Hyper-active RAS/MAPK introduces cancer-specific mitotic vulnerabilities

Significance

The next generation of cancer therapeutics will ideally inhibit cellular activities that are uniquely required by cancer cells, often referred to as an “addiction” to a certain protein. Toward this goal, we show that, while the mitotic protein BubR1 is required for the spindle assembly checkpoint in all cell types, its role in stabilizing kinetochore–microtubule attachments is often dispensable. We find that, relative to nontransformed cells, tumor cells with amplified RAS/MAPK signaling exhibit an enhanced requirement for BubR1 to align chromosomes during mitosis and prevent lethal chromosome segregation errors. Altogether, this work expands our understanding of how chromosome segregation and cancer biology intersect and highlights the need to explore mitotic processes in diverse cellular states.

Abstract

Aneuploidy, the incorrect number of whole chromosomes, is a common feature of tumors that contributes to their initiation and evolution. Preventing aneuploidy requires properly functioning kinetochores, which are large protein complexes assembled on centromeric DNA that link mitotic chromosomes to dynamic spindle microtubules and facilitate chromosome segregation. The kinetochore leverages at least two mechanisms to prevent aneuploidy: error correction and the spindle assembly checkpoint (SAC). BubR1, a factor involved in both processes, was identified as a cancer dependency and therapeutic target in multiple tumor types; however, it remains unclear what specific oncogenic pressures drive this enhanced dependency on BubR1 and whether it arises from BubR1’s regulation of the SAC or error-correction pathways. Here, we use a genetically controlled transformation model and glioblastoma tumor isolates to show that constitutive signaling by RAS or MAPK is necessary for cancer-specific BubR1 vulnerability. The MAPK pathway enzymatically hyperstimulates a network of kinetochore kinases that compromises chromosome segregation, rendering cells more dependent on two BubR1 activities: counteracting excessive kinetochore–microtubule turnover for error correction and maintaining the SAC. This work expands our understanding of how chromosome segregation adapts to different cellular states and reveals an oncogenic trigger of a cancer-specific defect.

To prevent aneuploidy and ensure equal chromosome segregation, each mitotic chromosome must build a multimeric protein complex called the kinetochore on its centromeric chromatin. The kinetochore links chromosomal DNA to dynamic microtubule polymers and powers chromosome movements. This requires that each kinetochore pair on duplicated chromosomes become attached to microtubules anchored at opposite ends of the mitotic spindle, which results in a “bioriented” state. In early mitosis, kinetochore–microtubule attachments form stochastically such that errors in attachment are common (1). To prevent the accumulation of erroneous attachments, kinetochore–microtubule turnover is high in early mitosis (2–4). Kinetochores also prevent aneuploidy by orchestrating the spindle assembly checkpoint (SAC), which ensures that chromosome segregation initiates only after every kinetochore pair becomes bioriented (5, 6). Together, these two mechanisms work to prevent chromosome segregation errors in mitosis.

The mitotic protein BubR1/BUB1B was first identified as an essential regulator of the SAC (7–9). Subsequently, it was found that BubR1 deletion was embryonic lethal in mice (10), and its depletion caused aneuploidy and/or cell death in cultured cells due to premature mitotic exit prior to chromosome alignment (11). However, in DLD-1 and HeLa cells treated with MG132 to prevent premature mitotic exit, it was demonstrated that BubR1/BUB1B also functions as an error correction factor by counteracting excessive kinetochore–microtubule turnover and promoting chromosome alignment at the spindle equator, independent of the SAC (12–14). Previously, we reported that all cell types tested required BubR1 to maintain the checkpoint, but nontransformed cells efficiently stabilized kinetochore–microtubule attachments independent of BubR1 (15). Furthermore, only a subset of cancer cell lines required BubR1 to properly align chromosomes, including those transformed with an HRAS mutant (15). Recently, a pair of studies suggested that aneuploidy drives a cancer-specific requirement for the SAC, and BubR1 contributes specifically to this activity rather than to kinetochore–microtubule attachment stabilization and chromosome alignment (16, 17). While much is understood about the effects of oncogenic RAS signaling on the transcriptome and the G1/S cell cycle checkpoint, little is known about its effects on mitotic processes that ensure chromosome segregation fidelity (18). To better understand BubR1’s cancer-specific behaviors and to identify which BubR1 activity has the most potential as a drug target, we set out to resolve the molecular mechanisms that underlie the cancer-specific requirement for BubR1 in chromosome alignment.

Results

Validating a Genetic Transformation Model to Dissect How Oncogenic Pathways Affect Mitosis.

To understand how RAS signaling alters kinetochore function, we focused on signaling pathways downstream of RAS in which components localize to or affect mitotic structures (Fig. 1A) (19–25). We generated an isogenic panel of laboratory-transformed cells where RAS and its downstream pathways, MAPK and PI3K/mTOR, are constitutively activated. We chose nontransformed retinal pigment epithelial cells (ARPE19), since they are amenable to standard culturing and experimental techniques (15, 26, 27). Since hyperactivation of RAS and other oncogenes induces senescence in primary cells (28, 29), we employed a validated retroviral transduction system that bypassed p53 and Rb tumor-suppressor pathways by simultaneously transducing and overexpressing TERT, mTP53^DD, CCND1, and CDK4^R24C (28). From this polyclonal population (ARPE19^T53D4), we created three more cell lines by introducing individual dominant oncogenic mutants of AKT, MAP2K1/MEK1, and HRAS that had been previously validated (29, 30) (Fig. 1B; detailed descriptions of each transgene are in the SI Appendix). Each oncogene vector included a drug-resistance gene such that only cells transduced with the oncogene survived drug selection, and only cells expressing TERT, mTP53^DD, CCND1, and CDK4^R24C escaped oncogene-induced senescence. This resulted in three polyclonal cell lines from a shared genetic background, ARPE19^T53D4-AKT, ARPE19^T53D4-MEK, and ARPE19^T53D4-RAS cells, and limited rare genetic events arising from clonal expansion (Fig. 1B).

Fig. 1.

Laboratory-transformed cells recapitulate RAS-dependent behaviors observed in tumor-isolated cells. (A) Schematic of signaling pathways downstream of RAS and their connections to mitosis. Proteins outlined in blue or brown localize to the kinetochore or mitotic spindle, respectively. (B) Schematic of viral-mediated genomic modifications to transform retinal pigment epithelial cells with different oncogenes. (C) Ability of genetically modified cells to proliferate as colonies suspended in a matrix of soft agar. Three biological replicates are shown with the SD indicated. (D) RNA-seq data from three biological replicates analyzed through PCA reveals parental cell lines ARPE19 and ARPE19^T53D4 are essentially indistinguishable from each other, and are clearly partitioned from the three oncogenically stressed cell lines. (E) Transcripts overexpressed in ARPE19^T53D4-AKT, ARPE19^T53D4-MEK, and ARPE19^T53D4-RAS cells relative to parental ARPE19 cells were tested for their enrichment in published gene signatures of AKT, MEK, and RAS activation. (F) Distances between kinetochores on duplicated chromosomes (interkinetochore distances) were measured in unperturbed cells and those treated for 1 h with MEK or ERK inhibitors. Rest length refers to mean interkinetochore distances of prophase kinetochore pairs prior to microtubule attachment. Scale bars are 5 μm (Top) and 1 μm (Bottom). Below the images, the points on the graph represent the average interkinetochore distance per cell, and bars represent mean and 95% CIs. For each condition, at least 20 kinetochore pairs were measured from at least 28 cells from three independent experiments. (G) Chromosome alignment assay with and without short-term inhibition of MEK using drug U0126 (Left). Three biological replicates were measured, and the SD is indicated (Right). All P values in figure were calculated with Tukey’s multiple comparisons test.

We found that neither ARPE19 nor ARPE19^T53D4 cells exhibited transformation phenotypes, including increased population-doubling rate (31), anchorage-independent growth (32), and changes to cell shape (33, 34) (Fig. 1C and SI Appendix, Fig. S1 A and B). As expected, ARPE19^T53D4-AKT and ARPE19^T53D4-RAS cells behaved like fully transformed tumor-isolated cells in these assays, but surprisingly, ARPE19^T53D4-MEK cells failed to grow independent of anchorage, suggesting they are only partially transformed (Fig. 1C and SI Appendix, Fig. S1 A and B). The expression of each transgene was validated by PCR amplification from complementary DNA, and oncogene activation was assayed by immunoblotting for phosphorylation of MEK and AKT targets (SI Appendix, Fig. S1 C and D). To gain a holistic view of transformation events, we performed RNA sequencing (RNA-seq) on each of the asynchronous cell populations (35). TERT, p53^DD, and each oncogene were highly transcribed in the appropriate cell lines, as evidenced by their transcript abundance after normalization and the presence of unique nucleotide polymorphisms (SI Appendix, Fig. S2 and Datasets S1 and S2). In contrast, CCND1 and CDK4 transcript abundance values varied among cell lines. This suggests that TERT and mTP53^DD contribute to senescence escape in the presence of all three oncogenes, but each oncogene differentially requires CCND1 and CDK4^R24C to bypass the Rb pathway, and thus those two transgenes may experience negative selection and/or gene silencing (SI Appendix, Fig. S2 A–C). The similarities in transformation behavior observed between ARPE19 and ARPE19^T53D4 cells were consistent with principal-component analysis (PCA) of their transcriptomes, where replicate experiments from ARPE19 and ARPE19^T53D4 cells are interleaved on the plot (Fig. 1D). PCA also illustrated dramatic divergences among ARPE19^T53D4-AKT, ARPE19^T53D4-MEK, and ARPE19^T53D4-RAS cells from their parental cell lines, suggesting the largest driver of transcriptional behavior was oncogene expression, rather than inhibition of tumor suppressors (Fig. 1D).

Finally, to confirm that each oncogenic transgene activated the expected transcriptional program observed in tumor-isolated cells, we identified differentially expressed genes in ARPE19^T53D4-AKT, ARPE19^T53D4-MEK, and ARPE19^T53D4-RAS cells in comparison with the parental ARPE19 population. We then determined whether these gene sets were overrepresented with published gene signatures of AKT (36), MEK (37–39), or RAS (40) pathway activation derived using clinical samples, tumor-isolated cell lines, and chemical inhibitors (SI Appendix, Fig. S3 and Datasets S2 and S3). Transcriptional targets of the AKT pathway reported in a breast-cancer study (36) did not predominate the changes we observed, suggesting tissue-specific differences affected the outcomes (Fig. 1E and SI Appendix, Figs. S3 and S4). In contrast, the MEK and RAS pathway gene signatures (37–40) were markedly enriched in our sets of overexpressed genes for each transformed line (Fig. 1E and SI Appendix, Figs. S3 and S4). Together, our data provide compelling evidence that introduction of the transgenes described above results in predicted transcriptional and phenotypic changes.

Oncogenic MAPK Signaling Is Necessary and Sufficient to Induce BubR1 Sensitivity.

We found previously that RAS-transformed cells exhibited two related kinetochore defects (15). First, in RAS-transformed cells, the distance between kinetochores on bioriented chromosome pairs (interkinetochore distance) was ∼100 nm shorter than nontransformed cells, suggesting they formed weaker attachments. Second, RAS-transformed cells depleted of BubR1 failed to form stable kinetochore–microtubule attachments, even when the SAC was artificially maintained, resulting in lethal chromosome alignment defects (15). To further explore RAS-dependent changes to mitosis, we determined whether our laboratory-transformed cells behaved similarly.

First, we measured the distances between kinetochore pairs on bioriented chromosomes in normally cycling cell populations. As expected, interkinetochore distances in ARPE19^T53D4-RAS cells were ∼100 nm shorter than in ARPE19 cells (1,110 and 1,229 nm, respectively) (Fig. 1F). By generating MEK- and AKT-transformed cells, we were able to map this kinetochore defect specifically to the MAPK pathway because ARPE19^T53D4-MEK, but not ARPE19^T53D4-AKT, cells exhibited the ∼100-nm decrease in interkinetochore distances (1,116 nm and 1,214 nm, respectively; Fig. 1F). We note that ARPE19^T53D4-AKT cells exhibited no reduction in interkinetochore distance, despite expressing TERT, mTP53^DD, CCND1, and CDK4^R24C, indicating that these genes were not responsible for the mitotic defect observed in ARPE19^T53D4-MEK or ARPE19^T53D4-RAS cells. Further illustrating this point, we also observed the same trend in TP53-null mouse embryonic fibroblasts transformed with oncogenic AKT, MEK, and RAS transgenes (SI Appendix, Fig. S5A). To ascertain whether MAPK signaling was responsible for the kinetochore defect, we measured interkinetochore distances in the presence of MAPK inhibitors. We found that treating cells with a MEK inhibitor (U0126) or an ERK inhibitor (pyrazolylpyrrole 6p, pp6p) (41) for 1 h prior to fixation restored interkinetochore distances to the length in parental ARPE19 cells (Fig. 1F and SI Appendix, Fig. S5B). Importantly, neither inhibitor affected the interkinetochore distances in ARPE19 or ARPE19^T53D4-AKT cells (Fig. 1F), which would become longer if the kinetochore–microtubule attachments were incorrectly stabilized, as occurs when a known attachment regulator, Aurora B kinase, is inhibited (42, 43) (SI Appendix, Fig. S5C). This suggests MAPK activity is not a primary regulator of kinetochore–microtubule attachments but aberrantly weakens them upon oncogenic activation.

We next assayed chromosome alignment after BubR1 depletion in ARPE19^T53D4-AKT, ARPE19^T53D4-MEK, and ARPE19^T53D4-RAS cells in the presence of MG132, which prevents premature mitotic exit, even when the SAC is defective. After treatment with BubR1 small interfering RNA, <50% of ARPE19^T53D4-RAS cells aligned their chromosomes, while >90% of nontransformed cells exhibited full chromosome alignment, consistent with our previous findings (15) (Fig. 1G and SI Appendix, Fig. S5D). As expected from interkinetochore distance measurements, ARPE19^T53D4-MEK cells also failed to align chromosomes after BubR1 depletion, while ARPE19^T53D4-AKT exhibited no alignment defects, indicating that the enhanced requirement for BubR1 in chromosome congression was induced by MAPK signaling downstream of RAS and not inhibition of tumor-suppressor pathways (Fig. 1G). We next tested whether U0126 was able to rescue the chromosome alignment defect in ARPE19^T53D4-MEK and ARPE19^T53D4-RAS cells. Indeed, in the presence of the MEK inhibitor, chromosome alignment was achieved in >90% of both ARPE19^T53D4-MEK and ARPE19^T53D4-RAS cells depleted of BubR1, similar to ARPE19 and ARPE19^T53D4-AKT cells (Fig. 1G). These results suggest that hyperactivation of the MAPK pathway regulates mitosis enzymatically rather than transcriptionally because the brief U0126 treatment affects only the most-rapid gene-regulation programs, which is unlikely during the G2/M transition, when transcription and translation are significantly reduced (44). Altogether, we observe kinetochore defects and an enhanced requirement for BubR1 in transformed ARPE19^T53D4-RAS cells and partially transformed ARPE19^T53D4-MEK cells, but not in transformed ARPE19^T53D4-AKT cells, which were both rescued by short-term MEK inhibition. This argues that unregulated MAPK signaling in mitosis, but not transformation nor the transgenes inhibiting tumor-suppressor pathways, produces these mitotic phenotypes.

Dependency on BubR1 to Recruit PP2A-B56α Underlies Phenotypes in RAS Cells.

RAS signaling through the MAPK pathway drives the requirement for BubR1 in chromosome alignment, but how does this signaling alter kinetochore function? BubR1 stabilizes kinetochore–microtubule attachments using its kinetochore attachment regulatory domain (KARD) to recruit the phosphatase PP2A-B56α to kinetochores, which counteracts the destabilizing activity of Aurora B kinase (45, 46). Due to their fully transformed state, we used ARPE19^T53D4-RAS cells to test whether oncogenic RAS/MAPK activity altered kinetochore phosphoregulation and caused cells to become dependent on BubR1-mediated recruitment of PP2A-B56α. Three BubR1 residues (S670, S676, and T680) must be phosphorylated to bind PP2A-B56α, which can be prevented by mutating each phospho-target site to alanine (EGFP-BubR1^3A, Fig. 2A) (45). We therefore assayed the ability of transiently expressed EGFP-BubR1^3A to support chromosome alignment after depletion of endogenous BubR1.

Fig. 2.

RAS-transformed cells require BubR1-mediated recruitment of PP2A-B56α. (A) Diagram of BubR1 mutants/fusions used to manipulate PP2A-B56α recruitment to kinetochores. KEN, KEN boxes; TPR, Tetratricopeptide Repeat motif; BUB-BD, BUB Binding Domains (Bub3 and Bub1); KARD, kinetochore attachment regulatory domain. (B) Chromosome alignment assay in ARPE19^T53D4-RAS cells depleted of endogenous BubR1 and rescued with transient expression of empty vector (E.V.), BubR1^WT, BubR1^3A that cannot bind PP2A-B56α, or Mis12-KARD^3D that constitutively recruits PP2A-B56α to kinetochores independent of BubR1. ARPE19 data are replotted from Fig. 1 (green bar) as a reminder that BubR1 is dispensable for chromosome alignment in these cells. Three biological replicates are shown, and the SD is indicated. Numbers above each bar in the graph indicate P value when compared to E.V. transfection alone. (C) Interkinetochore distances measured in ARPE19^T53D4-RAS cells when PP2A-B56α is constitutively recruited through expression of Mis12-KARD^3D. Points represent the average distance per cell (at least 20 kinetochores were measured per cell) collected from three biological replicates (n = 34 cells for control cells; n = 35 cells for Mis12-KARD^3D-expressing cells), and bars represent mean and 95% CIs. (D) Cell proliferation of ARPE19 and ARPE19^T53D4-RAS cells depleted of endogenous BubR1 and rescued with stable expression of E.V., BubR1^WT, or BubR1^3A. Three biological replicates are shown with the mean and SD reported. All P values in figure were calculated with Tukey’s multiple comparisons test.

EGFP-BubR1^WT or EGFP-BubR1^3A plasmids were expressed specifically in ARPE19^T53D4-RAS cells depleted of BubR1, but not ARPE19 cells, because they exhibit no defects in chromosome alignment when BubR1 is fully depleted (Figs. 1G and 2B). We found that EGFP-BubR1^WT rescued chromosome alignment, but EGFP-BubR1^3A did not, suggesting RAS-transformed cells specifically require BubR1 for its ability to recruit PP2A-B56α to kinetochores (Fig. 2B and SI Appendix, Fig. S6). To further test this, we constitutively recruited PP2A-B56α to kinetochores by expressing a phosphomimetic BubR1 KARD fused to the carboxyl terminus of kinetochore protein Mis12 (Mis12-KARD^3D, Fig. 2A). This protein fusion rescued chromosome alignment defects in BubR1-depleted ARPE19^T53D4-RAS cells, despite lacking all other BubR1 activities (Fig. 2B and SI Appendix, Fig. S6). Moreover, ectopic kinetochore recruitment of PP2A-B56α using the Mis12-KARD^3D construct was sufficient to rescue the short interkinetochore distances observed in ARPE19^T53D4-RAS cells, which increased from ∼1,110 nm to ∼1,235 nm (Fig. 2C). Altogether, these results suggest that ARPE19^T53D4-RAS cells require BubR1-mediated recruitment of PP2A-B56α to prevent lethal chromosome segregation errors, while this activity is dispensable or redundant in nontransformed cells. Indeed, when EGFP-BubR1^3A was stably transduced into cells and endogenous BubR1 depleted, we observed decreased cell proliferation for ARPE19^T53D4-RAS compared with ARPE19 cells (Fig. 2D).

Oncogenic MAPK Activity Stimulates a Network of Kinetochore Kinases and Error Correction.

Our finding that ARPE19^T53D4-RAS cells are particularly reliant on BubR1’s role in phosphatase recruitment suggested that phosphoregulation was unbalanced in these cells. To test this, we measured the phosphorylation status of a key PP2A-B56α and Aurora B substrate, Hec1/NDC80, in ARPE19 and ARPE19^T53D4-RAS cells (Fig. 3 A and B) (42, 47). Immunostaining with pan-Hec1 antibodies and antibodies recognizing Hec1 phosphorylated at serine 55 (pSer55) showed that phosphorylation of kinetochores was modestly increased in ARPE19^T53D4-RAS cells compared with nontransformed ARPE19 cells (Fig. 3B). This was the case in early mitosis (∼20% difference) and continued into metaphase (∼15% difference) (Fig. 3B). Hec1 phosphorylation levels have been shown to correlate with changes in interkinetochore distances (43), and we reason that a 10–20% increase in phosphorylation is consistent with a 100-nm decrease in interkinetochore distances, since few phosphorylation events (0.5–2 per molecule) are sufficient to achieve wild-type behavior (43, 48). To test whether MAPK activity was responsible for the increase in phosphorylation, we treated ARPE19^T53D4-RAS cells for 1 h with U0126 and found it reduced Hec1 phosphorylation to similar levels as exhibited by ARPE19 cells in both prometaphase and metaphase (Fig. 3B). We also found that U0126 had no effect on Hec1 phosphorylation status in ARPE19 cells (SI Appendix, Fig. S7A). Together, this suggests that aberrant MAPK signaling, and not inhibiting p53 or other tumor-suppressor pathways, results in changes in kinetochore phosphorylation; thus, for further studies, we compared only ARPE19 and ARPE19^T53D4-RAS ± U0126 cells.

Fig. 3.

MAPK signaling in ARPE19^T53D4-RAS cells hyperactivates a network of kinetochore kinases. (A) Published connections between BubR1/PP2A-B56α and RAS/MAPK that could be implicated in the observed chromosome alignment and interkinetochore distance phenotypes. (B–F) Immunostaining of prometaphase ARPE19 and ARPE19^T53D4-RAS cells ± MEK inhibitor U0126 with corresponding antibodies (Left), whose kinetochore levels were quantified and normalized to those measured in ARPE19^T53D4-RAS cells (Right). Points represent the average kinetochore intensity per cell, and bars represent mean and 95% CIs. For all panels, at least 30 kinetochores were measured per cell from at least 26 cells, totaled from three biological replicates. All P values were calculated with Tukey’s multiple comparisons test. For panel B, levels of pHec1/Hec1 in prometaphase and metaphase were independently normalized.

We confirmed that ARPE19^T53D4-RAS cells exhibited defects in kinetochore phosphoregulation by documenting increased phosphorylation of another kinetochore component and target of Aurora B kinase, Dsn1 (pDsn1 Ser100) (Fig. 3C) (45). We hypothesized that Hec1 and Dsn1 hyperphosphorylation could arise from either altered kinase activity or altered phosphatase activity (Fig. 3A). Therefore, we assessed kinetochore association or markers of activity for components from each pathway. Via immunostaining for BubR1 and PP2A-B56α at kinetochores, we found no changes to the phosphatase arm of the regulatory pathway (SI Appendix, Fig. S7 B and C). This suggested that hyperphosphorylation was a result of aberrant Aurora B kinase activity (Fig. 3A). Indeed, immunostaining with an antibody targeting an Aurora B autophosphorylation site (pABK Thr232), whose phosphorylation is required for kinase activity, revealed a significant increase in Aurora B activity in ARPE19^T53D4-RAS cells that was MEK dependent (Fig. 3D). Unlike other kinetochore phosphoproteins, we could not normalize pABK Thr232 to total Aurora B levels, because these antibodies recognize different subcellular populations of Aurora B (SI Appendix, Fig. S7D). However, we did not observe differences in centromere-associated Aurora B kinase between ARPE19 and ARPE19^T53D4-RAS cells (SI Appendix, Fig. S7D). In all cases, MEK inhibition restored Aurora B activity to levels measured in ARPE19 cells but did not reduce them further. Together with the results that MEK and ERK inhibition did not result in increased interkinetochore distances (Fig. 1F), our data suggest that MAPK is not a primary regulator of kinetochore–microtubule attachment stability during mitosis in healthy cells.

We hypothesized instead that oncogenic RAS/MAPK signaling inappropriately activates Aurora B through a network of kinetochore kinases, including Mps1 (49–51), which requires MAPK activity for kinetochore localization in Xenopus egg extracts (Fig. 3A) (25). We found that the recruitment of Mps1 to kinetochores was partially dependent on MAPK activity in human cells, as it was elevated in ARPE19^T53D4-RAS cells and reduced slightly below parental ARPE19 levels when ARPE19^T53D4-RAS cells were treated with U0126 (Fig. 3E). Mps1 phosphorylates the kinetochore scaffolding protein KNL1 to recruit Bub1 (as well as Bub3 and BubR1) (52). Interestingly, we also observed a MEK-dependent increase in the phosphorylation of Bub1 substrate histone H2A at Thr120 in ARPE19^T53D4-RAS cells (Fig. 3F). These data are consistent with MAPK signaling aberrantly stimulating kinetochore kinases rather than MAPK members directly phosphorylating Hec1, which we further validated by combining Aurora B and MEK inhibitors to find no additive behavior to Hec1 phosphorylation status (SI Appendix, Fig. S7E). Consistent with increased Mps1 recruitment and activity being the most upstream kinetochore defect, we found that treating ARPE19^T53D4-RAS cells with the Mps1 inhibitor Reversine (and MG132 to prevent premature mitotic exit) reversed the interkinetochore distance defect and relieved the requirement for BubR1 to align chromosomes, while not affecting ARPE19 cells (Fig. 4 A–C). Thus, oncogenic MAPK signaling destabilizes kinetochore–microtubule attachments by hyperactivating kinetochore kinases and rendering the standard levels of phosphatase recruitment insufficient to counteract the increased activity.

Fig. 4.

Inhibition of Mps1 relieves RAS/MAPK-dependent mitotic phenotypes. (A) lnterkinetochore distances in ARPE19^T53D4-RAS cells treated with increasing concentrations of the Mps1 inhibitor, Reversine. MG132 was included to prevent mitotic exit. The average interkinetochore distance measured in ARPE19 cells shown in Fig. 1 is indicated on the graph (green dotted line). For each concentration of Reversine, at least 20 kinetochore pairs were measured from at least 24 cells. (B) Interkinetochore distance values for ARPE19 and ARPE19^T53D4-RAS cells treated with 750 nM Reversine (data from ARPE19^T53D4-RAS cells replotted from A; data from ARPE19 cells with no drug are replotted from Fig. 1F). For each condition, at least 20 kinetochore pairs were measured from at least 33 cells. (C) Chromosome alignment phenotypes following small interfering RNA depletion of BubR1 in the presence or absence of 750 nM Reversine (n = 3 independent experiments).

Oncogenic MAPK Activity Misregulates Chromosome Alignment and Segregation.

Having identified kinetochore hyperphosphorylation as a likely mechanism by which RAS transformation induces the enhanced requirement for BubR1, we asked whether this challenged chromosome alignment efficiency, even when BubR1 was present. Indeed, by tracking chromosome morphology in H2B-EGFP–expressing cells, we found that RAS-transformed cells spent a longer time in mitosis as compared with ARPE19 cells (median durations 24 and 21 min, respectively). As with most mitotic phenotypes in ARPE19^T53D4-RAS cells, this was alleviated by treating cells with U0126, which had no effect on mitotic duration of ARPE19 cells (Fig. 5A). Further analysis revealed that ARPE19^T53D4-RAS cells required excess time specifically to align chromosomes (15 or 12 min for ARPE19^T53D4-RAS or ARPE19 cells, respectively), but not to silence the SAC (Fig. 5B). This 20% delay in alignment is likely inconsequential to ARPE19^T53D4-RAS cell proliferation unless the SAC is compromised, as has been observed with Mps1 inhibitors (53). Consistent with previous studies (18), we found that ∼4% of ARPE19^T53D4-RAS cells exhibited chromosome segregation errors (lagging chromosomes and/or formation of micronuclei), which was a threefold increase over ARPE19 cells (Fig. 5C). Chromosome segregation fidelity was restored through MEK inhibition with U0126, suggesting kinetochore hyperphosphorylation was responsible for the errors and that aberrant MAPK signaling likely contributes to aneuploidy and/or chromosome instability in these cells. Thus, RAS transformation and kinetochore hyperphosphorylation challenge but do not severely compromise chromosome segregation, which is consistent with the robust proliferation of ARPE19^T53D4-RAS cells. Instead, these minor defects in chromosome alignment and segregation indicate a stressed but functional kinetochore state that may be a therapeutic target.

Fig. 5.

ARPE19^T53D4-RAS cells exhibit chromosome alignment delays and chromosome segregation errors in a MAPK-dependent manner. (A) Mitotic duration (nuclear envelope breakdown [NEB] to anaphase onset) was quantified in asynchronous ARPE19 and ARPE19^T53D4-RAS cells expressing histone H2B-EGFP. The delay observed in ARPE19^T53D4-RAS cells (Top) was reduced when cells were treated with MEK inhibitor U0126 (Bottom). Points represent the percent of cells in mitosis binned by image acquisition intervals (3 min) from three biological replicates. Data are fitted with a sigmoidal curve, and dotted lines represent 95% CI for the fitted line. (B) Mitotic duration in A was separated into time to alignment (Left, NEB to chromosome alignment) and SAC silencing (Right, chromosome alignment to anaphase onset). Timing for each cell is represented as a point on the graph, and P values were calculated using Dunn’s multiple comparisons test. Cell n values are as follows for chromosome alignment timing: ARPE19, 175; ARPE19 + U0126, 166; ARPE19^T53D4-RAS, 515; ARPE19^T53D4-RAS + U0126, 251; and for SAC silencing timing: ARPE19, 170; ARPE19 + U0126, 165; ARPE19^T53D4-RAS, 515; and ARPE19^T53D4-RAS + U0126, 520. (C) Number of cell divisions with chromosome segregation errors in ARPE19 and ARPE19^T53D4-RAS cells when untreated or in the presence of MEK inhibitor U1026.

MAPK Activation Drives Enhanced BubR1 Requirement in Glioblastoma Isolates.

Our findings that hyperactive RAS/MAPK signaling contributes to chromosome missegregation and aneuploidy, as well as the enhanced requirement for BubR1 in chromosome alignment, have implications for cancer biology, particularly in glioblastoma tumors, where we first documented the BubR1 phenotype (15). Approximately 80% of glioblastoma tumors are aneuploid (54) and 67% require BubR1 for chromosome alignment (15, 55), yet RAS/RAF/MEK mutations occur in only ∼5% of central nervous system tumors (18). Thus, we wanted to understand whether nonmutagenic hyperactivation of RAS/MAPK signaling may be a common source of aneuploidy or the BubR1 vulnerability. We first searched for signatures of RAS/MAPK transcriptional activation in a list of genes overexpressed in aneuploid relative to diploid tumors derived from many different tissue types (17). We found a 1.9- to 2.4-fold enrichment for RAS/MEK/AKT-associated transcripts in aneuploid tumors (Fig. 6A), which could occur either because RAS/MAPK signaling instigates aneuploidy, as suggested from observations of chromosome missegregation in ARPE19^T53D4-RAS cells (Fig. 5C), or because excess chromosomes in aneuploid tumor cells encode RAS/MAPK genes, increasing the gene dosage of this pathway. In either situation, aneuploid tumors are more likely than diploid tumors to activate RAS and/or its downstream pathways, even absent activating mutations.

Fig. 6.

Oncogenic RAS/MAPK signaling is associated with BubR1 sensitivity in patient-derived glioblastoma stem-like cells. (A) Genes overexpressed in aneuploid compared with diploid tumor cells (17) were interrogated for enrichment of previously published signatures of RAS, MEK (37–39), and AKT activation. (B) Sets of genes expressed at elevated levels in BubR1-resistant (827 and 1502) and BubR1-sensitive (131, G155, and G179) glioblastoma patient isolates compared with nontransformed neural stem cells (CB660) were generated using published RNA-seq data (58) and represent each column. Fold enrichment and P values for the enrichment of previously published gene signatures of RAS, MEK, and AKT signaling within glioblastoma isolates are displayed in each row. The number of genes significantly elevated in patient isolates is listed in parentheses at the bottom of the plot. (C and D) Average metaphase interkinetochore distance ± MEK inhibition in (C) neural stem cells (CB660) and RAS-transformed neural stem cells (CB660^T53D4/RAS) and in (D) glioblastoma patient isolates with or without the enhanced BubR1 requirement. (C and D) Points represent the average distance per cell collected from three biological replicates, and bars represent mean and 95% CIs. P values were calculated with Tukey’s multiple comparisons test. For each condition in C and D, at least 20 kinetochore pairs were measured from at least 22 cells. (E) Copy number gains through aneuploidy and/or activating mutations of RAS/MAPK components alter the balance of phosphatase and kinase activity at kinetochores. This weakens kinetochore–microtubule attachments and renders the phosphatase recruitment activities of BubR1 a possible target for therapeutics.

To determine how nonmutagenic RAS/MAPK signaling contributes to the enhanced requirement for BubR1, we searched for signatures of RAS, MEK, or AKT activation in a set of glioblastoma patient isolates (1502, 827, 131, G166, and G179) for which the enhanced BubR1 requirement was known (15, 55). As in Fig. 1E, we determined whether RNA-seq profiles of glioblastoma isolates (56) were enriched for RAS-, MEK-, or AKT-induced gene signatures relative to a neural stem cell line, CB660. Patient isolates 131, G166, and G179 require BubR1 for chromosome alignment (15, 55) and showed a strong enrichment for signatures of RAS and MEK signaling, yet none harbored a known activating mutation in RAS or MAPK genes (Fig. 6B). Conversely, we found mixed results among cells that lack the enhanced BubR1 requirement. The patient-isolated 827 cell line showed no enrichment for signs of RAS activation, while the 1502 cell line did show enrichment (Fig. 6B). With this small sample of glioblastoma cells, the transcriptional signals of RAS/MAPK activation tended to correlate with an enhanced requirement for BubR1, but this analysis was limited by how many isolates had both datasets available. Thus, we directly tested whether MAPK activity was required for mitotic phenotypes in glioblastoma cells by inhibiting MEK with U0126 and measuring interkinetochore distances. The results from nontransformed neural stem cells (CB660) and those transformed with RAS in the same manner as Fig. 1 (CB660^T53D4-RAS cells) (26, 28) recapitulated our observations with ARPE19 cells, indicating this phenomenon was not unique to a single tissue type (Fig. 6C). Consistent with our expectations, of the five glioblastoma isolates tested, only the three with the enhanced requirement for BubR1 (131, G166, and G179) exhibited short interkinetochore distances (Fig. 6D). Strikingly, despite the complex genetic backgrounds and lack of clear oncogenic mutations in the pathway, inhibiting MEK for 1 h rescued the interkinetochore distances in these three isolates (Fig. 6D). This is strong evidence that our genetic transformation model accurately identified a causal oncogenic pressure driving the observed kinetochore dysfunction. Thus, we conclude that, independent of specific genetic mutations, ploidy, or other complex epigenetic changes, MAPK activity during G2/M destabilizes kinetochore–microtubule attachments, giving rise to an increased dependence on BubR1’s phosphatase recruitment activities.

Discussion

Here, we used a genetically controlled transformation model to recapitulate the biology observed in complex tumor isolates and found that the MAPK pathway directly influences mitotic behavior, which causes a cancer-specific vulnerability. Transforming cells with oncogenic AKT revealed that neither inactivating tumor-suppressor genes nor transformation itself was sufficient for the enhanced BubR1 requirement. More to this point, ARPE19^T53D4-MEK cells exhibited only partial transformation yet required BubR1 for chromosome alignment, suggesting that increased RAS/MAPK signaling caused kinetochore defects. This was further demonstrated via short-term chemical inhibition of MEK, which rescued all mitotic phenotypes. Altogether, numerous experiments made it clear that inhibiting the TP53 and Rb tumor-suppressor pathways did not contribute to the mitotic phenotypes we studied; thus, for this work, direct comparisons between ARPE19 cells and RAS-transformed cells were appropriate. However, it is important to note that, in future studies probing different biological questions, the more appropriate control may be ARPE19^T53D4 cells.

We propose here a molecular mechanism whereby hyperactive RAS inappropriately stimulates kinetochore kinases, including Aurora B. This results in hyperphosphorylation of kinetochores and decreased microtubule binding capacity, which compromises the chromosome alignment process (Fig. 6E). In these cells, BubR1 becomes essential to stabilize kinetochore–microtubule attachments, and its depletion results in a 20% increase in the time it takes to align chromosomes, which in turn makes cells more dependent on SAC signaling. These compounding vulnerabilities are likely why BubR1/BUB1B is a common hit in cancer-lethal RNAi screens (15–17, 57, 58).

We applied the lessons learned from our genetic transformation model to characterize cells derived from glioblastoma patient tumors. Both transcriptional gene signatures and the effect of MEK inhibition on interkinetochore distances suggest that inappropriate activation of the MAPK cascade may be a common oncogenic event, even in tumors lacking mutations or genomic amplifications. Moreover, we found that hyperactivation of RAS is strongly associated with aneuploidy. In ARPE19 cells, we observed RAS/MAPK-dependent chromosome segregation errors, and in aneuploid tumor cells, there was an enrichment for gene signatures of RAS activation. More work is required to dissect this connection, but it is likely that aneuploidy can be both a cause and consequence of RAS hyperactivation.

Altogether, our work emphasizes the adaptable nature of chromosome segregation mechanisms and the importance of studying these pathways in multiple cellular contexts. It also highlights the need to further explore BubR1’s kinetochore localization domain (Gle2-binding-sequence domain) (15) or KARD as potential therapeutic targets, as inhibiting these regions that regulate kinetochore–microtubule attachments may have preferable therapeutic outcomes compared with BubR1 depletion and subsequent loss of SAC activity.

Materials and Methods

Detailed descriptions of the following methodologies and reagents are included in the SI Appendix, Supplementary Methods and Tables S1–S3: cell culture, transformation behavior, transient and viral transductions, plasmids, proliferation assay, immunoblotting, immunofluorescence, image acquisition and analysis, RNA extraction, RT-PCR, RNA-seq library prep and sequencing, and RNA-seq data analysis.

Data, Materials, and Software Availability

[RNA seq] data have been deposited in [NCBI Gene Expression Omnibus (GEO) database] [GSE172512 (59)].