Landscape of somatic mutations and clonal evolution in mantle cell lymphoma

Landscape of Mutations in MCL.

We performed WGS and WES of 4 and 29 MCL, respectively. These patients were representative of the broad clinical and biological spectrum of the disease, including five patients with an indolent clinical evolution. In addition, we performed WES of six MCL cell lines. Selected mutated genes were investigated in a validation series of 172 MCL patients (SI Appendix, Tables S1–S6). We detected about 3,700 somatic mutations per tumor (1.2 per Mb) by WGS (SI Appendix, Figs. S1A and S2A and Dataset S1). The most common substitution was the transition C > T/G > A, usually occurring in a CpG context. Two IGHV-mutated MCL showed a higher proportion of A > C/T > G mutations than the two IGHV-unmutated cases (SI Appendix, Fig. S1B). The breakpoints of the t(11;14) translocation differed in the four cases (13) (SI Appendix, Table S7). We next investigated the presence of regional clustering of somatic mutations by constructing “rainfall plots” (SI Appendix, Fig. S2B). Foci of hypermutation or kataegis, a phenomenon recently described in breast cancer (14), were observed in three cases. They were more frequent in the two IGHV-mutated tumors. These clusters occurred around the 11q13 breakpoint of the t(11;14); the Ig genes at 2p11, 14q32.33, and 22q11.22; and the deleted 9p21.3 region, but we also observed this phenomenon in regions without apparent structural alterations and lacking coding genes (SI Appendix, Table S8). The same clusters of hypermutation were observed in the sequential sample of case M003, suggesting that kataegis in MCL may occur as an early phenomenon that remains stable during the evolution of the disease.

We further characterized the spectrum of mutations in 29 MCL by WES (SI Appendix, Tables S1, S3, and S9 and Dataset S2). All these cases were also analyzed by SNP array for copy number alterations (CNA) and copy number neutral loss of heterozygosity (SI Appendix, Fig. S3 and Table S10). We identified 652 protein-coding genes with somatic mutations affecting the structure of the encoded protein (nonsynonymous changes, frameshifts in the coding sequence, and mutations affecting canonical splicing sites) with a median of 20 mutations per case (range 8–47). Twenty-five of the 33 mutated genes in at least two samples were mutated at a rate significantly higher than expected by chance, and all tumors harbored mutations in at least 1 of these 25 genes (Fig. 1 and SI Appendix, Table S11). Similarly, five of the six MCL cell lines also had mutations in at least one of the recurrently mutated genes identified in primary tumors (Fig. 1, SI Appendix, and Dataset S3). Chromosomal CNAs were present in 26/29 (90%) cases (mean 11.1 ± 9.2 per altered case) (SI Appendix, Fig. S3 and Table S10). Tumors expressing SOX11 showed a significantly higher number of CNAs than SOX11-negative MCLs (mean 13 ± 9 versus 2 ± 2; P = 2.1 × 10⁻⁵), despite the similar number of somatic mutations (mean 24 ± 12 versus 17 ± 9; P = 0.141) (Fig. 1). Interestingly, five patients who did not need treatment (median 55 mo, range 4–147) had significantly fewer somatic protein-coding mutations (mean 11 ± 4 versus 25 ± 11, P = 3.4 × 10⁻⁵) and CNAs (mean 2 ± 3 versus 12 ± 9; P = 0.001) than patients who required treatment at diagnosis (n = 24).

• Download figure
• Open in new tab
• Download powerpoint

Fig. 1.

WES in 29 cases and 6 MCL cell lines. Heatmap with the mutation pattern of the 25 significantly recurrent mutated genes. Each row represents a gene and each column represents a primary tumor/cell line. Vertical bar graphs show the total number of somatic mutations by WES (blue) and somatic CNAs by SNP array (green) for primary tumors, and the total number of nonpolymorphic variants and CNAs for cell lines. The plot below the case label indicates sample characteristics (SOX11 expression and IGHV gene status).

Recurrent Somatic Mutations in MCL.

ATM, CCND1, and TP53, previously described as drivers in MCL, were among the most frequently mutated genes in this study. ATM mutations were found in 12 of the 22 (55%) tumors expressing SOX11, but in none of the SOX11-negative MCL (P = 0.023) (Table 1, Fig. 1, and SI Appendix, Table S9). Six of the mutations were associated with deletions of the wild-type allele, whereas five cases with no 11q loss had two different ATM mutations per case and only one was a single mutation with no 11q deletion (SI Appendix, Fig. S4A). These mutations mainly truncated or affected functional domains (SI Appendix, Fig. S5). CCND1 mutations were found predominantly in exon 1 (9 of 10 CCND1-mutated cases) (SI Appendix, Fig. S6) and were more frequent in SOX11-negative MCL [6/7 (86%) versus 4/22 (18%); P = 0.03] and with IGHV-mutated [7/12 (58%) versus 3/16 (19%), P = 0.05], suggesting their acquisition in the germinal center microenvironment (Fig. 1, Table 1, and SI Appendix, Table S9). TP53 mutations [8/29 (28%)] were associated with 17p alterations in six cases, and only one case without 17p alteration had two mutations. TP53 mutations were equally distributed in tumors regardless of SOX11 expression or IGHV mutations (Fig. 1).

We also identified recurrent mutations in WHSC1, MLL2, BIRC3, MEF2B, and TLR2 (Table 1; Fig. 1), as well as NOTCH2 in one case. WHSC1 encodes a histone 3 methyltransferase of lysine-36 (H3K36) that has not been found previously mutated in lymphomas. Two missense mutations (p.E1099K and p.T1150A) were recurrently found in two cases each. Both residues are in close proximity and affect two of the most conserved domains in exons 18 and 19. We further analyzed these exons in 101 additional tumors and confirmed the presence of the same mutations in nine more cases [total 13/130 (10%)] (Table 1, Fig. 2A, and SI Appendix, Table S9 and Fig. S7). Interestingly, WHSC1 (also named MMSET or NSD2) is the target of the IGH-translocation t(4;14) in plasma cell myeloma (PCM), where it is overexpressed and associated with an increase in H3K36 and a decrease in H3K27 methylation across the genome (15). Gene expression analysis of 8 WHSC1-mutated and 31 WHSC1-unmutated MCL identified 236 genes differentially expressed (false discovery rate <5%) with the majority of these genes [192/236 (81%)] up-regulated in the WHSC1-mutated cases (Fig. 3B; SI Appendix, Table S12). A gene set enrichment analysis (GSEA) using previously published lymphoid gene expression signatures (15) demonstrated that WHSC1-mutated cases displayed significant overexpression of several signatures related to proliferation and cell-cycle regulation (SI Appendix, Fig. S8) (16). Interestingly, WHSC1-mutated MCL showed a highly significant overexpression of the gene signature up-regulated in PCM with the t(4;14) translocation overexpressing WHSC1 (17). In addition, WHSC1-mutated MCL showed overexpression of genes up-regulated by wild-type WHSC1 or the gain-of-function exon 19 mutant WHSC1 in the KMS11 PCM cell line (15) (SI Appendix, Fig. S8).

• Download figure
• Open in new tab
• Download powerpoint

Fig. 2.

WHSC1, BIRC3, and TLR2 alterations in MCL. (A) WHSC1 mutations in MCL (above gene symbol) and PCM (below gene symbol) (41). Amino acid conservation of WHSC1 (blue plot), and multiple sequence alignment of the region containing the two mutations in residues E1099 and T1150. (B) Heatmap showing the 236 differentially expressed genes in WHSC1-mutated versus WHSC1-unmutated MCL cases. (C) BIRC3 mutations in MCL (Upper) and other hematologic neoplasms (Lower) (38, 39, 42). (D) Graphic representation of 11q deletions in BIRC3-mutated cases. Ten of 11 mutated MCL cases carried deletions encompassing BIRC3 gene (11q22.2). (E) Plots representing cytokine levels secreted by B cells from tumors and healthy donors. Only IL-1RA and IL-6 showed differences between TLR2-mutated and -unmutated samples. Percentage of increase or decrease of cytokine expression before and after stimulation with PGN-SA (blue bars) and PAM3 (red bars). *Significance comparing the two cases with p.D327V mutation versus TLR2-unmutated tumors (P = 0.037 and 0.026 for IL-1RA and IL-6, respectively). (F) Basal levels of IL-6 and IL-1RA in MCL, CLL, and normal B cells.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 3.

NOTCH2 alterations in mantle cell lymphoma. (A) NOTCH2 mutations in MCL (Upper) and other hematologic neoplasms (Lower) (37, 38, 43). (B) Heatmap showing the 841 differentially expressed genes in NOTCH2-mutated versus NOTCH2-unmutated MCL cases. (C) Actuarial probability of overall survival of MCL patients according to NOTCH2 mutation and NOTCH2 or NOTCH1 mutation.

In addition, the histone methyltransferase MLL2 was mutated in 4/29 primary tumors and 2/6 MCL cell lines. Four of the six mutations were truncating and two were missense changes and affected the conserved FYRN and FYRC domains (Table 1 and SI Appendix, Fig. S9). None of these cases had a deletion of the second allele. MEF2B carried the same p.K23R somatic mutation in exon 2 in 2/29 primary tumors, and a p.N49S mutation in exon 2 was found in the REC-1 cell line. We then expanded the study of exon 2 in 158 MCL cases and found the same p.K23R mutation in four additional tumors [total 6/187 (3.2%)] (Table 1 and SI Appendix, Figs. S7 and S10).Virtually all WHSC1, MLL2, and MEF2B mutations occurred in MCL expressing SOX11.

Deletions of 11q21–q23 are common alterations in MCL. In addition to ATM, BIRC3 is also located in this region (11q22.2). We found inactivating mutations in exon 9 in two MCL cases and a splice-site mutation in an additional patient (Table 1, Fig. 1, and Fig. 3C). We expanded the study and found mutations in 11/173 (6.4%) cases, and these cases had more frequent 11q deletions than BIRC3-unmutated MCL [10/11 (91%) versus 25/87 (29%), respectively; P = 1.1 × 10⁻⁴] (Fig. 2D).

We found two mutations of TLR2 in two SOX11-negative/IGHV-mutated MCL (Table 1). One of these mutations (p.D327V) had been previously identified in one IGHV-mutated chronic lymphocytic leukemia (CLL) (18). To determine the potential functional activity of these mutations, we stimulated primary cells of the two TLR2-mutated MCL cases (M021, M003), the mutated CLL case, and 11 unmutated controls (3 CLL, 4 MCL, and 4 normal B-cell samples) with TLR2 agonists and assessed the response of 25 cytokines. CLL and MCL cells carrying the p.D327V mutation showed a significant increased secretion of IL-1RA and IL-6 and, to a lesser extent, of IL-8 compared with TLR2-unmutated samples (Fig. 2E and SI Appendix, Fig. S11). The MCL carrying the p.Y298S mutation had extremely high basal levels of IL-6 compared with the basal or poststimulation levels of other tumors or normal B lymphocytes (Fig. 2F). Several TLR2 agonist stimuli (PGN-SA, LTA-SA, or PAM3) did not generate additional IL-6 increases in these cells.

Among the other mutated genes, we found mutations of the ubiquitin ligase UBR5, recently described in MCL patients (19), and an inactivating mutation affecting B2M (p.L13fs*10) in a MCL carrying a 15q12–q21.1 deletion, encompassing the B2M locus. However, we did not find mutations in an additional 97 MCL patients, including 3 with a similar 15q monoallelic deletion.

NOTCH2 and NOTCH1 Mutations in MCL.

The finding of a NOTCH2 mutation prompted us to expand the analysis of the HD, TAD, and PEST domains in additional tumors, and we found mutations in 9/172 MCL (5.2%) (Table 1; Fig. 3A). All these mutations generate a premature stop codon within the PEST domain. Gene expression analysis of two NOTCH2-mutated and 19 NOTCH2-unmutated MCL (also wild-type for NOTCH1) showed many differentially expressed genes (n = 841) (false discovery rate < 5%); 42% of them (n = 355) were up-regulated in NOTCH2-mutated cases and were significantly enriched in cell-cycle and metabolic pathways (Fig. 3B, SI Appendix, and Dataset S4). GSEA showed that NOTCH2 mutated samples displayed a significant overexpression of genes up-regulated by NOTCH activation in lymphoid cells (20) and also had a concordant modulation of gene signatures regulated by NOTCH inhibition in MCL cell lines (21⇓–23) (SI Appendix, Fig. S12). NOTCH2 mutations occurred more frequently in blastoid/pleomorphic MCL (66 versus 18%, P = 0.001) and conferred a dismal prognosis [3-y overall survival (OS): 0 versus 62%, P = 2.5 × 10⁻⁴] (Fig. 3C). TP53 mutations were found in 42/192 patients (22%) and were also associated with poor outcome. In a bivariate analysis, both NOTCH2 mutations (hazard ratio: 3.5; 95% confidence interval: 1.3–9.5; P = 0.017) and TP53 mutations (hazard ratio: 2.4; 95% confidence interval: 1.4–4.2; P = 0.003) were identified as independent risk factors for OS (SI Appendix, Fig. S13).

NOTCH1 mutations have been recently described in MCL (20). We investigated this gene and found truncating mutations in 8/172 (4.6%) MCL cases, as well as in MINO and REC-1 cells (Table 1 and SI Appendix, Fig. S14). NOTCH1-mutated tumors were predominantly blastoid/pleomorphic (67 versus 19%, P = 0.03) and showed shorter survival than NOTCH1-unmutated MCL (3-y OS: 33 versus 60%, P = 0.026) (SI Appendix, Fig. S15). NOTCH1 and NOTCH2 mutations occurred in different subsets of tumors because only 1 of the 16 patients with mutations in these genes had mutations in both. Taken together, NOTCH1/2 mutations were present in 9.5% of MCL and identified a subset of tumors with more adverse biological and clinical features including blastoid/pleomorphic morphology (67 versus 13%, P = 1.3 × 10⁻⁵) and a significant shorter survival (3-y OS: 24 versus 63%, P = 3.4 × 10⁻⁴) (Fig. 3C).

Sequencing Simultaneous and Subsequent MCL Samples Reveals Clonal Heterogeneity.

To explore the subclonal architecture of MCL, we sequenced a second tumor sample obtained simultaneously from two different topographic sites (n = 6) or at two time points (diagnosis and disease progression, n = 2). Analysis of the frequency of reads supporting somatic substitution allowed the estimation of major subclonal populations (24). Four of the six patients with simultaneous samples showed the same mutations and CNAs in the peripheral blood (PB) sample and corresponding lymphoid tissue, suggesting the presence of a single major clone at both sites. In contrast, two cases (M023 and M026) had a subset of common mutations in both topographic sites, but they also carried a subset of mutated genes that were exclusive of the PB or tissue tumor sample (Fig. 4 A and B). This pattern of alterations is consistent with two major subpopulations derived from an initial founder clone differentially represented in the two topographic sites. In addition, we also observed a different pattern of genomic alterations at diagnosis and progression in the two cases with sequential samples (Fig. 4C and D and SI Appendix, Table S10). Patient M003 had stable disease without treatment for more than 3 y and then rapidly progressed. The major clone identified at diagnosis gained new mutations and genomic complexity at the time of clinical progression. This new clone carried the same 17p deletion and somatic mutations observed in the initial sample, but had a 16% increase in the number of whole-genome mutations (from 3,928 to 4,665) (SI Appendix, Fig. S1A). These included nonsynonymous mutations in additional genes, a complex DNA copy number profile with 20 acquired CNAs, and chromothripsis involving chromosomes 4 and 12 (Fig. 4C and SI Appendix, Fig. S2A). Patient M002 was treated with chemotherapy after diagnosis and relapsed 3 y later. The major clone at relapse maintained a subset of CNAs and somatic mutations present at diagnosis, but 11 of the initial mutations had disappeared and other mutated genes had emerged (Fig. 4D). This pattern of evolution is consistent with the eradication of the major initial subclone by chemotherapy and the relapse of a new subclone after treatment.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 4.

Subclonal architecture in MCL. Representation of four informative MCL patients with two tumor samples analyzed. The mutation clusters are represented in the upper panel of each case, the shared and stable mutations are in green color, in blue the mutations specific of the first sample and in red the mutations identified only in the second sample. In the lower panel of each case, a detailed representation of the percentage of mutated reads in each of the samples (Left and Right) with the same color code, and with the significantly recurrent mutated genes highlighted in the same color. (A–B) Cases M023 and M026 have two major subclones derived from an initial founder clone differentially represented in simultaneous samples of lymph node (LN) and peripheral blood (PB). (C) Longitudinal analysis in patient M003 at diagnosis and at disease progression previous to treatment. (D) Longitudinal analysis in patient M002 at diagnosis and at first relapse.

Comparison of the allele frequency of mutations in the different simultaneous or evolving subclones may help to infer the dynamic architecture of somatic events in the evolution of tumors (24, 25). Eight of the 25 recurrently mutated genes in MCL (ATM, CCND1, MLL2, KCNC2, KIAA1671, PCSK2, TNRC6B, and TRPM6) were present at similar allelic frequency in the two subclones of different cases, suggesting that they represent early events. In contrast, four recurrently mutated genes (ABCA3, TLR2, TP53, and WHSC1) were seen in only one of the two simultaneous or emerging subclones at progression, supporting the notion that they might constitute later events. We further analyzed by Sanger sequencing 11 additional serial and 8 simultaneous tumor samples from different topographic sites (PB and lymphoid tissues) (SI Appendix, Fig. S16). Interestingly, we observed that BIRC3 mutations were absent at diagnosis in two cases that acquired the mutation in a posttreatment sample associated with the acquisition of an 11q22.1–q24.2 deletion in one of them. Another case showed the BIRC3 mutation in the PB but not in the simultaneous lymph node. Additionally, NOTCH1 was mutated in only one of the synchronic samples in two cases.