Spatial regulation of Aurora A activity during mitotic spindle assembly requires RHAMM to correctly localize TPX2

RHAMM localized to centrosome and non-centrosomal spindle microtubule assembly sites

To query a putative role for RHAMM during spindle assembly, we used immunofluorescence to localize RHAMM in prophase cells and found that it co-localized with a centrosome marker (gamma-tubulin, TUBG1) and decorated mitotic microtubules and spindle poles in cells at later stages of mitosis (Fig. 1A) (Fig. S1A).⁶ In prophase HeLa cells, we localized RHAMM to discrete foci within the chromosome volumes (Fig. 1A). Ectopically expressed GFP-RHAMM also localized to punctuate, non-centrosomal sites, which co-localized near to a kinetochore marker (Bub1-related kinase, BubR1) (Fig. 1A). This localization was more pronounced in the hematopoietic cell line, RPMI-8226 (Fig. 1B, arrows). We asked whether these foci may be sites for microtubule assembly using a regrowth assay in which microtubules in mitotic cells were depolymerized with nocodazole and allowed to regrow during a recovery phase. As shown at 2 m for recovery, microtubules (β-tubulin, TUBB) were first assembled at centrosomes, demarked by TUBG1 (Fig. 1C). By 6 min for recovery, mitotic microtubules were assembled at centrosomes as well as non-centrosome sites, which co-localized with BubR1 (Fig. 1C). In these microtubule-regrowth experiments in mitotic cells, RHAMM localized to both centrosome and non-centrosome sites of microtubule assembly (Fig. 1C).

RHAMM is required for normal mitotic spindle assembly kinetics and architecture

To query whether RHAMM contributes to microtubule assembly or is simply a passenger microtubule-associated protein at these sites, cells were separately treated with 3 siRNA duplexes that target the untranslated regions (UTRs) of RHAMM mRNA, which enabled us to silence RHAMM and rescue function through transient expression of the wild-type transgene tagged with GFP at the N-terminus (GFP-RHAMM). With the outlined protocol (Fig. S1B), endogenous RHAMM mRNA levels were reduced by 40–70% (Fig. S1C), which translated to a 70–80% reduction in the levels of RHAMM protein (Fig. 1D). Rescue with GFP-RHAMM resulted in transgene expression at levels that approximated endogenous RHAMM levels (top band is GFP-RHAMM in Fig. 1E), and we confirmed that GFP-RHAMM localized in a manner that is similar to that of the endogenous protein (Fig. S1D). Consistent with the previously established role for RHAMM as a structural mitotic spindle protein, the silencing of RHAMM resulted in a significant increase in both the proportion of cells with mitotic spindle defects, such as multipolar spindles and disorganized spindles (Fig. 1F) and multinucleated cells (Fig. 1G), which may result through the failure of cell division. Importantly, both of these depletion phenotypes could be rescued through the transient expression of GFP-RHAMM (Fig. 1F and G). Moreover, in RHAMM-silenced HeLa cells, we found a significant increase in the proportion of mitotic cells, which suggested that the silencing of RHAMM induced a mitotic arrest and/or delayed mitotic kinetics (Fig. 1H).

To help distinguish between mitotic arrest and delayed kinetics, living, non-synchronized HeLa cells were followed by time-lapse microscopy as they transited through mitosis (Fig. 2A and B; Videos S1–4). We quantified the time necessary for spindle assembly and mitosis completion by tracking the progression of mitotic cells (Fig. 2C). Spindle assembly duration was defined as the time from the onset of prophase (chromosome condensation) to the onset of metaphase (chromosome alignment), and mitosis duration was measured from the onset of prophase to the end of telophase (disappearance of the midzone spindle). In cells pretreated with scrambled control siRNA, the average duration for spindle assembly was 26 ± 9 min, and these cells completed mitosis in 106 ± 17 min (Fig. 2C; Video S1), which are similar kinetics to those of untreated cells (27 ± 13 min and 118 ± 24 min, respectively). Silencing of RHAMM, however, resulted in a significant delay in both the time necessary to assemble a bipolar spindle and the time necessary to complete mitosis (51 ± 17 min and 185 ± 32 min, respectively) (Fig. 2C; Videos S2 and 3); indeed, 43% of mitotic cells pretreated with siRNA duplexes targeting RHAMM were unable to complete mitosis during the imaging time, and these cells were excluded from the quantitation of kinetics (Fig. 2C). Moreover, we found that phenotypically normal mitotic spindles would rotate freely in cells pretreated with siRNA targeting RHAMM (Video S3), consistent with a necessary role for RHAMM in the establishment of mitotic spindle orientation.⁹ Mitotic defects that accompanied the silencing of RHAMM were attenuated when GFP-RHAMM was transiently expressed (Fig. 2B and C; Video S4), which indicates that defects in spindle assembly, both kinetics and structure, can be assigned to the specific loss of RHAMM.

RHAMM is required for TPX2 localization and stability

To dissect the contributions of RHAMM to mitotic spindle assembly, the kinetics of mitotic microtubule regrowth after nocodazole depolymerization were followed in prophase cells that were previously treated with a scrambled control siRNA duplex or with duplexes targeting RHAMM. In comparison to control-treated cells, the silencing of RHAMM delayed the assembly of microtubule asters and the formation of a bipolar spindle following nocodazole washout (Fig. 3A; Videos S5 and 6), suggesting that RHAMM may play a role in seeding or establishing microtubule assembly sites. The attenuation of assembly was quantified in fixed cells at 6 min following recovery, when both centrosome and non-centrosomal assembly sites were present in the control-treated cells. Silencing of RHAMM resulted in a significant reduction in the number of microtubule foci assembled at non-centrosomal sites, and rescue with GFP-RHAMM was sufficient to recover non-centrosome microtubule assembly in RHAMM-silenced mitotic cells (Fig. 3B).

In Xenopus meiotic cell extracts, immunodepletion of XRHAMM prevented TPX2 concentration to DMSO-induced microtubule spindle poles,¹⁰ so we examined whether the silencing of RHAMM in mitotic HeLa cells prevented the localization of TPX2 to sites of microtubule assembly. In RHAMM-silenced mitotic cells, we found significantly fewer TPX2-positive microtubule regrowth sites (Fig. 3C). Pre-treatment of mitotic cells with nocodazole may dramatically alter microtubule assembly kinetics due, in part, to residual drug. So, we examined the immunolocalization of TPX2 in non-synchronized mitotic HeLa cells and found that TPX2 was lost from spindle poles (i.e., regions of interest identified by TUBG1) in RHAMM-silenced mitotic cells (Fig. 3D).

In both of the described assays, the abundance of TPX2 appeared reduced by immunofluorescence in RHAMM-silenced cells. To test the possibility that the specific loss of RHAMM altered the stability of TPX2, the abundance of these proteins were determined in lysates from G₂/M synchronized cell populations, as each gene product is strongly cell cycle regulated.^6,7,26 Synchronized populations of RHAMM-silenced cells showed reduced abundance of RHAMM and TPX2, but not the GTP binding protein Ran (Fig. 3E). Pre-treatment of cells with the proteasome inhibitor MG132 was sufficient to recover the level of TPX2 in RHAMM-silenced cells (Fig. 3E), which implies that the loss of RHAMM augments proteasome-dependent turnover of TPX2.

We next measured the levels of Aurora A and active, autophosphorylated Aurora A (pAurora A) in these lysates. Like TPX2, the abundance of Aurora A and the levels of pAurora A were reduced in RHAMM-silenced G₂/M synchronized cell lysates (Fig. 3F). Proteasome inhibition was sufficient to recover the total level of Aurora A (Fig. 3F). However, proteasome inhibition did not completely recover pAurora A, which remained attenuated in the presence of the correct levels of total Aurora A and TPX2 (Fig. 3F). These data support a regulatory role for RHAMM in the activation of Aurora A. The silencing of RHAMM reduced the abundance of TPX2, but even when the abundance of TPX2 was restored through proteasome inhibition, the level of active pAurora A was sub-optimal due to the mislocalization of TPX2.

The C-terminal bZIP of RHAMM binds near the nuclear localization signal of TPX2 to correctly locate TPX2 to sites of microtubule assembly and control spindle length

To better define the RHAMM-TPX2 protein complex and map the domains necessary for this interaction, truncation mutants were constructed for the C-terminus of RHAMM (Fig. 4A). These variants for GFP-RHAMM were transiently expressed in cells depleted of endogenous RHAMM, and the ability of these variants to co-precipitate TPX2 was tested through immunoprecipitation with antibodies to GFP, or IgG control antibodies. TPX2 was efficiently co-precipitated by RHAMM variants that contained a minimal necessary binding domain of amino acids 623–679 (i.e., full-length, 1–679, and 623–724), which contains the bZIP motif (Fig. 4B). Indeed, mutation to arginines of the 3 leucines in the motif was sufficient to abolish TPX2 from the precipitates (Fig. 4B).

Reciprocal studies were performed with transient expression of truncation variants for human TPX2. Since there are essential protein-binding domains in both the N- and C terminus of TPX2, we designed various truncation mutants from wild-type transgene tagged with either mCherry at the N terminus (mCherry-TPX2) or GFP at the C terminus (TPX2-GFP) to map the domains necessary for RHAMM interaction. mCherry–TPX2 and TPX2–GFP truncation mutants were transiently expressed in HeLa cells (Fig. 4C) and immunoprecipitated with antibodies to GFP, antibodies to mCherry, or IgG control antibodies. mCherry-TPX2 (FL) co-precipitated RHAMM and Aurora A, which was included as a positive control (Fig. 4D). The Klp2-targeting domain (aa 319–783²⁷) was not sufficient to co-precipitate RHAMM (Fig. 4D). The N terminus of TPX2 (aa 1–319), however, was sufficient, and the deletion of the Aurora A binding domain (aa 1–40^13,27) did not alter the complex formation (Fig. 4D). Together, these data localize the RHAMM-binding domain in TPX2 to amino acids 40–319, which contain an importin-α binding motif.²⁸

Compressed mitotic spindles result directly from the loss of TPX2 at the pole, which attenuates kinetochore-mediated microtubule assembly and spindle length.^29,30 So, we measured TPX2 localization at spindle poles (Fig. 5A) and mitotic spindle length (Fig. 5B) to map the minimal necessary domain in RHAMM for TPX2 targeting and retention. Our observations, and findings by others⁹ demonstrate that mitotic spindle orientation is disturbed in RHAMM-silenced cells. Spindle rotation may result in normal length spindles appearing compressed when viewed along the z-axis. To avoid this potential confounding factor, we measured pole-to-pole distance along a plane that is perpendicular to the aligned chromosomes in a 3D image, using TUBG1 as a marker protein for spindle poles.

In RHAMM-silenced cells rescued with GFP-RHAMM (full-length) or GFP-RHAMM(1–679) variants, TPX2 localized to the poles and decorated phenotypically normal mitotic spindles (Fig. 5A) of typical lengths (Fig. 5B). However, the expression of GFP-RHAMM(1–623), which lacks the bZIP motif, or GFP-RHAMM(623–724) was not sufficient to correctly localize TPX2, and, consistently, mitotic spindle lengths were significantly compressed when compared with those in control-treated cells or cells expressing full-length or GFP-RHAMM(1–679) variants (Fig. 5B). Thus, the bZIP motif in RHAMM is required, but is not sufficient, for the correct targeting and retention of TPX2 at spindle poles.

To define the necessary domains in TPX2 that may target RHAMM, reciprocal targeting experiments were performed in HeLa cells treated with shRNA targeting TPX2 (Fig. S2A). We found that RHAMM remain localized to unfocused spindle poles in TPX2-silenced mitotic cells (Fig. 5C). However, we observed compressed spindles in TPX2-silenced cells, in a fashion similar to RHAMM-silenced mitotic cells. These compressed phenotypes were rescued only by the expression of full-length mCherry-TPX2 and not by truncation variants (Fig. 5D and E). Thus, TPX2 is not required for the correct targeting and retention of RHAMM at spindle poles.

RHAMM is required for the temporally and spatially regulated activity of Aurora A

A regulatory relationship between RHAMM and Aurora A activity during mitotic spindle assembly is implied from our prior results (Fig. 3F) but not yet demonstrated. Therefore, we designed a series of experiments to examine the effects of RHAMM depletion on the activity of Aurora A. In asynchronous mitotic cells, we measured by immunofluorescence the levels of pAurora A at mitotic spindle poles, as demarked by TUBG1 (Fig. 6A); as a control for these studies, mitotic cells were pretreated with an Aurora A-specific small-molecule inhibitor, MLN8237, or vehicle DMSO control (Fig. S2C). Depletion of RHAMM, or pre-treatment with MLN8237, significantly reduced the levels of pAurora A at spindle poles (Figs. 6A; Fig. S2C). For kinetic and spatial analysis of Aurora A activity at non-centrosome sites of assembly, we used mitotic microtubule regrowth experiments. First, we measured pAurora A abundance by immunofluorescence at sites for microtubule regrowth following 6 min for recovery. In scrambled siRNA-treated (control) cells, or vehicle-treated cells, non-centrosome sites for assembly were strongly positive for pAurora A (Figs. 6B; Fig. S2D). Depletion of RHAMM, or pre-treatment with MLN8237, reduced pAurora A intensity at these sites (Figs. 6B; Fig. S2D).

Non-centrosome sites of assembly following recovery from nocodazole were attenuated with RHAMM depletion, which hindered the measurement of pAurora A at these sites. Thus, we utilized a FRET-based probe for Polo-like kinase 1 (PLK1) activity that is targeted to kinetochores³¹ to indirectly measure the spatial and temporal activity of Aurora A. PLK1 is a substrate for Aurora A and depletion of Aurora A by siRNA was sufficient to attenuate PLK1 activity.^32-35 Following nocodazole washout, PLK1 activity is high in prometaphase cells and resolves with lower phosphorylation at metaphase alignment; the current iteration of the probe³¹ tracks these kinase dynamics with an increasing FRET ratio (reflecting a decreasing level of phosphorylated substrate) as the cell progresses towards metaphase (Fig. 6C). We confirmed that the FRET emission ratio of the sensor was increased in the presence of the PLK1-specific inhibitor, BI2536, within prometaphase cells (Fig. 6D). In metaphase cells, however, PLK1 inhibition did not further augment the emission ratio, indicating, as expected, a low level of kinase activity within cells at this stage of mitosis (Fig. 6D). Incubation with the Aurora A-specific inhibitor, MLN8237, was sufficient to augment the emission ratio in prometaphase cells as was the pre-treatment of cells with siRNA to RHAMM, but not scrambled siRNA controls (Fig. 6D). Together, these data suggest that the loss of RHAMM attenuates Aurora A activity at kinetochores and sites of microtubule assembly in prometaphase cells.

Cell culture and cell cycle synchronization

HeLa cells were purchased from ATCC and cultured in Dulbecco modified Eagle medium (DMEM, Thermo-Fisher) supplemented with 10% fetal bovine serum (Invitrogen). HeLa cells that stably express eGFP-α-Tubulin and mCherry-Histone-H2B were kindly provided (Gruneberg Lab, University of Liverpool)⁴⁴ and maintained in media with 0.3 μg/ml puromycin (Invitrogen) and 0.5 μg/ml blasticidin S (Invitrogen). RPMI 8226 cells were purchased from ATCC and cultured in ATCC-formulated RPMI-1640 Medium (ATCC) supplemented with 10% fetal bovine serum (Invitrogen). All cell lines were grown at 37 °C in a 5% (v/v) CO₂ incubator. For synchronization, cells were subjected to a double thymidine block with 2 mM thymidine (Sigma) as described.⁴¹

Reagents and antibodies

Proteasome inhibitor MG132 (Sigma), MLN8237 (Selleck Chemicals), and BI2536 (Selleck Chemicals) were used as indicated. The following antibodies were used for immunoblotting and immunofluorescence: anti-Aurora A/AIK (Cell Signaling), anti-Phospho-Aurora A (Cell Signaling), anti-actin (Sigma), anti-BubR1 (Cell Signaling), anti-eGFP (Abcam), anti-GAPDH (Fitzgerald), anti-mCherry (Abcam), anti-RHAMM (Epitomics), anti-TPX2 (Novus Biologicals), anti-TUBG1 (Sigma Aldrich), anti-Ran (Cell Signaling), and anti-β-tubulin (Cell Signaling). Secondary antibodies were conjugated to IRDye (Rockland), HRP (Sigma), or to Alexa Fluor 488, 594, or 679 (Invitrogen). DAPI was purchased from Sigma Aldrich.

Molecular cloning and site directed mutagenesis

Truncation variants of RHAMM and TPX2 were constructed using the Gateway cloning system by Invitrogen. DNA fragments of RHAMM(1–623), RHAMM(1–679), and RHAMM(623–725) were amplified from eGFPC1-RHAMM.⁶ TPX2(1–319) was amplified from mCherry-TPX2 (Origene), while TPX2(40–783) and TPX2(319–783) were amplified from TPX2-GFP (Origene). Products were cloned into pDONR223 and transferred into pLenti6/V5-DEST or pDEST57 expression vectors (Invitrogen). Site-directed mutagenesis was performed using the QuikChange II kit (Agilent). Primer sequences used for PCR were listed in Table S1.

Transfection (siRNA and plasmids)

On-target plus siRNA (Dharmacon) sequences are listed in Table S1. Scrambled siRNA was used as a negative control. Transfections of siRNA and plasmid used Lipofectamine 2000^™ (Invitrogen) following the manufacturer’s protocols. Message and protein expression levels were measured 48–72 h post-transfection.

Lentivirus-mediated shRNA knockdown and generation of sub-lines

Lentivirus packaging, envelope, and control non-hairpin (NHP) pLKO.1 plasmids (Addgene) were used with shRNA against TPX2 (Sigma) as described⁴¹ (Table S1). Transfected cells were selected with 0.5 µg/ml puromycin (GIBCO) for 1 wk and maintained with 0.3 µg/ml puromycin.

RT-PCR and qPCR

Total RNA was extracted from cells using RNeasy mini kit (Qiagen), quantified with NanoDrop (Thermo-Fisher), and 1 μg of total RNA was converted to cDNA using AccessQuick (VWR). qPCR reactions were run in triplicate in an 7000 series machine (Applied Biosystems). Analysis of qPCR results was done using the delta delta Ct method. Expression of RHAMM mRNA was normalized to the level of TATA box binding protein (TBP). Primers and conditions are listed in Table S1.

Western blot and immunoprecipitation

Cells were lysed at 0.5−1.0 × 10⁷ cells/ml in RIPA buffer (western blot) or 0.5% NP-40 immunoprecipitation buffer (50 mM TRIS-HCl, pH 7.4, 150 mM NaCl, 1 mM EDTA, 0.5% NP-40) supplemented with protease inhibitors (Roche). Cell lysates were clarified by centrifugation at 16 000 × g for 10 min at 4 °C, and protein concentration was determined using the BCA protein assay kit (Pierce). For immunoprecipitation, lysates were precleared with protein A/G PLUS-Agarose beads (Santa Cruz) for 2 h at 4 °C with rotation. Protein complexes were isolated by incubation with 2.5 μg antibodies for 6 h at 4 °C with rotation, and then with protein A/G PLUS-Agarose beads for 4 h at 4 °C with rotation. Isolated complexes were washed 3 times with lysis buffer. Lysates were mixed with SDS sample buffer, separated by SDS-PAGE, and blotted with indicated primary antibodies and detected by HRP- or IRDye-conjugated antibodies and enhanced chemiluminescence (GE Healthcare) or Odyssey IR imaging (LI-COR) system, respectively.

Microtubule regrowth assay

Microtubules were depolymerized by treating cells with 5 μM nocodazole (Sigma) for 1 h at 37 °C. Cells were then washed 3 times with PBS and incubated in warm fresh media at 37 °C to allow microtubule regrowth. Cells were fixed at specified time points or living cells were imaged.

Immunofluorescence and image acquisition

Cells grown on coverslips were fixed with cold methanol. Cells were permeabilized with PBS-0.25% Triton X-100. Antibodies were diluted in PBS with 0.1% Triton X-100 and 3% BSA. Cells were washed 3 times in PBS-0.5% Tween-20 before and after incubations. Coverslips were mounted with ProLong Gold Antifade Reagent containing DAPI (Invitrogen).

Fixed cells were imaged using a 60× 1.2NA oil objective with the Fluoview software (Olympus) on the Olympus Fluoview FV10i (Olympus). Image stacks of 9 optical sections with a spacing of 1.0 µm through the cell volume were taken. Maximum intensity projection of the fluorescent channels was performed in ImageJ v1.46j (National Institute of Health).

For live-cell imaging, the glass coverslip was mounted in a silicone gasket (Chamlide CMM) and placed in a 37 °C and 5% CO₂ environmental chamber (Precision Control). Images were taken using a 40× 0.75 NA dry objective with the MetaMorph 7.5 software (Molecular Devices Inc) on the Olympus IX81 epifluorecent microscope (Olympus). For mitosis analysis, images were taken every 15 min using 350-ms exposures 2 × 2 binned resolution, with 2% of full lamp intensity per channel, and 7 optical sections spaced 1.0 µm apart. For microtubule regrowth analysis, images were taken every 2 min using 350-ms exposures 2 × 2 binned resolution, with 2% of full lamp intensity, and 7 optical sections spaced 1.0 µm apart. Maximum intensity projection of the fluorescent channels was performed in MetaMorph (Molecular Devices Inc).

Live imaging of the FRET (fluorescence resonance energy transfer) biosensor (kindly provided by the Lampson lab, University of Pennsylvania)³¹ was performed using the Olympus IX81 epifluorescent microscope equipped with a 60× 1.35NA oil objective, a cooled charge coupled device camera (CoolSNAP HQ2; Photometrics) controlled by the MetaMorph 7.5 software (Molecular Devices Inc). TFP and YFP emissions were acquired simultaneously with a 505DCXR beam splitter (Dual-View; Optical Insights, LLC) with the following optical filters: 438/24 for TFP excitation; 480/30 for TFP and 535/40 for YFP emission, respectively. Images were taken every 2 min using 300-ms exposures at 2 × 2 binned resolution for 30 min after nocodazole washout, with 45% of lamp intensity (X-Cite exacte, Lumen Dynamics) and 5 optical sections spaced 1.0 µm apart. Post-acquisition processing of images was performed using ImageJ v1.46j (National Institute of Health), as follows. The TFP and YFP emission images were background corrected by subtracting the value obtained from the cytoplasm, which does not include the kinetochore regions. The YFP/TFP FRET emission ratio value was then calculated as an average for the region of interest that was demarcated by the whole cell. Pseudocolored FRET images were generated by viewing the ratiometric data using the Ratio LUT (Look Up Table) plugin in ImageJ.

Statistics

Data were expressed as mean ± standard deviation for all figures except Figure 3C and D, which are mean ± standard error of mean. Statistical analysis was performed by unpaired 2-tailed Student t test. The results were considered significant at P < 0.05.