Truncating mutation in the autophagy gene UVRAG confers oncogenic properties and chemosensitivity in colorectal cancers

Autophagy-related factors are implicated in metabolic adaptation and cancer metastasis. However, the role of autophagy factors in cancer progression and their effect in treatment response remain largely elusive. Recent studies have shown that UVRAG, a key autophagic tumour suppressor, is mutated in common human cancers. Here we demonstrate that the cancer-related UVRAG frameshift (FS), which does not result in a null mutation, is expressed as a truncated UVRAGFS in colorectal cancer (CRC) with microsatellite instability (MSI), and promotes tumorigenesis. UVRAGFS abrogates the normal functions of UVRAG, including autophagy, in a dominant-negative manner. Furthermore, expression of UVRAGFS can trigger CRC metastatic spread through Rac1 activation and epithelial-to-mesenchymal transition, independently of autophagy. Interestingly, UVRAGFS expression renders cells more sensitive to standard chemotherapy regimen due to a DNA repair defect. These results identify UVRAG as a new MSI target gene and provide a mechanism for UVRAG participation in CRC pathogenesis and treatment response.

C olorectal cancer (CRC) remains one of the most widespread malignancies worldwide 1 . Approximately 15% of sporadic CRC and 90% of Lynch syndrome (hereditary nonpolyposis colorectal cancer) exhibit a microsatellite instability (MSI) phenotype, caused by a deficiency in DNA mismatch repair (MMR) that progresses with a high rate of insertions/deletions to repetitive DNA sequences, termed microsatellites 2 . Increasing evidence suggests that MMR deficiency per se is not sufficient to drive cell transformation and tumorigenesis, but that microsatellite mutations in a limited number of target genes might be positively selected during tumour development and underlie MSI-associated pathogenesis and treatment response 3,4 . Frameshift (FS) mutations of several autophagy-related genes, including Atg2b, Atg5, Atg9b, Atg12 and UVRAG (ultraviolet irradiation resistance-associated gene) [5][6][7] , were recently reported in gastric cancer and CRC with MSI. Nevertheless, the functional consequences and key molecular events downstream of these mutations have not been extensively investigated.
Our previous studies have established UVRAG as a critical regulator of intracellular membrane trafficking, including autophagy and chromosomal stability 6,[8][9][10][11][12][13][14][15][16] . UVRAG contains four functional domains, that is, a proline-rich domain, a lipidbinding C2 domain, a Beclin1-binding coiled-coil domain (CCD) and a C-terminal domain presumed to be unstructured and involved in centrosome integrity and DNA damage repair (Supplementary Fig. 1a) 12,17 . Importantly, all the different activities of UVRAG are functionally independent, suggesting biological interaction and coordinated regulation of the different processes under diverse environmental cues. Although most cellular studies to date have considered UVRAG as a tumour suppressor in human cancers 18 , the genetic linkage of UVRAG mutations in major tumour types and the significance of these mutations in tumour pathogenesis remains less understood.
Here we show that MSI CRCs with the FS mutation in UVRAG express a truncated UVRAG protein, referred to here as UVRAG FS . In addition to losing the wild-type (WT) UVRAG functions, this nonsense mutant acts as a dominant-negative mutant and contributes to the oncogenesis and tumour metastasis of CRC, likely by antagonizing the activity of UVRAG WT as a tumour suppressor. UVRAG FS expression also increases the sensitivity to anticancer agents such as 5-fluorouracil (5-FU), oxaliplatin and irinotecan, routinely prescribed as adjuvant therapies for CRC patients. Our data thus identified the underlying pathogenic mechanisms beyond autophagy that are associated with UVRAG FS -positive cancers and suggest that expression of UVRAG FS might also be a predictive factor for chemotherapy response.

Results
UVRAG A 10 DNA microsatellite mutation in MSI CRC. The human UVRAG gene contains a tract of A 10 mononucleotide repeats in exon 8, spanning codons 234-237 (5 0 -AAA AAA AAA AGT-3 0 ; Supplementary Fig. 1a,b). Using seven MSI þ CRC cell lines (HCT15, HCT116, KM12, LIM2405, LS180, RKO and SW48) and genomic sequencing, we confirmed, as reported previously 6,7,16 , the heterozygous FS deletion of one nucleotide (A) in the UVRAG A 10 -coding repeat in most tested MSI þ CRC cells, with the exception of HCT15 and SW48. In contrast, MSS (microsatellite stable) cells, including COLO205, HCC2998, HT29, SW480 and SW620, contained only WT coding repeats (Fig. 1a). The FS mutation was predicted to produce a premature stop codon and therefore a truncated UVRAG 7 (referred here as UVRAG FS ; Supplementary Fig. 1a,b). To assess whether this mutation is indeed expressed in MSI cells, we generated an antibody specifically recognizing UVRAG FS , but not UVRAG WT , using the FS-derived neopeptide ( 234 KKKVNACS 241 ) as antigen ( Supplementary Fig. 1b,c). UVRAG FS expression was detected in all MSI cell lines carrying the FS mutation, but not in MSI or MSS cells that are WT for UVRAG (Fig. 1b). Notably, the overall expression of UVRAG WT was diminished in MSI cells with the FS mutation (Fig. 1b), and the levels of UVRAG FS were inversely correlated with the expression of UVRAG WT in all tested cell lines (Fig. 1c). This was consistent with the UVRAG expression profile from the CRC cell lines of the NCI-60 panel 19 . Therein, a significant reduction of UVRAG WT expression was detected in UVRAG FS -positive KM12 and HCT116 CRC cells compared with other CRC cells without UVRAG FS (Supplementary Fig. 1d). In addition, the UVRAG FS mutation was present in one of the four analysed cases of human primary CRC with MSI (fourth column in Fig. 1d), but not in primary MSS CRC or in normal colorectal mucosa (Fig. 1d, Supplementary Table 1). This is in line with a previous report 2a that evaluated the mutation frequencies in 137 genes in MSI cancers, revealing the high frequency of the A 10 UVRAG FS mutation that was found in 33% CRC, 8% endometrial and 7.8% gastric cancers with MSI ( Supplementary  Fig. 1e). Whole-genome sequencing analysis of a large cohort of gastric cancers (Pfizer and UHK; n ¼ 100) also confirmed the presence of the UVRAG FS mutation in MSI gastric cancer (40%) 20 . Collectively, these results indicate that the frameshift UVRAG mutation is likely selected and is expressed as a truncated UVRAG protein in MSI tumours.
Oncogenic property of the UVRAG FS mutation. To probe whether the UVRAG FS mutant represents a mere loss of WT function 11 as occurs with most other tumour suppressors, or imparts oncogenic properties, we established MSS SW480 and MSI HCT116 cell lines stably expressing Flag-tagged UVRAG WT and UVRAG FS at equivalent levels ( Supplementary Fig. 2a,d). UVRAG FS -transduced cells showed increased proliferation and enhanced anchorage-independent growth in soft agar ( Supplementary Fig. 2a-e), independently of the tissue of origin ( Supplementary Fig. 2f,g). Subcutaneous transplantation in athymic nude mice of UVRAG FS SW480 cells resulted in tumour formation with accelerated kinetics ( Supplementary  Fig. 2c). To further test whether expression of UVRAG FS is sufficient to transform noncancerous cells, we used NIH3T3 mouse embryonic fibroblasts stably expressing UVRAG WT or UVRAG FS (Fig. 2a). Compared with control (3T3.Vec), UVRAG FS -3T3 cells had elevated growth rate, formed larger colonies when plated at low density and induced anchorageindependent growth, whereas UVRAG WT exerted the opposite effects ( Fig. 2a-c). The tumour growth rate and mean tumour volume were drastically increased when 3T3-UVRAG FS cells were injected into nude mice (Fig. 2d). Immunohistological analyses of tumour xenografts showed UVRAG FS expression and enhanced mitotic index and number of Ki67 þ (proliferating) cells in UVRAG FS tumours (Fig. 2e). CRC primary tumours with the FS mutation also had increased Ki67 staining (Fig. 1d). Altogether, these data indicate a strong association of the cancer-derived UVRAG FS with a tumorigenic phenotype.
Dominant-negative effect of UVRAG FS on autophagy activation. UVRAG FS retains the N-terminal proline-rich and C2 domains, and the partial CCD required for Beclin1-mediated autophagy ( Supplementary Fig. 1a) 10,12,[21][22][23] . To determine whether UVRAG FS retained its autophagy activity, we measured the subcellular distribution of the autophagy marker greenfluorescent protein (GFP)-LC3 and the levels of the autophagosome-associated lipidated LC3 (LC3-II) 24,25 in noncancerous NIH3T3 cells. As shown previously 10,12,26 , UVRAG WT or rapamycin markedly promoted autophagy, as evidenced by increased GFP-LC3 puncta per cell, increased LC3-II conversion and increased response to the late-stage autophagy inhibitor Bafilomycin A 1 (Fig. 3a,b). In sharp contrast, UVRAG FS did not demonstrate any proautophagic activity. Furthermore, UVRAG WT autophagy-promoting activity was   Fig. 3a). UVRAG interacts with Beclin1 through their respective CCD, resulting in activation of Beclin1-associated Vps34 kinase 27 . On UVRAG FS expression, the endogenous association between UVRAG WT and Beclin1 was diminished, and UVRAG FS was able to sequester the Beclin1 and UVRAG proteins in vivo, in line with its dominant-negative effect ( Supplementary Fig. 3b, Fig. 3c). Accordingly, Vps34 enzymatic activity was significantly reduced in UVRAG FS cells (Fig. 3d), as illustrated by decreased punctate staining of the Vps34 kinase product, phosphatidylinositol 3-phosphate 28 . Impaired autophagy was also observed in vivo in NIH3T3 tumour xenografts expressing UVRAG FS (Fig. 2e), showing increased levels of p62, an autophagic substrate 29 . To explore whether autophagy inhibition underlies UVRAG FS -mediated oncogenesis, we examined the transforming effect of UVRAG FS in autophagy-null Atg5-deficient MEFs 29 .   UVRAG FS promoted cell proliferation (Fig. 3e) and colony growth in soft agar ( Fig. 3f-h), irrespective of the autophagy status. These data support a direct role of UVRAG FS in promoting tumorigenesis independently of autophagy.
UVRAG FS induces chromosomal instability and centrosome amplification. Because the role of UVRAG in cancer has been linked to its ability to maintain chromosomal stability 17   stability in genetically stable mouse embryonic stem cells. Spectral karyotyping analysis showed that, unlike control cells that were mostly diploid, UVRAG FS -embryonic stem cells were highly heterogeneous with respect to both structural and numerical aberrations as compared with the vector control ( Fig. 4a, Supplementary Fig. 4a) with a greater than sevenfold increase in aneuploidy in UVRAG FS cells ( Supplementary Fig. 4b).
These results indicate that UVRAG FS elicits severe chromosomal instability and aneuploidy. To validate this, we analysed the Pfizer and UHK cohort 20 of gastric cancers, and observed significantly enhanced chromosomal rearrangement in UVRAG FS MSI gastric cancers as compared with UVRAG WT MSI gastric cancers (Fig. 4b). In fact, UVRAG FS gastric cancers had substantially more protein-altering mutations and singlenucleotide variants than UVRAG WT MSI and MSS gastric cancers ( Supplementary Fig. 4c). Moreover, the FS mutation appeared to be more frequent in gastric cases with advanced tumour, node, metastasis stage ( Supplementary Fig. 4d). Thus, UVRAG FS may predispose MSI cancers to increased genetic instability and cancer progression. UVRAG WT has been shown to associate with the centrosome protein CEP63 (ref. 17), contributing to chromosomal stability by preventing centrosome overduplication 17 . UVRAG FS expression in SW480 cells was sufficient to induce a marked increase in the incidence and degree of centrosome amplification compared with control ( Fig. 4c). Consistent with the consensus that centrosome amplification causes erroneous chromosomal segregation 30 , we detected spindle malformation, chromosomal missegregation and prolonged mitosis in UVRAG FS clones, whereas UVRAG WT clones behaved in the opposite manner (Fig. 4d, Supplementary  Fig. 5a). Unlike WT, UVRAG FS was unable to associate with CEP63 ( Fig. 4e), failing to colocalize with CEP63 and the centrosome marker, g-Tubulin (Fig. 4f). UVRAG FS disrupted UVRAG WT -CEP63 interaction (Fig. 4g) and displaced UVRAG from the centrosome in a dominant-negative manner ( Supplementary Fig. 5b). These results indicate that centrosome amplification induced by UVRAG FS may play a role in UVRAG FS -associated chromosomal aneuploidies.

UVRAG FS promotes cell invasion and metastasis outgrowth.
Centrosome amplification per se has been shown to promote cell invasion through inappropriate microtubule nucleation and Rac-1 activation 31 , a small GTPase important for the control of cell invasiveness and metastasis 32,33 . Indeed, pull-down assay in UVRAG FS SW480 cells detected a more than twofold Rac1 activation, which could be blocked by Taxol, but not by the autophagy inhibitor chloroquine or the anticancer reagent 5-FU (Fig. 5a), indicating a requirement for dynamic microtubules. Consistent with increased Rac1 activation, UVRAG FS enhanced the cell motility of SW480 cells in a wound-healing assay, which was inhibited by Taxol (Fig. 5b). It also enhanced HCT116 cell migration through a collagen matrix, whereas UVRAG WT exerted an inhibitory effect ( Supplementary Fig. 6a,b). Spleen injection of non-metastatic SW480 cells expressing UVRAG FS into nude mice resulted in a higher incidence of liver metastasis and a greater number of colonization in the lungs, kidney and peritoneum, whereas no colonization was found in the control group (Fig. 5c,d, Supplementary Fig. 6c). UVRAG FS -induced tumour metastases were confirmed in an independent mouse metastasis model with SW480 cells expressing GFP-UVRAG FS , as determined by bioluminescence imaging of metastatic lesions ( Supplementary Fig. 6d). These results indicate that UVRAG FS enhances the metastatic capacity of CRC cells.
Autophagy has been postulated to be exploited by metastatic tumours to survive unfavourable conditions 34 . Nevertheless, UVRAG FS -metastatic tumours displayed higher levels of p62 than primary tumours, indicative of suppressed autophagy (Fig. 5e). Moreover, UVRAG FS metastatic tumours exhibited decreased apoptosis, as shown by decreased caspase 3 activation (Fig. 5e). Hence, in this context, autophagy is not the driving mechanism for metastatic colonization in CRC. Nevertheless, we observed other pathological differences that may account for increased metastasis on UVRAG FS expression. The colonized CRC tumours had reduced levels of the epithelial cell marker E-cadherin but increased levels of the mesenchymal markers, N-cadherin and vimentin (Fig. 5e), suggesting an induction of epithelial-mesenchymal transition (EMT) in the process of colonization. Indeed, expression of UVRAG FS in SW480 cells downregulated E-cadherin and upregulated N-cadherin and vimentin, whereas expression of UVRAG WT had the opposite effect (Fig. 5f). Importantly, UVRAG FS -associated EMT was efficiently reverted by Taxol without affecting Taxol-induced cell death (Fig. 5f, Supplementary Fig. 6f). Consistent with our in vitro observations, the primary MSI colon tumour with UVRAG FS exhibited elevated expression of N-cadherin and vimentin, along with significant reduction in E-cadherin levels, which were not detected in tumours with UVRAG WT (Supplementary Fig. 6e).
These results indicate that UVRAG FS expression, which triggers centrosome amplification and Rac1 activation, can activate the EMT program and promote cell invasion and tumour metastasis.
UVRAG FS affects CRC response to chemotherapy. We next investigated the possible clinical relevance of UVRAG FS by testing the response of CRC to 5-FU chemotherapy, the first-line treatment for CRC patients, using a tumour xenograft model. Surprisingly, UVRAG FS expression significantly increased tumour sensitivity to 5-FU treatment with an approximate 10-fold reduction in tumour volumes after a 4-week administration of 5-FU (Fig. 6a), compared with a less than twofold reduction in the control group (Fig. 6a-c). Histological analyses revealed a significant reduction in cell proliferation and an increase in the number of cells undergoing apoptosis in 5-FU-treated UVRAG FS tumours, in concordance with induced tumour shrinkage (Fig. 6d). In addition, UVRAG FS expression in CRC cells markedly increased their sensitivity to other DNA-based cytotoxic anticancer agents, including oxaliplatin and irinotecan, as shown by reduced rates of clonogenic survival, whereas UVRAG WT cells were resistant to the drugs (Fig. 6e). To examine the unexpected role of UVRAG FS in tumour chemosensitivity, we measured the levels of g-H2AX, a sensitive marker of double strand breaks (DSBs) 35 , and observed that UVRAG FS SW480-tumours accumulated higher levels of g-H2AX than the controls, which further increased with 5-FU that produces DNA strand breaks (Fig. 6d). Consistent with our observation in xenograft tumours, UVRAG FS expression resulted in a significant increase of g-H2AX foci and levels in SW480 CRC cells ( Supplementary Fig. 7a,b). Furthermore, the overall levels of g-H2AX were higher in MSI CRC cell lines expressing UVRAG FS compared with the WT counterparts, and likewise, were significantly different between UVRAG FS -positive and -negative primary tumours (Fig. 1b,d). Adding UVRAG WT to UVRAG FSpositive HCT116 and RKO cells at different doses clearly suppressed the levels of DSBs ( Supplementary Fig. 7c), highlighting a direct involvement of UVRAG FS in genetic stability. To determine whether the observed accumulation of DSB in UVRAG FS cells reflects impaired DNA repair, we measured unrepaired DSBs after ionizing radiation (IR) using the comet assay. We found that IR induced comparable levels of DNA damage in vector, UVRAG WT  post-IR in Fig. 7a). However, a high persistence of comet tails was observed 24 h post-irradiation in UVRAG FS cells, whereas UVRAG WT cells have repaired most of the damaged DNA.
These data indicate that UVRAG FS disrupts the rapid repair process of DSBs. The inhibitory effect of UVRAG FS on DSB repair was also detected in the autophagy-competent Atg3 þ / þ and the autophagy-null Atg3 À / À cells ( Supplementary Fig. 7d), suggesting minimal participation of autophagy in the elevated DNA damage induced by UVRAG FS expression.
UVRAG FS is defective in the repair of DNA damage. We then asked whether UVRAG FS -associated DNA damage results from suppression of UVRAG WT function, which is known to promote DSB repair by NHEJ (non-homologous end joining) through interaction with the Ku70/Ku80/DNA-PKcs complex 17 . Unlike with UVRAG WT , no physical interactions between UVRAG FS and DNA-PK proteins could be detected ( Supplementary  Fig. 7e). Moreover, UVRAG FS failed to translocate to sites of laser-induced DNA damage stripes containing g-H2AX, whereas UVRAG WT was enriched at the damaged sites of DSBs ( Supplementary Fig. 7f). As expected, ectopic expression of UVRAG FS blocked UVRAG-Ku70/Ku80 interaction, and disturbed Ku/DNA-PKcs complex formation after IR, concomitant with increased sequestration of UVRAG WT , again highlighting the dominant-negative effect of the FS mutation (Fig. 7b). To further establish a link between UVRAG FS and the DNA-damaging phenotype observed, we evaluated the DNA repair capacity in UVRAG FS cells, using a NHEJ repair reporter, the EJ5-GFP system 36 . Expression of UVRAG FS alone markedly reduced the rate of NHEJ repair by over 50%, whereas it had no discernable effect on DNA homologous recombination repair (Fig. 7c). Treating cells with Nu7441, a specific inhibitor of DNA-PK 37 , abolished the effect of UVRAG FS (Supplementary Fig. 7g). These results indicate that UVRAG FS -induced DNA damage is dependent on the inactivation of DNA-PK-mediated NHEJ, which renders tumour cells more sensitive to DNA-damaging chemotherapy.

Discussion
Microsatellite instability as a result of MMR deficiency has been widely observed in human CRC. However, little is known of the biological consequences and pathogenic mechanisms associated with the selective gene targeting by MSI. Herein, we demonstrate that the autophagic tumour suppressor UVRAG represents a new bona fide MSI target gene in CRC and, likely, other MSI-related tumours, and that the truncating mutation in UVRAG enhances cellular transformation and penetrance of CRC tumour by interfering with the tumour-suppressing functions of UVRAG WT in a dominant-negative manner. Furthermore, mutated UVRAG alleles sensitize CRC to DNA damage-inducing treatment, making the UVRAG FS genotype a possible predictive factor for the response to chemotherapy treatment.
In this study, we found that the heterozygous deletion of the UVRAG A 10 exonic DNA repeat resulted in the expression of a truncated protein using an antibody specifically recognizing UVRAG FS , and that it influences the expression and function of UVRAG WT in a series of CRC cell lines and primary CRCs. Contrary to our findings, a previous study 7 showed by immunoblotting that the levels of UVRAG WT appeared to be unaffected by the occurrence of the UVRAG FS mutation in three MSI CRC cell lines carrying the FS mutation (HCT116, LoVo and RKO), two of which having also been used in our study (Fig. 1b). While it is difficult to explain the discrepancy between this published work and ours, it might be due to differences in experimental design and/or to different sources or passage numbers of CRC cell lines used in both studies. Nonetheless, our results are consistent with the gene expression data retrieved from a GeneChip analysis of NCI-60 cancer cell lines from TSRI (The Scripps Research Institute; data are accessible at BioGPS: http://biogps.org), correlating reduced UVRAG WT expression in  a subset of CRC cell lines with the UVRAG FS mutation. Taken as a whole, our findings and those of others suggest that inactivation of UVRAG is selected for during the progression of colorectal tumours, and that UVRAG WT plays a suppressor role in colorectal tumorigenesis. Previous studies have indicated that autophagy protects genomic integrity presumably by removing aged or damaged proteins and organelles [38][39][40] . We observed a significant reduction of autophagy by UVRAG FS in CRC cells and primary tumours, which was even greater in the metastases. Of note, a previous study 7 argued that UVRAG FS lost Beclin1-binding activity due to the frameshift truncation. However, we found that even though UVRAG FS lost more than 50% of CCD of UVRAG WT , it still retains a small alpha-helix structure in the CCD and remains competent for UVRAG and Beclin1 interaction, thereby neutralizing their proautophagic effect in a dose-dependent manner. However, autophagy loss could not prevent the transformed phenotype induced by UVRAG FS , indicative of an autophagy-independent oncogenic mechanism associated with UVRAG FS , as previously suggested 7 .
We found that ectopic expression of UVRAG FS per se in both embryonic stem cells and cancer cells results in extensive centrosome amplification and concomitant aneuploidy. Indeed, this cancer-associated mutated UVRAG, which lacks CEP63binding ability, is more than just a relic of UVRAG inactivation, it instead disturbs the association of endogenous UVRAG WT with CEP63, presumably by displacing endogenous active UVRAG from the centrosome and/or by titrating out an unknown regulator into nonfunctional complexes. This is similar to what occurs with mutations in other tumour suppressors, such as p53. Certain mutated forms of p53 have not only lost their tumoursuppressive function, but have also gained a function as an oncogene 41 . Consistent with a previous study demonstrating that inappropriate microtubule nucleation due to centrosome amplification enables Rac1 activation and promotes cell invasion 31 , we found that UVRAG FS promotes metastatic outgrowth and EMT properties in a Rac1-dependent manner. Our mutational and integrative analyses emphasize the critical role of UVRAG FS and centrosomal stability in the context of metastatic CRC.
Despite increased oncogenic transformation, UVRAG FSexpressing tumours appear to be more responsive to chemotherapy that induces massive DNA damage and replication stress. Unlike UVRAG WT , UVRAG FS cannot translocate to DSB sites and its expression further interferes with a functional complex assembly of DNA-PK, a key effector in the NHEJ pathway. As NHEJ factors are considered as genome caretakers that guarantee genomic integrity through the proper repair of DNA lesions, our data thus provide a potential mechanism by which UVRAG FS elevates the levels of DNA damage via acting on NHEJ repair and sensitizes tumour cells to chemotherapy. Thus, UVRAG FS may represent an important determinant in the treatment response of CRC tumours.
In summary, we have demonstrated that a cancer-derived UVRAG truncated mutant plays a role in oncogenic transformation and tumour metastasis, which explains the selection for its expression in human CRC cell lines and primary tumours with MSI. This mutant impairs UVRAG WT function in autophagy and chromosomal stability. Our findings suggest that UVRAG FS expression contributes to chemosensitivity through direct repression of DNA damage repair and ensuing increased cell death. This regulatory circuit may partially explain the more favourable prognosis in patients with MSI tumours than in those with MSS tumours, as previously noted 42 . It may also have potential relevance for pharmacogenetic selection of MSI cancer patients for adjuvant chemotherapy.
Mutation analysis. Genomic DNA and cDNA from cell lines and primary tumours were amplified by PCR. The primer pair (forward and reverse, respectively) is: 5 0 -ATGTTTTAAGCCATTATTTA-3 0 and 5 0 -CGTTCCAGTTC ATTCTG-3 0 . PCR products from single clones from every sample were sequenced using an automated ABI PRISM 377 automatic DNA sequencer.
For multichannel imaging, fluorescent staining was imaged sequentially in line-interlace modes to eliminate crosstalk between the channels. The step size in the z axis varied from 0.2-0.5 mm to obtain 16 slices per imaged file. All experiments were independently repeated several times. The investigators conducted blind counting for quantification. Values indicate the mean ± s.d. of at least three independent experiments.
Histopathology and immunohistochemistry. Tissue sections from the indicated mouse models were fixed in 10% buffered formalin and embedded in paraffin. Tissue sections were routinely stained with haematoxylin and eosin. For immunohistochemistry staining, tissue slides were deparaffinized in xylene and rehydrated in alcohol. Endogenous peroxidase was blocked with 3% hydrogen peroxide. Antigen retrieval was achieved using a microwave and 10-mM citric sodium buffer (pH 6.0). Sections were then incubated overnight at 4°C with the primary antibody. Antibody binding was detected with Envision Dual Link System-HRP DAB kit (K4065, Dako). Sections were then counterstained with haematoxylin. For negative control, the primary antibody was replaced with the buffer. The mitotic index was quantified by viewing and photographing 10 random high-power field of each tissue section on a Nikon microscope, using a 40 Â objective. For evaluation and scoring of immunohistochemical data, we randomly selected 10 fields within the tumour area under high-power magnification ( Â 400) for evaluation. The investigators conducted blind counting for each quantification-related study.
Immunoblotting and immunoprecipitation. For immunoblotting, polypeptides were resolved by SDS-PAGE and transferred to a PVDF membrane (Bio-Rad). Membranes were blocked with 5% non-fat dry milk, and probed with the indicated antibodies. HRP-conjugated goat secondary antibodies were used (1:10,000, Invitrogen). Immunodetection was achieved with the Hyglo chemiluminescence reagent (Denville Scientific), and detected by a Fuji ECL machine (LAS-3000). For co-immunoprecipitation, cells were lysed in 1% NP40 lysis buffer (25 mM Tris pH 7.5; 300 mM NaCl, 1 mM EDTA, 1% NP40), supplemented with a complete protease inhibitor cocktail (Roche). After preclearing with protein A/G agarose beads for 1 hr at 4°C, whole-cell lysates were used for immunoprecipitation with the indicated antibodies. Generally, 1-4 mg commercial antibody was added to cell lysate, which was incubated at 4°C for 8-12 h. After addition of protein A/G agarose beads, incubation was continued for another 2 h. Immunoprecipitates were extensively washed with NP40 lysis buffer and eluted with SDS-PAGE loading buffer by boiling for 5 min before resolution by SDS-PAGE.
Soft agar anchorage-independent growth assay. To evaluate anchorageindependent colony formation, engineered cells (10 4 ) were suspended in complete medium containing 0.3% Nobel agar (Difco) supplemented with 2 mg ml À 1 puromycin and plated in 6-well plates over a basal layer of 0.5% agar in complete medium. Colonies were scored 21 days after plating and were photographed by phase-contrast microscopy. Images were captured with the QCapture software program. Clonogenicity was determined in triplicate experiments.
In vitro wound-healing assay. The cell invasive activity was determined using the wound-healing assay 45 . Briefly, cells (2.5 Â 10 5 ) were seeded in 12-well slide chambers and grown into a 100% confluent monolayer culture. The confluent cell monolayer was scratched with a pipette tip, followed by media replacement. After 24 h, the width of the mean wound distance was calculated using software connected to Nikon Eclipse digital inverted microscope. To evaluate the 'wound closure', 10 randomly selected points along each wound were marked, and the horizontal distance the migrating cells travelled into the wound was measured.
In vitro cell migration assays. A Transwell system (Corning, NY, USA) was used to evaluate cell migration. The upper and lower chambers were separated by a polycarbonate membrane with pores of 8-mm coated with fibronectin (BD Biosciences, CA, USA) on the lower surface. Cells (2 Â 10 5 ) suspended in 100 ml serum-free medium were seeded onto the upper chamber, and 800 ml of medium with 10% FBS was added to the lower chamber. After 24-h incubation at 37°C with 5% CO 2 , the medium was removed from the upper chamber. The non-invading cells on the upper side of the chamber were scraped off with a cotton swab. Cells on the bottom side of the membrane were fixed, stained with crystal violet and mounted. The migration activity of cancer cells was determined by counting cells in 10 different viewing fields using a microscope at Â 200 magnification. Each assay was repeated three times.
Clonogenic cell survival assay. The log-phased cells were plated in six-well plates overnight, allowing cells to attach to the plates. After chemotherapy drug treatment (24 h exposure), cells were trypsinized, counted and replated at appropriate dilutions for colony formation. After 10-14 days of incubation, colonies were fixed with methanol/acetic acid (3:1), stained with crystal violet and counted. Plating efficiency was determined for each individual cell line 46 and the surviving fraction (SF) was calculated based on the number of colonies that arose after treatment, expressed in terms of plating efficiency. Each experiment was repeated three times.
In vivo tumorigenicity assay. To measure in vivo tumorigenicity, engineered NIH3T3 and SW480 cells expressing WT or the mutant form of UVRAG (5 Â 10 6 ) were transplanted into the flanks of six-week-old female NCR nude mice (Charles River). Ten mice per cell line were used. Mice were monitored triweekly for the development of tumours, and necropsied after a 3-week observation period. The tumour growth was monitored by measurements of tumour length (L) and width (W) and tumour volume was calculated 47 using the following formula: Volume ¼ 4/3 Â p Â (1/2 width) 2 Â 1/2 length. All animal studies were performed in compliance with the University of Southern California Institutional Animal Care and Use Committee guidelines.
In vivo metastasis assay. A midline incision was made on the left flank, and the spleen was exteriorized. SW480 cells (10 6 cells) were injected into the spleen, after which the wound was closed with surgical metal clips. The mice were sacrificed after 8 weeks, and their spleen, liver, lungs and lymph nodes were removed and examined for tumour metastases. The organ specimens were formalin-fixed and paraffin-embedded for histological analysis. Alternatively, GFP-labelled cells can be tracked using bioluminescence imaging at the end of experiment. Briefly, mice were placed in the induction chamber with 2% isoflurane in oxygen. GFP activity was localized and quantified using an IVIS 200 image system. Images were taken with an excitation wavelength of 465 and emission wavelength ranging from 500 to 540. Imaging processing and analysis, including flat fielding, adaptive background subtraction and spectral unmixing were performed with Living Image 3.0 software.
Autophagy analyses. Quantitative GFP-LC3 light microscopy assay was performed in NIH3T3 cells expressing the WT or FS mutant of UVRAG, then transfected with a GFP-LC3-expressing plasmid 24 . Autophagy was then induced by 100 nM rapamycin (Sigma-Aldrich) for 2-6 h in DMEM containing 1% FBS. For autophagic flux, the rapamycin-treated cells were cultured in DMEM containing 100 nM Bafilomycin A 1 for 2 h. LC3 mobility shift and levels were detected by immunoblotting 12,48 .
Neutral comet assay. Neutral comet assay was performed using the CometAssay kit (Trevigen) following the manufacturer's instruction. Briefly, 10 ml of cell suspension (10 5 cells per ml) was carefully mixed with 90 ml of molten LMAgarose. After solidification, slides were immersed in Lysis Solution at 4°C for 1 h, and equilibrated in chilled neutral electrophoresis buffer for 30 min. Electrophoresis was performed in neutral electrophoresis buffer for 1 h with an electric field of 1 volt cm À 1 . Slides were further treated with DNA Precipitation Solution, followed by 70% ethanol for 30 min each at room temperature. After air drying, cells were stained with SYBR Green (1 mg ml À 1 ) or Propidium Iodide (1 mg ml À 1 ). Comet images were captured using an epifluorescence microscope (Nikon Eclipse C1). To analyse the images, cells were scored into three categories based on tail length (no tails, tail length shorter than 20 mm, tail length longer than 20 mm), and quantified.
Laser microirradiation. Laser microirradiation was done essentially as described before 49 . Cells grown on coverslips were incubated for 24 h in medium containing 10 mM BrdU (Sigma-Aldrich). Laser microirradiation was carried out with a Nikon C1 confocal microscope (Nikon) equipped with a 37°C CO 2 chamber and a diode laser (Melles Griot). DSBs restricted to the laser path were generated across the nuclei in 50 cells per coverslip, using the 100 Â oil objective and 30% of laser power (l ¼ 405 nm) for 100 scans. Cells were then returned to tissue culture incubator at 37°C, fixed 1 h later and analysed by immunofluorescence as described below. Laser-induced DNA damage was visualized with the g-H2AX antibody (Millipore) and the UVRAG antibody (Sigma). Images were taken with a Nikon C1 confocal microscope (Nikon) and Axio Imager 2 (Zeiss).
In vivo DNA DSB Repair. To measure the DNA DSB repair activity, a GFP-based chromosomally integrated reporter was utilized 50 . In brief, the HEK293 cells stably expressing EJ5-GFP reporter were transfected with empty vector or UVRAG FS plasmid. Two days later, a secondary transfection was performed with the same plasmids plus an I-SceI expression vector (pCBASce), together with pmCherry as a transfection indicator. Cells were collected after another 48 h, and analysed by standard flow cytometry. UVRAG expression was verified by western blotting. The repair activity of DSB generated by I-SceI was calculated by the percentage of GFP-positive (repaired) cells in the mCherry-positive cells (transfected).
Chromosomal analysis by SKY. SKY analysis of embryonic stem cells was performed. Briefly, metaphase chromosome were prepared from exponentially growing cells after treatment with colcemid (KaryoMAX, GIBCO) at 0.1 mg ml À 1 for 1 hr (ref. 51). Cells were swollen in prewarmed 0.56% KCl for 10 min at 37°C, then carefully fixed in methanol:acetic acid (3:1) overnight and kept at À 20°C. Metaphase spreads were prepared by dropping cells in the fixative onto chilled Superfrost glass slides (Fisher Scientific) at 25°C and 60% of humidity. After air drying and pepsin digestion, slides were denatured at 80°C for 5 min, hybridization was performed using SKY probe (Applied Spectral Imaging, San Diego) and fluorescence-conjugated secondary antibodies in accordance with the manufacturer's specification. Metaphase images were captured and analysed using a SpectraCube imaging system and software (Applied Spectral Imaging). At least 20 metaphases from each cell line were scored for chromosomal aberration.
Genomic analysis of publically available datasets. All data for UVRAG frameshift mutation in human CRC, gastric, and endometrial cancers with MSI were obtained from SelTarbase (http://www.seltarbase.org/) and primary public sources 6,52-54 . All data for DNA sequence alteration, chromosomal structure variants, and clinical information in gastric cancer were obtained from cBioportal (http://www.cbioportal.org) 55,56 and primary sources 57 . All statistical analyses were carried out using the R software package. Circos plots were carried out using Circos (http://circos.ca/). Statistical analysis. All experiments were independently repeated at least three times. To ensure adequate power and decrease estimation error, we used large sample sizes and multiple independent repeats by independent investigators. Multiple lines of experiments including different quantification methods were used for the consistent and mutually supportive results. The sample size was chosen according to the well-established rule in the literature as well as our ample experience in previous research. Data are presented as the mean ± s.d. Statistical significance was calculated using the Student's t-test or one-way analysis of variance test using GraphPad Prism 5.0 (GraphPad Software, Inc.), unless otherwise stated. A P value of r0.05 was considered statistically significant 58 .