Inactivating mutations of acetyltransferase genes in B-cell lymphoma

Abstract

Diffuse Large B-cell Lymphoma (DLBCL) represents the most common form of B-cell non-Hodgkin Lymphoma (B-NHL), accounting for ~30% of the de-novo diagnoses and also arising as a frequent clinical evolution of Follicular Lymphoma (FL)¹. The molecular pathogenesis of DLBCL is associated with multiple genetic lesions that segregate in part with individual phenotypic subtypes, namely germinal-center B-cell-like (GCB) and activated B-cell-like (ABC) DLBCL, suggesting the involvement of distinct oncogenic pathways^2–8. However, the full spectrum of lesions that contribute to malignant transformation remains unknown. Genome-wide efforts toward the identification and functional characterization of the entire set of structural alterations present in the DLBCL genome are required for a complete understanding of its pathogenesis⁹.

Toward this end, we have integrated next generation whole-exome sequencing analysis of 7 DLBCL cases and genome-wide high-density single nucleotide polymorphism (SNP) array analysis of 72 DLBCL cases. This combined approach led to the identification of >450 loci that are affected by somatic point mutations and/or by recurrent, focal gene copy number (CN) aberrations. Among those that have been so far independently validated, the most commonly involved regions were those harboring the acetyltransferase genes CREBBP (CBP) and EP300 (p300). CREBBP encodes a highly conserved and ubiquitously expressed nuclear phosphoprotein that, together with the closely related protein EP300, belongs to the KAT3 family of histone/protein lysine acetyltransferases^10,11. CREBBP and EP300 function as transcriptional coactivators for a large number of DNA-binding transcription factors involved in multiple signaling and developmental pathways, by modifying lysine residues on both histone and non-histone nuclear proteins^12,13. CREBBP and EP300 enhance transcription by multiple mechanisms, including: i) targeted acetylation of chromatin^12,13; ii) acetylation of transcriptional activators (e.g., the tumor suppressors p53 and GATA1)^14–17; and iii) acetylation-mediated inactivation of transcriptional repressors (e.g. the DLBCL-associated oncogene BCL6)¹⁸. Additionally, both molecules were found to exhibit ubiquitin ligase activity^19,20. Consistent with the involvement in critical cellular functions, homozygous null mice for either Crebbp or Ep300 are early embryonic lethal^21,22, and the same is true for the compound Crebbp/Ep300 double heterozygous mice^21,22. Haploinsufficiency of CREBBP (and, rarely, EP300) is responsible for the Rubinstein-Taybi syndrome (RTS), an autosomal congenital disorder characterized by mental and growth retardation, skeletal abnormalities and predisposition to tumor development^23–25. Somatic mutations of these two genes are exceedingly rare in epithelial cancers^26–28, and only three cases were reported to carry EP300, but not CREBBP mutations in hematologic malignancies^29,30. Additionally, chromosomal translocations of CREBBP are associated with an infrequent type of acute myeloid leukemia (AML) and with therapy-related AML and myelodysplastic syndrome, although the precise consequences of these lesions have not been fully elucidated^31–33.

Monoallelic lesions of CREBBP in DLBCL

Following initial observations from whole-exome sequencing analysis of 7 DLBCL cases and paired normal DNAs, we performed targeted re-sequencing of the entire CREBBP coding exons in 134 DLBCL samples representative of major phenotypic subtypes. This analysis revealed a total of 34 sequence variants distributed in 30 samples and whose somatic origin was documented by analysis of paired normal DNAs (available in 8 cases)(Fig. 1a,b and Table S1). Of these variants, 17 (50%) were inactivating events, including nonsense mutations (n=9), frameshift insertions/deletions (n=7) and mutations at consensus splice donor/splice acceptor sites (n=1), which generate aberrant transcripts carrying premature stop codons. Based on their distribution along the CREBBP protein, these mutations are all predicted to cause the elimination or truncation of the HAT domain (Fig. 1a and Table S1). The remaining variants included 3 in-frame deletions and 14 missense mutations, primarily within the HAT domain (Fig. 1a and Table S1), suggesting that they may be functionally important (see below). While CREBBP mutations were identified in both DLBCL phenotypic subtypes, their frequency was significantly higher in GCB-DLBCL, where they account for ~32% of the cases (n=21/65) as compared to 13% in ABC/non-GC DLBCL (n=9/69; p<0.01)(Fig. 1c).

Open in a separate window

Figure 1

The CREBBP gene is mutated in DLBCL

a, Schematic diagram of the CREBBP gene (top) and protein (bottom). Exons are color coded according to the corresponding protein functional domains (CH, cysteine-histidine rich; KIX, CREB-binding; Bromo, bromodomain; HAT, histone acetyltransferase; Q, poly glutamine stretch). Color-coded symbols depict distinct types of mutations. b, Sequencing traces of representative mutated DLBCL tumor samples and paired normal DNA; arrows point to the position of the nucleotide change (amino acid change shown at the bottom). c, Distribution of CREBBP mutations in major DLBCL subtypes; on top, actual number of mutated samples over total analyzed.

High density SNP array analysis, available for 72 samples from the same panel, and FISH analysis revealed the presence of monoallelic deletions encompassing or internal to the CREBBP locus in 8 additional cases (5 GCB-DLBCL and 3 ABC/non-GC-DLBCL), and a homozygous deletion in one patient (Fig. 2a,b and Fig. S1). Notably, the loss of genetic material was smaller than 240Kb in 4 cases and, in two (2147 and 2043), only involved a limited subset of CREBBP exons, thereby identifying this gene as the specific target (Fig. 2a and Table S2). In two additional samples for which CN data were not available, direct sequencing analysis revealed a hemizygous missense mutation, reflecting either the loss of the second allele or copy neutral loss of heterozygosity (Table S1). When combining the sequencing data with the CN data, CREBBP mutations and deletions were found to be mutually exclusive in most samples, revealing a predominantly monoallelic distribution (n=33/39 cases) (Fig. 2c). Rare instances of biallelic genetic lesions include the homozygous loss, a missense mutation with loss of heterozygosity (2 primary biopsies), biallelic nucleotide substitutions (2 cases), and a frameshift deletion with missense mutation of the second allele in the OCI-Ly8 cell line. Thus, 29% of all DLBCL patients (n=39/134), corresponding to 41.5% of GCB- and 17% of ABC-DLBCL, harbor genomic alterations affecting the CREBBP gene (Fig. 2d).

Open in a separate window

Figure 2

Mutations and deletions of CREBBP are predominantly monoallelic

a, Map of the genomic region encompassing CREBBP. Blue lines below the map indicate the extent of the deletions identified in 9 DLBCL samples, with the dark blue segment corresponding to a homozygous loss. b, dChipSNP heatmap showing median-smoothed log2 CN ratio for 8 DLBCL biopsies harboring CREBBP deletions, and two normal DNAs (N). A vertical blue bar indicates the location of the CREBBP locus; in the red-blue scale, white corresponds to a normal (diploid) CN log-ratio, blue is deletion and red is gain. c, Allelic distribution of CREBBP genetic lesions in individual DLBCL samples. d, Overall frequency of CREBBP structural alterations in DLBCL subtypes (mutations and deletions, combined).

Frequent mutations of CREBBP in FL

We next analyzed various types of mature B-NHL, including FL, Burkitt Lymphoma (BL), Marginal-Zone Lymphoma (MZL) and Chronic Lymphocytic Leukemia (CLL). Mutations analogous to those found in DLBCL were frequent in FL (32.6%, with 16 events distributed in 15/46 cases), but not in other lymphoma types, suggesting a specific role in the pathogenesis of these two diseases (Fig. S2). This analysis also revealed the existence of several mutational hotspots at specific codons within the HAT domain, including R1446, also mutated in B-cell acute lymphoblastic leukemia (B-ALL)³⁴, Y1503 and D1435; in addition, a 3bp in-frame deletion causing the loss of S1680 was observed in 5 cases, suggesting a functional role for this presently uncharacterized serine (Table S1). None of these recurrent changes were detected in paired normal DNA, thus excluding that they represent germline polymorphisms. While CN data were not available for the same FL panel, array-CGH analysis performed on a distinct dataset showed deletions spanning the CREBBP locus in only 1/68 cases (not shown). Collectively, these findings identify somatic mutations of CREBBP as a common event in FL.

EP300 mutations in DLBCL and FL

Given the significant structural and functional similarities between CREBBP and EP300, we investigated whether this second member of the KAT3 acetyltransferase family is also targeted by structural alterations in B-NHL. Mutational analysis of the same panel identified 19 sequence variants leading to amino acid changes (n=11), in-frame deletions (n=2), and premature stop codons due to frameshift deletions, aberrant splicing or nonsense mutations (n=6) (Fig. S3a and Table S3). These lesions were found in 10% of DLBCL (n=13/134) and 8.7% of FL samples (n=4/46), but were virtually absent in other B-NHLs (Fig. S3b). Seven additional DLBCLs harbored monoallelic deletions spanning, although not limited to, the EP300 locus (Fig. S3c, Table S4 and data not shown). Notably, structural alterations of CREBBP and EP300 co-exist in only a minority of the affected cases (n=6/53 DLBCL and 0/19 FL) (Fig. S3d), suggesting that inactivation of these loci is at least in part functionally equivalent. Thus, overall ~39% of all DLBCL and at least 41% of FL cases (based on mutations only) display structural alterations of KAT3 family genes.

CREBBP and EP300 protein expression

In order to compare the CREBBP and EP300 protein levels in normal and transformed B-cells, and to investigate the expression of the retained normal allele in cases carrying monoallelic genomic alterations, we examined the expression pattern of these two proteins in reactive human tonsils and in 78 DLBCL primary cases, of which 49 harbored both genes in wild-type configuration, by immunofluorescence (IF) and immunohistochemical (IHC) analysis. Consistent with their reported ubiquitous expression, CREBBP and EP300 were detectable in naïve B-cells within the mantle zone and, at higher levels, in germinal center (GC) B-cells (Fig 3a). Notably, most of the monoallelically-deleted DLBCL cases were positive for the two proteins, although at reduced levels, demonstrating that the residual wild-type allele is expressed (see Fig.3b,c for representative examples). Furthermore, RT-PCR amplification and direct sequencing of the ten CREBBP-mutated cell lines invariably showed the presence of the wild-type allele (Fig. S4), while western blot analysis using antibodies directed against the N-terminal portion of CREBBP revealed the expression of a full-length protein, corresponding to the wild-type allele, in most of the cell lines carrying truncating gene mutations, as well as in the monoallelically-deleted SUDHL5 cells (Fig. 3d, top panel). Slightly different findings were observed for EP300 where, in contrast with the primary biopsies, the presence of truncating mutations or deletions was associated with the absence of protein expression in four affected cell lines (Fig. 3d, second panel from top). With the exception of SUDHL2, both CREBBP and EP300 mRNAs were readily detected by northern blot analysis in all lines examined (Fig 3d, panel 4 and 5 from top).

Open in a separate window

Figure 3

CREBBP and EP300 expression in normal and transformed B-cells

a, Immunofluorescence analysis of reactive tonsils. BCL6 identifies GC B-cells, and Dapi is used to detect nuclei. b, c, Immunohistochemistry analysis of CREBBP (b) and EP300 (c) protein expression in representative DLBCL biopsies (genomic status as indicated; scale bar, 100µm). Sample 2147, which harbors a homozygous CREBBP deletion, serves as negative control. d, Western blot and northern blot analysis of DLBCL cell lines carrying wild-type or aberrant CREBBP and EP300 alleles (color coded as indicated). The aberrant band in SUDHL10 corresponds in size to the predicted ~220kD CREBBP truncated protein. * non-specific bands. Tubulin and GAPDH control for total protein and RNA loading, respectively. e, Overall proportion of DLBCL biopsies showing defective CREBBP/EP300 function due to genetic lesions (red scale) and/or lack of protein expression (blue scale).

Interestingly, a few additional lines were found to express very low to undetectable protein levels, despite the presence of mRNA and the absence of structural alterations in the corresponding gene (see CREBBP in BJAB and SUHDL2, or EP300 in SUDHL7) (Fig. 3d). Similarly, 6/78 (8%) DLBCL biopsies from patients with intact alleles appeared to lack expression of the two proteins, either simultaneously (3 cases) or individually (2 CREBBP+/EP300− and 1 CREBBP−/EP300+ cases)(Fig. 3c,e). Thus, the fraction of DLBCL patients with defective CREBBP and/or EP300 may be higher than that determined based on genetic lesions alone, suggesting that alternative mechanisms of KAT3 gene family inactivation may play a role in this disease.

CREBBP missense mutants fail to acetylate BCL6 and p53

While the presence of gross gene deletions and the distribution of truncating mutations clearly predict a complete gene inactivation or the loss of multiple key functional domains, including the HAT, the functional consequences of the numerous CREBBP missense mutations required direct experimental analysis. Notably, of the 30 total events identified, 27 cluster within HAT coding exons, suggesting a selective pressure to alter the CREBBP enzymatic activity (Table S1). In particular, 19 mutations were located within a 68 amino acids stretch that is 96% identical to EP300 and includes the contact surface for coenzyme A (CoA)³⁵ (Fig. S5). We therefore examined the effect of these mutations on the ability of CREBBP to acetylate known physiologic substrates. We selected BCL6 and p53 because of their biological relevance for GC development and lymphomagenesis^4,36. In fact, acetylation of the tumor suppressor p53 is indispensable for its transcriptional activity^14,15,37, while EP300-mediated acetylation of the proto-oncoprotein BCL6 leads to inactivation of its transcriptional repressor function¹⁸.

Transient transfection/co-immunoprecipitation assays confirmed that, analogous to EP300¹⁸, CREBBP binds to and acetylates BCL6, leading to a dose dependent impairment in its ability to repress a BCL6-responsive reporter gene (Fig. S6a,b). We then generated HA-tagged constructs for expression of 9 representative CREBBP alleles harboring missense mutations within (n=6) or immediately outside (n=2) the core HAT domain, as well as a premature stop codon (R1360X) (Fig. 4a). Notably, all of the HAT domain mutant proteins had lost their ability to acetylate BCL6 (Fig. 4b) and to interfere with its transrepression activity, consistent with the reported role of acetylation in inactivating BCL6 (Fig 4c). Conversely, no significant effects were observed from the C1240R and K1320R proteins or from two additional alleles (P1053L, Q1079H) harboring mutations outside the HAT domain (Table S1 and data not shown), although C1240R appeared to have lost its activity in the reporter assay, suggesting that alternative mechanisms may be involved.

Open in a separate window

Figure 4

CREBBP missense mutations impair its ability to acetylate BCL6 and p53

a, Schematic diagram of the CREBBP HAT and CH domains, with the CREBBP point mutations tested in b–d (in green, residues located immediately outside the HAT domain). b, Acetylation levels of exogenous BCL6 in Flag immunoprecipitates obtained from HEK293T cells co-transfected with wild-type or mutant CREBBP expression vectors. Actin, input loading control. c, Luciferase reporter assays using a synthetic 5X-BCL6 reporter. Results are shown as relative activity compared to the basal activity of the reporter, set as 1 (mean +/− SD, n=2). Bottom, BCL6 and CREBBP-HA protein levels in the same lysates. Note that the amount of transfected BCL6 and CREBBP-encoding plasmids was adjusted to achieve equal protein amounts. d, p53 acetylation in HEK293T cells co-transfected with the indicated CREBBP expression vectors. The anti-p53 antibody documents comparable amounts of p53 (exogenous+endogenous). GFP monitors for transfection efficiency, and actin is used as loading control.

When tested on the tumor suppressor p53, the same core HAT mutants were either impaired or severely attenuated (H1487Y and D1435E) in their acetylation activity, as compared to wild-type CREBBP or to C1240R and K1320R (Fig. 4d). It should be noted that C1240R and K1320R represent a second mutational event in cases carrying an additional truncating CREBBP mutation (see Table S1), suggesting the possibility that they represent passenger events or they confer more subtle functional consequences. Taken together, these findings demonstrate that CREBBP missense mutant alleles have been selected for their defective HAT activity. In particular, the impairment on BCL6 and p53 acetylation is consistent with specific effects of CREBBP genetic lesions in favoring the constitutive activity of the BCL6 oncogene over the function of the p53 tumor suppressor.

CREBBP mutations reduce affinity for Acetyl-CoA binding

To investigate the mechanisms responsible for the observed loss of function on BCL6 and p53, we examined the effect of CREBBP mutations on subcellular localization, enzyme-substrate complex formation, and enzymatic activity. All mutants tested were correctly localized in the nuclear compartment and could be efficiently co-immunoprecipitated with BCL6 or p53 (Fig. S7 and data not shown), indicating that the inability to acetylate these substrates in vivo was not due to mislocalization or to impaired physical interaction. Instead, in vitro acetylation assays using recombinant GST-p53 and highly purified, wild-type or mutated CREBBP proteins demonstrated a marked, Acetyl-CoA dose-dependent defect in the enzymatic activity of the core-HAT mutants (Fig. 5 and Fig. S8), suggesting that these changes impair the affinity for Acetyl-CoA (see also Fig. S5).

Open in a separate window

Figure 5

DLBCL-associated mutations in the CREBBP HAT domain decrease its affinity for Acetyl-CoA

a, Western blot analysis of in vitro acetyltransferase reactions performed using the wild-type or mutant Y1450C CREBBP-HA recombinant protein and a GST-p53 substrate in the presence of decreasing amounts of Acetyl-CoA. Anti-HA and anti-p53 antibodies document the presence of equivalent amounts of effector and substrate proteins in the reaction. b, In vitro acetyltransferase activity of the indicated CREBBP-HA mutant proteins in the same assay using 25nM Acetyl-CoA.

CREBBP missense mutants fail to rescue the phenotype of Crebbp/Ep300 null cells

To further investigate the consequences of CREBBP missense mutations on its function as a transcriptional co-activator, we examined the response of endogenous cAMP response element-binding protein (CREB) target genes in mouse embryonic fibroblasts carrying conditional Crebbp and Ep300 knock-out alleles (dKO MEFs), reconstituted with retroviral vectors expressing either wild-type or DLBCL-associated mutated Crebbp cDNAs (Fig. S9a), and treated with forskolin and IBMX (FI), two inducers of PKA activity. In normal cells, PKA-mediated phosphorylation of CREB is required for the recruitment of CREBBP, which in turn leads to transcriptional activation of cAMP-responsive genes³⁸. As previously observed³⁹, quantitative RT-PCR analysis of CREBBP/EP300-dependent CREB target genes confirmed their upregulation in dKO cells reconstituted with wild-type CREBBP expression constructs (Fig. S9b, column 2 from left). In contrast, all four DLBCL-derived CREBBP mutants tested were generally deficient for cAMP-responsive transcription, as was the HAT-dead control W1502A/Y1503S⁴⁰ (Fig. S9b and Fig. S10). These mutants were also associated with reduced endogenous histone H3K18 acetylation (Fig. S11). Finally, we tested whether the mutant CREBBP polypeptides were proficient in rescuing the proliferative defect of the dKO cells, as this system provides a specific readout for the biological activity of the CREBBP point mutants in a physiologic setting. Notably, cells transduced with the four DLBCL-associated alleles displayed significantly impaired cell growth compared to wild-type reconstituted cells, as measured in the YFP+ (dKO) population (Fig. S9c). Collectively, the data presented above provide direct experimental evidence for a role of HAT missense mutations in impairing CREBBP acetyltransferase activity and its function as a transcriptional coactivator.

Discussion

The results herein indicate that inactivation of CREBBP/EP300 represents a common event in the two most frequent forms of B-NHL, namely FL and DLBCL. Previous extensive surveys in malignancies of epithelial origin have reported inactivating mutations of EP300 and CREBBP in exceedingly rare cases (<2% of primary biopsies)^26–28. Considering their virtual absence in solid tumors, and the finding of recurrent mutations in B-ALL³⁴, our results point to a specific role for CREBBP/EP300 inactivation in the pathogenesis of malignancies derived from B-lymphocytes. Overall, CREBBP/EP300 lesions are among the most frequent structural alterations yet detected in FL and DLBCL, thus representing an important feature of the pathogenesis of these common diseases. Moreover, the observation of significantly reduced CREBBP and EP300 expression levels in a sizeable fraction of DLBCL samples, independent of genetic lesions (Fig. 3e and Fig. S12), suggests that additional epigenetic mechanisms may cause reduction of HAT dosage in a larger number of tumors.

One key observation of this study is that CREBBP/EP300 lesions are mostly detected in heterozygosity, suggesting a haploinsufficent role in tumor suppression. This notion is supported by several observations. First, congenital heterozygous mutations of CREBBP/EP300 are sufficient to cause significant pathologic and developmental phenotypes, including tumorigenesis, in patients with RTS^23–25, thereby confirming the deleterious effect of reduced HAT activity. Second, a fraction of mice with conditional deletion of Crebbp in mature B-cells showed reduced survival past 12-months of age⁴¹. Hematologic malignancies were also reported in 22% of constitutive 10–21 month-old Crebbp heterozygous mice and ~30% of chimeric animals after bone marrow or spleen cell transplantation from Crebbp heterozygotes⁴². Third, in contrast with the abundantly expressed histone deacetylases (HDACs), HATs are limiting in the cell, suggesting that small dosage variations can have severe biological consequences⁴³. Overall, these data provide direct evidence for CREBBP (and, more rarely, EP300) as haploinsufficient tumor suppressors, whose specific role in lymphomagenesis will have to be tested by conditional deletion of these alleles in GC B-cells.

Given the global involvement of CREBBP/EP300 on gene transcriptional regulation, it is difficult to predict which cellular targets/pathways may be critically affected by HAT reduction in lymphomagenesis. At this stage, our results demonstrate that mutant CREBBP and EP300 proteins are deficient in acetylating BCL6 and p53, leading to constitutive activation of the oncoprotein and to decreased p53 tumor suppressor activity. The balance between the activities of these two genes is critical for the regulation of DNA damage responses in mature GC cells during immunoglobulin gene remodeling^36,44. Thus, the consequences of BCL6 activity overriding p53 would be an increased tolerance for DNA damage in the context of diminished apoptotic and cell cycle arrest responses.

These results have important therapeutic implications in view of current attempts to utilize a variety of HDAC inhibitors as anti-cancer drugs. While the benefits of these compounds have been proven in certain malignancies, such as mature T-cell lymphoproliferative disorders, their efficacy in other common cancers, including B-NHL, is uncertain at this stage⁴⁵. The findings of this study suggest that the use of HDAC inhibitors has a rational basis in B-NHL, as it may contribute to re-establishing physiologic acetylation levels. On the other hand, their efficacy should be re-evaluated by stratifying patients based on the presence of HAT defects as well as by testing the numerous HDAC and sirtuin inhibitors with target specificity.

Methods Summary

Supplementary Material

Acknowledgements

Footnotes

References

References not cited in the main text