Crebbp loss cooperates with Bcl2 overexpression to promote lymphoma in mice

Abstract

CREBBP is targeted by inactivating mutations in follicular lymphoma (FL) and diffuse large B-cell lymphoma (DLBCL). Here, we provide evidence from transgenic mouse models that Crebbp deletion results in deficits in B-cell development and can cooperate with Bcl2 overexpression to promote B-cell lymphoma. Through transcriptional and epigenetic profiling of these B cells, we found that Crebbp inactivation was associated with broad transcriptional alterations, but no changes in the patterns of histone acetylation at the proximal regulatory regions of these genes. However, B cells with Crebbp inactivation showed high expression of Myc and patterns of altered histone acetylation that were localized to intragenic regions, enriched for Myc DNA binding motifs, and showed Myc binding. Through the analysis of CREBBP mutations from a large cohort of primary human FL and DLBCL, we show a significant difference in the spectrum of CREBBP mutations in these 2 diseases, with higher frequencies of nonsense/frameshift mutations in DLBCL compared with FL. Together, our data therefore provide important links between Crebbp inactivation and Bcl2 dependence and show a role for Crebbp inactivation in the induction of Myc expression. We suggest this may parallel the role of CREBBP frameshift/nonsense mutations in DLBCL that result in loss of the protein, but may contrast the role of missense mutations in the lysine acetyltransferase domain that are more frequently observed in FL and yield an inactive protein.

Figure 1.

Crebbp deletion leads to impairments in B-cell development. (A) An example of flow cytometry density dot plots showing B220 and IgM staining of bone marrow cells from mice older than 5 months. Plots are gated on viable single leukocytes based upon PI staining and scatter characteristics. Gating shows pre- or pro-B cells (B220low, IgM⁻), IgM⁺ B cells (B220⁺, IgM⁺), and B220^high B cells that can be defined as recirculating B cells based upon expression of IgM and IgD. A notable change in the IgM⁺ and B220^high B-cell populations can be observed across strains, with reduced numbers being associated with Crebbp deletion and being partially rescued by the EµBcl2 transgene. (B) A summary of total B-cell percentage in the bone marrow, measured as the percentage of B220⁺ cells among viable single cells, is shown for all strains. A 1-way ANOVA test showed a significant variance across strains in the dataset (P < .001). Post hoc testing (Tukey) revealed that this was driven by significantly higher numbers of total B cells in EµBcl2 mice compared with Cbp^WT/F (P = .011), Cbp^WT/Δ (P < .001), Cbp^Δ/Δ (P < .001), and Cbp^WT/Δ × EµBcl2 (P < .001) mice, but not between EµBcl2 and Cbp^Δ/Δ × EµBcl2 mice (P = .068). All other head-to-head comparisons were not significant (P > .05). (C) A summary of pre- or pro-B cells, measured as the percentage of B220^low IgM⁻ cells as a proportion of all B220⁺ cells, is shown for all strains from mice older than 5 months. There was no significant variance in this population across strains (1-way ANOVA, P = .652). (D) A summary of immature B-cell frequencies, measured as the percentage of B220⁺ IgM⁺ IgD⁻ cells as a proportion of all B220⁺ cells, is shown for all strains. A 1-way ANOVA test showed a significant variance across the dataset (P = .034) that was driven by significantly higher frequencies in the Cbp^Δ/Δ × EµBcl2 strain compared with the Cbp^WT/Δ strain (Tukey, P = .018). No other head-to-head comparisons were statistically significant in post hoc testing (P > .05). (E) A summary of recirculating B-cell frequencies, measured as the percentage of B220⁺ IgM⁺ IgD⁺ cells as a proportion of all B220⁺ cells, is shown for all strains. One-way ANOVA showed a significant variance across the strains in the dataset (P < .001). Post hoc testing (Tukey) revealed that this was driven by a significantly higher frequency of mature B cells in EµBcl2 mice compared with Cbp^WT/F (P = .001), Cbp^WT/Δ (P < .001), Cbp^Δ/Δ (P < .001), Cbp^WT/Δ × EµBcl2 (P < .001), and Cbp^Δ/Δ × EµBcl2 (P = .004) mice. (F) An example of flow cytometry density dot plots showing IgD and IgM staining of splenocytes from mice older than 5 months. Plots are gated on viable single leukocytes based upon PI staining and scatter characteristics. A reduction in the frequency of immature (IgM⁺IgD⁻), transitional (IgM^hiIgD⁺), and mature (IgM⁺IgD⁺) cells can be seen with Crebbp deletion, and this is not restored by the addition of the EµBcl2 transgene. Note that the frequencies for these populations, shown in panels I and J of and supplemental Table 2, are based upon additional gating for B220⁺ cells that are not shown in this figure. (G) Box plots show the B220⁺ cells, as a percent of all viable single cells, across the 6 strains. There was a significant variability among strains (1-way ANOVA, P = .007) that was driven by significantly lower frequencies in Cbp^WT/Δ compared with Cbp^WT/F (Tukey, P = .036) and Cbp^WT/Δ × EµBcl2 (Tukey, P = .048) strains. No other head-to-head comparison was significant. (H) Box plots show follicular (B220⁺CD21⁺CD23⁺) B cells, as a percent of B220⁺ cells, across the 6 strains. There was significant variance across the strains, driven by significantly lower frequencies in Cbp^WT/Δ, Cbp^Δ/Δ, Cbp^WT/Δ × EµBcl2, and Cbp^Δ/Δ × EµBcl2 mice compared with both Cbp^WT/F and EµBcl2 (Tukey, P < .01 for all head-to-head comparisons). (I) The frequency of immature (B220⁺IgM⁺IgD⁻) B cells, as a percentage of B220⁺ cells, are shown in box plots. There was significant variance across the strains (1-way ANOVA, P < .001) that included significantly lower frequencies in Cbp^WT/Δ (Tukey, P < .001) and Cbp^Δ/Δ (Tukey, P = .005) compared with Cbp^WT/F mice, and significantly lower frequencies in Cbp^WT/Δ × EµBcl2 (Tukey, P = .006), and Cbp^Δ/Δ × EµBcl2 (Tukey, P = .013) compared with EµBcl2 mice. (J) The frequencies of mature B cells (B220⁺IgM⁺IgD⁺), as a percentage of B220⁺ cells, are expressed in a box plot. There was significant variance across the strains (1-way ANOVA, P < .001) that was driven by significantly lower frequencies in Cbp^Δ/Δ, Cbp^WT/Δ × EµBcl2, and Cbp^Δ/Δ × EµBcl2 mice compared with both Cbp^WT/F and EµBcl2 mice (Tukey, P < .05 for all comparisons). There was no significant difference between Cbp^WT/Δ mice and either Cbp^WT/F (Tukey, P = .172) or EµBcl2 mice (Tukey, P = .082).

Figure 2.

GCB-cell formation following immunization. Mice from each strain were immunized with 1-2 × 10⁸ sheep red blood cells into the peritoneum, and the spleens analyzed for germinal center formation 10 days later. (A) Example contour plots show GCB-cell frequencies across the 6 strains that were analyzed. Plots are gated on viable (PI-negative) single B cells (B220⁺), and GCB cells are defined as peanut agglutinin (PNA)-positive and Fas-positive cells. (B) Box plots show the trends of GCB-cell development across 8 Cbp^WT/F, 8 EµBcl2, 5 Cbp^WT/Δ, 7 Cbp^Δ/Δ, 6 Cbp^WT/Δ × EµBcl2, and 6 Cbp^Δ/Δ × EµBcl2 mice. One-way ANOVA showed a significant variability across the strains (P < .001), but post hoc analysis (Tukey) showed that this was driven by significantly higher frequencies of GCB cells in mice carrying the EµBcl2 transgene compared with mice without the EµBcl2 transgene (P < .001 to .040). There was no significant difference between Cbp^WT/F and either Cbp^WT/Δ or Cbp^Δ/Δ mice, nor between EµBcl2 and either Cbp^WT/Δ × EµBcl2 or Cbp^Δ/Δ × EµBcl2 mice (P > .05).

Figure 3.

Crebbp deletion promotes B-cell lymphoma in tandem with Bcl2 overexpression. (A) A Kaplan-Meier plot shows the lymphoma-specific survival of 6 transgenic strains. It can be seen that deletion of 1 or both alleles of Crebbp leads to the development of lymphoma at a late time point in some mice, that the penetrance is increased and the latency is decreased with the addition of the EµBcl2 transgene. (B) An example of spleens and lymph nodes from Cbp^Δ/ΔxEµBcl2 mice that became ill, showing splenomegaly and lymphadenopathy that is indicative of lymphoma. (C) Hematoxylin and eosin staining from representative spleen and lymph node samples of age-matched tumor-free Cbp^WT/F and tumor-bearing Cbp^Δ/ΔxEuBcl2 mice. The B-cell lymphomas can be seen to be associated with a diffuse spread of centroblasts, with loss of normal architecture. Pathology review determined this specimen to be most similar to DLBCL histology.

Figure 4.

B-cell lymphomas in Crebbp transgenic mice are of GCB-cell origin. (A) An example of immunohistochemical staining for the spleen of an age-matched control mouse (Cbp^WT/F) and a lymphoma-involved spleen from a Cbp^Δ/ΔxEµBcl2 mouse. The control mouse shows normal benign follicles and an expected pattern of Pax5 and Bcl6 staining. The lymphoma-involved spleen shows diffuse histology with cells that are Pax5 and Bcl6 positive, supporting a GCB-cell origin. (B) Immunoglobulin rearrangements were assessed by PCR in DNA extracted from tumor-involved spleens and a spleen from age-matched control mice (Cbp^F/FxEµBcl2). Control mice showed a laddering pattern indicative of a polyclonal B-cell population, as shown in the left-most lane. In contrast, the majority of tumor samples from Cbp^WT/Δ, Cbp^Δ/Δ, Cbp^WT/ΔxEµBcl2, and Cbp^Δ/ΔxEµBcl2 mice showed a single dominant band that is indicative of a clonal B-cell population, as seen in these examples. Eight of these bands, highlighted in red boxes, were excised and cloned for sequencing. Analysis revealed the presence of somatic hypermutation in 7/8 tumors, with an average of 2% deviation (min = 0.8%, max = 4.2%) deviation from the germ-line V-gene sequence. This provides further evidence in support of the GCB-cell origin of these tumors, or that the B cells have previously transited through the germinal center. NA, not applicable; SHM, somatic hypermutation.

Figure 5.

Transcriptional changes associated with Crebbp deletion are not associated with gene-proximal H3K18Ac changes. (A) Gene expression profiling was performed on purified B cells from adult disease-free mice to determine early molecular alterations associated with disease etiology. Differential gene expression analysis revealed a signature of genes with significantly (false discovery rate < 0.25, fold change >1.2) reduced or increased expression associated with deletion of 1 or both alleles of Crebbp in the EµBcl2 background. This represents the intersection of head-to-head comparisons between mice with both alleles of Crebbp intact (Cbp^F/FxEµBcl2) compared with those with either 1 allele (Cbp^WT/ΔxEµBcl2) or 2 alleles (Cbp^Δ/ΔxEµBcl2) of Crebbp deleted and included increased expression of the Myc oncogene. The samples for which ChIP-seq was also performed are annotated at the bottom of the figure with roman numerals. (B) Increased expression of Myc was confirmed by immunohistochemical staining of spleens from Cbp^Δ/ΔxEuBcl2 mice compared with an age-matched control (Cbp^F/F). (C) Heat maps show the level of H3K18Ac from 10 kb upstream to 10 kb downstream from the TSS of genes with differential expression associated with Crebbp deletion. ChIP-seq was performed on the same samples that were interrogated by gene expression profiling, as annotated by corresponding roman numerals in panels A and C. The H3K18Ac level of each gene is aligned with the expression levels in panel A, and the average signal (line) ± the standard deviation (shaded region) is summarized for genes with reduced (blue) or increased (red) gene expression at the top of the heat map. There is a peak of H3K18Ac at the TSS of most genes with differential expression, showing our ability to detect H3K18Ac at TSSs. However, the changes in gene expression between strains were not associated with changes in H3K18Ac at the TSS ± 10 kb.

Figure 6.

Regions of differential H3K18Ac are primarily intragenic and are bound by Myc. (A) Ratio heat maps of the level of H3K18Ac from Cbp^Δ/ΔxEuBcl2 B cells compared with Cbp^F/FxEuBcl2 B cells from 2 unique biological and technical replicates. These heat maps show the intersection of significant differences identified in each replicate, which includes a small number of regions with significantly reduced peaks of H3K18Ac and a large number of regions with significantly increased peaks of H3K18Ac. (B) Regions of reduced and increased acetylation primarily affect intragenic regions that are distant from the nearest TSS. The distance from the center of the peak of significantly reduced (above) or increased (below) regions of acetylation to the TSS of the nearest gene is shown using a heat plot. It can be seen that the majority of the peaks of differential H3K18Ac lie very far from the nearest TSS, suggesting that they may be distant regulatory elements. (C) An example of motifs that were most significantly enriched in regions of differential H3K18Ac and showed homology to MYC binding sites is displayed. This provided evidence that regions of altered H3K18Ac may be bound by MYC. (D) Public ChIP-seq data for Myc from the Ch12 murine B-cell lymphoma and Mel murine erythroleukemia cell line show a peak of strong Myc binding at the same location as the peak of increased H3K18Ac observed in Cbp^Δ/ΔxEuBcl2 B-cells. Regions are aligned and show the same physical location as panel A. The average signal (line) ± 1 or 2 standard deviations (shaded region) is summarized at the top of the heat map. The strong peak at the top and Myc binding signal that aligns with the center of the peaks of increased acetylation in our transgenic mice support the binding of Myc to these regions and implicate Myc in the observed epigenetic alterations. (E) Due to the potential importance of Myc in the disease etiology, we confirmed that Myc was also expressed in tumor samples from our transgenic mice. All tumors showed Myc staining, but this was absent from age-matched control spleens. Illustrative examples are shown, with only rare positive cells visible in the control spleen (Cbp^WT/FxEµBcl2), but broad positive staining in the tumor-involved spleens from mice with 1 or both alleles of Crebbp deleted. Cent., center; cMyc, Myc protooncogene.

Figure 7.

Significantly different distributions of CREBBP mutations in human FL compared with DLBCL. (A) CREBBP mutation data were obtained for a total of 310 FL and 274 DLBCL in which both diseases were interrogated in the same study using the same approach. Mutations in CREBBP were identified in 151 FL and 51 DLBCL, and the relative representation of the position of these mutations are expressed as a fraction of all CREBBP mutations in that disease, relative to the protein position of CREBBP isoform 1. The KAT domain is shaded in green and defined as amino acids 1342 to 1649. Larger peaks, indicative of a higher fraction of all CREBBP mutations, are seen upstream of the KAT domain in DLBCL compared with FL. In addition, a dominant hotspot can be seen at arginine 1446 in FL that is significantly reduced in DLBCL. In contrast, other hotspots at tyrosine 1482 and 1503 are present at relatively similar frequencies in both FL and DLBCL. (B) Pie graphs show that 78% (118/151) of CREBBP mutations fall within the KAT domain in FL, as compared with only 43% (22/51) in DLBCL (Fisher’s exact test, P < .001). (C) Pie graphs show that only 17% (25/151) of CREBBP mutations in FL create a frameshift of premature stop codon, while the remainder creates single-amino-acid substitutions or insertion/deletions. In contrast, missense/frameshift mutations are present at greater than twice this frequency, 39% (20/51), in DLBCL (Fisher’s exact test, P = .0016). (D) Gene expression microarray data from DLBCL tumors with previously determined CREBBP mutational status were used to evaluate MYC expression. There was a significantly higher expression of MYC in tumors with CREBBP mutation compared with those with no CREBBP mutation (1-tailed Student t test, P = .026). This is despite the unknown MYC translocation status that may alter the expression of MYC in a subset of cases. (E) MYC translocation status was available for 54 cases with known CREBBP mutation status. We observed no overlap in CREBBP mutation and MYC translocation, although this was not statistically significant.