Chromatin activation profiling of stereotyped chronic lymphocytic leukemias reveals a subset 8–specific signature

Chronic lymphocytic leukemia (CLL) is the most common leukemia in the western world and shows a wide range of biological features with clinical associations.¹ A key feature of CLL pathogenesis is the B-cell receptor immunoglobulin (BcR-IG), which represents a critical disease driver providing survival and proliferation signals. The importance of the BcR-IG is highlighted by the identification of 2 major CLL molecular subtypes based on the somatic hypermutation (SHM) status of IGHV genes. Cases displaying no or limited SHM (“unmutated” CLL, U-CLL) generally show a more aggressive clinical course compared with those with a significant SHM burden (“mutated” CLL, M-CLL).² Another remarkable aspect underlining the significance of the BcR in CLL concerns the identification of subsets of patients expressing (quasi)identical, stereotyped BcR-IG sequences,³ strongly supporting a role for antigen selection in the natural history of CLL. Four major stereotyped subsets in CLL, that is, 1, 2, 4, and 8, which cumulatively account for ∼7% of all CLLs, mirror the heterogeneous nature of CLL as a whole. Also, these stereotyped subsets represent distinct disease variants with consistent clinico-biological features, such as different frequencies of genetic lesions (SF3B1 mutations in 2 or trisomy 12 in 8⁴) and different outcomes.^5,6 

Besides immunogenetics and genetics, epigenetic studies focused on DNA methylation have further identified differential features in certain stereotyped subsets, for example, subset 2 belongs to the intermediate epigenetic subgroup⁷ and subset 8 shows a specific DNA methylation signature.⁸ Despite the information gathered through these studies, an analysis of chromatin activation assessed by histone 3 lysine 27 acetylation (H3K27ac), represents a better approach to identify altered regulatory elements and transcription factor (TF) networks in CLL.^9-11 In this study, we have generated H3K27ac profiles of stereotyped CLL subsets 1, 2, 4, and 8. In comparison with compared nonstereotyped CLLs, we identified a remarkably distinct profile in stereotyped subset 8, potentially linked to its particularly high risk of Richter transformation (RT), which is the highest among all CLLs.¹² 

We performed chromatin-immunoprecipitation followed by sequencing (ChIP-seq) with an H3K27ac antibody in sorted B cells from 15 CLL cases from stereotyped subsets 1, 2, 4, and 8. In addition, H3K27ac ChIP-seq profiles from 31 CLLs (including 3 cases of subset 1, 2 cases of subset 4, and 1 case each of subsets 2 and 16) and 15 samples from 5 different normal B-cell subpopulations (n = 3 each) were included from a previous publication from our group.⁹ Overall, we studied ChIP-seq data from 22 stereotyped CLLs and 24 nonstereotyped CLLs (supplemental Table 1, available on the Blood website). To estimate the functional impact of differential H3K27ac, we integrated our data with RNA-seq data (n = 16) considering the 3D chromatin configuration (Hi-C data from GM12878 and a representative U-CLL case¹³) using InterTADs tool¹⁴ (Figure 1A; supplemental Figure 1; supplemental Table 1; supplemental Methods).

Initial unsupervised principal component analysis revealed that the main source of the variability (30%) differentiates CLL as a whole from normal B-cell subpopulations (Figure 1B; supplemental Figure 2). Although stereotyped and nonstereotyped CLLs overall clustered together, component 2 of the variability (13%) suggested that subset 8 cases are different from the other CLLs. Next, we performed supervised differential acetylation analyses between each CLL subset vs nonstereotyped CLLs matched by SHM status. Using a stringent false discovery rate (FDR) of 0.0001, we identified 2726 differential regions (DRs) in subset 8 vs U-CLL, 0 DRs in subset 1 vs U-CLL, 3 DRs in 2 vs M-CLL, 0 DRs in 4 vs M-CLL. Out of the 2726 DRs in subset 8, we selected 1185 robust DRs after repeating the test 100 times by applying random resampling (Figure 1C; supplemental Methods; supplemental Figure 3; supplemental Table 2). Overall, both unsupervised and supervised analyses suggest that subsets 1, 2, and 4 show few if any changes compared with IGHV-matched–nonstereotyped CLLs; although we cannot rule out the presence of small-effect changes that cannot be detected owing to the low sample size. However, subset 8 cases are markedly different.

Therefore, we focused the downstream analyses on further characterizing subset 8. Notably, subset 8 cases frequently (∼60%) show trisomy 12,⁵ which shows differential acetylation profiles in CLL.^9,15 To evaluate this relationship, we analyzed the overlap of the subset 8 signature and that established for trisomy 12.⁹ We detected only 30 DRs in common, which was a nonsignificant overlap (p-hypergeometric test = 0.66), and therefore, the subset 8 signature seemed to be unrelated to trisomy 12.

We subsequently characterized the 1185 DRs and identified 6 subclusters with different patterns, 2 hyperacetylated (clusters 1,5) and 4 hypoacetylated (clusters 2-4) in subset 8. It is worth highlighting that cluster 1 contained regions de novo acetylated in subset 8, whereas the cause of hyperacetylation in cluster 5 seemed to be related to reduced acetylation in nonsubset 8 CLLs as compared with subset 8 (Figure 1C-D). We then designed a strategy to annotate DR to target genes (Figure 1E). First, we linked a given DR to all genes located within a topologically associating domain (TAD) in GM12878, a lymphoblastoid B-cell line widely used to characterize epigenetic patterns in CLL.⁹ Second, we selected those genes that are differentially expressed between 8 and U-CLL (FDR < 0.01), mining RNA-seq data from partially the same cases.^8,9 Finally, we correlated the chromatin acetylation and gene expression levels in the same cases considering only the differentially expressed genes (FDR < 0.01) (Spearman correlation, P value < .05), expecting a positive correlation for target genes. Hyperacetylated regions from clusters 1 and 5 showed high correlation rates with gene expression (78/209 [37.3%] and 63/157 [40.1%]) (Figure 1F). In contrast, hypoacetylated regions showed a lower correlation rate. This analysis suggests that clusters associated with increased H3K27ac in subset 8 frequently contain genes upregulated in this CLL subtype. As cluster 1 was clearly associated with de novo chromatin activation (Figure 1C), we next focused on the 209 regions from this cluster, as they could likely be informative on the specific mechanisms underlying the pathogenesis of subset 8.

Having seen that subset 8 CLLs have a unique H3K27ac signature, we next wondered whether this signature may be related to their high risk to develop RT. To gain insight into this association, we aggregated recently published longitudinal H3K27ac ChIP-seq data from 4 nonsubset 8 U-CLLs that progressed to RT and included samples at diagnosis, progression, relapse, and RT.¹¹ This comparison revealed a shared pattern between them, both considering the 1185 DRs in subset 8 and the 209 de novo active regions from cluster 1 (Figure 2A; supplemental Figure 3E). These results highlight that the chromatin signature observed in subset 8 cases may potentially contribute to their pronounced tendency to develop RT.

As chromatin activation patterns are frequently associated with binding sites of particular TF,^9,10 we next performed a TF binding site analysis. The regions from cluster 1 were significantly enriched (adj-P value < .05) in binding sites of ZNF467, ZBTB7A, SALL4, MYF6, KLF15, GLI3, EBF1, E2F2, and AHR (Figure 2B) whose expression was upregulated in subset 8 in comparison with U-CLL. Most of them have previously been associated with CLL, such as MYF6 which is overexpressed in trisomy 12 cases,¹⁶ or other hematologic malignancies, such as ZBTB7A, which is mutated in acute myeloid leukemia.¹⁷ This analysis suggests that a few overexpressed TFs may be responsible for the de novo chromatin activation pattern observed in subset 8.

We next focused on the downstream effects of chromatin activation in subset 8 by characterizing the potential target genes. Of the 209 regions of cluster 1, 78 unique peaks were significantly correlated with the expression of 113 unique genes, which also were overexpressed in subset 8 (Figure 2C; supplemental Table 3). Among the top 5 differentially expressed genes (ie, HLA-DRA, ZDHHC19, LINGO3, NDUFA4L2, and GZMM), mitochondrial protein NDUFA4L2 regulates cell survival associated with poor prognosis in human cancers.¹⁸ Moreover, ZDHHC19 mediates palmitoylation and provides signals on STAT3 regulation,¹⁹ which plays a critical role in CLL cell survival.²⁰ 

Beyond individual gene functions, we observed that 11 out of 78 activated regions were correlated with more than 2 genes, suggesting that several genes within a TAD could be associated with subset 8 pathogenesis through local coupregulation (Figure 2D). Among them, the acetylation peak 61315 located in chromosome 9 showed a positive correlation with the expression of 7 transcripts (6 coding genes CLIC3, FBXW5, AJM1, MAMDC4, CCDC183, and TRAF2 as well as 1 unannotated transcript) within its host TAD (Figure 2D-E). In this case, although peak 61315 was the only significantly different one, we observed 7 additional peaks in the region with increased H3K27ac in subset 8 (supplemental Figure 4), suggesting that coupregulated genes in this area are not only affected by 1 peak but by the activation of several regulatory elements within the TAD. As controls, we used the TADs next to TAD1791 (TAD1790 and TAD792), which showed 1 differential H3K27ac peak each that was not correlated with the expression of the genes within their host TAD or the adjacent TAD1791. To examine the overexpressed genes in subset 8 regulated by peak-61315, TRAF2 was a remarkable example because this gene suppresses death-receptor–mediated apoptosis and cooperates with BCL2 promoting CLL/small lymphocytic lymphoma in mice.²¹ Further supporting the role of TRAF family, TRAF6, TRAF4, and TRAF1 have been described to be overexpressed in U-CLL.²² Additionally, other coupregulated genes within peak 61315, such as CLIC3, FBXW5, and CCDC183 have been associated with poor prognosis in human cancers.^23-25 

Taken together, our systematic ChIP-seq analysis in stereotyped and nonstereotyped CLLs reveals that subset 8 cases are characterized by a distinct chromatin activation signature. This finding further supports the distinct biological nature of subset 8 and may help to understand the particularly aggressive clinical behavior of this peculiar CLL variant, including its predisposition to develop RT.

This work was partially developed at the Centro Esther Koplowitz (Barcelona, Spain).

This research was funded by the European Research Council under the European Union’s Horizon 2020 research and innovation program (Project BCLLATLAS, grant agreement 810287), Fundació la Marató de TV3, Generalitat de Catalunya Suport Grups de Recerca AGAUR 2017-SGR-736, CIBERONC (CB16/12/00225), the Accelerator award CRUK/AIRC/AECC joint funder-partnership, the Hellenic Precision Medicine Network in Oncology, and the project ODYSSEAS (Intelligent and Automated Systems for enabling the Design, Simulation, and Development of Integrated Processes and Products) implemented under the “Action for the Strategic Development on the Research and Technological Sector,” funded by the Operational Programme “Competitiveness, Entrepreneurship and Innovation” (NSRF 2014-2020) and cofinanced by Greece and the European Union.

Contribution: M.T. designed the research, analyzed the data, performed the study, and wrote the manuscript; V.C., B.G.-T., N. Pechlivanis, and F.N. provided data and assisted in data interpretation; N.R. performed the ChIP-seq libraries; N. Papakonstantinou, A.C., F.P., and E.C., assisted in research, data analysis, and interpretation; N.S. provided patient samples; and J.I.M.-S. and K.S. designed, supervised the research, and wrote the manuscript.

Conflict-of-interest disclosure: The authors declare no competing financial interests.

Correspondence: José Ignacio Martín-Subero, Institut d'Investigacions Biomèdiques August Pi i Sunyer, Universitat de Barcelona, Rossello 153, 08036 Barcelona, Spain; e-mail: imartins@recerca.clinic.cat; and Maria Tsagiopoulou, Centro Nacional de Análisis Genómico, Carrer de Baldiri Reixac, 4, torre i, 08028 Barcelona, Spain; e-mail: maria.tsagiopoulou@cnag.crg.eu.