Evolutionarily conserved protein ERH controls CENP-E mRNA splicing and is required for the survival of KRAS mutant cancer cells

Abstract

Cancers with Ras mutations represent a major therapeutic problem. Recent RNAi screens have uncovered multiple nononcogene addiction pathways that are necessary for the survival of Ras mutant cells. Here, we identify the evolutionarily conserved gene enhancer of rudimentary homolog (ERH), in which depletion causes greater toxicity in cancer cells with mutations in the small GTPase KRAS compared with KRAS WT cells. ERH interacts with the spliceosome protein SNRPD3 and is required for the mRNA splicing of the mitotic motor protein CENP-E. Loss of ERH leads to loss of CENP-E and consequently, chromosome congression defects. Gene expression profiling indicates that ERH is required for the expression of multiple cell cycle genes, and the gene expression signature resulting from ERH down-regulation inversely correlates with KRAS signatures. Clinically, tumor ERH expression is inversely associated with survival of colorectal cancer patients whose tumors harbor KRAS mutations. Together, these findings identify a role of ERH in mRNA splicing and mitosis, and they provide evidence that KRAS mutant cancer cells are dependent on ERH for their survival.

The Ras family of small GTPases is mutated in a significant fraction of human cancers, with high frequencies of mutations in v-Ki-ras2 Kirsten rat sarcoma viral oncogene homolog (KRAS) found in colon, lung, and pancreatic cancers (1–4). Ras proteins are activated by growth factor receptors, and they, in turn, activate a number of downstream effector pathways to coordinate cell proliferation, survival, and migration. Somatic mutations in Ras frequently lead to its constitutive activation, which in turn, drives malignant growth. Cancer cells harboring mutations in KRAS often exhibit the classic behavior of oncogene addiction: they become dependent on the KRAS oncogene for growth and survival and therefore, are hypersensitive to the loss of KRAS protein (5). Efforts to pharmacologically inactivate mutant KRAS have been unsuccessful thus far. To identify additional genetic dependencies in Ras mutant cells, we previously conducted a genome-wide shRNA synthetic lethal screen in isogenic KRAS mutant and WT cells (6). In this screen, we identified a surprisingly diverse set of genes whose depletion causes greater toxicity in KRAS mutant cells compared with KRAS WT cells. Surprisingly, many of these genes do not directly partake in the Ras signaling network, but rather, they act to maintain cell viability by alleviating the stress phenotypes in cancer cells. We, therefore, proposed the concept of nononcogene addiction to explain the heightened dependency of Ras mutant cells on stress relief pathways for survival (7).

In the aforementioned screen, we identified genetic interactions between mutant KRAS and a network of mitotic genes, including the mitotic kinase polo-like kinase 1 (PLK1) and the E3 ligase anaphase-promoting complex (APC/C) that coordinately maintain the fidelity of chromosome segregation (6). Symmetrical distribution of chromosomes during mitosis is critical for genomic stability and cell survival (8, 9). During metaphase, chromosomes congression from spindle poles to the metaphase midplate is driven by the plus end-directed kinesin centromere protein E (CENP-E) (10). Unattached kinetochores activate spindle assembly checkpoint proteins such as budding uninhibited by benzimidazoles 1 homolog (Bub1), MAD3/BUB1-related protein kinase (BubR1), and mitotic arrest deficient 2-like protein 1 (MAD2), which in turn, inhibit the activity of APC/C to delay anaphase onset until all sister chromatids are bioriented and properly attached to opposite spindle poles (11). Many mitotic proteins are degraded by APC/C on mitotic exit. CENP-E is one such protein, and it is degraded on mitosis exit and resynthesized in the next S-phase (12). Thus, the proper expression and turnover of CENP-E during each cell cycle is necessary for chromosome congression and genomic stability (13, 14).

In this report, we identify a candidate Ras synthetic lethal gene, enhancer of rudimentary homolog (ERH). ERH is a highly conserved gene originally identified in Drosophila (15), and it has been implicated to play a role in nuclear gene expression (16–18). Here, we show that ERH interacts with the Sm protein SNRPD3, and it plays a critical role in the mRNA splicing and therefore, expression of CENP-E. KRAS mutant colorectal cancer (CRC) cells are more sensitive to the depletion of ERH protein. Consistent with this finding, low ERH expression is associated with better survival in cancer patients whose tumors harbor KRAS mutations. Our findings suggest that targeted inactivation of splicing machinery could be exploited to therapeutically restrict the malignancy of Ras-driven cancer.

Results

Ras Mutant Cells Are Hypersensitive to ERH Depletion.

We identified ERH as a candidate KRAS synthetic lethal gene from a genome-wide RNAi screen (6). ERH is a protein with 104 aa, and its molecular function is poorly understood. To validate the genetic interaction between ERH and the KRAS oncogene, we first tested an shRNA targeting ERH that scored in the screen using DLD-1 and HCT116 isogenic cells that are either WT or mutant for KRAS. These isogenic cells were derived by targeted deletion of the mutant KRAS^G13D allele (6, 19, 20). For simplicity, the parental cell line with KRAS^G13D/wt genotype is hereafter abbreviated as KRAS mutant line, and the KRAS^−/wt derivative cells are abbreviated as the KRAS WT line. We found that, in each isogenic pair, the KRAS mutant cells showed less viability on ERH knockdown compared with their respective KRAS WT counterpart (Fig. 1A). Western blot and quantitative RT-PCR (RT-qPCR) revealed a partial knockdown of ERH by this shRNA (Fig. 1B and Fig. S1A). Because this shRNA is the only effective shRNA against ERH in our library, we identified two additional ERH siRNAs: one targeting its coding sequence (siERH-3) and one targeting its 3′-UTR (siERH-5). Transfection of these siRNAs in these cell lines, as shown in Fig. 1C and Fig. S1B, also decreased the viability of the KRAS mutant cells more than the viability of the KRAS WT cells. Furthermore, on ERH depletion, we detected higher caspase activity in KRAS mutant cells compared with KRAS WT cells (Fig. S1C). Western blotting and RT-qPCR confirmed that ERH protein was efficiently depleted by both siRNAs (Fig. 1D and Fig. S1D). To ensure that cell viability change on ERH knockdown is an on-target effect, we constructed a C-terminal hemagglutinin (HA)-FLAG–tagged ERH cDNA and stably expressed it in DLD-1 and HCT116 cells. Because this cDNA construct lacks a 3′-UTR, it is resistant to knockdown by siERH-5, but it remains sensitive to siERH-3 (Fig. 1E). As expected, the HA-ERH construct rescued the lethality of siERH-5 but not siERH-3 (Fig. 1F).

Fig. 1.

Synthetic lethal interactions between ERH and the KRAS oncogene. (A) DLD-1 and HCT116 KRAS mutant cells show less viability compared with their respective WT control after retroviral ERH shRNA infection. Cell viability was assessed 4 d post-shRNA infection (error bars indicated SD of three independent experiments in all figures unless otherwise indicated). (B) Confirmation of ERH protein knockdown by ERH shRNA at 4 d post-shRNA infection. The number below each band indicates relative ERH protein level. (C) DLD-1 and HCT116 KRAS mutant cells show less viability compared with their respective KRAS WT control after ERH siRNA transfection. Cell viability was assessed 3 d post-siRNA transfection (assessment was the same for all siRNA viability experiment unless otherwise stated). (D) Confirmation of ERH protein knockdown by ERH siRNAs at 3 d post-siRNA transfection. (E) An HA-ERH cDNA rescue construct lacking the UTR regions of endogenous ERH remains sensitive to siERH-3, but it is resistant to siERH-5, which was confirmed by Western blot. (F) Stable expression of HA-ERH rescues the toxicity of siERH 5 in both DLD-1 and HCT116 KRAS mutant cells. (G) ERH depletion strongly decreases the ability of DLD-1 KRAS mutant cells to form colonies in soft agarose. Cells were transfected with indicated siRNAs and plated in soft agarose. Colonies were counted 14 d later. (H) Correlation of sensitivities to KRAS and ERH siRNAs in a panel of KRAS mutant (SW1116, SW620, SW403, LS123, and LOVO) and WT (RKO, CACO2, and SW48) CRC cell lines.

To test whether ERH is required for the transformed phenotype of the KRAS mutant cells, we tested ERH-depleted DLD-1 KRAS mutant in soft agarose colony assay. Loss of ERH nearly completely inhibited the anchorage-independent growth of these cells (Fig. 1G and Fig. S1E). We next tested whether the synthetic lethality of ERH depletion is selective for the KRAS oncogene. We depleted ERH in isogenic phosphatidylinositol 3-kinase (PI3K) mutant and WT DLD-1 cells (21) and observed no difference in their viabilities (Fig. S1F). Thus, ERH synthetic lethality is selective for the KRAS oncogene but not the PIK3CA oncogene.

To assess whether ERH is important for the survival of additional KRAS mutant CRC cell lines, we tested eight CRC cell lines for their sensitivity to ERH and KRAS depletion. These lines include five KRAS mutant cell lines (SW1116, SW620, SW403, LS123, and LOVO) and three KRAS WT cell lines (RKO, CACO2, and SW48). We found that four of five KRAS mutant lines are variably sensitive to KRAS knockdown, whereas all three KRAS WT lines are resistant. This finding is consistent with previous findings that KRAS mutant cell lines exhibit different degrees of KRAS dependency (22). We observed a strong correlation between these cells’ sensitivity to KRAS depletion and their sensitivity to ERH depletion (Fig. 1H and Fig. S1G). Thus, dependency on ERH is not confined within the isogenic KRAS cell lines, but it is evident in KRAS-dependent CRC cells in general.

ERH Regulates Chromosome Congression.

ERH is highly conserved through evolution, and it shares little sequence homology with other proteins (Fig. S2A). ERH depletion does not seem to affect the level of KRAS protein or signaling through the MAPK pathway (Fig. S2B). Because a number of mitotic genes have been identified in our previous screen (6), we investigated whether ERH might play a role in cell cycle regulation. Indeed, ERH knockdown causes a more pronounced G2/M arrest in KRAS mutant cells compared with KRAS WT cells (Fig. 2A and Fig. S2C). We next used immunofluorescence (IF) staining of serine-10 phosphorylated histone H3 to visualize mitotic cells to investigate the potential cause for the mitotic delay. ERH-depleted cells failed to align chromosomes at the metaphase midplate, and lagging chromosomes at spindle poles are abundantly evident (Fig. 2 B and C). Importantly, these mitotic defects can be rescued by the siRNA-resistant HA-ERH cDNA. To better understand the effect of ERH on chromosome congression, we depleted ERH in U2OS cells stably expressing GFP-H2B protein and tracked mitosis using live-cell fluorescence video microscopy. The time duration from nuclear envelope breakdown to anaphase onset in ERH-depleted cells was approximately sixfold longer than the time duration of control cells, reflecting difficulties in chromosome alignment in the absence of ERH (Fig. 2 D and E and Movies S1, S2, and S3).

Fig. 2.

ERH is required for chromosome congression. (A) Depletion of ERH leads to accumulation of G2/M cells. Cells were fixed 3 d post-siRNA transfection and stained with propidium iodide to quantify the fraction of G2/M cells (*P < 0.05). (B) IF staining of mitotic cells using phospho-H3 serine 10 (pH3S10) antibody, tubulin antibody, and DAPI revealed that ERH depletion leads to chromosome lagging at spindle pole (white arrowheads). Cells were imaged 3 d post-siRNA transfection. (Scale bar: 10 μm in all images unless otherwise indicated.) (C) Quantification of chromosome congression defects in DLD-1 and HCT116 cells as seen in B. (D) Live-cell video microscopy of U2OS cells stably expressing GFP-H2B undergoing mitosis with or without ERH depletion. Representative time frames are shown. Cells were imaged 3 d post-siRNA transfection for up to 12 h at 37 °C in an environment chamber. (E) Quantification of average mitosis duration of U2OS GFP-H2B cells as imaged in D. The mitosis duration is defined as the time lapse between nuclear envelop breakdown and anaphase onset. Numbers at the base of the graph indicate total mitotic events measured for each treatment.

We next investigated how ERH might regulate chromosome congression at the molecular level. In interphase cells, ERH is distributed throughout the nucleus, but it seems to be excluded from the nucleoli. In mitotic cells, ERH is distributed throughout the cells, but it seems to be excluded from chromatin. Interestingly, unlike many spindle checkpoint proteins, ERH does not localize to kinetochores (Fig. 3A). We noticed that the chromosome congression defects associated with ERH depletion resemble the defects of CENP-E loss (13, 14). We, therefore, investigated whether ERH depletion would influence CENP-E activity. During mitosis, CENP-E localizes to kinetochores [as marked by the inner kinetochore protein anticentromere antibody (ACA)]. Kinetochore localization of CENP-E was dramatically reduced on ERH depletion and restored on HA-ERH rescue (Fig. 3B). To better quantify the localization defect of CENP-E, we arrested cells in prometaphase with nocodazole (Fig. S2D) and scored the fraction of cells with or without visible CENP-E at kinetochores. ERH depletion resulted in a 60–70% decrease in the number of cells with CENP-E at kinetochores (Fig. 3C). Surprisingly, ERH depletion causes a dramatic loss of both CENP-E protein (Fig. 3D) and CENP-E mRNA (Fig. 3E), which explains the loss of CENP-E at kinetochores.

Fig. 3.

ERH is required for the expression of CENP-E. (A) Localization of ERH throughout the cell cycle using IF against the HA-tag in DLD-1 cells stably expressing HA-ERH. (B) ERH depletion causes loss of CENP-E from kinetochore. Asynchronous DLD-1 cells in mitosis were stained with antibodies against CENP-E and ACA (an inner kinetochore protein) with or without ERH depletion and rescue. (C) Quantification of CENP-E localization in DLD-1 and HCT116 cells arrested in prometaphase with ERH depletion and rescue. (D) Western blot reveals loss of CENP-E protein on ERH depletion in asynchronous DLD-1 cells. (E) Real-time qPCR using exon-spanning primers reveals loss of CENP-E mRNA on ERH knockdown and recovery of CENP-E mRNA on rescue. (F) IF staining against Bub1, BubR1, and MAD2 in mitotic cells shows normal localization of these proteins to kinetochores. (G) Western blot shows that the levels of Bub1, BubR1, MAD2, and PLK1 are not affected by ERH depletion.

Recruitment of CENP-E to kinetochores depends on the spindle assembly checkpoint (SAC) proteins, and RNAi depletion study indicates that localization of Bub1 to kinetochores in early prophase is necessary for the kinetochore localization of BubR1, CENP-E, and MAD2 (23, 24). We, therefore, investigated whether Bub1, BubR1 and MAD2 localization is also impaired by ERH depletion. All three proteins localized to the kinetochore normally (Fig. 3F), and show normal protein levels in ERH-depleted cells (Fig. 3G). Furthermore, depletion of ERH did not impair the cell’s ability to stably arrest in prometaphase in the presence of nocodazole (Fig. S2D), suggesting that the SAC remains intact in ERHp-depleted cells. Thus, ERH is required for the expression of CENP-E but not other SAC proteins.

ERH Interacts with SNRPD3 and Is Required for CENP-E mRNA Splicing.

Because ERH depletion led to loss of CENP-E mRNA, we investigated how ERH might affect CENP-E expression. ChIP of HA-tagged ERH protein did not reveal significant enrichment of ERH at the CENPE promoter (Fig. S3), suggesting that ERH might not directly regulate CENPE transcription initiation. Because the cellular function of ERH is poorly understood, we used stable isotope labeling by amino acids in cell culture MS to identify its binding partners (Fig. 4A). One candidate ERH binding protein is the small nuclear ribonucleoprotein Sm D3 (SNRPD3), a member of the Sm protein complex that is involved in snRNP assembly and pre-mRNA splicing (25, 26). Coimmunoprecipitation experiments confirmed the interaction between endogenous ERH and SNRPD3 in cells (Fig. 4B). Interestingly, siRNA depletion of SNRPD3 also resulted in CENP-E protein loss and chromosome congression defect (Fig. 4 C–E). These observations support the model that ERH functions as part of an mRNA splicing complex that is required for the splicing of CENP-E mRNA.

Fig. 4.

ERH interacts with the splicing factor SNRPD3 and is required for CENP-E mRNA splicing. (A) Schematics of stable isotope labeling by amino acids in cell culture MS to identify ERH interacting proteins. Proteins that immunoprecipitate selectively with HA-ERH have positive light-to-heavy (L/H) amino acid ratios. The top five candidate proteins are shown. (B) Reciprocal immunoprecipitation (IP) experiments in DLD-1 cells show that endogenous SNRPD3 coimmunoprecipitates with HA-tagged ERH and that endogenous ERH coimmunoprecipitates with endogenous SNRPD3. (C) Western blot verifying depletion of SNRPD3 by two independent siRNAs leads to loss of CENP-E protein. Cells were analyzed 2 d post-siRNA transfection. (D) IF shows that depletion of SNRPD3 in DLD-1 cells leads to chromosome congression defects similar to ERH depletion; 2 d post-siRNA transfection, cells were fixed and stained with pH3S10 antibody, tubulin antibody, and DAPI. (E) Quantification of the chromosomal congression defects in DLD-1 cells on SNRPD3 depletion as seen in D. (F) Real-time qPCR using exon-spanning and splice junction primers at three exon–intron junctions of the CENP-E pre-mRNA. The schematic indicates the location of PCR primers and the species of mRNA that they detect (EE, exon–exon PCR; EI, exon–intron PCR; IE, intron–exon PCR). Depletion of ERH leads to a loss of the spliced products (EE amplicons) and an increased in unspliced mRNA (EI and IE amplicons) at each exon–intron junction.

Human CENP-E is a very large protein with 2,663 aa that are encoded by 49 exons spanning a 92,604-nt pre-mRNA transcript. To test whether ERH is required for CENP-E mRNA splicing, we used intron-spanning primers and splice junction primers to compare the relative levels of CENP-E mature and pre-mRNA in cells with or without ERH knockdown. We tested three splicing junctions with large intervening introns and three junctions with small intervening introns (Fig. 4F and Fig. S4A). Real-time qPCR revealed that, on ERH depletion, there was a decrease in mature splicing products at all six exon–intron junctions, with a concomitant increase in unspliced mRNA (Fig. 4F and Fig. S4B). These results suggest that ERH is required for CENP-E pre-mRNA splicing. Consistent with our finding that ERH interacts with the Sm complex subunit SNRPD3, siRNA depletion of two other Sm complex subunits SNRPD1 and SNRPD2 also led to the loss of CENP-E mRNA (Fig. S4 C and D). Misspliced mRNAs are often rapidly removed in cell through the nonsense-mediated mRNA decay (NMD) pathway (27). Half-life measurement of spliced CENP-E mRNA in the presence of the transcription inhibitor actinomycin D indicates that ERH depletion did not increase the degradation rate of spliced CENP-E mRNA (Fig. S4E). siRNA depletion of up-frameshift suppressor 1 homolog (UPF1), a key protein in the NMD pathway (28), resulted in a partial rescue of CENP-E mRNA and a further increase in CENP-E pre-mRNA levels (Fig. S4F). Thus, the NMD pathway plays a role in degrading aberrantly spliced or unspliced CENP-E pre-mRNA in the absence of ERH.

ERH Regulates the Expression of a Subset of Genes Involved in Cell Cycle.

Because ERH is required for the splicing and expression of CENP-E, we asked whether CENP-E depletion would also constitute synthetic lethality with KRAS. Depletion of CENP-E in KRAS WT and mutant CRC cell lines, however, showed similar toxicity across all lines, with no clear distinction between KRAS WT and mutant cells (Fig. S5A). Thus, CENP-E alone does not mediate the synthetic lethal interaction between ERH and KRAS. We also tested whether pharmacological inhibition of aurora-B kinase, which is required for both CENP-E activity (29) and the SAC (30), might synergize with ERH siRNAs. Two structurally distinct aurora-B kinase inhibitors, AZD1152 (31) and ZM447439 (30), exhibited modest selectivity against KRAS mutant cells but were not synergistic with ERH siRNAs (Fig. S5B). Thus, the nature of mitotic stress in KRAS mutant cells is likely to be complex.

We next carried out gene expression profiling in DLD-1 cells to identify additional genes that might be regulated by ERH. To distinguish siRNA on-target effects from off-target effects, we compared gene expression profiles of DLD-1 cells transfected with siERH-5 with the profiles of control siRNA-transfected cells and cells transfected with siERH-5 together with HA-ERH rescue (Fig. 5A and Fig. S6A). We define on-target genes as those genes with expression that is altered by siERH-5 but restored to normal level on HA-ERH rescue. This analysis identified 64 genes that are down-regulated and 23 genes that are up-regulated on ERH depletion. Ingenuity analysis revealed a significant enrichment for genes involved in cell cycle and DNA replication/repair pathways that are dependent on ERH for their expression (Fig. 5B and Fig. S6B). These genes include CENPE, the condensin subunit SMC4, the mitotic kinesin and cancer testis antigen KIF20B, and DNA damage repair genes ATR, MRE11A, RAD50, RAD54B, and FANCM. Thus, ERH is likely to regulate cell cycle genes, and its synthetic lethal effect could be caused by the simultaneous deregulation of multiple genes.

Fig. 5.

ERH regulates the expression of multiple cell cycle genes. (A) Gene expression profiling comparison of DLD-1 cells transfected with siNEG, siERH-5, and siERH-5 plus HA-ERH rescue identified 87 genes with expressions that are regulated by ERH. (B) Ingenuity Pathway Analysis identified cell cycle and DNA replication/repair as the top two gene networks regulated by ERH. (C) The ERH gene expression signature and two KRAS signatures (Singh KRAS dependency signature and Bhattacharjee lung KRAS signature) were used to score colon and lung cancer cell lines [Cancer Cell Line Encyclopedia (CCLE) dataset] and colon [The Cancer Genome Atlas (TCGA) dataset] and lung (combined Tomida and Chitale dataset) tumor samples. Within each dataset, the ERH signature negatively correlated with the KRAS signature. (D) A model of the function of ERH in mRNA splicing. Our data support the notion that ERH is required for the mRNA splicing of CENP-E and potentially other cell cycle genes and therefore, their expression, and KRAS mutant cells are more dependent on the function of ERH for survival. Dashed lines indicate functional connections with unclear molecular steps that warrant additional investigation.

We next tested whether an ERH gene expression signature consisting of all 87 genes that are deregulated by ERH depletion negatively correlates with two KRAS signatures: the Singh KRAS dependency signature (22) and the Bhattacharjee lung KRAS signature (6). We first analyzed CRC and lung cancer cell lines that were recently characterized by the Cancer Cell Line Encyclopedia project (32). Cell lines were independently scored using each of the three signatures, and the correlations across the signatures were analyzed. Consistent with our hypothesis, ERH signature scores show a moderate but consistent negative correlation with the Singh KRAS dependency signature scores. In the lung cancer cell lines, ERH signature scores also negatively correlated with the Bhattacharjee lung KRAS signature scores (Fig. 5C and Fig. S6 C and D). We next carried out a similar analysis in a collection of CRC tumor samples characterized by The Cancer Genome Atlas (the TCGA colorectal tumors) (33) and a combined set of lung tumor samples (the Tomida_Chitale lung tumors) (1, 3). We again observed a modest but consistent negative correlation between the ERH and the Singh and Bhattacharjee signature scores in these samples (Fig. 5C and Fig. S6 E and F). Taken together, our data support the model that KRAS mutant cells are more dependent on ERH and the gene expression program ERH regulates for survival (Fig. 5D).

ERH Expression Is Associated with CRC Patient Survival.

Our in vitro analysis in CRC cell lines shows that KRAS mutant CRC cancer cells are more dependent on ERH expression for survival. This finding, in turn, suggests that ERH gene expression might be associated with malignancy of KRAS mutant tumors in patients. To evaluate this hypothesis, we analyzed the association between ERH mRNA expression and patient survival in two independent CRC patient cohorts at the National Taiwan University Hospital (NTUH cohort) and the National Cancer Institute (NCI cohort). The patient characteristics in these two cohorts are shown in Table S1. ERH expression is not significantly different between KRAS mutant and WT tumors (Fig. 6 A and B). However, whereas ERH expression was not associated with survival of patients with KRAS WT tumors (Fig. 6 C and D), in patients with KRAS mutant tumors, low ERH expression was associated with significantly better survival (Fig. 6 E and F).

Fig. 6.

Clinical association between ERH expression and survival in CRC patients with KRAS mutation tumors. (A and B) Tumor ERH expression in the NTUH and NCI cohorts. No difference in ERH expression was observed between KRAS WT and mutant tumors in either cohort. (C and D) In the subset of patients with KRAS WT tumors, ERH expression is not associated with survival. Patients were divided into ERH low and high groups based on median ERH expression, and their survival was plotting using the Kaplan–Meier analysis (uptick indicates censoring events). (E and F) In the subset of patients with KRAS mutant tumors, low ERH expression is significantly associated with better survival.

We also evaluated the clinical relevance of ERH expression in stage I/II lung adenocarcinoma using a publicly available gene expression dataset from the Gene Expression Omnibus database (accession no. GSE31210) (34). In this cohort, ERH expression was higher in tumors with EGF receptor (EGFR) mutations (Fig. S7A). In the subset of patients with KRAS mutant tumors, we again observed a strong trend in favor of longer survival with lower levels of tumor ERH expression, although the P value just failed to reach significance (Fig. S7B). Importantly, ERH expression was unrelated to prognosis in patients with EGFR mutations or patients whose tumors were WT for both EGFR and KRAS (Fig. S7 C and D).

Although the number of patients was relatively small in each of the three cohorts analyzed, our results are highly consistent across cohorts, and they support the model where KRAS mutant cancer cells are more dependent on ERH for their viability and low ERH expression could serve to constrain the malignancy of KRAS mutant colorectal and lung cancers.

Discussion

KRAS is a particularly potent oncogene, because it can activate a number of downstream effector pathways, such as the MAPK pathway, the PI3K pathway, and the Ral GTPases (2, 4). Cancer cells that are mutant in the KRAS gene exhibit the classical phenotype of oncogene addiction, which is reflected by their dependence on mutant KRAS for their transformed phenotype and survival (19, 22) (Fig. 1H). However, dissecting the downstream genetic dependencies that constitute Ras addiction has proved more complex. Several genome-wide screens against the KRAS oncogene have identified a surprisingly diverse set of genes whose depletion causes greater toxicity in KRAS mutant cells compared with KRAS WT cells (6, 35–39). We think that this finding reflects a broad dependency of Ras mutant cells on various stress relief pathways that constitute what we termed nononcogene addictions (7).

In this study, we characterized an Ras synthetic lethal gene ERH and provided evidence that ERH is critical for the expression of the mitotic motor protein CENP-E and consequently, chromosome congression. ERH interacts with the Sm protein SNRPD3 and is required for the splicing of CENP-E mRNA. Gene expression microarray revealed that siRNA depletion of ERH results in the loss of expression of many cell cycle and DNA replication/repair genes. The synthetic lethality between ERH and KRAS, thus, is likely to reflect the increased cell cycle stress in the KRAS mutant cells, which was documented previously by us and others (6, 40).

ERH is a highly conserved protein among metazoans and plants (Fig. S2A). Amino acid sequence similarity of ERH orthologs between human and Dictyostelium is 80%, and amino acid sequence similarity of ERH orthologs between human and Arabidopsis is 78%. Distant ERH orthologs exist in the fission yeasts Schizosaccharomyces pombe and S. japonicus but not in the budding yeast Saccharomyces cerevisiae. Such a high degree of conservation suggests that ERH plays an important role in cellular function. ERH has a unique protein fold (41) and has been implicated to play a role in nuclear gene expression (16–18) and cell growth (42, 43). Our study indicates that, through its interaction with SNRPD3, ERH is likely to function as a subunit of an mRNA splicing complex. This finding is consistent with a recent proteomics study identifying ERH as an interactome member of the survival of motor neuron complex that is required for spliceosome assembly (44). Although mRNA splicing factors are often inferred to play a housekeeping role, our data indicate that ERH depletion has a striking effect on chromosome congression through its selective regulation of CENP-E mRNA splicing. CENP-E is degraded on mitotic exit (12), and thus, its protein levels are sensitive to perturbation in its mRNA level. Several large-scale RNAi studies have implicated the role of mRNA splicing factors in mitosis (45–47). These factors include members of the U1, U2, and U5 U4/6 snRNPs and some SR proteins. The mechanisms by which these splicing factors affect mitosis, however, are not known. Recently, the ubiquitin specific peptidase 4 (USP4) was found to be required for the splicing of Bub1 (48), whereas depletion of the serine/arginine-rich splicing factor SON causes pleotropic mitotic defects (49). ERH is likely to operate in a distinct pathway, because it is required for the splicing of CENP-E pre-mRNA for the expression of Bub1, BubR1, MAD2, or PLK1. Accordingly, the ERH-depleted cells exhibit defective chromosomal congression but intact SAC, similar to Cenpe-null mouse embryonic fibroblasts (50). Interestingly, we were able to identify distant ERH orthologs in fission yeast but not budding yeast. S. pombe has considerably more introns in its genome compared with S. cerevisiae, and the S. pombe splicing machinery is more conserved with metazoans (51). Furthermore, during mitosis, chromosome congression occurs in S. pombe (52) but not S. cerevisiae (53). It is, therefore, an intriguing possibility that ERH might have coevolved with CENP-E–like kinesins to regulate chromosome congression. Our microarray analysis indicates that, in addition to CENPE, ERH is also required for the expression of additional cell cycle genes, including SMC4, ATR, and MRE11A. Whether ERH is also required for the splicing of these genes’ pre-mRNA requires additional investigation.

This study, together with previous work by us and others (6, 35), supports the notion that KRAS mutant cells experience elevated mitotic stress and therefore, are more dependent on the proper function of multiple mitotic genes. Our finding is consistent with studies showing that the expression of mutant Ras, either exogenously or from the endogenous locus, can induce aneuploidy in mouse fibroblasts (40, 54, 55). Downstream of Ras, the MAPK pathway has been implicated in this process (56, 57), although the molecular steps connecting Ras with genomic instability remain unclear and warrant additional investigation. Importantly, we did not observe synthetic lethality between ERH and the PIK3CA oncogene. Furthermore, ERH expression is inversely correlated with survival of patients whose tumors harbor KRAS mutations, but it is not associated with survival in those patients whose tumors are WT for KRAS or harbor EGFR mutations. Thus, dependency on ERH is likely a nononcogene addiction that is unique to KRAS-transformed tumor cells. Understanding the mechanisms underlying this addiction could help identify new avenues of therapeutic approaches for Ras-driven tumors, and our study suggests targeted disruption of the mRNA splicing machinery could present one such potential avenue for therapeutic exploration.

Materials and Methods

Cell lines and Reagents.

The KRAS mutant and WT isogenic DLD-1 and HCT116 cell lines were gifts from Bert Vogelstein (Johns Hopkins University School of Medicine, Baltimore, MD) and have been described previously (6, 19, 20). All CRC cell lines were maintained in McCoy’s 5A media with 10% fetal bovine serum. ERH shRNA was expressed using the MSCV-PM retroviral vector. All siRNAs were from Qiagen and Dharmacon and their sequence information is provided in SI Materials and Methods. Rabbit polyclonal anti-ERH antibody was generated using the peptide sequence QPTKRPEGRTYADYC (GenScript). MAD2 and BubR1 antibodies were gifts from Stephen Taylor (University of Manchester, Manchester, United Kingdom). All other antibodies were purchased from commercial sources as listed in SI Materials and Methods.

MS and Microarray Gene Expression Analysis.

The method of stable isotope labeling by amino acids in cell culture (SILAC) was used to label cells prior to affinity purification of HA-tagged ERH and its associated proteins from total cell lysates. Following trypsin digestion, peptide ions were detected in a data-dependent manner using an LTQOrbitrap Velos (Thermo Electron). The six most intense precursor ions from each full MS1 scan were selected for MS/MS by collision-induced dissociation. LC MS/MS data were searched using the MASCOT algorithm within Proteome Discoverer 1.2 (Thermo Electron). Gene expression profile was obtained using GeneChip Human Genome U133 Plus 2.0 Array (Affymetrix). GeneSpring software 11.5 (Agilent) was used for microarray data analysis. For gene expression signature analysis, a two-sided t statistic comparing the average of the signature up genes with the average of the signature down genes within each cell line/tumor was used to compute the gene signature score. Signature scores were correlated using the Pearson method.

Clinical Samples, Patient Data Analysis, and Ethics Statement.

Patient studies were approved by the Ethical Committee of the National Taiwan University Hospital (NTUH cohort) and the Institutional Review Board of the National Institutes of Health and the Institutional Review Board for Human Subject Research at the University of Maryland (NCI cohort), respectively. Written informed consent was received from all participants before inclusion in the study. Somatic mutation status in KRAS was determined by Sanger sequencing. ERH gene expression was determined by quantitative RT-PCR (NTUH cohort) or by microarrays (NCI cohort).

A full description of the materials and methods used for this work is described in SI Materials and Methods.