Centrosome Linker–induced Tetraploid Segregation Errors Link Rhabdoid Phenotypes and Lethal Colorectal Cancers

Discovery of CROCC mutations and centrosome anomalies

Two previously reported primary BRAF^V600E-mutant rhabdoid colorectal cancers (RC1 and RC2; refs. 9, 10), harboring MSI due to defective DNA mismatch repair (MMR) machinery caused by promoter methylation of the MLH1 gene, were subjected to whole-exome sequencing (WES) using DNA from FFPE-matched tumor/normal samples (Supplementary Table S1). We detected an exceptionally large number of somatic point mutations, 1,056 and 1,078 per 10⁶ bases for RC1 and RC2, respectively, which is consistent with the presence of MMR defects (13–15; Fig. 1A). About twenty percent of mutations occurred within CpG dinucleotide context as seen in classical colorectal cancers (14). Transitions were more frequent than transversions (71.8% vs. 28.2%; Supplementary Fig. S1A) with a dominance of C>T/G>A, T>C/A>G transitions (Supplementary Fig. S1B), which is characteristic of the mutational signature due to alterations of MMR mechanisms (signature 6; ref. 13). The most prevalent single-nucleotide variants (SNV) were nonsilent mutations (14), where over 90% of potentially damaging mutations were missense and around 10% were splicing, stop-gain, stop-loss, or, rarely, frameshift insertions or initiation codon mutations (Fig. 1A; Supplementary Fig. S1C and S1D). The two rhabdoid colorectal cancer cases shared 112 (10%) mutated genes (Supplementary Fig. S2A). By applying DrGaP computational tool (16), which allows inferring cancer driver genes, a number of potential candidate disease-causing genes were identified (Supplementary Table S2), approximately half of which (45%) were enriched for cytoskeleton/centrosome and microtubule biological functions (Supplementary Fig. S2B and S2C).

• Download figure
• Open in new tab
• Download powerpoint

Figure 1.

WES reveals mutations in CROCC, encoding an essential component of the centrosome linker. A, Representative rhabdoid colorectal cancer histopathologic images from RC1 and RC2 that were subjected to WES (H&E, hematoxylin and eosin). The graph indicates the total number of somatic mutations per tumor. The circo shows the distribution of nonsilent mutations and copy number variations (CNV); the outer ring indicate the chromosomes. B, Prevalence of alterations in the candidate genes harboring somatic mutations in both RC1 and RC2 in 224 colorectal cancers of the TCGA database. C, CROCC chromosome localization (1p36.13) and organization (from Ensembl, reference transcript ENST00000375541). All 37 exons are depicted as vertical bars and introns as horizontal lines. Solid circles indicate the mutations identified in RC1 and RC2. The “proteinaceous linker” is composed of CROCC filaments (black arrow) that physically connect the mother (M) and daughter (D) centrioles surrounded by the pericentriolar material (PCM). At the onset of mitosis (Mi), the linker is disassembled to support the formation of the bipolar mitotic spindle. D, Quantification of CROCC mRNA (qPCR) expression levels in tumor and adjacent normal mucosa. Data are mean ± SD; n = 5 biological replicates; **, P < 0.01, two-tailed Student t test). Representative images of CROCC immunostaining in nonneoplastic colon mucosa and a fallopian tube used as control (arrow). CROCC immunopositive centrosomes are reduced in number (arrow) or mispositioned (distant/separated from the nucleus) in RC1, (inset modeled image). Scale bars are reported in each microphotograph.

The search of candidate genes in The Cancer Genome Atlas (TCGA) database (13–15) comprising 224 sequenced classical colorectal cancers (http://www.cbioportal.org) revealed that the majority (65%) of the candidates had a low frequency of mutations (≤4% of cases). Among all candidate genes, only for CROCC mapping to 1p36.13 (11, 12) involved in centrosome cohesion and disjunction, no somatic mutations (0/224; 0%) were reported (Fig. 1B). Notably, 1p36 deletions are recurrent in neuroblastoma, Wilms tumor, and medulloblastoma. In our two rhabdoid colorectal cancer cases, CROCC harbored two missense mutations, p.Ala161Ser (c.481G>T, Exon 4) and p.Val1885Ala (c.5654T>C, Exon 35), and one prominent splicing mutation at the conserved 3′ acceptor splice site (c.3705-2A>G) in the intron between exons 25 and 26 (Fig. 1C). A review of multiple colorectal cancer sequencing datasets (N = 2070) revealed CROCC mutations in (1.4% of cases). However, although of unknown significance, none of the CROCC mutations were identified as putative driver mutations in colorectal cancer (Supplementary Fig. S2D). SMARCB1 and SMARCA4 mutations, which have been associated with rhabdoid phenotype (6), showed a trend of mutual exclusivity with CROCC alterations. However, only putative truncating driver mutations in SMARCB1 and/or SMARCA4 correlated with tumor poor differentiation and short-time metastatic progression. Therefore, we reasoned that the splicing mutation detected in RC1 might be causally correlated with rhabdoid phenotype. Indeed, the mutation reduced the strength of the physiologic acceptor site, causing a large deletion of the CROCC coding region involving exons 23–31 (17, 18) (Supplementary Fig. S3A). RT- PCR analysis (exons 5–7 and 33–35) revealed reduced CROCC mRNA in RC1 harboring the splicing mutation as compared to the normal mucosa. This suggested the alteration of the mature CROCC transcript by the utilization of cryptic splice sites or by the activation of the nonsense-mediated mRNA decay, which, in some cases can eliminate aberrant mRNA transcripts (Supplementary Fig. S3B; ref. 17). Unexpectedly, RC2 also displayed expression levels in tumor lower than in normal tissue, supporting a role for defective CROCC expression in rhabdoid tumors (Fig. 1D). We next used IHC and immunofluorescence at high magnification with an anti-CROCC antibody to analyze the centrosomes in rhabdoid cancers and matched normal tissues (Supplementary Table S3). We found that nearly 50% of tumor cells had no CROCC immunolabeling, and the presence of cells with a single and often abnormally shaped, larger (up to 6-fold greater than normal), or fragmented centrosomes, suggesting the presence of numerical and structural centrosome aberrations (Supplementary Fig. S3B). We also observed dramatic and uncommon cytologic defects, such as anucleated cells having larger centrosomes positive for CROCC associated with mitotic catastrophe in late telophase, particularly in RC1 harboring the splicing site mutation (Fig. 1d). We used the pericentriolar material (PCM) component γ-tubulin as our reference marker for immunolabeling experiments, because it consistently colocalizes with centriole markers, which are closely connected in interphase by the centrosomal linker (1, 2, 11, 12). Moreover, γ-tubulin has been proposed as a marker to identify spindle poles (19, 20). We observed a remarkable loss of cell polarity in interphase nuclei and abnormal mitotic figures, many of which included asymmetric bipolar or monopolar spindles. Rhabdoid cells showed a diffused staining of γ-tubulin into the cytoplasm and reduced centrosomal localization, a phenomenon described in tumors with high metastatic potential (ref. 19; Supplementary Fig. S3C and S3D). Double immunofluorescence analysis using antibodies directed against CROCC and γ-tubulin confirmed these observations and revealed cells either with fragmented/larger centrosomes or with a consistent loss of centrosome staining (Supplementary Fig. S3). These experiments indicated that genetic defects in CROCC and other centrosome components may compromise centrosome function in rhabdoid colorectal cancers.

A validation set of 10 additional rhabdoid colorectal cancers was studied, including 3 cases (RC7, RC9, RC11) with microsatellite instability due to MLH1 promoter methylation and 7 cases (RC3–6, RC8, RC10, RC12) with stable microsatellites (Fig. 2A; Supplementary Table S1). Targeted sequencing identified three CROCC mutations (p.Ser1320Ile, p.Arg1659His and p.Ala1510Thr) of unknown significance in additional 2 cases (RC9 and RC11) harboring microsatellite instability (1, 2, 11, 12). Indeed, the 24 CROCC mutations identified across cBioportal database were more recurrent in MSI (12/24; 50%) than in MSS (6/24; 25%) colorectal cancers. Notably, CROCC mutations were classified as missense (n = 22) or truncating mutations (n = 2) of unknown significance but associated with both well- differentiated and early-stage classical colorectal cancers. In our rhabdoid colorectal cancer dataset, 5 of the remaining cases for which sufficient material was available (cases RC3–RC7) harbored loss of heterozygosity (LOH) at the 1p36.13 locus, where CROCC resides, which was associated with mRNA below normal levels (Fig. 2B). In keeping with the findings in RC1 and RC2 cases, comparable levels of centrosome defects and a high prevalence of bizarre mitotic figures and/or cytomorphologic aberrations were evident in all tumors (Fig. 2C; Supplementary Table S3). Analysis of independent colorectal cancer databases (n = 1,387) for which copy number alterations were available (http://www.cbioportal.org) revealed no alteration at 1p36.13 locus, suggesting CROCC impairment as a consequence of reduced gene dosage (21) caused by allelic deletion. As centrosome anomalies are intimately connected with chromosome segregation errors (1, 3, 20), we assessed DNA content in tumor samples (cases RC1–RC7, Supplementary Table S3). Remarkably, we observed recurrent ploidy abnormalities mainly consisting of triploid or near-tetraploid cells ranging from 10% to 40% of tumor cells (Fig. 2D). Globally, these results indicated that centrosome defects underlie rhabdoid colorectal cancer pathogenesis.

• Download figure
• Open in new tab
• Download powerpoint

Figure 2.

Centrosome anomalies characterize colorectal cancer with rhabdoid phenotype. A, Histopathologic images from a subset of five additional prototypical rhabdoid colorectal cancers (RC), in which are evident rounded eosinophilic cytoplasmic inclusions, eccentric nuclei, and prominent nucleoli. Scale bar, 20 μm H&E images. B, Mutations for selected driver genes and CpG island methylation (CIMP) profile accompanied by loss of heterozygosity analysis at 1p36.13 locus. CROCC mRNA (qPCR) expression levels in tumors and adjacent normal mucosa. Data are mean ± SEM; (n = 5 biological replicates; *, P < 0.05; **, P < 0.01; two-tailed Student t test). C, Representative IHC analysis from case RC5: Cytokeratin-18 (CK18) marks intermediate filaments in an anucleated cell (arrow); Ki67 reveals abnormal chromosome structures (arrow); CROCC marks a multinucleated cell (arrow), anucleated cell (arrow) or it appears fragmented in a mitotic cell (monopolar spindle, arrow). Scale bar, 10 μm. Right, quantification of the centrosome phenotypes against CROCC observed in all RCs (n = 2 experiments, >500 cells/sample). D, Cytogenetic abnormalities (tetraploid signals, arrows) observed by FISH using the centromeric chromosome probes illustrated. Scale bars, 20 and 40 μm. Left, ploidy pattern in all RCs (for chromosomes 1, 12, and 17, n = 2 experiments, >500 cells per sample). Right, quantification of cells with polyploidy measured as ratio of triploid and tetraploid on diploid cells for each tumor.

Centrosome and genomic profiling of rhabdoid tumors of infancy

Insight into genetic characterization of rhabdoid neoplasms are limited to the so called extrarenal rhabdoid tumors arising in children, in which inactivating mutation and/or deletion of the chromatin remodeling gene SMARCB1 (INI1) and low mutation load have been reported (6, 22–24). We analyzed 7 cases of this tumor type, hereafter named rhabdoid of infancy, for centrosome and molecular anomalies (Fig. 3A; Supplementary Table S4). Compared with rhabdoid colorectal cancers, analysis in pediatric tumors was associated with much higher CROCC mRNA expression levels than matched normal tissues (P < 0.00001; Fig. 3B). Target sequencing identified no mutations (0/7 tumors) in the CROCC gene. Moreover, CROCC immunostaining was seen as a single and large signal adjacent to the nuclei in almost the totality of the cells (80%)—only 10% of cells had no centrosome staining (Fig. 3B; Supplementary Table S4). Consistent with literature, the genetic profile of rhabdoid of infancy revealed missense or truncating mutations in SMARCB1 (5/7, 71%; refs. 6, 23, 24) and/or TP53 (3/7, 42%) accompanied by a near diploid DNA content and less aggressive clinical course when compared with rhabdoid colorectal cancers (Fig. 3C–E). This suggested that rhabdoid of infancy, which is characterized by SMARCB1 (INI1) mutation, did not harbor any CROCC alteration and did not display the centrosomal defects observed in rhabdoid colorectal cancer. Inspection of an available database from pediatric rhabdoid cells (25) confirmed that both mutations and genetic deletion affecting CROCC locus were infrequent (2/20, 10%), whereas the transcript profile tended to be similar to our rhabdoid of infancy dataset (Fig. 3C). Therefore, rhabdoids arising in colorectal cancer, although morphologically indistinguishable from their pediatric counterparts, demonstrate distinct molecular, cytogenetic, and centrosomal aberrations.

• Download figure
• Open in new tab
• Download powerpoint

Figure 3.

Centrosome and cytogenetic aberrations comparison between colorectal and pediatric rhabdoid tumors. A, Representative hematoxylin and eosin (H&E) images of pediatric rhabdoid tumors. Scale bar, 50 μm. B, IHC and interphase FISH analysis for the indicated markers (right). Note that centrosomes are single, larger, uniform in size and close to the nuclei (black arrowhead). Left, quantification of CROCC IHC and centromeric (CEN) signals (Chr 1 and Chr 17) in pediatric rhabdoid tumors, (>500 cells per tumor) were evaluated, percentages represent mean values from three independent investigators. Quantification of CROCC mRNA (qPCR) expression levels. Each circle represents the mean value of five biological replicates from a single lesion (**, P < 0.01, two-tailed Student t test). C, CROCC expression in pediatric rhabdoid–derived cancer cell lines according to copy number alterations and mutational load (Novartis/broad cancer cell lines encyclopedia). D, The panel shows the distribution of nonsilent missense or truncating mutations for the indicated pathways in rhabdoid colorectal cancer (RC) and rhabdoid of infancy. E, Kaplan–Meier overall survival curve for rhabdoid of infancy (age class 2 months–19 years) and rhabdoid colorectal cancers (age class 49–83 years). The P value is obtained by the log-rank test is reported in the graph.

Centrosome and CROCC expression in classical colorectal cancers

We screened 242 primary classic colorectal cancers, comprising two independent series of 140 (dataset A; refs. 26, 27) and 102 (dataset B; refs. 28, 29) cases for CROCC mRNA and protein expression (Supplementary Table S5). CROCC mRNA expression levels were higher in colorectal cancers than in normal colonic mucosa. However, no significant protein expression change in tumor tissues compared with that in normal mucosae was detected (Supplementary Fig. S4A). In cohort A, using IHC and immunofluorescence against CROCC and γ-tubulin, we found round and uniform centrosomes, prevalently in normal number (1–2 per cell; 97/140, 69.3%), supernumerary (>2 per cell; 39/140, 27.8%), and only few (<1 per cell; 4/140, 2.9%) displayed reduced centrosome labeling (Supplementary Fig. S4B). Centrosome abnormalities, particularly supernumerary centrosomes, were more prevalent in advanced stage (stage III–IV) than in low stage (stage I–II) lesions (Supplementary Fig. S4C). In cohort B, which was enriched for stage III–IV tumors (79.5% of cases), we confirmed a high prevalence of supernumerary (60/102; 59%; refs. 1, 3, 20) or defective (11/102; 11%) centrosomes associated to poorer clinical course than those expressing a normal pattern [31/102; 30%, HR = 0.30; 95% confidence interval (CI), 0.21–0.81; P < 0.0001] in keeping with the notion that numerical centrosomal abnormalities are more common in invasive cancers (1, 3, 20; Supplementary Fig. S4C). To independently validate the pattern of gene expression changes detected in our datasets, we analyzed the patient-matched tumor–normal expression data available from the TCGA (14) and three independent datasets GSE20916 (30), GSE41258 (31), and GSE30540 (32), of classical colorectal cancer (Supplementary Fig. S4D). CROCC mRNA was upregulated in colorectal cancer compared with normal only in TCGA database. The analysis of other datasets revealed a heterogeneous expression pattern, while CROCC upregulation compared with normal control did not reach statistical significance. The analysis of GSE30540 (32) dataset, for which both transcriptomic data and degree of chromosome instability (CIN) were available, revealed that CROCC expression levels tended to be lower in CIN-high than in CIN-low tumors (Supplementary Fig. S4D). These data suggested the possibility that imbalanced genetic defects at CROCC locus may be related to marked anomalies in the fidelity of chromosome segregation.

CROCC depletion impairs mitosis and induces rhabdoid-like features

We next sought a genetic basis for the relation between CIN and rhabdoid phenotype by examining WES and transcriptomic data from The Cancer Cell Line Encyclopedia (CCLE; ref. 25). Unexpectedly, data from a collection of 60 colorectal cancer cell lines revealed that the deletions at 1p36.13 locus tended to be more prevalent in CIN-high (23.6%, 9/38) compared with CIN-low (9.1%, 2/22) cells (Supplementary Fig. S5A). However, cell lines with 1p36.13 deletion displayed neither rhabdoid phenotype nor BRAF mutations. As expected, compared with cells retaining 1p36.13 locus, those harboring the deletion revealed a gene expression signature significantly enriched for pathways implicated in chromosomal instability (refs. 33, 34; Supplementary Fig. S5B and S5C). In a panel of colorectal cancer cells, we then confirmed that both CROCC mRNA and protein expression levels were concordant and higher in CIN-low than in CIN-high cell lines (P < 0.05, Supplementary Fig. S6A). CIN-low cells showed centrosomes stained for CROCC and γ-tubulin that were structurally indistinguishable from those in normal human fibroblasts BJ, consistent with literature (ref. 35; Supplementary Fig. S6A and S6B). In line with this, CIN-high displayed a higher frequency of micronuclei and nuclear γH2AX foci than CIN-low cells (ref. 29; Supplementary Fig. S6C and S6D). This suggested that CROCC might be a CIN suppressor gene and its deletion in CIN-low/BRAF-mutant cell lines might be permissive for abnormal phenotypes. Therefore, we reasoned that RKO cells, sharing a near-diploid karyotype, BRAF^V600E mutation/MSI, and alterations in microtubule/centrosomal components with rhabdoid colorectal cancer (14), could be an excellent system to explore CROCC silencing in vitro. We found that the clone sh4, hereafter named CROCC^KD, provided a stable and consistent knockdown of CROCC transcript to more than 75% and protein to 3.3-fold lower then RKO cells transfected with control vector (shCon), achieving nearly comparable levels to those seen in vivo (Supplementary Fig. S7A). Previous studies have demonstrated that CROCC knockdown in nontransformed cells causes centriole splitting and increases centrosome separation (11, 12). By using γ-tubulin and centrin as reference markers, we observed a consistent PCM fragmentation after CROCC depletion, which resulted in abnormal chromosome segregation and higher frequency of monopolar spindles as compared with control (Fig. 4A; Supplementary Fig. S7B). Importantly, monopolar spindles displayed larger or “fragmented” centrosomes, which accounted for 85% of the abnormal phenotype (Fig. 4A–C; Supplementary Fig. S7C). Thus, BRAF-mutant/MSI colorectal cancer cell lines, in which centrosome and microtubule stability is damaged by genetic hypermutation, CROCC depletion may determine a major impact in the progression of mitotic errors (36–38). Consistently, an increased frequency of micronuclei (median 11% CROCC^KD vs. 1% ShCon cells P = 0.0003) and γH2AX nuclear foci (median 43% CROCC^KD vs. 18% ShCon cells P = 0.011) were observed (Fig. 4C; Supplementary Fig. S7D). Metaphase karyotyping revealed that CROCC deficiency leads to an increased number of tetraploid (4N) cells (median 13.3% CROCC^KD vs. 3.51% ShCon cells, P = 0.001) characterized by prominent and larger nuclei than diploid (2N) cells. Consistently, analysis of centromeric probes in intephase nuclei confirmed tetraploidy (Fig. 4C). The number of CROCC-deficient cells was reduced in G₀–G₁ or G₂–M phases when compared with the wild-type population (by 26%–45%, FACS analysis), suggesting an impaired cell-cycle progression as a consequence of misaligned chromosomes (1–3, 33; Supplementary Fig. S7D). In contrast, cells grown under replication stress conditions (serum deprivation) resulted in higher proliferation rate than control cells (ref. 34; Fig. 4D). Most strikingly, CROCC-deficient cells exhibited all cardinal signs of rhabdoid features (8–10), displaying huge nuclei pushed to the periphery of the cells with single or multiple large nucleoli associated with eosinophilic cytoplasmic inclusions and large cellular protrusions resembling the morphology observed in vivo (Fig. 4D; Supplementary Fig. S8A). These features resulted in the activation of prometastatic genes involved in epithelial mesenchymal transition accompanied by a dramatic change of spindle-shaped morphology (4, 7) consistent with the enhanced metastatic potential of rhabdoid phenotype (Supplementary Fig. S8A and S8B). Expression of exogenous GFP-tagged CROCC (1–2018 aa) rescued these phenotypic changes induced by depletion of endogenous CROCC (Supplementary Fig. S8B; ref. 11). In line with previous results (11, 12), we observed no alteration in cell-cycle profile or aberrant phenotypic changes after CROCC depletion in nontransformed BJ cells. Therefore, CROCC depletion in BRAF-mutant near-diploid cancer cells induces tetraploidization and rhabdoid phenotype in vitro.

• Download figure
• Open in new tab
• Download powerpoint

Figure 4.

CROCC depletion induces rhabdoid phenotype exacerbating DNA segregation errors. A, Images of RKO cells with stable CROCC depletion (CROCC KD) showing that in mitosis there is an abnormal spindle formation, “monopolar spindles,” as compared with control (CON, left). Large (white arrow) or fragmented (arrow) centrosomes are shown. Cells are stained using immunofluorescence antibodies as indicated (αTUB, anti-α-tubulin antibody; γTUB, anti-β-tubulin antibody) and nuclei are stained with DAPI (4,6-diamidino-2-phenylindole). Scale bar, 5 μm. Below is a schematic illustration of cells with aberrant spindles (85%). B, Anaphase bridges (white arrow), multinucleated (arrow), multilobulated nucleus (arrow), and fragmented centrosomes (arrows) associated with loss of cell polarity and large micronuclei (white arrow) in CROCC-depleted cells. Scale bar, 5 μm. C, The top left graph shows the percentage of micronuclei and monopolar spindles (>250 cells per cell line, triplicate experiments, **, P < 0.01; ***, P < 0.001 Mann–Whitney U test). Representative images of metaphase chromosome spreads and cells stained with DAPI and anti-centromere antibody (ACA). Scale bar, 10 μm. The bottom left graph shows the quantification of ploidy content at metaphase. Data are mean ± SEM; n = 5 biological replicates; **, P < 0.01, two-tailed Student t test). The bottom right graph shows tetraploid on diploid cells ratio. D, The top graphs report the survival assay with serum supplementation or under replication stress condition “serum deprivation.” Error bars represent mean ± SEM of five independent experiments (*, P < 0.05; **, P < 0.01; ***, P < 0.001, two-tailed Student t test. Below are representative cytomorphologic changes showing large polygonal cells and eccentric round nuclei with prominent nucleoli (black arrow) and eosinophilic hyaline cytoplasmic inclusions (arrow).

Tolerance to mitotic errors and tetraploidization promoted by 1p36.13 deletion

As colorectal cancer cells with driver mutations in CROCC have not been reported, to test the hypothesis that CROCC impacts tumor growth and centrosome-related mitotic errors, we analyzed metastatic colorectal cancer T84 cells harboring an allelic deletion at 1p36.13 locus (25). Although T84 are well-differentiated cancer cells and do not show rhabdoid morphology, they, however, exhibit some of the characteristics detected in RKO CROCC-depleted cells. Consistent with reduced CROCC endogenous activity, we observed an increased rate of micronuclei, tetraploid, or near-tetraploid cells and recurrent mitotic errors resulting essentially in “monopolar spindles,” which were more recurrent under replication stress conditions (Fig. 5A; Supplementary Fig. S8C and S8D). We then investigated the localization of CROCC in the centrosome by immunofluorescence. Almost half of the cells (40%) revealed a faint CROCC signal, which was consistently accompanied by atypical γ-tubulin aggregates prevalently in cells with mitotic anomalies (Fig. 5A). Most strikingly, such aberrations were rarely, if ever, detected in pediatric rhabdoid G401 or colon cancer cell lines with an intact 1p36.13 locus (Supplementary Fig. S8C and S8D). Therefore, we transfected CROCC-GFP and GFP alone (control) into T84 cells. Restoration of CROCC determined a dramatic decrease of cell viability (12 days later, 0%) as compared with control plasmid. Similarly, we detected a higher number of G₀–G₁ cells than control (41% vs. 26%; P = 0.0018; Fig. 5B and C). Gain of CROCC conferred a flat/adherent phenotype and formation of filament-like structures colocalizing with γ-tubulin resulting in an expression of mesenchymal genes lower than in control (ref. 4; Supplementary Fig. S9A). Accordingly, we detected a 4-fold decrease of tetraploid cells, and reduced γH2AX foci from 59% to 22% with respect to control cells (Fig. 5C) raising the possibility that the centrosome spindle pole integrity is strongly affected by 1p36.13 deletion. To see whether T84 cells are sensitive to mitotic drugs, we mined the data from the Genomics of Drug Sensitivity in Cancer project (Sanger panel). As shown in (Supplementary Fig. S9B), among 221 molecules tested, IGF1R inhibitor (linsitinib) and Epothilone B, a microtubule-stabilizing agent, were significantly effective in T84 lines. Accordingly, we observed a significant difference in the sensitivity to Epothilone B in 1p36.13-deleted cells as compared with cells with an intact 1p36.13 locus. Similar results were not reproduced comparing CIN-low and CIN-high colorectal cancer cell lines (Supplementary Fig. S9B). We used another cell line HT29 with an intact 1p36.13 locus to test CROCC restoration. Similarly to T84, we observed a significant decrease of micronuclei in HT29-CROCC-GFP⁺ cells compared with control. Although gain of CROCC in HT29 increased the cell death, it appeared an essential gene only for T84 cell survival lacking 1p36.13 (Supplementary Fig. S9C). This supported the hypothesis that a reduced CROCC dosage promotes defects in spindle-assembly checkpoint. When we repeated the experiments using a GFP-tagged truncated form of CROCC (1–494 aa; ref. 11), we observed that this mutant failed to rescue the aberrant growth phenotype and mitotic errors (Supplementary Fig. S9D). Consistent with previous findings (11), we did not detect phenotypic changes in nontransformed BJ cells transduced with the full-length CROCC construct. Thus, we conclude that DNA segregation errors resulting from impaired centrosome function are driven by reduced CROCC dosage at 1p36.13 locus.

• Download figure
• Open in new tab
• Download powerpoint

Figure 5.

CROCC abrogates centrosome-related mitotic errors in 1p36.13-deleted cancer cells. A, Images of T84 cells with a large micronucleus (white arrow) and significantly reduced CROCC staining (enlarged in insets), anaphase bridges (arrow), or monopolar spindle (arrow) associated to deficient or fragmented γ-tubulin dots (enlarged in insets). Scale bar, 5 μm. The top right graph reports the percentage of anaphase showing segregation errors and micronuclei in T84 cells with serum supplementation or serum deprivation (SD) at 12 hours. Error bars represent mean ± SEM; *, P < 0.05, two-tailed Student t test). B, Representative images of T84 cells transfected with full-length human CROCC-GFP or GFP alone, immunostained for γ-tubulin (enlarged in insets). The graph on the right shows the survival of T84 transfected with CROCC and matched control cells maintained in neomycin selection (0.6 μg/mL) for the indicated time. Viability was assessed by a colony formation assay. The GFP vector was used as a control. Cells were fixed, stained, and photographed after 6 and 12 days of culture. C, The left graph reports flow cytometry analysis after 6 days. Error bars represent mean ± SEM of five independent experiments; **, P < 0.01; ***, P < 0.001, two-tailed Student t test). Tetraploid on diploid cells ratio after 6 days quantified by metaphases spreads (16 independent experiments for each condition; **, P < 0.01, two-tailed Student t test). The right graph shows the percentage of γH2AX foci in prometaphase (>250 cells per cell line, ***, P < 0.001, Mann–Whitney U test). D, Schematic representation of rhabdoid colorectal cancer progression. In colorectal cancer with defective microtubule functions, BRAF^V600E mutation depletion of CROCC causes defective centrosome structure, abnormal mitotic progression, and lethal cancer phenotypes.