Germline Mutations in PALB2, BRCA1, and RAD51C, Which Regulate DNA Recombination Repair, in Patients With Gastric Cancer
2016; Elsevier BV; Volume: 152; Issue: 5 Linguagem: Inglês
10.1053/j.gastro.2016.12.010
ISSN1528-0012
AutoresRuta Sahasrabudhe, Paul C. Lott, Mábel Bohórquez, Ted Toal, Ana Estrada, John Suarez, Alejandro Brea‐Fernández, José Cameselle‐Teijeiro, Carla Pinto, Irma Ramos, Alejandra Mantilla, Rodrigo Prieto, Alejandro H. Corvalán, Enrique Norero, Carolina Álvarez, Teresa Tapia, Pilar Carvallo, Luz María González, Alicia Cock‐Rada, Ángela R. Solano, Florencia Neffa, Adriana Della Valle, Christopher Yau, Gabriela Soares, Alexander D. Borowsky, Nan Hu, Li-Ji He, Xiao-You Han, Philip R. Taylor, Alisa M. Goldstein, Javier Torres, Magdalena Echeverry, Clara Ruiz‐Ponte, Manuel R. Teixeira, Luis G. Carvajal‐Carmona, Magdalena Echeverry, Mábel Bohórquez, Rodrigo Prieto, John Suarez, Gilbert Mateus, María Mercedes Bravo, Fernando Bolaños, Alejandro Velez-Zea, Alejandro H. Corvalán, Pilar Carvallo, Javier Torres, Luis G. Carvajal‐Carmona,
Tópico(s)Helicobacter pylori-related gastroenterology studies
ResumoUp to 10% of cases of gastric cancer are familial, but so far, only mutations in CDH1 have been associated with gastric cancer risk. To identify genetic variants that affect risk for gastric cancer, we collected blood samples from 28 patients with hereditary diffuse gastric cancer (HDGC) not associated with mutations in CDH1 and performed whole-exome sequence analysis. We then analyzed sequences of candidate genes in 333 independent HDGC and non-HDGC cases. We identified 11 cases with mutations in PALB2, BRCA1, or RAD51C genes, which regulate homologous DNA recombination. We found these mutations in 2 of 31 patients with HDGC (6.5%) and 9 of 331 patients with sporadic gastric cancer (2.8%). Most of these mutations had been previously associated with other types of tumors and partially co-segregated with gastric cancer in our study. Tumors that developed in patients with these mutations had a mutation signature associated with somatic homologous recombination deficiency. Our findings indicate that defects in homologous recombination increase risk for gastric cancer. Up to 10% of cases of gastric cancer are familial, but so far, only mutations in CDH1 have been associated with gastric cancer risk. To identify genetic variants that affect risk for gastric cancer, we collected blood samples from 28 patients with hereditary diffuse gastric cancer (HDGC) not associated with mutations in CDH1 and performed whole-exome sequence analysis. We then analyzed sequences of candidate genes in 333 independent HDGC and non-HDGC cases. We identified 11 cases with mutations in PALB2, BRCA1, or RAD51C genes, which regulate homologous DNA recombination. We found these mutations in 2 of 31 patients with HDGC (6.5%) and 9 of 331 patients with sporadic gastric cancer (2.8%). Most of these mutations had been previously associated with other types of tumors and partially co-segregated with gastric cancer in our study. Tumors that developed in patients with these mutations had a mutation signature associated with somatic homologous recombination deficiency. Our findings indicate that defects in homologous recombination increase risk for gastric cancer. See editorial on page 926. See editorial on page 926. Worldwide, gastric cancer (GC) is the fifth most commonly diagnosed malignancy and the third cause of cancer-related deaths.1Ferlay et al.J. Int J Cancer. 2015; 136: E359-E386Crossref PubMed Scopus (22728) Google Scholar Up to 10% of cases show familial clustering, suggesting a genetic basis.2La Vecchia C. et al.Cancer. 1992; 70: 50-55Crossref PubMed Scopus (304) Google Scholar CDH1 mutations are a known cause of hereditary diffuse gastric cancer (HDGC), explaining approximately 40% of cases,3Guilford P. et al.Nature. 1998; 392: 402-405Crossref PubMed Scopus (1430) Google Scholar, 4Hansford S. et al.JAMA Oncol. 2015; 1: 23-32Crossref PubMed Scopus (491) Google Scholar but the genetics of non-HDGC remain largely unknown. To identify novel GC genes, we analyzed CDH1 mutation-negative HDGC cases using whole-exome sequencing (WES) followed by candidate gene targeted analyses in independent HDGC and non-HDGC cases. WES of 28 CDH1-negative European HDGC cases identified three with candidate causal variants (Table 1): nonsense (p.Arg414Ter) and splice site (c.3201+1G>T) PALB2 mutations, and a nonsense RAD51C (p.Arg237Ter) mutation. No deleterious mutations were seen in other known cancer genes (Supplementary Methods). PALB2 and RAD51C are both critical in homologous recombination (HR), a major DNA repair pathway.5Prakash R. et al.Cold Spring Harb Perspect Biol. 2015; 7: a016600Crossref PubMed Scopus (539) Google Scholar Both of the PALB2 mutations have been reported previously as pathogenic in breast cancer families6Antoniou A.C. et al.N Engl J Med. 2014; 371: 1651-1652Crossref PubMed Scopus (628) Google Scholar and RAD51C p.Arg237Ter is reported as pathogenic in ClinVar.7Landrum M.J. et al.Nucleic Acids Res. 2016; 44: D862-D868Crossref PubMed Scopus (1747) Google ScholarTable 1Details of Clinical Information of the Mutation CarriersMutation detailsIDAge of onsetSexHistologySatisfied HDGC criteria?Helicobacter pylori infectionHistory of smokingPALB2c.1240C>T, p.Arg414TerCG-12aIdentified by WES.,dLOH and mutational signature analyzed.69MIntestinalNoNANACG-008cIdentified by genotyping.48FDiffuseNANAYesGM03758946FNANoNegativeNoPALB2c.3201+1G>TCG-05aIdentified by WES.50MDiffuseYesNegativeNoPALB2c.1882_1890delAAGTCCTGC, p.Lys628_Cys630delCG-039bIdentified by targeted sequencing.47FDiffuseNANegativeNoCG-028cIdentified by genotyping.,dLOH and mutational signature analyzed.81MIntestinalNANegativeYesPALB2c.2753C>A, p.Pro918Gln3CG-103bIdentified by targeted sequencing.,dLOH and mutational signature analyzed.79FMixedNoNegativeYesBRCA1c.3331_3334delCAAG, p.Gln1111AsnfsCG-036bIdentified by targeted sequencing.67FDiffuseNoNANoCG-059bIdentified by targeted sequencing.54MDiffuseNoNANoBRCA1c.1674delA, p.Gly559ValfsCG-001cIdentified by genotyping.65MNANoPositiveYesRAD51Cc.709 C>T, p.Arg237TerGM022584aIdentified by WES.,dLOH and mutational signature analyzed.73MDiffuseYesNegativeNoNA, Not available.a Identified by WES.b Identified by targeted sequencing.c Identified by genotyping.d LOH and mutational signature analyzed. Open table in a new tab NA, Not available. We then performed targeted sequencing of PALB2 and RAD51C, their interaction partners BRCA1/2 and CDH1 in 173 additional Latin-American GC cases. Based on enrichment of HR mutations in our discovery cohort and a recent report showing multiple intestinal, diffuse, and mixed histology gastric tumors with a somatic HR deficiency signature,8Alexandrov L.B. et al.Nat Commun. 2015; 6: 8683Crossref PubMed Scopus (101) Google Scholar our validation cohort included both HDGC and non-HDGC cases of diffuse and nondiffuse histology (Supplementary Methods). Targeted sequencing identified 4 additional mutation carriers: 2 sharing a known Hispanic BRCA1 founder mutation (p.Gln1111Asnfs)9Torres D. et al.Breast Cancer Res Treat. 2007; 103: 225-232Crossref PubMed Scopus (79) Google Scholar and 2 with novel PALB2 mutations (p.Pro918Gln and p.Lys628_Cys630del) with predicted deleterious effects. Residue Pro918 falls in the PALB2 WD40 domain, which mediates interactions with BRCA2, RAD51, and RAD51C, whereas Lys628-Cys630 resides in the binding domain of MRG15, a transcription regulator and whose PALB2 interaction is required for homology-directed DNA double-strand break repair indicating potential pathogenicity of these 2 novel mutations.10Park J.Y. et al.Oncogene. 2014; 33: 4803-4812Crossref PubMed Scopus (101) Google Scholar, 11Sy S.M. et al.J Biol Chem. 2009; 284: 21127-21131Abstract Full Text Full Text PDF PubMed Scopus (56) Google Scholar In a third phase of the study, we genotyped all 6 PALB2, RAD51C, and BRCA1 mutations described plus 4 known Hispanic BRCA1/2 founder mutations (Supplementary Methods) in 160 independent Latin-American non-HDGC cases and found 3 additional mutation carriers, 1 with a BRCA1 mutation (p.Gly559Valfs) and 2 with PALB2 mutations (p.Lys628_Cys630del and p.Arg414Ter) (Table 1). Interestingly, during the preparation of this article, our clinic-based Portuguese collaborator (MRT and GS) identified 1 additional GC case (GM037589) with PALB2 p.Arg414Ter. None of the 7 PALB2, RAD51C, and BRCA1 mutations detected in 11 unrelated Caucasian and Latin-American cases was detected in 1,170 population-matched controls (see mutation details in Supplementary Table 1). Clinical details of our mutation carriers are presented in Table 1. Most of them had diffuse histology, 2 had HDGC syndrome (CG-05 and GM022584), and 1 reported a history of hereditary breast and ovarian cancer (case CG-36, not shown). These mutation carriers were predominantly nonsmokers and/or negative for Helicobacter pylori infection (Table 1), which suggest that GC risk in most of these cases was not driven by these 2 known environmental risk factors.12Cover T.L. et al.Gut Microbes. 2013; 4: 482-493Crossref PubMed Scopus (86) Google Scholar To obtain additional evidence of the causality of our HR gene mutations, we carried out loss of heterozygosity, mutational signature, and co-segregation analyses in available samples from tumors and relatives. For loss of heterozygosity and mutational signatures, we performed WES in 4 available tumor samples from 3 PALB2 (CG-12/p.Arg414Ter, CG-028/p.Lys628_Cys630del and 3CG-103/p.Pro918Gln) and RAD51C mutation carriers (Table 1). We found no loss of heterozygosity or compound heterozygosity in these tumor samples (not shown). Interestingly, when we analyzed the somatic WES data for mutational signatures, we found that all 4 tumors were enriched for a signature indicative of HR defects,13Alexandrov L.B. et al.Nature. 2013; 500: 415-421Crossref PubMed Scopus (5416) Google Scholar, 14Shiraishi Y. et al.PLoS Genet. 2015; 11: e1005657Crossref PubMed Scopus (70) Google Scholar providing evidence for the causality of these mutations (Supplementary Methods, Supplementary Figures 1 and 2). Figure 1 shows available pedigrees from mutation carriers. Case 3CG-103 and her daughter were both diagnosed with GC and carried the PALB2 p.Pro918Gln mutation (Figure 1A). GM037589, a PALB2 p.Arg414Ter carrier, developed GC and breast cancer and had a sister diagnosed with ovarian and endometrial cancer who also carried PALB2 p.Arg414Ter (Figure 1B). The RAD51C p.Arg237Ter carrier's son died of colon cancer but did not carry the mutation (Figure 1C). We found that GC was the predominantly diagnosed malignancy among unavailable relatives of these carriers (Figures 1A−D). Although we did not have access to samples from relatives of the PALB2 p.Lys628_Cys630del carriers, our local collaborators found this mutation co-segregating in an unrelated breast cancer family (unpublished data). Albeit limited, our co-segregation data partially support GC causality of PALB2 mutations. The RAD51C co-segregation data are, however, inconclusive, but the presence of a strong HR signature in the gastric tumor of this mutation carrier warrants further studies on RAD51C as a candidate GC gene. In summary, our study identified 11 cases with mutations in PALB2, BRCA1, and RAD51C, 3 closely related HR genes. Some of these mutations are known to be pathogenic in other cancer types. Out of 362 cases analyzed, 6.45% of the HDGC cases (2 of 31) and 2.7% (9 of 331) of non-HDGC cases had PALB2, BRCA1, or RAD51C mutations, suggesting that HR genes play a role in GC risk. Our data also provide evidence of a germline basis for the recently reported HR mutational signature in gastric tumors and strengthens the evidence for a causal role of these genes, specifically PALB2, in GC, as observed previously.4Hansford S. et al.JAMA Oncol. 2015; 1: 23-32Crossref PubMed Scopus (491) Google Scholar, 15Lu C. Xie M. Wendl M.C. et al.Nat Commun. 2015; 6: 10086Crossref PubMed Scopus (205) Google Scholar Future larger studies are needed to definitively assign causality and understand the penetrance and prevalence of HR gene mutations in GC and to further understand if and why some individuals from hereditary breast and ovarian cancer families with HR gene mutations develop GC. Further characterizations of the GC histology in HR gene mutation carriers are also needed, as we found instances where the same mutation was found in cases with different histologies (CG-12 and CG-008 with PALB2 p.Arg414ter and CG-039 and CG-028 with PALB2 p.Lys628_Cys630del; Table 1). CDH1 mutation−negative families might benefit from HR gene testing and increased endoscopic surveillance and targeted therapies, such as poly ADP ribose polymerase inhibitors.8Alexandrov L.B. et al.Nat Commun. 2015; 6: 8683Crossref PubMed Scopus (101) Google Scholar The authors thank all of the patients and their families for participating in the study and the multiple clinicians and research nurses who helped with patient recruitment. The authors are thankful to No Stomach for Cancer Inc and the University of California Davis for providing principal financial support to the study. The authors are grateful to former Carvajal-Carmona laboratory members John Williamson, Julian Halmai, Anna Marie Tuazon, and Cathy Wang for their support during the study, and to Drs Yuichi Shiraishi and Matthew Stephens for their support and stimulating discussions about mutational signature analyses. Latin American Gastric Cancer Genetics Collaborative Group: Magdalena Echeverry and Mabel Bohorquez (Universidad del Tolima, Ibagué, Colombia), Gilbert Mateus (Hospital Federico Lleras Acosta, Ibagué, Colombia), Maria Mercedes Bravo (Instituto Nacional de Cancerología, Bogotá, Colombia), Fernando Bolaños (Hospital Hernando Moncaleano Perdomo, Neiva, Colombia), Alejandro Vélez (Hospital Pablo Tobón Uribe, Medellín, Colombia), Alejandro Corvalan (Facultad de Medicina, Pontificia Universidad Católica de Chile, Santiago, Chile), Pilar Carvallo (Departamento de Biología Celular y Molecular, Facultad de Ciencias Biológicas, Pontificia Universidad Católica de Chile, Santiago, Chile), Javier Torres (Unidad de Investigación en Enfermedades Infecciosas, UMAE Pediatria, IMSS, México City, México), and Luis Carvajal-Carmona (University of California, Davis, CA). Author contributions: Study concept and design: RS, MRT, LGC-C. Acquisition of data: RS, PL, MB, APE, JJS, AB-F, CP, IR, AM, RP, JS, AC, EN, CA, TT, PC, LMG, AC-R, AS, FN, ADV, CY, GS, AB, NH, L-JH, X-YH, PRT, AMG, JT, ME, CR-P, MRT, LGC-C. Analysis and interpretation of data: PL, RS, TT, LGC-C. Drafting of the manuscript: RS, LGC-C. Critical revision of the manuscript for important intellectual content: RS, PL, TT, AG, JT, ME, JC-T, AB, CR-P, MRT, LGC-C. Statistical analysis: RS, PL, TT, LGC-C. Obtained funding: LGC-C. Administrative, technical, or material support: RS, PL, MRT, JT, CR-P, ME, LGC-C. Study supervision: RS, PL, MRT, JT, CR-P, ME, and LGC-C. Ruta Sahasrabudhe and Paul Lott contributed equally to this work. For WES analysis, we included 28 GC cases (and 6 relatives from 4 different families) with HDGC (defined according to the published guidelines1van der Post R.S. et al.J Med Genet. 2015; 52: 361-374Crossref PubMed Scopus (427) Google Scholar) recruited in the Portuguese Oncology Institute (University of Porto, Portugal) and in the Genomic Medicine group (Santiago de Compostela, Spain). Sample collection was undertaken with informed consent and ethical review board approval of the corresponding institution, in accordance with the tenets of the Declaration of Helsinki. All of these 28 index HDGC cases tested negative for CDH1 mutations at clinical laboratories in these 2 Portuguese and Spanish institutions. Mean age of HDGC index patients was 48.2 years (SD, 13.2 years). Fifteen of these patients were male and 13 were female. Interestingly, 1 of these patients (CG-12), who was initially included as an HDGC case, on histologic re-examination by 2 independent surgical pathologists (JC-T and AB) was reclassified as having intestinal histology. This case was therefore reclassified as a non-HDGC in our study. Samples were prepared for WES using Agilent SureSelect XT2 protocol. Briefly, up to 1 μg DNA was sheared using Covaris E220 sonicator. Fragments were end-repaired, A-tailed and Illumina-compatible adaptors were ligated at the ends. The fragments were then enriched using PCR. Eight multiplexed samples were hybridized to the bait set, washed, and captured fragments were amplified by PCR. Samples were then sequenced on an Illumina HiSeq2000 sequencer with 100PE sequencing. For data analysis, publically available tools as well as custom shell scripts were used. Raw data was trimmed for adaptors and sequence quality and then aligned to the human reference genome GRCh37 with decoy sequences using BWA-mem, version 0.7.12 (Scythe, version 0.994 Beta, 2011, https://github.com/vsbuffalo/scythe; Sickle, version 1.33, 2011, https://github.com/najoshi/sickle).2Li H. https://pdfs.semanticscholar.org/0ee3/a1f7a363b16ceda8f1053a8172f051fd8d4c.pdf. Accessed Accessed February 15, 2017.Google Scholar For WES, duplicates were removed with Picard, version 1.129 (http://picard.sourceforge.net). BAM files were locally realigned using GATK IndelRealigner v3.3 and recalibration of the quality scores was performed using GATK BaseRecalibrator, version 3.3.3McKenna A. et al.Genome Res. 2010; 20: 1297-1303Crossref PubMed Scopus (16856) Google Scholar Multiple callers were used to call variants: GATK HaplotypeCaller non-joint, version 3.34DePristo M.A. et al.Nat Genet. 2011; 43: 491-498Crossref PubMed Scopus (7736) Google Scholar; Freebayes, version 0.9.14–175Garrison EM, et al. Available at: https://arxiv.org/pdf/1207.3907.pdf. Accessed February 15, 2017.Google Scholar; SNVER6Wei Z. et al.Nucleic Acids Res. 2011; 39: e132Crossref PubMed Scopus (192) Google Scholar; Varscan, version 2.3.77Koboldt D.C. et al.Genome Res. 2012; 22: 568-576Crossref PubMed Scopus (3289) Google Scholar; and Samtools mpileup, version 1.2.8Li H. Handsaker B. et al.Bioinformatics. 2009; 25: 2078-2079Crossref PubMed Scopus (36209) Google Scholar Calls were filtered based on: coverage ≥10, number of reads supporting variant ≥5, minimum variant frequency ≥0.20, minimum frequency of variant reads present on opposite strand >0.10, and minimum average read quality ≥22. Variants were annotated using Annovar.9Wang K. et al.Nucleic Acids Res. 2010; 38: e164Crossref PubMed Scopus (9044) Google Scholar In addition, single-nucleotide polymorphisms and INDEL calling was performed using GATK HaplotypeCaller joint genotyping. Calling, variant filtering, and variant score recalibration were performed using GATK, version 3.3 Best Practices.4DePristo M.A. et al.Nat Genet. 2011; 43: 491-498Crossref PubMed Scopus (7736) Google Scholar, 10Van der Auwera G.A. et al.Curr Protoc Bioinform. 2013; 43 (10.1–33): 11PubMed Google Scholar Variants called by at least 2 different callers were considered for further analysis. To select the most informative single-nucleotide variants (SNVs), filtering of the initial data was performed to exclude all synonymous SNVs, SNVs that map to pseudo-genes, repeated regions, segmental duplications, and "dispensable" genes. The remaining protein sequence-altering variants were subjected to frequency filtering using data from publicly available data sets, such as the Exome Variant Server, the UK10K study, dbSNP, and the 1000 Genomes Project to exclude variants with >1% minor allele frequency. Of the remaining 7781 variants, SNVs in known cancer predisposition genes11Rahman N. et al.Nature. 2014; 505: 302-308Crossref PubMed Scopus (408) Google Scholar were identified (n = 45). Of those, 2 SNVs were protein-truncating (PALB2: p.Arg414Ter and RAD51C: p.Arg237Ter) with predicted deleterious amino acid substitutions (based on Polyphen, SIFT, MutationAssessor, and MutationTaster) and 1 variant resulted in disruption of a splice site. For the 3 candidate causal variants, pileups were visually inspected in the Integrative Genomics Viewer.12Robinson J.T. et al.Nat Biotechnol. 2011; 29: 24-26Crossref PubMed Scopus (8760) Google Scholar No truncating, deleterious mutations were seen in any other cancer genes. For WES replication by targeted sequencing, we included 14 Chilean GC cases recruited in a local cancer clinic, 4 of which satisfied HDGC criteria. Our study included a total of 31 HDGC index cases in the discovery (n = 27) and validation (n =4) phases. Of the remaining 10 Chilean non-HDGC cases, 5 had intestinal GC and 5 were of unknown histology. For targeted sequencing, we also included additional GC cases from Colombia (n = 90) and Mexico (n = 69), of which 104 cases had diffuse histology, 42 had mixed histology, 1 had intestinal histology, and in 12 cases histology was unknown. Together, 53 cases had early-onset GC (younger than 50 years). Chilean cases were recruited in Dr Sótero del Río Hospital, and Clinical Hospital Pontificia Universidad Cátólica (both in Santiago, Chile). The Ethics Committees of Dr Sótero del Río Hospital and Clinical Hospital Pontificia Universidad Cátólica de Chile approved the recruitment protocols. Colombian cases for validation phases 2 and 3 were recruited from a multicenter study in Colombia and in the Instituto Mexicano de Seguro Social following protocols approved by University of Tolima (Ibague, Colombia) and Instituto Mexicano de Seguro Social National Council for Research on Health (Mexico City, Mexico). Approximately 350-bp PCR amplicons covering the entire coding regions of BRCA1, BRCA2, CDH1, PALB2, and RAD51C were amplified from 50 ng genomic DNA using Fluidigm Access array system and libraries were sequenced on a MiSeq platform with 250PE reads. Sequence data analysis was performed with a bioinformatics pipeline similar to the one described for WES. Patient recruitment and genotyping: For genotyping, we included 160 non-HDGC cases from Colombia (n = 93) and Mexico (n = 67) that included 24 cases with diffuse histology, 117 with intestinal histology, 8 with mixed histology, and 11 with unknown of histology. All 6 sequence-identified PALB2, RAD51C, and BRCA1 mutations in phase 1 and 2 (see Phase 1 and Phase 2 description in supplementary methods and main text), as well as 4 additional known Hispanic BRCA1/2 founder mutations (c.5123C>T /p.Ala1708Val and c.1674delA/p.Gly559Valfs in BRCA1 and c.2808_2811delTAAA/p.Ala938Profs and c.4889C>G/p.Ser1630Ter in BRCA2) were included in phase 3 of genotyping. Genotyping of these 10 mutations was performed using competitive allele-specific PCR using KASP assays (LGC Genomics, Beverly, MA), following manufacturer's guidelines. All mutations identified using WES, targeted sequencing, and genotyping in phases 1, 2, and 3 were verified using Sanger sequencing. Details of the sequencing primers are as follows: PALB_p.Arg414Ter - Forward: TGAACTTGGTTGTCCTGTGC, Reverse: TGACACTCTTGATGGCAGGA. PALB2_c.3201+1G, Forward: TTTGCCCTCAGGTCCTACAG, Reverse: TGGTTTGTTGGAAGAATGTGA, PALB2_p.Lys628_Cys630del, Forward: CCTCCATTTCTGTATCCATGC, Reverse: AAGAGGATTCCCTTTCTTGGA, PALB2_p.Pro918Gln – Forward : CCAGCTGACAGAGACAAAGATG, Reverse: TCTGAGCCTTCAAATGATGAAA, BRCA1_p.Gln1111Asnf – Forward: GGGTGAAAGGGCTAGGACTC, Reverse: CAGAGGGCCAAAATTGAATG, BRCA1_p.Gly559Valfs – Forward: ACCAAACGGAGCAGAATGGT, Reverse: GCAATTCAGTACAATTAGGTGGGC, RAD51C_p.Arg237Ter - Forward: GGTCCCTGCTCTCTTGGAGA, Reverse: ACCAACCAAACGTAACTTTACTCAA. DNA was extracted, using a Qiagen tissue kit, from formalin-fixed paraffin-embedded tumor tissue samples from 4 cases: CG-12 (PALB2 nonsense mutation carrier), 3CG-103 (PALB2 missense mutation carrier), CG-028 (PALB2 in-frame deletion carrier), and GM022584 (RAD51C nonsense mutation carrier). WES was performed using KAPA and Agilent SureSelect XT kits following manufacturer's guidelines. Samples were sequenced on a HiSeq4000 using PE150 sequencing. Sequence data analysis was performed using GATK best practices as described, and somatic variants were called with GATK MuTect2.13Cibulskis K. et al.Nat Biotechnol. 2013; 31: 213-219Crossref PubMed Scopus (3219) Google Scholar Mutational signature analysis in somatic tissue is a recent field that is undergoing active development, improvement, and statistical grounding. The first general signature model for mutation signature analysis was developed by Alexandrov et al14Alexandrov L.B. et al.Cell Rep. 2013; 3: 246-259Abstract Full Text Full Text PDF PubMed Scopus (822) Google Scholar and was used to analyze The Cancer Genome Atlas dataset, leading to the first defined mutational signature resulting from defects in homologous recombinational DNA repair (HR), annotated as "Signature 3."15Alexandrov L.B. et al.Nature. 2013; 500: 415-421Crossref PubMed Scopus (6909) Google Scholar A conceptually different theoretical model of mutation signatures was developed by Shiraishi et al16Shiraishi Y. et al.PLoS Genet. 2015; 11: e1005657Crossref PubMed Scopus (84) Google Scholar with an accompanying computational framework called pmsignature. This model pools all mutations from all the samples and seeks signatures that occur relatively frequently in the mutation pool. The output from the analysis is a matrix of estimated signature parameters defining the signatures, and a membership weight matrix that estimates the relative contribution of each signature to the mutations in each sample. The number of signatures that is found, K, is a parameter that must be specified a priori. The Shiraishi signature model differs from the earlier model in that it assumes independence of the adjacent bases, so the number of parameters with a single surrounding base is far fewer than with the Alexandrov model, leading to more statistically stable parameter estimates. We combined the mutations of our 4 tumor samples with 40 TCGA GC whole exomes to increase the power to detect common GC signatures and to provide positive and negative HR signature controls. Of the 40 samples, 20 were selected from the 27 samples with non-zero value for "Signature 3" and 20 were selected from the remaining samples with a zero value.17Alexandrov L.B. et al.Nat Commun. 2015; 6: 8683Crossref PubMed Scopus (133) Google Scholar We configured the Shiraishi framework to use 5 bases of total context (the mutated base plus 2 bases upstream and 2 bases downstream) and to include the transcription strand as a mutation feature. The mutation signature analysis was done using the R language (http://www.R-project.org/). In order to detect an HR signature, we first determined which of the 27 Shiraishi signatures was most similar to the Alexandrov "signature 3" by using both Frobenius and cosine similarity measures. Heatmaps depicting the Frobenius and cosine similarity of each of the 27 Shiraishi cancer signatures to each of the 30 Alexandrov (COSMIC) cancer signatures are shown in Supplementary Figure 1A and B, respectively. For Frobenius similarity, Shiraishi signatures 16, 23, 24, and 25 all have similarity ≥0.7 to COSMIC signature 3. For cosine similarity, Shiraishi signatures 16, 23, and 25 all have similarity >0.7 to COSMIC signature 3. We have designated Shiraishi signatures 16 and 23−25 as HR signatures on heatmaps that show Shiraishi signatures. Knowing which Shiraishi signatures correspond to an HR signature, we proceeded to determine which signature, if any, of K signatures produced by our analysis, are similar to one of those Shiraishi HR signatures. We used Frobenius similarity in that case, because both signatures being compared are Shiraishi signatures, and the comparison is more reliable than the Alexandrov-Shiraishi comparison. Frobenius similarity showed that, at K = 3, signature #1 [noted as 1(HR)] was most similar to the Shiraishi HR signatures 16, 23, and 25 (full analysis, Supplementary Figure 1C). Tumor DNA from our study samples was derived from formalin-fixed paraffin-embedded tissue, and was expected to have a higher percentage of C: G>T: A mutations. Therefore, we analyzed mutational signatures after removing C:G>T:A from our study samples as well as from control samples (restricted analysis). Similar to the full analysis, we first identified signatures with high Frobenius similarity to Shiraishi HR signatures, using K = 3 (Supplementary Figure 1D). After optimizing the method, we proceeded to determine whether an HR signature was demonstrated by the 4 study samples where somatic WES data was available. As shown in Supplementary Figure 2, our study samples as well as the TCGA positive controls, at K = 3, in full and restricted analysis have a significantly higher relative contribution or membership weight for the HR signature compared to the negative controls. Interestingly, another hallmark of somatic HR deficiency is a high frequency of large indels.14Alexandrov L.B. et al.Cell Rep. 2013; 3: 246-259Abstract Full Text Full Text PDF PubMed Scopus (822) Google Scholar, 17Alexandrov L.B. et al.Nat Commun. 2015; 6: 8683Crossref PubMed Scopus (133) Google Scholar Consistently, similar to The Cancer Genome Atlas HR-positive controls, the mean deletion length found in the tumors from our 4 PALB2/RAD51C mutation carriers was higher than in TCGA non-HR GC cases (31.6 bp vs 15.4 bp; P = 3 × 10−7). Author names in bold designate shared co-first authorship.Supplementary Figure 2Analysis of mutational signatures in tumor samples. We used WES data from 4 PALB2 and RAD51C mutation carriers (GM022584, 3CG-103, CG-028, and CG-12) and from 40 HR-defective (TCGA_GC_HR, n = 20) and HR-proficient (TCGA_GC_non-HR, n = 20) cases from The Cancer Genome Atlas (TCGA) study. These analyses included all mutations (full analyses, A−C, left panel) and removal of C:G>T:A changes (restricted analyses, D−E, left panel) as our WES data was generated from archival tumors, which are known to accumulate artifactual C:G>T:A mutations. (A) and (D) Logos of somatic HR signatures. The central base represents the frequency of the mutation, which is surrounded by the frequency of bases at positions −2 and −1 (left) and +1 and +2 (right). The top right bars indicate the frequency of such mutations in the + and − transcription strand polarities (see Robinson et al12Robinson J.T. et al.Nat Biotechnol. 2011; 29: 24-26Crossref PubMed Scopus (8760) Google Scholar for more details). (B) and (E) Heatmaps of relative contribution or membership weights of each signature within each sample. Dark shading indicates low contribution of the mutation signature and light shading represents high contribution of the mutation signature. Our 4 samples had highest membership weight to signature #1 (the HR signature) and clustered in the full (which included all mutations, panel B, right) and restricted (which excluded C:G>T:A changes, panel E, left) analyses with the TCGA HR-positive cases. The pattern involving signatures #2 (unknown cases but very similar to a previously reported signature by Shiraishi et al16Shiraishi Y. et al.PLoS Genet. 2015; 11: e1005657Crossref PubMed Scopus (84) Google Scholar in gastric and colorectal tumors) and # 3 (cytosine deamination) showed stronger membership weights with the non-HR samples. The PALB2 nonsense mutation carrier and 5 TCGA_GC_non-HR samples were removed from the restricted analysis as they had few mutations after removal of C:G>T:A changes. (C) and (F) Tables indicating membership weights for each sample. Table indicates the estimated fraction of mutations associated with the HR signature pattern. Study sample mean indicates mean membership weight of HR signature. P value from Mann−Whitney 2-sample U test compares membership weight of the study sample mean or TCGA_GC_HR sample mean to TCGA_GC_non-HR sample mean (row 6 and 8 and row 7 and 8), respectively.View Large Image Figure ViewerDownload Hi-res image Download (PPT)Supplementary Table 1Details of Mutations Identified in the StudyChr position (genome assembly = GRCh37/hg19)RefAltGene nameTranscript IDComplementary DNA changeProtein change and effectPathogenicity predictionType, effect on proteinExAC frequency16: 23646627GAPALB2NM_024675.3c.1240C>Tp.Arg414TerReported Pathogenic in ClinVarNonsense, truncates proteinNA16: 23625324CAPALB2NM_024675.3c.3201+1G>TReported Pathogenic in ClinVarSplice-donor variantNA16: 23641585-23641593GCAGGACTT—PALB2NM_024675.3c.1882_1890delAAGTCCTGCp.Lys628_Cys630delReported as VUS in ClinVar,In-frame deletion, possible effect on recruitment to DNA damage site (see text)3.31 × 10−516: 23635411GTPALB2NM_024675.3c. 2753C>Ap.Pro918GlnReported as VUS in ClinVar, predicted deleterious in SIFT, PolyPhen, LRT and MutationTasterMissense, possible effect on protein−protein interaction1.742 × 10−517: 41244214-41244217CAAG—BRCA1NM_007294.3c. 3331_3334delCAAGp.Gln1111AsnfsPathogenicFrameshift deletion, truncates proteinNA17: 41245874A—BRCA1NM_007294.3c.1674delAp.Gly559ValfsReported Pathogenic in ClinVarFrameshift deletion, truncates proteinNA17: 56787223CTRAD51CNM_058216.2c.709C>Tp.Arg237TerReported Pathogenic in ClinVarNonsense, truncates protein8.23 × 10−6ExAC, exome aggregation consortium; NA, not available; VUS, variant of uncertain significance. Open table in a new tab ExAC, exome aggregation consortium; NA, not available; VUS, variant of uncertain significance. Gaining Ground in the Genetics of Gastric CancerGastroenterologyVol. 152Issue 5PreviewGastric cancer is the third most common malignancy worldwide.1 Outcomes are poor with 5-year survival estimates of 30% and 5% for all stages and for metastatic disease, respectively.2 Environmental factors, such as Helicobacter pylori infection, diet, and tobacco use, are strong risk factors that likely underlie many gastric cancer cases.3 In 5% to 10% of cases, there is familial clustering of gastric cancers, and, although shared environmental risk factors could explain some of this clustering, hereditary syndromes owing to genetic mutations are known to increase gastric cancer risk. Full-Text PDF
Referência(s)