Background Familial intestinal gastric cancer (FIGC) remains genetically unexplained and without testing/clinical criteria. Herein, we characterised the age of onset and disease spectrum of 50 FIGC families and searched for genetic causes potentially underlying a monogenic or an oligogenic/polygenic inheritance pattern.
Methods Normal and tumour DNA from 50 FIGC probands were sequenced using Illumina custom panels on MiSeq, and their respective germline and somatic landscapes were compared with corresponding landscapes from sporadic intestinal gastric cancer (SIGC) and hereditary diffuse gastric cancer cohorts.
Results The most prevalent phenotype in FIGC families was gastric cancer, detected in 138 of 208 patients (50 intestinal gastric cancer probands and 88 unknown gastric cancer histology relatives), followed by colorectal and breast cancers. After excluding benign and intronic variants lacking impact in splicing, 12 rare high-quality variants were found exclusively in 11 FIGC probands. Only two probands carried potentially deleterious variants, but lacked somatic second-hits, weakly supporting the monogenic hypothesis for FIGC. However, FIGC probands developed gastric cancer at least 10 years earlier and carried more TP53 germline common variants than SIGC (p=4.5E-03); FIGC and SIGC could be distinguished by specific germline and somatic variant profiles; there was an excess of FIGC tumours presenting microsatellite instability (38%); and FIGC tumours displayed significantly more somatic common variants than SIGC tumours (p=4.2E-06).
Conclusion This study proposed the first data-driven testing criteria for FIGC families, and supported FIGC as a genetically determined, likely polygenic, gastric cancer-predisposing disease, with earlier onset and distinct from patients with SIGC at the germline and somatic levels.
- cancer: gastric
- molecular genetics
- clinical genetics
- evidence based practice
Statistics from Altmetric.com
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.
Gastric cancer (GC) ranks as the fifth most commonly diagnosed and the third deadliest cancer worldwide, with a 5-year overall survival rate of less than 25%.1–3 The two main histotypes, intestinal and diffuse, are recognised by distinct morphological, molecular, aetiological, clinical and epidemiological features.4–6
While most GCs are sporadic, 10% show familial clustering. Among these, only 1%–3% are thought to be hereditary, falling into one of the following syndromes: hereditary diffuse gastric cancer (HDGC), gastric adenocarcinoma and proximal polyposis of the stomach (GAPPS), and familial intestinal gastric cancer (FIGC).7–9 Germline mutations and deletions within the E-cadherin gene (CDH1) are the main cause of HDGC and affect 14%–40% of families.10–12 Additionally, while α-E-catenin gene (CTNNA1) mutations have been proven to cause HDGC, germline variants in homologous recombination DNA repair genes, such as PALB2, await confirmation as potential causes of disease in mutation-negative HDGC families.13–15 Concerning GAPPS, APC promoter 1B point mutations are the underlying cause of this syndrome in several families.16 Unlike HDGC and GAPPS, FIGC remains genetically unexplained, despite the recent report of PALB2 germline mutations in three individuals with intestinal tumours but lacking family history of GC.14 17
FIGC is characterised by an autosomal dominant inheritance pattern of intestinal gastric cancer (IGC), without gastric polyposis, and is defined according to GC incidence, as agreed by the International Gastric Cancer Linkage Consortium.9 Therefore, in high incidence countries, the diagnostic criteria is analogous to the Amsterdam criteria for hereditary non-polyposis colorectal cancer (HNPCC): at least three relatives should have IGC and one of them should be a first-degree relative of the other two; at least two successive generations should be affected; and in one of the relatives, GC should be diagnosed before the age of 50. In countries with low incidence, the following criteria are used: at least two first-degree relatives (FDR) or second-degree relatives (SDR) affected by IGC, one diagnosed before the age of 50; or three or more relatives with IGC at any age.9 Because no novel data exist supporting familial aggregation of IGC, no specific tumour spectrum has been defined, and no data support a particular age of onset; hence, the above criteria have never been revisited or validated. Therefore, these families are often neglected and rarely followed in oncogenetic consultations.
GC also develops in the context of other inherited cancer predisposition syndromes.18 In particular, GC has been identified in the tumour spectrum of Lynch syndrome, Li-Fraumeni syndrome, Peutz-Jeghers syndrome, familial adenomatous polyposis, juvenile polyposis, and hereditary breast and ovarian cancer, among others.19–22 Therefore, genes causing hereditary cancer susceptibility syndromes, even if only slightly associated with GC susceptibility, would be good candidates to test as potential FIGC causal genes.
Herein, we used a next-generation sequencing approach to interrogate a panel of genes implicated in upper gastrointestinal tract cancer, or in cancer susceptibility syndromes, across 50 probands with familial aggregation of IGC from Tuscany, a region from Italy with high incidence of GC.23 The access to a highly homogeneous FIGC cohort, the largest ever studied, and its comparison with an HDGC series and a cohort of sporadic intestinal gastric cancer (SIGC) allowed us to define three objectives and to extend the current knowledge on FIGC predisposition: (1) characterise the age of cancer onset and disease spectrum of our FIGC cohort; (2) search for evidence for a Mendelian and monogenic pattern of inheritance; and (3) search for evidence of alternative oligogenic/polygenic modes of inheritance.
Herein, we gathered evidence that FIGC is likely a genetically determined, GC-predisposing disease, different at the clinical, germline and somatic levels from SIGC and HDGC. We further proposed the first testing criteria for FIGC families.
Fifty FIGC and 17 HDGC-CDH1 mutation-negative probands were admitted at the Division of General Surgery and Surgical Oncology, University of Siena, Italy. The selection of FIGC families was based on the following criteria: (1) proband presenting with GC of intestinal histology; (2) familial aggregation of GC; (3) family history of cancer, other than gastric; (4) negative genetic test for germline CDH1 coding sequence mutations (exclusion of HDGC); and (5) negative genetic test for germline for the promoter 1B of APC (exclusion of GAPPS). The 17 HDGC probands were negative for CDH1 germline coding mutations and selected as a control group. Forty-seven patients with SIGC were collected in Portugal.
Multigene panel sequencing, variant calling and filtering
DNA from normal gastric mucosa (germline) and tumour tissue from 50 FIGC and 17 HDGC-CDH1 mutation-negative probands were sequenced using three Illumina MiSeq custom panels: TruSeq Custom Amplicon Assay 1, TruSeq Custom Amplicon Assay 2 and Nextera custom panel (online supplementary table 1). The selection of genes deposited in each panel was based on their implication in upper gastrointestinal tract cancers or in cancer susceptibility syndromes identified through literature review (online supplementary table 2). FASTQ files were aligned to the RefSeq Human Genome GRCh38 using bwa-mem, and variants were called using Samtools.24 25 Called variants were defined as germline or somatic by normal-tumour pair comparison and annotated with Ensembl and Catalogue Of Somatic Mutations In Cancer (COSMIC (FATHMM- Functional Analysis through Hidden Markov Models).26 27 High-quality (HQ) germline or somatic variants were defined as presenting ≥20 reads per allele and genotype quality ≥90 and call quality ≥100. Next, all single nucleotide polymorphism database (dbSNP) identifiers available for FIGC germline variants (regardless of quality criteria) were screened in four European populations from 1000 Genomes: (1) 107 normal individuals from Tuscany (Italy, TSI); (2) 91 normal individuals from Great Britain (GBR); (3) 99 normal individuals from Finland (FIN); and (4) 107 normal individuals from Spain (IBS).28 Germline variants without dbSNP identifiers available in the 1000 Genomes were screened using Ensembl VEP for truncating consequences. Detected truncating variants presented on average less than four reads, that is, were of low quality and discarded. FIGC germline, rare HQ exclusive variants were selected if they (1) displayed genotypes in FIGCs distinct from GBR, FIN and IBS populations and below 1% in the TSI population; (2) presented ≥20 reads per allele, genotype quality ≥90 and call quality ≥100; (3) displayed genotypes distinct from HDGCs and SIGCs; and (4) presented allele frequency in ExAC and gnomAD populations below 1%.29
Validation of FIGC germline, rare HQ exclusive variants by Sanger sequencing
Twelve out of 32 FIGC germline, rare HQ exclusive variants were validated by PCR-Sanger sequencing. Briefly, 20–50 ng of DNA from normal and matched tumour was amplified using Multiplex PCR Kit (Qiagen) and custom primers flanking each variant. PCR products were purified with ExoSAP-IT Express (Applied Biosystems) and sequenced on an ABI3100 Genetic Analyzer using BigDye Terminator V.3.1 Cycle Sequencing Kit (Applied Biosystems).
Intronic germline variants were analysed using the splice site prediction software NetGene2 V.2.4.30
Somatic second-hit analysis
Loss of heterozygosity (LOH) and somatic second mutations were determined by calculating the variant allele frequency (VAF) and screening genes with FIGC germline, rare HQ exclusive variants, respectively. In particular, VAF was calculated by dividing the number of reads for the variant allele by the total number of reads both for the normal and for the corresponding tumour samples. LOH was defined when more than 20% increase of VAF over normal was observed.
Germline and somatic landscape analysis of 50 FIGC cases
FIGC germline and somatic landscapes were analysed on a per-variant and per-gene basis, considering the number of FIGC germline, rare HQ exclusive variants detected per proband (0, 1 or >1). The similarities/differences for the germline and somatic variant and gene landscapes per FIGC class were analysed using unsupervised hierarchical clustering using R package ggplot2 for heatmap and dendrogram construction.31 For somatic variant/gene landscape analysis, FIGC classes were also divided according to microsatellite instable status and compared using analysis of variance statistics with R. The number of microsatellite instable (MSI) and microsatellite stable (MSS) tumours per FIGC class was compared using Pearson’s χ2 test.
Comparison of germline and somatic landscapes for FIGC, SIGC and HDGC
VCF files obtained from whole genome sequencing (Complete Genomics platform) of 47 SIGCs and VCF files of 17 HDGCs were analysed to detect germline and somatic variants, using the same germline/somatic variant definition and sequencing quality criteria previously described for FIGC cases. Of note, due to the differential resolution between whole genome sequencing and targeted sequencing, only variants detected in the 47 SIGCs in the same regions targeted by the custom panels were selected for downstream analysis.
Germline and somatic landscapes of FIGC, SIGC and HDGC cases were performed on a per-gene basis: each gene was classified as presenting 0 or ≥1 germline/somatic variants. Germline and somatic joint landscape was defined by counting the number of germline and somatic variants for each gene, which was classified as displaying no germline or somatic variants; ≥1 germline and 0 somatic variants; 0 germline and ≥1 somatic variants; or ≥1 germline and ≥1 somatic variants. Results were plotted in a heatmap and a dendrogram, and principal component analysis was performed using R. The frequency of genes with germline/somatic variants in FIGCs, SIGCs and HDGCs was calculated, and genes with a frequency difference ≥50% were represented in a bar plot and in a heatmap using R.
Age of onset and disease spectrum in FIGC
Of the 50 FIGC probands (table 1), 18 were female and 32 were male. The mean age at diagnosis was 71.8±8.0 years. From the 50 families depicted in table 1, 5 (10%) had >1 FDR with GC (mean age: 68.8±7.5 years); 14 (28%) had concomitantly FDR and SDR or FDR and third-degree relatives with GC (mean age: 68.7±8.4 years); 29 (58%) had a single FDR with GC (mean age: 73.6±7.2 years); and 2 (4%) had only SDR affected with GC (mean: 74±15.6 years).
When considering the disease spectrum in these FIGC families, 19 different phenotypes have been observed affecting 208 family members (figure 1, table 1). The most prevalent phenotype was GC, detected in 138 of 208 (66.3%) family members: 50 probands with IGC and 88 additional patients with unknown GC histology. The second and third most prevalent phenotypes were colorectal/colon and breast cancer observed in nine patients from seven families. Of note, eight patients from six families were affected with gastric ulcer, a non-cancerous lesion, which is the third most common disease phenotype in this cohort. Besides these phenotypes, positive history of lung cancer was observed in six families; leukaemia in five families; laryngotracheal and hepatobiliary cancer in four families; osteosarcoma in three families; prostate, liver, melanoma, gynaecological, bladder and brain cancers were detected in two families each; and thyroid, kidney and oral cancer in one family. Moreover, 11 families had relatives affected by an unidentified type of cancer that often coexisted with other cancer types such as colon, leukaemia, breast, liver and prostate.
Germline and somatic variant discovery across FIGC probands
Multigene panel sequencing analysis of normal-tumour DNA of 50 FIGC probands revealed a total of 10 062 variants (≥1 read covering the alternative allele). Of these, 4998 (49.7%) were detected in normal DNA and defined as germline variants. The remaining 5064 (50.3%) were called as somatic variants due to exclusive presence in tumour DNA. We started by exploring germline variants, focusing on rare variants in single genes (monogenic hypothesis) or variants co-occurring in several genes, regardless of their population frequency (oligogenic/polygenic hypothesis).
Monogenic hypothesis: FIGC-associated rare germline variants and somatic second-hits
To identify rare germline FIGC-predisposing variants, we performed a systematic analysis of all germline variants, focusing on their frequency across normal populations and GC cohorts, and sequencing quality.
We identified 4998 germline variants in the 50 patients with FIGC (figure 2A). From the 4998 FIGC germline variants, the genotype frequency of 1038 (20.8%) was available for four 1000 Genomes European populations.28 From the 79.2% of variants absent from 1000 Genomes, only 1.3% (n=53) presented truncating effects, however supported on average by less than four reads, that is, of very low quality and hence confidently discarded. From the 1038 variants present in 1000 Genomes, 121 (11.7%) presented genotypes absent from the four populations screened. Of these 121 variants, only 60 presented the abovementioned sequencing quality criteria. From these, 43 variants were exclusively detected in FIGC comparing with HDGC-CDH1 mutation-negative and SIGC cohorts. With regard to the 17 discarded variants, all were found in at least one HDGC proband and none in SIGC.
From the 43 germline, rare and HQ FIGC-exclusive variants, 31 (72.1%) displayed very low allele frequency in all ExAC and gnomAD populations (figure 2A, online supplementary table 3), and were present in 21 of 50 (42%) FIGC probands (7 missense, 7 3’untranslated (UTR), 2 5’UTR, 12 intronic and 3 synonymous in 18 genes; online supplementary table 4). Fifteen probands carried a single variant and six exhibited co-occurrence of two or more variants (online supplementary table 5). After excluding variants classified as benign and predicted as intronic, synonymous or not impacting splicing, 12 variants were validated by Sanger sequencing (table 2).
A missense variant in PMS1 (c.224C>T), predicted as pathogenic, deleterious and probably damaging by FATHMM, SIFT and PolyPhen, respectively (table 2, online supplementary table 3), was found in family P1 (table 1, online supplementary table 4). The probands, who developed an MSS IGC at 59 years, had an FDR with GC at 80 and two other FDR and SDR with unidentified cancers at 50 and 75 years, respectively. The only supporting evidence for the role of this variant in FIGC was its COSMIC record as somatic in one GC sample (COSM6198026) (online supplementary table 3).
The proband of family P27 presented three germline variants of uncertain significance, two in SMAD4 (c.424+5G>A; c.454+38G>C) and one in PRSS1 (c.201-99G>C) (online supplementary table 4). Variants c.424+5G>A in SMAD4 and c.201–99G>C in PRSS1 were the only intronic variants predicted to disrupt RNA splicing (table 2, online supplementary tables 3 and 5,). In particular, SMAD4 variant c.424+5G>A decreases the confidence of a donor splice site, which may lead to intron 3 retention, a premature termination codon and generation of a 142 amino acid truncated protein. On the other hand, PRSS1 variant c.201-99G>C creates a new, high-confidence acceptor splice site within intron 2, which may lead to a truncated 69 amino acid protein. Proband P27 developed an MSS IGC at age 64 and had family history of GC, gastric ulcer, laryngotracheal, gynaecological and hepatobiliary cancers (table 1, online supplementary table 4). The presence of these phenotypes seems to exclude juvenile polyposis and hereditary pancreatitis as underlying syndromes of this family, but could support a potential role for SMAD4 together with PRSS1 in FIGC.
We then screened the primary tumours of P1 and P27 FIGC probands for somatic second-hit inactivating mechanisms (LOH, somatic mutation) in germline-affected genes. None of the two FIGC probands showed evidence of deleterious somatic variants nor LOH of the wild-type allele of the germline targeted genes (data not shown).
Although interesting, these findings are insufficient to support the monogenic hypothesis for FIGC and a potentially causal role for the abovementioned affected genes.
Oligogenic/polygenic hypothesis: co-occurrence of rare germline variants determines somatic landscapes of FIGC tumours
We then proceeded with the oligogenic/polygenic hypothesis, which takes into consideration the co-occurrence of germline variants, regardless of their population frequency, as a risk factor for this disease, which would determine the subsequent somatic events necessary for malignant transformation.
We categorised the 50 FIGC probands according to the presence of rare germline variants: families with no variants (n=30); families with a single variant (n=14); and families with multiple variants (n=6). To understand the germline and somatic variant burden for each of these three FIGC classes, we applied the previously described quality criteria obtaining 710 HQ germline variants and 344 HQ somatic variants. The average number of HQ germline variants was identical across the three classes of FIGC families (75.7, 77.4 and 74.5 for families without (0), with one (1) or more than one (>1) rare germline variants, respectively; figure 2B). Germline landscape unsupervised hierarchical clustering revealed no associations between variants or variant-bearing genes and a particular FIGC family class (figure 2C,D).
Concerning the somatic variant burden, no significant differences were observed across the three FIGC classes (15.0, 13.8 and 11.2 for families with 0, 1 or >1 rare germline variants, respectively; figure 3A). Again, no clustering of specific variants/genes and particular FIGC classes was observed (figure 3B,C).
We verified that 38% of the FIGC tumours in our series displayed the MSI phenotype, and further investigated whether MSI could influence the somatic variant burden and landscape in families with 0, 1 or >1 rare germline variants. After subdividing each FIGC class according to its MSI status, no significant differences were observed both in terms of somatic variant burden and landscape between categories (figure 3B–D). Nevertheless, we observed that among FIGC families with multiple rare germline variants (>1), MSI tumours showed an average number of HQ somatic variants twofold higher than that of MSS tumours (17 vs 10 HQ somatic variants per case, respectively; figure 3D, online supplementary figure 1A). This observation prompted us to explore the influence of rare germline variants, independently of their number, on tumour instability and consequent somatic variant burden. Despite the lack of statistical significance, we observed an enrichment of MSI tumours in FIGC families carrying rare germline variants comparing with MSI tumours from families lacking rare germline variants (online supplementary figure 1B). Concerning the average of somatic variants, whereas MSI and MSS tumours from FIGC lacking rare germline variants displayed a similar average number, there was a non-significant trend for higher average number of HQ somatic variants in MSI tumours versus MSS tumours from FIGC families with rare germline variants (≥1; online supplementary figure 1C).
Although our data did not support the hypothesis that co-occurrence of rare germline variants is a major determinant of FIGC-related somatic landscapes, these pinpointed a potential correlation between the coexistence of rare and common germline variants, high average number of somatic variants and MSI phenotype in FIGC.
FIGC is genetically distinct from SIGC and from HDGC-CDH1 mutation-negative
Since the late age of onset in FIGC probands and their relatives makes it hard to distinguish bona fide FIGCs from SIGCs, we compared the age of onset of FIGC probands with the age of onset of a series of SIGC cases. We found that FIGC probands developed GC approximately 10 years earlier than patients with SIGC (p=4.5E-03; figure 4E).
We next explored whether these FIGC and SIGC were also distinct at the germline and/or somatic levels. Principal component analysis revealed that certain genes were differentially associated with FIGCs and SIGCs (figure 4A,B). Specifically, common germline variants in TP53 were present in more than 50% of FIGC probands, while only 11% of SIGC cases presented these germline variants (figure 4A,C). At the somatic level, the frequency of BRCA2, ATM, FOXF1, FHIT, SDHB, MSH6, CTNNA1 and PXN could distinguish FIGC from SIGC tumours, with more than 50% of FIGC displaying common variants in these genes, as compared with very low frequencies in SIGC (figure 4B,C).
By combining all germline and somatic landscapes of 50 FIGCs and 47 SIGCs focusing only on the abovementioned genes, and using unsupervised hierarchical clustering, two main clusters were evidenced separating most FIGCs from SIGCs (figure 4D). Whereas FIGCs carried both germline and somatic variants in TP53, BRCA2, ATM, FOXF1, FHIT, SDHB, MSH6, CTNNA1 and PXN genes, SIGCs lacked TP53 and FHIT germline and somatic variants and mainly presented BRCA2, ATM, FOXF1, SDHB, MSH6, CTNNA1 and PXN somatic variants.
Further supporting that FIGC represents a different entity likely evolving for longer than SIGCs is the fact that FIGC tumours presented statistically significantly more somatic common variants than SIGC tumours (p=4.2E-06), even if arising from patients 10 years younger on average (figure 4E,F).
To further understand whether FIGC is a genetic entity also distinct from HDGC-CDH1 mutation-negative, we compared the germline and somatic landscapes of 7 FIGCs and 17 HDGCs sequenced with the same Next Generation Sequencing (NGS) panel. We verified that indeed FIGC and HDGC also display considerable differences between germline and somatic landscapes (online supplementary figure 2)(); however, the low number of FIGC cases possible to analyse, which was due to sequencing panel differences, hampers more formal conclusions.
Overall, our results suggest that FIGC, rather than a monogenic disease, is likely a polygenic disease with distinctive germline and somatic landscapes from SIGC and HDGC-CDH1-negative.
FIGC presents an autosomal dominant inheritance pattern of IGC, without gastric polyposis, and has been clinically defined by analogy to the Amsterdam criteria for HNPCC.9 However, lack of novel data supporting familial aggregation of IGC at a given age of onset as well as the non-existence of tumour spectrum descriptions have impeded the redefinition of FIGC testing criteria, useful for identification and management of these families.
The primary strength of this study is the use of a large homogeneous cohort of probands with IGC, familial aggregation of GC, detailed personal/family history, age of disease onset and disease spectrum. This series does not present clinical criteria compatible with any other gastrointestinal cancer-associated syndrome, is clearly enriched in GC and mainly of intestinal type, which suggests this is the first data-driven testing criteria for FIGC families. We propose that any family presenting two GC cases, one confirmed of intestinal histology, independently of age, and with or without colorectal cancer, breast cancer or gastric ulcers in other family members, could be considered FIGC.
Besides potential testing criteria, our study also reported the first large-scale sequencing analysis of the germline and somatic landscapes of FIGC and respective comparisons with comparable landscapes of SIGC and HDGC-CDH1 mutation-negative. We used these data to explore the unknown inherited nature of FIGC. Among the FIGC-exclusive germline rare variants found, the missense PMS1 c.224C>T variant was the only one predicted as pathogenic in family P1. Deleterious variants in this DNA mismatch repair protein (PMS1, OMIM:600258) can be found in HNPCC families, either alone or co-occurring with mutations in other HNPCC-related genes.32 33 However, the real contribution of PMS1 germline mutations for HNPCC predisposition is still debatable. Liu et al 33 detected PMS1 and MSH2 germline mutations in an HNPCC proband with an MSI tumour, and observed that only the MSH2 germline mutation was shared with another member of the family affected with colorectal cancer, thus demonstrating that MSH2 is the real predisposing gene to colorectal cancer in this family. Notwithstanding, they postulated that the PMS1 mutation could contribute to the unusual number of lung cancer cases in this HNPCC family.33 Our FIGC proband (P1) carrying a PMS1 germline variant displayed an MSI-low tumour, consistent with the fact that Pms1-deficient mice do not show an increased mutation rate (MSI) in the colonic epithelium.34 Although we lack full evidence for the potentially causative role of this PMS1 variant in family P1, namely a second-hit in the tumour and segregation analysis, this remains an open possibility. The same applied to family P27, where potentially truncating variants are simultaneously found in SMAD4 and PRSS1, but no second somatic-hits are found in these genes. Overall, these findings do not strongly support a monogenic nature for FIGC, at least as evident as that seen for CDH1-associated HDGC or GAPPS.
In the last decade, several studies have integrated large-scale normal and tumour sequencing data to ascertain the impact of germline variation on tumour evolution.35–38 For example, Carter et al 36 identified germline variants that can either dramatically increase the frequency of somatic mutations or influence the site where a tumour develops. Others have shown that rare germline truncations in cancer susceptibility genes, including BRCA1, BRCA2, FANCM and MSH6, are significantly associated with increased somatic mutation frequencies in specific cancer types, suggesting that germline and somatic levels are intrinsically linked.37 Our findings revealed that, independently of the presence of rare germline variants, FIGC families displayed similar germline and somatic variant burden and landscapes, suggesting that this type of inherited variation may not be a major determinant of tumour development in these families. Interestingly, we found that MSI and MSS tumours from FIGC families lacking rare germline variants displayed a similar somatic variant burden, while MSI tumours from families carrying single/multiple germline rare variants tend to harbour more somatic variants than MSS tumour-bearing families. Altogether, these findings suggest that rare germline defects involving the DNA repair system may extend to the somatic level, as previously demonstrated in other cancer types.37 38
Our study, as the previous ones, failed to find the monogenic factor that genetically determined the occurrence of FIGC; however, before excluding the possibility of considering our FIGC series as a sporadic cohort, we explored the average age of onset of probands, number of somatic variants, and their germline and somatic landscapes as compared with other GC entities. This analysis showed that FIGC probands developed GC at least 10 years earlier and carried more TP53 germline common variants than SIGC, that 38% of FIGC tumours were MSI, but also that FIGC tumours displayed significantly more somatic common variants than SIGC tumours, as well as a specific germline and somatic variant profile. In addition, this germline and somatic variant profile was also different from that presented by HDGC cases lacking CDH1 germline causal variants. Therefore, the analysis of the large-scale normal and tumour sequencing data from FIGC, SIGC and HDGC-CDH1 mutation-negative cases was instrumental to define FIGC as a distinct clinical and molecular entity.
Altogether, these data support the idea of a so far unrecognised genetically determined factor(s) that promotes IGC in probands and GC in their close relatives, with an apparent pattern of autosomal inheritance, and that despite late onset it presents earlier than SIGC. Further, FIGC seems to evolve through a different path from SIGC, starting the accumulation of somatic variants earlier and often triggering MSI, as part of their evolution.
Our study displayed some limitations, such as the fact that our custom NGS panels did not account for all possible cancer predisposition genes, hence other genes may contribute to FIGC risk; and the fact that normal-tumour pairs of several FIGC probands were sequenced with different panels. However, the comparison with whole genome sequencing data from SIGC allowed us to overcome most of these problems, highlighting important FIGC-specific features.
In summary, our study is the first to gather evidence that allows us to suggest testing criteria for FIGC families, and to state with some degree of confidence that FIGC is likely a genetically determined polygenic GC-predisposing disease, different at the clinical, germline and somatic levels from SIGC and HDGC.
This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.
JC and PO contributed equally.
DH and CO contributed equally.
Contributors JC: designed the study, collected, analysed and interpreted the data, wrote the initial draft of the manuscript including the text, figures and tables, and submitted the manuscript. PO: designed the study, performed bioinformatics and statistical analyses, interpreted the data, and wrote the initial draft of the manuscript including the text, figures and tables. JS, SH: performed the next-generation sequencing analyses. CSJ, SPT: validated the next-generation sequencing data. MF: participated in the bioinformatics analysis. GC: participated in the collection of patient material, collected and compiled the family history data. HP: helped with the next-generation sequencing analyses. DL: worked on the bioinformatics analyses. VP: participated in the collection of family history data. FR: sought institutional approval, performed the clinical evaluation and recruitment of patients, and collected patient material and family history data. DH: designed the gene panel sequencing, coordinated the next-generation sequencing, obtained funding and critically reviewed the manuscript. CO: had the original idea and designed the study, coordinated the study, interpreted the data, sought institutional approval and funding, critically reviewed the manuscript, and is responsible as guarantor for the overall content.
Funding IPATIMUP integrates the i3S Research Unit, which is partially supported by FCT, the Portuguese Foundation for Science and Technology. This work has been financed by (1) FEDER - Fundo Europeu de Desenvolvimento Regional funds through the COMPETE 2020 - Operacional Programme for Competitiveness and Internationalisation (POCI), Portugal 2020, and by Portuguese funds through FCT/Ministério da Ciência, Tecnologia e Inovação in the framework of the following projects: (1.1) 'Institute for Research and Innovation in Health Sciences' (POCI-01-0145-FEDER-007274) and (1.2) 3DChroMe (POCI-01-0145-FEDER-30164); (2) Norte Portugal Regional Programme (NORTE 2020), under the PORTUGAL 2020 Partnership Agreement, through the European Regional Development Fund (ERDF) for the following projects: (2.1) NORTE-01-0145-FEDER-000029 and (2.2) DOCnet (NORTE-01-0145-FEDER-000003); (3) 'NSFC - No Stomach for Cancer': 'Defining the Contribution of Mutations in CDH1 Non-Coding Regions and Other Known Susceptibility Genes to Hereditary Gastric Cancer'; (4) FCT Fellowships: SFRH/BPD/86543/2012 to JC; SFRH/BPD/89764/2012 to PO; SFRH/BPD/79499/2011 to HP; and SFRH/BD/140796/2018 to CSJ; and (5) POCI and Programas Operacionais Regionais de Lisboa e do Algarve e pela Fundação para a Ciência e a Tecnologia, Project GenomePT - Laboratório Nacional para a Sequenciação e Análise de Genomas (22184).
Competing interests None declared.
Patient consent for publication Not required.
Ethics approval Informed consent was obtained from all patients, and the study was approved by the local institutional review boards.
Provenance and peer review Not commissioned; externally peer reviewed.
Data availability statement All data relevant to the study are included in the article or uploaded as supplementary information.