Genomic and transcriptomic profiling reveals distinct molecular subsets associated with outcomes in mantle cell lymphoma

Abstract

Mantle cell lymphoma (MCL) is a phenotypically and genetically heterogeneous malignancy in which the genetic alterations determining clinical indications are not fully understood. Here, we performed a comprehensive whole-exome sequencing analysis of 152 primary samples derived from 134 MCL patients, including longitudinal samples from 16 patients and matched RNA-Seq data from 48 samples. We classified MCL into 4 robust clusters (C1-C4). C1 featured mutated immunoglobulin heavy variable (IGHV), CCND1 mutation, amp(11q13), and active B cell receptor (BCR) signaling. C2 was enriched with del(11q)/ATM mutations and upregulation of NF-κB and DNA repair pathways. C3 was characterized by mutations in SP140, NOTCH1, and NSD2, with downregulation of BCR signaling and MYC targets. C4 harbored del(17p)/TP53 mutations, del(13q), and del(9p), and active MYC pathway and hyperproliferation signatures. Patients in these 4 clusters had distinct outcomes (5-year overall survival [OS] rates for C1-C4 were 100%, 56.7%, 48.7%, and 14.2%, respectively). We also inferred the temporal order of genetic events and studied clonal evolution of 16 patients before treatment and at progression/relapse. Eleven of these samples showed drastic clonal evolution that was associated with inferior survival, while the other samples showed modest or no evolution. Our study thus identifies genetic subsets that clinically define this malignancy and delineates clonal evolution patterns and their impact on clinical outcomes.

Keywords: Genetic variation; Genetics; Lymphomas; Oncology.

Figure 1. Recurrent somatic genetic alterations and mutation signatures in MCL.

(A) Recurrent somatic mutations and CN alterations (rows) identified following WES of 134 primary samples (columns) obtained from patients with newly diagnosed (green) and relapsed (red) MCL. Samples were annotated for prior treatment, MIPI risk, IGHV status, and Sox11 expression level when collected. Left: blue labels, recurrent CN deletion; red labels, recurrent CN amplification; black labels, somatic mutations; bold labels, novel CN alterations/mutations. Right: percentage of samples mutated. Top: total number of genetic alterations across the cohort. (B) Contributions of individual mutations to the collective WAP score of TP53. The changes in WAP score P value due to removal of individual mutations are plotted as function of residue number. The radius of the circles around each point in the graphs represent the number of patients with that mutation. Color indicates SOX11 expression. (C) TP53 dimer bound to DNA fragment, PDB ID: 3IGK. One of the monomers is shown in yellow, the other in gray. DNA is shown in orange. The mutations observed in SOX11⁺ and SOX11^– patients are shown as magenta and green, respectively. (D) β Scores from genome-wide CRISPR/Cas9 screens of JeKo-1 of genes identified as having recurrent mutations.

Figure 2. Recurrent SCNAs, cooccurring genetic events, and clinical association.

(A) Significant CN amplifications (left, red) and deletions (right, blue). Left sides of the mirror plots show the incidence of significant focal CNA events. Right sides of the mirror plots show q values for each region. Genes located in the peak of relevant cytobands are listed. (B) Pairwise associations between recurrent genetic alterations found in the 134 MCL samples. Low and high cooccurrence are shown in blue and red, respectively. Intensity of the color reflects the odds ratio. Statistically significant association as determined by q value is marked by asterisks. (C) Number of samples with cooccurrence of the indicated genetic events in the cohort of 134 MCL samples. Significance of Fisher’s exact test indicated by q.

Figure 3. Associations of somatic mutations with clinical outcomes.

(A) Lollipop diagrams of selected putative driver genes showing mutation subtype, position, and frequency. Bottom: y axis indicates the number of identified mutations in the COSMIC database. (B) Kaplan-Meier plots (with log-rank P values) of PFS and OS associated with presence and absence of selected mutations. (C) Samples with SP140 mutations or deletions did not overlap in the cohort. (D) Deletion of SP140 affected its gene expression. SP140 expression TPM value was extracted and plotted from MCL samples with SP140 deletion, mutation, or WT. *P < 0.05. (E) Forest plots of the multivariate analysis of MIPI risk groups and individual genetic factors for PFS and OS in our MCL cohort.

Figure 4. Deletion of chromosome 9 was associated with poor survival.

(A) Chromosome 9 deletion in samples from our cohort. Top: blue line indicates percentage of MCL samples with chromosome 9 deletion at the location. Known tumor suppressors and oncogenes present on chromosome 9 are color coded based on their z score in the CRISPR/Cas9 screen in JeKo-1 cells. Bottom: deletions in 9p (purple), 9q (blue), or large regions (dark red) in samples from our cohort. Homozygous minimal 9p deletions are marked in red. CCF (Supplemental Methods) of chromosome 9 deletion is shown in gray scale. (B) Unsupervised clustering analysis of gene expression in chromosome 9 distinguishes MCL samples with deletions in different region. (C) Volcano plot of genes on chromosome 9 that are differentially expressed between MCL samples that have and do not have chromosome 9 deletions. Downregulated genes that were significantly associated with shorter PFS and OS are indicated in red (Cox’s regression HR <1, P < 0.05). (D) Kaplan-Meier plots of PFS and OS according to type of chromosome 9 deletion.

Figure 5. Coordinate genetic signatures group MCL into 4 clusters associated with clinical outcome.

(A) Nonnegative matrix factorization (NMF) consensus clustering was performed using all somatic mutations and SCNAs in the 134 MCL samples (columns). Clusters 1 to 4 are shown with their associated landmark genetic alterations (boxed for each cluster). Left bar graph shows the correlation of genetic alterations associated with each cluster (q value, Fisher’s exact test). Nonsynonymous mutations, black; low-level deletion (1.0 ≤ CN ≤ 1.7 copies), light blue; high-level deletion (CN ≤ 1.0 copies), dark blue; low-level amplification (3.7 ≥ CN ≥ 2.3 copies), orange; high-level amplification (CN ≥ 3.7 copies), red. Header shows cluster association (C1, black; C2, green; C3, blue; C4, red), clinical group (cMCL, yellow green; nnMCL, light green), Sox11 expression (negative, green; positive, brown), MIPI risk (high risk, dark pink; intermediate risk, median pink; low risk, light pink), pathology status (blastoid or pleomorphic, crimson; classic, bright lilac), and treatment regimen (standard cytarabine-based aggressive regimen, dark blue; other regimen, light blue). (B) Kaplan-Meier plots of PFS and OS of patients grouped into the 4 clusters. *P < 0.05, log-rank test.

Figure 6. Molecular cluster and gene expression signature validated in Barcelona cohort.

(A) Sample inclusion description in the validation cohort. (B) Projective nonnegative matrix factorization consensus clustering was performed using genetic alterations identified from our discovery cohort (Figure 5A). Clusters 1–4 are shown with their associated landmark genetic alterations (boxed for each cluster). Header shows cluster association (C1, black; C2, green; C3, blue; C4, red). (C) Kaplan-Meier plots of OS of patients grouped into the 4 clusters. P indicates log-rank test. Number indicates samples included in each cluster. (D) Integration of genetic and transcriptomic analyses identified gene expression signatures for each genetic subset. The heatmap was generated using normalized enrichment score (NES). Asterisks indicate the significance level of the enrichment.

Figure 7. Integrative analysis of genome and transcriptome reveals a unique gene expression signature in each cluster.

(A) Recurrent somatic mutations, SCNAs, and gene expression associated with SCNAs. Top panel: x axis shows the chromosome location of recurrent somatic mutations; y axis indicates the frequency of mutations detected in our MCL cohort (n = 134). Genes shown in purple have a mutation incidence of greater than 5%. Bottom panel: left y axis indicates proportions of CN deletion (DEL) and amplification (AMP). Each dot represents a gene at its chromosome location. Genes with absolute CN < 1.7 or > 2.3 were defined as deleted or amplified, respectively. Genes with a deletion incidence > 10% are shown in blue, and genes with an amplification incidence > 10% as red. (B) Integration of genetic and transcriptomic analyses identified unique gene expression signatures for each genetic subset. The Hallmark and KEGG gene sets and Signature database were used for Gene Set Enrichment Analysis. The heatmap was generated using normalized enrichment score (NES). Red indicates an upregulated pathway in the cluster compared with other clusters, while blue indicates a downregulated pathway. Asterisks indicate the significance level of the enrichment. (C) Proposed model for the 4 MCL subgroups. Clusters 1–4 were all associated with distinct genetic events and gene expression signatures. C1 had indolent disease and carried memory B cell gene signature. C2–C4 had more aggressive clinical courses and expressed CCR6-negative light zone or naive B cell gene signature.

Figure 8. Clonal driver events associated with clinical outcomes.

(A) CCF values for each sample affected by a recurrent genetic alteration across all 134 samples. Median CCF values are shown (top, bars represent the median and interquartile range for each genetic alteration). Alterations with a CCF value of greater than 0.9 were defined as a clonal event. The cumulative proportion of a recurrent genetic alteration found as clonal (blue) or subclonal (red) in the cohort is shown in bottom plot. (B) Computational inference of temporal order of genetic alterations in MCL. Arrows indicate when paired clonal and subclonal genetic alterations were found in the same sample. Dashed lines indicate the temporal order was found in 3 or more samples; solid lines that the temporal order was found in 5 or more samples. (C) Kaplan-Meier plot of PFS according to the number of clonal driver events.

Figure 9. Clonal evolution pattern in MCL and its association with clinical outcome.

(A–C) Depiction of tumor clonal evolution from diagnosis to relapse in a representative patient (MCL34). (A) Dynamic changes in genetic alterations during disease progression. Representative genetic alterations for each cluster are listed in the plot. (B) Clonal evolution estimated using PhylogicNDT. The mean CCF and 95% CI of each cluster are indicated. (C) Fish plot showing the clonal evolution process. The width of each time point indicates the clonal fractions of each subclone population. (D) Joint distributions of CCF values of genetic alterations across 2 (or more) time points (ND, newly diagnosed; P, progression; R, relapse; R1, first relapse; R2, second relapse) were estimated using clustering analysis. Each line corresponds to cluster of genetic alterations (range 3–33) and illustrates the dynamic changes in CCF at the different time points for clusters. We classified any CCF increase or decrease greater than 0.5 between 2 time points for any cluster as extreme evolution. CCF changes between 0.2 and 0.5 or less than 0.2 were classified as moderate evolution or no evolution, respectively. (E) Sample interval and number of clonal clusters in patients with either extreme evolution or with modest or no evolution. (F) Kaplan-Meier plot of survival from either first sampling (left) or second sampling (right).