The CREBBP Acetyltransferase Is a Haploinsufficient Tumor Suppressor in B-cell Lymphoma Free

Modulation of chromatin accessibility through histone modifications is a key step in the regulation of gene transcription, and disruption of histone-modifying enzymes by genetic lesions has emerged as a common feature of cancer, including lymphoma (1, 2). Among these lesions, inactivating mutations and deletions of the gene encoding the CREBBP acetyltransferase (or, less frequently, its paralogue EP300) are highly recurrent in diffuse large B-cell lymphoma (DLBCL; 30% of cases) and follicular lymphoma (FL; 60% of cases), the two most common types of germinal center (GC)–derived B-cell non-Hodgkin lymphoma, together accounting for over 70% of all diagnoses (3–8). CREBBP mutations have also been observed in a third malignancy originating from the GC, Burkitt lymphoma (9), whereas they are generally absent in non–GC-derived mature B-cell malignancies such as mantle cell lymphoma, suggesting a prominent role for the perturbation of epigenetic mechanisms in the malignant transformation of GC B cells.

Genetic-driven losses of CREBBP and EP300 are predominantly monoallelic, mutually exclusive, and accompanied by expression of the residual wild-type allele, a pattern consistent with a haploinsufficient tumor-suppressor role (6). Indeed, a pathogenic effect for dose reduction of CREBBP/EP300 is demonstrated by the fact that germline loss of a single CREBBP allele by mutation or deletion is the causative genetic event in Rubinstein–Taybi syndrome, a rare autosomal congenital disorder that is also associated with tumor predisposition (10). Interestingly, phylogenetic analysis of tumor evolution during FL progression and transformation to DLBCL indicates that genetic lesions in epigenetic modifiers, including CREBBP and the methyltransferase KMT2D, are already present in a common precursor clone before divergent evolution to FL or DLBCL, suggesting a role early in the history of tumor clonal expansion (5, 8, 11).

CREBBP and EP300 are highly conserved, ubiquitously expressed enzymes that belong to the KAT3 family of acetyltransferases. They interact with over 400 proteins (12) and function as global transcriptional coactivators through the modification of lysines on both histone and nonhistone nuclear proteins, also including well-known proto-oncogenes (e.g., the BCL6 transcriptional repressor; ref. 13) and tumor-suppressor genes (e.g., TP53; refs. 14–16). In accord with their involvement in multiple cellular processes, constitutional homozygous-null mice for either Crebbp or Ep300 are early embryonic lethal, and the same is true for the compound Crebbp/Ep300 double heterozygous mice (17), consistent with the notion that the combined amount of these two proteins is limiting in the cell. Furthermore, although a partially redundant function has been invoked for CREBBP and EP300 during development, studies using conditional knockout mice indicate that, in certain cellular contexts, they can exert distinct roles (18–21). Nonetheless, a comprehensive investigation of the tissue-specific requirement for CREBBP is lacking.

In DLBCL and FL, CREBBP mutations (both truncating and missense in the HAT domain) impair its ability to catalyze acetylation of TP53 as well as to acetylate and inactivate BCL6, providing one mechanism by which loss of CREBBP activity may favor the malignant transformation of GC B cells (6). However, it is conceivable that reduced expression of CREBBP will have broad repercussions on gene transcription. Although several studies have examined its role during hematopoiesis, including early B- and T-cell development (18–21), the transcriptional network regulated by CREBBP in the unique environment of the GC, and the mechanism by which genetic-driven inactivation of its function contributes to their malignant transformation remain unknown. The aim of this study was to explore the role of CREBBP loss in the biology of normal and transformed GC B cells, by integrating functional epigenomics in human cells and mouse genetics approaches.

In order to define the genome-wide binding pattern of CREBBP in the GC, we performed chromatin immunoprecipitation and massively parallel sequencing (ChIP-seq) in two independent pools of purified human GC B cells (n = 3–5 donors/pool) with antibodies directed against CREBBP and, in parallel, against specific histone modifications (H3K4me1, H3K4me3, H3K27me3, and H3K27Ac) denoting well-characterized functional states of the bound chromatin. CREBBP-mediated histone acetylation is expected to be genome-wide (22), and, consistently, we identified 16,215 genomic regions (6,494 unique genes) that were significantly and reproducibly enriched in CREBBP binding in both biological replicates (P < 10⁻¹²; Fig. 1A). The vast majority of these regions (n = 12,440, 76.7%) were localized distal from the transcription start site (TSS) of the closest gene (5,170 intragenic, 32.0%; and 7,270 intergenic, 44.7%), suggesting possible association with enhancers, whereas only 3,775 (23.3%) were represented by proximal promoter regions (−2/+1 kb from TSS; Fig. 1A and B). CREBBP-bound regions were enriched in epigenetic marks of transcriptionally active chromatin, consistent with the notion that CREBBP acts to promote transcription via its acetyltransferase activity (Fig. 1B; ref. 17). In particular, we found significant overlap between CREBBP occupancy and H3K27Ac (73% of all chromatin-bound CREBBP), along with either H3K4me3 at TSSs, indicative of active promoters (n = 3,135 peaks, 19%), or H3K4me1 in the absence of H3K4me3 at TSS-distal sites, a feature of active enhancers (n = 7,372 peaks, 45%; refs. 23, 24; Fig. 1C). This chromatin profile suggests that CREBBP accesses DNA in nucleosome-free promoter/enhancer regions, possibly via binding to specific transcription factors.

Notably, 2,285 (31%) of these “active” enhancer peaks were located within predicted GC-specific super-enhancers (SE), as determined by the ROSE algorithm based on ranked ordering of H3K27Ac enrichment (see Methods; ref. 25; Fig. 1D). Indeed, virtually all GC SE (n = 649/655, 99%) were found to be loaded by CREBBP (Supplementary Table S1), including many that could be assigned to master regulators of the GC (namely BCL6, MEF2B, and MEF2C, among others; refs. 26, 27) or to molecules (e.g., CIITA and CD74) involved in MHC class II antigen presentation, a program recently found to be disturbed in FLs lacking functional CREBBP alleles (ref. 11; Fig. 1E and Supplementary Fig. S1). Collectively, these data indicate that CREBBP is preferentially associated with enhancer regions and suggest an important function for this acetyltransferase as a modulator of enhancer networks in GC B cells.

CREBBP and EP300 are ubiquitously expressed during B-cell development (Supplementary Fig. S2A–S2D) and presumably regulate distinct sets of genes in a context-specific manner. In order to determine functionally relevant targets regulated by CREBBP in the GC, we integrated the human ChIP-seq data with gene expression profile analysis of GC B cells purified from mice (Crebbp^fl/fl, Crebbp^fl/+, and Crebbp^+/+; ref. 28) in which the Crebbp gene can be conditionally deleted as early as 2 days after the initiation of the GC reaction via the Cγ1-Cre recombinase (29), following immunization with the T cell–dependent antigen sheep red blood cells (SRBC). We did not detect apparent differences in the fraction of GC B cells (B220⁺CD95⁺PNA^hi) across the three genotypes, even though efficient ablation of CREBBP expression was documented at both the RNA (not shown) and protein level (Fig. 2A and B). However, supervised analysis of gene expression data identified 1,343 transcripts that were differentially expressed in CREBBP-proficient versus CREBBP-deficient cells, of which 680 were downregulated and 663 were upregulated in the knockout (Student t test, P < 0.05, FC ≥1.2, FDR < 0.5; Fig. 2C and Supplementary Table S2). Because CREBBP-mediated acetylation facilitates active transcription, we focused on genes whose mRNA levels were reduced in the absence of Crebbp, and intersected these data with the list of genes that are bound by CREBBP, decorated by H3K27Ac, and expressed in human GC B cells, as determined by RNA-sequencing analysis of bulk CD77⁺ cells or by gene expression profiling of cells purified from the dark zone (DZ) and light zone (LZ) compartment. This approach revealed a restricted group of 243 genes that presumably represent bona fide direct targets of CREBBP, at least in steady-state conditions (Fig. 2D and Supplementary Table S3).

When interrogated for signatures from the Lymphochip database (30) and MSigDB (31), these genes were significantly enriched for functional annotations related to signal transduction programs that are central to GC B-cell physiology and pathology, including positive regulators of the B-cell receptor (FYN) and CD40 receptor (CD40), as well as downstream targets involved in NF-κB activity (NFKBIE, TNFAIP3, NFKB2), chemokine/cytokine responses (IL6, IL4RA, JUNB), lymphocyte migration, and previously identified targets of CREBBP in mature B cells, such as CIITA and BCL2a1 (refs. 11, 21, 27; Fig. 2E and F; Supplementary Fig. S3A; and Supplementary Table S4). Thus, despite its genome-wide occupancy, CREBBP deficiency alone appears to modulate specific functions of relevance to GC biology.

In particular, the signaling pathways uncovered above are known to be engaged in the LZ of the GC, where B cells receive specific cues to differentiate into plasma cells or memory B cells, or to recirculate into the DZ for additional rounds of mutations and selection (27, 32). We thus used the more sensitive Gene Set Enrichment Analysis (GSEA) to ask whether the CREBBP targets are preferentially upregulated during the DZ to LZ transition. Analysis of both human and mouse transcriptional signatures using the top 500 most significantly bound genes or the 243 core targets confirmed the significant enrichment of the CREBBP cistrome along the ranked gene expression changes associated with the GC LZ, and the same was true in the reverse analysis, using the list of LZ-upregulated genes (ref. 33; Fig. 2G and H; Supplementary Fig. S3B; and Supplementary Table S5). These data suggest that CREBBP is implicated in LZ-related functions, including recirculation, differentiation, and exit from the GC. Indeed, CREBBP binding was also found at the regulatory sequences of known master regulators of plasma cell differentiation, including two well-characterized elements located upstream to the TSS of both PRDM1 isoforms alpha and beta (Fig. 2I), and two putative regulatory regions located upstream and downstream of the IRF4 promoter, supporting the involvement of CREBBP in the decision of LZ B cells to enter terminal differentiation.

To verify this hypothesis in vivo, we investigated the GC B-cell response in Crebbp^fl/fl Cγ1-Cre mice analyzed upon SRBC immunization. No significant differences were detected between Crebbp-deficient and Crebbp-proficient littermates in the overall percentage of GC B cells as well as in the ratio between DZ and LZ B cells (Fig. 2J, left, and Supplementary Fig. S4A–S4D). Because epigenetic reprogramming may require multiple rounds of cell divisions to allow the replacement of modified histones by nonmodified histones (34), and the developmental stage at which CREBBP mutations are introduced in the precursor tumor cell is not known, we then tested whether loss of Crebbp at earlier stages of B-cell development had measurable effects on GC formation, by using CD19-Cre conditional knockout mice (35). As reported (21), Crebbp^fl/fl CD19-Cre animals showed significant reduction in overall percentage and number of splenic B220⁺ cells (not shown), with consequently reduced numbers of all mature B-cell compartments (Supplementary Fig. S4E–S4H and Supplementary Fig. S5A–S5D for a complete characterization of the GC/post-GC compartment). However, a mild but statistically significant increase could be appreciated in the DZ:LZ B-cell ratio of the Crebbp knockout mice, as compared with their wild-type littermates (P < 0.05, one-way ANOVA; Fig. 2J, right, and Supplementary Fig. S4H), indicating a selective defect in their relative redistribution at the expense of LZ B cells. Taken together, these data are consistent with the involvement of CREBBP in the orderly activation of transcriptional programs that operate in the DZ to LZ transition during the GC reaction.

A relevant nonhistone substrate of CREBBP-mediated acetylation in GC B cells is the BCL6 proto-oncogene, where this posttranslational modification acts as a negative modulator by disrupting its ability to recruit HDACs and to function as a transcriptional repressor (6, 13). Yet, GSEA of the CREBBP-bound regions also identified a significant enrichment for previously reported BCL6 direct targets (Fig. 2E; Supplementary Fig. S3A; and Supplementary Table S4; ref. 26). Indeed, intersection of the CREBBP and BCL6 ChIP-seq data obtained from the same pools of GC B cells revealed that 87% (6,088/7,961) of the BCL6 peaks were consistently co-occupied by CREBBP (P << 0.0001), both at TSS proximal and at more distal sites that may represent cisregulatory elements with enhancer features. In particular, most BCL6-bound peaks clustered in close proximity (100–1,000 bp) to CREBBP-bound chromatin, whereas CREBBP binding also occurred at a large number of additional loci, as expected (Fig. 3A and B). Integration of these data with transcriptomic analysis of human GC versus naïve and memory B cells pointed to a number of biological programs that are modulated by BCL6 and CREBBP in opposite fashion via binding to the same regulatory regions, including signal transduction from the B-cell receptor (BCR), Toll-like receptor, and CD40 receptor, NF-κB and Galpha13 signaling, T cell–mediated B-cell activation, and plasma cell differentiation (Fig. 3C and D and Supplementary Table S6). In addition, genes implicated in DNA repair (ATM, TP53), protection from apoptosis (BCL2, BCL2L1), and cell proliferation (the cell-cycle modulators CCND2 and CDKN1A) were also occupied by both proteins at promoter and/or enhancer sequences. Of note, many of the validated BCL6 target genes were decorated by both repressive and activating histone marks in a pattern consistent with a poised state, in agreement with their lack of expression in GC centroblasts at steady-state conditions, and with their rapid induction upon specific signals.

Overall, the data described above suggest that CREBBP and BCL6 exert opposing functions in modulating a defined subset of transcripts with key roles in the GC reaction, and point to a dual mechanism by which CREBBP may restrain the potentially deleterious activity of BCL6 by both acetylating and inactivating its protein, and by directly facilitating the transcription of its transrepressed genes.

The finding of CREBBP occupancy in the regulatory sequences of PRDM1 and IRF4 prompted us to test whether reduced dosage of CREBBP affects the ability of GC B cells to differentiate into plasma cells. Because the GC reaction cannot be recapitulated in vitro, we first used an ex vivo system to assess the response of B220⁺ cells isolated from Crebbp^+/+, Crebbp^fl/+, and Crebbp^fl/fl CD19-Cre mice to LPS and IL4, a mitogenic stimulus that induces proliferation as well as differentiation (Fig. 4A). We selected the CD19-Cre strain because the Cγ1-Cre mouse model requires activation of the Cre recombinase ex vivo, and the kinetics of the experiment may be too short to allow implementation of the epigenetic modifications that are necessary to induce phenotypic effects. In agreement with previous studies, flow cytometric analysis of CD138 expression performed at days 3 and 4 after stimulation documented the appearance in the Crebbp^+/+ population of cells expressing high levels of CD138 and low levels of B220, consistent with a plasmablastic phenotype (on average, 8.5% of cells). In contrast, only ∼3% of the Crebbp^fl/+ and Crebbp^fl/fl cells had upregulated CD138 (P < 0.05; Student t test; Fig. 4B and C). This difference was not due to diminished proliferation, which is also required for plasma cell differentiation (36), as the percentage of cells that had undergone cell divisions in the Crebbp^fl/fl population was even higher compared with wild-type cells (see next paragraph). Interestingly, qRT-PCR analysis of PRDM1 mRNA performed at days 2 through 4 after stimulation revealed significantly reduced levels in both Crebbp^fl/+ and Crebbp^fl/fl B cells, consistent with a dose-dependent role of Crebbp in its transcriptional activation (Fig. 4D).

In line with these results, flow cytometric analysis of the plasma cell compartment in Crebbp^fl/+ and Crebbp^fl/fl Cγ1-Cre mice, compared with littermate controls, uncovered a mild yet significant decrease in the percentage of splenic CD138⁺B220⁻ cells (P < 0.05, one-way ANOVA), which was accompanied by a trend toward reduced levels of serum IgG1 (Fig. 4E and F; see also Supplementary Fig. S5C and S5D for the analysis of the CD19-Cre cohort). The inability to detect differences in bone marrow plasma cells indicates that the phenotypic changes imposed by Crebbp loss in vivo are subtle and may be explained by the involvement of compensatory mechanisms (e.g., extra-GC plasma cells), along with the difficulty to quantitatively measure the GC-exiting plasma cell population (<1% of total lymphocytes) in the absence of a trackable marker, also considering the early time point (primary response) and the presence of residual cells that fail to delete Crebbp in both the Cγ1-Cre and CD19-Cre models (see Fig. 2B and Supplementary Fig. S4E). Nonetheless, these data suggest that a full complement of Crebbp may be necessary for optimal terminal differentiation of GC B cells, which is attained in part by remodeling the chromatin landscape of plasma cell master regulators and facilitating their accessibility to the transcriptional machinery.

To further investigate the biological consequences of CREBBP haploinsufficiency in the mature B-cell compartment, we measured the proliferative capacity of B220⁺ cells isolated from the spleen of CD19-Cre mice and stimulated with anti-CD40+IL4 (mimicking the signals delivered in vivo during T cell–dependent B-cell responses) or LPS+IL4. Flow cytometric analysis of CellTrace Violet–labeled cells at days 3 and 4 after stimulation revealed that a significantly higher number of cells had undergone cell division in both homozygous and, to a minor extent, heterozygous-null Crebbp cultures, compared with wild-type (on average, 4.7% and 12.2% nondivided cells, respectively, vs. 17.5% in the control mice; P < 0.05; Fig. 5A and B). Consistently, viability was greater in the Crebbp^fl/fl and Crebbp^fl/+ cultures as measured based on their metabolic activity (Fig. 5C and not shown). This finding seems to be due to an increased propensity of the Crebbp-deficient population to initiate proliferation following mitogenic stimuli, as indicated by the almost complete absence of nondivided cells in this culture compared with the wild-type control, whereas analysis of BrdUrd incorporation revealed a subtle yet measurable difference in the percentage of cells residing in the G₂ phase of the cell cycle (Fig. 5D and E). Also consistent with enhanced proliferative activity, class switch recombination to IgG1—a process known to require cell division—was significantly increased in Crebbp^fl/+ and Crebbp^fl/fl cells overall, but was comparable to that of control cells when normalized by the number of cell divisions (not shown). Finally, although GC B cells from Crebbp^fl/fl and Crebbp^+/+ Cγ1-Cre mice were undistinguishable in vivo, a measurable increase in the fraction of BrdUrd⁺ GC B cells could be detected in the CD19-Cre cohort (not shown).

Together, these results indicate that loss of Crebbp provides proliferative advantage to B cells, reflecting in part an augmented ability to respond/start cell division in response to stimulation, and suggest an additional mechanism by which tumor-associated inactivating mutations might contribute to tumor development in the mature B-cell lineage. Intriguingly, none of these effects were observed in the Cγ1-Cre background, despite the fact that deletion of the Crebbp protein was efficiently achieved within 48 hours of ex vivo stimulation (not shown), supporting the notion that cells need to complete multiple cell divisions in order to establish the epigenetic changes underlying this phenotype (see Discussion).

To investigate whether GC-specific haploinsufficiency of Crebbp promotes the development of lymphoma in vivo—either individually or in cooperation with other oncogenic lesions—we monitored the Cγ1-Cre and CD19-Cre cohorts for tumor incidence and survival.

We did not observe significant differences in event-free survival between Crebbp^fl/fl and Crebbp^+/+ Cγ1-Cre littermates over a follow up of 18 months, and although 3 of 22 (14%) heterozygous animals developed mature B-cell lymphomas (one FL and two DLBCL), this difference did not reach statistical significance (Fig. 6A, right). Loss of Crebbp alone was not sufficient to drive oncogenic transformation even when engineered at earlier stages of B-cell development in the CD19-Cre mouse model (Fig. 6A, left), suggesting that additional oncogenic events are involved in the clonal expansion of cells carrying CREBBP/EP300 mutations.

To test this hypothesis, Crebbp^fl/fl Cγ1-Cre mice were crossed with VavP-Bcl2 transgenic mice (37), which carry a deregulated BCL2 allele, thus mimicking the translocation that frequently co-occurs with CREBBP mutations in human FL and DLBCL (∼50% of mutated cases; Supplementary Fig. S6A–S6C). The compound animals were undistinguishable from VavP-Bcl2 littermates in terms of overall survival (Supplementary Fig. S7A and S7B), likely due to the glomerulonephritis and autoimmune diseases developing in this strain at nearly 100% penetrance, as reported (37). However, when combined with BCL2 deregulation, Crebbp haploinsufficiency led to a significant increase in the incidence of B-cell malignancies recapitulating molecular and phenotypic features of the human FL, including their derivation from GC B cells, as shown by the expression of the GC-specific marker BCL6 in the absence of post-GC markers (IRF4 and CD138; Fig. 6B), and the evidence of clonally rearranged immunoglobulin genes harboring somatic mutations (Fig. 6C and Supplementary Table S7). Overall, 92% (22/24) of Crebbp^fl/+ VavP-Bcl2 compound animals presented with in situ or overt FL in the spleen, lymph nodes, and frequently in the Peyer's Patches, as opposed to 16 of 26 (61.5%) of VavP-Bcl2 mice. Also of interest was the observation of rare B220⁻PAX5⁻BCL6⁻ lymphoproliferations showing plasmacytic morphology and IRF4 positivity (Supplementary Fig. S7C) in the VavP-Bcl2 mice (n = 3/26), but not in their compound Crebbp^fl/+ littermates, lending further support to the hypothesis that reduced dosage of CREBBP disturbs terminal differentiation.

Interestingly, GSEA of human DLBCLs using the CREBBP signature identified by our integrated approach showed a significant underrepresentation of these genes in tumors carrying CREBBP or EP300 genetic lesions, compared with CREBBP–wild-type biopsies, independent of molecular subtype (Fig. 6D), supporting the relevance of our model to the human disease and confirming a role for perturbations of this program in DLBCL pathogenesis.

Collectively, these data document a role for Crebbp as a bona fide haploinsufficient tumor-suppressor gene in GC B cells, whose reduced levels cooperate with Bcl2 deregulation to facilitate the development of B-cell lymphomas.

CREBBP is one of the most commonly mutated genes in GC-derived B-cell lymphomas (3–9), but the specific role of this acetyltransferase in the GC B-cell compartment and the mechanism by which its inactivation contributes to lymphomagenesis remain unknown. The results reported here establish CREBBP as a major regulator of GC enhancer networks, identify the transcriptional program that is sensitive to its dose reduction in this compartment, and demonstrate that CREBBP is a haploinsufficient tumor suppressor of GC-derived lymphomagenesis. These findings have implications for normal GC B-cell physiology, for the mechanisms leading to FL and DLBCL, and for the therapeutic targeting of these cancers.

Our results show that CREBBP binds extensively to the genome of GC B cells (∼16,000 peaks), and that most of these regions exhibit features of transcriptionally active (or poised) enhancers (38), including virtually all predicted GC-specific SE. These clusters of enhancers are known to provide cell-specific sites of open chromatin that can be accessed by signal-dependent transcription factors, and are thus primarily responsible for cell type–specific transcriptional responses (25, 39). Thus, CREBBP may act as a major regulator of enhancer networks in the specialized microenvironment of the GC, analogous to what has been established for the CREBBP/EP300 complex in embryonic stem cells (25).

In apparent contrast with this genome-wide distribution, deletion of Crebbp had modest consequences on global gene expression, both in terms of number of differentially expressed genes and of magnitude of the observed transcriptional changes, suggesting that most genes can tolerate reduced availability of this protein, at least in “nonstress” conditions. This finding may reflect in part the compensatory role of the CREBBP paralogue EP300. Indeed, although the single GC-specific Ep300 knockout model shows an analogous modest phenotype, Crebbp/Ep300 double knockout animals are completely unable to form GCs upon antigenic stimulation (data not shown). Nonetheless, we were able to identify a restricted set of genes that were sensitive to Crebbp depletion in vivo; this network was preferentially enriched in signal transduction programs that are normally suppressed by BCL6 in the GC DZ but become engaged in the LZ of the GC (e.g., BCR, NF-κB; refs. 27, 32), supporting a role for CREBBP dosage in cell fate specification and terminal B-cell differentiation in this compartment.

In this context, the almost complete overlap between the genome-wide mapping of BCL6 (a transcriptional repressor) and CREBBP (a positive regulator of transcription) suggests that a major role of CREBBP in the GC is to counterbalance BCL6-suppressive effects by directly binding to its targets and licensing their transcription. This model is further reinforced by the known ability of CREBBP to regulate BCL6 activity through acetylation-mediated inactivation of its protein function (6, 13). Together, these two mechanisms may ensure a continued modulation on BCL6 activity, allowing signal-dependent rapid reinduction of BCL6 target genes during GC exit. Nonetheless, the fact that multiple proteins have been identified as substrates of CREBBP enzymatic activity (e.g., p53), and the observed reduction in H3K27 and H3K18 acetylation levels in Crebbp-deficient B cells (ref. 6 and Supplementary Fig. S7D and S7E) are consistent with the involvement of both BCL6-dependent and BCL6-independent effects in the tumor-suppressor activity of CREBBP.

As previously observed with KMT2D, a methyltransferase whose activity is also genetically reduced in FL and DLBCL (4, 6), deletion of Crebbp early during B-cell development led to a more pronounced phenotype compared with that of mice where deletion occurs upon initiation of the GC (40). This may in part be explained by the fact that multiple rounds of cell division are thought to be necessary to implement the substitution of modified histones with nonmodified histones upon loss of Crebbp; alternatively, the cumulative effect of small changes in gene expression may be amplified by additional time and cell divisions, consistent with a model whereby CREBBP mutations occur early in the tumor ancestor clone.

Indeed, although individual loss of Crebbp upon initiation of the GC reaction or in pro-B cells was insufficient to induce the development of lymphoma, its heterozygous deletion accelerated tumor formation when combined with BCL2 deregulation, establishing this acetyltransferase as a bona fide haploinsufficient tumor-suppressor gene in FL and DLBCL. These findings are again analogous to what we previously observed with KMT2D (40) and, together with the observation that both KMT2D- and CREBBP-inactivating mutations represent early events in lymphomagenesis, suggest that CREBBP acts as a nonconventional tumor suppressor, the loss of which facilitates oncogenic transformation by establishing a permissive epigenetic environment for the accumulation of additional oncogenic perturbations.

In the context of the GC, our in vitro data suggest that one mechanism by which dose reduction of Crebbp may contribute to lymphomagenesis is by favoring proliferation at the expense of differentiation. These alterations are reflected by the in vivo phenotype of Crebbp-depleted GCs, which show an expansion of the DZ (i.e., the site of intense proliferation), at the relative expense of the LZ, where cells are induced to differentiate. In addition, CREBBP inactivation may facilitate malignant transformation by weakening the transcriptional responses of several tumor suppressors associated with FL/DLBCL pathogenesis, including TNFAIP3, GNA13, and S1PR2, while perturbing antigen presentation processes that are mediated by MHC class II molecules (Supplementary Fig. S1).

The reason why FL and DLBCL almost invariably retain one normal CREBBP allele is unclear, but may be related to the need for a minimum dose of CREBBP/EP300 activity in the cell. This notion is supported by the analogous monoallelic distribution of EP300 somatic mutations (found in a smaller subset of DLBCL and FL; ref. 6) and germline alterations of CREBBP/EP300 in patients with Rubinstein–Taybi syndrome (10), as well as by their mutually exclusive pattern in these diseases. Interestingly, monoallelic CREBBP/EP300 alterations are often concurrent with inactivating mutations of the methyltransferase KMT2D (∼60% of FL and 40% of mutated DLBCL), and these two histone modifiers are often found at the same genomic locations in GC B cells (not shown), suggesting cross-talk between their enzymatic activity, as reported for EP300 and the SET1 methyltransferase (41).

Finally, the identification of the transcriptional program perturbed by reduced dosage of CREBBP may provide a framework for the development of combinatorial therapies aimed at reestablishing physiologic acetylation levels (e.g., by using deacetylase inhibitors, or compounds that can stimulate histone acetylation) while modulating signaling pathways (e.g., BCR and NF-κB) that are concurrently dysregulated in these tumors. Moreover, the EP300 addiction presumably created by CREBBP loss in tumor cells could be exploited for therapeutic targeting. In this context, the animal model described here represents a useful tool for the preclinical testing of novel therapeutic approaches.

Tonsils were obtained as residual material from routine tonsillectomies at the Children's Hospital of Columbia-Presbyterian Medical Center, in compliance with Regulatory Guideline 45 CFR 46.101 (b)(4) for Exempt Human Research Subjects of the U.S. Department of Health and Human Services, and according to protocols approved by the Institutional Ethics Committee. Human GC centroblasts, naïve B cells, and memory B cells were isolated from reactive tonsils as described (42).

Purified GC B cells were cross-linked with 1% formaldehyde for 10 minutes at room temperature, quenched by the addition of glycine to a final concentration of 0.125 mol/L, and frozen. To minimize potential biases between individual donors, ChIP was performed on two independent biological replicates, each consisting of 3 to 5 GC B-cell isolations. Cell lysis and nuclei isolation was performed using the truChIP Chromatin Shearing Kit High Cell SDS (Covaris). Nuclei were sonicated using the S220 Ultrasonicator (Covaris) in order to obtain chromatin fragments of 200 to 500 bp. Sheared chromatin was incubated overnight with 4 μg of anti-CREBBP (A-22 rabbit polyclonal; Santa Cruz Biotechnology), anti-H3K27Ac (Active Motif), anti-H3K4me3 (Abcam), or 2 μg of anti-H3K4me1 (Abcam). Protein A magnetic beads were then added for 4-hour incubation followed by sequential washes at increasing stringency and reverse cross-linking. Upon RNAse and proteinase K treatment, ChIP DNA was purified using the MiniElute Reaction Clean Up Kit (Qiagen) and quantified by Quant-iT PicoGreen dsDNA Reagent (Life Technologies). The specificity of the antibodies against H3K4 methylation and H3K27 acetylation has been previously documented (40).

ChIP-seq libraries were constructed starting from 4 ng of ChIP or Input DNA, as reported in ref. 40. Libraries were quantified using the KAPA SYBR FAST Universal qPCR Kit (KAPA Biosystems), normalized to 10 nmol/L, pooled, and sequenced in an Illumina HiSeq 2000 instrument as single-end 100 bp reads, obtaining on average 25 × 10⁶ reads/sample.

Sequencing data were acquired through the default Illumina pipeline using Casava V1.8. Reads were aligned to the human genome (UCSC hg19) using the Bowtie2 aligner v2.1.0 (43), allowing up to two mismatches to cope with human variation. Duplicate reads (i.e., reads of identical length mapping to exactly the same genomic locations) were removed with SAM tools v0.1.19 using the rmdup option (44). Reads were normalized to total reads aligned (counts per million). Peak detection was done using ChIPseeqer v2.0 (45) enforcing a minimum fold change of 2 between ChIP and input reads, a minimum peak width of 100 bp, and a minimum distance of 100 bp between peaks. The threshold for statistical significance of peaks was set at 10⁻¹² for CREBBP and BCL6, and 10⁻¹⁵ for H3K4me1, H3K4me3, and H3K27Ac. The significance of the overlap between CREBBP-occupied regions and H3K27-acetylated regions was assessed by binomial test considering the total size (in bp) of the significant peaks identified in the two ChIP-seq experiments (e.g., CREBBP vs. H3K27Ac) and the size of the overlapping region between the two experiments, relative to total genome size of the UCSC hg19 assembly. The BCL6 ChIP-seq analysis is described in ref. 46. The ChIP-seq data have been deposited in the Gene Expression Omnibus (GEO) database under accession number GSE89688.

Significant CREBBP-bound regions (peaks) were classified as active promoters if located within −2/+1 kb from the TSS of an annotated gene and enriched in H3K4me3, but not H3K27me3. Intergenic and intragenic (exonic or intronic) regions were considered as active enhancers if located distal to TSS sites and enriched in H3K4me1 and H3K27Ac, but not in H3K4me3 (38). For the identification of SE in GC B cells, we applied the ROSE algorithm (25) to our H3K27Ac ChIP-seq dataset. Occupancy of CREBBP at SE was then determined based on the overlap between CREBBP peaks and genomic regions identified by ROSE.

CREBBP-bound enhancers, identified as described above, were assigned to the nearest expressed and transcriptionally active gene (distance from enhancer center to TSS) as the most likely candidate target gene (40).

The conditional Crebbp knockout mouse model (mixed C57BL/6 × Sv129 background) has been reported (28). Deletion of Crebbp was directed to GC B cells or precursor B cells by breeding Crebbp^fl/+ mice with the Cγ1-Cre (29) or CD19-Cre (35) deleter strain (both in pure C57BL/6 background), followed by offspring intercrossing. Compound mice carrying the conditional Crebbp x Cγ1-Cre allele and the VavP-Bcl2 transgene were obtained by breeding with VavP-Bcl2 mice (C57BL/6 background; ref. 37). A sample size of ≥23 animals/genotype was calculated to ensure over 80% power to detect an increase in tumor formation in the Crebbp compound lines at a P < 0.05 significance. No randomization was used. All animal work was performed under NCI and Columbia University Institutional Animal Care and Use Committee–approved protocols. Genotyping was performed by PCR analysis, and the protocol is available upon request. Animals were monitored for tumor incidence and survival twice/week over a period of 18 months, and sacrificed for analysis when visibly ill or when they reached end point. Kaplan–Meier event-free survival curves were generated using the GraphPad Prism 5 software (GraphPad Software), and statistical significance was calculated using the log-rank (Mantel–Cox) test. The Fisher exact test (two-tailed) was used to assess whether differences in the incidence of lymphoid malignancies between Crebbp-deficient mice and control wild-type littermates were significant.

For the analysis of T cell–dependent responses, age-matched 10- to 16-week-old mice were immunized by intraperitoneal injection of 500 million SRBC (Cocalico Biologicals) in PBS and analyzed after 10 days (12 days for the BrdUrd in vivo labeling experiment). Alternatively, mice were immunized by 0.5 mg/mL 4-Hydroxy-3-nitrophenylacetyl hapten conjugated to Keyhole Limpet Hemocyanin (NP-KLH; Biosearch Technologies) along with Complete Freund's Adjuvant (Sigma-Aldrich), and analyzed after 12 and 28 days. To achieve a higher yield of GC B cells for gene expression profile analysis, mice were immunized with two sequential SRBC injections (day 0, 1 × 10⁸; day 5, 1 × 10⁹) and sacrificed at day 12.

Multicolor flow cytometric analysis of the B- and T-cell lymphoid compartment was performed at 3 and 6 months of age as previously reported (40), using 3 to 4 mice/genotype/experiment. The B-cell compartment was also examined in all animals from the tumor-watch cohort, when sacrificed because visibly ill or at end of the study. Briefly, single-cell suspensions prepared from various lymphoid organs and from the peritoneal cavity were stained using different combinations of the following fluorescent-labeled antibodies: anti-B220 (RA3-6B2), -IgM, -IgD, -IgK, -CD21, -CD23, -IgG1, -CD138, -CD95, -Mouse Early B Lineage, -CD5, -MAC1, -CD4, -CD8, -CD43, -BCL6 (all from BD Biosciences); anti-CXCR4 and anti-CD86 (eBioscience); and fluorescein-labeled Peanut Agglutinin (PNA; Vector). The complete list of antibodies used for FACS analysis can be found in Supplementary Table S8. Samples were acquired on a FACSCanto II or on a FACSCalibur (BD Biosciences) and analyzed using FlowJo software (Tree Star).

Total IgM, IgG1, IgG2b, IgG3, and IgA serum titers were measured by ELISA as previously described (47), using the following antibodies: anti-mouse Ig H+L (1010-01; Southern Biotech), anti-mouse IgM-AP (1020-04; Southern Biotech), anti-mouse IgG1-AP (1070-04; Southern Biotech), anti-mouse IgG2b-AP (1090-04; Southern Biotech), anti-mouse IgG3-AP (1100-04; Southern Biotech), and anti-mouse IgA-AP (1040-04; Southern Biotech). To quantify NP-specific antibodies, serum was collected at day 12 after a single intraperitoneal injection of NP-KLH (100 μg) in Complete Freund's Adjuvant (Sigma). Analysis in Fig. S5D was performed on 5 Crebbp^+/+, 6 Crebbp^fl/+, and 6 Crebbp^fl/fl CD19-Cre mice; for IgG3 and IgA, one Crebbp^fl/fl animal was censored from the analysis due to insufficient material. NP-specific high-affinity antibodies were captured on plates coated with 2.5 μg/mL NP₂₅-BSA (Southern Biotech), and bound antibodies were detected with alkaline phosphatase–conjugated IgM-, IgG1-, IgG2b-, IgG3-, or IgA-specific antibodies (Southern Biotech). To quantify the total antibody serum levels in unimmunized mice, serum was collected at day 0 from Crebbp^fl/fl, Crebbp^fl/+, and Crebbp^+/+ CD19-Cre mice (n = 3 per genotype). Pan-mouse immunoglobulins were captured on plates coated with 2.5 μg/mL of goat anti-mouse Ig (H+L; Southern Biotech). Each assay was performed in duplicates using 12 (NP-KLH–immunized mice) or 7 (unimmunized mice) serial dilutions.

Splenic murine B cells were isolated from 10- to 16-week-old mice using the mouse B-cell isolation kit (Miltenyi Biotec) according to the manufacturer's instructions, plated at 2 × 10⁶ cells/mL, and cultured for 4 days in RPMI 1640 medium supplemented with 15% FBS, 55 μmol/L β-mercaptoethanol, 50 ng/mL recombinant mouse IL4 (R&D systems), and either 1 μg/mL Hamster monoclonal anti-CD40 (clone HM40-3; BD Pharmingen) or 5 μg/mL LPS (L263; Sigma). Cells were harvested at days 2, 3, and 4 and processed for flow cytometric analysis and RNA or protein isolation.

The number of viable cells in the ex vivo cultures was determined by measuring the levels of ATP, which signals the presence of metabolically active cells, using the CellTiter-Glo Luminescent Cell Viability Assay (Promega), according to the manufacturer's instructions. Cell proliferation was analyzed with the CellTrace Violet Cell Proliferation Kit (Life Technologies), which monitors distinct generations of proliferating cells by a fluorescent dye dilution. Data were acquired at days 2 to 4 on a FACSCanto II (BD Biosciences) flow cytometer with 405 nm excitation and an emission filter in the 450 nm range. All experiments were performed at least twice in triplicate, and the genetic background of the animals was verified by amplification and sequencing of a fragment in the immunoglobulin heavy chain locus, which encompasses a known SNP discriminating between the 129Sv and the C57BL6 strain. Analysis of BrdUrd incorporation was performed using the APC BrdU Flow Kit (BD Pharmingen), according to the manufacturer's protocol.

GC B cells (B220⁺CD95⁺PNA^hi) were sorted from Crebbp^fl/fl, Crebbp^fl/+, and Crebbp^+/+ Cγ1-Cre splenocytes (n = 4–5 mice/genotype) in a BD FACSAria III cell sorter (BD Biosciences). Total RNA was extracted using the NucleoSpin RNA II Kit according to the manufacturer's instructions (MACHEREY-NAGEL), and samples with RNA integrity numbers >9, as assessed on a BioAnalyzer 2100 (Agilent), were processed for cDNA synthesis. From each sample, 20 ng were reverse transcribed and amplified using the Ovation RNA Amplification System (NuGEN), followed by labeling with the Encore Biotin Module (NuGEN). Labeled samples were hybridized to Affymetrix MG 430 2.0 arrays according to the manufacturer's protocol. Raw expression values were normalized using the Robust Multiarray Averaging algorithm in GenePattern (48), and multiple probes corresponding to the same gene were collapsed to a single probe based on maximum t-statistic. Differentially expressed genes were determined by the Student t test using a P < 0.05, absolute fold change ≥1.2, and FDR <0.5 after the Benjamini–Hochberg correction. This lenient cutoff was allowed because the transcriptional changes imposed by CREBBP loss are expected to be modest, as reported in a number of studies performed in other cells/tissues, and intersection of the gene expression data with the list of genes bound by CREBBP provided further filtering to the analysis. The GEO accession number for these data is GSE88799. For visualization of gene-expression intensity, all gene expression data were normalized by gene and row (z-score transformation) with the GenePattern tool.

Gene expression profile data from Crebbp^fl/fl, Crebbp^fl/+, and Crebbp^+/+ Cγ1-Cre GC B cells were analyzed for enrichment in predefined sets of genes with GSEA-2.0 (49), using 1,000 gene set permutations and the Canonical Pathway Molecular Signature Database gene sets collection, or in the Lymphoid Signature Database (30). Multiple probes corresponding to the same gene were collapsed to a single probe based on maximum t-statistic. Enrichments were considered significant if FDR ≤0.15 after correction for multiple hypothesis. GSEA of the CREBBP core targets in human and mouse DZ versus LZ GC B cells was performed using a list of 243 genes that were bound by CREBBP and marked by H3K27Ac in human tonsils, conserved in the mouse, and downregulated in Crebbp^fl/fl versus Crebbp^+/+ Cγ1-Cre cells. This list is reported in Supplementary Table S3, and the gene expression data of DZ and LZ GC B cells have been previously published (33). To interrogate the enrichment of LZ-upregulated genes in the rank of genes differentially expressed between Crebbp^fl/fl and Crebbp^+/+ Cγ1-Cre GC B cell, we used the list of genes differentially expressed in DZ versus LZ subpopulations, identified as described in ref. 33, whereas enrichment in FOXO1-bound genes was performed using separately the list of the top 300 genes bound by FOXO1 in human GC B cells, the top 300 promoter-bound genes, and the integrated list of genes bound by FOXO1 and downregulated in Foxo1-deficient GC B cells (46).

To determine whether genes bound by CREBBP at active or poised promoters/enhancers and downregulated in Crebbp-deficient GC B cells were enriched in annotated functional categories, we overlapped the list of downregulated genes obtained by supervised analysis of mouse GC B cells (P < 0.05 and fold change ≥1.2) with genes identified by the integrated CREBBP/H3K27Ac ChIP-seq analysis of human GC B cells. This unique list of overlapping genes was then analyzed in the GSEA software (49) and the DAVID software (50). Genes were interrogated for enrichment in all Canonical Pathways and Gene Ontology annotation for biological processes, and clusters with a significant P value (<0.05 after Benjamini–Hochberg correction) were retained. Pathway analysis using the DAVID software was also performed with the list of genes cobound by CREBBP and BCL6.

Whole-cell extracts were obtained from purified mouse B cells using NP-40 lysis buffer according to a previously described protocol (13). For the analysis of histone modifications, the chromatin pellet was subsequently resuspended in radioimmunoprecipitation assay (RIPA)-coimmunoprecipiation (coIP) buffer (Tris-HCl pH 8.0 50 mmol/L, NaCl 200 mmol/L, glycerol 5%, MgCl₂ 2.5 mmol/L, sodium deoxycholate 0.05%, SDS 0.05%) and mobilized by sonication in a Bioruptor Standard sonication device (Diagenode). For the analysis of histone modifications from mouse frozen tumor biopsies, the chromatin pellet was resuspended in 0.2 N HCl and incubated overnight at 4°C. Extracts were cleared by centrifugation at 12,000 rpm for 10 minutes.

Protein extracts were resolved on NuPAGE Tris-acetate 3%–8% gels (for CREBBP and EP300) or Tris-glycine 4%–20% gels (for histone H3; Life Technologies) and transferred to nitrocellulose membranes (GE Healthcare) according to the manufacturer's instructions. Antibodies used were rabbit polyclonal anti-CREBBP (A22; Santa Cruz Biotechnology), rabbit polyclonal anti-EP300 (N15; Santa Cruz Biotechnology), rabbit polyclonal anti-BCL6 (Cell Signaling Technology; #4242), mouse monoclonal anti–β-Actin (A5441; Sigma-Aldrich), mouse monoclonal anti–α-tubulin (clone DM1A; Sigma-Aldrich), rabbit polyclonal anti-H3K18Ac (Abcam; cat# ab1191), anti-H3K27Ac (Abcam; cat# ab4729), and rabbit monoclonal anti-Histone H3 (clone D1H2; Cell Signaling Technology). Loading of chromatin extracts (typically 0.5–3 μg) was adjusted to ensure comparable amounts of total histone H3. Quantitation of signal intensity was obtained in the ImageJ software by subtracting the background signal measured above each band from the signal measured in each band; areas of the same size (set on the image of the wild-type protein) were used for all measurements. Values are expressed as fold differences relative to the wild-type protein sample, set at 1, after normalization for the loading control.

Histologic analysis of mouse organs was performed on 4-μm-thick formalin-fixed paraffin-embedded (FFPE) tissue sections, stained with hematoxylin and eosin (Thermo Scientific) following standard procedures. The following primary antibodies were used for immunohistochemical analysis: anti-BCL6 (1:300; N3, rabbit polyclonal; Santa Cruz Biotechnology), anti-PAX5 (1:400; rabbit polyclonal; Neomarker), anti-CD3 (1:800; rabbit monoclonal, clone SP7; Neomarker), anti-B220 (1:400; rat monoclonal, clone RA3-6B2; BD Biosciences), anti-IRF4 (1:200; M-17, goat polyclonal; Santa Cruz Biotechnologies), anti-BCL2 human (1:100; rabbit polyclonal; Santa Cruz Biotechnologies), and anti-CD138 (1:200; rat monoclonal, clone 281-2; BD Biosciences).

Double-immunofluorescence analysis of CREBBP and PNA was performed on FFPE material from mouse spleens, using a combination of two anti-CREBBP antibodies (each at 1:400 dilution; A22 and C20; Santa Cruz Biotechnology) and biotinylated PNA (1:300 dilution; Vector Laboratories; cat# B-1075). A serial section was stained with anti-EP300 (1:200 dilution; N15; Santa Cruz Biotechnology). Detection of CREBBP and EP300 was obtained using the EnVision System–HRP-Rabbit antibody (Dako) followed by Tyramide Signal Amplification system (PerkinElmer); Avidin NeutrAvidin, FITC conjugate (cat# A2662; Invitrogen) was used at 1:300 dilution to detect PNA.

Lymphoproliferative diseases developing in the VavP-Bcl2/Crebbp/Cγ1-Cre cohort were diagnosed based on morphology and phenotype according to the following criteria, in analogy to the classification of human lymphoma: (i) early stages of FL (FL in situ), defined by the presence of oversized, often coalescent follicles with partial or absent mantle zone and loss of confinement in the context of a yet preserved tissue architecture; (ii) overt FL of various grades, characterized by the effacement of the nodal and/or splenic architecture by a proliferation of follicle center B cells with a follicular growth pattern occupying the medullary and/or paracortical areas; (iii) DLBCL, defined by the effacement of the lymphoid organ architecture due to the expansion of large cells, with occasional infiltration beyond the capsule into surrounding soft tissues. Lymphoid tumors with plasmacytoid features were also occasionally observed in the VavP-Bcl2 mice; these lymphomas did not express B220 and PAX5 as well as BCL6, but stained positive for IRF4 and showed clonally rearranged Ig genes, indicating their B lymphoid origin. The genotype of the animal was not disclosed to the pathologist (S. Hussein). Immunohistochemical analysis of B220/CD3, BCL6, PAX5, IRF4, BCL2, CD138, and Ki67 was performed as described in the previous section.

To assess the presence of clonal immunoglobulin gene rearrangements in lymphoid tissues presenting morphologic and histologic evidence of lymphoma, high molecular weight genomic DNA was extracted from frozen tissues by phenol-chloroform extraction and digested with EcoRI overnight at 37°C (4 μg). The digestion reaction was resolved on 0.8% agarose gel followed by transfer to nitrocellulose membrane according to standard procedures. Hybridization was performed at 37°C overnight using a JH4 specific probe, as reported (40).

The rearranged immunoglobulin VH sequences were amplified from mouse tumor DNAs by PCR as described (40). The Sμ region, a target of AID-mediated activity in cells undergoing class switch recombination, was also analyzed by PCR amplification and direct sequencing, using primers SμA (ATTCCACACAAAGACTCTGGACC) and SμD (CAGTCCAGTGTAGGCAGTAGA), and the following PCR conditions: 94°C for 30 seconds, 58°C for 30 seconds, and 72°C for 1 minute, for 30 cycles.

Total RNA was extracted from primary mouse lymphocytes by TRIzol (Life Technologies) and treated with DNAse prior to cDNA synthesis, which was performed using the SuperScript First-Strand Synthesis System (Life Technologies), according to the manufacturer's instructions. The ABsolute QPCR SYBR green mix (Thermo Scientific) was then utilized to amplify specific cDNA fragments in the 7300 Real Time PCR system (Applied Biosystems), with the following oligonucleotides: mGAPDH_qPCR_F1 (5′-CATGGCCTTCCGTGTTCCTA-3′); mGAPDH_qPCR_R1 (5′-GCGGCACGTCAGATCCA-3′); mPRDM1_qPCR_F2 (5′-GGCATTCTTGGGAACTGTGT-3′); mPRDM1_qPCR_R2 (5′-ACCAAGGAACCTGCTTTTCA-3′). Data were analyzed by the change-in-threshold (2^−ΔΔCT) method, using Gapdh as “housekeeping” reference gene. Results are represented as fold changes relative to the expression levels of the wild-type animals, arbitrarily set as 1.

The CREBBP mutation pattern in DLBCL, tFL, and FL was assessed by integrating data generated in our previous studies (6–8) with published information from exome sequencing and/or transcriptome sequencing studies (3–5, 11), resulting in a total of 243 DLBCL primary cases, 31 DLBCL cell lines, and 71 FL cases. Co-occurrence of CREBBP mutations/deletions and BCL2 translocations was assessed using only samples for which both data were available. Gene expression profile data were available for 104 DLBCL primary cases and were used for the GSEA analysis shown in Fig. 6D.

P values were calculated with the Student t test or the one-way ANOVA in the Graphpad Prism software, unless described otherwise. The significance of the overlap between genes downregulated in CREBBP-deficient GC B cells (n = 681) and genes bound by CREBBP, decorated by H3K27Ac, and represented in the mouse MG430.2 expression array (n = 243), was calculated by the hypergeometric distribution considering that the total number of unique genes represented in the MG430_2 array and mapped to human genes is 16,567.

The gene expression profile data of mouse GC B cells and the ChIP-seq data of human GC B cells have been deposited in the GEO database under accession numbers GSE67388, GSE67494, GSE88799, and GSE89688.

No potential conflicts of interest were disclosed.

Conception and design: R. Dalla-Favera, L. Pasqualucci

Development of methodology: J. Zhang, K. Basso

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): S. Vlasevska, V.A. Wells, S. Nataraj, R. Duval, S.N. Meyer, T. Mo, P.K. Brindle, L. Pasqualucci

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): J. Zhang, V.A. Wells, S. Nataraj, A.B. Holmes, R. Duval, S.N. Meyer, K. Basso, S. Hussein, L. Pasqualucci

Writing, review, and/or revision of the manuscript: J. Zhang, S. Vlasevska, V.A. Wells, A.B. Holmes, R. Duval, S.N. Meyer, K. Basso, P.K. Brindle, S. Hussein, R. Dalla-Favera, L. Pasqualucci

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): J. Zhang, S. Vlasevska, V.A. Wells, S. Nataraj, A.B. Holmes, R. Duval, L. Pasqualucci

Study supervision: L. Pasqualucci

We would like to thank U. Klein for discussions and critical reading of the manuscript; C. Scuoppo and N. De Silva for suggestions; W. Zheng for help with the initial analysis of CD19-Cre mice; H. Tang for assistance with the mouse analysis; the Flow Cytometry Shared Resource of the Herbert Irving Comprehensive Cancer Center at Columbia University for assistance with cell-sorting procedures; and Roxanne Ko at the JP Sulzberger Genome Center of Columbia University for help with high-throughput sequencing. We also thank S. Cory (Walter and Eliza Hall Institute of Medical Research, Parkville, Australia) for the VavP-Bcl2 mice, and K. Rajewsky (Max Delbrück Center for Molecular Medicine, Berlin, Germany) for the CD19-Cre and Cγ1-Cre mice.

L. Pasqualucci is supported by NIH grants R01-CA172492 and R21-CA192854, and a Columbia University Interprogrammatic Pilot Project award. R. Dalla-Favera is supported by NIH grants R01-CA164152, R35-CA210105, and P50-CA192937. J. Zhang was a fellow of the Lymphoma Research Foundation.