Fusion Kinases Identified by Genomic Analyses of Sporadic Microsatellite Instability −High Colorectal Cancers

Published OnlineFirst October 2, 2018; DOI: 10.1158/1078-0432.CCR-18-1574

Research Article Clinical Cancer Research Fusion Kinases Identiﬁed by Genomic Analyses of Sporadic Microsatellite Instability–High Colorectal Cancers Kazuhito Sato1,2, Masahito Kawazu3, Yoko Yamamoto1, Toshihide Ueno2, Shinya Kojima2, Genta Nagae4, Hiroyuki Abe5, Manabu Soda2, Takafumi Oga2, Shinji Kohsaka3, Eirin Sai3, Yoshihiro Yamashita2, Hisae Iinuma6, Masashi Fukayama5, Hiroyuki Aburatani4, Toshiaki Watanabe7,†, and Hiroyuki Mano2,8

Abstract

Purpose: Colorectal cancers with microsatellite instability– LS-associated) or somatic, Lynch-like mutations in mismatch high (MSI-H) status, due to mismatch repair deﬁciency, are repair genes. MM tumors had more insertions and deletions associated with poor patient outcomes after relapse. We aimed and more recurrent mutations in BRAF and RNF43 than to identify novel therapeutic targets for them. LS-associated or Lynch-like MSI-H tumors. Eleven fusion Experimental Design: We performed MSI analyses of over kinases were exclusively detected in MM MSI-H colorectal 2,800 surgically resected colorectal tumors obtained from cancers lacking oncogenic KRAS/BRAF missense mutations consecutive patients treated in Japan from 1998 through June and were associated with worse post-relapse prognosis. We 2016. Whole-exome sequencing, transcriptome sequencing, developed a simple method to identify MM tumors and and methylation analyses were performed on 149 of 162 applied it to a validation cohort of 28 MSI-H colorectal tumors showing MSI in BAT25 and BAT26 loci. We analyzed cancers, identifying 16 MM tumors and 2 fusion kinases. patient survival times using Bonferroni-adjusted log-rank Conclusions: We discovered that fusion kinases are tests. frequently observed among sporadic MM MSI-H colorectal Results: Sporadic MSI-H colorectal cancers with promoter cancers. The new method to identify MM tumors enables us methylation of MLH1 (called MM) had a clinicopathological to straightforwardly group MSI-H patients into candidates proﬁle that was distinct from that of colorectal cancers of LS or fusion kinase carriers. Clin Cancer Res; 1–12. 2018 of patients with germline mutations (Lynch syndrome, AACR.

Introduction are frequently altered (1–3). Most patients with MSI-H colorectal cancers have poor prognosis after relapse according to consensus Approximately 10% of all colorectal cancers exhibit a micro- molecular subtyping (4, 5), supporting the concept that MSI-H satellite instability–high (MSI-H) status, in which the number of cancers are resistant to 5-fluorouracil (6). The considerable num- mono-, di-, or tri-nucleotide repeats in microsatellite sequences ber of mutations resulting from DNA mismatch repair (MMR) deficiency has hampered the identification of driver oncogenes that play essential pathogenic roles in MSI-H colorectal cancers. 1Department of Surgical Oncology, Graduate School of Medicine, The University MSI-H colorectal cancers have been conventionally divided of Tokyo, Tokyo, Japan. 2Department of Cellular Signaling, Graduate School of into hereditary or sporadic. To identify cases potentially involving 3 Medicine, The University of Tokyo, Tokyo, Japan. Department of Medical Lynch syndrome (LS), we have to rely on patients' clinical infor- Genomics, Graduate School of Medicine, The University of Tokyo, Tokyo, Japan. mation, such as family history or age. It is well known that MSI-H 4Genome Science Division, Research Center for Advanced Science and Tech- nologies, The University of Tokyo, Tokyo, Japan. 5Department of Pathology, colorectal cancers are tightly linked to the CpG island methylator Graduate School of Medicine, The University of Tokyo, Tokyo, Japan. phenotype (CIMP; refs. 7, 8), which is characterized by the cancer- 6Department of Surgery, Teikyo University School of Medicine, Tokyo, Japan. specific methylation of a definite set of CpG islands. Rational 7Department of Surgical Oncology and Vascular Surgery, Graduate School of patient stratification based on the precise recognition of the 8 Medicine, The University of Tokyo, Tokyo, Japan. National Cancer Center etiology is essential for the appropriate management of patients Research Institute, Tokyo, Japan. with MSI-H colorectal cancers. Note: Supplementary data for this article are available at Clinical Cancer To address these issues, we conducted an integrated molecular Research Online (http://clincancerres.aacrjournals.org/). characterization of 149 MSI-H colorectal tumors. †Deceased. Corresponding Author: Masahito Kawazu, Graduate School of Medicine, The Material and Methods University of Tokyo, 7-3-1 Hongo, Bunkyo-ku, Tokyo 113-0033, Japan, Phone: 81- Ethics 3-3547-5201: E-mail: [email protected] Patients with colorectal cancer gave written informed consent doi: 10.1158/1078-0432.CCR-18-1574 prior to their participation in the study. This project was approved 2018 American Association for Cancer Research. by the institutional ethics committees of The University of Tokyo

www.aacrjournals.org OF1

Sato et al.

genome-wide DNA methylation profiles of 93 tumors with an Translational Relevance Infinium Human MethylationEPIC BeadChip (Illumina) and Microsatellite instability–high (MSI-H) colorectal cancers performed whole-transcriptome sequencing (RNA-seq) from have been conventionally divided into hereditary (Lynch syn- 111 MSI-H with HiSeq2500 (Illumina). drome, LS) and sporadic categories. This report provides a fi rational basis for further classi cation of sporadic MSI-H colo- DNA methylation analyses rectal cancers into those with somatic mutations in mismatch Genome-wide DNA methylation analysis. Infinium Human repair genes (Lynch-like, LL) and those with promoter meth- MethylationEPIC BeadChip (Illumina) was used in accordance fi ylation of MLH1 (MM). There were signi cant differences with the manufacturer's protocol. Beta values were normalized between the LS/LL and MM groups in clinicopathological using BMIQ function in the R package wateRmelon (http:// properties including tumor localization, number of inser- www.bioconductor.org/packages/release/bioc/html/wateRmel tions/deletions, and recurrent mutations of KRAS/APC and on.html). For downstream analysis, we selected probes that fi BRAF/RNF43. Such a classi cation would enable precise man- were designed on promoter-associated sites of autosomes. agement of patients with MSI-H colorectal cancer. Fusion Then, we performed consensus clustering with 3,073 probes kinases were detected only in MM MSI-H colorectal cancers (probes with variance rankedinthetop1%)usingtheR lacking oncogenic KRAS or BRAF mutations and were associ- package ConsensusClusterPlus (http://www.bioconductor. ated with worse prognosis after relapse. A new, convenient org/packages/release/bioc/html/ConsensusClusterPlus.html). method for detecting MM tumors makes it possible to straight- forwardlyidentifyLScandidatesorMSI-Htumorslikelytocarry Bisulfite sequencing of MLH1 promoter. Genomic DNA was fusion kinases that are therapeutic targets of kinase inhibitors. subjected to bisulfite conversion using an EpiTect Bisulfite Kit (Qiagen). Converted DNA fragments were amplified by PCR using a Kapa HiFi Uracilþ Kit (Kapa Biosystems) with the primer sets indicated in Supplementary Table S1. Amplified PCR pro- (The Human Genome, Gene Analysis Research Ethics Committee; ducts were subjected to sequencing using the MiSeq system G10063 and G3546) and Teikyo University (#14–197), and the (Illumina) to determine the proportion of methylated alleles at study was conducted in accordance with the Declaration of each cytosine residue. Helsinki. MLH1 promoter methylation assay with methylation-sensitive Sample collection restriction enzyme. MLH1 promoter methylation was assessed by Surgically resected colorectal tumors (n ¼2,800) were PCR after digestion of genomic DNA with methylation-sensitive obtained from consecutive patients treated at The University of restriction enzyme. Genomic DNA (200 or 25 ng) was digested in Tokyo Hospital and Teikyo University Hospital between 1998 a volume of 20 or 10 mL by Anza 22 SmaI (Thermo Fisher and June 2016. To validate a fusion-detection strategy, we added Scientific), followed by heat inactivation of restriction enzyme, 185 primary colorectal cancers, collected at The University of in accordance with the manufacturer's instructions. Digested DNA Tokyo Hospital from July 2016 through April 2017. (20 or 5 ng) was subjected to 25 or 27 cycles of multiplex PCR in a total volume of 25 mL using primeSTAR GXL DNA polymerase Microsatellite instability testing (Takara Bio), in accordance with the manufacturer's instructions. Tumor tissues and corresponding normal mucosae were BRAF was used as a positive control. The primer sets are indicated obtained from surgically resected specimens and were either in Supplementary Table S1. Methylation status was determined by snap-frozen in liquid nitrogen immediately after resection and 2% agarose gel electrophoresis of 12.5 mL of PCR products. stored at 80C or immersed in RNAlater Tissue Protect Tubes (Qiagen) overnight at 4 C followed by storage at 20 C until use. Whole-exome sequencing including mutation call, Genomic DNA was extracted from tissue samples with the DNeasy copy-number analysis, and signature analysis Blood and Tissue Kit or the QIAamp DNA Mini Kit (Qiagen) and Genomic DNA was isolated from each sample and underwent analyzed by PCR at two microsatellite loci, BAT25 and BAT26, enrichment of exonic fragments using SureSelect Human All Exon using the labeled primers indicated in Supplementary Table S1. Kit v5 (Agilent Technologies). Massively parallel sequencing of PCR products were electrophoresed on an ABI PRISM 3100, isolated fragments was performed with a HiSeq2500 (Illumina) an ABI PRISM 3700, or a 3130xl Genetic Analyzer (Applied using the paired-end option. Paired-end whole-exome sequenc- Biosystems), and fluorescent signals were analyzed using GeneS- ing (WES) reads were independently aligned to the human can 3.7, GeneScan 3.5, or GeneMapper 4.0, and Genotyper 2.1, reference genome (hg38) using BWA (9), Bowtie2 (http://bow Genotyper 3.6, or PeakScanner 1.0 (all from Applied Biosystems), tie-bio.sourceforge.net/bowtie2/index.shtml), and NovoAlign in accordance with the manufacturer's instructions. (http://www.novocraft.com/products/novoalign/). Somatic mutations were called using MuTect (http://www.broadinsti Comprehensive genomic and epigenomic analyses tute.org/cancer/cga/mutect), SomaticIndelDetector (http:// Genomic DNA from 149 MSI-H tumors (148 adenocarcinomas www.broadinstitute.org/cancer/cga/node/87), and VarScan and 1 adenoma in 146 patients, all chemotherapy-na€ve, 1 case (http://varscan.sourceforge.net). Mutations were discarded if: with preoperative radiotherapy) and corresponding paired-nor- (i) the read depth was <20 or the variant allele frequency (VAF) mal tissues consisting of adjacent, histologically normal tissues was <0.1, (ii) they were supported by only one strand of the resected at the time of surgery was subjected to whole-exome genome, or (iii) they were present in normal human genomes in sequencing (WES) with HiSeq2500 (Illumina). We analyzed the either the 1000 Genomes Project dataset (http://www.internatio

OF2 Clin Cancer Res; 2018 Clinical Cancer Research

Actionable Fusion Kinases in MSI-H CRCs

nalgenome.org/) or our in-house database. Gene mutations were (Agilent Technologies) or TapeStation (Agilent Technologies). annotated by SnpEff (http://snpeff.sourceforge.net). Copy-num- RNA with a high RNA integrity number underwent RNA-seq using ber status was analyzed using our in-house pipeline, which a NEBNext Ultra Directional RNA Library Prep Kit (New England determines the logR ratio (LRR) as follows: (i) we selected SNP BioLabs) in which complementary DNA (cDNA) was prepared positions in the 1000 Genomes Project database that were in a from polyA-selected RNA. RNA with a low RNA integrity number homozygous state (VAF 0.05 or 0.95) or a heterozygous state was subjected to RNA-seq using a TruSeq RNA Access Library Prep (VAF 0.4–0.6) in the genomes of respective normal samples, (ii) Kit (Illumina) in which cDNA was generated with random pri- normal and tumor read depths at the selected position were mers. Prepared RNA-seq libraries underwent next-generation adjusted based on GþC percentage of a 100-bp window flanking sequencing of 120 bp from both ends (paired-end reads). The ti the position (10), (iii) we calculated the LRR ¼ log2 , where ni and expression level of each gene was calculated using DESeq2 (http:// ni bioconductor.org/packages/release/bioc/html/DESeq2.html) ti are normal and tumor-adjusted depths at position i, and (iv) each representative LRR was determined by the median of a with VST transformation, and gene fusions were detected using a moving window (1 Mb) centered at position i. deFuse pipeline (https://bitbucket.org/dranew/defuse) and STAR LRR of the copy number of both alleles, that of the major allele, (https://github.com/alexdobin/STAR). and that of the minor allele were determined for every region of the genome. The P values for gain or loss of respective genomic Cloning of fusion genes regions were determined from the LRRs with the permutation test Complementary DNAs for wild-type, mutant, and fusion pro- fi (100,000 iterations) following the algorithm used in GISTIC teins were ampli ed by reverse transcription PCR (RT-PCR) from (11, 12). Q values were calculated from the P values using the RNA samples and then ligated into the pMXs retroviral vector fi R package q value (http://github.com/jdstorey/qvalue). (Cell Biolabs). The sequences of all cDNAs were veri ed by Sanger Mutational signatures were analyzed using the Wellcome Trust sequencing. Primer sequences are provided in Supplementary Sanger Institute Mutational Signature Framework (http://jp.math Table S1. works.com/matlabcentral/fileexchange/38724-wtsi-mutational- signature-framework). The optimal number of signatures was 3T3 cell transformation assay determined in accordance with the signature stabilities and aver- Human embryonic kidney (HEK) 293T cells and mouse 3T3 fi age Frobenius reconstruction errors. broblasts were obtained from the ATCC and maintained in DMEM-F12 supplemented with 10% FBS (both from Life Tech- nologies). Cell lines were propagated for less than 3 months after Immunohistochemical analysis of mismatch repair proteins initial plating. Cultured cells were tested for mycoplasma con- Immunohistochemistry of mismatch repair proteins (MLH1, tamination using a MycoAlert Mycoplasma Detection Kit MSH2, MSH6, and PMS2) was performed on whole sections of (Lonza). To obtain infectious virus particles, recombinant vectors each tumor. Formalin-fixed paraffin-embedded tumor blocks were introduced together with an ecotropic packaging plasmid were sliced into 3-mm-thick sections, which were then immunos- (Takara Bio) into HEK293T cells by transfection. For the focus tained with the Ventana Benchmark automated immunostainer formation assay, 3T3 cells were infected with ecotropic recombi- (Roche). Primary antibodies used were mouse monoclonal anti- nant retroviruses and cultured for 12 days in DMEM-F12 supple- MLH1 antibody (clone ES05, dilution 1:50; Leica), mouse mono- mented with 5% calf serum. Cell numbers were counted using a clonal anti-MSH2 antibody (clone FE11, dilution 1:50; Dako), luminometer and the CellTiter-Glo Luminescent Cell Viability rabbit monoclonal anti-MSH6 antibody (clone EPR3945, dilu- assay (Promega). Linsitinib (S1091), entrectinib (S7998), regor- tion 1:200; GeneTex), and mouse monoclonal anti-PMS2 anti- afenib (S1178), and PLX7904 (S7964) were obtained from body (clone EPR3947, no dilution; Roche). Selleck Chemicals (Houston). Immunostained slides were blindly evaluated by a board- certified pathologist (Hiroyuki Abe). Nuclear staining was eval- Animal experiment uated in each tumor. Epithelial cells in the proliferative zone of All animal experimental procedures were approved by the Insti- non-neoplastic colonic mucosa and lymphocytes in the germinal tutional Animal Care and Use Committee of The University of center of lymph follicles were used as internal positive controls. Tokyo. We also adhered to the standards articulated in the NC3Rs guidelines (Animal Research: Reporting of In Vivo Experiments). Pathway analysis 3T3 cells expressing SLC12A2–INSR and RUFY1-RET were sub- The Database for Annotation, Visualization and Integrated cutaneously inoculated into the flankof6-to8-week-oldfemale Discovery web-based tool (https://david.ncifcrf.gov) was used to nude mice (BALB/c-nu/nu; Charles River Laboratories Japan, Inc.) fi identify pathways. Pathways de ned in Gene Ontology (limited at 2 106 cells/200 mL of PBS. Mice carrying implanted tumors to the "biological pathway" category; http://www.geneontology. were divided randomly into two or three groups after confirming org), KEGG pathway (http://www.genome.jp/kegg/), BioCarta tumor growth in each experiment. Kinase inhibitors, dissolved in (https://cgap.nci.nih.gov/Pathways/BioCarta_Pathways), and the N-methyl-2-pyrrolidone (Nacalai Tesque) and polyethylene glycol Reactome Pathway Database (http://www.reactome.org) were (PEG; Sigma-Aldrich), were applied orally once daily as indicated. used for the analysis. Linsitinib (LCL #L-5814; LC Laboratories) was applied at a dose of 25 mg/kg body weight (8 tumors in 4 mice). Regorafenib Transcriptome sequencing, expression analysis, and detection (AK106990; Ark Pharm) was applied at a dose of 100 mg/kg of fusion genes (10 tumors in 5 mice) or 10 mg/kg (8 tumors in 4 mice). Tumor Total RNA was extracted with RNA Bee reagent (Tel-Test Inc.) diameter was measured using callipers, and tumor volume was and treated with DNase I (Qiagen) using an RNeasy Mini Kit determined by calculating the volume of an ellipsoid using the (Qiagen). RNA integrity was evaluated using either a Bioanalyzer following formula: length width2 0.5.

www.aacrjournals.org Clin Cancer Res; 2018 OF3

Sato et al.

Statistical analysis Data availability Numerical variables were summarized by median and range. Raw sequencing data were deposited in the Japanese Genotype- Comparisons of numerical variables between groups were Phenotype Archive (http://trace.ddbj.nig.ac.jp/jga), which is carried out using a nonparametric approach (Wilcoxon rank- hosted by the DNA Data Bank of Japan (13) under accession sum test, two-tailed). Comparisons of the distribution of cat- number JGAS00000000113 (NBDC number: hum0094). egorical variables in different groups were performed using the Fisher's exact test. The survival curve was established by the Results Kaplan–Meier method and compared by the log-rank test with Bonferroni adjustment. Cox proportional hazards model was Demographic characteristics also used to evaluate the effects of multiple variables. Statistical In the genomes of 162 tumors (5.8%), microsatellites were analysis was performed using the computing environment R unstable at both BAT25 and BAT26 loci, which we classiﬁed as (version 3.2.3). MSI-H in the strict sense (Fig. 1A and Supplementary Table S2;

Figure 1. Classification of microsatellite instability–high colorectal tumors. A, Number of samples analyzed in this study. B, Subtype classification flowchart. C, Tumor localization in the large intestine. The numbers of LS-associated, Lynch-like, and MLH1-methylated tumors within the indicated anatomical regions are shown in the pie charts. The size of the pie chart is proportional to the total number of tumors within the region. D, Age-of-operation distributions of microsatellite instability–high colorectal tumors according to subtype. A two-tailed Wilcoxon rank-sum test was used for statistical analysis. E, Survival after relapse of patients with microsatellite instability–high colorectal tumors, according to subtype. Survival curves were estimated using the Kaplan–Meier method and compared by two-sided log-rank test. WES, whole-exome sequencing; RNA-seq, transcriptome sequencing; MMR, mismatch repair; NA, not available; LS, Lynch syndrome- associated; LL, Lynch-like; MM, MLH1 promotor-methylated; C, cecum; A, ascending colon; T, transverse colon; D, descending colon; S, sigmoid colon; R, rectum.

OF4 Clin Cancer Res; 2018 Clinical Cancer Research

Actionable Fusion Kinases in MSI-H CRCs

ref. 14). The frequency of MSI-H was lower than expected (6), Further analysis with a larger population is required to evaluate probably because only two mononucleotide repeat markers were the difference in prognosis. used, while the probability of including MSI-low tumors was low (15–17). Clinicopathological characteristics of the MSI-H cases Mutations in MMR genes are summarized in Supplementary Table S3. Because MSI-H colorectal cancers often harbor disruptive somatic mutations within MMR genes that can in turn be affected Classification of MSI-H colorectal cancer by MMR deficiency, it is difficult to distinguish causative muta- To clarify the association between MSI-H status and the tions in MMR genes from resultant mutations. Because MLH1 CIMP (7, 8), consensus clustering was performed using data promoter methylation is regarded as the primary cause of MMR from methylation profile analyses. Sixty-four tumors (69%) deficiency (20), we considered that somatic mutations in MMR that were clustered in the CIMP branch were also found in a genes detected in MLH1 promoter-methylated tumors were likely subset of 65 tumors in which MLH1 expression was silenced resultant ones, and we excluded them from the causal MMR gene because of promoter methylation (Supplementary Fig. S1), mutation panel of LS/LL (Fig. 2). The excluded somatic mutations confirming almost complete overlap between these two clus- were as follows: MLH1 [(S193fs; n ¼ 1), MSH2(L229fs; n ¼ 3), ters. In this report, we define CIMP based on the methylation MSH6(R248fs) (n ¼ 4), and MSH6(P1087fs; n ¼ 41; located on status of the MLH1 promoter in order to avoid ambiguity (A)6, (A)7, (A)7, and (C)8 repeats, respectively]. Mutations in caused by the different clustering methods. Promoter methyl- MMR genes that were truncated or reported to be pathogenic ation of MMR genes other than MLH1, namely, MSH2, MSH6, were regarded as being causative (Supplementary Table S4). PMS1, MSH3, PMS2,andMLH3, was not detected (Supple- Next, we investigated how the function of MMR genes with a mentary Fig. S1). heterozygous mutation in LS was abolished (Fig. 2A, Supplemen- To complement the genome-wide methylation profile analysis, tary Fig. S3). Somatic uniparental disomy (UPD) in three MMR the promoter region of MLH1 (chr3: 36,993,202–36,993,864) in genes was observed in 12 LS tumors (MLH1, n ¼ 8; PMS2, n ¼ 3; the genome of 149 samples was analyzed by bisulfite sequencing. and MSH2, n ¼ 1). Additional somatic single-nucleotide varia- The methylation status determined by bisulfite sequencing was tions (SNV) or insertions/deletions (indels) in MMR genes were well correlated with that determined by genome-wide analysis observed in five LS tumors (MSH6, n ¼ 2; PMS2, n ¼ 2; and MSH2, (Supplementary Fig. S2). We refer to the MLH1-silenced MSI-H n ¼ 1). An LS tumor with a heterozygous germline mutation in colorectal cancers as MLH1 promoter-methylated (MM) tumors PMS2 lost the wild-type allele, and additional pathogenic muta- (Fig. 1B). tions were not identified in the remaining five LS tumors. Of the 23 tumor samples from patients with LS in our cohort, We further sought to identify the alterations of MMR genes in LL 11 LS tumors underwent methylation profile analysis, and we tumors. Sixteen LL tumors (16/27, 59%) harbored somatic muta- found that none of these tumors clustered in the CIMP branch tions within MMR genes, as previously reported (21). Six tumors (Supplementary Fig. S1). Bisulfite sequencing also revealed that had somatic SNVs/indels in MLH1 and one tumor had a somatic promoter methylation of MLH1 was absent in all 23 LS tumors. SNV in MSH2, both of which were accompanied by UPD of the Neither germline mutations in MMR genes nor promoter meth- respective gene. Biallelic mutations in MSH2 were observed in five ylation of MLH1 was observed in 27 tumor samples of MSI-H tumors. Heterozygous somatic mutations in MLH1 and MSH2 colorectal cancers, 17 of which were subjected to methylation were observed in three tumors and one tumor, respectively, profile analysis. Because the 17 tumors clustered together with LS without additional detectable events. However, in the remaining tumors in consensus clustering, we refer to these samples as 11 tumors, we did not identify pathogenic mutations in MMR "Lynch-like (LL)" tumors (Fig. 1B, Supplementary Fig. S1). As genes that would account for the MSI-H status, although six shown in the flowchart in Fig. 1B, our cohort of patients with MSI- variants of uncertain significance (VUS), listed in Supplementary H colorectal cancer was classified into three subgroups with Table S4, are not included in Fig. 2A. distinct genomic/epigenomic profiles: LS (n ¼ 23), LL (n ¼ We validated these results by immunohistochemistry (IHC) of 27), and MM (n ¼ 99) (Fig. 1B). MMR proteins, obtaining concordant results in 91% (32/35) of cases for which both the genotype data and protein expression Distinct clinical features of LS/LL tumors and MM tumors data were available (Fig. 2). In seven cases without genotype data, Interestingly, although MM tumors were more likely to develop the results of IHC revealed probable disrupted MMR genes. in the "right" side of the large intestine (consisting of the cecum, and the ascending and transverse colon), a finding that was Somatic SNVs and indels previously reported concerning MSI-H colorectal cancers in gen- With >20-fold coverage of at least 85% of the target regions, eral (1), LS/LL tumors were evenly distributed across the large whole-exome sequencing (WES) identified a median of 962 intestine (Fig. 1C). This observation is consistent with the hypoth- (range, 103–6,973) somatic nonsynonymous SNVs and a median esis that aberrant DNA methylation may develop in the specific of 130 (range, 5–308) somatic indels in coding regions. The microbiological environment of the "right" side of the large number of recurrent (>5% frequency) indels (n ¼ 684) was larger intestine (18, 19). Notably, and justifying our classification than that of recurrent (>5% frequency) SNVs (n ¼ 5). The approach, we found that the age of the patients at operation frequencies of SNVs did not differ significantly between LL/LS differed between the subtypes [one-way ANOVA, F(2,143) ¼ 54, P tumors and MM tumors (P ¼ 0.49), whereas the frequencies of < 2.0 10 16; Fig. 1D]. The patients with MM tumors had worse indels were higher in MM tumors than those found in LS/LL survival after relapse than those with LS/LL tumors (log-rank test, tumors (P ¼ 0.0028, Wilcoxon rank-sum test; Supplementary P ¼ 0.013; Fig. 1E), although the difference was not statistically Fig. S4). significant after adjusting for age and stage at operation (Cox To identify mutations that are likely relevant to MSI-H colo- proportional hazard model, HR, 0.36; 95% CI, 0.04–3.08). rectal cancer oncogenesis under conditions with a large number of

www.aacrjournals.org Clin Cancer Res; 2018 OF5

Sato et al.

Figure 2. Mutations in DNA mismatch repair genes and immunohistochemical analysis of DNA mismatch repair proteins. A, Mutational status of DNA mismatch repair genes is shown in association with the results of immunohistochemical analysis of DNA mismatch repair proteins. The columns in the table denote samples and the rows denote gene alleles (top) or proteins (bottom). B, Representative images of the immunohistochemical analysis. SNV, single-nucleotide variation; indel, insertion or deletion; UPD, uniparental disomy.

background mutations, the identified somatic variants were strat- respectively), whereas mutations in BRAF and RNF43 were more ified into three tiers (Supplementary Fig. S4): Tier 1 (n ¼ 78,187 in prevalent in MM tumors than in LS/LL tumors (P ¼ 5.5 10 12 18,646 genes), consisting of all somatic variants detected; Tier 2 (n and 9.4 10 8, respectively; Fig. 3A, Supplementary Table S7). ¼ 46,502 in 10,748 genes), consisting of those Tier 1 variations Mutations encoding BRAF(V600E) were not observed among LS/ that were either also detected in RNA-seq analysis or present LL tumors, in accordance with the well-established notion that recurrently (10 times) in the Catalogue of Somatic Mutations BRAF mutations and germline mutations in MMR genes are in Cancer database (http://cancer.sanger.ac.uk/cosmic/; Supple- mutually exclusive (25, 26). We did not observe statistically mentary Table S5); and Tier 3 (n ¼ 3,346 in 458 genes), consisting significant differences in mutation frequencies between LS and of those Tier 2 variations present in genes found in the Cancer LL tumors (Supplementary Table S7). Gene Census (22) or that were reported to be recurrent (30% Through signature analysis of SNVs (27), four mutational frequency) in a previous study of MSI-H colorectal cancer (23). signatures (termed Signatures A, B, C, and D) were detected (Fig. The 40 genes most frequently affected in Tier 3 are well- 3A and B). Signatures A, B, and C exhibited profiles similar to recognized targets in the MSI-H status, including ACVR2A, SEC63, signatures associated with MMR deficiency, whereas Signature D TGFBR2, and RNF43 (shown in Fig. 3A). We found that there were was an age-related signature (28). Signature C, which may be 209 frequently altered genes (>20% of samples) within Tier 2 derived from G-T mismatch, was abundant in MM tumors; the variants (Supplementary Table S5). We also obtained a list of 860 median percentage of signature C mutations in LS/LL tumors was significantly (Q < 0.01) mutated genes using MutSigCV (Supple- 13.9% (interquartile range: 7.04%–24.4%), while that in MM mentary Table S6, Supplementary Fig. S4; ref. 24). Notably, 153 tumors was 24.0% (interquartile range: 19.0%–29.8%; Fig. 3C). (73%) of the 209 most frequently altered Tier 2 genes were Signature A, which may be derived from G-T and A-C mismatches, included in the list of significantly mutated genes, partly justifying was enriched in tumors with PMS2 disruptions; the median the stratification strategy used in this study. Mutations in KRAS, percentage of signature A mutations in tumors with defective APC, and TCF7L2 were more prevalent in LS/LL tumors than in PMS2 was 73.2% (interquartile range: 65.0%–87.3%), while MM tumors (P ¼ 5.7 10 9, 6.5 10 8, and 1.1 10 3, that in the other tumors was 26.4% (interquartile range:

OF6 Clin Cancer Res; 2018 Clinical Cancer Research

Actionable Fusion Kinases in MSI-H CRCs

Figure 3. Summary of mutations in microsatellite instability–high colorectal tumors. A, The 40 most frequently mutated genes with their mutation status color-coded for each patient. The frequency of synonymous or nonsynonymous substitutions and insertions/deletions is shown at the top, and the percentages of the mutational signatures are shown at the bottom. The frequencies of mutations in each gene according to subtype are shown on the left. B, The four mutational signatures identified in this study. C, The percentage of Signature C according to subtype. D, The percentage of Signature A according to PMS2 mutational status. E, Allele-specific copy-number alterations in microsatellite instability–high colorectal tumors. The profiles of allele-specific copy-number alterations according to subtype are shown. Red and blue lines indicate the Q-value for gains of major allele and losses of minor allele, respectively. The y-axis indicates the frequency of the observed gain or loss. Representative genes affected by respective copy- number alteration are shown. A two- tailed Wilcoxon rank-sum test was used for statistical analysis. LS, Lynch syndrome; LL, Lynch-like; MM, MLH1-methylated.

21.5%–31.3%; Fig. 3D). These data may be useful for further specific copy number (CN) analysis using WES data revealed 41 exploration of the etiology of MSI-H colorectal cancers. recurrent (Q < 0.01) CNAs, 9 of which were involved in UPD (Fig. 3E, Supplementary Table S8). Notably, these CNA profiles Somatic copy-number alterations differed between LS/LL and MM tumors. UPD encompassing A previous study reported that somatic copy-number altera- TGFBR2 (3p24.1), MLH1 (3p22.2), and CTNNB1 (3p22.1) was tions (CNA) are infrequent in MSI-H colorectal cancers compared observed specifically in LS/LL tumors, and these genes have been with those in MSS colorectal cancers (29). In this study, allele- linked to the pathogenesis of MSI-H colorectal cancer. In contrast,

www.aacrjournals.org Clin Cancer Res; 2018 OF7

Sato et al.

Figure 4. Frequently affected pathways. Major components of the pathways extracted from the top 209 Tier 2 genes using The Database for Annotation, Visualization and Integrated Discovery web-based tool algorithm are shown. The frequencies of single-nucleotide variations and insertions/deletions in the genes are expressed as a percentage of all cases according to subtype. Warm colors denote activated genes, and cold colors denote inactivated genes. The frequencies are color-scaled. LS, Lynch syndrome; LL, Lynch-like; MM, MLH1-methylated.

we found that a CN gain within the long arm of chromosome 8 with MutSigCV (Supplementary Table S10), partly supporting our was recurrent in MM tumors. In addition, UPD within the short stratification of mutated genes. arm of chromosome 6 involving the major histocompatibility complex class I and class II genes was prevalent in both LS/LL and Oncogenic alterations in MSI-H colorectal cancer MM tumors, which may contribute to the avoidance of antitumor Considering that activating kinase fusion events are extremely immunity by tumor cells. rare in colorectal cancer and do not seem to cluster in a particular genomic or epigenomic subtype of the disease (32), we did not Affected pathways anticipate the existence of recurrent fusion genes; surprisingly, To better understand genetic alterations in the MSI-H colorectal using RNA-seq, we detected in-frame fusion transcripts encoding cancer genome in our cohort, we performed pathway analysis fusion-type kinases in 11 out of 111 tumors (9.9%). These 11 using the Database for Annotation, Visualization and Integrated tumors carrying fusion kinases were all found in the MM subtype, Discovery web-based tool (https://david.ncifcrf.gov; refs. 30, 31). appearing at a frequency of 11% (11/99). Considering the mutual Of the 209 most frequently altered Tier 2 genes, four molecular exclusivity of oncogenic KRAS/BRAF missense mutations and pathways were identified, specifically, DNA damage-sensing and fusion kinases, clinically actionable fusion kinases including histone H3 methylation-associated pathways, as well as Wnt novel insulin receptor gene fusion accounted for 55% (11/20) signaling and RAS/RAF/mitogen-activated protein kinase (MAPK) of MM tumors lacking oncogenic KRAS/BRAF missense mutations pathways (Fig. 4, Supplementary Table S9). A similar result was (Fig. 5A). In addition, we identified oncogenic mutants of ERBB2 obtained using the 860 significantly mutated genes identified (n ¼ 3), ERBB3 (n ¼ 5), HRAS (n ¼ 2), RRAS2 (n ¼ 1), and RAC1

OF8 Clin Cancer Res; 2018 Clinical Cancer Research

Actionable Fusion Kinases in MSI-H CRCs

Figure 5. Oncogenic fusion proteins in microsatellite instability–high colorectal tumors. A, Mutation plot of oncogenes with transforming potential. Mutations with transforming capacity are shown in darker colors, whereas mutations of unknown signiﬁcance are shown in lighter colors. B, Schematic diagrams of the fusion proteins identiﬁed in this study. Amino acid numbering on the fusion proteins refers to the sequences of the wild-type proteins. A novel fusion kinase (SLC12A2–INSR) is shown in orange. A novel fusion partner (RUFY1) of RET fusion kinase is shown in green. C, The transcript fusion point of SLC12A2–INSR complementary DNA determined by Sanger sequencing. D, A schematic representation of the gene rearrangement generating the SLC12A2–INSR fusion gene. The sequence of the genomic fusion point determined by Sanger sequencing is shown at the bottom. E, The response to kinase inhibitors of 3T3 cells expressing fusion proteins. Cells were treated with the indicated drugs at the indicated concentrations. Cellularity was measured 9 days after treatment and is plotted relative to that of cultures treated with the lowest concentration. Error bars indicate model-based standard errors. In the graphs showing the data obtained for cells expressing TPM3-NTRK1, KANK1-NTRK3, or EML4-NTRK3, the control v-Ras data are identical. A two-tailed student t test was used for statistical analysis. CC, coiled coil.

www.aacrjournals.org Clin Cancer Res; 2018 OF9

Sato et al.

(n ¼ 1; Supplementary Fig. S5, Supplementary Table S11), although these variants were not necessarily mutually exclusive regarding KRAS, BRAF, and fusion kinase genes. The detected fusion kinases involved INSR (n ¼ 1), RET (n ¼ 2), NTRK1 (n ¼ 2), NTRK3 (n ¼ 2), and BRAF (n ¼ 4; Fig. 5B). The expression of these fusion transcripts was confirmed by reverse- transcription PCR (Fig. 5C, Supplementary Table S12). We further identified the genomic fusion points of 10 fusion genes, including SLC12A2–INSR, all of which were confirmed to be somatic (Fig. 5D, Supplementary Figs. S6 and S7). Excluding AKAP9–BRAF, which was predicted to be 4,253 amino acids in length and technically difficult to manipulate, and TRIM24–BRAF, which has been well described previously (33), tumorigenicity of the fusion proteins identified in our study was confirmed using a 3T3 transformation assay (Supplementary Fig. S5). Luciferase reporter assay indicated that these fusion proteins activated the mitogen-activated protein kinase pathway (Supplementary Fig. S5). In addition, small-molecular inhibitors Figure 6. that suppress the activity of the kinases identified in our fusion Precise and simple detection of MLH1-methylated tumors. A comparison products significantly attenuated the malignant transformation of between extracted data of the degree of methylation in the MLH1 promoter region of 13 tumors by genome-wide methylation array (see Supplementary 3T3 cells in a concentration-dependent manner (Fig. 5E). Nota- Fig. S1C; top) and detection of MLH1 methylated tumors using our MLH1 bly, linsitinib, which is usually used for the main purpose of methylation assay (bottom). Amplicon of BRAF is a loading control and antagonizing IGF1R signaling (34), inhibited the growth of 3T3 used as a template of Sanger sequencing for detecting BRAF(V600E) mutation. cells expressing SLC12A2–INSR. Although it may be difficult to M, 50-bp DNA ladder marker; RKO, positive control; DLD1, negative control; target oncogenic BRAF fusion proteins with specific BRAF inhi- NTC, no template control. bitors that are currently available, it is anticipated that tumors with BRAF fusion proteins could be targeted with a combination method to detect MLH1 promoter methylation utilizing methyl- of MAPK kinase and PI3K inhibitors (35–37). ation-sensitive restriction enzyme. As shown in Fig. 6, the meth- It has been reported that alectinib and entrectinib inhibited the ylation status determined by our MLH1 promoter methylation in vivo growth of the Lc-2/ad lung cancer cell line carrying CCDC6- assay was well correlated with that determined by genome-wide RET and the KM12 colorectal cancer cell line carrying TPM3– analysis. Tumors in a validation cohort were subjected to MSI NTRK1, respectively (38, 39). Because we could not identify MSI- testing, and the methylation status of MSI-H tumors was deter- H colorectal cancer cell lines harboring SLC12A2–INSR or mined by our MLH1 methylation assay. Subsequently, we con- RUFY1–RET, we evaluated the therapeutic efficacy using an in ducted RNA-seq in MM MSI-H colorectal cancers lacking onco- vivo mouse model, in which 3T3 cells transformed by the fusion genic KRAS/BRAF missense mutations. We succeeded in detecting kinases were subcutaneously inoculated. Linsitinib and regorafe- two fusion kinases (NCOA4-RET and CUL1–BRAF; ref. 40) out of nib substantially suppressed growth of the transformed 3T3 cells four MM MSI-H colorectal cancers lacking oncogenic KRAS/BRAF expressing SLC12A2 –INSR and RUFY1–RET, respectively (Sup- missense mutations (Supplementary Fig. S9). We used BRAF as a plementary Fig. S8). control for PCR because we detected CNAs in neither chromosome 7, including BRAF as previously reported (29) nor the SmaI Comparison with MSS site. We analyzed the MSS colorectal cancer RNA-seq database from the Cancer Genome Atlas Network (29). Twenty-six paired-end Prognosis and 181 single-end RNA sequences (median 36,213,120 reads It has been recognized that mutation in BRAF is a poor prog- per sample, 76 bp) were available. Only one fusion kinase (HLA- nostic factor in metastatic or relapsed MSI-H colorectal cancer A-RET) was detected in 189 patients. To adjust the conditions of (4, 41, 42, 43). We analyzed the survival data in our cohort, given analysis, the 36 randomly selected mega reads per sample from the considerable number of fusion kinases observed in MSI-H our RNA-seq data (median 206,727,876 reads per sample, colorectal cancers lacking oncogenic KRAS/BRAF missense muta- 120 bp) were trimmed down to 76 bp and subjected to analysis, tions, as mentioned above. When VUS was included in the wild- identifying all of the fusion transcripts described above except for type group, patients harboring BRAF (V600E) had worse overall that encoding ARMC10-BRAF. Taken together, oncogenic fusion survival than patients in the BRAF-wild-type group (P ¼ 2.3 kinases were more prevalent in MSI-H colorectal cancers than in 2 10 ; Supplementary Fig. S10A) and patients in the BRAF-wild- MSS colorectal cancers. –3 type non-fusion group (P ¼ 9.5 10 ; Supplementary Fig. S10B). The survival after relapse of patients harboring fusion Effective strategy for detection of fusion kinases kinases was shorter than that of patients harboring oncogenic Despite the considerable number of fusion kinases in MSI-H 3 KRAS mutations (P ¼ 8.2 10 ; Supplementary Fig. S10C). colorectal cancers, it is expensive to perform next-generation sequencing for all MSI-H samples in clinical practice. In addition, Discussion bisulfite-treated DNA and particular devices such as pyrosequen- cer are required for precise MLH1 promoter methylation assays. In this study, we performed a comprehensive analysis of genetic To effectively detect fusion kinases, we first developed a simple alterations in MSI-H colorectal cancer, revealing several important

OF10 Clin Cancer Res; 2018 Clinical Cancer Research

Actionable Fusion Kinases in MSI-H CRCs

pathological aspects of MSI-H colorectal cancers that have clinical and that chromosomal aberrations are relatively infrequent in implications. MSI-H colorectal cancers (29). A precise understanding of the First, we showed that MSI-H colorectal cancers could be clas- functional consequences of mutations in the DNA damage-sens- sified into three subtypes. These classifications were validated ing pathway in MSI-H colorectal cancers may lead to the devel- through several observations: (i) there was a greater number of opment of novel diagnostic and/or therapeutic approaches for indels in the MM subtype than in the LS/LL subtype; (ii) one of the patients with MSI-H colorectal cancer. mutational signatures associated with MMR deficiency was Taken together, our genomic and epigenomic analyses of MSI-H enriched in the MM subtype; (iii) the frequencies of mutations colorectal cancer provide clinically relevant findings that may in several genes differed between the LS/LL and MM subtypes; (iv) impact on patient management and care in association with the LS/LL and MM subtypes exhibited distinctive CNA profiles; cancer predisposition, molecular-targeted therapy, and cytotoxic and (v) fusion genes encoding oncogenic kinases were enriched in chemotherapy. the MM subtype. However, a caveat of our findings is that the LL subtype may contain misdiagnosed LS tumors because some Disclosure of Potential Conflicts of Interest forms of germline mutations in MMR genes are difficult to detect No potential conflicts of interest were disclosed. (44). Although it is not clear whether the LS and LL subtypes have distinctive clinicopathological characteristics, further investiga- Authors' Contributions tion is warranted. Conception and design: K. Sato, T. Watanabe, H. Mano Second, we showed that the classification described above Development of methodology: K. Sato, Y. Yamamoto Acquisition of data (provided animals, acquired and managed patients, has important clinical implications. We also propose a feasible provided facilities, etc.): K. Sato, Y. Yamamoto, G. Nagae, H. Abe, M. Soda, method using a methylation-sensitive restriction enzyme for S. Kohsaka, Y. Yamashita, H. Iinuma, M. Fukayama, H. Aburatani, T. Watanabe, the detection of MM tumors. Importantly, we chose the H. Mano genomic region, in which the methylation status is most strictly Analysis and interpretation of data (e.g., statistical analysis, biostatistics, associated with the silencing of MLH1 based on the genome- computational analysis): M. Kawazu, T. Ueno, S. Kojima, M. Soda wide methylation analysis (Supplementary Fig. S2). Among the Writing, review, and/or revision of the manuscript: K. Sato, M. Kawazu, Y. Yamamoto, H. Abe, M. Soda, S. Kohsaka, Y. Yamashita, H. Iinuma, MSI-H colorectal cancers, MM tumors that lack KRAS or BRAF M. Fukayama, T. Watanabe, H. Mano mutation are expected to carry a targetable fusion kinase with Administrative, technical, or material support (i.e., reporting or organizing high probability. Determination of the MM subtype may data, constructing databases): T. Oga, E. Sai, M. Fukayama, T. Watanabe, help to identify patients who can benefitfromanintensive H. Mano search for fusion kinases. Despite the remaining ambiguity Study supervision: M. Kawazu, H. Aburatani, T. Watanabe between LS and LL subtype classifications, recognition of the LLsubtypeisalsoclinicallyvaluableinpracticewhen Acknowledgments considering patients of whom germline mutations in MMR The authors thank Miki Tamura, Reina Takeyama, Akane Maruyama, Junko Tamura, Riyo Kakimoto, Dr. Takamitsu Kanazawa, Dr. Yuzo Nagai, and genes may escape detection. Dr. Takashi Kobunai for their technical assistance, and Dr. Yoichi Furukawa Third, we identified oncogenic fusion genes that have trans- for his advice and discussion. We are grateful to all of the patients and families forming activity and are potential therapeutic targets. Using our who contributed to this study. We thank all members of the Colorectal Group of MLH1 methylation assay and considering their mutual exclusiv- The University of Tokyo and Teikyo University for their support of this research. ity, it is possible to develop a simple diagnostic sequencing We also thank Edanz Group (www.edanzediting.com/ac) for editing a draft of fi strategy for the efficient detection of LS and fusion kinases to this article. This study was supported in part by grants-in-aid for Scienti c Research (KAKENHI, grant numbers 17J00386, 26430106, 16K07143, and provide improved personalized therapeutic opportunities for 16H02672) from the Japan Society for the Promotion of Science (JSPS) and patients with MSI-H colorectal cancer. Although MSI-H colorectal by grants from Leading Advanced Projects for Medical Innovation (LEAP, cancers carry multiple driver oncogenes, knockdown of TPM3– JP17am0001001; to H. Mano) and from the Project for Cancer Research and NTRK1 in MSI-H colorectal cancer cells reduced their prolifera- Therapeutic Evolution (P-CREATE, JP17cm0106502; to M. Kawazu, M. Soda, tion, further supporting the role of NTRK1 fusions as clinically H. Aburatani, and T. Watanabe) from the Japan Agency for Medical Research actionable (45). and Development. K. Sato is a research fellow of JSPS. Fourth, we detected those biological pathways that were affect- fi The costs of publication of this article were defrayed in part by the payment of ed by many mutations. A unique nding of our study was the page charges. This article must therefore be hereby marked advertisement in fi identi cation of the DNA damage-sensing pathway based on the accordance with 18 U.S.C. Section 1734 solely to indicate this fact. genetic alterations in MSI-H colorectal cancer genomes, which is of particular interest given that the DNA double-strand break Received May 20, 2018; revised July 31, 2018; accepted September 27, 2018; repair machinery is closely associated with MMR machinery (46) published first October 2, 2018.

References 1. Thibodeau S, Bren G, Schaid D. Microsatellite instability in cancer of the KRAS with survival after recurrence in stage III colon cancers: a second- proximal colon. Science 1993;260:816–9. ary analysis of 2 randomized clinical trials. JAMA Oncol 2017;3: 2. Aaltonen L, Peltomaki P, Leach F, Sistonen P, Pylkkanen L, Mecklin J, et al. 472–80. Clues to the pathogenesis of familial colorectal cancer. Science 1993; 5. Guinney J, Dienstmann R, Wang X, de Reynies A, Schlicker A, Soneson C, 260:812–6. et al. The consensus molecular subtypes of colorectal cancer. Nat Med 3. Boland CR, Goel A. Microsatellite instability in colorectal cancer. Gastro- 2015;21:1350–6. enterology 2010;138:2073–2087.e3. 6. Arnold CN, Goel A, Boland CR. Role of hMLH1 promoter hypermethyla- 4. Sinicrope FA, Shi Q, Allegra CJ, Smyrk TC, Thibodeau SN, Goldberg RM, tion in drug resistance to 5-ﬂuorouracil in colorectal cancer cell lines. et al. Association of DNA mismatch repair and mutations in BRAF and Int J Cancer 2003;106:66–73.

www.aacrjournals.org Clin Cancer Res; 2018 OF11

Sato et al.

7. Toyota M, Ho C, Ahuja N. Identification of differentially methylated 27. Alexandrov LB, Stratton MR. Mutational signatures: the patterns of somatic sequences in colorectal cancer by methylated CpG island amplification. mutations hidden in cancer genomes. Curr Opin Genet Dev 2014;24: Cancer Res 1999;59:2307–12. 52–60. 8. Weisenberger DJ, Siegmund KD, Campan M, Young J, Long TI, Faasse MA, 28. Helleday T, Eshtad S, Nik-Zainal S. Mechanisms underlying mutational et al. CpG island methylator phenotype underlies sporadic microsatellite signatures in human cancers. Nat Rev Genet 2014;15:585–98. instability and is tightly associated with BRAF mutation in colorectal 29. The Cancer Genome Atlas Network. Comprehensive molecular character- cancer. Nat Genet 2006;38:787–93. ization of human colon and rectal cancer. Nature 2012;487:330–7. 9. Li H, Durbin R. Fast and accurate short read alignment with Burrows- 30. Huang DW, Sherman BT, Lempicki RA. Systematic and integrative analysis Wheeler transform. Bioinformatics 2009;25:1754–60. of large gene lists using DAVID bioinformatics resources. Nat Protoc 10. Yoon S, Xuan Z, Makarov V, Ye K, Sebat J. Sensitive and accurate detection 2009;4:44–57. of copy number variants using read depth of coverage. Genome Res 31. Huang DW, Sherman BT, Lempicki RA. Bioinformatics enrichment tools: 2009;19:1586–92. paths toward the comprehensive functional analysis of large gene lists. 11. Beroukhim R, Getz G, Nghiemphu L, Barretina J, Hsueh T, Linhart D, et al. Nucleic Acids Res 2009;37:1–13. Assessing the significance of chromosomal aberrations in cancer: 32. Dienstmann R, Vermeulen L, Guinney J, Kopetz S, Tejpar S, Tabernero J. methodology and application to glioma. Proc Natl Acad Sci U S A 2007; Consensus molecular subtypes and the evolution of precision medicine in 104:20007–12. colorectal cancer. Nat Rev Cancer 2017;17:79–92. 12. Mermel CH, Schumacher SE, Hill B, Meyerson ML, Beroukhim R, Getz G. 33. Nakaoku T, Tsuta K, Ichikawa H, Shiraishi K, Sakamoto H, Enari M, et al. GISTIC2.0 facilitates sensitive and confident localization of the targets of Druggable oncogene fusions in invasive mucinous lung adenocarcinoma. focal somatic copy-number alteration in human cancers. Genome Biol Clin Cancer Res 2014;20:3087–93. 2011;12:1–14. 34. Mulvihill MJ, Cooke A, Rosenfeld-Franklin M, Buck E, Foreman K, Landfair 13. Kodama Y, Mashima J, Kosuge T, Katayama T, Fujisawa T, Kaminuma E, D, et al. Discovery of OSI-906: a selective and orally efficacious dual et al. The DDBJ Japanese genotype-phenotype archive for genetic and inhibitor of the IGF-1 receptor and insulin receptor. Future Med Chem phenotypic human data. Nucleic Acids Res 2015;43:D18–22. 2009;1:1153–71. 14. Jass JR, Walsh MD, Barker M, Simms LA, Young J, Leggett BA. Distinction 35. Kim HS, Jung M, Kang HN, Kim H, Park CW, Kim SM, et al. Oncogenic between familial and sporadic forms of colorectal cancer showing DNA BRAF fusions in mucosal melanomas activate the MAPK pathway and are microsatellite instability. Eur J Cancer 2002;38:858–66. sensitive to MEK/PI3K inhibition or MEK/CDK4/6 inhibition. Oncogene 15. Wright CM, Dent OF, Newland RC, Barker M, Chapuis PH, Bokey EL, et al. 2017;36:3334–45. Low level microsatellite instability may be associated with reduced cancer 36. Hutchinson KE, Lipson D, Stephens PJ, Otto G, Lehmann BD, Lyle PL, et al. specific survival in sporadic stage C colorectal carcinoma. Gut 2005; BRAF fusions define a distinct molecular subset of melanomas with 54:103–8. potential sensitivity to MEK Inhibition. Clin Cancer Res 2013;19: 16. Lee SY, Kim D-W, Lee HS, Ihn MH, Oh H-K, Min BS, et al. Low-level 6696–702. microsatellite instability as a potential prognostic factor in sporadic colo- 37. Turski ML, Vidwans SJ, Janku F, Garrido-Laguna I, Munoz J, Schwab R, et al. rectal cancer. Medicine 2015;94:e2260. Genomically driven tumors and actionability across histologies: BRAF- 17. Bocker T, R~aschoff J. Diagnostic microsatellite instability: definition and mutant cancers as a paradigm. Mol Cancer Ther 2016;15:533–47. correlation with mismatch repair protein expression. Cancer Res 1997; 38. Kodama T, Tsukaguchi T, Satoh Y, Yoshida M, Watanabe Y, Kondoh O, et al. 57:4749–56. Alectinib Shows Potent Antitumor Activity against RET-Rearranged Non- 18. Tahara T, Yamamoto E, Suzuki H, Maruyama R, Chung W, Garriga J, et al. Small Cell Lung Cancer. Mol Cancer Ther 2014;13:2910–8. Fusobacterium in colonic flora and molecular features of colorectal 39. Ardini E, Menichincheri M, Banfi P, Bosotti R, DePonti C, Pulci R, et al. carcinoma. Cancer Res 2014;74:1311–8. Entrectinib, a Pan-TRK, ROS1, and ALK inhibitor with activity in multiple 19. Mima K, Cao Y, Chan AT, Qian ZR, Nowak JA, Masugi Y, et al. Fusobacter- molecularly defined cancer indications. Mol Cancer Ther 2016:15:628–39. ium nucleatum in colorectal carcinoma tissue according to tumor location. 40. Grisham RN, Sylvester BE, Won H, McDermott G, DeLair D, Ramirez R, Clin Transl Gastroenterol 2016;7:e200. et al. Extreme outlier analysis identifies occult mitogen-activated protein 20. Herman JG, Umar A, Polyak K, Graff JR, Ahuja N, Issa JP, et al. Incidence kinase pathway mutations in patients with low-grade serous ovarian and functional consequences of hMLH1 promoter hypermethylation in cancer. J Clin Oncol 2015;33:4099–105. colorectal carcinoma. Proc Natl Acad Sci U S A 1998;95:6870–5. 41. Tran B, Kopetz S, Tie J, Gibbs P, Jiang ZQ, Lieu CH, et al. Impact of BRAF 21. Mensenkamp AR, Vogelaar IP, Van Zelst-Stams WAG, Goossens M, mutation and microsatellite instability on the pattern of metastatic spread Ouchene H, Hendriks-Cornelissen SJB, et al. Somatic mutations in MLH1 and prognosis in metastatic colorectal cancer. Cancer 2011;117:4623–32. and MSH2 are a frequent cause of mismatch-repair deficiency in lynch 42. Godstein J, Tran B, Ensor J, Gibbs P, Wong HL, Wong SF, et al. Multicenter syndrome-like tumors. Gastroenterology 2014;146:643–646. retrospective analysis of metastatic colorectal cancer (CR) with high-level 22. Futreal P, Coin L, Marshall L, Down T, Hubbard T, Wooster T, et al. A census microsatellite instability (MSI-H). Ann Oncol 2014;25:1032–8. of human cancer genes. Nat Rev Cancer 2004;4:177–83. 43. Venderbosch S, Nagtegaal ID, Maughan TS, Smith CG, Cheadle JP, Fisher 23. Kim T, Laird PW, Park PJ. The landscape of microsatellite instability in D, et al. Mismatch repair status and BRAF mutation status in metastatic colorectal and endometrial cancer genomes. Cell 2013;155:858–68. colorectal cancer patients: a pooled analysis of the CAIRO, CAIRO2, COIN, 24. Lawrence MS, Stojanov P, Polak P, Kryukov G V.Cibulskis K, Sivachenko A, and FOCUS studies. Clin Cancer Res 2014;20:5322–30. et al. Mutational heterogeneity in cancer and the search for new cancer- 44. Taylor CF, Charlton RS, Burn J, Sheridan E, Taylor GR. Genomic deletions associated genes. Nature 2013;499:214–8. in MSH2 or MLH1 are a frequent cause of hereditary non-polyposis 25. Bessa X, Balleste B, Andreu M, Castells A, Bellosillo B, Balaguer F, et al. A colorectal cancer: Identification of novel and recurrent deletions by MLPA. prospective, multicenter, population-based study of BRAF mutational Hum Mutat 2003;22:428–33. analysis for lynch syndrome screening. Clin Gastroenterol Hepatol 45. Vaishnavi A, Capelletti M, Le AT, Kako S, Butaney M, Ercan D, et al. 2008;6:206–14. Oncogenic and drug-sensitive NTRK1 rearrangements in lung cancer. 26. Wang L, Cunningham JM, Winters JL, Guenther JC, French AJ, Boardman Nat Med 2013;19:1469–72. LA, et al. BRAF Mutations in colon cancer are not likely attributable to 46. Spies M, Fishel R. Mismatch repair during homologous and homeologous defective DNA mismatch repair. Cancer Res 2003;63:5209–12. recombination. Cold Spring Harb Perspect Biol 2015;7:a022657.

OF12 Clin Cancer Res; 2018 Clinical Cancer Research

Fusion Kinases Identified by Genomic Analyses of Sporadic Microsatellite Instability −High Colorectal Cancers

Kazuhito Sato, Masahito Kawazu, Yoko Yamamoto, et al.

Clin Cancer Res Published OnlineFirst October 2, 2018.

Updated version Access the most recent version of this article at: doi:10.1158/1078-0432.CCR-18-1574

Supplementary Access the most recent supplemental material at: Material http://clincancerres.aacrjournals.org/content/suppl/2018/10/02/1078-0432.CCR-18-1574.DC1

E-mail alerts Sign up to receive free email-alerts related to this article or journal.

Reprints and To order reprints of this article or to subscribe to the journal, contact the AACR Publications Subscriptions Department at [email protected].

Permissions To request permission to re-use all or part of this article, use this link http://clincancerres.aacrjournals.org/content/early/2018/11/13/1078-0432.CCR-18-1574. Click on "Request Permissions" which will take you to the Copyright Clearance Center's (CCC) Rightslink site.