Rae1 interaction with NuMA is required for bipolar spindle formation

Abstract

In eukaryotic cells, the faithful segregation of daughter chromosomes during cell division depends on formation of a microtubule (MT)-based bipolar spindle apparatus. The Nuclear Mitotic Apparatus protein (NuMA) is recruited from interphase nuclei to spindle MTs during mitosis. The carboxy terminal domain of NuMA binds MTs, allowing a NuMA dimer to function as a “divalent” crosslinker that bundles MTs. The messenger RNA export factor, Rae1, also binds to MTs. Lowering Rae1 or increasing NuMA levels in cells results in spindle abnormalities. We have identified a mitotic-specific interaction between Rae1 and NuMA and have explored the relationship between Rae1 and NuMA in spindle formation. We have mapped a specific binding site for Rae1 on NuMA that would convert a NuMA dimer to a “tetravalent” crosslinker of MTs. In mitosis, reducing Rae1 or increasing NuMA concentration would be expected to alter the valency of NuMA toward MTs; the “density” of NuMA-MT crosslinks in these conditions would be diminished, even though a threshold number of crosslinks sufficient to stabilize aberrant multipolar spindles may form. Consistent with this interpretation, we found that coupling NuMA overexpression to Rae1 overexpression or coupling Rae1 depletion to NuMA depletion prevented the formation of aberrant spindles. Likewise, we found that overexpression of the specific Rae1-binding domain of NuMA in HeLa cells led to aberrant spindle formation. These data point to the Rae1–NuMA interaction as a critical element for normal spindle formation in mitosis.

In eukaryotic cells upon entry into mitosis, interphase microtubules (MTs) are reorganized into the spindle apparatus, a complex and dynamic macromolecular machine composed of polymerized tubulin and many interacting proteins (1). MTs are polymers of α-β-tubulin dimers with distinct plus and minus ends. The typical metaphase spindle apparatus contains two poles at centrosomes with γ-tubulin at the minus ends of MTs. Bipolar spindle assembly requires organization of MTs and their selective local stabilization. Chromatin and kinetochores stabilize the plus ends of MTs and become aligned in the center of the spindle awaiting successful biorientation of all sister chromatids before anaphase. In some settings, notably plant cells and oocytes, spindles form in the absence of centrosomes by MT nucleation on chromatin followed by bundling at the minus ends (2–4).

The Nuclear Mitotic Apparatus protein (NuMA) is a 237-kDa protein with an ≈1,500-aa discontinuous coiled-coil spacer between N- and C-terminal globular domains (5, 6) that can form parallel coiled-coil dimers ≈200 nm in length (6). The C-terminal domain of NuMA contains a site for multimerization (6), a nuclear localization sequence that interacts with karyopherin α (7), a MT-binding site that overlaps with a binding site for LGN (8) (a leucine-glycine-asparagine-repeat containing protein) and a site for binding the polyADP-ribose polymerase, tankyrase (9). The N-terminal domain of NuMA is believed to contain a binding site for dynactin that acts as an adaptor for dynein, a minus end-directed motor known to target NuMA to the spindle pole (10). The WD (tryptophan-aspartic acid) repeat β propeller protein Rae1 (11), also known as gle2 (12) or mrnp41 (13), is one of ≈30 different proteins (14) (nucleoporins or nups) found in the nuclear pore complex. Rae1 has been shown to bind to the nucleoporin Nup98 (15) and the mitotic checkpoint kinase Bub1 (16) through their so-called GLEBS (Gle2-binding site) domains (17) and to function with Nup98 in securin degradation (18). The vesicular stomatitis virus M protein blocks host cell mRNA export by binding to Rae1 (19). Although Rae1 has been reported to bind to MTs (20, 21), these binding sites have not been mapped. Interestingly, several nucleoporins uniquely localize to the spindle (22), including the Nup107–160 complex recently shown to be required for spindle assembly (23), but the mechanistic aspects and functional relevance of these mitotic redistributions are largely unknown.

Aberrant expression of either Rae1 or NuMA has been linked to formation of multipolar spindles. In the case of NuMA, multipolarity is linked to overexpression, whereas in the case of Rae1, multipolarity is linked to its depletion (21, 24). Here we identify a mitotic interaction between Rae1 and NuMA, map this interaction to a specific domain of NuMA, and demonstrate that a balance of these two proteins is required for bipolar spindle formation. We propose a model wherein Rae1 modulates the MT crosslinking valency of NuMA in mitotic spindles to prevent chromosome segregation defects that are commonly found in cancer cells.

Results

Rae1 and NuMA Form a Transient Complex During Mitosis.

To elucidate in greater depth the specific role of mitotic Rae1, we analyzed the composition of purified Rae1 complexes in mitotic HeLa cells. The major Rae1-associated proteins from mitotic IP were subjected to MALDI mass spectrometry after trypsin digestion, leading to the identification of NuMA (data not shown). By immunoblotting of anti-Rae1 immunoprecipitates, we detected coprecipitating NuMA along with Nup98 and dynein (Fig. 1A). Conversely, using anti-NuMA antibodies, we immunoprecipitated Rae1 and dynein but not Nup98 (Fig. 1B). These data suggested that Rae1 and NuMA interact. To further define the specificity for this mitotic Rae1–NuMA interaction, we prepared extracts from HeLa cells synchronized using a double thymidine block followed by release into and out of the MT destabilizer nocodazole. HeLa cells were released from an S phase double thymidine block into nocodazole for 12 h and arrested in mitosis. At this time, mitotic cells were collected by shake-off and released out of nocodazole for 4 h. These experiments revealed a transient association of Rae1 and NuMA during mitosis (Fig. 1C). Consistent with the IP data, we found that Rae1 and NuMA colocalized transiently on HeLa cell spindle poles from prophase to anaphase (Fig. 1D and SI Fig. 6).

Fig. 1.

Rae1 and NuMA form a transient complex during mitosis. (A and B) IP from mitotic HeLa extracts with α-Rae1 and α-NuMA or control antibodies (IgG), followed by immunoblotting with α-NuMA, α-Nup98, α-dynein, and α-Rae1. In lanes marked “2% input,” 5 μl of 250 μl extract used per IP was analyzed directly. (C) Synchronized HeLa cells were collected at the indicated time points, and extracts were analyzed by immunoblotting directly (Input 2%) or after IP with α-Rae1. Anti-phospho-Histone H3 and α-tubulin were used as mitotic index and loading controls. (D) Asynchronous HeLa cells costained with α-Rae1 (green) and α-NuMA (red); chromatin was visualized using DAPI (blue). The large yellow arrow points to metaphase cell, small white arrowpoints to interphase, and the large white arrowpoints to late telophase. (Scale bar, 25 μm.)

Simultaneous Depletion of Rae1 and NuMA Rescues Bipolarity.

Because NuMA or Rae1 are known to individually impact spindle formation, we altered their balance in vivo by modulating their concentrations using RNAi and overexpression strategies and assayed the effect on spindle polarity. Consistent with previous observations (21), reduction of Rae1 by RNAi (SI Fig. 7) led to the formation of multipolar spindles (Fig. 2A). We quantified the mitotic defects at 72 h after transfection with siRNA duplexes targeting Rae1 and found a high proportion (>30%) of cells displayed strikingly altered spindle morphology compared with controls treated with buffer alone (transfection efficiency >90%; see Fig. 2A and Table 1). The extra spindle poles appeared to pull chromosomes away from the main spindle, contributing to serious chromosome-alignment defects. The Rae1 siRNA-treated cells displayed NuMA localization to spindle poles (data not shown) and stained positive for the centrosomal markers pericentrin and γ-tubulin (Fig. 2 B and C). Given our observation of a mitotic Rae1–NuMA interaction, we were interested in exploring the effect of NuMA down-regulation on the multipolar spindle phenotype of Rae1-depleted cells. NuMA overexpression is also linked to multipolar spindle formation that may be rescued by reduction of NuMA levels (24). Indeed, when NuMA and Rae1 levels were reduced simultaneously by siRNA, the incidence of multipolar spindles was greatly reduced (Fig. 2A and Table 1).

Fig. 2.

Simultaneous depletion of Rae1 and NuMA rescues bipolarity. (A) HeLa cells were transfected with either siRNA duplexes against Rae1 (Left) or Rae1 and NuMA together (Right). After 72 h, cells were stained with α-tubulin antibody (red) and analyzed by confocal laser microscopy. Chromatin was stained with DAPI (blue). [Scale bars, 25 μm (Upper); 5 μm (Lower).] (B and C) Representative figures of cells treated with Rae1 siRNA, fixed, and stained with anti-pericentrin and either γ-tubulin (B) or α-tubulin (C) antibodies. DNA is counterstained with DAPI.

Table 1.

RNAi and protein overexpression spindle phenotypes

	Mitotic cells	Percent bipolar	Percent monopolar	Percent multipolar
Control	300	98.5 ± 3	0	1.5 ± 1
Rae1 siRNA	300	67 ± 7	0	33 ± 3
(Rae1 + NuMA) siRNA	300	91 ± 3	0	9 ± 6
Control	200	98 ± 2	0	2 ± 1
GFP-NuMA	200	59 ± 5	12 ± 3	29 ± 4
GFP-NuMA + HA-Rae1	200	83 ± 2	4 ± 2	13 ± 4
GFP	200	98 ± 3	0	2 ± 2
GFP-NuMA325–829	200	65 ± 2	13 ± 3	22 ± 2

Quantitation of spindle defects represented in Figs. 2 and 3. n = three independent experiments.

Simultaneous Overexpression of NuMA and Rae1 Rescues Bipolarity.

To further test our hypothesis that mitotic Rae1 can bind to NuMA and influence spindle formation, we explored the effect of overexpressing Rae1 in cells overexpressing NuMA and displaying the multipolar spindle phenotype. First we transfected GFP-NuMA into HeLa cells and 72 h later observed the cells using confocal microscopy. A variety of phenotypes was observed: in 29% of the cells (n = 300), additional poles were observed and in 12% of the cells monopolar spindles formed (Fig. 3 and SI Fig. 8). Interestingly, if we cotransfected NuMA with Rae1, we found that additional poles were observed in only 13% of the cells (n = 300); 4% of the cells remained monopolar; and most of the cells appeared normal during prometaphase/metaphase (83%; Table 1, Fig. 3 and SI Fig. 8). These data therefore further suggest that normal spindle pole formation requires balanced concentrations of NuMA and Rae1 during mitosis.

Fig. 3.

Simultaneous overexpression of Rae1 and NuMA rescues bipolarity. Representative figures of HeLa cells transfected with plasmids overexpressing either GFP-NuMA or GFP-NuMA and Rae1-HA together. After 24 h, cells were fixed, stained with α-tubulin antibody (red in overlay; GFP is green), and analyzed by confocal laser microscopy. Chromatin was stained with DAPI (blue). (Scale bar, 5 μm.)

Mapping the Rae1 Interaction Domain on NuMA.

To investigate the basis of the Rae1 NuMA interaction at mitosis, we generated a series of NuMA deletion mutants and tested them for their ability to interact with Rae1. We coexpressed Rae1 and various fragments of NuMA (see Fig. 4A) in a cell-free reticulocyte translation system. Only the fragment NuMA_325–829 interacted directly with Rae1 (Fig. 4B). This fragment of NuMA contains a potential GLEBS-like motif (SI Fig. 9). The Nup43 or Seh1 WD-repeat β propellers did not interact with NuMA_325–829 (data not shown). We therefore conclude that a fragment of NuMA spanning amino acids 325–829, at the N-terminal end of the coiled-coil domain, interacts specifically with Rae1.

Fig. 4.

Mapping the Rae1 interaction domain on NuMA. (A) Schematic of NuMA and five FLAG-tagged fragments of NuMA. Numbers on the left refer to amino acids (aa); all fragments are continuous (e.g., NuMA2 ends at amino acid 829 and NuMA3 starts at amino acid 829). (B) Autoradiograph of [³⁵S]methionine-labeled Rae1 and NuMA-FLAG fragments coexpressed in vitro, affinity-purified, and separated by SDS/PAGE. Rae1 is untagged. Asterisks indicate the five FLAG-tag NuMA fragments expressed in varying amounts using this system. Numbers indicate molecular weight markers in kilodaltons. (C) Immunoblotting of α-GFP IPs from either GFP- or GFP-NuMA_325–829-expressing HeLa cells. IPs are blotted with α-Rae1 or α-NuMA [using BD Biosciences clone 22 monoclonal that recognizes an epitope (amino acids 658–691) within NuMA_325–829]. (D) HeLa cells overexpressing GFP-NuMA_325–829 (green) costained with tubulin (red) and DAPI (blue).

To test whether NuMA_325–829 interacts with Rae1 in vivo, we expressed NuMA_325–829-GFP in HeLa cells. We then used anti-GFP antibodies for IPs from extracts of cells expressing either NuMA_325–829-GFP or GFP alone and tested for the presence of Rae1. As seen in Fig. 4C, Rae1 was specifically coimmunoprecipitated with NuMA_325–829-GFP but not GFP alone (asterisks). We could also detect a small amount of full-length NuMA (arrow) in the IPs, suggesting that full-length NuMA associates with the NuMA_325–829 fragment (arrowhead). This observation implies that at least some dimerization determinants reside between amino acids 325 and 829 at the beginning of the predicted coiled-coil region of NuMA.

Because the NuMA_325–829 fragment seems to bind Rae1 in cells, we examined the effect of expressing this domain in HeLa cells. NuMA_325–829-GFP localized in part to spindles (upper row of Fig. 4D) and colocalized with tubulin (lower row of Fig. 4D). We found that additional poles were observed in 30% of these cells (n = 200; Fig. 4D and Table 1). Although we did not observe any gross mislocalization of Rae1 in these cells (not shown), a possible interpretation of these results is that NuMA_325–829 binds to and sequesters some Rae1, making it unavailable to interact productively with full length NuMA. This would then be analogous to the multipolar spindle phenotype observed after reduction of Rae1 by RNAi. The NuMA_325–829 fragment could also dimerize with full-length NuMA, and the resulting hybrid NuMA-NuMA_325–829 heterodimers would lack one C-terminal domain and would potentially have reduced ability to link MTs.

Discussion

Spindle assembly requires the temporal and spatial coordination of multiple overlapping pathways involving MT nucleation and stabilization, pole formation and attachment and alignment of chromosomes (3, 25). MTs are dynamically unstable structures that are stabilized by a variety of MT-associated proteins. In addition, “crosslinking” among MTs is required to bundle them at their minus end at the spindle pole. The C-terminal domain of NuMA has been shown to bundle MTs. If NuMA is a dimer (ref. 6 and Fig. 5), it could act as a divalent crosslinker of MTs. In addition, any NuMA-associated protein that also binds MTs could function in the MT crosslinking. Rae1 has previously been shown to bind to MTs (21). Here we show that Rae1 also interacts with NuMA, and we have mapped this interaction to the N-terminal end of the coiled-coil domain. To our knowledge, this is a previously undescribed biochemical mapping of a specific interaction between a nucleoporin and any component of the mitotic spindle. We suggest that by interacting with NuMA and MTs, Rae1 could increase the MT crosslinking valency of NuMA (Fig. 5) and further stabilize MTs at their minus ends. We propose this interaction is critically required for normal bipolar spindle formation. We observed the presence of centrosomal markers in the aberrant spindle poles of Rae1-depleted cells, and the same has been reported in the case of NuMA overexpression in cancer cell lines (24); therefore, we cannot preclude the possibility that Rae1 and NuMA are involved in centrosome duplication or stabilization.

Fig. 5.

A “valency” model of MTs interacting with NuMA and Rae1. NuMA is assumed to be a dimer (6) with the C-terminal (C) indicated to directly interact with MTs (7). A region (residues 325–829) at the N-terminal end of the coiled coil of NuMA interacts with Rae1 (data in this paper) and therefore converts NuMA from a divalent to a tetravalent MT “crosslinker.”

Our data suggest that a balance between NuMA and Rae1 is critical for bipolar spindle formation. Specifically, in the case of overexpression of NuMA coupled with concomitant overexpression of Rae1 and, conversely, in the case of depletion of Rae1 coupled with concomitant depletion of NuMA, formation of supernumerary spindle poles is suppressed (Figs. 3 and 4D). We speculate that the additional crosslinking valency of NuMA, by virtue of its interaction with MT-bound Rae1, increases the “density” of crosslinks and therefore enhances the bundling of MTs at their minus ends. In the case of NuMA overexpression or Rae1 depletion, many of the crosslinks at the minus end of MTs would be divalent rather than tetravalent (Fig. 5). Nevertheless, if a critical number of these divalent crosslinks has been established, the resulting supernumerary poles may be sufficiently stable to persist, because none of the spindle poles have accumulated enough of the tetravalent crosslinks to compete with each other for stability. In this scenario, the minus ends of MTs are capped with γ-tubulin and surrounded by pericentrin. In contrast, if NuMA overexpression is accompanied by Rae1 overexpression, the formation of a high density of tetravalent crosslinks may kinetically favor the formation of a bipolar spindle destabilizing any supernumerary poles that lack a critical density of tetravalent crosslinks. A spindle with a higher density of tetravalent crosslinks could successfully compete for NuMA and Rae1 against spindles with low-density crosslinks that are less stable. Our results with the overexpression of a Rae1-binding fragment of NuMA (NuMA_325–829) would be entirely consistent with our speculations regarding the valency and density of NuMA-mediated minus-end MT bundling.

The kinetics of spindle formation are influenced by many other components, including MT-based motors (26), MT dynamics (27), gradients of Ran and kinases (28), and polyADP-ribosylation (29), that will likely modify the effects we reported here for Rae1 and NuMA. In any case, our results should provide a useful framework for further testing the dynamics of MT bundling in mitosis and elucidating the role of a Rae1–NuMA imbalance in chromosome segregation defects leading to aneuploidy.

Materials and Methods

Plasmids.

The plasmid-encoding full length human Rae1 (Image ID LIFESEQ95168410; Open Biosystems, Huntsville, AL) was subcloned into pcDNA3 with HA tag and pET28a. The NuMA domains were subcloned by PCR from pCDNA3-GFP-NuMA into pET28a with a C-terminal simian virus 40 T antigen nuclear localization sequence and a FLAG tag. All constructs were confirmed by DNA sequencing.

Cell Culture, Transfections, and Synchronization.

HeLa cells were transfected with Rae1 and NuMA siRNAs using Oligofetamine and with GFP-NuMA and HA-Rae1 plasmids using Lipofectamine 2000 following the manufacturer's protocol (Invitrogen, Carlsbad, CA). Cells were synchronized in S phase by double thymidine block (30) using 2 mM thymidine with the following modifications. In experiments involving siRNA oligos [Fig. 2 and supporting information (SI) Fig. 7], the cells were transfected 24 h before the initiation of the first thymidine block and collected or imaged after 72 h. In experiments involving plasmid-mediated protein overexpression (Figs. 2 and 3 and SI Fig. 8), the transfection was initiated before the second thymidine block for 4 h. For experiments represented in Fig. 1C, cells were released into 30 ng/ml Nocodazole after the second thymidine block, harvested for analysis at hourly intervals for 12 h during Nocodazole incubation, and then collected by mitotic shakeoff, replated in fresh medium, and harvested for analysis at hourly intervals for 4 h. For the RNAi experiments, siRNA duplexes targeting Rae1 [5′-GCAGUAACCAAGCGAUACA-3′] (21) or NuMA [5′-GGCGUGGCAGGAGAAGUUC-3′] (31) were purchased from Integrated DNA Technologies (Coralville, IA). Mock transfection was with buffer alone (control). Transfection efficiency was monitored with Block-iT (Invitrogen).

Antibodies and Immunofluorescence.

In initial experiments, anti-Rae1 antibodies from K. Weis (University of California, Berkeley, CA) (21) and J. van Deursen (Mayo Clinic, Rochester, MN) (32) were used (Fig. 1 A and B). For all subsequent experiments, peptides based on human Rae1 residues 313–327 (FYNPQKKNYIFLRNAAEE) (21), with N-terminal acetylation and C-terminal amidation, were injected in rabbits (Cocalico Biologicals, Reamstown, PA). Antibodies were affinity-purified before use. Anti-NuMA polyclonal antibody used in Fig. 1 A and B was from D. Compton (Dartmouth Medical School, Hanover, NH) (33); anti-NuMA monoclonal antibody (clone 22) from BD Biosciences (Franklin Lakes, NJ) was used in all other experiments. DM1A monoclonal α-tubulin antibody and γ-tubulin antibody were from Sigma–Aldrich (St. Louis, MO). The dynein monoclonal antibody (clone 74.1) was from Chemicon (Temecula, CA), Phospho-Histone H3 antibody was from Upstate Biotechnology (Lake Placid, NY). α-HA, α-GFP, and α-pericentrin antibodies were from Abcam (Cambridge, U.K.). Secondary antibodies were from Molecular Probes (Eugene, OR).

For immunofluorescence, synchronized HeLa cells were washed in PBS and fixed for 10 min in methanol at −20°C. Cells were then permeabilized with 0.2% Triton X-100 in PBS for 10 min at room temperature. Samples were examined on a Zeiss (Oberkochen, Germany) LSM510 MEGA confocal microscope, and all images were acquired by using a plan-Apochromat 100 × 1.4-N.A. objective.

Immunoprecipitations (IPs).

For IPs, ≈10⁷ cells were seeded and synchronized as described above. Mitotic HeLa cells were collected, washed with PBS, spun at 400 × g for 10 min, and lysed in 1 ml of cold Lysis buffer (50 mM Tris·HCl, pH 7.2/250 mM NaCl/0.1% Nonidet P-40/2 mM EDTA/10% glycerol) containing 1× protease inhibitor mixture (Roche, Indianapolis, IN) and 1 mM PMSF. Lysates were centrifuged for 30 min at 4°C at 14,000 × g. The resulting lysate supernatants were precleared with 50 μl of Protein A/G bead slurry (Santa Cruz Biotechnology, Santa Cruz, CA), mixed with 5–10 μl of various antibodies as specified, and incubated for 1 h at 4°C with rocking. The beads were then washed five times with 500 μl of Lysis buffer. After the last wash, 50 μl of 1× SDS/PAGE blue loading buffer (New England Biolabs, Ipswich, MA) was added to the bead pellet before loading.

In Vitro Binding Assays.

Proteins were expressed by using the Promega (Madison, WI) TNT coupled transcription/translation system according to the manufacturer's protocol. Five microliters of Flag beads (ANTI-FLAG M1 Agarose Affinity Gel from Sigma–Aldrich) were washed three times with binding buffer (20 mM Hepes, pH 7.5/100 mM KCl/5 mM MgCl₂/0.1% Tween 20/20% glycerol/0.01% BSA/1 mM DTT/1 mM PMSF/1× complete protease inhibitor mixture), preblocked for 10 min with 10 μl of nonspecific rabbit serum, washed with binding buffer, and resuspended in 60 μl of binding buffer. Then, 10 μl of in vitro transcribed and translated [³⁵S]methionine labeled Rae1 and NuMA-Flag mutants were added to the beads, and the mixture was incubated at 4°C for 1 h. Beads were washed six times with binding buffer and boiled in 15 μl of SDS/PAGE sample buffer. Samples were analyzed by SDS/PAGE (4–20% Tris-glycine gels; Invitrogen), followed by autoradiography.

Abbreviations

MT

microtubule

IP

immunoprecipitation.