The comparison of samples with and without mutations revealed that the presence of mutations is strongly associated with BCL2 expression, with a median fold change (FC) = 5 and FC = 1.2, in PCAWG BNHL and CLL samples, respectively, and FC = 9.5 in HMF BNHL samples (Fig. 2A). Interestingly, basal BCL2 levels were approximately 10 times higher in CLL than BNHL, potentially explaining the smaller effect of the mutations in CLL. To further validate our findings, we analyzed 1955 cell lines from the Broad Institute Cancer Cell Line Encyclopedia (CCLE). The analysis identified 83 BCL2hot1 mutations, mostly in lymphoid cell lines (29/36, 80.5%), including DLBCL cell lines (15/29, 51.7%), and rarely in other cell lines (7/1955, 0.4%). Consistent with the previous analyses, lymphoid cell lines with BCL2hot1 mutations exhibited significantly higher BCL2 expression than wild-type lines (FC = 1.3, p = 2.9e-4). The effect was even more pronounced in DLBCL samples analyzed separately (FC = 2.5, p = 6.7e-4; Fig. 2B).
Fig. 2: Associations of BCL2hot1 mutations in lymphoid malignancies.The alternative text for this image may have been generated using AI.
A BCL2 levels in PCAWG BNHL, HMF BNHL, and PCAWG CLL cohorts with (+) and without (−) BCL2hot1 mutations. Sample sizes are indicated below the charts (also in B–D, E, G). B BCL2 levels (log2 Transcripts Per Million (TPM)) in Broad Institute CCLE lymphoid and DLBCL cell lines with and without BCL2hot1 mutations. Note that, as CCLE utilizes whole-exome sequencing, the identified mutations primarily localize to the BCL2 first exon. C BCL2 levels in PCAWG BNHL, and PCAWG CLL samples stratified by the presence of BCL2hot1 mutations and BCL2 copy numbers. NI (not identified) indicates groups where no samples were found (also in D). D BCL2 levels in BNHL samples stratified by the presence of BCL2hot1 mutations and t(14;18). The overall ANOVA and adjusted post-hoc p-values are indicated. E The impact of BCL2hot1 mutations on BCL2 isoform levels. (Above) Scheme of BCL2 isoforms showing t(14;18) breakpoints (black bars) and miR-15a/16-1 binding sites (purple arrowhead). (Below) BCL2 isoform levels in BNHL samples with and without BCL2hot1 mutations, with the nominal (unadjusted for multiple comparisons) p-values. (Right) Stacked-bar chart of isoform proportions in samples with and without BCL2hot1 mutations. F BCL2 allelic imbalance in 8 DLBCL HMF samples with BCL2hot1 mutations. (Above) RNA-seq coverage of the BCL2 constitutive exonic sequence within BCL2hot1 in two representative samples. Light-gray bars indicate nucleotide coverage; colored bars indicate coverage of mutant alleles; dark-gray bars indicate coverage of wild-type and unannotated variant alleles. (Below) Correlation of mutant allelic frequencies at DNA (x-axis) and RNA (y-axis) levels. Colors indicate mutations identified in different samples. G Kaplan–Meier survival analysis of DLBCL patients across mutually exclusive (at sample level) cohorts: without BCL2hot1 mutations and t(14;18) (dark blue line), with BCL2hot1 mutations and t(14;18) (green line), and with BCL2hot1 mutations only (red line). Significance was assessed using the log-rank test; vertical ticks indicate censoring times. H Volcano plot of differentially expressed genes in PCAWG BNHL samples with vs. without BCL2hot1 mutations. I GSEA of differentially expressed genes shown in H (NES Normalized Enrichment Score, FDR False Discovery Rate).
Subsequent analyses showed that the effect of BCL2hot1 mutations (rpartial = 0.6, p < 0.0001) on BCL2 level is independent of and stronger than the effects of BCL2 copy number (rpartial = 0.3, p = 0.0026; multiple regression p < 0.0001) (Fig. 2C) and t(14;18) (Fig. 2D). Notably, 15 PCAWG BNHL samples with BCL2hot1 mutations exhibited even higher BCL2 levels than samples with mutations and the translocation, which is consistent with11, where the authors reported BCL2 overexpression in the absence of t(14;18) in FL samples. These findings challenge the suggested “enhancer hijacking” effect, which attributes BCL2 overexpression to its juxtaposition with the Eµ enhancer of the immunoglobulin heavy chain 6 (IGHJ6) gene. Although the effect of the translocation on the expression of surrounding genes cannot be completely ruled out, analysis of IGHJ6 (on chromosome 14) and genes flanking BCL2 also does not show a distinctive effect of the translocation (Fig. S3). On the other hand, the translocation may disrupt miR-15/16 binding sites in the BCL2 3′UTR1, relieving BCL2 repression and potentially explaining the clustering of breakpoints upstream of miR-15/16 binding sites (Fig. 2E[above]). We also found that the increase of BCL2 expression is independent of the number of mutations in BCL2hot1 (Fig. S4).
The comparison of genomic and transcriptomic allelic fractions of BCL2hot1 mutations located in the constitutive exonic sequence (chr18:60,985,314-60,986,045) in 8 HMF DLBCL samples revealed a very strong allelic imbalance in favor of the mutated alleles (Fig. 2F). The majority of mutations showed very strong allelic enrichment, representing >90% of transcript reads, with a median enrichment of mutated alleles at the RNA vs. DNA level across all mutations of 11 (ranging from 2.5 to 88 in individual samples). Moreover, whole-genome allelic-specific expression analysis using phASER12 identified BCL2 with very high confidence (−log10p(adjusted) from 54 to 320 across individual samples) as the top-ranked gene exhibiting allelic imbalance. The allelic imbalance analysis demonstrated that the effect of BCL2hot1 mutations is allele-specific.
Further analysis of transcript levels in the HMF dataset suggests that BCL2hot1 mutations may have an unequal effect on different BCL2 splicing isoforms (Fig. 2E[below]). The observed trend indicates an increase in ENST00000589955 (FC = 11.8), the isoform missing the 2nd exon encoding the C-terminal domain responsible for interaction with the mitochondrial membrane and activation of apoptosis, but retaining the N-terminal domain taking part in angiogenesis13, at the expense of all other isoforms.
The analysis of the effect of mutation occurrence on the overall survival of DLBCL patients revealed that BCL2hot1 mutations are associated with reduced survival, independent of t(14;18) (Fig. 2G), consistent with findings from the GOYA study, which demonstrated a similar effect for BCL2 coding mutations14. Due to an insufficient number of samples with available clinical data in specific subgroups, such as the lack of control samples without BCL2hot1 mutations in FL, survival analysis could not be performed for other lymphoma subtypes.
The differential transcriptomic analysis of PCAWG BNHL samples with BCL2hot1 mutations revealed 652 upregulated (including BCL2) and 549 downregulated genes (Fig. 2H). This analysis identified a robust transcriptomic shift toward cell cycle suppression and reduced DNA replication. Key drivers of cell cycle progression were among the most deeply downregulated genes, including E2F2, MYC, CDC25A, and E2F1. The suppression of E2F1 is particularly interesting, as it binds the BCL2 5′UTR15,16, suggesting a feedback loop regulation of BCL2 expression. Furthermore, essential DNA replication machinery was significantly suppressed, including components of the Minichromosome Maintenance complex (MCM4 and MCM2), the Origin Recognition Complex (ORC6), and the standard replication marker PCNA. Conversely, we observed a slight upregulation of well-established cell cycle inhibitors and classical markers of G0/G1 arrest, such as CDKN1B and RBL2. In addition to these proliferation dynamics, other notable cancer-related genes, such as TERT and AID encoding AICDA, were significantly downregulated (Fig. 2H). Gene Set Enrichment Analysis (GSEA) revealed pronounced downregulation of cell cycle and DNA replication pathways, and upregulation of cell adhesion and hematopoietic differentiation in BCL2hot1 samples (Fig. 2I).
