Integrated Molecular Characterization of the Lethal

Home , Esophageal cancer

Author Manuscript Published OnlineFirst on December 12, 2017; DOI: 10.1158/0008-5472.CAN-17-2581 Author manuscripts have been peer reviewed and accepted for publication but have not yet been edited.

1 Integrated molecular characterization of the lethal pediatric cancer 2 pancreatoblastoma 3 4 Tomoya Isobe1, Masafumi Seki1, Kenichi Yoshida2, Masahiro Sekiguchi1, Yusuke Shiozawa1,2, 5 Yuichi Shiraishi3, Shunsuke Kimura1,4, Misa Yoshida1,5, Yoshikage Inoue2, Akira Yokoyama2, 6 Nobuyuki Kakiuchi2, Hiromichi Suzuki2, Keisuke Kataoka2, Yusuke Sato2, Tomoko Kawai6, 7 Kenichi Chiba3, Hiroko Tanaka3, Teppei Shimamura7, Motohiro Kato8, Akihiro Iguchi9, 8 Asahito Hama10, Tomoaki Taguchi11, Masaharu Akiyama12, Junya Fujimura13, Akiko Inoue14, 9 Tsuyoshi Ito15, Takao Deguchi16, Chikako Kiyotani8, Tomoko Iehara17, Hajime Hosoi17, 10 Akira Oka1, Masashi Sanada18, Yukichi Tanaka5, Kenichiro Hata6, Satoru Miyano3, Seishi 11 Ogawa2, Junko Takita1 12 13 1Department of Pediatrics, Graduate School of Medicine, The University of Tokyo, Tokyo, 14 Japan; 2Department of Pathology and Tumor Biology, Graduate School of Medicine, Kyoto 15 University, Kyoto, Japan; 3Laboratory of DNA Information Analysis, Human Genome Center, 16 Institute of Medical Science, The University of Tokyo, Tokyo, Japan; 4Department of 17 Pediatrics, Hiroshima University Graduate School of Biomedical Sciences, Hiroshima, 18 Japan; 5Department of Pathology, Kanagawa Children’s Medical Center, Yokohama, Japan; 19 6Department of Maternal-Fetal Biology, National Research Institute for Child Health and 20 Development, Tokyo, Japan; 7Division of Systems Biology, Nagoya University Graduate 21 School of Medicine, Nagoya, Japan; 8Department of Pediatric Hematology and Oncology 22 Research, National Research Institute for Child Health and Development, Tokyo, Japan; 23 9Department of Pediatrics, Hokkaido University, Sapporo, Japan; 10Department of 24 Pediatrics, Nagoya University Graduate School of Medicine, Nagoya, Japan; 11Department 25 of Pediatric Surgery, Graduate School of Medicine, Kyushu University, Fukuoka, Japan; 26 12Department of Pediatrics, Jikei University School of Medicine, Tokyo, Japan; 27 13Department of Pediatrics, Juntendo University School of Medicine, Tokyo, Japan; 28 14Department of Pediatrics, Osaka Medical College, Osaka, Japan; 15Department of 29 Pediatrics, Toyohashi Municipal Hospital, Toyohashi, Japan; 16Department of Pediatrics, 30 Mie University Graduate School of Medicine, Tsu, Japan; 17Department of Pediatrics, 31 Kyoto Prefectural University of Medicine, Graduate School of Medical Science, Kyoto, 32 Japan; 18Clinical Research Center, National Hospital Organization Nagoya Medical Center, 33 Nagoya, Japan. 34 35 Running Title 36 Integrated Molecular Characterization of PBL 37 1

Downloaded from cancerres.aacrjournals.org on September 30, 2021. © 2017 American Association for Cancer Research. Author Manuscript Published OnlineFirst on December 12, 2017; DOI: 10.1158/0008-5472.CAN-17-2581 Author manuscripts have been peer reviewed and accepted for publication but have not yet been edited.

38 Corresponding Author 39 Junko Takita 40 Department of Pediatrics, 41 The University of Tokyo, 42 Hongo 7-3-1. Bunkyo-ku, 43 Tokyo, 113-8655 Japan 44 TEL: +81-3-3815-5411 (Ext. 33462) 45 FAX: +81-3-3816-4108 46 E-mail: [email protected] 47 48 Disclosure of Potential Conflicts of Interest 49 The authors declare no potential conflicts of interest. 50 51 Notes 52 Our manuscript contains 147 words of abstract, 5058 words of main text, 5 figures, no 53 tables, 10 supplementary figures, and 13 supplementary tables. 54 55

56 Abstract 57 Pancreatoblastoma (PBL) is a rare pediatric pancreatic malignancy for which the 58 molecular pathogenesis is not understood. In this study, we report the findings of an 59 integrated multi-omics study of whole exome and RNA sequencing as well as 60 genome-wide copy number and methylation analyses of 10 PBL cases. The PBL genome 61 was characterized by a high frequency of aberrant activation of the Wnt signaling 62 pathway, either via somatic mutations of CTNNB1 (90%) and copy-neutral loss of 63 heterozygosity (CN-LOH) of APC (10%). In addition, imprinting dysregulation of IGF2 as a 64 consequence of CN-LOH (80%), gain of paternal allele (10%), and gain of methylation 65 (10%) were universally detected. At the transcriptome level, PBL exhibited an expression 66 profile characteristic of early pancreas progenitor-like cells along with upregulation of 67 the R-spondin/LGR5/RNF43 module. Our results offer a comprehensive description of the 68 molecular basis for PBL and highlight rational therapeutic targets for its treatment.

70 Introduction

71 Pancreatoblastoma (PBL) is a very rare pancreatic solid tumor that typically affects 72 young children, with a median age at diagnosis of 5 years. Including much less frequent 73 adult cases, PBL comprises less than 1% of all pancreatic non-endocrine tumors; despite 74 its rarity, it is the most common malignant pancreatic tumor in children younger than 10 75 years (1). Morphologically, PBL resembles fetal pancreatic tissue at gestational age of 8 76 weeks, manifesting multi-lineage elements with acinar, ductal and endocrine 77 differentiation, which suggests a primitive cellular origin of PBL (2). Owing to its rarity, 78 optimized therapeutic strategies, including targeted therapies, have not been 79 established for PBL; only the complete surgical resection is of proven prognostic value. 80 Thus, children with unresectable or relapsed disease still have a very poor prognosis (3), 81 which prompts a need for novel therapeutic modalities based on a better understanding 82 of its molecular pathogenesis.

83 Several case studies have revealed recurrent alterations affecting the Wnt/β-catenin 84 pathway genes, including APC and CTNNB1 (encoding β-catenin), and a loss of 85 heterozygosity (LOH) on chromosome 11p in a subset of sporadic PBL (4,5). 86 Correspondingly, PBL cases associated with familial adenomatous polyposis and 87 Beckwith-Wiedemann syndrome (BWS) have also been reported (4,6). By contrast, except 88 for only a few adult PBL cases exhibiting inactivation of SMAD4 (7), mutations affecting 89 KRAS, TP53, CDKN2A, and SMAD4, the four common targets of oncogenic events in 90 pancreatic ductal adenocarcinoma (PDAC), have not been reported in PBL (4). These

91 clinicopathological and genetic features suggest a distinctive oncogenic mechanism of 92 this disease. However, because no comprehensive genomic studies have been reported 93 thus far, our knowledge about the molecular basis of PBL is still very limited.

94 In this study, we analyzed 16 primary or metastatic tumor specimens from 10 cases 95 with PBL using targeted deep sequencing of 155 cancer driver genes, whole exome 96 sequencing (WES), RNA sequencing, and genome-wide copy number and methylation 97 analyses. This multi-omics approach allowed us to obtain a comprehensive registry of the 98 molecular lesions underlying the pathogenesis of PBL.

100 Materials and Methods

101 Subjects and samples 102 This study comprised 16 tumor samples from 10 Japanese patients with PBL, among 103 which fresh frozen tumors were obtained from 6 patients, and for the remaining 4 104 patients, formalin-fixed paraffin-embedded (FFPE) samples were procured. Written 105 informed consent was obtained according to protocols approved by the Human Genome, 106 Gene Analysis Research Ethics Committee of the University of Tokyo, and other 107 participating institutes. Matched peripheral blood samples were obtained and used as 108 germline controls for WES and amplicon deep sequencing in 6 cases. For PBL001, a total 109 of 7 samples (one from the primary, two from the second, and four from the third 110 surgery) were included. The diagnosis of PBL was centrally reviewed and confirmed; 111 staging was done based on WHO classification. All patients underwent surgery with or 112 without chemotherapy and radiation therapy. One case (PBL001) had been treated for the 113 progressive disease with several different regimens, including temozolomide (TMZ), 114 between the first and second surgeries. See Supplementary Table S1 for clinical 115 characteristics and experimental design summary. 116 117 Targeted deep sequencing and mutation calling 118 For fresh frozen samples, sequencing libraries were constructed using a SureSelect-XT 119 kit (Agilent Technologies) according to the manufacturer’s protocol. For FFPE samples, a 120 KAPA Hyper Prep kit (Kapa Biosystems) was used with a modified protocol to utilize the 121 adapter and PCR primers of an Agilent SureSelect-XT kit. Target enrichment was 122 performed using a SureSelect custom bait library that we previously designed to 123 investigate gene mutations in gastrointestinal cancers (Inoue, et al., in preparation). The 124 bait library was designed to include all coding exons of 127 genes, hotspot regions of 24 125 genes, and promoter regions of TERT, PLEKHS1, SDHD, and WDR74 (Supplementary Table

126 S2). The total of 155 targeted genes included (i) genes that were recurrently mutated or 127 affected by copy number alterations in colorectal cancer (8-11), esophageal cancer (12), 128 and hepatocellular carcinoma (13,14); (ii) genes that were reported to interact with or be 129 in the same family as recurrently altered genes; (iii) genes whose promoter regions were 130 reported to be recurrently mutated in human cancers (15); and (iv) genes that were 131 reported to be involved in growth signaling or epigenetic regulation. 132 Sequence alignment and mutation calling were performed using the Empirical 133 Bayesian Mutation Calling (EBCall) algorithm (16) implemented in our in-house Genomon 134 v2.3.0 pipeline with the following parameters: (i) Mapping Quality Score ≥20; (ii) Base 135 Quality Score ≥15; (iii) depths in both tumor and normal ≥8; (iv) number of variant read 136 pairs in the tumor ≥4; (v) variant allele frequencies (VAFs) in tumor samples ≥0.10; and 137 (vi) EBCall P-value ≤ 10−5. The following were further excluded: (i) synonymous single 138 nucleotide variants (SNVs); (ii) mutations detected on only either the plus or minus 139 strand; (iii) mutations in HLA genes; (iv) SNVs listed in the NCBI dbSNP build 131 or our 140 in-house single nucleotide polymorphism (SNP) database (SNVs with more than ten entries 141 in the Catalogue Of Somatic Mutations In Cancer (COSMIC) v70 were not excluded); (v) 142 variants with 0.45 ≤ VAF ≤ 0.55 in copy neutral regions; and (vi) variants detected in 143 paired normal samples, if available, by Sanger sequencing. 144 145 Whole-exome sequencing and mutation calling 146 For WES, DNA libraries were prepared using SureSelect Human All Exon V5 (Agilent 147 Technologies) according to the manufacturer’s protocol. Enriched exome libraries were 148 sequenced on an Illumina Hiseq 2000 platform using 100-bp paired-end mode. Somatic 149 variants were called by Genomon v2.3.0 with the following parameters: (i) Mapping 150 Quality Score ≥20; (ii) Base Quality Score ≥15; (iii) depths in both tumor and normal ≥8; 151 (iv) number of variant reads in the tumor ≥4; (v) variant allele frequencies (VAFs) in 152 tumor samples ≥0.02; (vi) VAFs in normal samples <0.10; (vii) EBCall P-value ≤ 10−5; and 153 (viii) Fisher’s exact P-value ≤ 10−2. We further excluded (i) synonymous SNVs; and (ii) SNVs 154 listed in the NCBI dbSNP build 131 or our in-house SNP database. For subpopulation 155 detection and phylogenetic analysis of multiple samples from PBL001, synonymous SNVs 156 were retained and also subjected to the following validation steps in this case. 157 158 Validation of candidate mutations 159 Candidate mutations were validated by PCR-based amplicon deep sequencing (depth 160 ≥1,000×) using matched normal samples. For PCR amplification, a NotI restriction site 161 was attached to each primer as a linker sequence. Amplified products were digested with

162 NotI, ligated, fragmented, and then used for deep sequencing library preparation as 163 previously described (17). Candidate variants that satisfied all the following criteria were 164 considered to be validated: (i) VAFs in tumor samples ≥0.02; (ii) VAFs in the paired normal 165 samples <0.02; (iii) the VAF in a tumor sample was at least 5 times the VAF in the paired 166 normal sample; (iv) the sequencing depth ≥1000. In total, 311 of 324 candidate non-silent 167 mutations (96.0%) were validated. 168 169 Multi-regional analysis of PBL001 170 Intra-sample subpopulations were estimated using a hierarchical bayesian model 171 implemented in PyClone software (18). Mutations on chromosomal regions with copy 172 number gains were not used for the analysis because mutant cell fractions may be 173 underestimated. The maximum parsimony tree representing the clonal evolution in 174 PBL001 was constructed using MEGA v6.0.6 software with branch-and-bound algorithm 175 (19). Mutational signatures were extracted from validated somatic mutations. 176 177 Targeted deep sequencing-based copy number analysis 178 Our custom bait library for targeted deep sequencing also included additional baits for 179 1,760 SNPs to generate genome-wide allele-specific copy number profiles 180 (Supplementary Table S3). Allele-specific copy numbers were calculated by allele 181 frequencies and sequenced depths of SNPs using CNACS (Shiozawa, et al., in preparation) 182 as previously described (20). Total copy numbers were computed using off-target reads 183 from targeted deep sequencing data as implemented in the Bioconductor package 184 CopywriteR (21). The copy number heatmap was generated using the Integrative 185 Genomics Viewer (IGV). 186 187 SNP array analysis 188 SNP array was performed on fresh frozen tumors and paired normal samples from 6 189 patients using CytoScan HD Array (Affymetrix) according to the manufacturer’s protocol. 190 The array data were analyzed for copy number alterations by using CNAG software 191 (22,23). 192 193 Bisulfite conversion and Sanger sequencing of ICR1 194 Bisulfite conversion of 200–500 ng of DNA samples from PBL and normal pancreatic 195 tissues was performed using the EpiTect Fast Bisulfite Conversion Kits (Qiagen) according 196 to the manufacturer’s instructions. By using bisulfite-treated DNA as PCR templates, ICR1 197 locus (chr11: 2020978-2021251, hg19) was amplified with the following primers:

198 5’-GATGGTAYGGAATTGGTTGTAG-3’ (forward) and 5’-AACTTAAATCCCAAACCATAACA-3’ 199 (reverse). PCR was performed with the following thermal cycling conditions: a 95°C for 5 200 min; 35 cycles of 95°C for 30 s, 58°C for 30 s and 72°C for 30 s; and a 72°C for 7 min. The 201 same primer pair was also used for Sanger sequencing. For all samples examined, the 202 tumor purity was estimated from targeted deep sequencing or, if available, WES data 203 using the R package FACETS (24). Accounting for the estimated purity, copy-neutral LOH 204 (CN-LOH) positive samples with C/T signal ratios of 2×purity+(1-purity)

1-purity 205 or higher were considered to be homozygously methylated. 206 207 Semi-quantitative reverse transcription PCR of IGF2 208 Complementary DNA (cDNA) was synthesized from 200 ng of total RNA samples using 209 SuperScript VILO cDNA Synthesis Kit (Thermo Fisher Scientific) according to the 210 manufacturer’s protocol. From 20 µL of diluted cDNA, 1 µL of cDNA was used as PCR 211 template. IGF2 and ACTB (as a reference gene) were amplified with the following 212 primers: 5’-CTGTTCGGTTTGCGACAC-3’ (forward) and 5’-AAGCACCAGCATCGACTTC-3’ 213 (reverse) for IGF2; 5’-AACCGCGAGAAGATGACC-3’ (forward) and 214 5’-AGGCGTACAGGGATAGCAC-3’ (reverse) for ACTB. PCR was performed with the 215 following thermal cycling conditions: a 95°C for 5 min; 35 cycles of 95°C for 30 s, 58°C for 216 30 s and 72°C for 30 s; and a 72°C for 7 min. DNA bands were quantified using Image Lab 217 software (Bio-Rad). 218 219 Immunohistochemistry 220 FFPE tumor specimens were deparaffinized, rehydrated, and immersed in 0.3% 221 hydrogen peroxide in methanol for 30 minutes to block the endogenous peroxidase 222 activity. The 14/Beta-Catenin monoclonal antibody (BD Transduction Laboratories) and 223 peroxidase-conjugated anti-mouse immunoglobulin antibody (Medical & Biological 224 Laboratories) were used as primary and secondary antibodies for detecting β-catenin. 225 Antigen retrieval was performed by microwave treatment for two 10-minute cycles with 226 180 W of power in 1 mM EDTA (pH 8.0). Final visualization was carried out by development 227 in PBS (pH 7.6) containing 20 mg/100 mL of diaminobenzidine tetrahydrochloride and

228 0.1% H2O2 for 10 minutes. Sections incubated with PBS instead of the primary antibodies 229 served as negative controls. 230 231 Methylation Array

232 DNA methylation profiles were analyzed using Infinium MethylationEPIC BeadChip 233 (Illumina), according to the manufacturer’s protocol. Two genomic DNA products 234 (ZYAGEN: #HG-313; BioChain: #D1234188-A712202) were used as normal pancreatic 235 control. Raw signal intensity data were processed using the Greedycut algorithm 236 implemented in the Bioconductor package RnBeads (25). β-values were normalized with 237 beta-mixture quantile normalization method and global distributions of β-values were 238 plotted for PBL and normal pancreas, respectively. After evaluating promoter differential 239 methylation using M-values, gene promoters were ranked on the basis of significance of 240 the differential methylation and over-represented gene ontology (GO) biological 241 process/molecular function terms were identified by hypergeometric test using RnBeads. 242 GO terms with hypergeometric P ≤ 0.01 were considered significantly differentially 243 methylated. 244 245 RNA sequencing 246 Total RNA was extracted from fresh frozen tumor samples and was assessed for 247 integrity and concentration using an Agilent 2100 Bioanalyzer or an Agilent 4200 248 TapeStation system. For 7 samples from 5 cases with RNA integrity number higher than 249 5.0, RNA sequencing libraries were constructed using TruSeq RNA Library Preparation Kit 250 v2 (Illumina) or NEBNext Ultra RNA Library Prep kit for Illumina (New England BioLabs). 251 Alignment to the reference genome (hg19) and fusion detection were conducted by 252 Genomon v2.3.0 with the following criteria: (i) fusion transcripts of two different genes; 253 (ii) at least three spanning reads; (iii) with junctions located at known exon–intron 254 boundaries. Detected fusions were validated by RT-PCR and Sanger sequencing. For 255 expression analysis, mapped reads were counted for each gene by our in-house 256 GenomonExpression pipeline 257 (http://github.com/Genomon-Project/GenomonExpression). Gene expression level 258 normalization and differential expression analysis were conducted using the Bioconductor 259 package DESeq2 (26). For the 5 normal controls, we sequenced 3 pancreatic total RNA 260 products (Agilent: #540023; BioChain: #R1234188-50-B712067 and 261 #R1234188-50-B103010) from different individuals and also utilized controlled-access 262 normal pancreatic RNA sequencing data of 2 different people provided by the 263 Genotype-Tissue Expression (GTEx) Portal. Genes with Benjamini-Hochberg adjusted P 264 values ≤ 0.10 were considered significantly differentially expressed. As gene-set 265 enrichment scores, weighted Kolmogorov–Smirnov-like statistics were calculated and 266 empirical permutation tests by shuffling samples’ group labels were performed to 267 evaluate the significance levels (P values) of enrichment scores as implemented in the

268 Bioconductor package SeqGSEA (27). Gene sets with P ≤ 0.01 were considered 269 significantly enriched. Pathway enrichment in hierarchically clustered gene groups was 270 evaluated using Metascape pathway analysis resource (http://metascape.org). Principal 271 component analysis was performed using the 500 genes with the highest 272 sample-to-sample variances and plotted using R. 273 274 Data availability 275 All WES, RNA sequencing, SNP array, and methylation array data have been deposited 276 in the Japanese Genotype-phenotype Archive (JGA), which is hosted by the DNA Data 277 Bank of Japan (DDBJ), under accession code JGAS00000000088. 278

279 Results

280 Mutational landscape of PBL

281 Our PBL cohort comprised 10 Japanese patients, with a median age at diagnosis of 3 282 years (range, 3 days-12 years) and a male:female ratio of 1.5:1. All 3 patients with 283 distant metastases had died of the disease. The remaining 7 patients with localized 284 disease achieved complete surgical resections of the tumors; none of the 7 patients had 285 died during the follow-up period. Additional clinical features are summarized in 286 Supplementary Table S1. To identify driver mutations in PBL, we first performed targeted 287 deep sequencing for 155 known gastrointestinal cancer driver genes (Supplementary 288 Table S2), on 9 primary and 1 relapsed tumors from the 10 PBL cases. The relapsed tumor 289 sequenced was from PBL009, for whom only a relapsed tumor was available. The mean 290 sequencing coverage was 677×, with which 99% of the targeted regions were sequenced 291 at a depth more than 100× (Supplementary Fig. S1A and B). Using our in-house Genomon 292 v2.3.0 pipeline (https://github.com/Genomon-Project/GenomonPipeline), 14 mutations 293 in 6 genes were identified (Supplementary Table S4), of which mutated in 9 of the 10 294 cases, CTNNB1 was the only recurrently mutated gene. All the CTNNB1 mutations were 295 located in exon 3 (Fig. 1A), where glycogen synthase kinase 3 (GSK3) and caseine kinase 1 296 (CK1) recognize and phosphorylate β-catenin, leading to the stabilization of β-catenin, as 297 it prevents from interacting with the APC/Axin destruction complex (28). Accordingly, 298 immunohistochemistry demonstrated the aberrant nuclear accumulation of β-catenin, a 299 molecular hallmark of activated Wnt/β-catenin signaling, in all CTNNB1 300 mutation-positive cases (Fig. 1B).

301 To further explore additional mutational hits in PBL, we performed WES on paired 302 tumor-normal DNA samples, which were available in 6 of the 10 cases. In a case with

303 refractory disease (PBL001), 7 spatially separated tumor samples resected at 3 different 304 time points were analyzed (Supplementary Table S1). The mean sequencing depth was 305 129×, with which 98% of the exome regions were analyzed at a depth more than 20× 306 (Supplementary Fig. S2A and B). In total, 324 non-silent somatic alterations were 307 detected, of which 311 in 240 genes (96%) were validated by amplicon deep sequencing 308 (Supplementary Table S5). As seen in other pediatric cancers, the mean number of 309 mutations in primary PBL samples was quite low (8.7 mutations per case; Supplementary 310 Fig. S2C). Conspicuously, PBL002, a case of congenital PBL, exhibited a CTNNB1 hotspot 311 mutation (p.T41A) as the only mutation in the whole exome. Overall, even with the 312 exome-wide approach, no recurrent alterations other than CTNNB1 mutations were 313 identified across the cases, reinforcing its pathogenic significance in PBL.

314 Contrary to the low mutational burden in primary tumors, all metastatic samples from 315 PBL001 had substantially elevated numbers of mutations (45 mutations per sample; 316 Supplementary Fig. S2D). Multi-regional sequencing analysis at different time points 317 allowed us to delineate the clonal history of these tumor cells. In this particular case, 303 318 mutations were detected in at least one region, and a short phylogenetic trunk containing 319 13 mutations (4%) represented the major population of the primary tumor, from which all 320 metastatic tumors seemed to diverge (Fig. 2A and B). Rapidly acquiring intra- and 321 inter-tumor heterogeneity, 283 out of 303 mutations (93%) were private to a single region 322 (Supplementary Fig. S3; Supplementary Table S6). Noteworthy is a novel nonsense 323 mutation of KDM6A (p.E92X), observed as one of the 13 truncal mutations in PBL001. 324 Considering that inactivation of KDM6A is associated with a poor-prognosis subtype of 325 PDAC (29), this putative driver mutation may have conferred an aggressive trait on this 326 case.

327 As an overall WES-based mutational signature, C>T transitions at CpG sites were the 328 most frequent in the 6 cases of primary samples, while the metastatic samples 329 accumulated C>T substitutions at CpC and CpT dinucleotides (Supplementary Fig. S4A and 330 B), which is reported as a TMZ treatment-associated mutational signature (30). This 331 result may reflect the treatment of metastatic tumors with TMZ before the second and 332 third surgeries for PBL001, though it should be addressed in future studies.

333

334 IGF2 imprinting ubiquitously disrupted by different genetic/epigenetic mechanisms in 335 PBL

336 Next we investigated genome-wide copy number alterations in PBL using either or both 337 SNP array and targeted deep sequencing data (Fig. 3A; Supplementary Fig. S5; 10

338 Supplementary Table S3; Methods). We detected a total of 232 segments showing 339 abnormal copy number or allelic imbalance, of which most frequently affected was 340 chromosome 11p (9/10), by CN-LOH (n = 8) or copy number gain (n = 1), followed by gain 341 of whole or partial chromosomes 8, 10, and 12. These CN-LOH and copy number gain 342 invariably affected an imprinted gene cluster region at 11p15.5 (Fig. 3B; Supplementary 343 Fig. S6A and B), where the paternal allele expresses IGF2 and the maternal allele 344 expresses H19 via differential methylation and the following binding of CCCTC-binding 345 factor (CTCF) on Imprinting control region 1 (ICR1; (31)). Bisulfite sequencing and tumor 346 purity-based methylated allele zygosity estimation (see Methods) revealed a homozygous 347 methylation of ICR1 in all 4 CN-LOH (+) cases tested (Fig. 3C; Supplementary Fig. S7A), 348 indicating that these CN-LOH cause paternal uniparental disomy implicated in 349 overexpression of IGF2 in BWS (31). Furthermore, bisulfite sequencing of the case with 350 gain of chromosome 11 (PBL009) demonstrated a gain of ICR1 methylated allele 351 (Supplementary Fig. S7B). Because partial trisomy 11 which causes paternal duplication 352 of 11p15.5 without loss of maternal allele confers BWS phenotype (32), gain of ICR1 353 methylated allele in PBL009 can be considered equivalent to CN-LOH at 11p15.5 observed 354 in other cases. Interestingly, homozygous methylation of ICR1 was also found in the 355 remaining one case (PBL005), who had no CN-LOH or copy number alterations at 11p15.5 356 (Fig. 3D). In the two cases without CN-LOH at 11p15.5 (PBL005 and PBL009), the elevated 357 expression of IGF2 was confirmed by semi-quantitative reverse transcription PCR 358 (Supplementary Fig. S8A and B). Taken together, different mechanisms of genetic and 359 epigenetic alterations targeting the gene imprinting on 11p15.5 were demonstrated in all 360 10 cases (100%), suggesting that aberrant imprinting at this locus play an essential role in 361 the development of PBL, most likely through overexpression of IGF2.

362 Additionally, copy number analysis in PBL009, the only case without CTNNB1 mutations, 363 identified a CN-LOH of 5q22 involving the APC locus (Supplementary Fig. S9A and B). 364 Although APC is a tumor suppressor gene thought to require bi-allelic inactivation for 365 constitutional activation of the Wnt signaling pathway, PBL009 exhibited neither 366 additional pathogenic germline nor somatic inactivating mutations in APC. As was the 367 case with CTNNB1 mutation-positive samples; however, immunohistochemistry on PBL009 368 demonstrated the nuclear accumulation of β-catenin (Supplementary Fig. S9C). This 369 affirmative evidence of Wnt/β-catenin signal activation suggested a different mechanism 370 serving as a ‘second hit’ on APC in this case.

371 As shown above, aberrant DNA methylation could play an important role in the 372 pathogenesis of PBL. Thus, we investigated genome-wide DNA methylation profiles in 4 373 primary PBL samples, in which a sufficient amount of DNA was available, using Infinium 11

374 MethylationEPIC BeadChip (Illumina). Basically, PBL samples (n = 4) manifested global 375 DNA hypomethylation profiles compared to normal pancreatic cells (n = 2; Supplementary 376 Fig. S10A and B). In addition, significant promoter hypomethylation (hypergeometric P ≤ 377 0.01) on the β-catenin/TCF transcription complex member genes and other downstream 378 DNA replication/cell division-related genes (Supplementary Table S7) was observed, 379 which may reflect the pathological activation of Wnt/β-catenin signaling. Since global 380 hypomethylation is reported in a variety of human cancers to contribute to cancer cell 381 development (33), these results suggest that global DNA hypomethylation in concert with 382 reduced DNA methylation on Wnt/β-catenin target genes, should drive the tumorigenesis 383 of PBL.

384

385 Gene expression profiling and signature-based cellular origin of PBL

386 To further illustrate the molecular basis of PBL, we conducted RNA sequencing on 7 387 tumors from 5 cases for which extracted RNA samples were of a decent enough quality 388 with integrity numbers greater than 5.0. Our fusion detection algorithm (Methods) 389 identified 2 fusion transcripts in the metastatic samples of PBL001 that were absent from 390 the corresponding primary sample (Supplementary Table S8), although the functional 391 significance of these fusions is still elusive. RNA sequencing-based gene expression 392 analysis identified 1274 genes with a highly significant differential expression (log2 fold 393 change ≤−2.5 or ≥2.5 and false discovery rate q ≤ 0.001; Fig. 4A; Supplementary Table S9). 394 Significantly overexpressed genes included negative-feedback regulators of Wnt pathway, 395 such as NOTUM and NKD1, as well as Wnt pathway effectors, LEF1 and TCF7. While 396 aberrantly imprinted IGF2 was validated as being overexpressed, KCNQ1, another 397 imprinted gene on 11p15.5, showed significantly reduced expression in PBL, though its 398 functional significance in the pathogenesis of PBL is still unclear. Among the genes listed, 399 of particular interest is LGR5, the third most significantly overexpressed gene on the 400 entire gene list. Because LGR5, a key molecule of the R-spondin/LGR5/RNF43 module, 401 positively enhances Wnt signal activity to govern cellular stemness (34,35), 402 overexpression of LGR5 can be considered to cause further intensification of mutant 403 β-catenin-driven Wnt signal activation. Importantly, other R-spondin/LGR5/RNF43 404 module genes (RSPO2, RSPO4, RNF43, and ZNRF3) were also markedly up-regulated in PBL 405 (Fig. 4A and Supplementary Table S9). Gene set enrichment analysis demonstrated 406 significant enrichments of cell cycle-related gene sets, cancer-related gene sets, and 407 stem cell regulator gene sets, as well as Wnt signaling pathway (permutation P ≤ 0.01; 408 Supplementary Table S10). In addition to significant overexpression of IGF2 and

409 pathway-level up-regulation of Wnt/β-catenin target genes, which are consistent with 410 genetic and epigenetic lesions, a significant enrichment of stem cell core modules is 411 accordant with the R-spondin/LGR5/RNF43 module overexpression (Fig. 4B).

412 To validate the molecular similarity between PBL and stem/progenitor cells, we also 413 compared their gene expression profiles using previously reported RNA sequencing data 414 of human embryonic stem cells (hESCs; (36)) and pancreatic multipotent progenitor cells 415 (pMPCs; (37)). The expression patterns of stemness-associated genes were similar among 416 PBL, hESCs and pMPCs (Fig. 4C), affirming the primitive cellular nature of PBL. As 417 previously well established in somatic stem cells which reside in adult organs/tissues such 418 as intestines (34), the R-spondin/LGR5/RNF43 module genes were highly expressed in 419 hESCs and pMPCs compared to mature pancreatic cells; of note, PBL exhibited an even 420 higher enhancement of these genes than in hESCs and pMPCs (Fig. 4D).

421 Finally, in order to gain a deeper insight into the identity of cellular origin of PBL, we 422 performed cluster and principal component analyses. Hierarchical clustering and 423 pathway enrichment analysis characterized the PBL transcriptome by identifying 6 gene 424 clusters: (i) leukocyte activation and early pancreas development-related genes (Cluster 425 1); (ii) genes reflecting later-stage pancreatic differentiation (Cluster 2); (iii) genes 426 related to the earliest stage of embryonic development (Cluster 3); (iv) genes 427 down-regulated in PBL (Cluster 4); (v) stem cell module genes (Cluster 5); and (vi) genes 428 up-regulated in PBL (Cluster 6) (Fig. 5A; Supplementary Table S11). In addition to 429 leukocyte-related genes which were not expressed in hESCs and enriched in Cluster 1 430 (Supplementary Table S12), PBL expressed core pMPC-specific transcription factors, 431 including PDX1, PTF1A, HNF1A, SOX6, and MEIS1 as Cluster 1 genes (37). Other pMPC 432 markers such as carboxypeptidases CPA1/CPA2 were also included in Cluster 1. Conversely, 433 PBL lacked the expression of secretory enzymes and hormones, such as AMY2B (encoding 434 Amylase alpha 2B) and INS (encoding Insulin), positivity of which represents the mature 435 cell populations of the pancreas (Cluster 2). Notably, besides these terminal 436 differentiation markers, Cluster 2 included the lineage-specific transcription factors, 437 BHLHA15 (also known as MIST1), ONECUT1 (also known as HNF6), and ISL1 (encoding Islet 438 1); the expression of these genes is observed from earlier development stages and 439 typifies the commitment to acinar, ductal, and endocrine lineages, respectively (38). 440 Clusters 3 and 5 further characterized PBL transcriptome in the context of the earliest 441 stage of development. Including SOX2 and NANOG, the most primitive stem cell markers, 442 Cluster 3 was strongly expressed in hESCs but not in pMPCs or PBL. On the other hand, 443 including ID1, BIRC5 (encoding apoptosis inhibitor Survivin) and also LGR5, Cluster 5 444 defined a set of common stemness-associated genes, which was substantially expressed 13

445 in hESCs and pMPCs as well as in PBL. Collectively, in spite of the PBL-specific aberrant 446 gene expression (Clusters 4 and 6) including the Wnt signal activation and IGF2 447 overexpression, the expression signature patterns of the remaining 4 gene clusters 448 (up-regulation of Clusters 1 and 5; down-regulation of Clusters 2 and 3) demonstrated the 449 transcriptome-level resemblance between PBL and pMPCs, rather than between PBL and 450 hESCs (Fig. 5A).

451 In principal component analysis, the first principal component (PC1) segregated the 452 differential development stages of pancreas (Fig. 5B), and the Cluster 2 genes and their 453 homologues showed the highest absolute coefficients of PC1 (Supplementary Table S13), 454 signifying that PC1 is an axis representing the pancreas development. On this axis, PBL 455 and pMPCs were close to each other with the mean (± standard deviation) 456 PC1-coordinates of 24.1 (± 10.6) in PBL and 22.2 (± 0.1) in pMPCs, in contrast to the mean 457 PC1-coordinates of −67.2 (± 4.6) in the normal pancreas and 85.5 (± 4.8) in hESCs. Overall, 458 since the pMPC transcriptomes that we analyzed were from human pancreatic bud at 459 Carnegie stage (CS) 16-18 (37), our results demonstrated that PBL and CS 16-18 pMPCs 460 share a similar gene expression profiles with regard to pancreatic development (Fig. 5C).

461

462 Discussion

463 To the best of our knowledge, this is the first report providing a genome-wide 464 molecular portrait of PBL, characterized by uniform pMPC-like gene expression patterns 465 based on an aberrant activation of the Wnt/β-catenin pathway with concurrent 466 imprinting dysregulation of IGF2. Since Horie (39) first named this unusual pediatric 467 pancreatic tumor ‘pancreatoblastoma (PBL)’, there have been reported only 468 approximately 400 cases of PBL in the literature. Although most of them were isolated 469 case reports, two molecular studies implicated the Wnt/β-catenin pathway and 470 chromosome 11p in the pathogenesis of PBL, where Abraham (4) identified 471 Wnt/β-catenin pathway mutations and LOH at chromosome 11p in 78% and 86%, 472 respectively; Tanaka (5) reported CTNNB1 mutations in 40%, but β-catenin nuclear 473 accumulation in 100%. In our cohort, as a result of an extensive molecular study, CTNNB1 474 mutations (90%) and a CN-LOH of APC (10%), all resulting in Wnt/β-catenin signal 475 activation, as well as aberrant imprinting of IGF2 including CN-LOH at 11p15.5 (80%), gain 476 of paternal allele (10%), and gain of methylation at ICR1 (10%) were identified, although 477 in a case (PBL009) with focal CN-LOH involving APC locus, but without another hit on the 478 healthy allele of APC gene, additional oncogenic events may be involved. Overall, our 479 comprehensive approach broadens the present understanding of PBL pathogenesis by

480 demonstrating that not only genetic alterations but also an aberrant epigenetic 481 modification intensively affect the Wnt/β-catenin pathway and IGF2 imprinting, making 482 the frequencies of these alterations virtually 100% (Fig. 4A).

483 Despite these homogeneous genetic/epigenetic features, the current study also 484 demonstrates a substantial intra- and inter-tumor heterogeneity in PBL. In PBL001, an 485 additional candidate driver mutation in KDM6A and the escalating numbers of mutations 486 due to TMZ treatment may be associated with an aggressive and deadly disease 487 phenotype (Fig. 2A and B). KDM6A is a specific histone lysine demethylase for the di- and 488 tri-methylation of lysine 27 on histone H3, which plays an important role in regulating the 489 balance between self-renewal and differentiation in hESCs (40). Because KDM6A is 490 located on chromosome X, this male case of PBL (PBL001) had only a mutated allele. In 491 stark contrast, PBL002, a congenital case diagnosed after a curative surgery at 3 days old, 492 exhibited a CTNNB1 hotspot mutation and CN-LOH limited to 11p15.5 as only genetic 493 abnormalities, and presented with a less aggressive tumor, not requiring adjuvant 494 chemotherapies to maintain complete remission. These 2 extreme cases reinforce the 495 pathogenic significance of up-regulation of Wnt/β-catenin and IGF2 axes in PBL as a set of 496 virtually necessary and sufficient conditions, as well as suggest the mechanisms of PBL to 497 acquire more malignant characteristics. Larger-scale studies are required for a more 498 detailed catalogue of driver mutations, which could be applied to genetic risk 499 stratification in PBL.

500 At lower frequencies than in PBL, Wnt/IGF2 co-activation has been reported in 501 pediatric embryonal tumors such as hepatoblastoma (41) and Wilms tumor (42). 502 Overexpression of R-spondin/LGR5/RNF43 module genes (LGR5 and RNF43) compared to 503 normal liver have also been reported in hepatoblastoma (43). In addition, recent studies 504 on colorectal cancer have defined the concurrent Wnt/IGF2 activation-positive colorectal 505 cancer as a consensus molecular subtype which is characterized by intestinal stem 506 cell-like phenotype and enriched LGR5 signature (44,45). Consistently, our transcriptome 507 analysis demonstrates that the R-spondin/LGR5/RNF43 module is highly enhanced in PBL 508 beyond the physiological up-regulation in hESCs and pMPCs (Fig. 4D). Taken together, the 509 triad of molecular signatures (Wnt signal activation, dysregulated IGF2 imprinting, and 510 the R-spondin/LGR5/RNF43 module enhancement) may define a unique malignant 511 disease entity originating from primitive cellular populations, regardless of the affected 512 organs and tissues. Additionally, because the higher expression of LGR5 has been 513 associated with the poorer prognosis in several human cancers (46,47), our data suggests 514 the enhanced up-regulation of the R-spondin/LGR5/RNF43 module may reflect the 515 intractability of PBL. Moreover, by identifying developmental stage-specific gene 15

516 expression signatures, our cluster and principal component analyses implicate pMPCs as 517 possible cellular origin of PBL (Fig. 5A and B). Recent studies have suggested that adult 518 pancreatic tumors, such as PDAC (48,49), acinar cell carcinoma (50), and solid 519 pseudopapillary neoplasm (51), arise from committed precursors or mature cells of 520 acinar/ductal lineages through the mechanisms including dedifferentiation and 521 acinar-to-ductal metaplasia, which in contrast, accentuates the distinctive biology of PBL 522 among primary pancreatic malignancies (Fig. 5C). Further analyses utilizing different 523 stages of developing pancreatic cells would help more accurate estimation of originating 524 cells of PBL.

525 In summary, our results, for the first time, illustrate a remarkable uniformity for the 526 molecular basis of PBL and shed a light on the ubiquitously activated Wnt/β-catenin 527 pathway and IGF2, as well as enhanced R-spondin/LGR5/RNF43 module, as potentially 528 druggable molecules for PBL.

529

530 Acknowledgments 531 We gratefully acknowledge the Genotype-Tissue Expression (GTEx) Project and all its 532 contributing investigators for making the invaluable data publically available. The 533 datasets used for the analyses described in this manuscript were obtained from dbGaP at 534 http://www.ncbi.nlm.nih.gov/gap through dbGaP accession number phs000424.v6.p1. 535 We are also grateful to Ms. Matsumura, Ms. Hoshino, Ms. Yin, Ms. Saito, Ms. Mizota, Ms. 536 Nakamura, and Ms. Iijima for their excellent technical assistance. We also wish to express 537 our appreciation to Dr. J. Mitsui, and Dr. S. Tsuji, The University of Tokyo, for next 538 generation sequencing. This work was supported by KAKENHI (26293242 (J.Takita)) from 539 Japan Society for the Promotion of Science; by Research on Measures for Intractable 540 Diseases, Health and Labor Sciences Research Grants, Ministry of Health, Labor and 541 Welfare (J.Takita); by Research on Health Sciences focusing on Drug Innovation (J.Takita); 542 by the Japan Health Sciences Foundation (J.Takita); by Core Research for Evolutional 543 Science and Technology, Japan Science and Technology Agency (J.Takita); by Japan 544 Agency for Medical Research and Development (AMED) Project for Cancer Research and 545 Therapeutic Evolution (P-CREATE) (J.Takita); and by Pancreas Research Foundation of 546 Japan (T. Isobe).

547 References 548 549 1. Glick RD, Pashankar FD, Pappo A, Laquaglia MP. Management of pancreatoblastoma in 550 children and young adults. J Pediatr Hematol Oncol. 2012;34 Suppl 2:S47–50.

551 2. Dhebri AR, Connor S, Campbell F, Ghaneh P, Sutton R, Neoptolemos JP, et al. Diagnosis, 552 treatment and outcome of pancreatoblastoma. Pancreatology. 2004;4:441–53.

553 3. Bien E, Godzinski J, Dall’Igna P, Defachelles A-S, Stachowicz-Stencel T, Orbach D, et al. 554 Pancreatoblastoma: a report from the European cooperative study group for paediatric 555 rare tumours (EXPeRT). Eur J Cancer. 2011;47:2347–52.

556 4. Abraham SC, Wu T-T, Klimstra DS, Finn LS, Lee J-H, Yeo CJ, et al. Distinctive Molecular 557 Genetic Alterations in Sporadic and Familial Adenomatous Polyposis-Associated 558 Pancreatoblastomas. Am J Pathol. 2001;159:1619–27.

559 5. Tanaka Y, Kato K, Notohara K, Nakatani Y, Miyake T, Ijiri R, et al. Significance of 560 aberrant (cytoplasmic/nuclear) expression of beta-catenin in pancreatoblastoma. J 561 Pathol. 2003;199:185–90.

562 6. Muguerza R, Rodriguez A, Formigo E, Montero M, Vázquez JL, Páramo C, et al. 563 Pancreatoblastoma associated with incomplete Beckwith-Wiedemann syndrome: case 564 report and review of the literature. J Pediatr Surg. 2005;40:1341–4.

565 7. Jiao Y, Yonescu R, Offerhaus GJA, Klimstra DS, Maitra A, Eshleman JR, et al. 566 Whole-exome sequencing of pancreatic neoplasms with acinar differentiation. J Pathol. 567 2014;232:428–35.

568 8. Cancer Genome Atlas Network. Comprehensive molecular characterization of human 569 colon and rectal cancer. Nature. 2012;487:330–7.

570 9. Seshagiri S, Stawiski EW, Durinck S, Modrusan Z, Storm EE, Conboy CB, et al. Recurrent 571 R-spondin fusions in colon cancer. Nature. 2012;488:660–4.

572 10. Fearon ER. Molecular genetics of colorectal cancer. Annu Rev Pathol. 2011;6:479–507.

573 11. Leary RJ, Lin JC, Cummins J, Boca S, Wood LD, Parsons DW, et al. Integrated analysis 574 of homozygous deletions, focal amplifications, and sequence alterations in breast and 575 colorectal cancers. Proc Natl Acad Sci USA. 2008;105:16224–9.

576 12. Lin D-C, Hao J-J, Nagata Y, Xu L, Shang L, Meng X, et al. Genomic and molecular 577 characterization of esophageal squamous cell carcinoma. Nat Genet. 2014;46:467–73.

578 13. Schulze K, Imbeaud S, Letouzé E, Alexandrov LB, Calderaro J, Rebouissou S, et al. 579 Exome sequencing of hepatocellular carcinomas identifies new mutational signatures 580 and potential therapeutic targets. Nat Genet. 2015;47:505–11.

581 14. Li M, Zhao H, Zhang X, Wood LD, Anders RA, Choti MA, et al. Inactivating mutations of 582 the chromatin remodeling gene ARID2 in hepatocellular carcinoma. Nat Genet. 583 2011;43:828–9.

584 15. Weinhold N, Jacobsen A, Schultz N, Sander C, Lee W. Genome-wide analysis of 585 noncoding regulatory mutations in cancer. Nat Genet. 2014;46:1160–5.

586 16. Shiraishi Y, Sato Y, Chiba K, Okuno Y, Nagata Y, Yoshida K, et al. An empirical Bayesian 587 framework for somatic mutation detection from cancer genome sequencing data. 588 Nucleic Acids Res. 2013;41:e89–9. 18

589 17. Yoshida K, Sanada M, Shiraishi Y, Nowak D, Nagata Y, Yamamoto R, et al. Frequent 590 pathway mutations of splicing machinery in myelodysplasia. Nature. 2011;478:64–9.

591 18. Roth A, Khattra J, Yap D, Wan A, Laks E, Biele J, et al. PyClone: statistical inference of 592 clonal population structure in cancer. Nat Methods. 2014;11:396–8.

593 19. Tamura K, Stecher G, Peterson D, Filipski A, Kumar S. MEGA6: Molecular Evolutionary 594 Genetics Analysis version 6.0. Mol Biol Evol. 2013;30:2725–9.

595 20. Yoshizato T, Nannya Y, Atsuta Y, Shiozawa Y, Iijima-Yamashita Y, Yoshida K, et al. 596 Genetic abnormalities in myelodysplasia and secondary acute myeloid leukemia: 597 impact on outcome of stem cell transplantation. Blood. 2017;129:2347–58.

598 21. Kuilman T, Velds A, Kemper K, Ranzani M, Bombardelli L, Hoogstraat M, et al. 599 CopywriteR: DNA copy number detection from off-target sequence data. Genome Biol. 600 2015;16:R163–15.

601 22. Nannya Y, Sanada M, Nakazaki K, Hosoya N, Wang L, Hangaishi A, et al. A robust 602 algorithm for copy number detection using high-density oligonucleotide single 603 nucleotide polymorphism genotyping arrays. Cancer Res. 2005;65:6071–9.

604 23. Yamamoto G, Nannya Y, Kato M, Sanada M, Levine RL, Kawamata N, et al. Highly 605 sensitive method for genomewide detection of allelic composition in nonpaired, 606 primary tumor specimens by use of affymetrix single-nucleotide-polymorphism 607 genotyping microarrays. Am J Hum Genet. 2007;81:114–26.

608 24. Shen R, Seshan VE. FACETS: allele-specific copy number and clonal heterogeneity 609 analysis tool for high-throughput DNA sequencing. Nucleic Acids Res. 2016;44:e131–1.

610 25. Assenov Y, Müller F, Lutsik P, Walter J, Lengauer T, Bock C. Comprehensive analysis of 611 DNA methylation data with RnBeads. Nat Methods. 2014;11:1138–40.

612 26. Love MI, Huber W, Anders S. Moderated estimation of fold change and dispersion for 613 RNA-seq data with DESeq2. Genome Biol. 2014;15:31–50.

614 27. Wang X, Cairns MJ. Gene set enrichment analysis of RNA-Seq data: integrating 615 differential expression and splicing. BMC Bioinformatics. 2013;14 Suppl 5:S16.

616 28. Clevers H. Wnt/β-Catenin Signaling in Development and Disease. Cell. 617 2006;127:469–80.

618 29. Bailey P, Chang DK, Nones K, Johns AL, Patch A-M, Gingras M-C, et al. Genomic 619 analyses identify molecular subtypes of pancreatic cancer. Nature. 2016;531:47–52.

620 30. Johnson BE, Mazor T, Hong C, Barnes M, Aihara K, McLean CY, et al. Mutational analysis 621 reveals the origin and therapy-driven evolution of recurrent glioma. Science. 622 2014;343:189–93.

623 31. Soejima H, Higashimoto K. Epigenetic and genetic alterations of the imprinting 624 disorder Beckwith-Wiedemann syndrome and related disorders. J Hum Genet. 625 2013;58:402–9.

626 32. Slavotinek A, Gaunt L, Donnai D. Paternally inherited duplications of 11p15.5 and 627 Beckwith-Wiedemann syndrome. J Med Genet. 1997;34:819–26.

628 33. Esteller M. Epigenetics in cancer. N Engl J Med. 2008;358:1148–59.

629 34. de Lau W, Barker N, Low TY, Koo B-K, Li VSW, Teunissen H, et al. Lgr5 homologues 630 associate with Wnt receptors and mediate R-spondin signalling. Nature. 631 2011;476:293–7.

632 35. Koo B-K, Spit M, Jordens I, Low TY, Stange DE, van de Wetering M, et al. Tumour 633 suppressor RNF43 is a stem-cell E3 ligase that induces endocytosis of Wnt receptors. 634 Nature. 2012;488:665–9.

635 36. Wu JQ, Habegger L, Noisa P, Szekely A, Qiu C, Hutchison S, et al. Dynamic 636 transcriptomes during neural differentiation of human embryonic stem cells revealed 637 by short, long, and paired-end sequencing. Proc Natl Acad Sci USA. 2010;107:5254–9.

638 37. Cebola I, Rodríguez-Seguí SA, Cho CH-H, Bessa J, Rovira M, Luengo M, et al. TEAD and 639 YAP regulate the enhancer network of human embryonic pancreatic progenitors. Nat 640 Cell Biol. 2015;17:615–26.

641 38. Pan FC, Wright C. Pancreas organogenesis: From bud to plexus to gland. Dev Dyn. 642 2011;240:530–65.

643 39. Horie A, Yano Y, Kotoo Y, Miwa A. Morphogenesis of pancreatoblastoma, infantile 644 carcinoma of the pancreas: report of two cases. Cancer. 1977;39:247–54.

645 40. Agger K, Cloos PAC, Christensen J, Pasini D, Rose S, Rappsilber J, et al. UTX and JMJD3 646 are histone H3K27 demethylases involved in HOX gene regulation and development. 647 Nature. 2007;449:731–4.

648 41. Honda S, Arai Y, Haruta M, Sasaki F, Ohira M, Yamaoka H, et al. Loss of imprinting of 649 IGF2 correlates with hypermethylation of the H19 differentially methylated region in 650 hepatoblastoma. Br J Cancer. 2008;99:1891–9.

651 42. Gadd S, Huff V, Huang C-C, Ruteshouser EC, Dome JS, Grundy PE, et al. Clinically 652 Relevant Subsets Identified by Gene Expression Patterns Support a Revised Ontogenic 653 Model of Wilms Tumor: A Children's Oncology Group Study. Neoplasia. 2012;14:742–56.

654 43. Cairo S, Armengol C, de Reyniès A, Wei Y, Thomas E, Renard C-A, et al. Hepatic 655 stem-like phenotype and interplay of Wnt/beta-catenin and Myc signaling in aggressive 656 childhood liver cancer. Cancer Cell. 2008;14:471–84.

657 44. Guinney J, Dienstmann R, Wang X, de Reyniès A, Schlicker A, Soneson C, et al. The 658 consensus molecular subtypes of colorectal cancer. Nat Med. 2015;21:1350–6.

659 45. Isella C, Brundu F, Bellomo SE, Galimi F, Zanella E, Porporato R, et al. Selective 660 analysis of cancer-cell intrinsic transcriptional traits defines novel clinically relevant 661 subtypes of colorectal cancer. Nat Commun. 2017;8:15107.

662 46. Yang L, Tang H, Kong Y, Xie X, Chen J, Song C, et al. LGR5 Promotes Breast Cancer 663 Progression and Maintains Stem-Like Cells Through Activation of Wnt/β-Catenin 664 Signaling. Stem Cells. 2015;33:2913–24.

665 47. Xi HQ, Cai AZ, Wu XS, Cui JX, Shen WS, Bian SB, et al. Leucine-rich repeat-containing 666 G-protein-coupled receptor 5 is associated with invasion, metastasis, and could be a 667 potential therapeutic target in human gastric cancer. Br J Cancer. 2014;110:2011–20.

668 48. Figura von G, Fukuda A, Roy N, Liku ME, Morris JP IV, Kim GE, et al. The chromatin 669 regulator Brg1 suppresses formation of intraductal papillary mucinous neoplasm and 670 pancreatic ductal adenocarcinoma. Nat Cell Biol. 2014;16:255–67.

671 49. Storz P. Acinar cell plasticity and development of pancreatic ductal adenocarcinoma. 672 Nat Rev Gastroenterol Hepatol. 2017;14:296–304.

673 50. Kong B, Cheng T, Qian C, Wu W, Steiger K, Cao J, et al. Pancreas-specific activation of 674 mTOR and loss of p53 induce tumors reminiscent of acinar cell carcinoma. Mol Cancer. 675 2015;14:212.

676 51. Heiser PW, Cano DA, Landsman L, Kim GE, Kench JG, Klimstra DS, et al. Stabilization of 677 β-Catenin Induces Pancreas Tumor Formation. Gastroenterology. 2008;135:1288–300.

678

679

680 Figure Legends 681 682 683 Figure 1. 684 Aberrant activation of Wnt/β-catenin pathway in PBL. A, Distribution of non-silent 685 somatic mutations in CTNNB1. AA, amino acid. B, Immunohistochemistry of β-catenin on 686 diagnostic or metastatic specimens from the CTNNB1 mutation-positive cases. For PBL007, 687 only a needle biopsy specimen was available and was not evaluable by 688 immunohistochemical analysis. 689 690 691 Figure 2. 692 Multi-regional analysis of PBL001. A, Landscape of somatic mutations in the primary and 693 metastatic samples from PBL001. The heights of bars represent variant allele frequencies 694 (VAFs) obtained by PCR-based amplicon deep sequencing. P, primary sample; M1-M6, 695 metastatic samples. B, Phylogenetic tree constructed using MEGA6 software based on the 696 mutational landscape (Methods). UPD, uniparental disomy. 697 698 699 Figure 3. 700 Copy number alterations and loss of imprinting of IGF2. A, Genome-wide copy number 701 alterations in 10 PBL cases identified by sequencing-based copy number analysis. CN, 702 copy number. B, Recurrent CN-LOH involving chromosome 11. Each green bar corresponds 703 to the affected region identified in each case. The allele-specific red/green signals of 704 PBL001, whose CN-LOH was limited to the 11p15.5 locus, are also shown. AsCN, 705 allele-specific copy number. C-D, Bisulfite sequencing of ICR1 in normal pancreas, PBL001 706 (C); and PBL005 (D). Tumor purities were estimated using targeted deep sequencing or 707 WES data (Methods). CpG sites are indicated by arrows. See Supplementary Fig. S7 for 708 sequencing results of the remaining samples examined. 709 710 711 Figure 4. 712 Comprehensive molecular signature of PBL. A, Integrated view of molecular and 713 clinicopathological characteristics of PBL. Upper panel shows clinicopathological and 714 genetic features of 10 patients. Lower panel shows the normalized expression levels of 715 genes with a highly significant differential expression (with log2 fold change ≤−2.5 or ≥2.5 716 and false discovery rate q ≤ 0.001. CR, complete remission; IHC, immunohistochemistry; 717 GOM, gain of methylation; GOP, gain of paternally methylated allele. B, Significant 22

718 enrichment of undifferentiated/stem cell gene modules identified by RNA 719 sequencing-based gene set enrichment analysis comparing PBL (n = 5) to normal 720 pancreatic tissues (n = 5). C, Gene expression profiles of the significantly enriched gene 721 sets. The colors represent relative log2 fold changes from the mean expression level in 722 the normal pancreas. N, normal pancreas; MP, pMPC; ES, hESC; Log2FC, log2 fold change. 723 D, Expression levels of the R-spondin/LGR5/RNF43 module genes in normal pancreas, PBL, 724 pMPC, and hESC. Significant differential expression (P ≤ 0.01, Wald test) between PBL and 725 the other groups is marked with an asterisk (*). 726 727 728 Figure 5. 729 Gene expression clustering and putative cellular origin of PBL. A, Hierarchical clustering 730 by expression of the top 1000 genes up- or down-regulated in the normal pancreas, PBL, 731 and hESCs. Gene expression data were transformed into z-scores and clustered using 732 Euclidean distance. RNA sequencing data from the Genotype-Tissue Expression (GTEx) 733 registry were included for expression data of the normal pancreas (see Methods). 734 Previously reported RNA sequencing data of hESCs (GSE20301; (36)) and pMPCs 735 (E-MTAB-3061; (37)) were also utilized. B, Principal component analysis of the 736 transcriptomes in the normal pancreas, PBL, pMPCs, and hESCs. PC1 and PC2 represent 737 the first two principal components accounting for 60% of the total variance. The black 738 arrows indicate the normal pancreatic development and suggested PBL initiation from 739 pMPCs. C, Schematic modeling of PBL initiation from pMPCs through acquisition of the 740 common genetic/epigenetic alterations, in contrast to other histological types of 741 pancreatic tumors originating from mature pancreatic cells. Genes expressed (indicated 742 by plus (+) signs or a remark of ‘HIGH’ for a higher level of expression) or unexpressed 743 (indicated by minus (−) signs) at each developmental stage are also denoted. SPN, solid 744 pseudopapillary neoplasm; IPMN, intraductal papillary mucinous neoplasm; PanIN, 745 pancreatic intraepithelial neoplasia; ACC, acinar cell carcinoma. 746

Integrated molecular characterization of the lethal pediatric cancer pancreatoblastoma

Tomoya Isobe, Masafumi Seki, Kenichi Yoshida, et al.

Cancer Res Published OnlineFirst December 12, 2017.

Updated version Access the most recent version of this article at: doi:10.1158/0008-5472.CAN-17-2581

Supplementary Access the most recent supplemental material at: Material http://cancerres.aacrjournals.org/content/suppl/2019/01/25/0008-5472.CAN-17-2581.DC1

Author Author manuscripts have been peer reviewed and accepted for publication but have not yet Manuscript been edited.

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://cancerres.aacrjournals.org/content/early/2017/12/12/0008-5472.CAN-17-2581. Click on "Request Permissions" which will take you to the Copyright Clearance Center's (CCC) Rightslink site.