Germline predisposition in multiple myeloma

Abstract

We present a study of rare germline predisposition variants in 954 unrelated individuals with multiple myeloma (MM) and 82 MM families. Using a candidate gene approach, we identified such variants across all age groups in 9.1% of sporadic and 18% of familial cases. Implicated genes included genes suggested in other MM risk studies as potential risk genes (DIS3, EP300, KDM1A, and USP45); genes involved in predisposition to other cancers (ATM, BRCA1/2, CHEK2, PMS2, POT1, PRF1, and TP53); and BRIP1, EP300, and FANCM in individuals of African ancestry. Variants were characterized using loss of heterozygosity (LOH), biallelic events, and gene expression analyses, revealing 31 variants in 3.25% of sporadic cases for which pathogenicity was supported by multiple lines of evidence. Our results suggest that the disruption of DNA damage repair pathways may play a role in MM susceptibility. These results will inform improved surveillance in high-risk groups and potential therapeutic strategies.

Keywords: Cancer; Genetics; Molecular biology.

Graphical abstract

Figure 1

Discovery of germline predisposition variants in 954 MM cases from the MMRF cohort and 82 families (A) Recruitment of research subjects and sample collection details for both the Multiple Myeloma Research Foundation (MMRF) and University of Chicago familial datasets. (B) Characteristics of the 954 MMRF and 99 familial samples, including sample size, female ratio, age distribution, and genetic ancestry for each cohort. Age is depicted as the average age +/− one standard deviation (MEAN +/− SD). Female ratio is the percentage (%) of female individuals in each cohort. Genetic ancestry is estimated from WES data by training a random forest classifier on variants detected in each cohort overlapping with the 1000 Genomes dataset, classifying samples into African (AFR); Ad Mixed American (AMR); East Asian (EAS); European (EUR); or South Asian (SAS). Accuracy on the test set was >99% for both datasets (see Figure S1). (C) Summary of germline variant calling and CharGer results for both datasets, showing the percentage of affected cases by ancestry group. Variants passing manual review are used in downstream analyses.

Figure 2

Distribution of rare germline predisposition variants across genes (A) Sum of unique P, LP, and PVUS per gene in each dataset, represented by stacked bars. (B) Number of cases (represented by dot size) affected by P/LP/PVUS across genes in each dataset. (C) Burden test results for MMRF and Familial datasets against the gnomAD non-cancer cohort. Results from our TCGA germline study by Huang et al. were included for a pan-cancer level comparison. The numbers in each box indicate the percentage (%) of carriers (carrier frequency) of P/LP variants of each gene per cohort. The black outline indicates significant (FDR≤0.05) enrichment for P/LP variants of that gene; the gray outline indicates suggestive (FDR≤0.15) enrichment. Only variants in the 158 candidate CPGs and other MM-related genes are represented. See Table S2.

Figure 3

LOH and biallelic events in the MMRF dataset (A) Comparison of variant allele frequencies (VAFs) in tumor and normal samples reveals events undergoing LOH in the tumor. Dots represent variants; diagonal line indicates equal tumor and normal VAFs (i.e., neutral selection); green represents suggestive LOH (FDR≤0.15); red represents significant LOH (FDR≤0.05); blue represents events not statistically significant. (B) Number of variants showing different types of LOH classified based on somatic copy number changes; only significant LOH events were classified. We highlight the LOH of an ATM P stop-gain variant (p.R2598∗) due to copy number deletion of the wild-type allele (shown in red). Data are represented as the total number of variants. (C) Lolliplot representing a candidate biallelic event of the same ATM variant coupled with a somatic event (p.T2853R) in a 45 year/o, female MM patient with no family history of cancer. See Table S8. (D–F) Trans germline-germline (D) and germline-somatic events (E and F). Germline-germline events were also evaluated for the families. Dots represent a trans event, colored by sample. See Table S5.

Figure 4

Effect of candidate predisposition variants on gene expression in the MMRF (A) Genes significantly associated (FDR≤0.05, linear regression) with higher or lower expression in carriers of P/LP/PVUS. Significance is represented as -log10(FDR) (y axis), and the estimated change in gene expression level is given as log2 fold change (coefficient on the x axis). Dots represent genes. (B) Gene expression distribution in carriers of P/LP/PVUS. Dots indicate gene expression percentile in the carrier relative to other cases, depicted in the y axis. Variants in oncogenes associated with >50% expression are labeled, and those associated with >75% expression are written in red. Variants in TSGs associated with <50% expression are labeled, and those associated <25% expression are written in blue. Variants in genes not classified as tumor suppressor genes or oncogenes are also labeled.

Figure 5

Germline predisposition variants supported by multiple lines of evidence in the MMRF cohort UpSet plot showing variants prioritized by CharGer that present additional evidence of pathogenicity through analyses of LOH, expression association, and co-occurrence with somatic mutations and copy-number variants.

Figure 6

Rare germline copy number variants (gCNVs) in MM (A) Rare gCNVs (AF <0.6% considering 50% overlaps) detected from WES using XHMM. CNV value is represented by the normalized read depth of the genomic region (x axis). (B) Percentage of gCNVs affecting single vs. multiple genes. (C) Distribution and pathogenicity of rare gCNVs across 158 CPGs and other MM-related genes. (D) gCNVs along with its CNV value and expression percentile. Only gCNVs validated by qPCR in the families are shown. For the MMRF, only gCNVs for which the expected transcriptional effect was observed are shown. (E) Expression quantile associated with each gCNV in MMRF. Dots represent gCNVs, colored by pathogenicity. See Table S9.

Figure 7

Segregating variants in patients with familial MM (A) Pedigree of familial MM kindred carrying BRCA2 p.Val2179fs mutation. (B) Pedigree of MM kindred carrying ATM p.Ile2179Thr mutation across 3 generations. (C) Pedigree of familial MM kindred carrying CHEK2 p.Ile200Thr mutation across 2 generations. See Table S10.