Timing the initiation of multiple myeloma

Abstract

The evolution and progression of multiple myeloma and its precursors over time is poorly understood. Here, we investigate the landscape and timing of mutational processes shaping multiple myeloma evolution in a large cohort of 89 whole genomes and 973 exomes. We identify eight processes, including a mutational signature caused by exposure to melphalan. Reconstructing the chronological activity of each mutational signature, we estimate that the initial transformation of a germinal center B-cell usually occurred during the first 2^nd-3^rd decades of life. We define four main patterns of activation-induced deaminase (AID) and apolipoprotein B mRNA editing catalytic polypeptide-like (APOBEC) mutagenesis over time, including a subset of patients with evidence of prolonged AID activity during the pre-malignant phase, indicating antigen-responsiveness and germinal center reentry. Our findings provide a framework to study the etiology of multiple myeloma and explore strategies for prevention and early detection.

Fig. 1. Multiple myeloma mutational signature landscape.

a The 8 mutational signatures extracted from 89 WGSs from 52 MM patients. b, c The relative (b) and absolute (c) contribution of each mutational signature for each patient. The x-axis labels are colored according to disease stage and treatment history: red = treatment naive; blue = relapsed cases without a full clinical annotation; black = relapsed cases after transplant with high-dose melphalan; purple = relapsed cases never exposed to transplant with high-dose melphalan; gray = missing clinical data. d Relative contribution of all eight MM mutational signatures on exome data collected at diagnosis (ND = newly diagnosed), relapsed, and when the disease was refractory to both bortezomib and lenalidomide (i.e., double-ref). e Transcriptional strand bias profile of patients with (top) and without (bottom) evidence of SBS-MM1.

Fig. 2. APOBEC3A and APOBEC3B activity in multiple myeloma and other cancers.

a, b The ratio between APOBEC3A/APOBEC3B in all multiple myeloma whole genomes (a) and in CoMMpass exomes with high APOBEC mutational burden and available structural variant data (b). Red and black x-labels were used to highlight patients with and without MAF translocations, respectively. c Differential APOBEC3A/APOBEC3B ratio between patients with (n = 40) and without MAF (n = 652) translocations (p < 0.0001 by wilcoxon.text R function). Boxplots show the median and interquartile range; observations outside this interval are shown as dots. d The APOBEC3A/APOBEC3B ratio in our MM cases and all tumors included in the PCAWG consortium with evidence of APOBEC activity. Red dots reflected the median APOBEC3A/3B ratio for each cancer type. e Correlation between APOBEC mutational burden (log scale) and APOBEC3A/APOBEC3B ratio. R-squared and p values (p < 0.0001) were estimated using linear regression (lm R function).

Fig. 3. Kataegis in multiple myeloma.

a The distribution of all kataegis events extracted across 52 MM patients, each color reflecting a distinct patient. b Number of kataegis events per patient. c In kataegis events close to at least one SV breakpoint, APOBEC (SBS2 and SBS13) was the main mutational process. d Other kataegis events not associated with any SV breakpoints were dominated by nc-AID. e Example of APOBEC-mediated kataegis associated with chromothripsis. f Example of nc-AID kateagis not associated with SV. In e, f, black dots represent the chromosome ploidy status. Vertical black, blue, green, and red lines reflect translocations, inversions, tandem duplications, and deletion, respectively. Below each copy number/structural variant plot, we reported the inter-mutational distance of all SNVs, color-coded by class (blue: C>A, black C>G, red C>T, gray T>A, green T>C, pink T>G).

Fig. 4. Chronological reconstruction of mutational processes in multiple myeloma.

a Heatmap showing the relative contribution of mutational signatures during the three evolutionary phases: early clonal, late clonal, and subclonal. Only patients with more than one sample collected at different time points were considered (n = 26). b Boxplots of single-cell gene expression for specific genes across 20,586 single bone marrow plasma cells from 29 newly diagnosed patients and 11 control donors. Each box represents 0.25–0.75 percentile of unique molecular identifier (UMI) count with line extension to 0.1–0.9 percentile; dot represents the mean UMI count. Patients are color-coded by disease (gray—controls from hip replacement surgery (Hip CTRL), yellow—MGUS, light red—SMM, red—MM, dark red—AL amyloidosis). c Mutational signature composition of each phylogenetic tree of patients with multiple samples collected at different time points and evidence of SBS-MM1. Asterisks represent the presence of transcriptional strand bias in SBS-MM1 single-, tri- and penta-nucleotide contexts. d Mutational signature contributions in CoMMpass patients with samples collected at baseline and first relapse and investigated by both whole exome and low coverage long insert WGS (n = 24). Mutations were grouped according to their position in each patient phylogenetic tree. x-axis labels colored by transplant and high-dose melphalan status: black for high-dose melphalan (HDM) exposed and purple for non-exposed. e, f The relative (e) and absolute (f) contribution of APOBEC (SBS2 and SBS13) to clonal and subclonal mutations from 788 CoMMpass MM patients. p Values (p < 0.001) were calculated using the wilcoxon.text R function. g Mutational signature contributions for non-coding compared with nonsynonymous (Nonsyn) mutations in newly diagnosed MM patients from both CoMMpass and WGS data.

Fig. 5. AID and APOBEC activity during early phases of cancer development.

a Relative contribution of SBS2 and SBS9 before and after large chromosomal gains. Each patient’s clonal mutation catalog was divided into two groups: mutations acquired before the gain (VAF corrected for the normal contamination = 66%), mutations present on one single allele (VAF corrected for the normal contamination 33%). p Values were estimated with the wilcoxon.text R function. b Heatmap showing the relative contribution of all mutational signatures across different time windows for each patient. c 96-class mutational profile of all mutations acquired before (top) and after the copy neutral LOH (bottom). d 96-class mutational profile of all mutations acquired on one duplicated allele before its second duplication. e Schematic representation of the four main mutational signature patterns over time. In the bottom right panel, APOBEC is colored in brown to reflect the high APOBEC contribution.

Fig. 6. Clock-like properties of SBS5 in multiple myeloma.

a Linear regression model (p < 0.0001 with lm R function) for SBS5 mutational burden and age in 764 CoMMpass whole-exome sequencing cases. b Linear mixed-effects model for SBS5 mutation rate in 72 CoMMpass patients with sequential samples collected at different time points (p < 0.0001 with lmer R function). Points represent observed SBS5 mutational burden in phylogenetic branches, colored by patient. Colored lines represent patient-specific SBS5 mutation rates (i.e., slopes), with the population average as a black line surrounded by a shaded 95% confidence interval. c Linear mixed-effects model for SBS5 mutation rate in our cohort of multiple myeloma genomes; legend as in b (p < 0.0001 with lmer R function). d Linear regression model for SBS5 mutational burden and age in non-hypermutated B cell lymphomas and CLL WGSs included in the PCAWG consortium (p < 0.0001 with lm R function).

Fig. 7. Timing the first multi-gain event in multiple myeloma.

a SBS5 mutation burden on duplicated (CN = 2) and non-duplicated (CN = 1) alleles of large chromosomal gains occurring in the same time window (top), and molecular time estimates based on the corrected ratio of duplicated and non-duplicated mutations (dots below). 95% confidence intervals in the upper panel are based on the uncertainty of mutational signature fitting, which are propagated to different estimates (colors) in the panel at the bottom. 95% confidence interval bars in the lower panel represent the uncertainty of molecular time estimation given a number of duplicated and non-duplicated SBS5 mutations. b Estimated patient age at the first (dark green) and second (green) multi-gain events with 95% CIs. Blue and red dots represent age at the MRCA emergence and first sampling, respectively. MRCA timing is only shown for patients with multiple samples.

Fig. 8. Single-cell expansion model for melphalan and GC-related mutational signatures.

a Cartoon summarizing the mutagenic impact of melphalan in MM patients. In the case of tumor cell post-transplant engraftment, MM cells will not have any melphalan-induced mutations (top). When a cancer cell is exposed to melphalan, it acquires unique mutations that will be detectable only in case of a single tumor cell expansion (center). In the absence of single-cell expansion, melphalan-induced mutations will be present in each exposed tumor cell but undetectable due to the low cancer cell fraction. b Proposed model of GC-dependent clonal expansion and AID-mediated hypermutation during the pre-malignant phase of MM development. Finally, a single cell becomes independent of the GC, going on to expand and give rise to the MM clone.