Combined analysis of chromatin accessibility, promoter interactions and whole genome sequencing solved the missing heritability in gastric cancer

31 Missing heritability in h ereditary diffuse gastric cancer (HDGC) ranges from 60 to 90%. 32 These HDGC-like families, despite complying with HDGC clinical criteria, lack CDH1 and 33 CTNNA1 actionable germline variants, and are not offered HDGC -targeted life-saving 34 disease prevention measures. Herein, we explored novel HDGC predisposition mechanisms 35 affecting the CDH1-regulatory network. We called single-nucleotide (SNV) and copy-number 36 variants (CNV) from 19 HDGC -like probands from whole -genome sequencing data and 37 performed gene -ontology analysis . Chromatin enhancer marks and CDH1 promoter 38 interactions were evaluated in normal stomach by ChIP-seq, ATAC-seq and 4C-seq, variant 39 causality was assessed by RT -PCR, immunohistochemistry and microsatellite instability 40 (MSI) analysis in tumours . Functional analysis was performed using CRISPR -Cas9, RT-41 PCR and flow cytometry in cell lines, and enhancer assays using mouse embryos. Within 42 the CDH1 topologically associating domain (TAD), we found two deletions in Family F4 and 43 F9. F4 carried a heterozygous CDH3 20kb-CNV triggering CDH1 mRNA/protein loss in 44 homozygosity by CRISPR-Cas9 editing, similarly to a CDH1 coding deletion . This 20kb 45 sequence encloses two hypomorphic ti ssue-specific regulatory elements (REs), each 46 contributing 50% to CDH1 expression regulation. F9 carried a heterozygous 39bp-intergenic 47 CNV downstream of CDH1, triggering CDH1 mRNA/protein loss by CRISPR-Cas9. F15, 48 presenting gastric but not colorectal cancer, carried an MLH1 heterozygous 2.7Kb germline 49 CNV overlapping a stomach -specific RE found by ChIP-seq. The gastric tumour of mixed 50 histology displayed Microsatellite instability ( MSI), reduced MLH1 mRNA and protein, and 51 reduced CDH1 and E -cadherin protein . CRISPR -Cas9 clones mimicking the MLH1 52 heterozygous CNV, triggered loss of MLH1 and CDH1/E-cadherin mRNA and protein, similar 53 to a coding deletion . Beyond the CDH1 TAD and tumour r isk syndrome genes, multiple 54 deletions of stomach accessible chromatin sequences were found in particularly young-55 affected individuals from additional 6 families. This oligogenic pattern impaired specifically 56 mucin genes and multiple immune-related pathways. Herein, we pinpointed novel 57 mechanisms behind HDGC predisposition. One involves deletions of CDH1-REs in the TAD 58 or stomach-specific CDH1-REs in the MLH1 locus. The second involves multiple deletions 59 of stomach REs affecting mucin and immune-related genes, favouring a gastric immune-60 deficient phenotype. Altogether, by combining stomach-specific chromatin accessibility and 61 promoter interactions with whole genome sequencing, we solved the missing heritability in 62 47% of HDGC-like families within our cohort. 63 All rights reserved. No reuse allowed without permission. perpetuity. preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in The copyright holder for thisthis version posted December 2, 2024. ; https://doi.org/10.1101/2024.11.27.24317972doi: medRxiv preprint 3

64 Missing heritability in hereditary diffuse gastric cancer (HDGC) ranges from 60 to 90%. This 65 corresponds to the proportion of families, that despite complying with clinical criteria for 66 HDGC (Blair et al. 2020) , cannot be explained by genetic variants in bonafide HDGC-67 associated genes, CDH1 (ENSG00000039068) and CTNNA1 (ENSG00000044115), and 68 are called HDGC -like (Oliveira et al. 2009b; Blair et al. 2020) . While 10- 40% of HDGC 69 families bear heterozygous coding truncating variants (actionable) in CDH1 (Oliveira et al. 70 2009b, 2015), only 2% carry CTNNA1 truncating variants (Lobo et al. 2021) . HDGC-like 71 families remain genetically undiagnosed, even after deep analysis using candidate gene, 72 whole exome and whole genome sequencing (WGS) (Hansford et al. 2015; Oliveira et al. 73 2015; Tedaldi et al. 2019). 74 Carriers of actionable variants in CDH1 have increased risk for early-onset diffuse gastric 75 cancer (DGC) and lobular breast cancer (LBC), the HDGC-associated phenotypes (Guilford 76 et al. 1998; Oliveira et al. 2009a; Garcia -Pelaez et al. 2023). Truncating SNVs in CTNNA1 77 have also been associated with DGC predisposition in a small fraction of HDGC families 78 (Majewski et al. 2013; Lobo et al. 2021). Complete CDH1 and CTNNA1 loss of function, due 79 to somatic inactivation of the remaining wild -type allele , results in loss of cell adhesion, 80 disruption of epithelial architecture and decreased differentiation, features of signet-ring cells 81 characteristic of HDGC-related tumours (Oliveira et al. 2009c; Benusiglio et al. 2019; Gullo 82 et al. 2020). 83 Identification of CDH1 and CTNNA1 actionable variants in families fulfilling HDGC clinical 84 criteria, triggers cascade genetic testing and intensive surveillance through endoscopy with 85 multiple biopsies , and/or prophylactic gastrectomy in asymptomatic carriers. For 86 asymptomatic CDH1 female carriers breast intensive surveillance using m agnetic 87 resonance imaging (MRI) and computed tomography scan (CT-scan) is recommended, 88 alongside the option of (Blair et al. 2020). 89 Missing heritability hampers the use of personalized life-saving measures, and may lead to 90 prophylactic surgery in individuals lacking genetic diagnosis, solely based on family history 91 (São José et al. 2023a). HDGC-like families, clinically undistinguishable from CDH1-HDGC 92 families, often exhibit germline mono-allelic CDH1 RNA expression in normal tissues and 93 loss of E-cadherin expression in tumours, despite the lack of germline coding SNVs or CNVs 94 in CDH1 (Pinheiro et al. 2010). These observations pinpoint a likely causality for CDH1 and 95 argues towards the presence of undiagnosed germline alterations within noncoding cis-96 regulatory elements (REs), controlling CDH1 expression and function. 97 Enhancer-promoter communication within the nucleus tightly regulates cell -type-specific 98 gene expression (Spielmann et al. 2018). REs are instrumental for proper gene expression 99 in time and space and enhancers in particular, play an important role in controlling 100 expression of disease-causing genes (Cova et al. 2023) . Their mode of action is usually 101 dose-dependent, redundant, orientation-independent, acting through binding of tissue -102 specific transcription factors and controlling tissue-specific gene expression (Will et al. 2017; 103 Bordeira-Carriço et al. 2022) . Enhancer-promoter communication may occur within large 104 genomic distances, which are brought in close spatial proximity upon 3D-chromatin folding 105 All rights reserved. No reuse allowed without permission. perpetuity. preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in The copyright holder for thisthis version posted December 2, 2024. ; https://doi.org/10.1101/2024.11.27.24317972doi: medRxiv preprint 4 (Bonev and Cavalli 2016; Furlong and Levine 2018) . Topologically associating domains 106 (TADs) are the functional units of 3D -chromatin architecture, restricting enhancers’ 107 interaction with their target promoters (Rao et al. 2014; Dixon et al. 2016) . CNVs affecting 108 TADs’ structure by affecting the boundary between two separated TADs may trigger gene 109 misexpression patterns and cause disease (Franke et al. 2016; Lupiáñez et al. 2016) . 110 Assigning function and pathogenicity to noncoding variants found in enhancers and other 111 REs, in the next-generation sequencing era, remains challenging. The integration of multiple 112 omics technologies and functional models is essential to elucidate the impact of RE 113 perturbation and to establish a link to disease predisposition. Herein, we aim to identify and 114 characterize novel HDGC-causing mechanisms involving REs regulating CDH1 expression 115 to solve the missing heritability in HDGC -like families , resourcing to WGS, chromatin 116 analysis technologies and functional assays in cells and mice. 117 118 119 120 All rights reserved. No reuse allowed without permission. perpetuity. preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in The copyright holder for thisthis version posted December 2, 2024. ; https://doi.org/10.1101/2024.11.27.24317972doi: medRxiv preprint 5

121 Patient selection 122 Twenty-one patients from 19 HDGC-like families from various genetic backgrounds were 123 admitted to BC Cancer in British Columbia, Canada and fulfilled the 1999 HDGC clinical 124 criteria (Caldas et al. 1999). Families were re-classified according to (Blair et al. 2020) and 125 HDGC-like families were considered if fulfilled criteria 1 and 2. Gastric cancers with diffuse 126 component, with or without an intestinal component (mixed-type MGC) were considered for 127 criteria 1. All patients tested negative for CDH1 SNVs and CNVs. The present project was 128 approved by the Ethical Committee of Centro Hospitalar Universitário de São João, with 129 internal references 445-20 and 305-19. 130 131 Whole genome sequencing 132 Genomic DNA extracted from peripheral blood of 21 patients was sequenced using Illumina 133 platform. Sequencing reads were mapped to the human reference genome (GRCh38) using 134 BWA mem (v.0.7.15). Duplicated reads were marked, and their base quality score was 135 recalibrated with GATK (v4.1.9.0, HTSJDK 2.23.0 and Picard 2.23.3). 136 137 Variant calling 138 SNVs were called with GATK – HaplotypeCaller tool (v4.1.9.0, HTSJDK 2.23.0 and Picard 139 2.23.3), and Depp Variant (Docker image: 1.3.0) (Poplin et al. 2018) . Called SNVs were 140 filtered according to quality control metrics, read depth (≥8) and genotype quality (≥20), and 141 annotated with VEP (v.107). SNVs were prioritized if called by (1) both callers, from which 142 HaplotypeCaller had high quality (≥200), (2A) Depp Variant, or (2B) HaplotypeCaller with 143 extremely high quality (≥500). CNVs calling was performed with LUMPY (v 0.2.13) (Rausch 144 et al. 2012; Layer et al. 2014) , DELLY (v.0.9.1) and GRIDSS (v.2.13.2) (Cameron et al. 145 2017) , and annotated with AnnotSV (v. 2.5.2, Tcl (v. 8.6)) (Geoffroy et al. 2018). Structural 146 variants with ≥1Mb length, inter-chromosomal events and variants lacking ‘PASS’ flag were 147 discarded. CNVs were prioritized based on the following criteria: ( Tier 1) called by LUMPY 148 and DELLY with ≥90% overlap, and GRIDSS confirmed breakpoints with ≤4bp difference, 149 (Tier 2A) called by DELLY and GRIDSS confirmed breakpoints with ≤4bp difference, ( Tier 150 2B) called by LUMPY and GRIDSS confirmed breakpoints with ≤4bp difference, and ( Tier 151 2C) LUMPY and DELLY with ≥90% overlap. Rare CNVs within the CDH1 TAD (hg38: chr16: 152 68550001-69180000) were analysed to identify potential CDH1 REs (AF<0.01). Rare coding 153 CNVs and truncating SNVs in genes associated with gastrointestinal syndromes (AF<0.01; 154 Supplementary Table 1) were assessed to exclude variants in other cancer -predisposing 155 genes and further identify CDH1 regulatory elements. For the polygenic approach, ultrarare 156 deletions and truncating SNVs (AF<0.00001) observed in fewer than 6/21 patients (<24% 157 of the cohort) were selected for analysis. Deletions in F4, F9 and F15 were validated by 158 Sanger sequencing and SVs in F4 were validated by Pacbio sequencing as well. 159 160 All rights reserved. No reuse allowed without permission. perpetuity. preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in The copyright holder for thisthis version posted December 2, 2024. ; https://doi.org/10.1101/2024.11.27.24317972doi: medRxiv preprint 6 Chromatin availability data 161 ATAC-seq (ENCSR337UIU, ENCSR970UNF) and ChIP-seq (ENCSR492BHN, 162 ENCSR009RJD, ENCSR751BHO, ENCSR313DUH, ENCSR574USP, ENCSR078LIZ) data 163 from normal stomach and colonic mucosa was collected from phase 3 ENCODE project 164 (Abascal et al. 2020) . GeneHancer enhancers and eQTLs were collected from the 165 GeneHancer project (Fishilevich et al. 2017) and cCREs were collected from the ENCODE 166 database (Moore et al. 2020). 167 168 4C-seq 169 Fresh gastric tissue from bariatric surgeries was collected and washed in HBSS 1x (Gibco). 170 Gastric mucosa cells were scraped and enzymatically dissociated with collagenase type I 171 (Merck), dispase (Merck), trypsin inhibitor (Sigma), BSA (Nzytech), dithiothreitol (Invitrogen) 172 and HBSS 1x (Gibco) at 37ºC and 150rpm for at least 1h. 4C-seq libraries were generated, 173 as previously described (van de Werken et al. 2012; Krijger et al. 2020) , containing 1x107 174 cells, cross-linked in 2% paraformaldehyde and lysed. Nuclei suspensions were digested 175 with DnpII (New England Biolabs) as primary and Csp6I (New England Biolabs) as 176 secondary restriction enzymes and re -circularized with T4 DNA Ligase (Thermo Fisher 177 Scientific). 4C-seq libraries were purified using Amicon Ultra-15 10 kDa (MWCO) (Millipore) 178 and PCR-amplified with 3.2µg per reaction (supplementary table 2). Samples were paired-179 end sequenced on Illumina HiSeqX technology (150bp reads) according to standard 180 protocols. CDH1 interactions on a genome scale were mapped from sequenced 4C libraries, 181 using a bioinformatics pipeline based on Pipe4C (Krijger et al. 2020) , PeakC for cis 182 interactions (Geeven et al. 2018) with window size 2, alpha fdr 0.1 and minimal distance 500 183 and fourSig for trans interactions with window size 5, 1000 iterations, 0.001 fdr and 0.01 fdr 184 probability and considered interactions in categories 1 and 2. 185 186 Immunohistochemistry (IHC) 187 Paraffin-embedded block s containing adequate representation of the advanced tumour 188 were selected for IHC testing. Parallel sections 5 μm thick were cut and stained with E-189 cadherin and MLH1 (1:150; clone G168 -728, BD Biosciences), or P-cadherin, using 190 standard automated techniques. For each antibody, positive and negative external controls 191 were placed on the same slide or on a separate control slide. Hematoxylin-eosin– and IHC-192 stained slides were analysed by two experienced pathologists. 193 194 gDNA and RNA paraffin extraction 195 Tumour/normal tissue was marked by an experienced pathologist and gDNA/RNA was 196 extracted from six 10 μm thick slides . Tumour areas with at least 75% tumour cells were 197 selected for extraction, using MagMax FFPE DNA/RNA Ultra Kit (Applied Biosystems, Life 198 technologies), according to the manufacturer’s instructions. Briefly, sections of paraffin -199 embedded tumour samples were de -paraffinized and protease K-digested. DNA and RNA 200 All rights reserved. No reuse allowed without permission. perpetuity. preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in The copyright holder for thisthis version posted December 2, 2024. ; https://doi.org/10.1101/2024.11.27.24317972doi: medRxiv preprint 7 were bonded to magnetic beads, pelleted against a magnetic stand, and the supernatant 201 containing DNA/RNA washed and eluted. 202 203 Microsatellites instability analysis (MSI) 204 Tumour tissue of the MLH1 proband was studied for MSI using a panel of at least five 205 dinucleotide repeat sequences, as described previously (Velho et al. 2008) . Tumour was 206 classified as 1) MSI-H if presented instability in ≥ 40% of the markers, 2) MSI-L if presented 207 instability in <40% of the markers and 3) MSS if showed no instability in the tested markers. 208 209 DNA Methylation Analysis 210 gDNA extracted from tumour and normal counterpart were treated with bisulfite using Epitect 211 Bisulfite kit (Qiagen ). MLH1 promoter was amplified by PCR using Multiplex PCR kit 212 (Qiagen) and the primers depicted in supplementary table 3. PCR products were analysed 213 in gel electrophoresis and Sanger sequenced on an ABI -3130 Genetic Analyzer (Applied 214 Biosystems). 215 216 Cell lines and culture conditions 217 Human cell line MKN74 was purchased from the JCRB Cell Bank. MKN74 cell line and 218 isogenic clones were cultured in RPMI medium (Gibco) supplemented with 10% foetal 219 bovine serum (Biowest) and 1% Penicillin-Streptomycin (Gibco). HEK293T was cultured in 220 DMEM medium supplemented with 10% foetal bovine serum (Biowest) and 1% Penicillin-221 Streptomycin (Gibco). Cell s were maintained at 37ºC and 5% CO 2 in a high humidity 222 atmosphere. Cell identification was confirmed by STR analysis and cells were confirmed to 223 be free of mycoplasma contamination. 224 225 In vitro functional assay by CRISPR-Cas9 226 To test enhancer function, each region was targeted with CRISPR -Cas9. sgRNAs were 227 designed using Benchling online platform ( supplementary table 3). Individual sgRNAs 228 (Invitrogen) were cloned in LentiCRISPRv2GFP (addgene 82416) or LentiCRISPRv2 -229 mCherry (addgene 99154) vectors using BsmBIv2 (New England Biolabs). Plasmids were 230 transformed into Stbl3 competent cells and colonies were sequenced ( supplementary 231 table 3). Lentiviral particles were produced resourcing to HEK293T cell line with pMD2.G 232 (addgene 12259) and pCMV -dR8.91 (addgene) vectors, following Lipofectamine 3000 233 manufacture’s protocol (Invitrogen) and collected at 48h. MKN74 was infected with pairs of 234 lentivirus particles in medium supplemented with 10 μg/μl hexadimethrine bromide (Merck 235 Life Science S.L.U.) for 48h. Transduced cells were selected for GFP and mCherry positive 236 expression at 7 days post-infection using FACS ARIA (BD Biosciences). 237 238 Genotyping of edited clones 239 All rights reserved. No reuse allowed without permission. perpetuity. preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in The copyright holder for thisthis version posted December 2, 2024. ; https://doi.org/10.1101/2024.11.27.24317972doi: medRxiv preprint 8 gDNA was extracted using NZY Tissue gDNA isolation kit (NZYTech), according to the 240 manufacturers’ protocol. gDNA was amplified using primers flanking the edition sites 241 (supplementary table 3) and Multiplex PCR kit (Qiagen). PCR products were analysed in 242 gel electrophoresis and Sanger sequenced on an ABI -3130 Genetic Analyzer (Applied 243 Biosystems). 244 245 CDH1/E-cadherin and CDH3/P-cadherin expression analysis 246 CDH1/3 and MLH1 mRNA expression was assessed by qPCR in triplicates for the MKN74 247 isogenic clones and in a single experiment for tumour/normal tissues from probands . RNA 248 was extracted using mirVana RNA Isolation Kit (Invitrogen), according to manufacturers’ 249 protocol. cDNA was synthesized using SuperScriptII reverse transcriptase (Invitrogen), 250 according to the manufacturers’ protocol. mRNA expression was analyzed by qPCR with 251 KAPA PROBE FAST qPCR Master Mix (2X) Kit (Sigma -Aldrich) and probes for CDH1 252 (Hs.PT.58.3324071, TaqMan), CDH3 (Hs.PT.51.5028751, IDT), MLH1 (custom assay, IDT) 253 and 18S (custom assay, IDT). Reactions were sequenced on a 7500 Real -Time PCR 254 System (Applied Biosystems). Relative expression was normalized for the endogenous 18S 255 control and quantified using the 2−∆∆Ct method. 256 E-cadherin, P-cadherin and MLH1 expression was assessed by flow cytometry in triplicates. 257 Cells were detached with Versene (Gibco) and blocked with 3% bovine serum albumin -258 phosphatase buffer saline (Gibco). Cells were incubated with primary mouse monoclonal 259 antibody HECD-1, P-cadherin or MLH1 (1:100 dilution; 1h at 4ºC; Invitrogen), washed and 260 incubated with secondary antibody anti -mouse Alexa Fluor 647 (Invitrogen). Fluorescence 261 was measured using FACS Fortessa or Canto II (BD Biosciences) and Flow Jo version 10 262 software was used to analyse the data. 263 264 Statistical analysis 265 Statistical analysis was performed using GraphPad Prism version 7.00 software (GraphPad 266 Software Inc.). A t-student test was used for comparison analysis, assuming equal variance 267 between clones and parental samples. Differences were considered significant when p -268 value<0.05. 269 270 Enhancer-reporter assays in vivo 271 Enhancer activity was analyzed in transgenic mouse embryos using a site -specific 272 integration protocol adapted for mouse embryonic stem cells (mESCs ), as described 273 previously (Phan et al. 2024). Briefly, the PhiC31 system (Chi et al. 2019) facilitated precise 274 recombination between two att sites: an attP site inserted in a landing pad at the H11 safe-275 harbor genomic locus and an attB site within a donor vector containing the candidate 276 enhancer and a puromycin selection marker (Sigma-Aldrich, P8833). 277 First, a mESC line was generated by CRISPR/Cas9 (FuGENE technology, Promega) to 278 insert the H11 landing pad, which contained the Hsp68 promoter upstream of the LacZ 279 All rights reserved. No reuse allowed without permission. perpetuity. preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in The copyright holder for thisthis version posted December 2, 2024. ; https://doi.org/10.1101/2024.11.27.24317972doi: medRxiv preprint 9 reporter gene (Hsp68::LacZ). This cell line served as a negative control to assess 280

staining. Candidate enhancer regions were PCR -amplified from human 281 genomic DNA using PrimeSTAR GXL DNA Polymerase (Takara Bio) and primers with 282 appropriate overhang s (supplementary table 4). Donor vectors were cloned by Gibson 283 assembly (New England Biolabs; (Gibson et al. 2009)). 284 Subsequently, each donor plasmid was co -transfected with a PhiC31 plasmid into the 285 Hsp68::LacZ mESC line using Lipofectamine LTX (Invitrogen) . Transfected cells were 286 selected with puromycin, and genetically modified mESCs were used to generate transgenic 287 embryos via morula aggregation (Artus and Hadjantonakis 2011). CD1 female mice served 288 as foster mothers. 289 At embryonic day E14.5, the embryos were harvested and processed for beta-galactosidase 290 activity. Embryonic t issues were dissected in iced -cold 1X PBS , fixed in 4% 291 paraformaldehyde (PFA) for 10 minutes and stained in beta-galactosidase solution at 37°C 292 for a several hours to overnight, following Lobe et al. (1999). Dissected organs were imaged 293 using a ZEISS SteREO Discovery.V12 microscope with a cold light source CL9000 and a 294 Leica DFC420 digital camera (Leica Microsystems). Embryos were stored in 4% PFA in 1x 295 PBS at 4°C. 296 297 298 299 All rights reserved. No reuse allowed without permission. perpetuity. preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in The copyright holder for thisthis version posted December 2, 2024. ; https://doi.org/10.1101/2024.11.27.24317972doi: medRxiv preprint 10

300 We whole genome sequenced blood gDNA from 21 HDGC-like patients. Description of the 301 cohort is depicted in supplementary figure 1. Following variant calling, we prioritized rare 302 SNV, CNV and SV (AF<0.01), based on low frequency in cancer-free individuals in the CDH1 303 TAD and in cancer syndrome -associated genes (supplementary table 1), and ultrarare 304 deletions and truncating SNVs (AF<0.00001) that were common in our cohort (<6 patients, 305 <24%). 306 307 Variants within REs in the CDH1 TAD solve missing heritability in two HDGC families 308 We identified potential REs within the CDH1 TAD by extracting biochemical and chromatin 309 annotations, such as histone modifications from ChIP-seq data in normal stomach. These 310 marks were overlapped with regions that physically interact with the CDH1 promoter in the 311 same tissue, obtained through 4C -seq, and that display accessible chromatin identified 312 through ATAC-seq. The list of potential REs was then crossed with CNVs and SVs found in 313 HDGC families through WGS. We found a heterozygous deletion encompassing CDH3 314 (encoding the P-cadherin protein) exons 1 and 2, intron 1 and part of intron 2, as well as two 315 inversions (SVs) overlapping each side of the deletion boundaries in the proband of family 316 F4 (figure 1A,B). These structural variants, validated by Pacbio long-read sequencing and 317 Sanger sequencing , revealed a 20kb -deleted allele ( a), and two inversions flanking the 318 deletion, retaining most of the deleted sequence ( b). However, the 5’ and 3’ breakpoints of 319 this allele were not determined (b) (figure 1B). In addition, we found evidence for a wild-320 type sequence including the deleted region with non-inverted flanking regions ( c) (figure 321 1B). 322 We focused on the CDH3-CNV impact in CDH1 expression, by CRISPR-Cas9 deleting the 323 CDH3 20kb sequence in a stomach-derived cell line. The homozygous deletion of this region 324 led to complete loss of CDH1 mRNA and E-cadherin protein (figure 1C), similar to a CDH1 325 coding deletion (São José et al. 2023b) . The homozygous CDH3-CNV also triggered 326 complete loss of CDH3/P-cadherin expressio n ( supplementary figure 3 A, B), as the 327 deleted sequence includes the CDH3 promoter, exon 1 and exon 2 (figure 1A). The tumour 328 from family F4 proband presented mixed-type histology with signet ring cells, characteristic 329 of HDGC tumours, and E-cadherin expression loss (figure 1D, E). Altogether, these results 330 suggest that CDH1 expression is regulated by REs lying within the CDH3 20kb deleted 331 sequence. 332 As dramatic expression changes often depend on multiple REs being perturbed, we looked 333 for REs within the 20kb deleted region. We found two potential CDH1 REs within the 334 deletion, namely a sequence interacting with the CDH1 promoter in normal stomach 335 epithelia found by 4C -seq, that displays accessible chromatin in a gastric cell line with 336 normal CDH1 expression (RE1), and a second sequence encompassing a CDH1-controlling 337 expression quantitative trait loci (eQTL, RE2) (Fishilevich et al. 2017) (figure 1B). CRISPR-338 Cas9 homozygous deletion of RE1 alone led to 60% expression loss in CDH1 mRNA and 339 E-cadherin protein levels, while RE2 homozygous deletion alone caused 40% expression 340 loss in CDH1 mRNA and E-cadherin protein levels. The sum of the perturbation of RE1 and 341 All rights reserved. No reuse allowed without permission. perpetuity. preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in The copyright holder for thisthis version posted December 2, 2024. ; https://doi.org/10.1101/2024.11.27.24317972doi: medRxiv preprint 11 RE2 independently, mimicked the impact of deleting the complete 20kb sequence (figure 342 1C). Deletion of the intronic RE1 did not interfere with CDH3 mRNA nor with P-cadherin 343 protein expression, while deletion of the intronic RE2 led to 30% expression loss in CDH3/P-344 cadherin ( supplementary figure 3 A,B). These data suggest that RE2 modulated both 345 CDH1 and CDH3 expression, while RE1 modulated CDH1 expression only. Enhancer 346 reporter assays in mouse embryos revealed strong enhancer activity in the forebrain for RE1 347 and RE2 sequences (figure 1F). While RE1 showed strong enhancer activity in the closing 348 palate, teeth, whiskers and eye, RE2 had mild activity in the stomach (figure 1F), correlating 349 well with tissues known to express CDH1 during development (Stemmler et al. 2003) . 350 Inversion of the 20Kb deleted region in F4, does not impact CDH1 expression (data not 351 shown). As for the allele bearing the two inversions, the 5’ -inversion places CDH3 exon 2 352 ahead of exon 1 and part of the promoter, which is predicted to disrupt CDH3 transcription 353 and translation initiation . A similar effect is expected for the 3’ -inversion that inverts the 354 region spanning exon 3 -8. The impact of these inversions in CDH1 expression was not 355 assessed, yet the CDH3-CNV correlates well with CDH1 expression loss and HDGC-related 356 phenotypes observed in family F4. 357 The tumour spectrum of Family F4 is broad, enclosing three gastric cancers (ages <40s), a 358 central nervous system tumour (age 40s), an ovarian cancer (age 20s), a skin cancer (age 359 50s) and a breast cancer (figure 2A). This young age of onset and broader tumour spectrum 360 may derive from the combined perturbation of CDH1 and CDH3, impacting tissues beyond 361 the classically associated with CDH1 impairment, and the activity of the deleted REs in 362 different tissues, as illustrated by the strong forebrain enhancer activity found in mice and 363 the presence of a central nervous system tumour (figure 1F, 2A). 364 We found that family F9, presenting with four gastric cancers (one in age 40s and three in 365 age 60s), three of which after age 60, and one colorectal cancer in 50s (figure 2B), carried 366 a 40bp deletion within the CDH1-TANGO6 intergenic region, bearing CDH1-promotor 367 interactions in normal stomach (figure 1A, B). Homozygous deletion of the entire CDH1-368 TANGO6 intergenic region led to 50% reduction loss in CDH1 mRNA and E-cadherin protein 369 levels (figure 1C). Th e deleted region showed enhancer activity in mouse embryos 370 forebrain, but not in the stomach or colon (figure 1F). Given that, in homozygosity, this 371 deletion triggers half CDH1 expression loss, a third event may be required for gastric cancer 372 (GC) development, which can explain the latter age of cancer onset seen in this family. 373 Altogether, we reveal a novel mechanism impacting CDH1 expression, through the 374 impairment of hypomorphic and dose -dependent enhancers within the CDH1 TAD, likely 375 explaining cancer predisposition in two HDGC-like families. 376 377 378 A variant within a long-distance stomach-specific RE solves missing heritability in one HDGC 379 family 380 To further solve the missing heritability in the remaining HDGC-like families, we inquired the 381 WGS data from the remaining HDGC-like patients for rare (AF<0.01) SNVs, CNVs and SVs 382 in cancer syndrome -associated genes depicted in supplementary table 1. We found a 383 All rights reserved. No reuse allowed without permission. perpetuity. preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in The copyright holder for thisthis version posted December 2, 2024. ; https://doi.org/10.1101/2024.11.27.24317972doi: medRxiv preprint 12 2.787-bp deletion encompassing the MLH1 out-of-frame exon 13 and part of introns 12 and 384 13 in family F15 (figure 3A). MLH1 is a Lynch syndrome mismatch repair gene associated 385 with Lynch syndrome, and its loss in cancer cells leads to microsatellite instability (MSI). 386 Indeed, the F15 proband’s tumour showed MLH1 expression loss, MSI and lacked MLH1 387 promoter methylation (Figure 3B-D, supplementary figure 3 E), arguing toward s the 388 causality of the MLH1-CNV in HDGC predisposition in this family. 389 This family presents a proband with gastric carcinoma with lymphoid stroma and an evident 390 diffuse component at age 40s, and another two early-onset GCs (ages 30s and 40s) in a 391 sibling and their mother, suggesting variant segregation with disease through the maternal 392 side (figure 3E). GC preponderance and lack of colorectal cancers in this family was 393 unexpected, and raised the hypothesis that stomach-specific REs may reside within the 394 deleted region. Indeed, t he MLH1 deletion overlaps H3K4me1 peaks, chromatin marks of 395 active enhancers (RE3), that are present in normal stomach , but not in normal colon 396 epithelia (figure 3F). RE3 lacks H3K4me3 promoter chromatin marks in both stomach and 397 colon (figure 3F), suggesting this element may act as an enhancer, rather than a promoter. 398 We resorted to 4C-seq data from normal stomach epithelia and found evidence for a long-399 distance interaction between CDH1 and MLH1 promoters (figure 3F), which could indicate 400 cross-regulation between these two genes. Supporting that the MLH1 stomach-specific RE 401 could modulate CDH1 expression, was the finding that E-cadherin was lost or mislocalized 402 in the diffuse component of the proband’s tumour (figure 3G, H) and that CDH1 mRNA was 403 reduced by 80% (figure 3I). A CRISPR/Cas9 clone mimicking the MLH1 heterozygous CNV 404 triggered a reduction of 50% in the MLH1 mRNA and protein expression levels 405 (supplementary figure 3 C,D), and of 40% in the CDH1 mRNA and protein expression 406 levels (figure 3J, K). 407 Altogether, our results pinpoint an additional molecular mechanism, involving the germline 408 perturbation of stomach-specific REs at the MLH1 locus, driving somatic CDH1/E-cadherin 409 downregulation and tumour MSI. This HDGC-like family may still be a Lynch syndrome 410 family with a particular tropism to the stomach, due to a deletion of a stomach -specific RE 411 with secondary impact in CDH1 expression. Yet, the potential development of Lynch 412 syndrome-associated tumours may not be disregarded in the clinic. 413 414 415 Multiple germline SNVs and CNV-deletions affecting stomach-specific regulatory elements 416 and immune-associated pathways occur in 50% of HDGC-like families 417 We next explored the potential of a polygenic nature for the remaining HDGC families, by 418 evaluating the co-occurrence of variants across the genome. We searched the WGS data 419 for ultra-rare coding deletion-CNVs and truncating SNV (AF<0.00001), prioritized by at least 420 two callers , common in the cohort ( ≥6 patients, ≥24%) and absent in 13 controls from 421 1000genomes project. Overall, in 21 patients, we found 771 truncating SNVs and 11.860 422 CNVs, of which 1.547 were coding deletions. Each patient carried between 12 and 72 SNVs 423 and 22 and 209 CNVs. When combining all genes affected by the prioritized variants in 424 pathway analyses (Figure 4A, supplementary figure 4), we found a pathway-cluster that 425 All rights reserved. No reuse allowed without permission. perpetuity. preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in The copyright holder for thisthis version posted December 2, 2024. ; https://doi.org/10.1101/2024.11.27.24317972doi: medRxiv preprint 13 was impacted in 10 HDGC-like patients mainly presenting DGC before age 50 (8/21) (figure 426 4A, B, supplementary figure 4). This cluster included i mmune-associated pathways, 427 encompassing antigen-presenting through major histocompatibility complex (MHC) class I, 428 natural killer, tumour cell killing and T cell mediated immunity (figure 4A, B, supplementary 429 figure 4). These pathways were not called for most families when considering SNVs alone, 430 but remained represented when considering CNV only (supplementary figure 5), which 431 are the main genetic event behind impairment of REs. We then asked whether deletions 432 overlapping stomach REs support GC predisposition. By focusing on deletions overlapping 433 accessible chromatin regions in normal stomach tissue by ATAC-seq, from 6.642 deletions, 434 415 overlapp ed accessible chromatin regions. Indeed, the same cluster of immune -435 associated pathways seen in HDGC-like patients based on overall CNVs and SNVs, was 436 mimicked, despite the reduction in number of deletions overlapping stomach REs (figure 437 4C, supplementary figure 6 A). There were 39 genes recurrently impaired by deletions or 438 truncating SNVs, and five genes recurrently deleted in their regulatory sequence, in at least 439 three patients (figure 6A). We plotted these together with the monogenic events described 440 above in Figure 6 , as well as the number of patients that presented immune -pathways 441 affected. Families F15 and F9 , which presented germline MLH1 and CDH1-TANGO6 442 intergenic CNVs, had no other genes impaired. Six probands and F19_3 family member 443 (invasive ductal breast cancer at age 50s) lacked variants in these pathways, mimicking the 444 controls. Besides the CDH3 germline CNV and SVs, F4 also displayed impaired immune-445 associated genes/pathways. The remaining nine probands and additional member of F19 446 (F19_1 with intestinal GC at 3 0s, F19_2 with DGC at 6 0s) presented between 3 and 27 447 deleted genes (figure 6A). Additionally, from 10 probands with impaired germline landscape 448 in immune-related genes, eight had age of onset below 50 years of age (mean age of onset: 449 47,2 years±16,8) (figure 6B). While 7 individuals with deletions overlapping stomach REs 450 had age of onset below 50 (mean age of onset: 38,4 years±8,5) from 12 individuals (figure 451 6B). 452 453 All rights reserved. No reuse allowed without permission. perpetuity. preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in The copyright holder for thisthis version posted December 2, 2024. ; https://doi.org/10.1101/2024.11.27.24317972doi: medRxiv preprint 14

454 More than half of HDGC suspected families remain molecularly undiagnosed (HDGC-like), 455 many of them displaying germline monoallelic CDH1 downregulation and E-cadherin loss in 456 the tumour (Pinheiro et al. 2010) . These data pinpoint CDH1 as the main gene triggering 457 HDGC. Herein, we worked under the hypothesis that perturbation of REs controlling CDH1 458 expression drive HDGC predisposition in HDGC-like families. 459 This study presents compelling evidence supporting the conclusion that cancer 460 predisposition in 16% (3/19) of HDGC-like families within our cohort is explained by 461 alterations occurring in the CDH1 regulatory network, within and beyond the CDH1 TAD. 462 Importantly, the high frequency of HDGC-like families (10/19 – 53%) bearing ultra-rare 463 deletions overlapping stomach REs and impairing immune-associated pathways, calls for a 464 multigenic regulatory landscape contributing to HDGC predisposition. 465 The CDH1 TAD is expected to be the most enriched region of the genome in REs modulating 466 the CDH1 locus (Zuin et al. 2022). Within the CDH1 TAD, enclosing eight genes and a total 467 of 630kb, we found a CDH3 20-kb deletion encompassing two noncoding hypomorphic and 468 dose-dependent enhancers controlling CDH1 expression, which likely represent a novel 469 biological mechanism of HDGC predisposition in family F4. The CDH1/E-cadherin loss in 470 the diffuse component of the tumour further argues towards the role of this deletion in HDGC 471 predisposition. The combined deletion of two CDH1 enhancers, each contributing to half 472 CDH1 expression loss, represents a classical mechanism of enhancer -driven expression 473 control in CDH1-expressing tissues (Stemmler et al. 2005; Will et al. 2017) . In this family, 474 this heterozygous CDH3-CNV produces an expression loss similar to that induced by a 475 heterozygous coding truncating CDH1 variants and is expected to lead to germline CDH1 476 monoallelic expression, which would explain the occurrence of GC with a diffuse component 477 in the proband, her brother and mother . Indeed, CD44-Cre/Cdh1fl/fl mice develop stage 1 478 mixed-type gastric cancer (Decourtye-Espiard and Guilford 2023) , correlating well with a 479 tumour with mixed-type histology reduced CDH1 expression. Although we could not assess 480 the impact of the CDH3 inversions, likely inherited from the paternal side, these are more 481 likely to impact CDH3 expression rather than CDH1 expression, as the enhancers are 482 orientation-independent. In fact, we found that inversion of the 20 kb deleted region in F4, 483 does not impact CDH1 expression (data not shown ). The compound heterozygous state 484 found for CDH3 (allele 1: deletion + allele 2: inversions) in F4 proband, leading to nearly 485 complete loss of CDH3 expression, supports the tumour spectrum found in this family, which 486 goes beyond the classical HDGC disease spectrum (Garcia-Pelaez et al. 2023). In contrast 487 to normal stomach epithelia, which does not express CDH3, ovary and skin do (Cao et al. 488 2020), which may explain why biallelic loss of CDH3 may predispose to the ovarian and skin 489 tumours observed in the family. Even if only one of the CDH3 alleles has been inherited by 490 other affected family members, it may still cause cancer predisposition, as combined loss of 491 function occurs for both CDH1 and CDH3 (allele 1: deletion) with a single allele being 492 targeted. Furthermore, the central nervous system tumour occurrence correlates well with 493 the forebrain enhancer activity observed for the RE1 and RE2 contained within the CDH3-494 CNV. CDH3 loss of expression through combined deletion of the CDH3 promoter, exon 1 495 and 2, and/or an inversion of the 5’-region of CDH3 may also trigger ectodermal dysplasia, 496 All rights reserved. No reuse allowed without permission. perpetuity. preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in The copyright holder for thisthis version posted December 2, 2024. ; https://doi.org/10.1101/2024.11.27.24317972doi: medRxiv preprint 15 ectrodactyly and macular dystrophy syndrome or congenital hypotrichosis with juvenile 497 macular dystrophy syndrome, both autosomal recessive (OMIM ID 601553). However, as 498 far as we know, these diseases have not been reported for this family. 499 The CDH1-TANGO6 intergenic sequence also encompasses a noncoding hypomorphic 500 enhancer, but homozygous deletion triggers loss of only half CDH1 expression. In addition, 501 the mean cancer age of onset in family F9 (63.7 years of age) is higher than what is usually 502 found for HDGC families (42±12,16 years of age) (Garcia-Pelaez et al. 2023) . Taken 503 together, a third inactivation event may be required for GC development in F9 carriers, which 504 would take more time to occur. This 3 -hit hypothesis would in part explain the later cancer 505 onset in this family. Deletion of this intergenic region in combination with a 3’-coding deletion 506 of the CDH1 has been shown to trigger extremely early GC onset and high disease 507 penetrance (São José et al. 2023a) . That deletion, in contrast to the current one, strongly 508 reduced CDH1 expression and led to re-wiring of the 3D chromatin structure, impairing the 509 transcriptome, immune, proliferation and adhesion-associated pathways, as compared to a 510 deletion of CDH1 alone (São José et al. 2023a). Thus, we propose that deletion of REs lying 511 in the CDH1-TANGO6 intergenic sequence occurring in both families , modulates CDH1 512 expression and predisposes carriers to GC development. This likely occurs later in life in F9, 513 because the CDH1 coding region remains wild-type (São José et al. 2023a). 514 In a knowledge -driven approach, we next explored rare CNVs and SNVs in cancer -515 syndrome-associated genes and found a germline MLH1 deletion, that has not been 516 previously identified in Lynch syndrome families to date, according to the InSiGHT database. 517 Although MLH1 loss of function predisposes individuals to Lynch syndrome, preferentially 518 increasing the risk of colorectal and endometrial cancers, intestinal-type GC may also occur, 519 with an estimated risk of 13% at age 80 (Møller et al. 2018). In family F15, the clinical history 520 is restricted to GC, but the proband’s tumour presents not only diffuse, but also intestinal 521 histology, which is consistent with MLH1 expression loss and MSI (Møller et al. 2018) . 522 Further supporting the hypothesis that GC predisposition is associated with MLH1 in this 523 family, somatic MLH1 promoter methylation, a hallmark of MSI in sporadic colorectal and 524 GCs, was negative in the proband ’s tumour (Shen et al. 2018) . Our data raise d the 525 hypothesis that CDH1 expression disruption was involved not only in generating the diffuse 526 component of this tumour, but also in shifting the MLH1-related tumour spectrum to the 527 stomach. Indeed, not only were CDH1/E-cadherin mRNA and protein found to be 528 downregulated in the tumour , but also the deleted sequence encompassed enhancer 529 chromatin marks present in normal stomach tissue and absent from normal colon mucosa. 530 Also, an interaction found by 4C-seq between CDH1 and MLH1 promoters further supported 531 a cross -regulation between both genes. Rare i nter-chromosomal interactions have been 532 observed among neighbouring chromosomal territories , such as CISTR-ACT in 533 chromosome 12 and SOX9 in chromosome 17 (Quinodoz et al. 2018; Milad Mokhtaridoost 534 et al. 2024). Altogether, these data indicate that the MLH1 exon 13 sequence, and its intronic 535 vicinity, act as a tissue -specific RE, controlling the expression of CDH1, specifically in the 536 stomach. This regulatory module and the loss of MLH1 expression, as a consequence of 537 deleting an out-of-frame exon, would concomitantly trigger MSI in the emerging GCs. To our 538 All rights reserved. No reuse allowed without permission. perpetuity. preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in The copyright holder for thisthis version posted December 2, 2024. ; https://doi.org/10.1101/2024.11.27.24317972doi: medRxiv preprint 16 knowledge, this is the first report associating a MLH1 exon 13 deletion with CDH1 loss, and 539 its associated stomach-specific REs with GC predisposition. 540 The novel genetic alterations described above , as potential causes for three HDGC -like 541 families, do not affect the structure of the CDH1 locus, but rather involve CDH1 enhancers 542 located either within the CDH1 TAD or located in another chromosome. In all instances, the 543 downregulation of CDH1 occurs due to transcriptional impairment caused by deletion of 544 enhancers for which long distance interactions have been identified herein. Similar 545 mechanisms have been described fo r several genes predisposing to developmental 546 disorders (Will et al. 2017). 547 It is also wor th mentioning that probands from HDGC -like families F4 (CDH3 CNVs and 548 SVs) and F15 ( MLH1 CNV) present mixed GC with a diffuse component, rather than pure 549 DGC, which suggests that impairment of CDH1 REs outside of the CDH1 locus may have 550 wider impacts than the impairment of the CDH1 locus alone. This becomes evident not only 551 in the different histology of GCs, but also in the diversity of cancers in family F4. 552 Given the lack of obvious variants in the remaining 16 families, we explored t he germline 553 regulatory landscape of HDGC undiagnosed families. The germline regulatory landscape of 554 undiagnosed families encompasses mainly downregulation of immune -associated 555 pathways. A specific cluster was found for younger affected individuals, with downregulation 556 of pathways associated with MHC I and natural killer immunity, absent in healthy controls, 557 which are known to play a role in GC immune evasion (Shen et al. 2005; Na et al. 2021) . 558 Interestingly, earlier onset DGC patients display a higher number of deletions in genome -559 wide open chromatin regions in normal stomach tissue , compared to patients with older 560 cancer onset . These “younger-onset”-related regions encompass immune -affected 561 pathways, recapitulating the germline regulatory landscape and pinpointing stomach REs 562 as possible genetic drivers. Altogether, the observed immune impairment in HDGC families 563 may result in lower immunological surveillance in early stages of intramucosal signet -ring 564 cell carcinoma foci, characteristic of HDGC tumours. Indeed, variants in immune 565 surveillance genes, such as MHC class I, have been suggested to play a role in intramucosal 566 signet-ring cell carcinoma foci (Decourtye-Espiard and Guilford 2023) , yet their 567 consolidation is requires additional research. Interestingly, families F15 and F9 lack 568 additional alterations to the MLH1 CNV and the CDH1-TANGO6 intergenic CNVs, 569 suggesting these were sufficient to trigger the HDGC phenotypes. Indeed, F15 (MLH1 CNV) 570 tumour presents high degree of inflammation, highlighting that the lack of germline defects 571 in immune-related pathways allows the immune system to recognize tumor cells. In this 572 particular case, the F15 proband would benefit from immunotherapy treatment (Chao et al. 573 2021). In contrast, family F4 displays impaired immune-associated genes/pathways as well 574 as the CDH3 CNV and SVs. 575 Immune-associated genes and several mucins were also found impaired in other HDGC-576 like families. Supporting the role of MUC4 in this context is the reduced expression described 577 in poorly differentiated GC or signet-ring cell carcinoma (Tamura et al. 2012) , also 578 characteristic of HDGC (Carneiro et al. 1995; van der Post et al. 2016) . Downregulation of 579 the stomach -specific MUC6 gene (Babu et al. 2006) was found to be associated with 580 advanced stages of GC and poor prognosis (Shi and Xi 2021) , supporting an association 581 All rights reserved. No reuse allowed without permission. perpetuity. preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in The copyright holder for thisthis version posted December 2, 2024. ; https://doi.org/10.1101/2024.11.27.24317972doi: medRxiv preprint 17 with HDGC. Indeed, homozygous deletion of MUC6 causes spontaneous GC development 582 with accumulation of CD45+ T-cells in normal and tumour tissue (Arai et al. 2024) . 583 Impairment of T-cell regulatory mechanisms and MHC -related immune surveillance in the 584 germline with co -occurrence of mucins reduced expression may pinpoint a polygenic 585 mechanism for GC development in HDGC families. 586 Altogether, impairment of immune- and mucins-associated pathways/genes may constitute 587 a germline genetic background capable of regulating HDGC development in the presence 588 of a primary unknown triggering mechanism. 589 Taken together, our data resolves the missing heritability in 3 out of 19 HDGC undiagnosed 590 families, corresponding to 1 6% of the entire cohort , and pinpointed novel mechanisms of 591 HDGC predisposition related with CDH1 REs as regulators of CDH1/E-cadherin expression. 592 The finding of a causal mechanism promotes re -engagement with the family , genetic 593 counselling and disease risk management of individuals at risk. The germline landscape of 594 younger DGC patients includes downregulation of immune -associated pathways and 595 several mucins, which may play a role in immune surveillance and DGC development when 596 triggered by a still unknown mechanism. 597 598 599 All rights reserved. No reuse allowed without permission. perpetuity. preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in The copyright holder for thisthis version posted December 2, 2024. ; https://doi.org/10.1101/2024.11.27.24317972doi: medRxiv preprint 18 Acknowledgments 600 This work received funding from: (1) the European Union’s Horizon 2020 research and 601 innovation programme under grant agreement No 779257 (Solve -RD: H2020-SC1-2017-602 Single-Stage-RTD), which provided Ph.D. salaries to JG -P; (2) The European Regional 603 Development Fund (ERDF) through the COMPETE 2020 -Operacional Programme for 604 Competitiveness and Internationalisation (POCI), Portugal 2020, and by Portuguese funds 605 through Fundacao para a Ci encia e Tecnologia (FCT)/Ministerio da Ci encia, Tecnologia e 606 Inovação (3DChroMe: PTDC/BTM-TEC/30164/2017, LEGOH: PTDC/BTM-TEC/6706/2020, 607 DGCoding: 2023.11994.PEX); (3) FCT with salary support for PhD studentships to CSJ (Ref. 608 SFRH/BD/140796/2018), MF (Ref. 2020.05763.BD) and SL (Ref. 2020.05773.BD). Further, 609 this study has received funding from the European Union’s Horizon Europe Coordination 610 and Support action under the Grant Agreement nº 101095483 (PREVENTABLE). Views and 611 opinions expressed are however those of the author(s) only and do not necessarily reflect 612 those of the European Union or The Health and Digital Executive Agency (HaDEA). Neither 613 the European Union nor the granting authority can be held responsible for them. 614 This work was supported by: (1) Doctoral Programme in Biomedicine - Faculty of Medicine, 615 University of Porto, to CSJ and JG-P; (2) Doctoral Programme in Computational Sciences, 616 Faculty of Sciences, University of Porto, to MF ; (3) Doctoral Programme on Cellular and 617 Molecular Biotechnology Applied to Health Sciences , School of Medicine and Biomedical 618 Sciences, University of Porto, Portugal, to SL. 619 620 Author Contributions 621 Carla Oliveira: supervision, Project administration, funding acquisition; Celina S. José and 622 Carla Oliveira: conceptualization, formal analysis; José García-Pelaez, Janine Sanz, Lilian 623 Cordova, Pardeep Kourah: acquisition of clinical and pathological data; Irene Gullo: 624 pathological analysis; Celina S. José, Ana Pedro, Ana André , Silvana Lobo : in vitro 625 experimental work; Marta Ferreira and Celina S. José: bioinformatic analysis; Celina S. 626 José: integration of genetics, clinical and functional data; Celina S. José, Fiona Puntieri and 627 Juliane Glaser: in vivo experimental work; All authors: writing, review, editing and approval 628 of the manuscript. 629 630

631 Abascal F, Acosta R, Addleman NJ, et al (2020) Expanded encyclopaedias of DNA elements 632 in the human and mouse genomes. Nature 583:699 –710. 633 https://doi.org/10.1038/S41586-020-2493-4 634 Arai J, Hayakawa Y, Tateno H, et al (2024) The role of gastric mucins and mucin -related 635 glycans in gastric cancers. Cancer Sci 115:2853 –2861. 636 https://doi.org/10.1111/cas.16282 637 Artus J, Hadjantonakis A -K (2011) Generation of chimeras by aggregation of embryonic 638 stem cells with diploid or tetraploid mouse embryos. Methods Mol Biol 693:37 –56. 639 https://doi.org/10.1007/978-1-60761-974-1_3 640 All rights reserved. No reuse allowed without permission. perpetuity. preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in The copyright holder for thisthis version posted December 2, 2024. ; https://doi.org/10.1101/2024.11.27.24317972doi: medRxiv preprint 19 Babu SD, Jayanthi V, Devaraj N, et al (2006) Expression profile of mucins (MUC2, MUC5AC 641 and MUC6) in Helicobacter pylori infected pre-neoplastic and neoplastic human gastric 642 epithelium. Mol Cancer 5:. https://doi.org/10.1186/1476-4598-5-10 643 Benusiglio PR, Colas C, Guillerm E, et al (2019) Clinical implications of CTNNA1 germline 644 mutations in asymptomatic carriers. Gastric Cancer 22:899 –903. 645 https://doi.org/10.1007/s10120-018-00907-7 646 Blair VR, McLeod M, Carneiro F, et al (2020) Hereditary diffuse gastric cancer: updated 647 clinical practice guidelines. Lancet Oncol 21:e386 –e397. 648 https://doi.org/10.1016/S1470-2045(20)30219-9 649 Bonev B, Cavalli G (2016) Organization and function of the 3D genome. Nat Rev Genet 650 17:661–678. https://doi.org/10.1038/NRG.2016.112 651 Bordeira-Carriço R, Teixeira J, Duque M, et al (2022) Multidimensional chromatin profiling 652 of zebrafish pancreas to uncover and investigate disease -relevant enhancers. Nat 653 Commun 13:1945. https://doi.org/10.1038/s41467-022-29551-7 654 Caldas C, Carneiro F, Lynch HT, et al (1999) Familial gastric cancer: overview and guidelines 655 for management. J Med Genet 36:873–880 656 Cameron DL, Schröder J, Penington JS, et al (2017) GRIDSS: sensitive and specific 657 genomic rearrangement detection using positional de Bruijn graph assembly. Genome 658 Res 27:2050–2060. https://doi.org/10.1101/GR.222109.117 659 Cao J, O’Day DR, Pliner HA, et al (2020) A human cell atlas of fetal gene expression. 660 Science 370:. https://doi.org/10.1126/SCIENCE.ABA7721 661 Carneiro F, Seixas M, Sobrinho -Simões M (1995) New elements for an updated 662 classification of the carcinomas of the stomach. Pathol Res Pract 191:571 –584. 663 https://doi.org/10.1016/S0344-0338(11)80878-2 664 Chao J, Fuchs CS, Shitara K, et al (2021) Assessment of Pembrolizumab Therapy for the 665 Treatment of Microsatellite Instability –High Gastric or Gastroesophageal Junction 666 Cancer Among Patients in the KEYNOTE -059, KEYNOTE -061, and KEYNOTE -062 667 Clinical Trials. JAMA Oncol 7:895. https://doi.org/10.1001/jamaoncol.2021.0275 668 Chi X, Zheng Q, Jiang R, et al (2019) A system for site-specific integration of transgenes in 669 mammalian cells. PLoS One 14:e0219842. 670 https://doi.org/10.1371/journal.pone.0219842 671 Cova G, Glaser J, Schöpflin R, et al (2023) Combinatorial effects on gene expression at the 672 Lbx1/Fgf8 locus resolve split -hand/foot malformation type 3. Nat Commun 14:. 673 https://doi.org/10.1038/S41467-023-37057-Z 674 Decourtye-Espiard L, Guilford P (2023) Hereditary Diffuse Gastric Cancer. Gastroenterology 675 164:719–735. https://doi.org/10.1053/J.GASTRO.2023.01.038 676 Dixon JR, Gorkin DU, Ren B (2016) Chromatin Domains: The Unit of Chromosome 677 Organization. Mol Cell 62:668–680. https://doi.org/10.1016/j.molcel.2016.05.018 678 Fishilevich S, Nudel R, Rappaport N, et al (2017) GeneHancer: genome-wide integration of 679 enhancers and target genes in GeneCards. Database (Oxford) 2017:. 680 https://doi.org/10.1093/database/bax028 681 Franke M, Ibrahim DM, Andrey G, et al (2016) Formation of new chromatin domains 682 determines pathogenicity of genomic duplications. Nature 538:265 –269. 683 https://doi.org/10.1038/nature19800 684 Furlong EEM, Levine M (2018) Developmental enhancers and chromosome topology. 685 Science (1979) 361:1341–1345. https://doi.org/10.1126/science.aau0320 686 Garcia-Pelaez J, Barbosa-Matos R, Lobo S, et al (2023) Genotype-first approach to identify 687 associations between CDH1 germline variants and cancer phenotypes: a multicentre 688 study by the European Reference Network on Genetic Tumour Risk Syndromes. Lancet 689 Oncol 24:91–106. https://doi.org/10.1016/S1470-2045(22)00643-X 690 All rights reserved. No reuse allowed without permission. perpetuity. preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in The copyright holder for thisthis version posted December 2, 2024. ; https://doi.org/10.1101/2024.11.27.24317972doi: medRxiv preprint 20 Geeven G, Teunissen H, de Laat W, de Wit E (2018) peakC: a flexible, non-parametric peak 691 calling package for 4C and Capture -C data. Nucleic Acids Res 46:e91. 692 https://doi.org/10.1093/nar/gky443 693 Geoffroy V, Herenger Y, Kress A, et al (2018) AnnotSV: an integrated tool for structural 694 variations annotation. Bioinformatics 34:3572–3574. 695 https://doi.org/10.1093/bioinformatics/bty304 696 Gibson DG, Young L, Chuang R-Y, et al (2009) Enzymatic assembly of DNA molecules up 697 to several hundred kilobases. Nat Methods 6:343 –345. 698 https://doi.org/10.1038/nmeth.1318 699 Guilford P, Hopkins J, Harraway J, et al (1998) E -cadherin germline mutations in familial 700 gastric cancer. Nature 392:402–405. https://doi.org/10.1038/32918 701 Gullo I, Grillo F, Mastracci L, et al (2020) Precancerous lesions of the stomach, gastric 702 cancer and hereditary gastric cancer syndromes. Pathologica 112:166 –185. 703 https://doi.org/10.32074/1591-951X-166 704 Hansford S, Kaurah P, Li -Chang H, et al (2015) Hereditary Diffuse Gastric Cancer 705 Syndrome: CDH1 Mutations and Beyond. JAMA Oncol 1:23 –32. 706 https://doi.org/10.1001/jamaoncol.2014.168 707 Krijger PHL, Geeven G, Bianchi V, et al (2020) 4C -seq from beginning to end: A detailed 708 protocol for sample preparation and data analysis. Methods 170:17 –32. 709 https://doi.org/10.1016/J.YMETH.2019.07.014 710 Layer RM, Chiang C, Quinlan AR, Hall IM (2014) LUMPY: a probabilistic framework for 711 structural variant discovery. Genome Biol 15:. https://doi.org/10.1186/GB -2014-15-6-712 R84 713 Lobe CG, Koop KE, Kreppner W, et al (1999) Z/AP, a double reporter for cre -mediated 714 recombination. Dev Biol 208:281–292. https://doi.org/10.1006/dbio.1999.9209 715 Lobo S, Benusiglio PR, Coulet F, et al (2021) Cancer predisposition and germline CTNNA1 716 variants. Eur J Med Genet 64:104316. https://doi.org/10.1016/j.ejmg.2021.104316 717 Lupiáñez DG, Spielmann M, Mundlos S (2016) Breaking TADs: How Alterations of 718 Chromatin Domains Result in Disease. Trends in Genetics 32:225 –237. 719 https://doi.org/10.1016/j.tig.2016.01.003 720 Majewski IJ, Kluijt I, Cats A, et al (2013) An α-E-catenin (CTNNA1) mutation in hereditary 721 diffuse gastric cancer. J Pathol 229:621–629. https://doi.org/10.1002/path.4152 722 Milad Mokhtaridoost, Jordan J Chalmers, Marzieh Soleimanpoor, et al (2024) Inter -723 chromosomal contacts demarcate genome topology along a spatial gradient. Nat 724 Commun 15(1):9813: 725 Møller P, Seppälä TT, Bernstein I, et al (2018) Cancer risk and survival in path_MMR carriers 726 by gene and gender up to 75 years of age: a report from the Prospective Lynch 727 Syndrome Database. Gut 67:1306 –1316. https://doi.org/10.1136/GUTJNL -2017-728 314057 729 Moore JE, Purcaro MJ, Pratt HE, et al (2020) Expanded encyclopaedias of DNA elements 730 in the human and mouse genomes. Nature 583:699 –710. 731 https://doi.org/10.1038/s41586-020-2493-4 732 Na HY, Park Y, Nam SK, et al (2021) Prognostic significance of natural killer cell-associated 733 markers in gastric cancer: quantitative analysis using multiplex immunohistochemistry. 734 J Transl Med 19:. https://doi.org/10.1186/S12967-021-03203-8 735 Oliveira C, Pinheiro H, Figueiredo J, et al (2015) Familial gastric cancer: genetic 736 susceptibility, pathology, and implications for management. Lancet Oncol 16:e60 -70. 737 https://doi.org/10.1016/s1470-2045(14)71016-2 738 Oliveira C, Senz J, Kaurah P, et al (2009a) Germline CDH1 deletions in hereditary diffuse 739 gastric cancer families. Hum Mol Genet 18:1545 –1555. 740 https://doi.org/10.1093/hmg/ddp046 741 All rights reserved. No reuse allowed without permission. perpetuity. preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in The copyright holder for thisthis version posted December 2, 2024. ; https://doi.org/10.1101/2024.11.27.24317972doi: medRxiv preprint 21 Oliveira C, Seruca R, Carneiro F (2009b) Hereditary gastric cancer. Best Pract Res Clin 742 Gastroenterol 23:147–157. https://doi.org/10.1016/j.bpg.2009.02.003 743 Oliveira C, Sousa S, Pinheiro H, et al (2009c) Quantification of epigenetic and genetic 2nd 744 hits in CDH1 during hereditary diffuse gastric cancer syndrome progression. 745 Gastroenterology 136:2137–2148. https://doi.org/10.1053/j.gastro.2009.02.065 746 Phan MH, Zehnder TM, Puntieri F, et al (2024) Conservation of Regulatory Elements with 747 Highly Diverged Sequences Across Large Evolutionary Distances 748 Pinheiro H, Bordeira-Carrico R, Seixas S, et al (2010) Allele-specific CDH1 downregulation 749 and hereditary diffuse gastric cancer. Hum Mol Genet 19:943 –952. 750 https://doi.org/10.1093/hmg/ddp537 751 Poplin R, Chang PC, Alexander D, et al (2018) A universal SNP and small-indel variant caller 752 using deep neural networks. Nat Biotechnol 36:983. https://doi.org/10.1038/NBT.4235 753 Quinodoz SA, Ollikainen N, Tabak B, et al (2018) Higher -Order Inter-chromosomal Hubs 754 Shape 3D Genome Organization in the Nucleus. Cell 174:744 -757.e24. 755 https://doi.org/10.1016/j.cell.2018.05.024 756 Rao SSP, Huntley MH, Durand NC, et al (2014) A 3D map of the human genome at kilobase 757 resolution reveals principles of chromatin looping. Cell 159:1665 –1680. 758 https://doi.org/10.1016/J.CELL.2014.11.021 759 Rausch T, Zichner T, Schlattl A, et al (2012) DELLY: structural variant discovery by integrated 760 paired-end and split -read analysis. Bioinformatics 28:. 761 https://doi.org/10.1093/BIOINFORMATICS/BTS378 762 São José C, Garcia -Pelaez J, Ferreira M, et al (2023a) Combined loss of CDH1 and 763 downstream regulatory sequences drive early-onset diffuse gastric cancer and increase 764 penetrance of hereditary diffuse gastric cancer. Gastric Cancer. 765 https://doi.org/10.1007/S10120-023-01395-0 766 São José C, Pereira C, Ferreira M, et al (2023b) 3D Chromatin Architecture Re -Wiring at 767 the CDH3/CDH1 Loci Contributes to E -Cadherin to P -Cadherin Expression Switch in 768 Gastric Cancer. Biology (Basel) 12:803. https://doi.org/10.3390/biology12060803 769 Shen S, Chen X, Li H, et al (2018) MLH1 Promoter Methylation and Prediction/Prognosis of 770 Gastric Cancer: A Systematic Review and Meta and Bioinformatic Analysis. J Cancer 771 9:1932–1942. https://doi.org/10.7150/jca.23284 772 Shen YQ, Zhang JQ, Miao FQ, et al (2005) Relationship between the downregulation of HLA 773 class I antigen and clinicopathological significance in gastric cancer. World J 774 Gastroenterol 11:3628–3631. https://doi.org/10.3748/WJG.V11.I23.3628 775 Shi D, Xi X (2021) Regulation of MUC6 Methylation Correlates with Progression of Gastric 776 Cancer. Yonsei Med J 62:1005. https://doi.org/10.3349/ymj.2021.62.11.1005 777 Spielmann M, Lupiáñez DG, Mundlos S (2018) Structural variation in the 3D genome. Nat 778 Rev Genet 19:453–467. https://doi.org/10.1038/S41576-018-0007-0 779 Stemmler MP, Hecht A, Kemler R (2005) E -cadherin intron 2 contains cis -regulatory 780 elements essential for gene expression. Development 132:965 –76. 781 https://doi.org/10.1242/dev.01662 782 Stemmler MP, Hecht A, Kinzel B, Kemler R (2003) Analysis of regulatory elements of E -783 cadherin with reporter gene constructs in transgenic mouse embryos. Developmental 784 Dynamics 227:238–245. https://doi.org/10.1002/dvdy.10301 785 Tamura Y, Higashi M, Kitamoto S, et al (2012) MUC4 and MUC1 expression in 786 adenocarcinoma of the stomach correlates with vessel invasion and lymph node 787 metastasis: an immunohistochemical study of early gastric cancer. PLoS One 7:. 788 https://doi.org/10.1371/JOURNAL.PONE.0049251 789 Tedaldi G, Pirini F, Tebaldi M, et al (2019) Multigene Panel Testing Increases the Number of 790 Loci Associated with Gastric Cancer Predisposition. Cancers (Basel) 11:. 791 https://doi.org/10.3390/CANCERS11091340 792 All rights reserved. No reuse allowed without permission. perpetuity. preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in The copyright holder for thisthis version posted December 2, 2024. ; https://doi.org/10.1101/2024.11.27.24317972doi: medRxiv preprint 22 van de Werken HJ, de Vree PJ, Splinter E, et al (2012) 4C technology: protocols and data 793 analysis. Methods Enzymol 513:89 –112. https://doi.org/10.1016/b978 -0-12-391938-794 0.00004-5 795 van der Post RS, Gullo I, Oliveira C, et al (2016) Histopathological, Molecular, and Genetic 796 Profile of Hereditary Diffuse Gastric Cancer: Current Knowledge and Challenges for the 797 Future. Adv Exp Med Biol 908:371–391. https://doi.org/10.1007/978-3-319-41388-4_18 798 Velho S, Moutinho C, Cirnes L, et al (2008) BRAF, KRAS and PIK3CA mutations in colorectal 799 serrated polyps and cancer: Primary or secondary genetic events in colorectal 800 carcinogenesis? BMC Cancer 8:255. https://doi.org/10.1186/1471-2407-8-255 801 Will AJ, Cova G, Osterwalder M, et al (2017) Composition and dosage of a multipartite 802 enhancer cluster control developmental expression of Ihh (Indian hedgehog). Nat 803 Genet 49:1539–1545. https://doi.org/10.1038/ng.3939 804 Zuin J, Roth G, Zhan Y, et al (2022) Nonlinear control of transcription through enhancer –805 promoter interactions. Nature 604:571 –577. https://doi.org/10.1038/s41586 -022-806 04570-y 807 808 All rights reserved. No reuse allowed without permission. perpetuity. preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in The copyright holder for thisthis version posted December 2, 2024. ; https://doi.org/10.1101/2024.11.27.24317972doi: medRxiv preprint D E A F B C Family Gene Syndrome Location hg38 SV size (bp) SV type Perturbed region AF F4 CDH3 Cancer chr16:6864534 6-68665730 20384 DEL Prom, exon1-2, intron 1-2 novel F4 CDH3 Cancer chr16:6861614 9-68647665 31516 INV Prom, exon1-2, intron 1-2 novel F4 CDH3 Cancer chr16:6866349 2-68680068 16576 INV Intron 2-7 exon 3-8 novel F9 CDH1-TANGO6 intergenic - chr16:6883749 4-68837533 39 DEL Intergenic novel Magnification: 40x Magnification: 40x Figure 1. Genomic CNVs in CDH1 TAD. A) Description of identified CNVs, B) Candidate REs, chromatin marks, and CDH1 interactors in normal stomach tissue, C) CDH1 mRNA and protein expression of edited MKN74 clones, D) Hematoxylin and eosin of tumour tissue from proband with CDH3 CNV, E) E- cadherin expression of tumour tissue from proband with CDH3 CNV (immunohistochemistry), F) Enhancer assay of candidate REs in mouse embryos, blue precipitate represent β-galactosidase expression and tissue- specific enhancer activity. GeneHancer: gradient of red and grey represent promoters and enhancers, respectively, grading from low to high activity. No enhancer control All rights reserved. No reuse allowed without permission. perpetuity. preprint (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in The copyright holder for thisthis version posted December 2, 2024. ; https://doi.org/10.1101/2024.11.27.24317972doi: medRxiv preprint I1 GC 30s II1 OC 20sSkC50s 2 MGC 30s 2 CNS 40s 34 BCGC I II IIIDGC 60sGC 40s C 80sGC 60sEC 60sGC 70s CRC 50s GC 60s A B F4 F9 Legend:Gastric cancer (GC), mixed (MGC), diffuse (DGC)Central nervous system (CNS)Breast cancer (BC)Skin cancer (SkC)Ovarian cancer (OC)Esophageal cancer (EC)Colorectal cancer (CRC)Other cancer types (C) 3 DAC 50s Duodenum adenocarcinoma (DAC) CLP Cleft lip/palate (CLP) Figure 2. Family pedigrees. A) Representative family pedigree of F4 carrying the CDH3 CNV, B) Representative family pedigree of F9 carrying the CDH1-TANGO6 intergenic CNV. Black fill represents gastric cancer, red represent central nervous system cancer, light blue represents breast cancer, yellow represent skin cancer, green represent ovarian cancer, grey represent endometrial cancer and blue represent colorectal cancer. Arrow indicates the proband. DGC: diffuse gastric cancer, MGC: mixed gastric cancer, GC: gastric cancer, DAC: adenocarcinoma of the duodenum, CNS: central nervous system cancer, BC: breast cancer, SC: skin cancer, OC: ovarian cancer, EC: Esophageal cancer, CRC: colorectal cancer, CLP: cleft lip/palate. 1 1234567 1 A I I II F D G H JK EC B F15 FamilyGeneSyndromeLocation hg38SV size (bp)SV typePerturbed regionAFF15MLH1Lynch syndromechr3:37039111-370418982787DELIntron 13-14, exon 14novel Legend:Gastric cancer (GC), mixed (MGC), diffuse (DGC) Magnification: 40x Magnification: 40x Magnification: 40xMagnification: 40x Figure 3. Genomic CNVs in cancer associated syndromes. A) Description of CNV identified, B) Microsatellite instability (MSI) analysis, C) Hematoxylin and eosin of tumour tissue from proband with MLH1 CNV, D) MLH1 expression of tumour tissue from proband with MLH1CNV (immunohistochemistry), E) Representative family pedigree of F15. Black fill represents gastric cancer development. DGC: diffuse gastric cancer, GC: gastric cancer. F) Chromatin marks and accessibility in normal stomach and colon tissue from ENCODE, G) E-cadherin expression of tumour tissue (intestinal component) from proband with MLH1 CNV (immunohistochemistry), H) E-cadherin expression of tumour tissue (diffuse component) from proband with MLH1 CNV (immunohistochemistry), I) CDH1 mRNA expression in the tumour, J-K) CDH1 mRNA and protein expression of edited MKN74 clones. GC 40s 1 MGC 40s 1 DGC 30s 2 A C Misregulated pathways:Immune responseT-cell regulationT-cell-mediated cytotoxicityT-cell response to tumour cellCell killingNatural killer regulationMHC-I and MHC-II regulation ControlsHDGC-like families Misregulated pathways:Immune responseT-cell regulationLeukocyte cytotoxicityT-cell response to tumour cellCell killingNatural killer regulationMHC-I and MHC-II regulationInterferon gamma production F4F5F2F8F19_1F19_2F11F16F14F10F19_3F6F9F12F3F18F13F1F17F7F15Controls Coding CNVs and truncating SNVs F16F2F11F7F14ControlControlF18ControlF9ControlF6 F17F19_3F9_2F19_1F8F4F5F10F3F1Controls Deletions overlapping accessible chromatin Figure 4. CNVs and SNVs misregulated pathways in HDGC undiagnosed patients. A) Heatmap representing all impaired pathways in HDGC undiagnosed patients, B) Cluster of immune-associated impaired pathways in HDGC undiagnosed patients and absent in control individuals, C) Cluster of CNVs overlapping accessible chromatin regions in immune-associated pathways. Red represents probands with cancer age of onset ≤50 years of age. B Legend:CNV in cancer-associated genesCNVs in CDH1 TADCoding CNVs and truncating SNVsDeletions overlapping stomach accessible chromatin B A FamilyGermline defectsFamily historyHDGC 2020 criteriaMeet criteria for other syndromesF15MLH1+/-MGC 40s, DGC 30s, GC 40s1F9CDH1-TANGO6 intergenic+/-DGC 60s, GC 60s, GC 40s, GC 60s, ESC 60s, CRC 50s1 F4CDH3+/-, Immune-related genesMGC 30s,GC 30s, GC 30s, BC, CNS 40s, CRC 30s, Sk 50s, OC 20s1LS (CRC < 50), HBOC (BC + OC)F5Immune-related genesDGC 20s, unknown 50s, unknown 50s4F16Immune-related genesDGC 20s, GC 60s, SARC 70s, Leuk 10s, Sk 40s, 50s1 F19Immune-related genesIGC 30s, UtC 20s, *DGC 60s, GC 60s, 60s, **IDC 50s, UtC 50s, BrC , PancC 80s, LC 80s, LC, BlC 60s, 80s, 80s, PC 80s, 80s, BSO, CRC 60s, CRC, CC 20s1HBOC (BC + PancC) F18Immune-related genesDGC 30s, GC 40s, BC 70s, LBC 60s, Lymph 60s1F2Immune-related genesDGC 40s, GC 40s, BC 20s, Sk , HNC, unknown 80s, -, BrC, LC, CRC 40s1LS (CRC < 50), HBOC (BC < 45)F14Immune-related genesDGC 40s, BC 40s, DGC 30s, gastric LMS 30s, IDC 50s, OC or DGC 30s, CRC 30s1F11Immune-related genesDGC 40s, GC 40s, GC 60s, GC, LBC 40s, LBC 50s, IDC 50s, BC 50s, BC1, 2, 7HBOC (BC < 45)F1Immune-related genesDGC 60s, PRSC 50s, GC, GC, GC, 60s, DGC 60s, BC 70s, MM 70s, OC 30s, CRC 40s, NHL 20s1F8Immune-related genesMGC 70s,GC 30s, GC 70s, PC 60s, CRC 70sNoF6DGC 30s, DGC 30s, Leuk 60s, BlC 60s, PancC 70s, OC 40s1F13DGC 40s, DGC 40s, GC 70s, BC, BlC, LC 60s, 60s, 50s, PC 70s, 70s, Sk 50s, CC1F17DGC 50s, GC 40s, BC 70s, CRC 80s, RC, BoC1F7DGC 50s, GC 80s, IDC 50s, CRC, CRC, CRC, CRC4LS (>3 CRCs)F12DGC 50s, GC 40s, CRC 80s, 50s, ESC 70s1F10LBC 60s, GC 50s, GC 80s, GC, BC 50s, PancC, HNC , MM 40s, TC 40sNoHBOC (BC + PancC)F3DGC 60s, GC 30s, GC 70s, MGC 70s, PRSC 60s, LC smoker 50s, CRC 70s1 Figure 5. Main findings solving the missing heritability in HDGC patients. A) Heatmap representing misregulated genes in HDGC undiagnosed families, Blue represents ultra-rare (AF<0.00001) coding CNVs and truncating SNVs found from WGS sequencing of HDGC patients. Yellow represent CNVs found overlapping CDH1 TAD. Orange square represents pathways overlapping accessible chromatin in normal stomach, B) Description of the genetic defects, family history, clinical criteria and ancestry of HDGC undiagnosed families. GC: gastric cancer, DGC: diffuse gastric cancer, BC: breast cancer, LBC: lobular breast cancer, PRSS: prostate cancer, MGC: mixed-type gastric cancer, IGC: intestinal gastric cancer, IDC: invasive ductal breast cancer, CRC: colorectal cancer, EsoC: esophageal cancer, RC: renal cancer, Panc: pancreatic cancer, BldC: blader cancer, Sarc: sarcoma, OvC: ovarian cancer, Cerv: cervical cancer, UtC: uterine cancer, SkC: skin cancer, MM: melanoma, BrC: CNS: central nervous system cancer, HN: head and neck, Leuk: leukemia, NHL: non-hodkins lymphoma, Lymp: lymphoma, LC: lung cancer. In F19, * represent phenotype in individual F19_2 and ** represent phenotype in individual F19_3.