MAP9 Loss Triggers Chromosomal Instability, Initiates Colorectal Tumorigenesis, and Is Associated with Poor Survival of Patients with Colorectal Cancer

Colorectal cancer is the third most common cancer worldwide and the second most common overall cause of cancer death (1, 2). Nevertheless, the mechanism leading to colorectal cancer initiation and development remains elusive. Chromosomal instability (CIN) is a major type of genomic instability, in which either the whole or parts of chromosomes are duplicated or deleted, leading to a range of karyotypic abnormalities (3). One main form of CIN, the recurrent missegregation of whole chromosomes during cell division, leads to aneuploidy, a hallmark of most solid tumors especially colorectal cancer (4). CIN can arise from a myriad of events, such as failed correction of chromosome error, an elevated formation of improper chromosome spindle attachment caused by abnormal spindle microtubule dynamics and/or spindle geometry defects (3). CIN contributes to disease progression at multiple stages of tumor evolution and a potential therapeutic target for intervention (5). Despite the importance of CIN to tumorigenesis, underlying mechanisms that drive CIN and its precise role in cancer initiation and development remain poorly understood.

Through a genome-wide screening, we identified that microtubule-associated protein 9 (MAP9), also known as aster-associated protein, was frequently inactivated in colorectal cancer. MAP9 localizes to microtubules in interphase and associates with the mitotic spindle during the whole process of mitosis, and the loss of MAP9 induced severe mitotic defects with delayed mitotic progression (6). However, the role of MAP9 in induction of CIN and oncogenic transformation is unknown. In this study, we demonstrated for the first time that MAP9 is an indispensable spindle microtubule stabilizer and chromosomal fidelity in the colonic epithelium. Knockout of Map9 triggers mitosis failure, aneuploidy, and CIN, leading to spontaneous colorectal cancer development in mice. MAP9 is commonly silenced in human colorectal cancer and is associated with poor survival of patients with colorectal cancer.

A total of 141 patients with histologically confirmed colorectal cancer who underwent surgery at the Beijing University Cancer Hospital (Beijing, China) were enrolled in the study. Biopsy samples from primary colorectal cancer tumor and adjacent normal were obtained from patients with colorectal cancer prior to any therapeutic intervention. In addition, 20 age-matched normal colon mucosae from healthy subjects were collected as normal control. All the patients provided informed consent for collecting specimens for study. The study protocols were approved by the Clinical Research Ethics Committee of Prince of Wales Hospital and the Chinese University of Hong Kong. This study was carried out in accordance with the Declaration of Helsinki of the World Medical Association. For survival analysis, the tumor/normal ratio of MAP9 mRNA was calculated as follows: |${{\rm{2}}^{( {\Delta \Delta {C_{\rm{t}}}( {{\rm{Tumor }}( {\Delta {C_{\rm{t}}}( {MAP9\, - \,\,\beta \hyphen ACTIN} )} ){\rm{ }} - {\rm{ Normal}}( {\Delta {C_{\rm{t}}}( {MAP9{\rm{ }} - {\rm{ }}\beta \hyphen ACTIN} )} )} )} )}}$|⁠. If tumor/normal ratio < 1, patients were defined as low MAP9 expression group. If tumor/normal ratio > 1, patients were defined as high MAP9 expression group.

Spectral karyotyping (SKY) analysis was used to uncover chromosomal abnormalities (7). Briefly, metaphase spreads were prepared on glass slides, denatured, and hybridized with denatured SKY probes (SkyPaint mixture, Applied Spectral Imaging).

Microtubule regrowth assay was performed as described previously (8). To completely depolymerize microtubules, 3 μmol/L nocodazole (Sigma-Aldrich) was added to cells for 3 hours. Nocodazole was then washed out with prewarmed culture medium and regrowth of microtubules and spindle reassembly were observed by live-cell imaging.

Microtubule cold-stable assay was performed as described previously (9). Briefly, cells were arrested in mitosis with 0.32 μmol/L nocodazole (Sigma-Aldrich) for 2.5 hours and released into 10 μmol/L MG132 (Sigma-Aldrich) for 40 minutes to enrich metaphase cells. Then cells were incubated for 10 minutes on ice before fixation and immunofluorescence staining.

Kinetochore microtubule turnover was determined as described previously (10). Briefly, cells stably expressing photoactivatable GFP (PA-GFP)-tubulin were incubated with 5 μmol/L DRAQ5 DNA dye (Abcam) for 5 minutes, which is a cell-permeable far-red fluorescent DNA dye for live cells. Spindles with metaphase plate were identified with the DRAQ5 signal. PA-GFP-tubulin was activated in thin stripes (2 pixels wide) on one side of the metaphase plate for 15 seconds from a 405 nm laser (100%). GFP fluorescence was captured every 5 seconds for 300 seconds. The poleward microtubule flux rate was manually calculated.

Map9-deficient mice containing a “knockout-first” allele were generated by Beijing Biocytogen Co., Ltd (Supplementary Fig. S1A and S1B; ref. 11). The “knockout-first” conditional allele was generated by inserting a gene trap cassette through homologous recombination in mouse embryonic stem cells into intron 4 of mouse Map9 gene, leading to disruption of Map9 gene. Cassettes were removed by Flp-mediated excision by crossing with mice expressing Flp (B6.129S4-Gt(ROSA)26Sortm1(FLP1)Dym/RainJ; Beijing Biocytogen Co., Ltd), resulting in a floxed Map9 allele, in which the critical exon 5 of Map9 is flanked by loxP sites (Map9^flox/+). To generate constitutive Map9-knockout mice (Map9^Δ/+ and Map9^Δ/Δ), germline Cre-mediated excision was performed by cross-mating Map9^flox/flox mice with mice harboring the cytomegalovirus (CMV) promoter−Cre (CMV-Cre; B6.C-Tg(CMV-cre)1Cgn/J; Beijing Biocytogen Co. Ltd). Genotyping was performed by PCR of tail-snip DNA using genotyping primers (Supplementary Fig. S1C; Supplementary Table S1). Absence of Map9 mRNA in Map9^Δ/Δ mice was confirmed by qPCR (Supplementary Fig. S1D). All mice were housed in a pathogen-free barrier environment for the duration of the study. All experimental procedures were approved by the Animal Ethics Committee of the Army Medical University (Chongqing, China).

The isolation of mouse embryonic fibroblasts (MEF) was performed as described previously (12, 13). Map9^Δ/+ and Map9^Δ/Δ MEFs were isolated from embryonic day 13–14 fetuses after intercross of constitutive Map9^Δ/+ mice. Wild-type (WT) MEFs were obtained by cross-mating WT mice.

CDX2-CreER^T2 transgenic mice was purchased from The Jackson Laboratory, which express Cre-ER^T2 in adult epithelium of the distal intestinal tract under the direction of caudal type homeobox 2 (CDX2) gene promoter/enhancer regions. Cre-ER^T2 can only gain access to the nuclear compartment after exposure to tamoxifen. Map9^flox/flox mice were bred to CDX2-CreER^T2 to generate Map9^flox/+CDX2-CreER^T2 mice and Map9^flox/floxCDX2-CreER^T2 mice (intestinal-specific Map9^−/+ and Map9^−/− mice; Supplementary Fig. S1A–S1C). All experimental procedures were approved by the Animal Ethics Committee of the Army Medical University (Chongquing, China).

To induce colon epithelium–specific deletion of Map9, 7-week-old intestinal-specific Map9^−/+ and Map9^−/− mice were intraperitoneally injected with tamoxifen (Sigma-Aldrich) dissolved in corn oil (100 mg/kg) for 4 consecutive days. Two weeks after first tamoxifen dose, mice were injected intraperitoneally with 10 mg/kg azoxymethane (AOM; Sigma-Aldrich) once a week for 6 consecutive weeks (14). Mice were sacrificed 25 weeks after the first dosing of AOM and subjected to a complete necropsy. Intestine examination, tumor measurement, and histology inspection were performed. All experimental procedures were approved by the Animal Ethics Committee of the Army Medical University (Chongquing, China).

Other additional experimental procedures are provided in the Supplementary Information.

MAP9 mRNA expression was downregulated in 68% (96/141) of primary colorectal cancer tissues as compared with their adjacent normal tissues (P < 0.0001; Supplementary Fig. S2A). We analyzed MAP9 mRNA levels in the Broad Institute Cancer Cell Line Encyclopedia (CCLE) database. Among 58 colorectal cancer cell lines in CCLE, there are only 4 colorectal cancer cell lines with MAP9 mRNA reads per kilobase per million mapped reads > 1 (SNU503, SNUC4, SNU1033, CW2; 6.9%; Supplementary Fig. S2B). By qRT-PCR, MAP9 mRNA was silenced in all 7 commercial human colorectal cancer cell lines (CL-14, DLD-1, HCT116, HT-29, LS 180 SW480, and SW620) and in all 5 primary human-derived colorectal cancer cells (colorectal cancer spheroid POP92 and POP66, colorectal cancer organoid PDO 828, PDO 1482, PDO 816), but was readily expressed in normal human colon tissues and two immortalized normal human colonic epithelial cells HCEC 1CT (1CT), HCEC 2CT (2CT; Supplementary Fig. S2C). The downregulation of MAP9 expression was validated in The Cancer Genome Atlas (TCGA) cohort of patients with colorectal cancer (Supplementary Fig. S2D).

Given it is downregulated in colorectal cancer, we hypothesized that MAP9 might function as a tumor suppressor in colorectal cancer. We generated intestinal epithelium–specific, inducible Map9-knockout (Map9^−/−) mice by crossing Map9^flox/flox mice to CDX2-CreER^T2 mice (Fig. 1A; Supplementary Fig. S1A–S1C; ref. 15). Tamoxifen injection activates CreER^T2, leading to excision of floxed exon 5 in the intestinal epithelium (Supplementary Fig. S1A). Cre-mediated excision of the floxed exon 5 results in a frameshift and a premature stop codon in exon 6, generating truncated MAP9 protein lacking C terminus, a domain essential for microtubule-binding ability of MAP9 (Supplementary Fig. S1B; ref. 6). At 15 months postinjection, intestinal-specific Map9^−/− mice spontaneously developed colonic dysplasia (5/22, 22.7%) and colon cancer (2/22, 9.1%; Fig. 1B and C). Moreover, intestinal-specific heterozygous Map9^+/− mice also developed colonic dysplasia (4/21, 19.0%), whereas their 23 WT littermates (Map9^+/+) showed normal colon histology at the same age (Fig. 1C).

To confirm the role of Map9 deficiency in promoting colorectal tumorigenesis, constitutive Map9-knockout (Map9^Δ/Δ) mice were cross-mated with Apc^Min/+ mice, a mouse model of colorectal tumorigenesis (Fig. 1A). As Apc^Min/+ mice typically develop intestinal adenomas by 3 months of age, we investigated whether Map9 deficiency can promote tumorigenesis in the Apc^Min/+ mice model (Fig. 1A). Macroscopic examination of mice at 3 months of age revealed that WT (Apc^+/+), Map9^Δ/+ (Map9^Δ/+Apc^+/+), or Map9^Δ/Δ (Map9^Δ/ΔApc^+/+) mice developed no visible tumors, whereas 50% Apc^Min/+ mice (5/10) developed tumors (3/10 with multiple tumors; Fig. 1D and E; Supplementary Fig. S3; Supplementary Table S2). Importantly, nearly all age-matched Map9^Δ/+Apc^Min/+ and Map9^Δ/ΔApc^Min/+ mice developed intestinal tumors (Map9^Δ/+ Apc^Min/+: 2/10 with multiple tumors; Map9^Δ/ΔApc^Min/+: 5/8 with multiple tumors; Fig. 1D and E; Supplementary Table S2). Map9^Δ/ΔApc^Min/+ mice had 3-fold more tumors than Apc^Min/+ mice, accompanied by significant body weight loss (Fig. 1D). Average tumor burden (sum of all tumor size per mouse) in Map9^Δ/ΔApc^Min/+ mice was two times greater than in Apc^Min/+ mice (Fig. 1D; ref. 14). Map9^Δ/ΔApc^Min/+ mice showed markedly decreased overall survival compared with Apc^Min/+ mice (P = 0.016; Fig. 1F).

Intestinal-specific Map9^−/−, Map9^−/+, and WT mice were treated with colonic carcinogen AOM to induce colorectal carcinogenesis (Fig. 1A and G; ref. 14). Tumor incidence was significantly higher in Map9^−/− mice (100%) and Map9^−/+ mice (100%) compared with WT mice (67%; Supplementary Table S2). Both intestinal-specific Map9^−/+ and Map9^−/− mice had a significant increase in tumor number and burden compared with control mice (Fig. 1H and I). These findings further confirmed that Map9 knockout promoted colon tumorigenesis in mice.

In light of the observed tumor-promoting effect of Map9 deficiency in mice, we investigated the molecular mechanisms of MAP9 in colon cancer. The subcellular localization of MAP9 was examined in the immortalized normal human colonic epithelial cells 1CT, 2CT, and immortalized normal human gastric epithelial GES1 cells using Flag-tagged MAP9. MAP9 colocalized with α-tubulin in interphase as well as during the whole process of mitosis in all three cell lines (Fig. 2A; Supplementary Fig. S4A). The colocalization of MAP9 with microtubules was confirmed by super-resolution fluorescence imaging analysis under stochastic optical reconstitution microscopy (STORM; Fig. 2B). In addition, MAP9 closely surrounded the pericentriolar γ-tubulin ring complex (γ-TuRC; Fig. 2C), an essential component of pericentriolar material (PCM) targeted to centrosomes (16). From these findings, we inferred that MAP9 might function as a microtubule-associated protein involved in chromosomal microtubule assembly and dynamics. We thus tested the function of MAP9 in mitosis. Knockdown of MAP9 in 1CT, 2CT, and GES1 cells delayed mitotic progression to varying extents as evidenced by increased G₂–M phase cells' accumulation (Fig. 2D; Supplementary Fig. S4B).

We next investigated underlying role of MAP9 on cell mitosis in 1CT, 2CT, and GES1 cells. We observed that bipolar spindles in metaphase were significantly shorter in MAP9-knockdown cells (Fig. 3A); conversely, ectopic expression of MAP9 led to longer bipolar spindles (Fig. 3B), indicating that MAP9 has microtubule-stabilizing activity (17). Notably, monopolar spindle formation was induced after ectopic expression of MAP9 in metaphase cells (1CT: 62.6%; GES1: 53.8%). This phenomenon was also observed in a previous study (6), which is probably due to the toxic effect for the cells under high constitutive expression of MAP9. As MAP9 colocalized with chromosomal microtubules during mitosis, we carried out a microtubule regrowth assay to examine its role in spindle assembly by live-cell imaging using α-tubulin-monomeric red fluorescent protein (mRFP)–expressing cells. In control cells, more than two microtubule asters were formed within 5 minutes regrowth after nocodazole release, which then quickly coalesced and progressively organized into a bipolar spindle (Fig. 3C and D; Supplementary Fig. S4C; Supplementary Video S1). However, in MAP9-knockdown cells, two or less microtubule asters appeared within same time frame with less densely packed and short microtubules, thus forming a much smaller spindle (Fig. 3C and D; Supplementary Fig. S4C; Supplementary Video S2). In addition, in approximately 20% of MAP9-knockdown GES1 cells, the spindle collapsed along with PCM fragmentation (Supplementary Fig. S4C). Hence, MAP9 is required for proper spindle assembly and spindle stability.

We next examined the consequence of MAP9 loss on mitosis following spindle assembly by live-cell imaging. Control cells quickly assembled a bipolar spindle and completed mitosis within 1.5 hours (Fig. 3E and F; Supplementary Videos S3 and S4). The spindle length of control cells was kept stable in metaphase. Mitotic chromosomes then underwent condensation and became fully condensed prior to the segregation of chromosomes. In contrast, MAP9-knockdown cells remained arrested in mitosis during the 4-hour observation period (Fig. 3E and F; Supplementary Videos S5 and S6). The spindle length of MAP9-knockdown cells in metaphase was approximately half of that in control cells. Although the formation of bipolar spindle and the alignment and condensation of chromosomes were also observed in MAP9-knockdown cells, these spindles were highly unstable and underwent repeated cycles of collapse and reassembly (Fig. 3E and F).

Consistent with the compromised spindle dynamics, more than half of the bipolar spindles in MAP9-knockdown cells were abnormal, presented as asymmetrically curved bending of kinetochore fibers (K-fibers; Supplementary Fig. S5A). We further assessed whether MAP9 affects correction of improper chromosome–spindle attachment using monastrol washout assay. The Eg5 inhibitor monastrol induces monopolar spindles along with many syntelic attachments, a major type of erroneous kinetochore attachments during spindle assembly (18, 19). The percentage of cells without complete chromosome alignment on the metaphase plate was elevated upon release from monastrol in MAP9-knockdown cells (Supplementary Fig. S5B).

K-fibers are microtubule bundles, of which the minus ends are associated with γ-TuRC in pericentriolar matrix and undergo controlled depolymerization at spindle poles to modulate poleward movement of chromosomes during mitosis (20, 21). Because γ-TuRC was closely surrounded by MAP9, we proposed that MAP9 might affect K-fiber formation and stability. Indeed, we observed that K-fiber length in monopolar spindles assembled in the presence of monastrol was substantially reduced in MAP9-knockdown cells (Fig. 4A and B; Supplementary Fig. S5C). Correspondingly, the percentage of cells with extremely short K-fibers was dramatically increased in MAP9-knockdown cells (Fig. 4B), indicating that MAP9 is required for K-fiber minus-end stability. Consistently, microtubule cold-stable assay showed that K-fibers depolymerized much faster upon cold exposure at their minus ends in MAP9-knockdown cells compared with control cells, with approximately 20% of MAP9-knockdown cells having less than 20 microtubules left, compared with <10% of control cells (Fig. 4C and D; Supplementary Fig. S5D). The minus ends of K-fibers attaching to spindle poles depolymerize continuously while the plus ends undergo polymerization, leading to a characteristic tubulin flux from the plus end toward the minus end (Fig. 4E, left; ref. 22). Using cells stably expressing PA-GFP-tubulin, microtubule poleward flux was measured by tracking marks photobleached into green fluorescent spindles in metaphase (Fig. 4E). As shown in Fig. 4F, poleward flux in metaphase spindles in most of MAP9-knockdown cells quickly moved from the metaphase plate to spindle pole within 90 seconds, whereas no obvious poleward flux in metaphase spindles was observed in control cells. Quantitative evaluation showed that poleward flux rates were significantly higher in the metaphase spindles of MAP9-knockdown cells (Fig. 4G). Thus, in the absence of MAP9, K-fiber minus ends are more prone to depolymerization, leading to unstable and shortened K-fibers undergoing a faster flux (Fig. 4E, right). MAP9 likely has no effect on plus ends, as its disruption would abolish microtubule poleward flux (8). Finally, monastrol washout assay demonstrated a significant increase in spindle pole defocusing and PCM fragmentation in MAP9-knockdown cells (Fig. 4H–J; Supplementary Fig. S5E), further confirming the critical role of MAP9 in spindle pole stability and integrity.

MAP9-knockdown 1CT, 2CT, and GES1 cells exhibited high rate of lagging chromosomes with increased chromosome segregation errors (Fig. 5A). Because aneuploidy is an obligate outcome of CIN, conventional karyotyping was performed to assess MAP9-knockdown–induced aneuploidy. All control cells had a normal modal chromosome copy number with a low percentage of mode deviation (Supplementary Fig. S6A and S6B). In contrast, MAP9-knockdown cells showed a significantly higher percentage of deviation from the mode compared with control cells (Supplementary Fig. S6A and S6B). To further identify numerical and structural chromosome abnormalities induced by MAP9 insufficiency, we performed SKY analysis on 1CT and 2CT cells after stable MAP9 knockdown. As shown in Fig. 5B–D, the aneuploidy frequency was nine times more in MAP9-knockdown 1CT cells (59.3% vs. 6.0%) and two times more in MAP9-knockdown 2CT cells (30.0% vs. 10.7%) compared with control cells. Thus, MAP9 insufficiency in cells clearly results in CIN.

To test the effect on CIN under varying MAP9-knockdown efficiencies in vitro, we transfected the 1CT cells with different amounts of MAP9 siRNA. Karyotype analysis revealed no changes in chromosome abnormalities with 20% MAP9 knockdown (Supplementary Fig. S6C). Approximately 50% knockdown led to a moderate increase in aneuploidy, and to a greater extent in cells with >80% knockdown of MAP9 (Supplementary Fig. S6C). Thus, approximately 50% loss of MAP9 could initiate genome abnormalities in normal colon epithelial cells.

We further examined the effects of the loss of Map9 on CIN by using MEFs derived from Map9^Δ/+, Map9^Δ/Δ, and WT (Map9^+/+) mice (Supplementary Fig. S7A–S7C). Conventional karyotyping showed that a large percentage of Map9^Δ/+ and Map9^Δ/Δ MEFs were aneuploid (70%–80%; Supplementary Fig. S7D; Supplementary Table S3). SKY data confirmed that the percentage of Map9^Δ/Δ MEFs with numerical aberrations was almost three times higher than that of WT MEFs (Fig. 6A–C). Increased CIN induced by knockout of Map9 was also evidenced by elevated rates of lagging chromosomes in anaphase of Map9^Δ/+ and Map9^Δ/Δ MEFs (Fig. 6D and E). Moreover, Map9^Δ/+ and Map9^Δ/Δ MEFs showed a significant increase in the percentage of cells with detached centrosomes and PCM fragmentation compared with WT MEFs (Fig. 6D and E). Of note, MEF cells appear to be more sensitive to effects of Map9 loss compared with intestinal epithelial cells. The heterozygous knockout led to striking changes on CIN and spindle pole stability, which nearly matches homozygous knockout in MEFs. We reasoned that this may be attributed to differential response of MEF cells from intestinal epithelial cells to Map9 loss, for example, differences in DNA damage repair ability. Collectively, these findings indicate that loss of Map9 is sufficient to induce chromosomal abnormalities in MEF cells.

To evaluate the role of Map9 in the context of loss of p53 function, we further investigated the effect of Map9 deficiency on CIN in splenocytes from Map9^Δ/+, Map9^Δ/Δ, p53^Δ/+, Map9^Δ/+p53^Δ/+, Map9^Δ/Δp53^Δ/+, and WT mice at 3–4 months of age. Map9 expression was inversely correlated with aneuploidy rate in mouse splenocytes (Fig. 6F; Supplementary Table S3). The aneuploidy frequency of Map9^Δ/Δp53^Δ/+ compound mutant mice (58%) was significantly higher than either Map9^Δ/Δ (39%) or p53^Δ/+ (23%) single mutant mice (Fig. 6F). The same trend was observed in Map9^Δ/+p53^Δ/+ mice (Fig. 6F), suggesting the loss of p53 function exacerbates CIN induced by Map9 knockout. We observed an aneuploidy frequency of 17% in WT splenocytes, similar to that reported by others (23–29).

CIN has been shown to drive tumorigenic phenotypes such as anchorage-independent growth and metastasis by selectively eliminating or acquiring chromosome segments that harbor the key regulator genes (30, 31). Protumorigenic effect of CIN is a long-term effect, as it requires adaptation to aneuploidy and subsequent tumor evolution (32). We thus first examined the effect of long-term MAP9 silencing (>50 days) on 1CT and 2CT cells. Anchorage-independent growth is a hallmark of cellular transformation (33). Although parental 1CT and 2CT cells did not grow in soft agar (33), multiple colonies were formed in long-term MAP9-knockdown 1CT and 2CT cells (Fig. 7A). Matrigel cell invasion assays also showed that long-term MAP9 knockdown enhanced invasiveness of 1CT and 2CT cells (Fig. 7B). In contrast, short-term MAP9 silencing (3–4 days) failed to elicit neither of these effects (Supplementary Fig. S8A and S8B), suggesting that MAP9 loss led to transformation of normal human colon epithelial cells via CIN. Nevertheless, our data do not exclude potential contributions from CIN-independent effect of MAP9.

We utilized the Human Cancer PathwayFinder RT² Profiler PCR Array to profile the expression of 84 genes involved in transformation and tumorigenesis in long-term MAP9-knockdown 1CT cells. The data demonstrated that long-term knockdown of MAP9 altered gene expression involved in DNA damage, apoptosis, cell cycle, cellular senescence, and angiogenesis pathways (Supplementary Fig. S8C; Supplementary Table S4). Gene set enrichment analysis of RNA sequencing data from TCGA colorectal cancer cohort revealed that genes promoting cell survival and growth such as telomere extension, mRNA capping, and activation of prereplicative complex and DNA repair pathways are upregulated in MAP9-low (<25th percentile) compared with MAP9-high tumors (>75th percentile; Supplementary Fig. S8D; Supplementary Table S5), consistent with the expression profiles in 1CT cells with long-term knockdown of MAP9. Taken together, MAP9 loss in normal colon epithelial induces CIN and oncogenic signaling, thereby causing cellular transformation.

To elucidate the effect of Map9 knockout on the colorectal cancer genome, we performed whole genome sequencing (WGS) on three colon tumors from Map9^Δ/ΔApc^Min/+ mice and two colon tumors from Apc^Min/+ mice. We observed increased trend of somatic copy-number variation (CNV) sizes in colon tumors from Map9^Δ/ΔApc^Min/+ mice compared with Apc^Min/+ mice (Fig. 7C and D; Supplementary Table S6); however, there was no significant difference due to the limited number of mice assessed and the big CNV size variations (Supplementary Fig. S8E). More importantly, colon tumors from Map9^Δ/ΔApc^Min/+ mice had significantly more somatic structural variations (SV; Fig. 7E and F; Supplementary Table S6). The number of interchromosomal translocations (CTX) in Map9^Δ/ΔApc^Min/+ tumors was twice as much as Apc^Min/+ tumors, whereas intrachromosomal translocations (ITX) were only found in Map9^Δ/ΔApc^Min/+ tumors (Fig. 7E). However, no significant difference was observed for somatic mutations (Fig. 7G; Supplementary Table S6), suggesting the effect of MAP9 is specific to chromosomal abnormalities. Collectively, Map9 knockout in vivo generates CIN, which in turn drives colorectal tumorigenesis.

We evaluated the clinical impact of MAP9 in 141 patients with colorectal cancer (Supplementary Table S7). Low MAP9 mRNA expression (tumor/normal ratio) was associated with poorer survival of patients with colorectal cancer after adjustment for age, gender, and tumor–node–metastasis (TNM) stage (HR, 2.36; 95% confidence interval, 1.05–5.34; P = 0.04). Kaplan–Meier survival curves showed that overall survival of patients with colorectal cancer with low MAP9 mRNA expression was significantly shorter than that of other patients (P = 0.03, log-rank test; Fig. 7H). After stratification by TNM stage, patients with colorectal cancer with low MAP9 mRNA expression had significantly shorter survival in early stage (I/II/III; P = 0.02), but not advanced stage (IV) colorectal cancer (Supplementary Fig. S8F).

A majority of sporadic colorectal cancer arise through the CIN pathway (65%–70%); however, its underlying molecular mechanism remains poorly understood. In this study, we identified a novel spindle-associated protein MAP9 that is frequently inactivated in colorectal cancer, inferring that MAP9 may have tumor-suppressive function. We showed that MAP9 expression is essential for maintaining chromosomal integrity, and its loss drives CIN in the colon epithelium and induces cellular transformation. Deletion of Map9 in the colonic epithelium triggers spontaneous colorectal cancer and accelerates carcinogen- or Apc^Min/+-driven colorectal cancer, implying that the silencing of MAP9 is a driver event in CIN-associated colorectal cancer development.

Despite the high prevalence of CIN in colorectal cancer, mutations in mitosis-associated genes are rare. Here, we uncovered that inappropriate silencing of MAP9 may be an important mechanism by which CIN can be triggered in colorectal cancer. MAP9 is downregulated in a majority of primary colorectal cancer (∼70%). MAP9 knockdown in human normal colon epithelial cells (1CT/2CT) and MEFs induced severe chromosome segregation error, aneuploidy, and a CIN phenotype. Long-term MAP9 silencing in 1CT/2CT cells led to acquisition of anchorage-independent growth capacity and enhanced invasiveness, key hallmarks of cellular transformation. Our in vitro data were recapitulated in vivo using Map9-knockout mice. Intestinal-specific knockout of Map9 in mice resulted in the spontaneous formation of colonic dysplasia and colorectal cancer. Moreover, Map9 loss also significantly increased tumor incidence, tumor number, and tumor load in genetically (Apc^Min/+) and carcinogen (AOM)-driven colorectal cancer mouse models. WGS revealed an increased CNVs and chromosomal translocations in colon tumors from Map9-knockout Apc^Min/+ mice compared with those from control Apc^Min/+ mice, consistent with in vitro data. These transgenic mice results demonstrated that Map9 inactivation promoted colon tumorigenesis in vivo. Given the gatekeeper role of MAP9 in chromosomal integrity, the transcriptional silence of MAP9 likely has a significant role-promoting colorectal cancer through induction of CIN. Consistent with tumor-suppressive function of MAP9 in colorectal cancer, patients with colorectal cancer with low MAP9 expression were associated with poorer outcome. Hence, our data provide solid evidence that Map9 inactivation plays a causal role in colorectal tumorigenesis through the induction of CIN.

Around 50% MAP9 loss could induce genome instability in vitro. On the other hand, tumor-promoting effect of heterozygous versus homozygous Map9 knockout varies according to the model utilized in vivo. In AOM or spontaneous model, heterozygous Map9 knockout could increase tumor burden or induce transformation; whereas only homozygous knockout of Map9 enhanced tumorigenesis in Apc^Min/+ mice. This discrepancy might be attributed to the short duration of the study (∼3 months) compared with the other two (>6 months). In our cohort of patients with colorectal cancer, we analyzed the distribution of MAP9 expression levels by tumor/normal ratio, revealing that more than 40% of patients have at least approximately 50% loss of MAP9 expression (Supplementary Fig. S8G). This indicates that a large number of human colorectal cancers harbor MAP9 loss that may reach the threshold to cause CIN.

MAP9 belongs to MAP family that regulates microtubule dynamics (6). We observed that MAP9 is localized to spindle microtubules during mitosis, suggesting that it might play a role in the mitosis processes. Indeed, MAP9 knockdown caused shortened spindles in metaphase that were extremely unstable and repeatedly collapsed and reassembled. Conversely, the ectopic expression of MAP9 led to spindle lengthening. We further demonstrated that disrupted spindle dynamics caused by MAP9 deficiency initiates CIN via the following pathways: (i) spindle pole defocusing and PCM fragmentation; (ii) chromosomes lagging; and (iii) failure to correct K-fiber misattachment. Mechanistically, MAP9 tightly surrounds γ-TuRC to protect K-fiber from depolymerization at their minus end at spindle poles. This is evidenced by shortened and curved K-fibers in MAP9-knockdown cells, whereas MAP9 overexpression has a stabilizing effect on microtubules. Consistent with our data, other MAP proteins have a wide range of functions that revolve around microtubules, including stabilization of microtubules by inducing polymerization or inhibiting depolymerization, destabilization of microtubules by preventing their assembly, and mediating interactions of microtubules with other proteins (34). This in turn controls spindle length during mitosis, which is critical for proper segregation of chromosomes (35). Taken together, our results suggested that MAP9 acts as a microtubule stabilizer and contributed to spindle stability by interacting and stabilizing spindle microtubules.

In summary, we identified a novel tumor suppressor, MAP9, in colorectal cancer. Loss of MAP9 led to compromised spindle stability and proper chromosomal segregation, contributing to CIN, aneuploidy, and ultimately colorectal tumorigenesis (Supplementary Fig. S9). This study established the molecular mechanisms by which this key CIN regulator normally maintains chromosomal stability and prevents cancer.

No potential conflicts of interest were disclosed.

Conception and design: S. Wang, J. Huang, J. Yu

Development of methodology: S. Wang, J. Huang, J. Zhai, W. Deng, Y. Zhang, G. Wang, J.W. Shay, J. Yu

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): S. Wang, J. Huang, L. Zhao, J. Zhai, Y. Zhou, W. Deng, S. Gao, X.Y. Guan, J. Yu

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): S. Wang, J. Huang, C.C. Wong, W. Deng, Y. Zeng, H.H. He

Writing, review, and/or revision of the manuscript: S. Wang, J. Huang, C.C. Wong, H.H. He, J.W. Shay, J. Yu

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): W. Deng, H. Wei, S.H. Wong, J.W. Shay

Study supervision: J. Yu

Other (animal experiment): C. Li

This study was supported by HMRF Hong Kong (03140856), National Key R&D Program of China (2017YFE0190700), Science and Technology Program Grant Shenzhen (JCYJ20170413161534162), Science and Technology Program Grant Shenzhen (JCYJ20170413161534162), RGC-GRF Hong Kong (14111216, 14163817), Vice-Chancellor's Discretionary Fund CUHK, CUHK direct grant, and Shenzhen Virtual University Park Support Scheme to CUHK Shenzhen Research Institute.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.