Protein 4.1R, a Microtubule-associated Protein Involved in Microtubule Aster Assembly in Mammalian Mitotic Extract*

Non-erythroid protein 4.1R (4.1R) consists of a complex family of isoforms. We have shown that 4.1R isoforms localize at the mitotic spindle/spindle poles and associate in a complex with the mitotic-spindle organization proteins Nuclear Mitotic Apparatus protein (NuMA), dynein, and dynactin. We addressed the mitotic function of 4.1R by investigating its association with microtubules, the main component of the mitotic spindles, and its role in mitotic aster assembly in vitro . 4.1R appears to partially co-localize with microtubules throughout the mitotic stages of the cell cycle. In vitro sedimentation assays showed that 4.1R isoforms directly interact with microtubules. Glutathione S -trans-ferase (GST) pull-down assays using GST-4.1R fusions and mitotic cell extracts further showed that the association of 4.1R with tubulin results from both the mem-brane-binding domain and C-terminal domain of 4.1R. microtubule-associated protein; associates with microtubules in the microtubule anti-HP, anti-dynactin, anti-dynein, anti- (cid:1) -tubulin, and anti-actin antibodies.

Protein 4.1R is an 80-kDa structural component of the red blood cell cytoskeleton. 4.1R is critical to the formation of the spectrin/actin/4.1R junctional complex, whose integrity is vital for the mechanical stability and elasticity of the red blood cells (reviewed in Ref. 1). The 80-kDa 4.1R of mature red blood cells is the prototype of a large array of 4.1R isoforms arising from a single gene by utilizing different promoters (2), alternative mRNA splicing (3,4), two translation initiation sites (5,6), and post-translational modifications (7). Three homologues of 4.1R, namely 4.1G, 4.1N, and 4.1B, have been identified (8 -11). They share a common structural organization with 4.1R: highly conserved protein interaction domains (the 30-kDa membranebinding domain (MBD 1 ), the 10-kDa spectrin/actin-binding domain (SAB), and the 22-to 24-kDa C-terminal domain (CTD)) interspersed with unique domains. The N-terminal extension domain (head-piece, or HP) and exon 13 in the 16-kDa domain are unique to 4.1R. Thus, 4.1R function(s) in non-erythroid cells may be specific or sustainable by redundant systems involving other family members.
Regulated alternative splicing is a characteristic protein 4.1 family feature (3,4,12). Multiple isoforms of 4.1R are present in many non-erythroid cell types. The prevalence and homology of these proteins in many species implies their basic importance to cell structure and function. 4.1R isoforms appear to be critical components of many important supramolecular complexes. 4.1R isoforms have been shown to interact with interphase microtubules in human T cells (13), the components of the contractile apparatus in skeletal myofibers (14), and the tight junction proteins in epithelial cells (15). 4.1R isoforms have also been shown to be the components of the nuclear matrix (16,17) and may be involved in splicing processes (18). 4.1R is reorganized during the cell cycle; it localizes in the nucleus and cytoplasm of interphase cells and rapidly redistributes to the developing spindle poles when the nuclear envelope dissembles in prometaphase (16,17,19,20). In keeping with the key transitions in the cell cycle controlled by the cyclin-dependent kinases, the redistribution of 4.1R to the mitotic spindle and spindle poles is regulated by a phosphorylation event that occurs at Thr-60 and Ser-679 sites by cyclin-dependent cdc2 kinase. 2 We (20) previously showed that a 135-kDa 4.1R isoform may participate in mitosis via its association with the non-centrosomal proteins NuMA, dynein, and dynactin, which are essential to the organization of functional spindle poles and tethering of the centrosomes to the spindle pole (21). From these results, we hypothesize that 4.1R plays an important role in organizing mitotic spindle and spindle poles.
A central event during cell division is the transformation of an interphase network of microtubules into a bipolar spindle that mediates the accurate segregation of chromosomes during both mitosis and meiosis. Microtubules in the spindle are organized such that the minus ends are at or near the poles while the plus ends extend toward the cell cortex or chromosomes (22,23). Recent experiments have begun to define the mechanisms and molecules involved in organizing microtubules into spindles (reviewed in Ref. 24). In somatic cells, centrosomes are the dominant centers of microtubule nucleation. However, centrosomes and the microtubule arrays nucleated from them are not sufficient to act as functional spindle poles (25). The organization of microtubules into spindles is governed largely by the interaction of microtubules and microtubule ends with accessory proteins that regulate microtubule dynamics. Several noncentrosomal proteins such as cytoplasmic dynein (26), dynactin (27), Eg5 (28), HSET (29), ch-TOP (30), cohesin (31), and astrin (32), as well as the structural protein NuMA (33) are involved in focusing microtubules at spindle poles. Researchers are now actively searching for additional proteins that may associate only transiently with spindle poles and for the proteins that are minor components of the spindle poles.
In this study, we investigate the association of 4.1R with microtubules, the main component of the mitotic spindles, and the role of 4.1R in mitotic aster assembly in vitro. We demonstrate that 4.1R is a mitotic microtubule-associated protein and that 4.1R is critical to the organization of microtubules into asters in HeLa mitotic extracts. Our findings suggest that a non-erythroid 135-kDa 4.1R is an important component of cell division by participating in the formation of mitotic spindles and spindle poles through its interaction with mitotic microtubules.

EXPERIMENTAL PROCEDURES
Antibody Production-Antisera to synthetic peptides of exon 13 of 4.1R (DSADRSPRPTSAPAI) (anti-exon 13) and the recombinant HP of 4.1R (anti-HP) were prepared as described previously (20). The sera were purified on immunoaffinity columns using the Sulfolink kit (Pierce) covalently cross-linked to the synthetic peptides or fusion peptides used for immunization. The purified antibodies were eluted from the columns with 0.1 N glycine (pH 2.8), neutralized with Tris-HCl, pH 9.5, and dialyzed into phosphate-buffered saline. Anti-NuMA mAb was purchased from Oncogene. Anti-dynein mAb (intermediate chain, clone 70.1), anti-actin mAb, and anti-␣-tubulin mAb DM1A were purchased from Sigma. Anti-p150 glued mAb (the largest subunit of dynactin) was purchased from BD Transduction.
In Vitro Microtubule Binding and GST-pull-down Assay-The microtubule binding assay was performed using the Microtubule Binding Protein Assay kit (Cytoskeleton, Inc.) according to the manufacturer's instructions. Equivalent aliquots of supernatants and pellets were analyzed on Coomassie-stained gels.
For the GST-pull-down assay, mitotic cell lysates were prepared in lysis buffer (25 mM Tris-HCl, 100 mM NaCl, 0.2 mM EDTA, 1 mM DTT, 1.0% Nonidet P-40, pH 7.5) supplemented with protease inhibitors and phosphatase inhibitors (5 mM ␤-glycerophosphate, 10 mM NaF, 10 mM Na 4 P 2 O 7 ). The cell lysate was then pre-cleared by mixing with 100 l of GST-Sepharose beads at 4°C for 1 h. Equal amounts of recombinant 135-kDa/GST, 80-kDa/GST, HP/GST, MBD/GST, 16-kDa/GST, SAB/ GST, and CTD/GST fusion proteins bound to glutathione-Sepharose were mixed with 0.5 mg of GST-Sepharose bead pre-cleared mitotic cell lysates at 4°C for 16 h. Beads were washed six times in 4°C with lysis buffer and immunoblotted with antibodies to ␣-tubulin. To verify the quantity and quality of GST and GST-4.1R fusions used in the experiments, a gel loaded with the same amount of fusion proteins was stained with Coomassie Blue.
In Vitro Mitotic Aster Assembly Assay-HeLa (ATCC CCL 2, human cervix epithelioid carcinoma) cells were grown in Dulbecco's modified Eagle's medium supplemented with 10% heat-inactivated fetal bovine serum (Sigma). Highly enriched mitotic HeLa cells were prepared by synchronization of the cells at the G 1 /S boundary of the cell cycle by a double thymidine block (36). Cells were grown in complete medium after being released from the second thymidine block for 4 h, and the mitotic population was enriched by addition of 60 ng/ml of nocodazole for 6 h. The in vitro aster assembly assay was performed using an established protocol (25,30). Briefly, mitotic cells were collected by shake-off and incubated with 20 g/ml cytochalasin B for 30 min at 37°C. Cells were then washed with phosphate-buffered saline and resuspended at a concentration of ϳ3 ϫ10 7 cells/ml in KHM buffer (78 mM KCl, 50 mM Hepes, pH 7.0, 4 mM MgCl 2 , 2 mM EGTA, 1 mM DTT) containing cytochalasin B and protease inhibitors. Cells were Douncehomogenized, and the crude extract was subjected to ultracentrifugation at 100,000 ϫ g for 15 min at 4°C. The supernatant was collected, and latrunculin B (5 g/ml) was added to reduce actin polymerization and the contamination of microtubule pellets with actin and actinassociated proteins. The supernatant was then supplemented with 10 M taxol and 2.5 mM ATP, and microtubule asters were assembled by incubation at 30°C for 60 min. After incubation, samples were processed for indirect immunofluorescence microscopy. The remaining samples were centrifuged through 50% sucrose cushions prepared in KHM at 100,000 ϫ g for 2 h at 4°C to collect mitotic microtubuleassociated proteins in the pellet. Both supernatant and insoluble pellet were collected directly in Laemmli sample buffer for Western blot analysis.
Immunodepletion and Reconstitution Assay-latrunculin B-treated HeLa mitotic extracts described above were subjected to immunodepletion. Thirty micrograms of pre-immune rabbit IgG or 4.1R affinitypurified anti-HP antibodies were coupled to protein A beads and incubated with mitotic extracts for 20 min at 4°C. The beads were spun down, the supernatants were collected, and the depletion process was repeated twice. Both beads and supernatants in each successive depletion were subjected to Western blotting to verify the effectiveness of depletion of 4.1R and the co-depletion of NuMA and tubulin in these extracts. The final supernatants were subjected to the in vitro aster assembly as described under "In Vitro Mitotic Aster Assembly Assay" above. For reconstitution assays, recombinant 4.1Rs were added to 4.1R-immunodepeleted mitotic extracts and subjected to an in vitro aster assembly as described above.
Coimmunoprecipitation and Immunoblotting-Coimmunoprecipitation of 4.1R, tubulin, and NuMA was performed using HeLa mitotic extracts as described previously (20). In brief, mitotic HeLa cells were homogenized in coimmunoprecipitation buffer (25 mM Tris-HCl, pH 7.5, 100 mM NaCl, 1 mg/ml bovine serum albumin, 0.2 mM EDTA, 5 mM iodoacetamide, 2% CHAPS, protease inhibitors, phosphatase inhibitors), and given 20 strokes in a tight-fitting glass homogenizer. The homogenate was centrifuged at 10,000 ϫ g for 10 min at 4°C, and the supernatant was pre-cleared with pre-immune rabbit IgG. Subse-quently, the supernatant was immunoprecipitated with pre-immune rabbit IgG, affinity-purified anti-HP Ab, anti-NuMA mAb, or anti-␣tubulin mAb. The immunoprecipitated samples were examined for the presence of 4.1R, NuMA, or tubulin using its respective antibody. Some chemiluminograms were scanned using the Adobe Photoshop software (Adobe Systems, Inc.), and the protein bands of interest were quantitated by using the National Institutes of Health Image software for the Apple Macintosh computer.
Indirect Immunofluorescence and Imaging-HeLa cells were grown on poly-D-lysine-coated coverslips and subjected to immunofluorescence staining as described previously (20). Mitotic asters assembled in vitro were spotted on poly-D-lysine-coated coverslips, fixed in methanol, and then subjected to immunofluorescent staining as described before (25). The samples were viewed with an Axiovert 200M inverted microscope (Zeiss, Inc.). Asters were viewed under a ϫ100 oil objective. Images were collected using SlideBook4 software and processed using Photoshop. (20) reported earlier that a 135-kDa 4.1R isoform specifically interacts with NuMA and associates in a complex containing NuMA, dynein, and dynactin. The cytoplasmic complex dynein/dynactin/NuMA is known to localize to the spindle poles and tether centrosomes to spindle microtubules (21). NuMA binds directly to tubulin and organizes the spindle poles by stable cross-linking to the microtubule fibers at the poles (37). Because NuMA co-localizes and interacts with both 4.1R and tubulin, we tested the association of 4.1R with tubulin. To examine their topography with respect to the mitotic apparatus, we used antibodies specific to 4.1R and ␣-tubulin to examine their localization at different stages of cell cycle.

Protein 4.1R and Tubulin Partially Co-localize Throughout the Cell Cycle-We
Immunofluorescence microscopy of HeLa cells during interphase revealed that 4.1R was diffusely distributed throughout the cytoplasm but predominantly resided in the nucleus (Fig. 1, Interphase, 4.1R), whereas tubulin localized mainly in the cytoplasm ( Fig. 1, Interphase, Tubulin). Co-localization of 4.1R and tubulin in interphase cytoplasm is most intense around the nucleus (yellow areas), which is produced by superimposing green and red ( Fig. 1, Interphase, Merged). At early prophase, 4.1R and microtubules are concentrated at the vertices of the developing spindle poles (Fig. 1, Early prophase). From late prophase throughout anaphase, 4.1R and tubulin are enriched in a crescent-shaped area and are intensely stained at the spindle and spindle poles (Fig. 1, Late prophase, Metaphase, Anaphase). At telophase, 4.1R was apparent in the cytoplasm as well as in the newly developed nuclei, whereas tubulin was mainly associated with midzone microtubules (Fig. 1, Telophase). No labeling was apparent when primary antibodies were replaced by pre-immune rabbit or mouse IgG or when anti-HP antibodies were pre-absorbed with its antigen (data not shown). We examined the endogenous 4.1R in ϳ100 interphase cells and ϳ50 cells in each stage of mitosis in three separate experiments. The anti-HP Ab consistently stained the endogenous 4.1R as shown in Fig. 1. Thus, the results suggest that 4.1R partially and reversibly co-localizes with tubulin in the cytoplasm of interphase cells and in the spindle and spindle poles during mitosis.
Protein 4.1R Interacts with Tubulin in Vitro and in Mitotic Homogenates-The co-localization of 4.1R and tubulin at the spindle and spindle poles of mitotic cells prompted us to investigate whether mitotic 4.1R isoforms interact with microtubules. Our analysis of 4.1R mRNA expression in HeLa cells 2 indicated that a 135-kDa isoform (4.1R 135 ), containing all of the alternatively spliced exons except exons 3, 14, 15, 17a, and 17b, is a predominant isoform that is expressed in mitotic cells. To determine whether 4.1R 135 interacts directly with microtubules, we used purified recombinant 4.1R 135 protein and pure microtubules in microtubule sedimentation assays to analyze the interaction between 4.1R 135 and tubulin. An 80-kDa 4.1R isoform (4.1R 80 ) that has been shown to interact with microtubules (38) served as a positive control. Taxol-polymerized microtubules were incubated with recombinant 4.1R 135 or 4.1R 80 and sedimented through a 50% glycerol cushion. Although 4.1R 135 and 4.1R 80 alone did not sediment to the pellet (data not shown), both forms were readily detected in the pellet fraction with polymerized microtubules (Fig. 2B, pellet) but not in the supernatant fraction (Fig. 2B, sup), suggesting their interaction with microtubules in vitro. This interaction was specific: a well characterized MAP2 was bound to microtubules, whereas bovine serum albumin did not bind (data not shown).
To confirm the specificity of the interaction between 4.1R and native tubulin in mitotic cells and to identify the domains of 4.1R responsible for the interaction, we performed a GST pull-down assay using homogenates of mitotic HeLa cells. We expressed either the full-length (135 and 80 kDa) of 4.1R or its individual domain (HP, MBD, 16 kDa, SAB, and CTD) ( Fig. 2A) as GST fusion proteins (Fig. 2C). GST-HP migrated with an apparent molecular mass of ϳ80 kDa instead of the calculated ϳ55 kDa, most likely because of post-translational modification. Equivalent amounts of the GST fusions and control GST were bound to glutathione matrices and incubated with homogenates of HeLa mitotic lysates. After extensive washing, we used Western blotting to examine the matrix-bound GST fusion proteins for their ability to absorb native tubulin. Consistent with the in vitro sedimentation analysis, this blotting revealed that tubulin was specifically associated with the GST-4.1R 135 and GST-4.1R 80 (Fig. 2D). When individual domains were tested, only GST-MBD and GST-CTD specifically retained native tubulin (Fig. 2D), suggesting that the MBD and CTD of 4.1R are responsible for its binding to tubulin in mitotic HeLa cells. The binding between the MBD or CTD with native mitotic tubulin was specific in that GST alone did not interact with tubulin. At this time, we do not know what type of post-translational modification is involved in GST-HP. However, it appears that the modification at the HP region does not contribute to the tubulin binding ability of 4.1R, because both the 80-kDa isoform, without the HP, and the 135-kDa isoform, with the addition of the HP, bind to tubulin efficiently.

Protein 4.1R Is a Component of Mitotic Asters Assembled in a
Cell-free Mitotic Extract-To determine whether 4.1R plays an active role in organizing the mitotic spindle assembly, we first asked whether 4.1R is a component of mitotic asters assembled in a well established in vitro mitotic aster assembly system (25,30,32). In addition to the spectrin-actin-4.1R ternary complex interaction (39, 40), 4.1R exhibits a binary interaction with spectrin (39,41) or actin (42). We reduced actin polymerization and eliminated the possible interaction between actin and 4.1R by the addition of latrunculin B to mitotic extracts prior to the assembly process. Microtubules were induced to polymerize with 10 M taxol in the presence of 2.5 mM ATP in the HeLa mitotic extract. Under incubation at 30°C for 60 min, mitotic asters organized into aster-like arrays in a centrosome-independent process that is driven by microtubule motors and structural proteins. Mitotic asters assembled in this extract are composed of microtubules arranged in a radial array that contain NuMA at the central core (Fig. 3A, ϩATP) (25). Consistent with the report of Gaglio and coworkers (33), the microtubules were arranged in poorly organized aggregates rather than astral arrays in the absence of ATP (Fig. 3A, ϪATP). NuMA did not concentrate at the central core in these aggregates.
We separated the asters from the soluble components of the extract by centrifugation through a 50% sucrose cushion, which allows for the sedimentation of microtubule asters with sufficient mass (32). Under this condition, a portion of 4.1R, but not actin, was co-sedimented with NuMA, dynactin, and tubulin (Fig. 3B, lane P); the latter three proteins are known to associate with microtubules in the microtubule pellet. The inclusion of latrunculin B and the use of 50% sucrose cushions eliminated the contamination of actin in the insoluble pellet (Fig.  3B, actin) and ruled out the possibility of 4.1R interacting with actin. These results demonstrate that 4.1R is a mitotic microtubule-associated protein.
The efficiency of the aster formation is calculated as the conversion of both NuMA and tubulin from a soluble form to an insoluble form (33) as judged by the amount of NuMA and tubulin sedimented into the pellet of the 50% sucrose cushion. It is worth noting that there is a high correlation between the morphology of the aster and the efficiency of aster formation. We standardized our assay conditions as described in the "Ex-perimental Procedures" section and analyzed the asters formed from extracts that exhibit assembly efficiencies ranging between 70 to 85%. With such efficiency, we usually observed ϳ20 clear asters per field under a ϫ100 oil objective. In each separate experiment, we viewed ϳ20 fields. Thus, a total of ϳ400 asters examined display approximately the same size and morphology as shown in Fig. 3A.
Protein 4.1R Is Required for Mitotic Aster Assembly in Vitro -To determine whether the formation of mitotic microtubule asters in synchronized mitotic HeLa extracts requires 4.1R, we examined whether the assembly of mitotic asters was perturbed by the depletion of 4.1R from the extract. In three successive depletions prior to the assembly reaction, the extracts were immunodepleted with 30 g of affinity-purified anti-HP antibodies or purified pre-immune rabbit IgG. We used Western blotting to examine the degree to which 4.1R, as well as its associated proteins NuMA and tubulin, were depleted from the extracts. Pre-immune IgG did not immunodeplete any 4.1R, NuMA, or tubulin from the cell lysate, because

FIG. 3. Mitotic asters assembled in a cell-free mitotic extract.
A, mitotic asters assembled under optimal conditions with taxol and with (ϩATP) or without (ϪATP) 2.5 mM ATP for 60 min at 30°C and processed for indirect immunofluorescence using anti-␣-tubulin (green) or anti-NuMA (red) antibodies. Co-localization of tubulin and NuMA is demonstrated through the superimposition of green and red. Bar, 5 m. B, 4.1R is a mitotic microtubule-associated protein in the microtubule pellet. After a 60-min incubation, the extracts were subjected to centrifugation through a 50% sucrose cushion, separated into soluble (S), and insoluble (P) fractions, and subjected to immunoblotting using anti-NuMA, anti-HP, anti-dynactin, anti-dynein, anti-␣-tubulin, and anti-actin antibodies. each successive depletion showed that these proteins were absent in protein A Sepharose-bound fractions (Fig. 4A, Preimmune depleted, lanes P-1, P-2, and P-3). Analysis of the supernatant fractions confirmed that 4.1R, NuMA, and tubulin were present in the supernatant in all three pre-immune IgG depletions (Fig. 4A, Pre-immune depleted, lanes Sup-1, Sup-2, and Sup-3). Anti-HP antibodies efficiently depleted ϳ95% of 4.1R and ϳ5% each of NuMA and tubulin in the first depletion (Fig. 4B, 4.1R-depleted, lane P-1). The second depletion removed the remaining ϳ5% of 4.1R and a small fraction of NuMA as well as tubulin (Fig. 4A, 4.1R-depleted, lane P-2). The third depletion did not remove any 4.1R, NuMA, or tubulin (Fig. 4A, 4.1R-depleted, lane P-3). 4.1R was completely absent from the final supernatant, whereas ϳ95% of NuMA and tubulin were still present in the final supernatant (Fig. 4A, 4.1Rdepleted, lane Sup-3). The in vitro aster assembly was carried out using the supernatant from the third immunoprecipitation.
We spotted equal amounts of assembly reactions from preimmune-depleted and 4.1R-depleted extracts on coverslips and co-stained them for ␣-tubulin and NuMA to examine the efficiency of aster assembly. Three separate aster assemblies were performed. Treatment of the extracts with pre-immune rabbit IgG did not alter either the morphology of the mitotic asters or the distribution of the NuMA protein at the core of these structures (Fig. 4B, Pre-immune). The asters assembled from the pre-immune depleted lysates have similar size and morphology to the untreated control lysates. The assembly efficiency is also comparable to that of the control, in which ϳ20 asters were observed per field. We routinely examined ϳ400 asters per experiment.
However, depletion of 4.1R from the extract with the anti-HP antibody inhibited the assembly of the mitotic asters and yielded randomly arranged microtubules (Fig. 4B, 4.1R-depleted). There were no visible organized asters in all three experiments performed. The failure of aster assembly in the 4.1R-depleted extracts was apparently not due to indirect depletion of NuMA or ␣-tubulin, because Western blotting showed that the amounts of NuMA and ␣-tubulin were comparable in the control (Fig. 4A, Pre-immune depleted, lane  and experimental (Fig. 4A, 4.1R-depleted, lane Sup-3) extracts. These results demonstrated that organization of microtubules into asters required 4.1R or a 4.1R-containing complex in a cell-free system.
To determine whether 4.1R is the functional component de-pleted from the extract using the anti-HP antibody, we tested whether the purified recombinant 4.1R protein would reconstitute mitotic aster formation in the 4.1R-depleted extracts. We first verified the amount of 4.1R to be added into the reconstruction assay by examining the number of copies of 4.1R 135 in a single mitotic cell. For this reason, we immunoblotted known amounts of recombinant 4.1R-HP and serial dilutions of mitotic HeLa extract with anti-HP antibody (Fig. 5A). We used Photoshop to scan the anti-HP chemiluminograms and quantified the protein bands with the NIH Image software for the Macintosh computer. We estimate that each mitotic HeLa cell has ϳ5 ϫ 10 4 copies of 4.1R 135 (Fig. 5A). The amount of recombinant 4.1R 135 at a concentration of 0.35 g is equal to that of endogenous 4.1R 135 from 3 ϫ 10 7 mitotic HeLa cells. The reconstitution assays were performed using extracts isolated from 3 ϫ 10 7 mitotic HeLa cells that have been subjected to immunodepletion with anti-HP antibodies. Supplementation of the 4.1R-depleted extract with purified recombinant 4.1R 135 at concentrations equivalent to 1-to 10-fold of endogenous 4.1R 135 restored mitotic aster formation (Fig. 5B,  1X). The reconstituted assembly reactions exhibited ϳ75% assembly efficiency and resulted in ϳ20 asters per field. The addition of recombinant 4.1R 135 to the 4.1R-depleted extract at quantities exceeding normal endogenous levels (5 times and 10 times molar excess) did not appear to increase the number of aster formations or alter the morphology of the asters (Fig. 5B,  5X and 10X). When centrifuged through a 50% sucrose cushion, recombinant 4.1R 135 was recovered along with NuMA and tubulin from the mitotic microtubule pellets (Fig. 5C). These results suggest that a 135-kDa 4.1R isoform is required for mitotic aster formation in a cell-free system.

Protein 4.1R and Tubulin Associate in Vivo -
The co-localization of 4.1R and tubulin at the spindle and spindle poles, the interaction of 4.1R with microtubules in vitro, and the involvement of 4.1R in mitotic aster assembly prompted us to investigate whether the native 4.1R can associate with native tubulin in vivo. We performed co-immunoprecipitation assays using mitotic extracts from highly synchronized HeLa cells with anti-HP or anti-␣-tubulin antibodies. Anti-NuMA mAb, known to co-immunoprecipitate with 4.1R (20) and tubulin (37), and pre-immune rabbit IgG served as positive and negative controls, respectively. Immunoblot staining using anti-4.1R (anti-HP and anti-exon 13), anti-␣-tubulin, or anti-NuMA antibodies revealed co-precipitated polypeptides (Fig. 6, Bound, upper   FIG. 4. Depletion of 4.1R from mitotic extract inhibited mitotic aster assembly in vitro. A, Western blot analysis of immunodepleted mitotic extracts. Extracts were successively depleted three times with 30 g of pre-immune IgG or anti-HP antibodies. The degree of 4.1R depletion and co-depletion of NuMA and tubulin was determined by the presence of its respective protein in protein A-Sepharose-bound fractions as indicated (protein A-Sepharose-bound fractions are indicated by P-1, P-2, and P-3; unbound fractions by Sup-1, Sup-2, and Sup-3). For control, 25 g of mitotic cell lysates was loaded. B, efficiency of aster assembly in mitotic extracts immunodepleted with rabbit preimmune IgG (Pre-immune) and anti-HP IgG (4.1R-depleted). The immunodepleted extracts were used to assemble asters in vitro. The same volumes of reaction mixtures were placed on coverslips and processed for indirect immunofluorescence using anti-␣-tubulin (green) and anti-NuMA (red) antibodies. The merged yellow color indicates the superimposition of tubulin and NuMA. Bar, 5 m. panel) and their supernatants (Fig. 6, Unbound, lower panel). Anti-HP antibody efficiently immunoprecipitated a 135-kDa 4.1R isoform (Fig. 6A, ␣-HP, lane ␣-HP), whereas control rabbit IgG did not precipitate any immunoreactive band (Fig. 6A, ␣-HP, lane Pre-immune). Analysis of anti-tubulin and anti-NuMA immunoprecipitates for the presence of 4.1R allowed ready detection of a 135-kDa 4.1R isoform in both immunoprecipitates, although at much less intensity (Fig. 6A, ␣-HP, lanes ␣-tubulin and ␣-NuMA). These results suggest that a 135-kDa 4.1R isoform occurs in vivo in a complex with NuMA and tubulin.
We further confirmed the association of 4.1R, NuMA, and tubulin by blotting the immunoprecipitates with anti-␣-tubulin or anti-NuMA antibodies. Analysis of a duplicated co-precipitated blot using anti-␣-tubulin antibody revealed a strong immunoreactive tubulin band from anti-tubulin immunoprecipitates and a less intense tubulin band from both anti-HP and anti-NuMA precipitates (Fig. 6B, ␣-tubulin, upper panel). Similarly, the anti-NuMA antibody detected a strong immunoreactive NuMA band from anti-NuMA precipitates and a less intense band from both anti-HP and anti-tubulin precipitates (Fig. 6C, ␣-NuMA, upper panel). In keeping with our earlier report (20) and that of Gregson and associates (31), our results indicate that NuMA and 4.1R as well as NuMA and tubulin co-precipitated in a complex in vivo. These results also demonstrate that 4.1R, tubulin, and NuMA associate together during the mitotic stage of the cell cycle. The co-precipitation of tubulin by anti-HP antibodies, the detection of 4.1R that co-precipitated with tubulin by anti-HP antibodies, and the molecular mass of the 4.1R isoform that co-precipitates with tubulin all suggest that a 135-kDa 4.1R isoform derived from the upstream translation initiation site associates with tubulin in vivo.
To examine the possible interaction of other isoforms of 4.1R with tubulin, we used a second antibody specific to 4.1R to examine immunoprecipitates of anti-tubulin mAb. A constitutively expressed exon 13 within the 16-kDa domain has been shown to be unique to 4.1R. Exon 13 does not exhibit sequence homology with other 4.1 family members (8 -11), so we used an exon 13-specific antibody to examine 4.1R isoforms that interact with tubulin. We included rabbit pre-immune IgG and anti-NuMA mAb immunoprecipitates as controls. Anti-exon 13 antibody did not detect any band from the pre-immune precipitates, whereas anti-HP precipitates showed a strong 135-kDa 4.1R band (Fig. 6D, ␣-exon 13, lanes Pre-immune and ␣-HP, upper panel). Consistent with our previous report (20) that a 135-kDa 4.1R isoform associates with NuMA in a mitotic complex, anti-exon 13 detected a 135-kDa 4.1R in NuMA immunoprecipitates (Fig. 6D, ␣-exon 13, lane ␣-NuMA, upper panel). Analysis of anti-tubulin immunoprecipitates with the antiexon 13 Ab (Fig. 6D, ␣-exon 13, lane ␣-tubulin, upper panel) revealed an immunoreactive band of 135 kDa. These data further suggest that a 135-kDa isoform of 4.1R interacts with tubulin and NuMA in mitotic HeLa cells.
The results obtained so far suggest that a small fraction of a 135-kDa 4.1R isoform occurs in vivo in a complex with a small fraction of the mitotic apparatus proteins tubulin and NuMA. To estimate what fractions of 4.1R and tubulin associate together, we determined the efficiencies of immunoprecipitation and co-precipitation by analyzing both the protein A Sepharose-bound (Fig. 6, Bound) and unbound (Fig. 6, Unbound) fractions. We used cell lysates from 10 7 mitotic cells with 10 g each of anti-HP, anti-␣-tubulin, and anti-NuMA IgGs for each immunoprecipitation assay. Quantitation from three separate experiments shows that ϳ80% of 4.1R was precipitated by anti-HP Ab that co-precipitated ϳ1% of tubulin and that ϳ50% of tubulin was precipitated by anti-tubulin mAb, which in turn brought down ϳ5% of 4.1R. Consistent with our previous report (20), 4.1R and NuMA co-immunoprecipitated with limited efficiency; ϳ50% of NuMA was precipitated by anti-NuMA mAb that brought down ϳ5% 4.1R. The lower efficiency of co-immunoprecipitation compared with immunoprecipitation suggests that only a fraction of these molecules associate together in vivo. This is consistent with our immunofluorescent staining results, suggesting that subpopulations of 4.1R and tubulin partially co-localize together. DISCUSSION Protein 4.1R has been known to localize at the spindle and spindle poles of mitotic cells (17,19,20); however, little is known about the topography and relation of 4.1R to the major constituent of the mitotic apparatus, the microtubules, or about its function in spindle/spindle poles. In this report, we demonstrate that 4.1R co-localizes with microtubules at the mitotic spindle and spindle poles, co-sediments with microtubules in vitro, co-immunoprecipitates with tubulin in vivo, and associates with tubulin in mitotic HeLa cell extracts through its MBD and CTD. Furthermore, we show that 4.1R is a mitotic microtubule-associated protein and is essential to mitotic aster assembly in vitro in a cell-free extract prepared from synchronized HeLa cells. Based on these results, we propose that non-erythroid 4.1R may contribute to the organization of mitotic spindles.
The cell cycle-dependent localization of 4.1R is unique; it localizes to the nucleus and cytoplasm during interphase and redistributes to the spindle poles during mitosis, at which point 4.1R partially co-localizes with two major constituents of the mitotic apparatus, NuMA, and microtubules. We (20) previously showed that 4.1R directly interacts with NuMA through its C-terminal domain and forms a complex with the spindle pole-organizing proteins NuMA, dynein, and dynactin during cell division. We now show that, in addition to its interaction with NuMA, 4.1R interacts with tubulin in mitotic cells and is a mitotic microtubule-associated protein. It has been previously documented that purified 4.1R from red blood cells interacts specifically with the C-terminal domain of tubulin (38). In human T cells, 4.1R isoforms lacking exon 16 co-localize and interact with interphase microtubules through a 22-amino acid sequence located within exon 10 of the MBD domain (13). Consistent with these reports, our immunofluorescent staining showed a partial co-localization of 4.1R and tubulin in interphase HeLa cells, and in vitro sedimentation experiments confirmed the interaction of an erythrocyte 4.1R 80 isoform with purified microtubules. Moreover, we showed that a major HeLa 4.1R 135 isoform containing all of the alternatively spliced exons except exons 3, 14, 15, 17a, and 17b 2 also sedimented with purified microtubules. Furthermore, we demonstrated that both 4.1R 80 and 4.1R 135 isoforms associate with tubulin in a GST pull-down assay using HeLa mitotic extracts. In contrast to the interaction of 4.1R isoforms with interphase T-cell tubulin (13), an interaction that requires only the MBD, we found that both the MBD and the CTD of 4.1R contribute to its binding to tubulin in HeLa mitotic extracts. Post-translational modifications of tubulin have been shown to regulate its association with microtubule-associated proteins (43). Polyglutamylated tubulin was detected in proliferating cells of different origins (HeLa, KE37, and NIH 3T3), where it associates with the centrioles, the spindle, and the midbody (44). The localization of 4.1R coincides with that of tubulin during mitosis; thus, it is possible that the post-translational modification of tubulin in mitotic cells might facilitate their selective recruitment of 4.1R isoforms through their MBD as well as their CTD into distinct microtubule populations, hence modulating their functional properties.
Alternatively, a third protein may be involved in the indirect interaction of mitotic tubulin and the CTD of 4.1R. A good candidate protein is NuMA, because NuMA interacts with tubulin through amino acids 1868 -1967 (37) and with the CTD of 4.1R through amino acids 1788 -1810 (20). This suggests that the same population of NuMA may link 4.1R molecules to tubulin; thus, these proteins can form a three-way complex. Using a complementary assay, co-immunoprecipitation, we also demonstrated that 4.1R appears to interact in vivo with NuMA as well as with tubulin. However, binary interactions between NuMA and tubulin also occur; the depletion of 4.1R from the mitotic HeLa extracts did not dissociate the interaction of NuMA and tubulin, as shown through their association in randomly dispersed microtubules assembled in 4.1R-immunodepleted extracts (see Fig. 4B, 4.1R-depleted). Further experiments are required to determine whether 4.1R associates with NuMA and tubulin simultaneously or individually. Nevertheless, the association of 4.1R, NuMA, and tubulin suggests that 4.1R may play a role in organizing mitotic spindles.
Centrosome-free spindle-pole formation that depends on the action of non-centrosomal structural and microtubule motor proteins was directly demonstrated during spindle assembly in extracts from metaphase-arrested frog eggs (45). Dynein-dependent and centrosome-free spindle-pole formation was mimicked by the induction of microtubule asters in the presence of the drug taxol in both Xenopus (46) and mammalian systems (25). An in vitro aster assembly assay using HeLa mitotic extracts, an assay that recapitulates many of the structural events in mitosis, has been used widely to characterize the function of a number of proteins localized at the spindle poles (29 -33). We used this assay to show that mitotic aster assembly was inhibited in a 4.1R depletion-specific manner and that assembly could be re-initiated with purified recombinant 4.1R 135 . These results suggest that 4.1R is essential to the assembly of mitotic asters in vitro.
Our immunoprecipitation/co-precipitation assays showed that only a subpopulation of NuMA and ␣-tubulin interacts with 4.1R. Consistent with these results, the immunodepletion assays further confirmed that 4.1R antibodies failed to codeplete significant amounts of NuMA and ␣-tubulin. These results imply that aster assembly is not abolished because of the lack of NuMA or ␣-tubulin. This idea is further supported by the observation that NuMA still associates with the irregularly shaped microtubules in the 4.1R-depleted assembly reactions. Thus, depletion of 4.1R does not seem to affect the association of NuMA with tubulin. Rather, recruitment of 4.1R to the astral arrays by tubulin and/or NuMA could be functionally necessary for assembly or maintenance of astral arrays.
The biological significance of 4.1R in centrosome-dependent spindle-pole formation in somatic cells is unclear. 4.1R might FIG. 6. Co-immunoprecipitation of 4.1R, tubulin, and NuMA from HeLa mitotic extracts. HeLa mitotic extracts were prepared as described under "Experimental Procedures" and were subjected to immunoprecipitation using rabbit preimmune IgG, anti-HP, anti-␣-tubulin mAb, and anti-NuMA mAb. Immunoprecipitate analysis entailed immunoblotting using anti-HP Ab (A), anti-␣-tubulin mAb (B), anti-NuMA mAb (C), or anti-exon 13 Ab (D). The arrows denote the migration position of the 135-kDa 4.1R isoform (A and D), tubulin (B), and NuMA (C). Onethird of the immunoprecipitates (Bound) and one-eighth of the supernatant fractions (Unbound) were loaded onto the lanes as indicated. For control, 25 g of cell lysates was loaded. stabilize the interaction among NuMA, dynein, dynactin, and microtubules in a manner analogous to its role in the stabilization of the association of spectrin, actin, and integral proteins in red blood cells. Viable homozygous 4.1R knockout mice have been generated (47). The viability of these mice is perplexing, given the wide tissue distribution of 4.1R and its possible involvement in cell division. However, it is not unprecedented that germ line alterations that activate or inactivate genes of interest might have effects other than somatic mutations because of developmental compensation (48,49). For example, the acute loss of Rb in primary quiescent cells has phenotypic consequences that differ from germ line loss of Rb function (50). In cell division, the function of 4.1R in 4.1R null mice may be sustained by redundant systems and might involve other 4.1 family members. It is worth noting that 4.1B does not express in HeLa cells, and 4.1N expresses in minute amounts. 4.1G is highly expressed in HeLa. Therefore, 4.1G would be the most likely candidate for redundancy. We 2 note that a complicated transcription coupling splicing mechanism exists in 4.1 family members. Further identification of the 4.1G or N isoforms most likely to be involved in mitosis is necessary to assess the possible functional redundancy of these 4.1 family members. Regardless, the ability of 4.1R to interact with mitotic tubulin and the fact that mitotic aster assembly in vitro requires 4.1R suggest that 4.1R may be pivotal to the structural organization and maintenance of the mitotic apparatus during cell division. This idea is supported by an in vitro aster assembly assay using Xenopus egg extracts (51) that indicates that the SAB and CTD peptides significantly deranged the normal symmetrical microtubule array and dispersed the tight focus of NuMA. Furthermore, small interference RNA-caused knockout of 4.1R in HeLa cells also resulted in poorly focused spindle poles in mitotic cells. 2 Additional work will establish the precise role of cytoskeletal 4.1R in the mechanochemistry of the mitotic spindle. This report, however, suggests that 4.1R is crucial and is related to mitotic tubulin binding.