Abstract
Whole-genome sequencing of longitudinal tumor pairs representing transformation of follicular lymphoma to high-grade B cell lymphoma with MYC and BCL2 rearrangements (double-hit lymphoma) identified coding and noncoding genomic alterations acquired during lymphoma progression. Many of these transformation-associated alterations recurrently and focally occur at topologically associating domain resident regulatory DNA elements, including H3K4me3 promoter marks located within H3K27ac super-enhancer clusters in B cell non-Hodgkin lymphoma. One region found to undergo recurrent alteration upon transformation overlaps a super-enhancer affecting the expression of the PAX5/ZCCHC7 gene pair. ZCCHC7 encodes a subunit of the Trf4/5-Air1/2-Mtr4 polyadenylation-like complex and demonstrated copy number gain, chromosomal translocation and enhancer retargeting-mediated transcriptional upregulation upon lymphoma transformation. Consequently, lymphoma cells demonstrate nucleolar dysregulation via altered noncoding 5.8S ribosomal RNA processing. We find that a noncoding mutation acquired during lymphoma progression affects noncoding rRNA processing, thereby rewiring protein synthesis leading to oncogenic changes in the lymphoma proteome.
Similar content being viewed by others
Main
B cells undergo a series of programmed genomic alterations that enable the immunoglobulin light and heavy chain loci to generate high-affinity antibodies against invading pathogens. First, B cells undergo variability, diversity and joining (VDJ) recombination in the bone marrow with subsequent somatic hypermutation (SHM) and class switch recombination (CSR) occurring within lymphoid follicles once the cells traffic to secondary or tertiary lymphoid organs1,2. Both CSR and SHM require the essential activity of the enzyme activation-induced cytidine deaminase (AID) that incorporates mutations via single-strand DNA nicks at variable region genes and introduces DNA double-strand breaks at switch sequences to initiate the process of SHM and CSR, respectively3,4,5,6. VDJ recombination, SHM and CSR all can also lead to DNA alterations outside the boundaries of the immunoglobulin gene loci, many of which promote lymphomagenesis7,8. The mechanism by which AID recognizes its target DNA sequences in the B cell genome is incompletely understood5,9,10,11,12,13,14. In this context, a better understanding of DNA targeting by AID, specifically in models of lymphoma progression, would be a significant advance. Furthermore, the consequences of AID-mediated nonimmunoglobulin locus-associated somatic mutation, so-called aberrant somatic hypermutation (aSHM) identified in mice and humans at coding and noncoding sequences, are only beginning to be evaluated4,7,15. The landscape of coding-region mutations observed in lymphoma does not account for the numerous alterations in gene expression required for lymphomagenesis. Therefore, it is possible that aSHM affecting gene regulatory regions greatly contributes to perturbations in gene expression at the transcriptional and translational levels, beyond altering single specific genes at or adjacent to the sites of aSHM. This role for aSHM could have important implications for our understanding of the pathophysiology of lymphoid malignancies and potentially neoplasia in general.
Genomic alterations acquired during lymphoma transformation
aSHM within both coding and noncoding regions has been observed in several classes of B cell non-Hodgkin lymphoma (B-NHL), particularly those originating from germinal center B cells16,17. Most low-grade B-NHLs are relatively indolent and, while often incurable, are not associated with heightened mortality. However, low-grade lymphomas can transform into more aggressive lymphomas18. For example, 25–35% of patients with low-grade follicular lymphoma (FL) experience transformation from a clinically indolent state to an aggressive and frequently fatal diffuse large B cell lymphoma (DLBCL)19,20. Prior genomic studies have investigated changes occurring upon FL transformation21,22,23,24,25,26,27. Several clinical and molecular prognostic indices for risk stratification of FL and prediction of transformation also have been proposed28,29. In our study, we have focused on transformation-associated DNA alterations observed in an important subset of B-NHL—‘double-hit’ lymphomas that harbor MYC rearrangements in addition to the BCL2 rearrangement characteristically observed in FL and which are highly aggressive and difficult to treat (Extended Data Fig. 1a).
In using longitudinal samples from the same patient, we sought to characterize aSHM events occurring at different stages during lymphoma progression and identify mutations that appear specific to FL transformation as opposed to those incurred during the development of de novo DLBCL. Because transformation to double-hit lymphoma (DHL), by definition, includes acquisition of an AID-dependent MYC translocation25, we felt that the role of AID in lymphoma transformation would be well illustrated through these samples. We identified a series of eight patients (clinical information described in Supplementary Fig. 1) diagnosed with DHL and for which preceding FL specimens were also available, with time to transformation ranging from 6 to 161 months (Fig. 1a). Longitudinal FL/DHL samples and, when available, nontumor DNA from nonneoplastic specimens for the same patient (for example, bone marrow and appendix) were subjected to whole-genome sequencing (WGS; see Supplementary Tables 1–3 for mutation information). As expected, characteristic translocations of the 3′ end of BCL2 to the IGH locus were detected in all lymphomas, with MYC translocations to various partner loci seen upon transformation to DHL (Fig. 1b). In addition to the acquisition of MYC translocations, transformation-specific changes included both increasing aSHM at the BCL2 promoter and increasing variant allele frequency of the BCL2 translocation observed in FL (Fig. 1c–e). Detailed evaluation identified break-end insertions at BCL2 translocation breakpoints (with signature insertions via TdT enzyme) and blunt end joining at the MYC locus (Extended Data Fig. 1b–e), indicating that BCL2 translocations are recombination activating gene (RAG)-endonuclease complex-dependent whereas MYC translocations are AID-dependent. These findings support the acquisition of IGH-BCL2 translocations in immature, RAG-expressing B cells (Extended Data Fig. 1f) followed by subsequent oncogenic mutations and ultimately MYC translocation upon the development of DHL30,31.
a, Eight cases of DHL with an available preceding FL sample diagnosed in the previous 17 years were identified. All DHL fulfilled WHO diagnostic criteria66, showing large B cell lymphoma morphology and harboring both MYC and BCL2 translocations identified in fluorescence in situ hybridization studies. For all patients, the FL sample preceded the DHL sample by months to years (average time to transformation 59 months, range 6–161 months). For one patient (P8), two FL samples were available in addition to the subsequent DHL. b, Schematic describing the BCL2 and MYC rearrangements observed for each patient. From WGS data, we identified all rearrangement breakpoints in the BCL2 and MYC loci. On the BCL2 locus, all tumor samples were found to harbor translocations to IGH, with breakpoints clustered in the 3′ UTR and downstream of BCL2. MYC rearrangements were mainly observed as DHL tumor-specific, with the exception of P4, which harbored an MYC translocation in both the FL and DHL samples. All three patients harboring MYC-IGH translocations (blue font) show breakpoints close to the first intron of the MYC gene. c, Mapping of BCL2-IGH and MYC-IGH translocations, and aSHM at BCL2, MYC and corresponding IGH loci. The occurrence of aSHM on BCL2 and MYC was significantly associated with IGH as the translocation partner (two-sided Fisher’s exact test P = 1.27 × 10−6). d,e, Plots showing increasing AF for BCL2 (d) and MYC (e) structural variants with transformation from FL to DHL (n = 8 patients). The MYC translocation in P4 was detected in both FL and DHL. P values were calculated by one-sided Wilcoxon signed-rank tests. The boxplots display 25th and 75th percentiles and the median of each group. Whiskers represent the highest and lowest values within 1.5× interquartile range. f, Comparison of mutational burden of FL compared to DHL in relation to the TSS, normalized and averaged over the cohort. A substantial concentration of mutations was observed within a 1,000 bp proximity of the TSS particularly within DHL. Only samples for which a nontumor DNA sequence was available are included in this analysis. AF, allele frequency.
DHL-specific mutations occur within SE-embedded promoters
In mouse B cells, AID-associated chromosomal translocations occur at promoters and inside gene bodies15. By analyzing paired patient samples by WGS, we find that FL and DHL harbor mutations in both coding and noncoding sequences (Figs. 1f and 2a), with a higher coding-region mutational burden observed in DHL relative to FL (analyses in Fig. 2b). Many mutations are observed in known B-NHL oncogenes including KMT2D, CREBBP, TNFRSF14, TP53, CCND3, EZH2, MED12 and SF3B1 (Fig. 2c)16,24,32,33,34. In addition to coding-region mutations, numerous mutations are observed at noncoding DNA sequences (often intragenic but some intergenic). Strikingly, many mutations acquired upon transformation to DHL cluster specifically within 2 kb of the transcription start sites (TSS) of genes (Fig. 1f), including several previously found to be mutated in B cell lymphomas (Fig. 2d). Many of these mutations occur within noncoding sequences, often in the first intron of genes known to undergo AID-mediated aSHM (for example, aSHM at MYC and BCL2 loci; Fig. 1c). In addition to single-nucleotide variants, we observe recurrent copy number gains at the ZCCHC7/PAX5 and MDM2 loci and recurrent losses at CDKN2A/B in DHL (Fig. 3a,b). Copy number gains at the ZCCHC7/PAX5 locus are surprisingly recurrent, acquired upon transformation to DHL in 6 of 8 patients (Figs. 2b and 3c).
a, Percentages of somatic point mutations observed in coding and noncoding regions within FL and DHL (left) and breakdown of genic mutations observed (right), showing more frequent coding region and 5′ UTR mutations in DHL relative to FL. The P value was calculated by the two-sided chi-squared test. b, Summary of alterations in coding and noncoding regions surrounding highlighted genic sequences observed in the cohort, including single-nucleotide variants, short indels, copy number variants and structural variants. Genes included in this figure are selected based on mutation frequency, mutation impact, COSMIC annotation, OncoKB information and a published study of DLBCL genomes16. c, Lollipop plots of nonsynonymous mutations in well-recognized oncogenes and tumor suppressor genes. The asterisks indicate a translation stop codon. d, Noncoding mutations seen in the six patients that could be analyzed (due to availability of nontumor DNA sequences) showing consistently increasing mutational frequency in the noncoding (intronic or promoter associated) DNA sequences of each gene upon transformation.
a,b, Copy number changes, that is, losses and gains seen in the genomes of (a) FL and (b) DHL samples. The GISTIC plots show significant focal copy number variation (including recurrent CDKN2A/B loss, MDM2 gain and PAX5/ZCCHC7 gain) occurring during transformation to DHL. c, Copy number and single-nucleotide variants involving the PAX5/ZCCHC7 locus. Most cases demonstrate a gain of ZCCHC7 upon transformation to DHL (Fisher’s test P = 0.0023). TAD, CTCF, H3K4me1, H3K4me3 and H3K27ac ChIP–seq data overlap the PAX5/ZCCHC7 locus. The area of copy number gain of PAX5/ZCCHC7 mostly overlaps the entirety of the TAD. Transformation-associated aSHM at PAX5 is enriched in the first intron and overlaps with H3K4me3 and H3K27ac peaks (gray vertical box).
Super-enhancers (SEs) are regulatory regions that often control the expression of lineage-specific genes to activate rapid transcription during cell differentiation. Many important B cell lineage-defining genes such as AID, RAG1 and PAX5 are regulated by neighboring SE sequences35. aSHM within SE clusters has been observed in human lymphomas and mouse B cells10,11,13,35,36; however, the evolution of SE mutations during lymphoma progression, their recurrent occurrence at SE-embedded promoters (in contrast to some reports that they occur surrounding enhancers) and their effects on gene expression in lymphoma cells are incompletely understood. Evaluation of FL/DHL pairs demonstrates accumulation of mutation clusters with short intermutational distances. These mutational signatures are similar to the kataegis mutagenesis seen in tumors due to the mutator activity of the AID/APOBEC family of proteins (Fig. 4a)13,37,38. A substantial number of mutations acquired upon transformation to DHL (11.74%) are geographically clustered at H3K27ac-enriched sites in the B cell genome (Fig. 4a). The enrichment of H3K27ac marks defines these sites as enhancer/SEs. Many of the genes we find to have transformation-associated enhancer/SE mutations, such as BCL6, CIITA, IRF8 and ZFP36L1, have well-known mechanistic roles in B cell development and lymphomagenesis (Fig. 4b). The canonical immunoglobulin targets of SHM (IGH, IGL and IGK) also continue to acquire mutations during transformation to DHL (Fig. 4a,b). Transformation-associated noncoding mutations in the CIITA and IRF8 genes are embedded in H3K27ac-enriched regions of the genome and adjacent to the TSS in the first intron of both genes (Fig. 4c), and observed recurrently in 5/8 and 6/8 DHL, respectively. Other examples of transformation-associated point mutations overlapping an SE occur at the PIM1, RHOH and CXCR4 loci in 3/8, 4/8 and 5/8 tumors, respectively (Extended Data Fig. 2a–c). Noncoding RNAs, including miR-142, are also found to be recurrently mutated upon transformation to DHL (4/8 cases; Supplementary Fig. 2a–c).
a, Rainfall plots of the intermutational distances across the genome in FL and DHL demonstrate that the majority of the mutation-enriched clusters, some showing a kataegis pattern, overlap SE regions. Hypermutated genes in both FL and DHL are labeled in green, and DHL-specific hypermutated genes are labeled in magenta. b, Unbiased analysis highlighting genes enriched for acquired mutations (mutations occurring upon transformation to DHL). After separating the genome into 3 kb segments and comparing the somatic mutation profile for DHL versus FL in high mutation load segments with at least five mutations, 30 of 204 segments show significantly more mutations in DHL. All these segments overlap with SE (labeled in red) or enhancer (labeled in blue) regions. These regions were then labeled with the names of nearby genes. c, Examples of regions recurrently mutated upon FL to DHL transformation. Both CIITA and IRF8 show hypermutation clusters overlapping SE regions and close to the TSS. d, Normalized and averaged somatic mutation density of FL and DHL in relation to H3K27ac marks representing active enhancers, H3K4me3 marks representing active promoters near TSS and H3K4me1 marks representing primed enhancers. The three panels on the right show mutation density in H3K27ac regions outside H3K4me3 areas, H3K4me3 regions outside H3K27ac marks representing promoters outside enhancers and intersection areas of H3K27ac marks and H3K4me3 marks representing promoters present within active enhancer regions.
SEs contain both regulatory DNA sequences (enhancers) and promoters of genes10,39. We find that transformation-associated aSHM occurs predominantly at promoters within SEs. We find that aSHM is distributed within wide regions covered by H3K27ac marks (representing enhancers) and H3K4me1 marks (representing poised enhancers) and relatively more tightly around promoter marks (H3K4me3; Fig. 4d). Comparative analyses of mutation density in H3K27ac regions outside of H3K4me3 areas, H3K4me3 regions outside of those with H3K27ac marks and in areas representing the intersection of H3K27ac marks and H3K4me3 marks suggest that sequences surrounding or overlapping promoters embedded in H3K27ac-marked SEs are most frequently mutated during lymphoma progression (Fig. 4d and Extended Data Fig. 2d). Furthermore, the percentage of mutations overlapping H3K4me3 and H3K27ac regions increases at many important loci, including BCL2, IGH, BCL6, CIITA, BCL7A, DTX1 and PAX5/ZCCHC7, upon transformation of FL to DHL (Fig. 5a). Because topologically associating domains (TADs) contain both SEs and their target genes40, and because loop extrusion and genome architecture-related proteins are implicated in aSHM35, we investigated whether SE-resident aSHM clusters reside within TADs. At many loci, we find lymphoma transformation-associated aSHM cluster around boundaries of TADs containing known aSHM targets including DTX1 (Fig. 5b), BCL2 (Fig. 5c) and PAX5 (Fig. 5d). Genome-wide analysis of SE-associated aSHM locations shows that a high proportion of SE-associated aSHM observed upon transformation to DHL is located in compartment A, the spatial region of the genome containing open and active chromatin41 (Fig. 5e), consistent with the fact that a higher fraction of SEs is located in compartment A (Fig. 5f). aSHM tends to accumulate close to TAD boundaries where cohesin and CTCF proteins localize (Fig. 5g). Micro-insertions (4% of all somatic mutations) and microdeletions (6% of all somatic mutations) occur more frequently in H3K27ac-marked and H3K4me3-marked intersecting regions in DHL than in FL. The most frequent of these alterations occur at SEs of known aSHM targets, such as IGH, CIITA, PAX5, BCL6 and BCL2 (Extended Data Fig. 3a,b).
a, Plots demonstrating percentages of mutations overlapping H3K27ac peaks (left, n = 35,770) and H3K4me3 peaks (right, n = 25,740) in DHL compared to FL. Each pair of points represents an H3K4me3/H3K27ac peak region. The 15 most frequently mutated peaks are highlighted and labeled with the names of nearby genes. The P values were calculated by two-sided Wilcoxon signed-rank test. b,c, In DTX1 (b) and BCL2 (c) transformation-associated aSHM overlaps H3K4me3 and H3K27ac sites and occurs near the boundaries of TADs. d, In the PAX5/ZCCHC7 region, transformation-associated aSHM overlaps H3K4me3 and H3K27ac sites and occurs near the boundaries of the TAD. The increased read depth of the PAX5/ZCCHC7 regions in P2.DHL is shown. e, Occurrence of SE mutations in compartment A versus compartment B in FL (left, n = 7) and DHL (right, n = 6) tumors with each pair of dots connected by a gray line representing one lymphoma sample. The P values were calculated using two-sided Wilcoxon signed-rank tests. The horizontal line in each group represents its mean value, and the corresponding vertical line represents its s.d. bar. f, Fraction of SE regions in compartment A versus compartment B. The bar plot shows size fraction of SE and non-SE in compartment A and compartment B, respectively. The overall total size of SEs across the whole genome is approximately 37 Mb; 31 Mb are located within compartment A and represent 2.6% of the overall compartment A region. g, Normalized and averaged somatic mutation density of FL and DHL in relation to TAD boundaries. Starting from the center of every TAD boundary (gray dashed line), the left side represents regions inside of TADs and the right side represents areas outside of TADs. Intra-TAD boundaries are not considered in this analysis.
Different types of genetic alterations occur at the same locus
A large set of recurrently acquired copy number gains as well as some recurrently acquired losses (Fig. 3a,b), including loss of CDKN2A, which has been previously implicated in transformation of FL and other types of lymphomas24,42, occurs in the course of transformation to DHL. The recurrent transformation-associated copy number gain at the PAX5/ZCCHC7 region extends from the 5′ region of the PAX5 gene to the 3′ end of the neighboring ZCCHC7 gene. Additionally, several transformation-associated mutations, representing an aSHM hotspot, occur near the promoter of the PAX5 gene (Fig. 3c), overlapping an H3K4me3 peak and inside an H3K27ac cluster. Analyses of whole-exome sequencing and WGS data from other series of DLBCL34 and chronic lymphocytic leukemia43 also show mutations in the same region of the PAX5/ZCCHC7 locus, providing further evidence of recurrent aSHM events in this region in B-NHL (Extended Data Fig. 3c,d).
We next sought to understand how aSHM at PAX5/ZCCHC7 affects expression of one or both genes in DHL. A common set of enhancers has been found to control the expression of both ZCCHC7 and PAX5 (refs. 44,45), and we suspected that ‘enhancer retargeting (ER),’ whereby functional loss of a promoter results in subsequent preferential targeting of a different promoter35,45, could occur following aSHM at promoter regions during lymphoma progression. Supporting this possibility, we observe that PAX5 and ZCCHC7 gene promoters exist in the same TAD in mice (Extended Data Fig. 3e) and humans (Fig. 5d)35 and that PAX5 and ZCCHC7 expression is negatively correlated in DLBCL cell lines (Extended Data Fig. 4a). Bioinformatic prediction of ER was carried out using genomes of the 8 DHLs in our study and 39 published DLBCLs34 (Extended Data Fig. 4b). In total, 143 gene pairs from 25 hypermutated loci were investigated. We identified several gene pairs that might undergo ER, including the PAX5-ZCCHC7 locus (Fig. 6a), IKZF3-STARD3 locus, FAM102A-SLC25A25-AS1, BCL7A-B3GNT4 and others (Extended Data Fig. 4c–e). To evaluate potential ER in the PAX5/ZCCHC7 locus, we deleted the PAX5 promoter alternate TSS (PAX5-TSS2), which was found to be mutated in our DHL cohort (Extended Data Fig. 5a,b) in the SUDHL10 cell line, using CRISPR/Cas9 homology-directed repair/mutagenesis (HDR), and observed a resulting increase in ZCCHC7 mRNA expression (Extended Data Fig. 5c, left). Additionally, several DLBCL cell lines and primary human DLBCLs harbor PAX5-TSS2 mutations (Extended Data Fig. 5d and Supplementary Table 4) and on average demonstrate increased ZCCHC7 mRNA levels relative to cell lines and tumors without such mutations (Extended Data Fig. 5e,f), suggesting that PAX5 promoter mutations might promote ZCCHC7 overexpression in lymphoma cells. In 4C assays using two baits within the ZCCHC7 locus, the enhancer regions located close to PAX5-TSS2 (Extended Data Fig. 6a) and within the PAX5/ZCCHC7 SE (sites 2, 3 and 4) show stronger interaction with the ZCCHC7 gene following deletion of the PAX5-TSS2 (ΔPAX5-TSS2) region (Extended Data Fig. 6a–c). Next, we incorporated the recurrent PAX5-TSS2 mutations (Chr9:37,026,299–37,026,327:GC to AT conversion labeled as PAX5-TSS2mut; Extended Data Fig. 6d) into the SUDHL10 cell line using homology-directed genome editing46 and found that the enhancer regions (sites 1, 2 and 3) in the PAX5-ZCCHC7 SE cluster interact more efficiently with the ZCCHC7 promoter (Fig. 6b,c), compared to unmutated cells. Comparison of 4C assays performed with ΔPAX5-TSS2 and PAX5-TSS2mut identified overlapping interaction regions in the PAX5-ZCCHC7 SE that loop to the ZCCHC7 promoter. Given that both deletion and point mutation of PAX5-TSS2 lead to stronger interactions between the PAX5 enhancer region and the ZCCHC7 promoter and a corresponding increase in ZCCHC7 mRNA expression (Extended Data Fig. 5c), we postulate that PAX5 promoter mutations in lymphoma cells increase ZCCHC7 expression. Notably, this mechanism could explain the ZCCHC7 overexpression that we observe in lymphomas that do not harbor PAX5/ZCCHC7 copy number gains but have PAX5-ZCCHC7 SE mutations. Consistently, we found several DLBCL cell lines and primary human DLBCLs to harbor PAX5-TSS2 mutations (Extended Data Fig. 5d).
a, Distribution of Spearman’s correlation coefficient between RNA expression of two genes located within the same TAD. In total, 19,676 gene pairs were investigated by analyzing transcriptomic data of 11 DLBCL cell lines to identify candidates of ER. The ZCCHC7–PAX5 pair is highlighted. b, The 4C-seq experiments from the ZCCHC7 promoter measuring the interaction with H3K27ac+ H3K4me3+ peaks, including those at the PAX5 promoter. In SUDHL10 cells where the PAX5-TSS2 has been replaced with a mutation identified in DHL ‘chr9: 37,026,315–37,026,316 GC->AT’ (mutated region shown with orange arrow; SUDHL10 PAX5-TSS2mut), the PAX5 intragenic enhancer regions (sites 1, 2 and 3) interact with the promoter of ZCCHC7 more strongly than in SUDHL10 cells without the PAX5-TSS2 mutation (see thickness of black lines). c, Normalized coverage of the 4C interactions for each 4C-seq viewpoint is shown. The TSS2 mutant region does not overlap the ZCCHC7 interaction domain of the PAX5 promoter. Data are presented as mean values ± s.d. Two-sided Welch’s t test was used; n = 3 for each group. d, Tissue microarrays containing triplicate samples of 33 DLBCL and 9 DHL with controls including benign spleen, lymph node, tonsil and low-grade FL were stained by immunohistochemistry for ZCCHC7. ZCCHC7 was visualized predominantly in nucleoli and sometimes in the nuclei. Scale bar, 20 µm. Examples of stained tumors and controls are shown. e, The DHL cell line SUDHL6 was separated on denaturing acrylamide gels and analyzed by northern blotting. CD77+ tonsillar B cells were used as controls. As additional controls, and to unambiguously map the various 5′ end extended forms of 5.8S, HeLa cells depleted of RNA exosome subunits (EXOSC2 and EXOSC10) or of a cofactor (SKIV2L2) are used (left). HeLa cells depleted of (− lane), or overexpressing (+ lane), ZCCHC7 were also included. Results represent two or more independently performed biological replicates. f, Total RNA extracted from HeLa cells, SUDHL10 cells overexpressing ZCCHC7 (SUDHL10-OE), parental SUDHL10 (WT), PAX5-TSS2mut (SUDHL10 cells with PAX5 promoter DHL mutation) and two sets of PAX5-TSS2mut transfected with Exosc10 were processed as in e. Equal loading is shown with the U1 RNA control (bottom). Results represent two independently performed experiments. RPM, reads per million.
As in humans, the mouse Zcchc7 gene resides within a TAD (Extended Data Fig. 3e). Translocation capture sequencing experiments identify AID-induced translocations overlapping the Pax5 promoter, similar to our observations in the human PAX5-TSS2, showing that this sequence accumulates AID-mediated genomic alterations in both human and mouse genomes. Furthermore, RNA-seq experiments performed in mouse B cells show sense/antisense transcription overlapping the Pax5 promoter region and deletion of RNA exosome activity (DIS3C/C B cells) leading to accumulation of these sense/antisense transcripts, a mechanism resulting in increased AID-mediated mutagenesis12,47,48. DNA/RNA hybrid immunoprecipitation followed by high-throughput sequencing demonstrates accumulation of R-loops at this site, and therefore generation of AID substrates. H3K27ac chromatin immunoprecipitation followed by sequencing (ChIP–seq) identifies the overlapping SE region (Extended Data Fig. 3e). In summary, as AID-mediated SHM occurs within SEs that are marked with overlapping sense/antisense transcription, and preferentially targets non-B DNA regions5,10,11,12,13,14, our findings support AID-mediated aSHM at the PAX5 promoter sequence. Many of the PAX5 promoter mutations observed in our cohort demonstrate the classical AID mutational signature (Extended Data Fig. 6d)49. Furthermore, sequencing of DHL samples has shown that ZCCHC7 represents a frequent translocation partner of MYC50,51 (such a translocation was also identified in one of our DHL samples separate from the longitudinal cohort, patient 11; Extended Data Fig. 7a). Additionally, in individual cell lines and tumors, ZCCHC7 translocates with PVT1 (Extended Data Fig. 7b,c), underscoring the susceptibility of this locus to AID-mediated genomic alterations.
PAX5-ZCCHC7 structural variations are also observed in a subset of B-lymphoblastic leukemia (B-ALL), a neoplasm that demonstrates robust ZCCHC7 expression overall (Extended Data Fig. 7d), and especially in treatment-refractory cases, with PAX5-ZCCHC7 positive cases showing the highest levels of ZCCHC7 mRNA expression (Extended Data Fig. 7e). It is likely that in B-ALL this PAX5-ZCCHC7 fusion is caused by a different mechanism driven by the RAG enzymes, although AID could also have a role52,53.
PAX5/ZCCHC7 alterations affect pre-rRNA processing
To investigate the degree to which FL to DHL transformation alters ZCCHC7 expression, we performed immunohistochemistry for ZCCHC7 using tissue microarrays (TMAs) containing 33 DLBCLs and 9 DHL samples and using pairs of lymphomas representing FL transformation to DLBCL from nine patients (these stained pairs were distinct from our sequenced DHL cohort; Extended Data Fig. 8a,b). We find that ZCCHC7 is more highly expressed in most DHL and DLBCL samples compared to benign lymphoid tissue and FL and that expression usually increases upon transformation to DLBCL (Fig. 6d and Extended Data Fig. 8a,b). ZCCHC7 is observed predominantly within the NPM1-stained nucleolus of human, nonneoplastic CL-01 B cells (Extended Data Fig. 8c), but this nucleolar localization is perturbed in SUDHL6 DHL cells that harbor a ZCCHC7-MYC translocation and show diffuse nuclear staining in addition to nucleolar staining (Extended Data Fig. 8c; nucleolar/nucleoplasmic ZCCHC7 distribution shown in Extended Data Fig. 8d). These observations raise the possibility of nucleolar dysregulation with changes in rRNA processing and ribosome biogenesis. The formation of the 3′ end of 5.8S rRNA is a complex multistep process involving sequential exoribonucleolytic digestion of internal transcribed spacer 2 (ITS2) by the RNA exosome (Extended Data Fig. 8e). Multiple subunits of the exosome are involved, with the trimmed precursors ‘handed over’ from one subunit to another, aided by cofactors54,55,56. The yeast Trf4/5-Air1/2-Mtr4 polyadenylation (TRAMP) complex participates in RNA degradation via interactions with the exosome57. ZCCHC7 potentially represents part of a human TRAMP-like complex, interacting with the noncanonical poly(A) polymerase, PAPD5 and the RNA helicase, MTR4 (Extended Data Fig. 8f) with RNA threading through the complex as observed in an AlphaFold58 reconstituted model (Extended Data Fig. 8g). We therefore hypothesized that ZCCHC7 could regulate the nucleolar function of the RNA exosome and also function independently in rRNA processing with Exosc10 (refs. 55,59,60). We were therefore interested to learn if pre-rRNA processing and ITS2 maturation, in particular, are altered in lymphoma cells upon ZCCHC7 overexpression. Total RNA extracted from DHL and control cells was analyzed by high-resolution northern blotting with probes detecting major 3′ extended precursor forms of 5.8S rRNA (Fig. 6e). The 5.8S + 40 is a normal pre-rRNA precursor, which is converted into 5.8S by the RNA exosome subunit EXOSC10 (ref. 60). The ZCCHC7-overexpressing SUDHL6 DHL cell line displays strikingly elevated levels of 5.8S + 40 relative to control CD77+ human tonsillar B cells (Fig. 6e, right). On the other hand, HeLa cells, used as an additional control, also do not show elevated levels of 5.8S + 40 (Fig. 6e, right). However, when ZCCHC7 is overexpressed in HeLa cells, 5.8S + 40 shows a marked increase similar to that observed in DHL cell lines (Fig. 6e). Specific silencers that deplete ZCCHC7 do not greatly affect 5.8S + 40 levels in HeLa cells (Fig. 6e, lane 2, right). The effect of ZCCHC7 overexpression on 5.8S + 40 rRNA processing in Hela cells or in SUDHL6 is specific to this step of rRNA processing as no major changes are observed at other steps of rRNA processing; in particular, production of mature 18S and 28S rRNA is not affected (Extended Data Fig. 9a–c). Thus, the overexpression of ZCCHC7 appears to particularly impact ITS2 processing, resulting in the accumulation of 5.8S + 40 rRNA. Next, we checked whether PAX5-TSS2mut incorporated in SUDHL10 lymphoma cells, using Cas9-driven HDR, would cause 5.8S + 40 rRNA accumulation. We find that unmutated SUDHL10 cells have an inherent level of 5.8S + 40 rRNA, but as seen either following ZCCHC7 overexpression (Fig. 6f, lane 2) or following targeted incorporation of PAX5-TSS2 mutations (Fig. 6f, lane 4) there is a significant increase in the 5.8S + 40 rRNA level. Maturation of 5.8S + 40 rRNA occurs via the 3′ end RNA trimming activity of Exosc10, whose deletion also results in accumulation of 5.8S + 40 rRNA60 (Fig. 6e). Overexpression of Exosc10 in PAX5-TSS2mut cells leads to rescue of 5.8S rRNA processing (Fig. 6f and Extended Data Fig. 9d). Taken together, ZCCHC7-mediated alterations of pre-rRNA processing kinetics may lead to production of a distinct ribosomal population, with repercussions for protein synthesis, efficiency and fidelity.
PAX5/ZCCHC7 alterations affect nascent protein synthesis
Next, we wanted to investigate potential changes in nascent protein synthesis caused by the PAX5-TSS2 mutation. We performed O-propargyl puromycin-mediated identification (OPP-ID)61 of proteomic changes occurring within a short 2-h pulse of OPP. Briefly, in this assay, OPP permeates cells and labels nascent elongating polypeptides, which are then captured by click chemistry on streptavidin beads to allow for the identification of global changes in active/nascent protein translation (Fig. 7a). We performed three replicates of OPP-ID in SUDHL10 cells and those overexpressing ZCCHC7 (Fig. 7b) or with PAX5-TSS2mut incorporation (Fig. 7c; robustness highlighted in PCA plot in Fig. 7d, with detailed OPP proteomics source data provided in Supplementary Table 5). We observe that de novo translation of many proteins is suppressed or stimulated as a consequence of the DHL-associated PAX5 promoter mutation and ensuing ER as well as upon overexpression of ZCCHC7, and that many alterations in protein translation are common to these two conditions (Fig. 7e). Heatmaps demonstrating groups of proteins with altered synthesis and associated gene ontology enrichment analyses are provided in Extended Data Fig. 10a–c, suggesting effects on protein synthesis, DNA damage/mutational repair and processing, as well as other cellular mechanisms. Indeed, a group of oncogenes are translated rapidly (Fig. 7e) while the translation of several tumor suppressors is attenuated (Fig. 7e), including IKZF3 (which was also found to be decreased in steady-state protein level analyses (Extended Data Fig. 10d)). Using polysome analyses (Fig. 7f and Extended Data Fig. 10e), to fractionate mRNAs based on their translational efficiency62, we confirm that IKZF3 mRNA indeed has lower translational efficiency in both PAX5-TSS2mut lymphoma cells (Fig. 7g) and in ZCCHC7-OE lymphoma cells (Extended Data Fig. 10f,g), compared to parental SUDHL10 lymphoma cells. Thus, we postulate that increased ZCCHC7 expression in lymphoma cells results in altered kinetics of 5.8S rRNA biogenesis, driving changes in protein synthesis and consequently remodeling of the lymphoma proteome. Interestingly, some of the nascent polypeptide changes seen in PAX5-TSS2 mutant cells via the OPP-ID translation assay identify molecular targets of currently available therapeutics (Extended Data Fig. 10h)
a, Schematic of nascent polypeptide synthesis analyses using OPP-ID61 (created with BioRender.com). b,c, Volcano plots showing differences in abundance of OPP+ proteins, in SUDHL10 cells with ZCCHC7-OE (b) or PAX5-TSS2mut (c), compared to the SUDHL10 WT cells. The proteins that are significantly upregulated or downregulated in ZCCHC7-OE or PAX5-TSS2mut are highlighted in red or blue, respectively. d, Principal component analysis plot of the proteomics data from SUDHL10 WT cells (no inherent mutation or alterations in ZCCHC7 or PAX5-TSS2), ZCCHC7-OE and PAX5-TSS2mut, with six biological replicates per group. Three replicates each were cultured with or without OPP (OPP+ and OPP−, respectively). Protein signals detected in the OPP+ group are shown as the union of OPP-labeled nascent proteins and nonspecifically captured background proteins. The OPP− data, representing protein signals from nonspecifically captured background proteins, were used as control to distinguish nascent proteins from background proteins. e, Two-way ANOVA test to evaluate the effect of genetic alteration (ZCCHC7-OE or PAX5-TSS2mut) and OPP label on proteomic intensity. The plot shows FDR adjusted P values of the two factors (x axis for genetic alteration and y axis for OPP label) in the two-way ANOVA test. Each dot represents a protein and shows the abundance of corresponding proteins that decreased (left side) or increased (right side) in the ZCCHC7-OE and PAX5-TSS2mut cells. The upregulated/downregulated proteins were chosen by setting the adjusted genetic alteration P value < 0.01. Oncoproteins that increased and tumor suppressor proteins (TSPs) that decreased in SUDHL10 ZCCHC7-OE and SUDHL10 PAX5-TSS2mut cells are highlighted in red and blue, respectively. Two-sided Fisher’s exact tests were conducted to evaluate the relationship between oncoproteins/TSPs and upregulation/downregulation. The results show significant association of TSPs and downregulation with P = 0.0016. f,g, Polysome analyses of SUDHL10 cells and SUDHL10 PAX5-TSS2mut cells. The distribution of IKZF3 mRNA (g) to heavier (highly translated fractions) decreases following PAX5-TSS2 mutation. Gross proteome-wide translational perturbation is not observed in PAX5-TSS2mut cells based on comparable polysome profiles (f), but specific mRNAs (for example, IKZF3) are translated differently (g). The data represent conclusions drawn from two independent biological replicates. ANOVA, analysis of variance. FDR, false discovery rate.
Discussion
The critical availability of multiple patient-derived paired longitudinal samples representing the transformation of FL to DHL provided a unique opportunity to answer an important question about the timing of SE mutations that occur in DLBCL/DHL. Furthermore, we demonstrated the consequences of these changes with respect to lymphoma biology. Our findings clarify the location of aSHM occurring in lymphoma transformation, with most of the observed mutations occurring near promoter-proximal H3K4me3-marked regions of lymphoma-related genes embedded in clusters of H3K4me1- and H3K27ac-marked SEs. The presence of mutations in promoter regions led us to perform studies that ultimately revealed a mechanism by which mutations in noncoding gene regulatory elements, such as SEs and associated promoters, can alter gene expression in lymphomas in unexpected ways, for example, via changes in the transcription of neighboring genes as opposed to simply altering transcription of cognate genes as expected. We demonstrate this phenomenon at the PAX5/ZCCHC7 locus; however, it is possible that AID-mediated mutations of other gene regulatory elements will lead to similar outcomes.
In addition to highlighting other consequential AID-mediated mutations in gene regulatory elements, future studies might further clarify the mechanisms and parameters causing AID-mediated mutations specifically to these non-Ig regions of the lymphoma genome. Of note, aSHM mutational sites do not always directly overlap H3K27ac sites, thus making it unlikely that these marks or the associated enhancers are sufficient for recruitment of AID during progression to DHL. Majority of aSHM sites are situated ±2 kb from genic TSSs present inside H3K4me3 hubs, indicating divergently transcribed promoters (with genic TSS or alternative genic TSS) embedded in SEs, and thus promoter-associated transcriptional, chromatin and local DNA topological properties are important for AID recruitment. In addition, properties of TADs and SEs surrounding promoters are likely to be responsible for the recruitment of AID protein and/or AID cofactors at aSHM sites.
At the FL stage, SE mutations occur at low frequencies, whereas at the DHL stage, the frequencies are much higher. Phylogenetic analyses of aSHM in our FL/DHL paired samples indicate divergent evolution of FL and DHL from a common progenitor cell24,63. Given that the sequenced FL and DHL are clonally related, our findings support a role for SE-associated aSHM predominantly occurring during progression from a shared or common progenitor to DHL.
Although PAX5 has a well-established integral role in B cell development and lymphomagenesis, the adjacent gene, ZCCHC7, has not been the subject of the same attention in B-NHL. In our longitudinal cohort, we find that ZCCHC7 overexpression can be altered by copy number gain and/or ER. We also sequenced DHL cases with high ZCCHC7 expression based on immunohistochemical (IHC) staining and identified one DHL (P11) harboring MYC-LPP and BCL2-IGH translocations as well as an MYC/PVT1-ZCCHC7 translocation. This translocation has been reported previously in a subset of DHL50 and is found in the SUDHL6 cell line that overexpresses ZCCHC7 mRNA. Taken together, our results provide evidence of the following three different mechanisms of ZCCHC7 overexpression in DHL: copy number gain, ER and translocation, and a potential role for ZCCHC7 overexpression in lymphoma cell survival. It remains to be determined whether the identified mechanisms underlie transformation of FL in general and if they also foster the progression of other types of low-grade B cell lymphomas.
Finally, we reveal the cellular function of ZCCHC7 and the consequences of ZCCHC7 alteration in lymphoma cells. ZCCHC7 overexpression via copy number gain or via ER interferes with normal 5.8S rRNA processing in DHL, with the resultant accumulation of 5.8S + 40 pre-rRNA leading to critical rewiring of the lymphoma proteome. We also provide a mechanistic role for ZCCHC7 in 5.8S rRNA processing, potentially acting as a part of the human TRAMP-like complex and thereby regulating the 3′–5′ exo-RNase Exosc10 in 5.8S rRNA processing. In addition to direct perturbation of homeostasis of tumor suppressors and/or oncogenes, alterations in rRNA biogenesis might cause translational stress that in turn might impact lymphomagenesis/lymphoma transformation (as has been previously implicated in the development of other malignancies64,65). We expect that in addition to the role of ZCCHC7, other mechanisms leading to rRNA processing defects in lymphoma likely will be identified in the future, presenting potential opportunities for the treatment of aggressive lymphomas and/or secondary prevention of lymphoma transformation.
In summary, our study highlights mechanisms by which gene and protein expression are broadly altered in lymphoma, via ER and effects on protein synthesis due to aberrant ribosome biogenesis, respectively (summarized schematically in Fig. 8). Future work will be required to understand the mechanisms through which AID is targeted to gene regulatory elements during lymphoma progression and to identify other cellular processes affected by ER in lymphoma cells. Finally, ZCCHC7-related changes to ribosome biogenesis present a mechanism through which the lymphoma proteome changes over time, with important implications for lymphoma biology and clinical management.
A schematic representation of the contribution of aSHM to transformation from FL to DHL, along with various mechanisms of PAX5/ZCCHC7 locus alterations that lead to aberrant rRNA processing and rewiring of lymphomagenic protein synthesis (created with BioRender.com).
Methods
Case selection
The study was performed according to the principles of the Declaration of Helsinki and in compliance with protocols approved by the Institutional Review Boards (IRB) of Columbia University and the University of Pittsburgh. For samples that underwent WGS, departmental databases at CUIMC and UPMC were searched for longitudinal specimens representing DHL preceded by FL, diagnosed in the same individual and during the past 17 years. Samples from eight patients (four female and four male, ages 55–79 years, median 68 years), including formalin-fixed paraffin-embedded (FFPE) tissue (18 samples) or archival DNA extracted from fresh tissue for purposes of clinical molecular testing (five samples), were retrieved for WGS studies. When available, nontumor tissue or DNA from the same patients also was identified and retrieved from pathology archives (6/8 patients). Additionally, fresh tumor tissue from the two unpaired DHL found to demonstrate the highest degree of ZCCHC7 expression based on IHC staining of TMAs (described below) was retrieved for WGS. For samples used in TMA creation and immunohistochemistry, the pathology department database at CUIMC was searched for specimens representing FL, DLBCL and DHL, including paired specimens representing longitudinal FL/DLBCL samples from the same individual, and diagnosed over the past 15 years. All selected cases fulfilled the morphologic, immunophenotypic and cytogenetic features of FL, DLBCL and high-grade B cell lymphoma with MYC and BCL2 rearrangements according to the current WHO classification66. Clinical information and laboratory data for each individual were obtained through review of electronic health records. The study was carried out using de-identified, residual banked and archival tissue/nucleic acids originally collected for clinical diagnostic purposes and remaining after completion of diagnostic work-up. The requirement for informed consent was waived, as approved by the IRB, because the specimens used in this study are from patients who were diagnosed with an aggressive disease associated with a high mortality rate, many of whom were deceased at the time of the study, and whose residual diagnostic specimens were used retrospectively years after diagnosis. Obtaining informed consent would not be practical. Furthermore, no intervention was performed, and the study involved no more than minimal risk to the subjects.
Creation of TMAs
Hematoxylin and eosin-stained sections derived from the paraffin blocks containing DLBCL and DHL were examined, and representative areas of interest were identified. Sector maps were designed using Microsoft Excel spreadsheets to identify the location of each specimen on the array blocks. Specifically, 33 DLBCL (15 female and 18 male, ages 12–85 years, median 71 years), 9 DHL (six female and three male, ages 42–76 years, median 66 years) and one each of low-grade FL (68 years, female), benign lymph node, tonsil and spleen were sampled in triplicate using 2 mm tissue cores. The DHL cases in the TMA and stained by IHC were distinct from those sequenced in the longitudinal cohort; however, a DHL in the TMA that showed high expression of ZCCHC7 based on IHC, from patient 11 (Extended Data Fig. 8a), was subsequently sequenced and analyzed separately from the main cohort. Each of the three lymphoma TMA blocks created included at least two samples (in triplicate) of benign lymphoid tissue controls. TMAs were created in the Experimental Molecular Pathology Core Facility of the Herbert Irving Comprehensive Cancer Center of Columbia University using a Beecher Instruments Manual Tissue Arrayer. Multiple four-micron sections from the tissue array blocks were cut and placed on charged polylysine-coated slides. These sections were used for IHC staining.
IHC staining and scoring of slides
IHC staining of TMA and lymphoma tissue sections was performed using standard methods with a polyclonal antibody against ZCCHC7 (Novus Biologicals, NBP1-89175). Stained sections were evaluated by a hematopathologist (R.J.L.-N.) and assigned an IHC H score. H score = 3 × percentage of strongly positive cells + 2 × percentage of moderately positive cells + 1 × percentage of weakly positive cells.
WGS
DNA was extracted from tissue samples using the QIAamp mini kit or the QIAamp FFPE kit (Qiagen) according to the manufacturer’s instructions. Samples underwent library preparation and sequencing on the Illumina Hi-Seq (8/10 patients, 19/26 samples) or DNBseq (2/10 patients, 7/26 samples) platform. Briefly, libraries sequenced on the Illumina platform were prepared using the NEBNext Ultra DNA Library Prep Kit according to the manufacturer’s recommendations. Libraries were sequenced on the Illumina HiSeq using 2× 150 bp paired-end configuration. Image analysis and base calling were carried out using HiSeq Control Software. Libraries sequenced on the DNBSeq by BGI technology were prepared using BGI’s in-house library preparation kit according to the manufacturer’s instructions. Libraries were sequenced on the BGISEQ-G400 platform.
Antibodies
Antibodies (dilutions in parentheses) for western blot (WB) and indirect immunofluorescence (IF) were as follows: rabbit polyclonal anti-ZCCHC7 (Novus Biologicals, NBP1-89175; 1:500 for IF), rabbit polyclonal anti-ZCCHC7 (ABclonal, A28251; 1:1,000 for WB), mouse monoclonal anti-GAPDH (Proteintech, 60004-1-Ig; 1:20,000 for WB), Alexa Fluor 647 donkey anti-mouse IgG (Thermo Fisher Scientific, A-31571; secondary 1:500 for IF), Alexa Fluor 488 donkey anti-rabbit IgG (Thermo Fisher Scientific, A-21206; secondary 1:500 for IF), IRDye 800CW donkey anti-Mouse IgG (Licor, 926-32212; secondary 1:10,000 for WB) and IRDye 680RD donkey anti-rabbit IgG (Licor, 926-68073; secondary 1:10,000 for WB).
Indirect IF
In total, 500,000 cells of each cell type (SUDHL6 obtained from American Type Culture Collection (ATCC), CRL-2959 and CL-01 obtained from Novus Biologicals, NBP1-49595) were incubated in PBS at 37 °C on poly-l-lysine-coated coverslips before fixation in 4% paraformaldehyde in PBS for 20 min, permeabilization with 1% Triton X-100 in PBS for 5 min and blocking with 1% powdered milk in PBS (IF-blocking buffer) for 60 min. The cells were then incubated for 2 h with primary antibodies in IF-blocking buffer, washed and incubated for 1 h with secondary antibodies in IF-blocking buffer in the dark. This was followed by washing and nuclear staining with 4′,6-diamidino-2-phenylidone (1 µg ml−1 in PBS). Coverslips were mounted on glass slides using ProLong Diamond Antifade Mount (Thermo Fisher Scientific, P36962). Spinning-disk confocal microscopy was performed on a Nikon TiE Eclipse inverted microscope (Nikon) equipped with a CSU-X1 spinning-disk unit (Yokogawa) and controlled with NIS Elements software (Nikon). A ×100/1.45 Plan Apo Lambda objective lens was used (Nikon). Fluorescence was excited with lasers at 405, 488, 561 and 647 nm, and emission was collected through standard filters for blue, green, red and far-red fluorophores. Z-stack images in 200-nm steps were acquired with a Zyla 4.2 CMOS camera (Andor Technology). Maximum projections were generated using ImageJ (National Institutes of Health (NIH)). Images for figures were cropped and adjusted using Photoshop (Adobe). To compare the different cell images, all images within the same panels and of the same antigens were acquired and adjusted identically.
Generation of Exosc10-overexpressing SUDHL10 cells
SUDHL10 cells (CRL-2963) were purchased from the ATCC, cultured in the laboratory and modified through HDR to form the SUDHL10 PAX5-TSSmut cell line, as described elsewhere in this paper. The HDR cells were transduced with a plasmid purchased from GeneCopoeia expressing human Exosc10 (transcript variant 1)-IRES2-mCherry-IRES-puromycin (EX-G0202-Lv213) or the empty vector without the Exosc10 transcript (EX-NEG-Lv213) using a Lonza 4D nucleofector. Forty-eight hours post-transfection, the cells were analyzed on a Becton Dickinson (BD Biosciences) SORP FACSAria running BD FACSDiva software in the Columbia University Stem Cell Initiative Flow Cytometry Core Facility. Forward scatter (FSC-A) and side scatter (SSC-A) were used to gate on live cells, followed by gating on singlets through forward scatter area relative to forward scatter height (FSC-A versus FSC-H). A distinct population of mCherry-expressing cells was observed, described as ‘Exosc10-OE’ in Figs. 6f and 7b–e and Extended Data Figs. 9d and 10a–g for those receiving the Exosc10-expressing plasmid. These cells then were subjected to western blot.
Pre-rRNA processing analysis in lymphoma cells
Five microliters of RNA were separated by migration on high-resolution denaturing acrylamide (for low-molecular-weight RNA analysis) or agarose (for high-molecular-weight RNA analysis) gels, and the gels were transferred to a nylon membrane and probed with radioactively labeled oligonucleotide probes (see ref. 60 for details). Mature rRNAs were visualized by ethidium bromide staining and northern blot probing. The probe sequences are shown in Supplementary Table 10.
Transcripts of interest were depleted for 3 d in HeLa cells (obtained from ATCC, CCL-2) transfected by RNAiMax (Invitrogen, 13778030) with silencers (used at 10 nM final concentration). ZCCHC7 was overexpressed following transfection with construct CMVp-hZCCHC7-Flag-IRES-GFP. The residual levels of ZCCHC7 after depletion (none was detected) or overexpression were established by western blotting with a specific antibody (ABClonal, A18251). As loading control, blots were probed for β-actin (Santa Cruz Biotechnology, SC-69879).
4C-seq protocol
To investigate chromatin interaction patterns of the PAX5 promoter region, we analyzed the Hi-C heatmap from published Hi-C (GSE63525_GM12878_insitu_DpnII67) and HiChIP (GSE80820 (ref. 68)) data from the human B lymphocyte cell line, GM12878. We focused on two interacting regions around the ZCCHC7 promoter locus that were found to strongly interact with the PAX5 promoter. We selected one of the two strongly interacting regions for use as a bait after evaluating the efficiency of the primer sets available for 4C studies.
4C-seq47,69 was carried out as follows: after crosslinking of 107 cells with 2% formaldehyde, HindIII (NEB, R3104S) was added to extracted nuclei, followed by overnight incubation, heat inactivation and washing with 1× T4 DNA ligase buffer (NEB, M0202S). The samples were resuspended in a 1.2 ml ligation mix (1× T4 DNA ligase, 1× BSA, 50 µl 20% Triton X-100, 5 µl T4 DNA ligase, H2O to 1.2 ml) and allowed to incubate at room temperature while rotating for 4–5 h, with addition of more ligase after 2 h. Reverse crosslinking was performed by adding 15 µl Proteinase K (Viagen Biotech, 501PK; 20 mg ml−1), followed by an overnight 65 °C incubation in phenol:chloroform:isoamyl alcohol (Sigma-Aldrich, P2069-100ML). The precipitated DNA was resuspended in 450 µl H2O. Samples were then digested with 50U DpnII (NEB, R0543S), followed by heat inactivation and purification with phenol:chloroform:isoamyl alcohol and resuspension in ~120 µl H2O. 4C PCR was carried out using Phusion High-Fidelity DNA Polymerase (NEB, M0530S) and two sets of viewpoint primers containing the Illumina sequencing adaptor sequences (see sequences in Supplementary Table 10). PCR parameters were as follows: 98 °C for 30 s, 16 cycles (98 °C for 10 s, 60 °C for 30 s, 72 °C for 2 min), 72 °C for 10 min. The products were cleaned with 0.8× Ampure XP beads (Beckman Coulter, A63880) as indicated in the referenced protocol. Addition of indexes and enrichment of adaptors containing first PCR product were carried out according to the manufacturer’s instructions (NEB, E7600/E7645). After quality control using the Bioanalyzer DNA HS chip, libraries were pooled and loaded on the Illumina MiniSeq using the MiniSeq High Output Reagent Kit (Illumina, FC-420-1002; 150-cycles). Both SUDHL10 PAX5-TSS2 deletion and point mutation clones were compared to control SUDHL10 using the same 4C-seq protocol.
Preprocessing of WGS data
Fastp (v0.23) was used in preprocessing FASTQ files. Low-quality reads containing over 40% poor-quality bases (base quality < 15) were filtered out. After removing adapters, clean reads with at least 45 bp were used for subsequent analyses. These clean reads were then mapped to hg38 and sorted by coordinates using BWA (v0.7.15) and SAMtools (v1.2), respectively. Picard (v2.23.9) was used to mark duplicates and generate bam files for mutation identification.
Somatic mutation calling in longitudinal samples
Somatic mutations in tumor samples were identified with the mutation caller SAVI (v2.0)70. For patient samples with available nontumor DNA, all primary and recurrent tumor samples were simultaneously compared to matched nontumor DNA for the identification of candidate somatic mutations. Variants exclusively identified in tumor samples (that is, those with P < 10−6 by the empirical Bayesian method and with <2 supporting reads in nontumor samples) were then annotated using multiple databases including dbSNP (https://www.ncbi.nlm.nih.gov/snp/), gnomAD (https://gnomad.broadinstitute.org/) and TOPMed (https://topmed.nhlbi.nih.gov/). After eliminating known common SNPs, the remaining candidates were subsequently filtered by eliminating (1) variants in low-complexity regions; (2) variants supported by <5 high-quality reads or supported by reads with strong strand biases; (3) variants with low allele frequencies (<10% for P2-FL and P8A-FL due to low tumor purity, <20% for others); (4) variants with total read depth of <15 in nontumor DNA; (5) variants only supported by the edge of reads (the distance from variants to all supporting reads’ 5′/3′ <25% of the read length). For patients without associated nontumor samples (P4 and P5) only mutations that differed between FL and DHL were used in analyses related to somatic mutations. For the same samples, known hotspot mutations found in the COSMIC (https://cancer.sanger.ac.uk/cosmic) database were listed as potentially pathogenic alterations.
Copy number variant (CNV) detection
CNVkit (v0.9.9)71 was used to detect CNVs. Only exonic regions were considered in the analysis. For patients with matched nontumor DNA (P1, P2, P6, P7, P8 and P9), tumor samples were compared to their matched reference. For others (P4 and P5), a generic copy number reference with neutral expected coverage was constructed in CNVkit and used for comparison. GISTIC (v2.0) was used to compute significant focal copy number variations.
Structural variant (SV) identification
Manta (v1.4.0)72 was used to identify somatic rearrangements involving BCL2 and MYC. The somatic SVs called from Manta were initially filtered with default settings and then manually screened using the Integrative Genomics Viewer (v2.7.2). The allele frequency was estimated based on supporting reads containing the breakpoints and whole read coverage of two loci.
Epigenetic annotation
The CTCF, H3K4me3 and H3K27ac ChIP–seq data from SUDHL6 and OCI-LY-1 cell lines were downloaded from ENCODE (ENCSR125DKL, ENCSR494LJG, ENCSR307DQT, ENCSR072EUE, ENCSR422JNY and ENCSR597UDW). The H3K4me1 ChIP–seq data are only available for OCI-LY-1 (ENCSR184QUS). SE regions were determined from the H3K27ac data of SUDHL6 and OCI-LY-1 via the ROSE method67. The TAD bed file, TAD boundary bed file, compartment information and Hi-C matrix data of the GM12878 cell line were downloaded from ref. 73 and used in the mutational annotation.
Identification of potential ER gene pairs in the B cell lymphoma genome
A TAD annotation bed file generated from Hi-C data of the GM12878 cell line was used to narrow down potential ER gene pairs. Any two genes with both of their promoters located within the same TAD were considered as a gene pair with ER potential. Using published transcriptome data of 11 DLBCL cell lines34, expression profiles of the 19,676 gene pairs were investigated. These gene pair candidates were then ranked by the P values that represent ranking consistently high (or low) on multiple lists including the rank of mean expression of gene A, the rank of mean expression of gene B and the rank of Spearman’s correlation coefficient between gene A and gene B. These asymptotic P values were estimated via the central limit theorem using the ‘rankPvalue’ function of the WGCNA (v1.72-1) R package. In addition, all gene pairs were annotated with data regarding hypermutation in our DHL cohort and from ref. 34 (n = 39 DLBCL), as well as with data regarding gene pairs previously reported in ref. 44 to experience ER. Finally, by filtering out nonhypermutated gene pairs, the prioritized potential ER gene pairs affected by aSHM in the B cell lymphoma genome were acquired.
Statistics and reproducibility
All statistical tests were computed with R v4.2.0, and details of statistical tests are indicated in all figures and corresponding figure legends. This is a retrospective study of all available FL/DHL pairs that were collected as part of diagnostic work in the University of Pittsburgh Medical Center and Columbia University Medical Center. Thus, no statistical test was used to predetermine sample size. One patient named ‘P3’ developed FL and DHL at the same time point, which may represent an early transformation. P3 was excluded from all analyses to remove any confusion.
Patients ‘P4’ and ‘P5’ were excluded from all somatic mutation burden analyses due to lack of matched nontumor DNA data. The experiments were not randomized. The investigators were not blinded to allocation during experiments and outcome assessment.
Additional software used
STAR (v2.7.3a) was used to map RNA-seq data. FeatureCounts (v2.0.0) was used to calculate gene expression levels. Bedtools (v2.26.0) was used to do genome annotation. ProteinPaint was used to draw lollipop mutation diagrams. WashU Epigenome Browser (v54.0.4) was used to visualize Hi-C matrix data. PROMO (v3.0.2) was used for transcription factor binding motif analysis. CTCFBSDB (v2.0) was used to predict potential CTCF binding sites on a given sequence. Arriba 2.3.0 was used to calculate gene fusion events via RNA-seq data. R package clusterProfiler (v4.4.4) was used to do gene set enrichment analysis.
Reporting summary
Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.
Data availability
Raw longitudinal WGS data from nine samples derived from five patients and collected before 2015 are available in dbGAP (phs003398.v1.p1). This IRB-approved study included a waiver of the requirement for informed consent. Per NIH policy, samples obtained after January 2015 cannot be uploaded to dbGAP without specific patient consent to do so. To access data from samples obtained after January 2015, investigators may contact the corresponding authors (U.B. or J.W.) and/or submit a request to the Columbia University Sponsored Projects Administration, via this form (https://cumc.co1.qualtrics.com/jfe/form/SV_29rqFAm9Dh4xX6Z), to obtain a data use agreement between Columbia University and the requesting institution. Following IRB approval of data sharing with each requesting institution, the data will be transferred electronically. Raw WGS data and RNA-seq data from 39 primary DLBCL cases and 11 DLBCL cell lines were downloaded from dbGaP—phs000235.v20.p6, and sequence read archive (SRA)—PRJNA523380. HiC data of GM12878 was downloaded from 4DN data portal (https://data.4dnucleome.org/) under accession 4DNES3JX38V5. 4C-Seq data was deposited in the Gene Expression Omnibus database with accession GSE210888. Raw WGS data and RNA-seq data from 29 DLBCL cell lines were downloaded from dbGaP—phs000328.v3.p1, SRA—PRJNA854968 and PRJNA523380. Proteomics data of SUDHL10 have been provided as source data in this paper. Source data are provided with this paper.
Code availability
All code developed for this study is available at https://github.com/ForceField17/WatchDHL.
References
Victora, G. D. & Nussenzweig, M. C. Germinal centers. Annu. Rev. Immunol. 40, 413–442 (2022).
Schatz, D. G. & Ji, Y. Recombination centres and the orchestration of V(D)J recombination. Nat. Rev. Immunol. 11, 251–263 (2011).
Alt, F. W., Zhang, Y., Meng, F.-L., Guo, C. & Schwer, B. Mechanisms of programmed DNA lesions and genomic instability in the immune system. Cell 152, 417–429 (2013).
Keim, C., Kazadi, D., Rothschild, G. & Basu, U. Regulation of AID, the B-cell genome mutator. Genes Dev. 27, 1–17 (2013).
Yeap, L.-S. & Meng, F.-L. Cis- and trans-factors affecting AID targeting and mutagenic outcomes in antibody diversification. Adv. Immunol. 141, 51–103 (2019).
Sun, J., Rothschild, G., Pefanis, E. & Basu, U. Transcriptional stalling in B-lymphocytes: a mechanism for antibody diversification and maintenance of genomic integrity. Transcription 4, 127–135 (2013).
Robbiani, D. F. & Nussenzweig, M. C. Chromosome translocation, B cell lymphoma, and activation-induced cytidine deaminase. Annu. Rev. Pathol. 8, 79–103 (2013).
Shaffer, A. L. III, Young, R. M. & Staudt, L. M. Pathogenesis of human B cell lymphomas. Annu. Rev. Immunol. 30, 565–610 (2012).
Feng, Y., Seija, N., Di Noia, J. M. & Martin, A. AID in antibody diversification: there and back again. Trends Immunol. 41, 586–600 (2020).
Meng, F.-L. et al. Convergent transcription at intragenic super-enhancers targets AID-initiated genomic instability. Cell 159, 1538–1548 (2014).
Pefanis, E. et al. RNA exosome-regulated long non-coding RNA transcription controls super-enhancer activity. Cell 161, 774–789 (2015).
Pefanis, E. et al. Noncoding RNA transcription targets AID to divergently transcribed loci in B cells. Nature 514, 389–393 (2014).
Qian, J. et al. B cell super-enhancers and regulatory clusters recruit AID tumorigenic activity. Cell 159, 1524–1537 (2014).
Wang, Q. et al. Epigenetic targeting of activation-induced cytidine deaminase. Proc. Natl Acad. Sci. USA 111, 18667–18672 (2014).
Casellas, R. et al. Mutations, kataegis and translocations in B cells: understanding AID promiscuous activity. Nat. Rev. Immunol. 16, 164–176 (2016).
Chapuy, B. et al. Molecular subtypes of diffuse large B cell lymphoma are associated with distinct pathogenic mechanisms and outcomes. Nat. Med. 24, 679–690 (2018).
Panea, R. I. et al. The whole-genome landscape of Burkitt lymphoma subtypes. Blood 134, 1598–1607 (2019).
Horning, S. J. & Rosenberg, S. A. The natural history of initially untreated low-grade non-Hodgkin’s lymphomas. N. Engl. J. Med. 311, 1471–1475 (1984).
Bastion, Y. et al. Incidence, predictive factors, and outcome of lymphoma transformation in follicular lymphoma patients. J. Clin. Oncol. 15, 1587–1594 (1997).
Gallagher, C. J. et al. Follicular lymphoma: prognostic factors for response and survival. J. Clin. Oncol. 4, 1470–1480 (1986).
Davies, A. J. et al. Transformation of follicular lymphoma to diffuse large B-cell lymphoma proceeds by distinct oncogenic mechanisms. Br. J. Haematol. 136, 286–293 (2007).
Gonzalez-Rincon, J. et al. Unraveling transformation of follicular lymphoma to diffuse large B-cell lymphoma. PLoS ONE 14, e0212813 (2019).
Okosun, J. et al. Integrated genomic analysis identifies recurrent mutations and evolution patterns driving the initiation and progression of follicular lymphoma. Nat. Genet. 46, 176–181 (2014).
Pasqualucci, L. et al. Genetics of follicular lymphoma transformation. Cell Rep. 6, 130–140 (2014).
Hilton, L. K. et al. The double-hit signature identifies double-hit diffuse large B-cell lymphoma with genetic events cryptic to FISH. Blood 134, 1528–1532 (2019).
Kridel, R. et al. Histological transformation and progression in follicular lymphoma: a clonal evolution study. PLoS Med. 13, e1002197 (2016).
Tsukamoto, T. et al. High-risk follicular lymphomas harbour more somatic mutations including those in the AID-motif. Sci. Rep. 7, 14039 (2017).
Boughan, K. M. & Caimi, P. F. Follicular lymphoma: diagnostic and prognostic considerations in initial treatment approach. Curr. Oncol. Rep. 21, 63 (2019).
Dreyling, M. et al. Newly diagnosed and relapsed follicular lymphoma: ESMO Clinical Practice Guidelines for diagnosis, treatment and follow-up. Ann. Oncol. 32, 298–308 (2021).
Cleary, M. L. & Sklar, J. Nucleotide sequence of a t(14;18) chromosomal breakpoint in follicular lymphoma and demonstration of a breakpoint-cluster region near a transcriptionally active locus on chromosome 18. Proc. Natl Acad. Sci. USA 82, 7439–7443 (1985).
Liu, D. & Lieber, M. R. The mechanisms of human lymphoid chromosomal translocations and their medical relevance. Crit. Rev. Biochem. Mol. Biol. 57, 227–243 (2022).
Kämpjärvi, K. et al. Somatic MED12 mutations are associated with poor prognosis markers in chronic lymphocytic leukemia. Oncotarget 6, 1884–1888 (2015).
Morin, R. D. et al. Frequent mutation of histone-modifying genes in non-Hodgkin lymphoma. Nature 476, 298–303 (2011).
Morin, R. D. et al. Mutational and structural analysis of diffuse large B-cell lymphoma using whole-genome sequencing. Blood 122, 1256–1265 (2013).
Senigl, F. et al. Topologically associated domains delineate susceptibility to somatic hypermutation. Cell Rep. 29, 3902–3915 (2019).
Bal, E. et al. Super-enhancer hypermutation alters oncogene expression in B cell lymphoma. Nature 607, 808–815 (2022).
Ye, X. et al. Genome-wide mutational signatures revealed distinct developmental paths for human B cell lymphomas. J. Exp. Med. 218, e20200573 (2021).
Mertz, T. M., Collins, C. D., Dennis, M., Coxon, M. & Roberts, S. A. APOBEC-induced mutagenesis in cancer. Annu. Rev. Genet. 56, 229–252 (2022).
Hnisz, D. et al. Super-enhancers in the control of cell identity and disease. Cell 155, 934–947 (2013).
Andrey, G. & Mundlos, S. The three-dimensional genome: regulating gene expression during pluripotency and development. Development 144, 3646–3658 (2017).
Lieberman-Aiden, E. et al. Comprehensive mapping of long-range interactions reveals folding principles of the human genome. Science 326, 289–293 (2009).
Fabbri, G. et al. Genetic lesions associated with chronic lymphocytic leukemia transformation to Richter syndrome. J. Exp. Med. 210, 2273–2288 (2013).
Puente, X. S. et al. Non-coding recurrent mutations in chronic lymphocytic leukaemia. Nature 526, 519–524 (2015).
Oh, S. et al. Enhancer release and retargeting activates disease-susceptibility genes. Nature 595, 735–740 (2021).
Chapuy, B. et al. Discovery and characterization of super-enhancer-associated dependencies in diffuse large B cell lymphoma. Cancer Cell 24, 777–790 (2013).
Liu, M. et al. Methodologies for improving HDR efficiency. Front. Genet. 9, 691 (2018).
Laffleur, B. et al. Noncoding RNA processing by DIS3 regulates chromosomal architecture and somatic hypermutation in B cells. Nat. Genet. 53, 230–242 (2021).
Pefanis, E. & Basu, U. RNA exosome regulates AID DNA mutator activity in the B cell genome. Adv. Immunol. 127, 257–308 (2015).
Rogozin, I. B. & Diaz, M. Cutting edge: DGYW/WRCH is a better predictor of mutability at G:C bases in Ig hypermutation than the widely accepted RGYW/WRCY motif and probably reflects a two-step activation-induced cytidine deaminase-triggered process. J. Immunol. 172, 3382–3384 (2004).
Chong, L. C. et al. High-resolution architecture and partner genes of MYC rearrangements in lymphoma with DLBCL morphology. Blood Adv. 2, 2755–2765 (2018).
Bertrand, P. et al. Mapping of MYC breakpoints in 8q24 rearrangements involving non-immunoglobulin partners in B-cell lymphomas. Leukemia 21, 515–523 (2007).
Rodriguez-Hernandez, G. et al. Infectious stimuli promote malignant B-cell acute lymphoblastic leukemia in the absence of AID. Nat. Commun. 10, 5563 (2019).
Swaminathan, S. et al. Mechanisms of clonal evolution in childhood acute lymphoblastic leukemia. Nat. Immunol. 16, 766–774 (2015).
Kilchert, C., Wittmann, S. & Vasiljeva, L. The regulation and functions of the nuclear RNA exosome complex. Nat. Rev. Mol. Cell Biol. 17, 227–239 (2016).
Puno, M. R., Weick, E. M., Das, M. & Lima, C. D. SnapShot: the RNA exosome. Cell 179, 282–282.e1 (2019).
Allmang, C., Mitchell, P., Petfalski, E. & Tollervey, D. Degradation of ribosomal RNA precursors by the exosome. Nucleic Acids Res. 28, 1684–1691 (2000).
LaCava, J. et al. RNA degradation by the exosome is promoted by a nuclear polyadenylation complex. Cell 121, 713–724 (2005).
Jumper, J. et al. Highly accurate protein structure prediction with AlphaFold. Nature 596, 583–589 (2021).
Lubas, M. et al. Interaction profiling identifies the human nuclear exosome targeting complex. Mol. Cell 43, 624–637 (2011).
Tafforeau, L. et al. The complexity of human ribosome biogenesis revealed by systematic nucleolar screening of Pre-rRNA processing factors. Mol. Cell 51, 539–551 (2013).
Forester, C. M. et al. Revealing nascent proteomics in signaling pathways and cell differentiation. Proc. Natl Acad. Sci. USA 115, 2353–2358 (2018).
Panda, A. C., Martindale, J. L. & Gorospe, M. Polysome fractionation to analyze mRNA distribution profiles. Bio Protoc. 7, e2126 (2017).
Carlotti, E. et al. Transformation of follicular lymphoma to diffuse large B-cell lymphoma may occur by divergent evolution from a common progenitor cell or by direct evolution from the follicular lymphoma clone. Blood 113, 3553–3557 (2009).
Marcel, V. et al. p53 acts as a safeguard of translational control by regulating fibrillarin and rRNA methylation in cancer. Cancer Cell 24, 318–330 (2013).
Pelletier, J., Thomas, G. & Volarevic, S. Ribosome biogenesis in cancer: new players and therapeutic avenues. Nat. Rev. Cancer 18, 51–63 (2018).
Swerdlow, S. H. et al. (eds) WHO Classification of Tumours of Haematopoietic and Lymphoid Tissues revised 4th edn, Vol 2 (IARC, 2017).
Demichev, V., Messner, C. B., Vernardis, S. I., Lilley, K. S. & Ralser, M. DIA-NN: neural networks and interference correction enable deep proteome coverage in high throughput. Nat. Methods 17, 41–44 (2020).
Mumbach, M. R. et al. HiChIRP reveals RNA-associated chromosome conformation. Nat. Methods 16, 489–492 (2019).
Rothschild, G. et al. Noncoding RNA transcription alters chromosomal topology to promote isotype-specific class switch recombination. Sci. Immunol. 5, eaay5864 (2020).
Tarasov, A., Vilella, A. J., Cuppen, E., Nijman, I. J. & Prins, P. Sambamba: fast processing of NGS alignment formats. Bioinformatics 31, 2032–2034 (2015).
Talevich, E., Shain, A. H., Botton, T. & Bastian, B. C. CNVkit: genome-wide copy number detection and visualization from targeted DNA sequencing. PLoS Comput. Biol. 12, e1004873 (2016).
Chen, X. et al. Manta: rapid detection of structural variants and indels for germline and cancer sequencing applications. Bioinformatics 32, 1220–1222 (2016).
Rao, S. S. P. et al. A 3D map of the human genome at kilobase resolution reveals principles of chromatin looping. Cell 159, 1665–1680 (2014).
Acknowledgements
Research in the Basu Lab is supported by grants to U.B. (1R01AI099195, 1RO1AI143897-01A1 and R01AI134988) and to R.J.L.-N. (DOD W81XWH-18-1-0394). J.B. was funded by NIH 5T32AI148099. Research in the Wang Lab is supported by the National Natural Science Foundation of China (NSFC) Excellent Young Scientists Fund (31922088), Hong Kong Research Grants Council (16101021, CRS_HKUST605/22), Hong Kong Innovation and Technology Commission (MHP/004/19, ITCPD/17-9), Padma Harilela Professorship and Project of Hetao Shenzhen-Hong Kong Science and Technology Innovation Cooperation Zone (HZQB-KCZYB-2020083). Research in the Lafontaine Lab is funded by the Belgian Fonds de la Recherche Scientifique (F.R.S./FNRS), the Université libre de Bruxelles (ULB), the European Joint Program on Rare Diseases (EJP-RD) ‘RiboEurope’ and ‘DBAGeneCure’; the Région Wallonne (SPW EER; ‘RIBOcancer’ FSO grant 1810070; POC grant 1880014), the EOS CD-INFLADIS (grant 40007512). Research in the L.E.B. Lab is funded by the Hirschl Family Trust and NIH grant R35 GM124633. J.L. is funded by NRF-2021R1A2C1012477. D.E.W. is funded by T32 CA265828. We acknowledge the support from Hong Kong RGC-CRF equipment grant C5033-19E (to Q.Z.). We thank the Columbia University Flow Cytometry Core of the Stem Cell Initiative and Department of Microbiology and Immunology cell sorting facility (both for FACS and cell sorting) and Columbia University Genome Center (for high-throughput genomics). Confocal microscopy was performed in the Confocal and Specialized Microscopy Shared Resource, with tissue processing, TMA creation and staining carried out in the Molecular Pathology Shared Resource of the Herbert Irving Comprehensive Cancer Center at Columbia University, supported by NIH/NCI Cancer Center Support Grant P30CA013696. We thank G. Zhang of the Weill Cornell Medicine Proteomics and Metabolomics Core and R. K. Soni of the Proteomics and Macromolecular Crystallography Shared Resource, Columbia University for their help with mass spectrometry experiments.
Author information
Authors and Affiliations
Contributions
J.B. performed microscopy experiments. D.S., Y.C. and W.Z. performed bioinformatic analyses of all genomics data. L. Wu performed 4C assay with guidance from J.L. A.Y.S., K.G., G.R. and M.C.L. generated cell lines and performed biochemical experiments. R.J.L.-N., S.B., G.B. and S.H.S. obtained appropriate FL/DHL for WGS. S.S. generated the predicted model of the TRAMP complex. R.J.L.-N. created TMAs and analyzed ZCCHC7 immunohistochemistry. L. Wacheul, G.R. and D.L.J.L. performed and interpreted the rRNA processing experiment. G.R., D.E.W. and L.E.B. performed the polysome analysis. Y.Y. and Q.Z. performed the OPP-ID nascent proteomics analysis. R.J.L.-N., D.S., J.W. and U.B. wrote the paper with input from all authors. G.R., L.E.B., S.H.S., D.L.J.L. and G.B. edited the paper. J.W. and U.B. directed the project. U.B. and J.W. contributed equally.
Corresponding authors
Ethics declarations
Competing interests
The authors declare no competing interests.
Additional information
Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Extended data
Extended Data Fig. 1 Features of BCL2 and MYC alterations.
(a) Examples of longitudinal samples of follicular lymphoma (FL, left) and double hit lymphoma (DHL, right) obtained from the same patient (100X) (top). Fluorescence in situ hybridization (FISH) performed on a DHL, using break-apart probes, showing rearrangements involving BCL2 (left) and MYC (right) (bottom). (b) Examples of BCL2 and MYC translocations identified via whole genome sequencing (WGS). The split reads that support BCL2-IGH in P1 show a 10bp break-end insertion. Supporting reads of IGH-MYC in P1 are fully mapped to the IGH-MYC junction. (c) BCL2 and MYC translocation breakpoints in IGH loci. All three IGH-MYC breakpoints are located at the intergenic region of IGH C genes. All BCL2-IGH breakpoints are clustered within IGH J genes. (d) Sequence pattern analysis of BCL2-IGH (n = 8) and IGH-MYC (n = 3) breakpoints. BCL2-IGH breakpoints always locate no more than 2bp from CpG sites in BCL2 but far from CpG sites in IGH loci, while breakpoints of IGH-MYC are close to CpG sites in both IGH and MYC. P-values were calculated by two-sided Wilcoxon signed rank tests. The boxplots display the 25th and 75th percentiles and median of each group. Whiskers represent highest and lowest values within 1.5 \({\rm{\times }}\) the inter-quartile range. (e) Comparison of break-end insertions between BCL2 fusions (n = 8) and MYC fusions (n = 8). All BCL2 translocations present template-independent break-end insertions with lengths ranging from 5–25 bp. None of the MYC translocations show break-end insertions. P-values were calculated by two-sided Wilcoxon rank-sum tests. The boxplots display the 25th and 75th percentiles and median of each group. Whiskers represent the highest and lowest values within 1.5 \({\rm{\times }}\) the inter-quartile range. (f) Models of BCL2-IGH in FL and IGH-MYC in DHL. Double strand breaks (DSB) in BCL2 are caused by activation induced cytidine deaminase (AID) activity at CpG sites in follicular lymphoma and/or immature B cells and are joined with the breakpoints of IGH resulting from RAG activity during VDJ recombination. Subsequently, the break-end joining process is triggered by TdT, a specialized DNA polymerase that adds nucleotides at junctions. In DHL, AID-associated DSBs are detected in both MYC and IGH C genes. No break-end insertions are observed.
Extended Data Fig. 2 Examples of recurrent foci of transformation-associated aSHM.
(a-c) Examples of recurrent foci of transformation-associated aberrant somatic hypermutation (aSHM), including PIM1 and RHOH genes (a), and aSHM occurring in the first intron of the CXCR4 gene, shown on the rainfall plot in (b) and in relation to H3K4me3 and H3K27ac peaks in (c). (d) Statistical evaluation of mutation density at various histone marks and equal size random areas for follicular lymphoma (left, n = 7 samples) and DHL (right, n = 6 samples). Each dot represents normalized mutation density in a specific genomic region for each tumor sample. The horizontal line in each group represents its mean value, and the corresponding vertical line represents its standard deviation bar. The group ‘Random A’ is composed of a series of randomly generated genomic regions with each single region size and total region size equal to the average single size and total size of all H3K27ac, H3K4me3, and H3K4me1 peaks respectively. The group ‘Random B’ is composed of a series of randomly generated genomic regions with each single region size and total region size equal to the average single size and total size of all H3K27ac outside H3K4me3, H3K4me3 outside H3K27ac, and H3K4me3 overlapping H3K27ac peaks respectively. P-values in each figure were calculated using two-sided Wilcoxon signed rank tests paired by tumor samples with false discovery rate (FDR) correction.
Extended Data Fig. 3 Micro-insertions and microdeletions observed in follicular lymphoma and double hit lymphoma, PAX5/ZCCHC7 mutational hotspot in lymphoma, and the PAX5/ZCCHC7 locus in mice.
(a) Overall frequency of micro-insertions (left) and micro-deletions (right) observed in paired follicular lymphoma and DHL samples from n = 6 patients. Each pair of dots connected by a grey line represents a pair of follicular lymphoma and DHL samples from a single patient. The mutational frequencies on the y-axis are expressed as percentage of micro-insertions or micro-deletions out of all mutations. The P-values were calculated by two-sided Wilcoxon signed rank tests. The boxplots display the 25th and 75th percentiles and the median of each group of data. The whiskers represent the highest and lowest values within 1.5 \({\rm{\times }}\) the inter-quartile range. (b) The number of micro-indels overlapping H3K4me3 and H3K27ac intersecting regions in DHL compared to follicular lymphoma. Each pair of points represents an intersection segment of H3K4me3 peak and H3K27ac peak. All n = 21,283 segments were ranked in descending order by the sum of micro-indels observed in DHL. Five segments that harbor over three micro-indels are highlighted and labeled with nearby genes. The P-value was calculated by two-sided Wilcoxon signed-rank test. (c) Microdeletions and micro-insertions observed in the PAX5-TSS2 region in three DHL tumors, two diffuse large B cell lymphoma (DLBCL) cell lines, as well as two primary DLBCL tumors. In IGV, WGS data from these seven samples show eight micro-deletions (blue arrows) and one micro-insertion (red arrow) in this region. (d) Mutational hotspots in the PAX5/ZCCHC7 locus previously reported in chronic lymphocytic leukemia (CLL)43 (top panel), DLBCL34 (middle panel), and in the current DHL series (bottom panel). All p-values were calculated by two-sided Wilcoxon signed-rank test. (e) Hi-C overlap of the PAX5/ZCCHC7 locus shows bidirectional transcription (Dis3C/C represents cells depleted of RNA exosome activity to stabilize bidirectional ncRNA expression), accumulation of non-B DNA associated DNA/RNA hybrids (in Dis3C/C cells via DRIP-seq), H3K27ac peaks and AID target site identification via translocation capture sequencing (via HTGTS-seq). These observations demonstrate that processes relating to DNA targeting by AID occur at the PAX5 promoter in mouse B cells.
Extended Data Fig. 4 Prediction of enhancer retargeting.
(a) Significant negative correlation between RNA expression of PAX5 and ZCCHC7 in 11 DLBCL cell lines. The P-value was calculated by two-sided Spearman’s correlation test. (b) Schematic of the pipeline used to identify potential targets of enhancer retargeting in the B cell lymphoma genome. Overall results of this analysis are shown in Fig. 4b. The transcriptomic data from 11 DLBCL cell lines were reported by Morin et al.34. The list of 367 enhancer retargeting (ER) gene pairs was previously reported by Oh et al.44. (c) Top 10 potential enhancer retargeting gene pair candidates ranked by the pipeline and filtered by ‘Hypermutated’ criteria. The P-values were estimated for ranking consistently high (or low) on multiple lists (see Methods). (d) Comparison of Spearman’s correlation coefficient between predicted gene pairs experiencing enhancer retargeting and affected in opposite directions (n = 10,158) versus those affected in the same direction (n = 9,518). The P-value is calculated by two-sided Wilcoxon rank sum test. The boxplots display the 25th and 75th percentiles and the median of each group of data. The whiskers represent the highest and lowest values within 1.5 \(\times\) the inter-quartile range. (e) Proximity of FAM102A and SLC25A25-AS1 and somatic mutations occurring in the promotor of FAM102A.
Extended Data Fig. 5 Overexpression of ZCCHC7 in DLBCL cell lines and primary tumors that harbor ZCCHC7 fusion, ZCCHC7 copy number gain, or PAX5-TSS2 mutation.
(a) Transcript reconstitution of PAX5 demonstrates two TSS, canonical TSS and PAX5-TSS2. The distribution of H3K4me3, H3K27ac, and aSHM overlapping the two TSS is shown. (b) PAX5-TSS2 hotspot mutations and deletions seen in four of the DHLs in this study (P1, P2, P6, P9) and in other studies34. The most frequent mutations, that characterize the PAX5-TSS2mut in Fig. 6 are marked by a red frame. (c) In the SUDHL10 cells, deletion of PAX5-TSS2 (left) or incorporation of TSS2mut (right) leads to marginally decreased expression of PAX5 mRNA but significantly increased expression of ZCCHC7. Student’s 2-tailed T-test (paired) used for statistical analysis in both panels. (d) Mutation heatmap of 29 DLBCL cell lines (left panel) and 24 DLBCL primary tumors (right panel) at ZCCHC7 and PAX5 TSS2. Cell line WGS/RNAseq data were obtained from CCLE and SRA: PRJNA854968. Tumor WGS/RNAseq data were obtained from dbGaP: phs000235.v20.p6. Cell lines or tumors showing ZCCHC7 copy number gain, ZCCHC7 fusion, or mutation in any of four PAX5-TSS2 hotspots (as shown in (b)) are included in the ‘Mutant’ group. Others, displaying no alteration in these regions, compose the ‘WT’ group. All of the ZCCHC7 fusion events were detected via WGS or RNAseq data and summarized in Extended Fig. 7. All of the hotspot mutations were confirmed by checking corresponding WGS data or RNAseq data in IGV, with at least five supporting reads. Six primary tumors without WGS data but showing ZCCHC7 fusion or PAX5-TSS2 hotspot mutation in RNAseq data were also included. ‘WT’ samples showing poor read coverage (depth < 10) at the four mutational hotspots were discarded. Z-score transformed RNA expression of ZCCHC7 and PAX5 are shown at the bottom. (e and f) RNA expression of ZCCHC7 and PAX5 in the ‘WT’ and ‘Mutant’ groups of DLBCL cell lines (e) and tumors (f) described in (d). P-values were calculated by two-sided Wilcoxon rank-sum test. Boxplots display the 25th and 75th percentiles and median of each group. Whiskers represent the highest and lowest values within 1.5 \({\rm{\times }}\) the inter-quartile range.
Extended Data Fig. 6 Enhancer retargeting at the PAX5/ZCCHC7 locus resulting from aSHM at the PAX5 alternative promoter/TSS.
(a) Four independent 4C-seq experiments measuring interactions between the ZCCHC7 promoter and intragenic enhancer regions adjacent to PAX5. In ΔPAX5-TSS2 SUDHL10 (PAX5-TSS2 deletion shown with black arrow), the PAX5 intragenic enhancer region interacts more strongly with the promoter of ZCCHC7 relative to cells with intact PAX5-TSS2 (see thickness of black lines). (b) Comparison of 4C-seq data from PAX5 TSS2WT versus PAX5 TSS2del cells, demonstrating that similar regions of the ZCCHC7-PAX5 SE show increased interaction with the ZCCHC7 gene. (c) The normalized coverage of the interactions (site 1, site 2, site 3, and site 4) in (a and b) is shown in the bar graph. Error bars represent the standard deviations. Two-sided Welch’s t-test were used; n = 4 for interaction between site 1 and 4C bait 1, n = 2 for other interactions. (d) Sanger sequencing of chr9:37,026,299-37,026,327 in SUDHL10 PAX5TSSmut cells. The two mutations are located at chr9:37,026,315-37,026,316 (as in Extended Data Fig. 5b). (e) Summary of PAX5 H3K4me3 region mutations acquired upon transformation to DHL. The weblogos show the classical AID mutational signature, C→T/G→A with WRC/GYW motif (W = A/T, R = A/G, and Y = T/C), that dominates the DHL-specific mutations within this 3kb region of PAX5. Specifically, 11 C→T/G→A mutations out of 28 DHL-associated mutations in this region reflect the WRC/GYW motif (labeled in red).
Extended Data Fig. 7 Structural variants involving ZCCHC7.
(a) MYC-ZCCHC7 translocation identified in the DHL from P11. P11 represents a de novo DHL (no prior follicular lymphoma) selected for sequencing due to a high degree of ZCCHC7 expression observed on IHC, but not included in the main cohort of paired longitudinal samples. The translocation junction is observed between MYC downstream and ZCCHC7 downstream. (b) PVT1-ZCCHC7 translocation and CD274-ZCCHC7 rearrangement identified in SUDHL6 and RCK8 cell lines, respectively. The breakpoints in SUDHL6 cells were observed within the two genes, while the RING Zn-finger domain (light green) of ZCCHC7 remained in the fused transcript. Break points of the CD274-ZCCHC7 fusion in the RCK8 cell line are located in the 3′ UTR of CD274 and 32 kb downstream of ZCCHC7. The fusion-supporting junction reads display ZCCHC7 in an antisense direction to CD274. (c) Three ZCCHC7 translocations detected in RNAseq data from primary DLBCL tumors reported by Morin et al.34. The PVT1-ZCCHC7 translocations were observed in two DHL patients. The majority of the functional domain (teal) of ZCCHC7 is preserved in the two fusion transcripts. The ZCCHC7-IGL translocation, which joined downstream of ZCCHC7 and IGLL5 exon 3, was detected in a DLBCL. All the RNAseq data supported gene fusions were detected by Arriba 2.3.0 and checked in IGV. (d) Investigation of the occurrence of the PAX5-ZCCHC7 gene fusion among n = 1,173 hematopoietic malignancies (HM) samples reported in the St. Jude Cloud ProteinPaint portal. The PAX5-ZCCHC7 gene fusions that connect intron 6 of PAX5 and intron 2 of ZCCHC7 in an antisense direction were reported in 17 cases of B-ALL. (e) Comparison of ZCCHC7 expression among four types of HM. B-ALL and B-ALL with known PAX5-ZCCHC7 fusions show the highest expression levels of ZCCHC7 compared to other types. The P-values were calculated using two-sided Wilcoxon rank-sum tests. The boxplots display the 25th and 75th percentiles and the median of each group of data. The whiskers represent the highest and lowest values within 1.5 \(\times\) the inter-quartile range.
Extended Data Fig. 8 ZCCHC7 protein expression in lymphoma tissue, ZCCHC7 interaction network, reconstituted structure of the TRAMP complex using alpha-fold, and ZCCHC7 localization in SUDHL6 cells.
(a) Tissue microarrays (TMAs) were immunohistochemically stained for ZCCHC7. DLBCL (N = 33, p = 0.0084) and DHL (N = 9, p = 0.03) showed a higher average degree of ZCCHC7 expression than individual benign lymph node, spleen, or tonsil, all of which showed a very low degree of ZCCHC7 expression. Mean H-score and standard deviation for each group are shown, with p-values calculated by two-sided Wilcoxon rank-sum test. (b) Longitudinal samples representing transformation of follicular lymphoma to DLBCL in 10 patients (separate from the sequenced cohort) were stained for ZCCHC7. ZCCHC7 expression usually increased upon lymphoma transformation. (c) Comparison of endogenous ZCCHC7 in the non-neoplastic human B cell line CL-01 and SUDHL6 (a DHL line that overexpresses ZCCHC7). NPM1 defines the nucleolus and DAPI counterstains nuclei (blue). ZCCHC7 accumulates in nucleoli of CL01 cells while also present in the nucleoplasm of SUDHL6 (scale bar 5 µm). (d) The ratio of nucleolar to total nuclear signal is significantly reduced in SUDHL6 compared to CL-01. Statistical significance was assessed using an unpaired two-tailed t-test, and the number of cells analyzed in each case is indicated. (e) Pre-rRNA processing pathway. Three out of four mature rRNAs, the 18S, 5.8S, and 28S are encoded as a long polycistronic precursor synthesized by RNA polymerase I, the 47S. The mature rRNA sequences are produced by extensive processing (cleavage sites indicated in blue). In the 47S, the mature rRNAs are interspersed by 5′ and 3′ external transcribed spacers (ETS) and internal transcribed spacers (ITS) 1 and 2. The probes used in northern blotting (LD1828, LD1844, LD2079, and LD2122) are highlighted. (f) FLAG-tagged ZCCHC7 protein and a FLAG-tag peptide were overexpressed in 293T cells to purify ZCCHC7 interacting factors via Mass spectrometry. The components of the RNA exosome complex essential cofactor NEXT directly interact with ZCCHC7, as does a set of nucleolar proteins. (g) Predicted structure model of the human TRAMP-like complex. The complex consists of the RNA helicase MTR4 (PDB ID: 7S7B, yellow), RNA (PDB ID: 7S7B, blue), the zinc-knuckle protein ZCCHC7 (AlphaFold prediction, green), and the noncanonical poly(A) polymerase PAPD5 (AlphaFold prediction, red). Ribbon representation of the model (left) and 90° rotation (right) highlighting binding of RNA and its recruitment to the RNA channel in MTR4.
Extended Data Fig. 9 Evaluating the effect of ZCCHC7 overexpression on rRNA processing.
(a-c) Total RNA extracted from the indicated cells was separated by denaturing agarose gel electrophoresis, the gel was stained with ethidium bromide (panel a), transferred to a nylon membrane and probed with radioactively-labeled anti-sense oligonucleotides specific to ITS1 (b) or 5′ ETS (c) SCR#1 is a non-targeting scrambled siRNA used to control HeLa depleted of EXOSC2, EXOSC10, and SKIV2L2. SCR#2 is non-targeting scrambled siRNA used to control HeLa depleted of ZCCHC7. The experiments represent results from 3 independently performed experiments. (d) Overexpression of Exosc10 (Exosc10-OE) in PAX5-TSS2mut SUDHL10 cells (PAX5-TSS2mut) rescues 5.8S + 40 rRNA processing. PAX5-TSS2mut SUDHL10 cells were transfected with a control vector that does not contain Exosc10 or with an Exosc10 expression vector also containing green fluorescent protein (GFP) (See western blot showing Exosc10 expression) and cells were sorted by red fluorescent protein (RFP) expression. The flow cytometric experiment was repeated three times. Total RNA was isolated from GFP positive cells and separated on a 7% acrylamide gel, transferred to a membrane and blotted with a probe against 5.8S+40 rRNA. The level of 5.8S+40 rRNA decreases following Exosc10 overexpression (lane 3), as compared to untransfected PAX5-TSS2mut cells (lane 1), PAX5-TSS2mut cells containing the control vector (lane 2), or unmutated and untransfected SUDHL10 cells (lane 4).
Extended Data Fig. 10 Proteomic changes resulting from ZCCHC7 alteration.
(a and b) Ranked gene set enrichment analysis (GSEA) of O-propargyl puromycin-mediated identification (OPP-ID) data using KEGG pathways collection. Genes (n = 5,176) corresponding to altered proteins were ranked in ascending order by two-way ANOVA test P-value of the genetic alteration factor on the x-axis in Fig. 7e. Pathways that show FDR adjusted P-value < 0.05 are shown in (a). The sign of normalized enrichment score (NES) suggests downregulation (negative NES) or upregulation (positive NES) in the SUDHL10 ZCCHC7-OE cells and SUDHL10 PAX5 TSS2Mut cells compared to SUDHL10 WT cells. GSEA shows downregulation of proteins in ribosome, DNA replication, mismatch repair, and non-homologous end joining pathways (b). (c) Fold changes of nascent proteins in ZCCHC7-OE or PAX5 TSS2Mut SUDHL10 cells compared to the wild type SUDHL10 cells. Each dot represents a protein with significantly higher abundance in OPP+ than OPP- in a two-way ANOVA (Fig. 7e). 1,690 proteins in total were captured via OPP and evaluated for nascent protein level. P-value and rho of Spearman’s correlation test between the x-axis and y-axis show a significant positive correlation between fold change of proteins in ZCCHC7-OE and PAX5 TSS2Mut. Tumor suppressor proteins including IKZF3, CHEK1, etc show decreasing translation in both ZCCHC7-OE cells and PAX5 TSS2Mut cells. (d) Independent whole cell proteomics for triplicate cultures of CL-01 (non-neoplastic B cells), SUDHL6 (which overexpress ZCCHC7 due to chromosomal translocation), SUDHL10 cells (no inherent alterations in PAX5-TSS2 or ZCCHC7) and ZCCHC7-OE (SUDHL10 cells overexpressing ZCCHC7). Of the 3428 proteins with different protein expression between SUDHL6 and CL-01, 215 showed alterations also observed in ZCCHC7-OE compared to SUDHL10. Some of these proteins are tumor suppressor proteins in B-NHL (for example IKZF3, TRAF3, etc.). (e and f) Polysome profiling of SUDHL10 WT and SUDHL10 ZCCHC7-OE (e) and analyses of IKZF3 mRNA distribution on the polysome (f). (g) ZCCHC7 transgene was introduced into SUDHL10, and stable lines isolated. ZCCHC7 protein expression is compared in ZCCHC7-OE and the SUDHL10 parental cell line. Three biological replicates were performed. (h) Evaluation of nascent levels of 20 proteins targeted by available therapeutics. The expression heatmap of the proteins in SUDHL10 and SUDHL10 PAX5-TSS2mut lymphoma cells is shown.
Supplementary information
Supplementary Information
Supplementary Methods and Figs. 1–4.
Supplementary Tables 1–10
Supplementary Tables 1–3: Detailed lists of all genomic alterations, including point mutations, SV and CNV seen in follicular lymphoma and DHL from the eight patients analyzed in this study (relevant to Figs. 1–5). Supplementary Table 4: DLBCL cell lines that harbor or lack genomic alterations of the PAX5/ZCCHC7 locus. Supplementary Table 5: Proteomics intensity data (in OPP-ID) of SUDHL10 relevant for Fig. 7 and Extended Data Fig. 10. Supplementary Tables 6 and 7: Detailed information about BCL2 rearrangements in follicular lymphoma and DHL from the eight patients in the cohort. Supplementary Tables 8 and 9: Detailed information about BCL2 rearrangements in follicular lymphoma and DHL from the eight patients in the cohort. Supplementary Table 10: Sequences of probes and oligonucleotides used.
Source data
Source Data Figs. 1–5 and 7 and Extended Data Figs. 2, 3, 5, 8, 7 and 10
Source data for Figs. 1d–f, 2d, 3a,b, 4b,d, 5a,e and 7b,c,e,g and Extended Data Figs. 2d, 3b, 5c, 8a, 7e and 10a,c,f.
Source Data Fig. 6
Northern blots for Fig. 6e,f. Source Data Extended Data Fig. 9 Northern and western blots for Extended Data Fig. 9d. Source Data Extended Data Fig. 10 Western blots for Extended Data Fig. 10g.
Rights and permissions
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons license, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this license, visit http://creativecommons.org/licenses/by/4.0/.
About this article
Cite this article
Leeman-Neill, R.J., Song, D., Bizarro, J. et al. Noncoding mutations cause super-enhancer retargeting resulting in protein synthesis dysregulation during B cell lymphoma progression. Nat Genet 55, 2160–2174 (2023). https://doi.org/10.1038/s41588-023-01561-1
Received:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1038/s41588-023-01561-1
