BCL6 suppression of BCL2 via Miz1 and its disruption in diffuse large B cell lymphoma

To identify the full set of BCL6 direct target genes, we performed a genomewide integrated transcriptional (gene expression profiling), biochemical (ChIP-chip), and bioinformatic (ARACNe algorithm) (36) analysis in purified normal GC B cells. This approach identified BCL2 as a novel candidate target gene. As shown in Fig. 1A, BCL6 binds to BCL2 in a region between the P1 and P2 promoters (37). This region contains numerous canonical BCL6 binding sites (B6BS) (1) and several putative Inr elements that could mediate BCL6 binding to BCL2 via Miz1, as previously described for the CDKN1A promoter (32). Indeed, a luciferase reporter construct driven by BCL2 promoter sequences spanning this region was efficiently repressed by BCL6 in transient transfection assays (Fig. 1B, solid bars). However, BCL6 could still repress a reporter construct where all canonical BCL6 binding sites had been mutated (Fig. 1B, shaded bars), suggesting that these sites are not the critical mediators of BCL6 transsuppression. We therefore tested whether transcriptional repression of BCL2 by BCL6 is mediated by Miz1, a known activator of BCL2 (38). Fig. 1C shows that BCL6 was capable of suppressing Miz1-induced activation of BCL2 in a dose-dependent manner, and this activity requires the presence of both the BCL6 transrepression domain and the ZF domain, which mediates BCL6-Miz1 physical interaction (32), suggesting that transcriptional repression of BCL2 by BCL6 acts through Miz1. Note that mutation of the Inr elements, useful to demonstrate their direct involvement in the BCL6/Miz1 regulation, was not informative because it abrogates BCL2 transcription completely. Notably, shRNA-mediated silencing of Miz1 in B cells impairs the ability of BCL6 to bind the BCL2 promoter (Fig. 1D, Right), despite the presence of comparable BCL6 protein levels (Fig. 1D, Left). Taken together, these results suggest that BCL6 binds to the BCL2 promoter in vivo via Miz1.

To demonstrate the physiologic significance of BCL2 suppression by BCL6, we analyzed the relative levels of BCL6 and BCL2 mRNA and protein in mature B cells by gene expression profile analysis of purified naive, GC, and memory B cells and double immunofluorescence analysis of normal lymphoid tissue sections. Consistent with previous data, BCL2 and BCL6 mRNA expression is mutually exclusive in B cells (Fig. 2A) (39). Accordingly, GC B cells, which uniformly express BCL6 at high levels, lack BCL2 protein expression, even when analyzed at the single-cell level (Fig. 2B). In addition, downregulation of BCL6 by CD40 and BCR activation in these cells leads to loss of BCL6 binding to the BCL2 promoter and upregulation of BCL2 expression (Fig. 2 C and D). This effect specifically depends on BCL6 because its shRNA-mediated silencing was sufficient to induce BCL2 expression in Ramos B cells (Fig. 2E). Overall, these results establish a physiologic and functionally relevant inverse relationship between BCL6 and BCL2 in GC B cells, supporting a direct mechanism for suppression of BCL2 by BCL6.

We then examined whether BCL6-mediated suppression of BCL2 is conserved in DLBCLs by characterizing the expression pattern (mRNA and protein) of these 2 factors in 148 samples representative of BCL2 translocated and nontranslocated cases. Immunohistochemical and Western blot analysis showed that BCL6 and BCL2 are coexpressed in a large fraction of DLBCLs, including 9/20 (45%) cell lines and 61/128 (48%) primary biopsies. Pathologic BCL2/BCL6 coexpression was found in the vast majority of cases carrying BCL2 translocations (7/9 lines and 16/21 biopsies) [supporting information (SI) Fig. S1 and Fig. 3], suggesting a role for the juxtaposed Ig enhancers in overriding BCL6-mediated repression (40). However, BCL6/BCL2 coexpression is also observed in 45/107 (42%) DLBCL cases lacking BCL2 translocations (Fig. 3B). These findings suggest that BCL6-mediated suppression of BCL2 is frequently disrupted in DLBCL, but this event can be explained by the juxtaposition of Ig regulatory elements only in a fraction of cases.

To characterize the BCL2 locus in cases apparently resistant to BCL6-mediated suppression, we proceeded to a detailed structural analysis of the BCL2 genomic sequences by DNA amplification and direct sequencing of ≈3 kb spanning the BCL6-bound region and its coding exon 2. This analysis revealed that, with rare exceptions, including 1 sample in which BCL2 was not expressed and 1 sample carrying a non-IgH translocation, all translocated cases display a massive load of mutations, whose frequency (0.77%) exceeds by far the one (0.11%) previously reported for the BCL2 coding region in FL and transformed DLBCL (Fig. 4 and Fig. S2A) (41, 42). Mutations include mostly single base pair substitutions, but also deletions, duplications, and insertions, and are not detected in matched normal DNA from the same individuals, thus constituting somatic events clonally represented in the tumor (see Table S1 for a detailed list). The characteristics of the mutations, and in particular their distribution downstream of the promoter, the positive transition/transversion ratio, and the preferential hotspot (RGYW) targeting, strongly suggest that they are caused by the SHM mechanism (43) (Table S2). In 3 informative cases, RT-PCR amplification and sequencing using specific primers documented that the mutations were restricted to the translocated allele, while the nontranslocated allele was unmutated and not expressed (not shown), suggesting a role for the juxtaposed Ig enhancer in recruiting the SHM machinery to these alleles (40). Interestingly, a significant, although smaller number of mutations with analogous features were also detectable in cases lacking BCL2 translocations (n = 19/94; mutation frequency, 0.027%) (Fig. 4 and Fig. S2A). Because no variants were observed in normal GC B cells (consistent with their lack of BCL2 transcription) (Fig. S2B), these findings indicate that BCL2 mutations in nontranslocated cases represent a tumor-associated phenomenon, presumably due to the aberrant activity of the SHM mechanism, present in most DLBCL (44). Of note, in 20/121 (14.2%) samples sequenced, mutations extend to the BCL2 coding domain, leading to amino acid substitutions with potential functional consequences. Thus, one-third (41/121) of DLBCLs, including both BCL2 translocated and nontranslocated samples, display somatic mutations of the BCL2 regulatory region.

To dissect the mechanism leading to BCL6 and BCL2 coexpression in DLBCL, we first analyzed whether BCL6 can still bind to the BCL2 promoter in these cases. Quantitative ChIP (qChIP) assays performed in 5 BCL6+BCL2+ DLBCL cell lines revealed the absence of BCL6 binding in 3 samples, as opposed to 2 control cell lines carrying an intact BCL2 locus and lacking BCL2 expression (Fig. 5A, Fig. S1). Interestingly, no Miz1 protein expression was observed in these 3 cases (Fig. 5A, Right), indicating that the inability of BCL6 to bind the BCL2 promoter was conceivably because of the absence of its interacting partner. Cases expressing significantly lower amounts of Miz1 mRNA were found also among BCL6+BCL2+ DLBCL primary biopsies (n = 5/61) (Fig. 5B), thereby identifying a subset of DLBCLs in which loss of BCL2 repression by BCL6 may be caused by the lack of Miz1 expression, preventing BCL6 from reaching the BCL2 promoter.

Analysis of Miz1 mRNA expression in primary DLBCLs also revealed the existence of cases where pathologic BCL2 and BCL6 coexpression is associated with significantly higher levels of Miz1 mRNA (Fig. 5B) (P < 0.001 as compared to cells displaying a normal inverse correlation between BCL6 and BCL2 expression). This subset was largely restricted to translocation-negative, BCL2 unmutated cases (n = 14/32, 43.8%, corresponding to ≈23% of all BCL2+BCL6+ DLBCLs), suggesting that exceedingly high levels of Miz1 may stimulate BCL2 expression via out-titrating BCL6. Thus, loss of BCL2 suppression by BCL6 may be due to both Miz1 downregulation and overexpression, the cause of which is presently unknown.

To directly assess whether BCL2 mutations affect BCL6/Miz1-mediated suppression of BCL2, we analyzed the response of DLBCL-derived mutant alleles in transient cotransfection/reporter gene assays. Alleles were selected among BCL2/BCL6 double-positive cases showing normal Miz1 expression and preserved BCL2 binding. Although the relative ability of BCL6 to repress Miz1-induced activation was comparable in all 6 alleles tested, 3 of them (2062, 2106, and SUDHL6) displayed significantly enhanced responses to Miz1-mediated transcriptional activation (2- to 3-fold), resulting in the overall increased expression of the reporter, as further evidenced by using increasing doses of BCL6 (Fig. 5C). Because the only variable in the experiment is the presence of mutations in the tested alleles, these results indicate that a subset of BCL2 mutations leads to increased response to Miz1. Given the complexity of the mutation pattern in these alleles, some of which carry >50 mutational events, it is not possible at this stage to assess the contribution of each mutation to the observed abnormal response. However, this result indicates that mutations leading to an increased response to Miz1 transactivation can contribute to BCL2 deregulated expression in a subset of DLBCLs (Fig. S3).

Previous reports have indicated that BCL6 can suppress target gene expression by directly binding to specific DNA sequences (1, 28) or by interacting with Miz1 (32) or AP1 factors (45), which in turn bind DNA directly. The results herein indicate that, in normal GC B cells, BCL6 occupies the BCL2 promoter and suppresses BCL2 transcription, and this function requires the presence of the Miz1 transcriptional activator. One important implication of this finding is that Miz1 levels can influence the ability of BCL6 to suppress BCL2, with low Miz1 levels limiting BCL6 function by preventing access to target promoters, and excessive Miz1 levels leading to the same net effect by out-titrating BCL6. Thus, the mechanism by which BCL6 suppresses BCL2 is analogous to the BCL6-mediated suppression of CDKN1A (32).

The ability of BCL6 to suppress the anti-apoptotic function of BCL2 is consistent with the complex biology of the GC, where cells must be ready to undergo apoptosis if not rescued based on their affinity for the antigen. Consistent with this scenario, the transcriptional profile of GC centroblasts is characterized by downregulation of multiple anti-apoptotic genes and expression of pro-apoptotic genes (39, 46). While other genes with anti-apoptotic function may also be targeted by BCL6 directly or indirectly, its specific activity on BCL2 represents a critical function in ensuring that GC cells die if not rescued by the antigen and by other cellular signals. At the end of the GC reaction, engagement of the BCR by the antigen, signals from T cells (e.g., CD40 activation), and the accumulation of genotoxic stress lead to BCL6 downregulation and consequent release of BCL2 expression, which is typical for memory and plasma cells (46).

Previous reports have indicated that the translocation t(14;18) leads to ectopic activation of BCL2 in most FL and up to one-third of DLBCL cases (34) and implied this event as the main mechanism leading to BCL2 expression in these lymphoma types (37, 40). The results herein modify these notions in several ways. First, a fraction of DLBCLs was identified in which BCL2 is coexpressed with BCL6 in the absence of BCL2 gross structural alterations, suggesting that additional mechanisms exist for BCL2 induction in DLBCL. Second, these results show that chromosomal translocations and amplifications are not the only genetic lesions affecting the BCL2 locus in DLBCL, because a significant load of mutations was found in both translocation-positive and -negative DLBCL cases. While the former likely derive from the spreading of SHM from adjacent Ig enhancers, the mechanism causing mutations in nontranslocated alleles is unknown and may be part of the aberrant SHM activity that targets multiple genes in DLBCL (44).

Although the complexity and case-to-case variability of the mutational pattern prevent a simple interpretation or analysis of their role in each case, our results suggest at least 3 mechanisms by which BCL2 may escape BCL6-mediated suppression in DLBCL, beside or together with BCL2 chromosomal translocations. In a minority of cases, insufficient Miz1 levels may prevent BCL6 from reaching the BCL2 promoter and repressing its transcription (7/18 translocation-positive and 2/32 translocation-negative cases) (Fig. 5 A and B and Fig. S3). This mechanism may complement the deregulating activity of the Ig enhancers in cases with structurally altered BCL2 genes. Alternatively, abnormally high levels of Miz1, as observed in over one third of translocation-negative BCL2+BCL6+ DLBCLs, may out-titrate the ability of BCL6 to suppress BCL2 by generating BCL6-free Miz1 molecules. This scenario seems largely restricted to BCL2 unmutated cases, suggesting that abnormally high levels of Miz1, unchecked by BCL6, may substitute for Ig enhancer elements in transactivating BCL2. Finally, a subset of mutations occurring in both translocation-positive and -negative DLBCLs may cause abnormally high Miz1 responses. Together, these 3 mechanisms may account for at least 50% of cases displaying pathologic coexpression of BCL6 and BCL2, independent of chromosomal translocations (Fig. S3). It is important to note that not all mutations may have functional significance, because the SHM mechanism displays a semirandom distribution, in which functionally relevant and irrelevant sequence changes may be coselected during tumorigenesis. Mutations were also found within the BCL2 coding domain, as previously reported in t(14;18)+ FL and transformed DLBCL cases. These mutations did not affect the ability of BCL2 to synergize with BCL6 and induce cell transformation in soft agar/colony formation assays (Fig. S4). Overall, these results underscore the complexity of the mechanisms responsible for BCL2 activation in DLBCL and the need for further studies, including testing on the role of mutations and the mechanisms leading to Miz1 under- and overexpression.

The pathologic coexpression of BCL6 and BCL2 in DLBCL was previously reported and may identify patients with unfavorable prognosis, although this issue is controversial (47). Abnormal BCL2 expression may allow GC B cells to survive in the presence of death signals (e.g., Fas) and in the absence of survival signals (e.g., BCR, CD40), thus contributing to lymphomagenesis by increasing the pool of cells that can be targeted by additional genetic alterations in the GC environment. Preliminary evidence for the pathologic effect of BCL6 and BCL2 expression was obtained in Rat1 cells, where cotransfection of both genes led to a synergistic transformation effect (Fig. S4). This result will have to be corroborated by directing the coexpression of BCL6 and BCL2 to GC B cells in transgenic mice. The results herein suggest that combination modalities targeting the multiple oncogenic activities of BCL6 and the anti-apoptotic function of BCL2 may represent a rational approach for the treatment of a subset of DLBCLs (48, 49).

Transient transfection/reporter assays were performed on 293T cells as described, using the indicated constructs (1, 12). All experiments were performed in duplicate, and luciferase activities were measured 48 h posttransfection using the dual-luciferase reporter assay kit (Promega) according to the manufacturer's protocol.

Genomic DNA was extracted by standard methods, and 6 overlapping PCR products were used to amplify the BCL2 genomic sequences spanning position ≈−800 to ≈+2700 from the first exon of the BCL2 transcript variant alpha (NM_000633.2) (see Table S3 for PCR primers and conditions). Purified amplicons were sequenced directly from both strands (Genewiz) and analyzed as described (44). The somatic origin of the mutations was confirmed by analysis of matched normal DNA where available. Mutation analysis of BCL2 in normal B-cell subpopulations (naive and CB) and in the IMR91 fibroblast cell line was done as reported (44).