (not shown)
yielded 1.5-2.0 mg of purified 58K.
and cell lines by immunofluorescencemicroscopy(Fig.7).
16086
A Protein That Associates with Microtubules and the Golgi
FIG.2. Specificity of monoclonal antibodies to 58K. A hoanalyzed by SDS-PAGE ( l a n e I ) and
mogenate of ratliverwas
immunoblotting with a mixture of all five monoclonal antibodies to
58K ( l a n e 2). A single immunoreactive band of M,58,000 was observed.
Hence, it seems likelythat each of the five antibodies recognizes a distinct epitope on 58K. We refer to thefive clonesas
anti-58K-2, -4, -7, -9, and -12.
58K Is a Novel Protein-The electrophoretic mobility of
58K falls within the range of that for the tau MAPs of brain,
which suggestedto us that 58K might be a liver form of tau
protein. To test for that possibility, as well as for the chance
that 58K might be a variant of some other similarly sized
cytoskeletal protein, such as tubulin, vimentin, or keratin,
peptide mapping and immunoblotting experiments were performed.
Fig. 3A illustrates the results of a peptide mapping experiment, in which protease V8 was used for limited proteolysis
of equivalentamounts of electrophoreticallypurified tau, 58K,
a-tubulin, and/3-tubulin during a second electrophoresis step
(Cleveland et al., 1977). The peptide map of 58K was clearly
unique, but bore some resemblanceto the fragment patterns
of the tau proteins. This result suggested that 58K might be
structurally related to tau, prompting us to compare the two
proteins by two additional sets of peptide mapping experiments. First, we repeated the experiment just described but
substituted chymotrypsin for protease V8. The peptide maps
of 58K and the tau proteins were substantially dissimilar in
that case (Fig. 3B). Next, equivalent amounts of purified,
native tau, or 58K were mixed in solution with proteolytic
enzymes, and time courses of digestion were analyzed by SDSPAGE (Fig. 3C). Protease V8 digested tau progressively and
nearly completely over
the course of 54 min (lanes 2-61, during
which time 58K remained practically unaffected by the enzyme (lanes 7-1 1 ). A similar time course usingchymotrypsin
resulted in the complete digestionof tau within 2 min (lanes
13-16) and the gradual cleavage of 58K (lanes 17-20) during
nearly 1h of exposure to theprotease. These peptide mapping
studies strongly suggested that 58K is not simply a previously
unrecognized formof tau or tubulin.
Two additional sets of data indicate that 58K is a novel
protein. First, five distinct monoclonal antibodies to tubulin,
two monoclonals specific for tau (Binder et al., 1985; Papasozomenos and Binder, 1987), and polyclonal antibodies to
vimentin and keratin all failed to cross-react with 58K on
immunoblots (not shown). Conversely,none of the five monoclonal anti-58Ks reacted on immunoblots with tubulin, tau,
vimentin, or keratin (see Figs. 2 and 4, for example). Next,
and most importantly, eight tryptic peptides of 58K were
purified and sequenced. The sequences, whichtotal 127 amino
acidresidues, were compared with thedata bank of the
Protein Identification Resource of the National Biomedical
Research Foundation and a recently published complete sequence of tau (Lee et al., 1988).The net result of these efforts
is that the 58K sequences bear only random resemblance to
any others that have been reported. Collectively, the structural studies just described provide compelling evidence
that
58K is a novel protein.
In Vitro Interactions of 58K with Microtubules-The purification protocolfor 58K rests in part on its binding to
preformed microtubules in cytosolicextracts of liver (seeFig.
1). To study the specificity of this binding in greater depth
and to analyze the binding quantitatively, further experiments
were performed. First, we found the sedimentability of 58K
out of liver extracts to be dependent on the presence of
exogenous microtubules and proportional to their concentration (Fig. 4). When taxol alone was added to anice-cold liver
extract that was subsequently centrifuged, no pellet was observed, and immunoblotting indicated that all of the 58K
remained in the supernatant (lanes 1 and 2). When brain
tubulin preassembled with taxol was added to comparable
aliquots of liver extract, though, 58K became sedimentable,
as judgedby immunoblotting (lanes 3-12). Although 58K
comigrated with a-tubulin on this gel, immunoblotting indicated that the amount of 58K recovered in the pellets increased in proportion to theamount of exogenous assembled
tubulin, and all of the detectable 58K was pelleted at added
tubulin concentrations of 0.5 mg/ml or greater. Note also the
enrichment of other non-tubulin proteins, such as dynein, in
the microtubule pellets.
The stimulation of tubulin assembly is a hallmark feature
of most, if not all MAPs that have been examinedso far. This
property is also exhibited by 58K, as demonstrated in Fig. 5.
In this experiment, GTP was added to purified bovine brain
tubulin in the presence or absence of 58K, and the mixture
was incubated at 37 "Cand centrifuged. When58K was omitted, nearly all of the tubulin remained soluble (lanes 1 and 2),
while a substantial pellet was obtained when 58K was present
(lanes 3 and 4 ) . Microtubules assembled from purified 58K
and tubulin were also examined by immunofluorescence microscopyto verify that 58K had actually bound to polymerized
tubulin (Fig. 6a).These biochemical and immunofluorescence
results indicate that 58K binds directly to the outer surface
of the tubulin backbone of microtubules. When observed by
electron microscopy, polymers composedof tubulin and 58K
exhibited the typical, smooth-walled appearance of microtubules lacking high molecular weight MAPs (Fig.
6, b and c).
The stimulation of tubulin assembly by 58K was also studied quantitatively by incubating 0.8 mg/ml tubulin with 1mM
GTP in the presence of 58K ranging in concentration from 0
to 0.5 mg/ml. After a 15-min incubation at 37 "C, the samples
were centrifuged,and theamounts of tubulin and 58K in each
microtubule pellet were measured by quantitative SDS-PAGE
(data not shown). In all samples tested, -50% of the 58K was
sedimentable,while the proportion of tubulin that had polymerized increased linearly from 3% in the absence of 58K to
50% in the presence of 0.5 mg/ml 58K. Assuming a native
molecular weightof 58,000 for 58K, the molar ratio of 58K to
A Protein That Associates with Microtubules and the Golgi
B
-
A
_.
1
2
3
16087
4
5
1
2
3
C
I
2-6
-17-20
13-16
,
FIG. 3. Comparative peptide mapping of 58K, tau, and tubulin. A , purified preparations of rat liver 58K
and bovine brain tau andtubulin were subjected to SDS-PAGE. Five electrophoretically distinct tau variants were
excised as a group and mounted in a single lane on a separate SDS-polyacrylamide gel with the slowest migrating
species on the left. Gel chips containing 58K, a-tubulin, or 8-tubulin were mounted in other lanes. 4 pg of total
protein was loaded into each lane, and thegel chips were overlaid with 1.25 pg of protease V8. Limited proteolysis
during electrophoresis was then performed by the method of Cleveland et al. (1977). The lanes andtheir
corresponding proteins were: 1, tau; 2, 58K, 3, a-tubulin; 4, 8-tubulin; 5, protease V8 alone. Note the slight
similarity between the peptide maps of 58K and the tauproteins. B, this experiment was performed exactly like
that in A , except that 1.25 pg of chymotrypsin was used in place of protease V8. The proteins loaded in each lane
were: 1, chymotrypsin alone; 2, 58K 3,tau. The peptide maps of 58K and tau were very distinct. C, native bovine
brain tau or rat liver 58K at 50 pg/ml each were mixed in solution for varying periods of time with either protease
V8 or chymotrypsin at 5 pg/ml each. Proteolysis was halted a t each time point by the addition of an SDScontaining solution and boiling of the samples, and SDS-PAGE was used to analyze the samples. Lane 1 , protease
V8 alone; lunes 2-6, tau digested with protease V8 for 0,2,6, 18, and 54 min, respectively; lanes 7 -11,58K exposed
to protease V8 for 0,2,6,18, and 54 min; lane 12, chymotrypsin alone; lanes 13-16, tau incubated with chymotrypsin
for 2, 6, 18, and 54 min; lanes 17-20, 58K mixed with chymotrypsin for 2, 6, 18, and 54 min. 58K was far more
resistant to digestion by both proteases than was tau. The gels in all three sections of this figure were stained by
an ammoniacal silver method (Wray et al., 1981).
A Protein
That
Associates
16088
with Microtubules and the Golgi
1 2 3 4
-0
/58K
>T
1 2
34
5 6
78
".,.-.._
910 1112
*
c
+
2
.
.
"
"
.
-
FIG. 4. The sedimentability of 58K out of rat liver cytosol
depends on the presence of microtubules and is proportional
to their concentration. A concentration series of MAP-free, taxolstabilized microtubules was added to aliquots of rat liver cytosol, and
the solutions were centrifuged. The supernatants (lanes 1, 3,5, 7, 9,
11) and pellets (lanes 2, 4,6, 8, 10, 12) were then analyzed by SDSPAGE (top) and immunoblotting (bottom) with a mixture of clones
2, 7, and 12 of monoclonal anti-58K antibodies. Only the region of
the blot in the immediate vicinity of 58K is shown. No other immunoreactive material was observed. The final concentrations of exogenous microtubules (expressed as mg/ml) were 0.0 (lanes 1 , 2 ) , 0.125
(lanes 3,4 ) , 0.25 (lanes 5, 6 ) , 0.5 (lanes 7, 8 ) , 1.0 (lanes 9,l o ) , and
2.0 (lanes I1,12). Because 58K and a-tubulin comigrated on this gel,
the levels of 58K in each lane can be judged most readily on the
immunoblot. Note that: 1) 58K remained completely soluble in the
absence of added microtubules; 2) there was a progressive decrease in
the amount of 58K that remained soluble as the concentration of
added microtubules was raised between 0.125 and 0.5 mg/ml; 3 ) at
exogenous microtubules a t 0.5 mg/ml or higher all of the detectable
58K pelleted; and 4) the levels of other liver MAPs, such as dynein,
rose in the microtubule pellets as the concentrationof added microtubules increased. The positions of 58K, tubulin (T),
and dynein (D)
are indicated to the right.
tubulin in the microtubules was relatively invariant andclose
to unity (1.15 f 0.31) at all concentrations of 58K that were
tested. Compared to MAPs that have been purified from other
sources (reviewed in Olmsted, 1986; Vallee, 1986; Vallee and
Collins, 1986), 58K is a weak stimulator of tubulin assembly
but saturates microtubules at a high stoichiometric level relative to tubulin. Apparently, therefore, the binding sites on
microtubules for 58K are unusually large in number, but low
in affinity compared to those for conventional MAPs. These
results, as well as our findings concerning the intracellular
distribution of 58K (see below), distinguish 58K from traditional MAPs and imply that its in vivo functions do not
include regulation of tubulin assembly.
58K Is Associated with the Cytoplasmic Surface of the Golgi
Apparatus-The subcellular distribution of 58K was studied
initially by immunofluorescence microscopy of three cultured
FIG.5. Purified 58K stimulates tubulin polymerization. Purified tubulin (T)at 0.8 mg/ml was incubated in the presence of 1
mM GTP for 15 min a t 37 "C in the absence (lanes 1 , 2 ) or presence
(lanes 3, 4 ) of 0.5 mg/ml 58K. The samples were then centrifuged,
(lanes I, 3)and pellets (lanes 2,4)were analyzed
and the supernatants
by SDS-PAGE. Note that tubulin pelleted appreciably only when
58K was present.
hepatoma cell lines, Faza (rat), Hepa-1-6-J (mouse),and HepG2 (human). Each of the five anti-58K monoclonals were
tested, and three clones were found to react with the cells.
Anti-58K-2 reacted with Faza cells and yielded perinuclear
staining (Fig. 7a). The same type of stained structures were
also observed in all three hepatoma lines incubated with anti58K-9, and essentially identical patterns were seen in the
mouse and human hepatoma cells treated with anti-58K-12.
The structures stained by the anti-58Ks were colocalized
extensively with cytoplasmic regions decorated by fluorescently labeled wheat germ agglutinin (see Fig. 7, b and c),
which specifically binds to the Golgi apparatus (Tartakoff
and Vassalli, 1983). Colocalization of 58K and binding sites
for wheat germ agglutinin was observed additionally in cells
grown in the presence of taxol (Fig. 7, d and e) or colchicine
(Fig. 7, f and g), both of which disrupt the organization of
Golgi elements (Wehland et al., 1983). These observations,
combined with our finding that 58K binds to microtubules
(Figs. 1 and 4-6), raised the possibility that the intracellular
function of 58K is to cross-link the Golgi apparatus tomicrotubules.
To accomplish that function, it would benecessary for 58K
to be exposed on thecytoplasmic side of the Golgi. Evidence
that this is, indeed, the case was obtained by isolating and
subfractionating rat liver Golgi membranes and by analyzing
the samples by immunoblotting with anti-58K(Fig. 8).A well
establishedmethod optimized for separating Golgi membranes from other particulate fractions inrat liver (Leelavathi
et ul., 1970) was used and, as can be seen (lane I), readily
detectable levels of 58K were present in theGolgi isolates.
Additional experiments were conducted to determine specifically where 58K is located within the isolated Golgi membranes (Fig. 8). First, most of the 58K (-65-90%) was found
to partition into the
aqueous phase when the membranes were
A Protein That Associates with Microtubules and the Golgi
16089
apparatus. Further evidence for this hypothesis was obtained
by demonstrating that exposure of isolated Golgi membranes
to chymotrypsin in either the presence or absence of Triton
X-100 resulted in substantialdegradation of 58K (lanes 6-9).
In order to have been accessible to theprotease in theabsence
of the detergent, it is likely that 58K was present onthe outer
or cytoplasmic surface of the Golgi membranes.
DISCUSSION
The structural integrity and intracellular location of the
Golgi apparatus appears to be dependent upon the presence
and organization of microtubules (Kupfer et al., 1982, 1983;
Nemere et al., 1985; Rogalski and Singer, 1984; Tassin et al.,
1985; Wehland et aL, 1983). To enable the behavior of these
distinct cytoplasmic structures tobe so closely coordinated, it
is likely that specific proteins areinvolved in linkingthe Golgi
to microtubules. Here, we describe the identification, purification, and characterization of a novel liver protein that may
fulfill such a task. The protein, 58K, was purified to homogeneity by a procedure that involves its binding to microtubules in aliver extract as the
principal enrichmentstep. Once
purified, 58K can stimulate polymerization of tubulin and
bind to preassembled tubulin. Surprisingly, though, 58K was
not found by immunofluorescence microscopy to be distributed uniformly along microtubules in cultured cells. Instead,
the Golgi apparatus was the primary immunoreactive site in
the cells. Consistent with this observation was our finding
that 58K was also readily detectable by immunoblotting in
Golgi membranes isolated from liver. Within these preparations, furthermore, 58K behaved as a peripheral membrane
protein exposed on the cytoplasmic face of the Golgi. This
places 58K in a location that potentially permits the protein
to bind microtubules, even while it remains associated with
the Golgi.
At least two mechanisms can be envisioned for connecting
the Golgi to microtubules by a cross-bridging protein. One of
these would have the cross-linker firmly bound to microtubules and, distributed throughoutthe entirenetwork of cytoplasmic microtubules. This would permit attachmentof Golgi
membranes to microtubules virtually anywhere in the cytoplasm, so it is hard toimagine how such a mechanism could
confine the Golgi to its characteristic location in the vicinity
of the MTOC. A means of accomplishing this, though, might
be provided by an alternative model, in which the crossbridging protein would have a primary association with the
outer surface of Golgi membranes. If the linking protein in
this case interacted weakly with microtubules, the Golgi might
FIG. 6. Immunofluorescent and electron microscopic ap- tend to become situated where the number of multiple weak
pearance of microtubules composed of 58K and tubulin. Microtubules assembled from purified preparations of 58K and tubulin interactions with microtubules could be maximized. This lowere examined by immunofluorescent microscopy using anti-58K-2 cation would likely be centered around the MTOC, where
microtubules are most densely packed in the cell. 58K is a
( a ) ,and negative stain ( b ) and thin section ( c ) electron microscopy.
The immunofluorescent image indicates that 58K can bind all along protein that apparently fulfills two key requirements of the
the outer walls of microtubules, whereas such microtubules appear
second model, a primaryassociation in the cell with the Golgi
smooth-walled at theelectron microscopic level. The bars indicate 10 apparatus and relatively weak binding to microtubules. It
pm ( a ) , 35 nm ( b ) ,and 50 nm (c).
seems possible, therefore, that the in vivo function of 58K is
to anchor the Golgi apparatus to microtubules by a mechaextracted with Triton X-114 (lanes 2 and 3). This result nism which passively directs the Golgi to reside near the
suggested that 58K is either a peripheral membrane protein MTOC.
or a soluble component of the interior of the Golgi (Bordier,
Using antibodies made against Golgi-enriched fractions,
1981). To distinguish between these possibilities, Golgi mem- two groups have discovered and begun to characterize Golgibranes were extracted with 0.5 M KC1. This led to the solu- associated proteins in themolecular weight range of 58K. One
bilization of -50-90% of the 58K, the exact amount having of these, a myeloma cell protein of M , 54,000 in SDS-PAGE,
varied among different experiments,while leaving most of the is cytoplasmically oriented and restricted to medial Golgi
Golgi protein in a sedimentableform (lanes4 and 5 ) . A likely stacks (Chicheportiche et al., 1984). The otherprotein, an Mr
explanation for this result is that 58K is a peripheral mem- 58,000 species of pancreatic origin, was found to be localized
brane protein exposed to the cytoplasmic side of the Golgi in the cis-Golgi (Sasarte et al., 1987). One additional Golgi
16090
A Protein That Associates with Microtubules and the
Golgi
1
FIG.7. Colocalization of 58K and the Golgi apparatus in hepatoma cells.
When labeled for immunofluorescence with monoclonalantibodies to58K and tetramethyl rhodamineisothiocyanate-labeled second antibodies,
cultured hepatoma cells exhibited a Golgi-like staining pattern, as observed for the rat Faza line ( a ) . Indeed, the
staining by anti-58K coincided with that obtained using fluorescently labeled wheat germ agglutinin, a specific
cytological marker for the Golgi apparatus (Tartakoff and Vassalli, 1983). This is illustrated in human Hep-G2
cells double-labeled with anti-58K-9 (b, d, f ) and fluorescein isothiocyanate-labeled wheat germ agglutinin ( c , e,
A Protein
That
Associates with Microtubules
and
I
58K-W
1
2 3
4
5
6
7
8
9
FIG. 8. 58K codistributes with cytoplasmically oriented,
peripheral membrane proteins in isolated Golgi apparatus.
Golgi membranes were isolated from rat liver (Leelavathi et al.,1970)
and subfractionated, and the products were analyzed by SDS-PAGE
(top) and immunoblotting (bottom) with a mixture anti-58K-4, -7,
-9, and -12. Only the region of 58K is shown for the immunoblot, and
no other immunoreactive bands were observed. Isolated Golgi membranes were found to contain readily detectable levels of 58K (lane
I ). When themembranes were extracted with Triton X-114 (Bordier,
19811, the majority of the 58K present partitioned into the aqueous
phase (lane 2 ) , while most of the protein remained in the detergent
phase ( l a n e 3 ) . This indicated that 58K is either a peripheral membrane or a soluble, intracisternal protein (Bordier, 1981). To distinguish between these possibilities, Golgi membranes were extracted
with 0.5 M KCI. At least half of the 58K was solubilized (lane 4 ) ,
although most Golgi proteins remained insoluble (lane 5).This result
suggested that 58K is a peripheral membrane protein exposed to the
cytoplasmic face of the Golgi. Further evidence for this was obtained
by incubating isolatedGolgi membranes (14 pg of protein per aliquot)
with or without chymotrypsin (50 pg/ml) a t 37 "C for 1 h in the
absence or presence of 0.5% Triton X-100. When chymotrypsin was
absent, virtually all of the Golgi proteins, including 58K, remained
intact in both the absence (lane 6) and presence (lane 7) of the
detergent. When chymotrypsin was added, though, substantial degradation of 58K and other Golgi proteins occurred when Triton X100 was either absent (lane 8 ) or present (lane 9).The accessibility
of 58K to theprotease in the absence of the detergentindicated that
58K faces the outer or cytoplasmic side of the Golgi.
protein in the size range of 58K has also been described. This
M , 55,000 species was determined to be a major component
of actin gels isolated from brain tissue and was found to be
associated with the Golgi in cultured cells by immunofluorescence microscopy (Sobrio, 1985). What structuralsimilarities,
if any, exist among these three proteins and 58K remain to
be determined.
Besides 58K, a t least one other protein, a 110-kDa species
which is immunologically cross-reactive with MAP2, has also
been proposed to participate in anchoring the Golgi to microtubules (Allan and Kreis, 1986). The 110-kDa protein was
shown to be present in cultured non-neuronal cells, where it
the Golgi
16091
colocalizes with the Golgi, and tocofractionate partially with
rat liver Golgi stacks andmicrotubules isolated from cultured
chick embryo fibroblasts. The accessibility to proteases of the
110-kDa protein in isolated Golgi membranes and itssolubility properties in such isolates suggested that, like 58K, the
protein is exposed to thecytoplasmic face of the Golgi. Direct
binding of the 110-kDa protein to polymerized tubulin was
not demonstrated, though, leaving open the question of how
it associates with microtubules. It also remains to be established whether or not the 110-kDa species is the same as an
identically sized Golgi protein described earlier (Linand
Queally, 1982). The results presented here do not necessarily
conflict with those reported for the 110-kDa protein. It is
possible, for example, that both 58K and the110-kDa species
belong to a complex of proteins involved in forming crossbridges between the cytoplasmic surface of the Golgi and the
tubulin backbone of microtubules. Further studies of the Golgi
in rat liver, where both the 110-kDa protein (Allan and Kreis,
1986) and 58K (this paper) have been found, will be needed
to define the functional relationship between these two proteins.
One of the most intriguing featuresof 58K is that itexhibits
several hallmark properties of MAPs in vitro but does not
appear by immunofluorescence microscopy to be associated
with microtubules in the cell. This raises the question of
whether or not 58K should be classified as a MAP. Although
there is no universally accepted definition of a MAP, both
biochemical and cytological criteria have been invoked to
place proteins in that category. The major enrichment step
for 58K is a typical taxol-based procedure (Bloom et al.,
1985b) for isolating MAP-containing microtubules from cytosol. Like virtually all MAPs that have been examined
(Olmsted, 1986; Vallee, 1986; Vallee and Collins, 1986), 58K
remains associated with taxol-stabilized microtubules in the
presence of low,but nothigh salt, stimulates tubulinassembly,
and binds to microtubules in a saturable manner (Figs. 1 and
4-6). Biochemically, therefore, 58K seems to be atypical
MAP in many respects.
Upon closer examination, though, 58K can be distinguished
biochemically from typical MAPs in two fundamental ways.
First, 58K is not a potent promotor of microtubule assembly,
as relatively high levels of the protein are required to enable
tubulin to polymerize. Second, in contrast to standard
MAPs,
58K saturates microtubules at levels equimolar, rather than
substoichiometric to the assembled tubulin. A potential explanation for these observations is provided by the recently
described complete sequences of tau (Lee et al., 1988) and
MAP2 (Lewis et al., 1988). Three copies of closely related 18residue sequences were found in both proteins and each of
these sequences has been hypothesized to represent abinding
site for a single tubulin molecule within a microtubule. The
%fold repeat present in MAP2 and tau may enable those
proteins to bind microtubules by spanning multiple tubulin
subunits and account for the potency with which they stimulate tubulin assembly and stabilize microtubules (Lewis et
al., 1988). Maybe 58K, in contrast to tau and
MAP2, contains
only a single microtubule binding domain. That might explain
how 58Kstimulates tubulinassembly only modestly and binds
microtubules with equimolar stoichiometry to tubulin. Another potentially relevant consequence of this model is that
g). 58K and Golgi components were colocalized in control (drug-free) cultures ( b , c), as well as in cultures treated
taxol ( d , e ) or 10 p~ colchicine ( f ,g). Cells that had been double-labeled with anti-58K-7, which
for 4 h with10 ~ L M
did not react with Hep-G2 cells, and fluorescein isothiocyanate-wheat germ agglutinin exhibited background levels
of rhodamine fluorescence ( b ) ,while Golgi components were readily visible in the fluorescein channel (i). The bar
in h indicates 10 pm.
16092
A Protein
That
Associates with Microtubules and the Golgi
the presence of only one microtubule binding site per 58K electron microscopy; and Ying-Chun Lin for isolating and charactermolecule might result inrelatively weak binding to theprotein izing rat liver Golgi membranes.
to microtubules compared to tau and MAPS. Whether these
REFERENCES
interpretations are correct or not will require further investi- Aebersold, R. H., Leavitt, J., Saavedra, R. A,, Hood, L. E., and Kent, S. B. H.
(1987)Proc. Natl. Acad. Sei. U. S. A. 84,6970-6974
gation, but there clearly are significant biochemical differ- Allan,
V. J., and Kreis, T. E. (1986)J. Cell Biol. 103, 2229-2239
Andrews. P. C.. and Dixon. J. E. (1987)Anal. Biochem. IC' "*"
""-528
ences between 58K and previously described MAPs.
Binder,
L. I., Frankfurter, A., and'Rebhun, L. I. (1985)J. Cell Biol. 101,1371The intracellular distribution of 58K further distinguishes
137R
the protein from traditional MAPs. Nearly all MAPs that BlakeiM. S., Johnston, K. H., Russell-Jones, G. J., andGotschlich, E. C. (1984)
Anal. Biochem. 136,175-179
have been examined immunocytochemically have been shown Bloom,
G. S., Schoenfeld, T. A,, and Vallee, R. B. (1984a)J. Cell Biol. 98,320330
to be extensively colocalized with microtubules in the cell
Bloom, G. S., Luca, F. C., and Vallee, R. B. (1984b)J. Cell Biol. 98,331-340
(Olmsted, 1986). In contrast,58K appears by immunofluores- Bloom, G. S., Luca, F. C., and Vallee, R. B. (1985a)Proc. Natl. Acad. Sei. U. S.
cence microscopy to be most concentrated on the Golgi ap- B Al o. ?82,5404-5408
~ G.
, S., Luca, F. C., and Vallee, R. B. (1985b)Biochemistry 24,4185paratus (Fig. 7). Even though 58Kmay be able to bind
41Yl
Bloom,
G.
S., Wagner, M. C., Pfister, K. K., and Brady, S. T. (1988)Biochemcytoplasmic microtubules while maintaining a connection to
istry 27,3409-3416
Bordier,
C.
(1981)J. Biol. Chem. 256,1604-1607
the outer surface of the Golgi, the primary association of this
M. M. (1976)Anal. Biochem. 72,248-254
protein with a structure other thanmicrotubules represents a Bradford
Chichepdrtiche, Y., Vassalli, P., and Tartakoff, A.M. (1984)J. Cell Biol. 99,
2200-2210
profound distinction between 58K and traditionalMAPs.
Cleveland, D. W., Fischer, S. G., Kirschner, M. W., and Laemmli, U. K. (1977)
If one considers MAPs to be proteins whose principal
J. Biol. Chem. 252, 1102-1106
N., Pfister, K. K., Yorifuji, H., Wagner, M. C., Brady, S. T., and
affiliation in the cell is with microtubules, then 58K is clearly Hirokawa,
Bloom G. S. (1989)Cell 56,867-878
not a MAP. Instead, 58K appears to be a protein that resides Kim, H.,'Binder, L. I., and Rosenbaum, J. L. (1979)J. Cell Biol. 80,266-276
S., Murofushi, H., Maekawa, S., Aizawa, H., and Sakai, H. (1988).J.
on a membrane-bounded organelle known to interact with Kotani,
Biol. Chem. 263,5385-5389
A., Louvard, D., and Singer, S. J. (1982)Proc. Natl. Acad. Sci. U. S. A.
microtubules, and whose role in the cell might require simul- Kupfer,
79.
.- , 2- - m
- - --2 - 1. ~
taneous binding to theorganelle and themicrotubule. In that Kupfer, A., Dennert, G., and Singer, S. J. (1983)Proc. Natl. Acad. Sci. U. S. A .
80, 7224-7228.
and several other regards, 58K resembles kinesin, a microtu- Kuznetsov,
S. A., and Gelfand, V. I. (1986)Proc. Natl. Acad. ScI. U. S. A. 83,
853n-m4
bule-stimulated ATPase (Kuznetsovand Gelfand, 1986; Wag"_"
Kuznetsov, S. A,, Vaisher , E A ,Shanina, N. A,, Magretova, N. N., Chernyak,
ner et al., 1989), and a likely motor protein for organelle
V. Y., and Gelfand, V. f (1988) EMBO J. 7,353-356
transport along microtubules. Like 58K, kinesin is purified Kuznetsov, S. A,, Vaisberg, Y. A,, Rothwell, S. W., Murphy, D. B., and Gelfand,
V. I. (1989)J.Biol. Chem. 264,589-595
by procedures involving a microtubule affinity step (Kuznet- Laemmli U. K. (1970)Nature 227,680-685
Langford
G. M. (1983)J. Ultrastrut.Res. 85,l-10
sov and Gelfand, 1986; Scholey et al., 1985; Vale et al., 1985; Lee, G., dowan,
N., and Kirschner, M. (1988)Science 239,285-289
Wagner et al., 1989), but does not appear by immunofluores- Leelavathi, D. E., Estes, L.W., Feingold, D. S., and Lombardi, B. (1970)
Bfflchtm.
B
K
X
Acta
S . 211.124-138
cence to be distributed evenly along microtubules in the cell. Lewis S. A., w a nI ~ VD.,
and Cowan, N. J. (1988)Science 242,936-939
Using a library of monoclonal antibodies to kinesin heavy and L n , J: J.-c., a nd ueall ,S. A. (1982)J. Cell Biol. 92, 108-112
Murphy, D. B. (1982)idethods Cell Biol. 24.31-49
light chains, we have shown by immunofluorescence micros- Nemere, I., Kupfer, A,, and Singer, S. J. (1985)Cell Motil. 5,17-29
J. B. (1986)Annu. Reu. Cell Biol. 2,421-457
copy that kinesin is maximally concentrated in cells on mem- Olmsted,
Papasozomenos, S. C., and Bmder, L. I. (1987)Cell Motil. Cytoskeleton 8, 210226
brane-bounded organelles that apparently move along microL. M., Asnes, C. F., and Olmsted, J. B. (1984a)J. Cell Biol. 99,1309tubules (Pfisteret al., 1989).Despite the functional differences PaGsek,
1315
between kinesin and 58K, both proteins can be thought of as Parysek, L. M., Wolosewick, J. J., and Olmsted, J. B. (1984b)J. Cell Biol. 99,
2287-2296
organelle-bound, microtubule-interacting proteins (Allan and Paschal,
B. M., Shpetner, H. S., and Vallee, R. B. (1987)J. Cell Bwl. 105,
-1273-1
- .- --2R3
Kreis, 1986), which integrate the behavior of membranePfister, K. K., Wagner, M. C., Stenoien, D. L.,Brady, S. T., and Bloom, G. S.
bounded organelles and microtubules. We (Bloom et al., 1988;
(1989)J. Cell Biol. 108, 1453-1463
Rogalski, A. A., and Singer, S. J. ( 1 9 8 4 ) J. Cell Biol., 99, 1092-1100
Hirokawa et al., 1989; Pfister et al., 1989) and others (Kuz- Saborro,
J. L., Dlaz-Bamga, F., Duran, G., Tsutsum~,V., and Palmer,E. (1985)
J. Bid. Chem. 260,8627-8636
netsov et al., 1988,1989; Scholey et al., 1989; Yang et al., 1989)
Sasarte, J., Palade, G. E., and Farquhar, M. G. (1987)J. Cell Biol. 105,2021have recently made significant strides toward understanding
9nvo
the function of kinesin in terms of its submolecular structure. S z o i i y , J. M., Porter, M.E., Grissom, P. M., and McIntosh, J. R. (1985)
Nature 318,483-486
The challenge of achieving that goal for 58K now awaits us.
Scholey, J. M., Heuser, J., Yang, J. T., and Goldstein, L. S. B. (1989)Nature
* " I ,
"I_
&
"
Acknowledgments-We would like to thankDr. Matthew Suffness
of the National Institutes of Health for kindly supplying us with
taxol, Dr. Lester Binder of the University of Alabama a t Birmingham
for his gifts of monoclonal anti-tau, and Dr. Robert Goldman of
Northwestern University for generously providing polyclonal antibodies to vimentin and keratin. Our appreciated is extended to Dr.
Clive A. Slaughter, Kim Orth, and Carolyn R. Moomaw ofthe Howard
Hughes Medical Institute branch at the University of Texas Southwestern Medical Center for preparing, purifying, and sequencing 58Kderived peptides. In addition, we are grateful to several members of
our department, including Drs. William Snell, Scott Brady, Fred
Grinnell, Kevin Pfister, and Richard Anderson, for their helpful
comments and thoughtful advice; Tammy Burgess for assistance with
338,355-357
Smith, P. K., Krohn, R. I., Hermanson, G. T.,Mallia, A. K., Gartner, E. H.,
Provenzano, M. D.,Fujimoto, E. K., Gaeke, N. M., Olson, B. J., and Klenk,
D. C.(1985)Anal. Biochem, 150,76-85
Tartakoff, A. M., and Vassalh, P. (1983)J. Cell Biol. 97,1243-1248
Tassin, A. M., Paintrand, M., Berger, E. G., and Bornens, M. (1985)J. Cell
Biol. 101,630-638
Vale R. D., Reese, T. S., and Sheetz, M. P. (1985)Cell 42,39-50
Vallie, R. B. (1982)J. Cell Biol. 92,435-442
Vallee, R. B. (1986)Methods Enzymol. 134,104-115
Vallee R. B. and Borisy G. G. (1978)J. Biol. Chem. 263,2834-2845
Vallee' R. 8.'and Collin;, C. A. (1986)Methods Enzymol. 134,116-127
Wa i r , M. k., Pfister, K. K., Bloom, G. S., and Brady, S. T.(1989)Cell Motil.
&oskeleton 12, 195-215
Wehland, J., Henkart, M., Klausner, R., and Sandoval, I. V. (1983)Proc. Natl.
Acad. Sci. U. S. A . 80.4286-4290
Wray, W., Boulikas, T., Wray, V. P., and Hancock, R. (1981)A d . Biochem.
118,197-203
Yang, J. T., Laymon, R. A., and Goldstein, L. S. B. (1989)Cell 56,879-889