The Expanding Role of the BCL6 Oncoprotein as a Cancer Therapeutic Target

BCL6 (B-cell lymphoma 6) is emerging as a key oncoprotein and therapeutic target. BCL6 was first identified as a locus affected by chromosomal translocations in diffuse large B-cell lymphomas (DLBCL; ref. 1). However, it is now known to be broadly expressed in many lymphomas regardless of genetic lesions. Its role in lymphomagenesis stems from its function in the humoral immune system, where upregulation of BCL6 is required for the formation of germinal centers (GC) during the humoral immune response (2–4). GCs are transient structures that form in response to antigen stimulation. Within GCs, B cells tolerate massive proliferation and the mutagenic effect of the DNA-editing enzyme AICDA to undergo immunoglobulin affinity maturation (5). All of this is orchestrated by and dependent on BCL6, a powerful transcriptional repressor that silences hundreds of genes. Some of these target genes control DNA damage sensing (i.e., ATR, CHEK1, TP53, ARF, etc.) and proliferation checkpoints (CDKN1A, CDKN1B, CDKN2A, CDKN2B, PTEN, etc.; ref. 6). BCL6 also represses genes required for exit from the GC reaction and plasma cell differentiation (e.g., IRF4, PRDM1; ref. 6). This ensures that GC B cells have sufficient time to acquire somatic hypermutation of their immunoglobulin genes. It is thus easy to visualize how deregulated suppression of these target genes could result in malignant transformation of B cells. Indeed constitutive expression of BCL6 in GC B cells drives the development of DLBCL in mice (7–9).

BCL6 also represses numerous oncogenes in GC B cells, including MYC, BCL2, BMI1, CCND1, and various others (10, 11). Through this function, BCL6 may mitigate its own pro-oncogenic checkpoint repression effect and, thus, reduce the potential for malignant transformation of GC B cells. This effect is abrogated in the presence of BCL2 or MYC translocations, which drive expression of these oncogenes through aberrant regulatory elements. The presence of both MYC and/or BCL2 together with BCL6 (regardless of translocations) is clearly deleterious. It provides B cells with simultaneous suppression of checkpoints through BCL6, along with the progrowth and survival effects of MYC and BCL6. Not surprisingly, the combination of MYC and/or BCL2 with BCL6 in DLBCL has been linked to unfavorable clinical outcomes (12).

In the normal immune response, BCL6 function is terminated by the disruption of BCL6 transcriptional complexes through CD40-induced ERK signaling and downregulation of BCL6 mRNA by IRF4 and PRDM1 (13–15). Termination of BCL6 function is required for B cells to exit the GC reaction. Yet in DLBCLs, a variety of mechanisms contribute to aberrant persistence of BCL6 expression. These include fusion of the BCL6 coding region to heterologous promoters via chromosomal translocations and somatic mutation of binding sites for repressors of BCL6 expression, such as IRF4, and BCL6 itself (15, 16). Somatic mutations of the BCL6 ubiquitin ligase FBXO11 can enhance the half-life of BCL6 protein in DLBCL (17). Induction of Hsp90 activation, which occurs almost universally in DLBCL, forms a positive feedback loop whereby (i) HSP90 maintains BCL6 mRNA and protein stability; (ii) HSP90 enhances BCL6 repressor function by directly forming a complex on chromatin; and (iii) BCL6 repression of EP300 prevents acetylation and inactivation of HSP90, thus further enhancing BCL6 protein expression (18, 19). BCL6 expression can also be aberrantly maintained by hypermethylation of regulatory CpGs contained in the BCL6 first intron (20). The powerful tumorigenic activity of BCL6 and the myriad ways that lymphoma cells maintain its activity have fueled interest in the development of BCL6 inhibitors.

To understand BCL6 as a therapeutic target, it is first necessary to consider how it mediates its biological actions. BCL6 has a trimodular structure consisting of an N-terminal BTB/POZ domain that mediates transcriptional repression, an unstructured middle region containing a second repression domain (RD2), and a series of six C2H2 zinc fingers at the C-terminus that bind to DNA and other proteins (6). The biochemical and biological functions, together with the partner proteins of each of the BCL6 domains, are summarized in Fig. 1. One approach to targeting BCL6 is to completely abrogate its functions, for example, using molecules that could block its zinc fingers from binding to DNA or by destroying or downregulating the entire protein using approaches such as RNAi, antisense molecules, or small molecules that target proteins for proteolytic destruction (e.g., degronomids). Yet in the case of BCL6, such an effect is not desirable. This is because BCL6 knockout mice, in addition to failure to form GCs, also manifest a severe and lethal systemic inflammatory disease driven by T cells and macrophages (21). BCL6-deficient mice die within a few weeks of birth and exhibit massive tissue infiltration of inflammatory cells in lung, heart, and other tissues (2–4). Loss of BCL6 in macrophages causes accelerated atherosclerosis in mice (22).

Fortunately, it is possible to avoid these adverse effects by targeting specific sites on BCL6 that drive its cancer functions but leave its anti-inflammatory functions intact. Structure–function studies show that the BCL6 BTB domain mediates repression by recruiting the corepressor proteins SMRT, NCOR, and BCOR to an extended groove motif that forms along the BTB dimer interface (Figs. 1 and 2A and B; refs. 23, 24). BCL6 can bind two corepressors at a time because there are grooves on either side of the dimer. The corepressors bind through an 18 amino acid BCL6-binding domain (BBD) that is highly conserved between SMRT and NCOR, but which is completely different in the case of BCOR (23, 24). Importantly, the BCL6 lateral groove residues that make contact with the BBD are unique to BCL6 and not other BTB proteins (23), which could facilitate the development of specific inhibitors that do not affect other BTB domains. Point mutation of the lateral groove abolishes the repressor function of the BCL6 BTB domain (23).

The repressive mechanism of action of the BCL6 BTB domain is quite intricate. NCOR and SMRT have related structures and form similar chromatin-modifying complexes that deacetylate histones (25, 26), whereas BCOR is completely different and forms a variant Polycomb PRC1 complex with multiple distinct effects on chromatin (27). On one hand, BCL6 preferentially represses gene promoters by forming a complex with BCOR (28). However, the BCOR complex can only stably assemble at promoters that are marked with histone 3 lysine 27 trimethylation (H3K27me3) by the Polycomb PRC2 protein EZH2 (29). H3K27me3 forms a binding site for CBX8, one of the components of the BCOR complex. Hence, BCL6-mediated promoter repression in GC and lymphoma cells requires “combinatorial tethering” of the BCOR complex through the simultaneous actions of BCL6 and EZH2 (29). On the other hand, BCL6 also represses gene enhancers, but this effect is linked to preferential interaction with SMRT and NCOR complexes, which include HDAC3 (28). BCL6 enhancer repression is driven through this indirect recruitment of HDAC3 to remove the activating H3K27acetyl mark. Notably, the genes that BCL6 represses by forming promoter repression complexes are mostly different from those that it represses by forming enhancer complexes (28).

Remarkably, mice expressing a mutated form of BCL6, where the BTB domain cannot bind to SMRT, NCOR, and BCOR, live normal, healthy lives without any evidence of inflammatory syndrome. These mice are nonetheless unable to form GCs, specifically due to failure of GC B cells to proliferate and survive (30). This is because BCL6 largely represses DNA damage and proliferation checkpoint genes through the BTB domain, which may explain (as discussed below) why this domain is also important in other tumor types. Indeed, mutation of the BTB domain abrogated the ability of the lymphoma oncogene EZH2^Y641 to drive preneoplastic lymphoproliferation in vivo (29). Hence, the BCL6 BTB domain is required for GC formation and lymphomagenesis, but not anti-inflammatory effects. Instead, the anti-inflammatory effect is due to direct competition of the BCL6 zinc fingers with STAT proteins for binding to inflammatory chemokines in macrophages (30). Finally, the minimal BCL6 RD2 domain consists of approximately 40 amino acids and is required for interactions with MTA3 and HDAC2 (31, 32). Mice engineered to express RD2-mutant BCL6 also live normal, healthy lives and have no inflammation (32). They do lack GCs, although in this case, the mechanism involves partial loss of function of T-follicular helper (TFH) cells and disruption of B-cell homing to the GC (32). This extraordinary cell-context biochemical specialization of BCL6 makes it possible to target specific functions relevant to tumor cells and, at the same time, avoid disrupting other critical functions.

The most straightforward approach to targeting BCL6 is disrupting the interaction between the BCL6 BTB domain lateral groove and its corepressors because (i) this protein surface mediates BCL6 oncogenic effects; (ii) the protein interaction is structurally characterized; (iii) it involves a unique interface not conserved in other proteins; and (iv) targeting the BTB lateral groove will not induce deleterious inflammatory effects caused by complete loss of BCL6. Targeting the BTB dimer interface would likely be unsuitable, as BTB domains aggregate and are degraded if they cannot dimerize, which could unleash the inflammatory effects of BCL6 deficiency. The BTB domain has other features, such as a charged pocked motif and a hydrophobic surface for oligomerization, which have not been functionally characterized (Fig. 2A; ref. 23).

Protein–protein interactions have long been thought to be “undruggable” (33) given that many protein interactions involve large surfaces that may be difficult to disrupt with small molecules. However, given sufficient knowledge of the protein interface, it is proving possible to design drugs that exploit vulnerable “sweet spots” (34). The first BCL6 inhibitor contained the 17 amino acids from SMRT-BBD, along with a protein transduction domain for cell penetration (35). This BCL6 peptide inhibitor (BPI) disrupted BCL6 repression complexes, induced expression of BCL6 target genes, killed DLBCL cells in vitro, and phenocopied the BCL6 BTB domain mutant phenotype in vivo (i.e., loss of GC formation without inflammation; Fig. 3). Although BPI was active at low micromolar concentrations, it was also readily degraded by proteases (35). The bulk of intermolecular contacts between the BBD and BCL6 are confined to a 9 amino acid sequence, which was used as the warhead for a second-generation peptide using D-amino acids resistant to proteases (36). The peptide sequence was inverted to maintain proper stereochemistry and included a fusogenic motif for superior uptake within cells. The resulting retro-inverso (RI)-BPI peptide retained specificity for BCL6, but not other BTB proteins; exhibited superior stability; and did not induce immunogenicity (36). RI-BPI killed BCL6-dependent DLBCLs at an average concentration of 16 μmol/L, similar to BPI (Fig. 3; ref. 36). Pharmacokinetic studies in mice revealed peak RI-BPI intratumor concentrations after 30 minutes and persistence in the nuclei of lymphoma cells for >24 hours (36). RI-BPI was nontoxic to mice, even after 1 year of continuous administration (36). These peptides thus appear suitable for use in humans. An alternative peptidomimetic strategy employed aptamer screening to identify Apt48, which binds to the BCL6 BTB domain in a different manner than BPI (Fig. 3; ref. 37). Apt48 promoted reexpression of BCL6 target genes and cell-cycle arrest in lymphoma cells, suggesting interference with BCL6 function.

The BCL6 BTB lateral groove–BBD complex was used for in silico screening of small-molecule libraries. These studies yielded a family of compounds with reproducible and specific BCL6 inhibition. A representative compound called “79-6” was shown to bind an aromatic pocket situated in the BTB domain lateral groove (38). 79-6 induced disruption of BCL6 transcriptional complexes, reactivation of BCL6 target genes, and selective killing of BCL6-dependent DLBCL cells (Fig. 3; ref. 38). Even though this small molecule was active in animal models, its binding affinity of approximately 138 μmol/L was lower than the endogenous BBD (∼20 μmol/L; ref. 38). A novel in silico–based fragment-based screening method called “SILCS” (39, 40) was subsequently used to design the most potent BCL6 inhibitor to date, FX1 (41). FX1 is the first reported inhibitor with higher affinity for the BCL6 BTB domain than the endogenous corepressor proteins (Kd ∼ 4 μmol/L), demonstrating the feasibility of blocking large protein interactions with small molecules (Fig. 3; ref. 41). SILCS modeling combined with NMR analysis revealed a more favorable orientation of the compound within the lateral groove aromatic pocket. FX1 outperformed all previous BCL6 inhibitors in cell-based pharmacodynamic assays while maintaining specificity for BCL6. The molecule exhibited favorable pharmacokinetics in vivo and lack of toxicity similar to previous generations of BCL6 inhibitors (41). Other small molecules with potential BCL6-binding activity include the natural product resveratrol (42), which can induce antiproliferative and anti-inflammatory activities in different types of cancer cells through various mechanisms (Fig. 3; ref. 43). Ansamycin antibiotics may also bind to BCL6. The structure of one of the members of this family, rifabutin, with the BCL6 BTB domain showed binding to the lateral groove similar to 79-6, although with a lower binding affinity of approximately 1 mmol/L (Fig. 3; ref. 44).

High BCL6 expression is usually associated with germinal center B (GCB)-type DLBCLs (45), which are the most obvious target disease for BCL6 inhibitors. Accordingly, daily administration of RI-BPI at a dose of 7.5 mg/kg/day reduced the growth of established BCL6-dependent GCB-DLBCL xenografts by 60% to 80% but did not affect BCL6-independent DLBCL xenografts (36). Activity was concentration dependent, with doses of 50 mg/kg/day leading to complete regression of GCB-DLBCLs (6). In spite of its weak binding, the small molecule 79-6 concentrates to high levels within lymphoma cells, which enhances its activity (38). Administration of 50 mg/kg/day 79-6 to SCID mice induced a 65% to 70% reduction in the size of established BCL6-dependent DLBCL xenografts but had no effect on BCL6-independent DLBCLs (38). In contrast, the potent small molecule FX1 induced complete regression of GCB-DLBCL xenografts at 25 mg/kg but did not suppress BCL6-independent lymphomas (41). In all of these studies, the various inhibitors induced BCL6 target gene reexpression, proliferation arrest, and apoptosis in DLBCL tumor xenografts. More importantly, all three compounds were active in killing primary human DLBCL specimens cultured ex vivo, underlining the potential for BCL6 inhibitors to have activity clinical trials (36, 38, 41).

From the predictive biomarker perspective, it would be useful to classify patients as having BCL6-dependent or BCL6-independent lymphomas. The simple expression of BCL6 may not be sufficient to indicate BCL6 dependence, as a few GCB-DLBCL cell lines express BCL6 but are not affected by genetic or pharmacologic BCL6 inhibition (36, 38, 41, 46). A recent study overexpressing BCL6 in hematopoietic cells in mice showed development of lymphomas that were BCL6 independent. These data suggest that BCL6 might sometimes act in a hit-and-run manner (47). Expression of BCL6 is, therefore, necessary but perhaps not in all cases sufficient to predict response to BCL6 inhibitors. Gene expression profiles might help to identify BCL6-dependent lymphomas. Indeed, one study identified a putative BCL6 functional signature by analyzing transcriptional profiles in primary DLBCL patient specimens. When applied to DLBCL cell lines, this signature accurately predicted which ones would be BCL6 dependent and which ones are BCL6 independent based on their response to BCL6 inhibitor (46). Perhaps functionally relevant biomarkers such as these could aid in patient selection for BCL6 inhibitor trials.

The relevance of BCL6 is not limited to GCB-DLBCLs. In fact, BCL6 is expressed in most activated B-cell (ABC)-type DLBCLs, albeit at lower levels (41, 48). Indeed, BCL6 translocation and amplification is largely restricted to ABC-DLBCLs, providing genetic evidence for the importance of BCL6 in these tumors (47). It is worth noting that BCL6 transcript or protein abundance is not associated with the degree of biological dependency on BCL6. Along these lines, BCL6 shRNA knockdown of BCL6 was just as deleterious to ABC-DLBCL cells as to GCB-DLBCL cells (41). ABC-DLBCL cells respond to similar doses of BCL6 inhibitor, such as FX1, which also suppressed the growth of ABC-DLBCL xenografts in vivo. Most importantly, primary human ABC- and GCB-DLBCL cells respond equally well to FX1 ex vivo (41). Given that ABC-DLBCLs manifest inferior outcome, it is warranted to develop BCL6-targeted therapy for these patients. Follicular lymphomas manifest features of GCB cells and, thus, generally express BCL6. We have found that primary low-grade follicular lymphoma cells are killed by the BCL6 inhibitor RI-BPI (E. Valls and A.M. Melnick; unpublished data). Burkitt lymphomas arise from GC B cells, and BL cell lines are potently killed by BCL6 inhibitors in vitro and in vivo (35). Although BCL6 is not usually expressed in multiple myeloma, withdrawal of microenvironment survival signals could induce BCL6 expression (49), raising the possibility that these B-cell neoplasms could be sensitive to BCL6 inhibitors as part of combination therapy regimens.

BCL6 is implicated in an expanding list of tumors. Table 1 summarizes the spectrum of BCL6-dependent tumors, the way in which BCL6 was inhibited, and the phenotype associated with the inhibition. For example, BCL6 is a direct transcriptional target of MLL fusion proteins in 11q23-rearranged B-acute lymphoblastic leukemias (B-ALL; ref. 50). Binding of MLL fusion protein to the BCL6 promoter was associated with hypomethylation and induction of BCL6 expression. Treatment with RI-BPI reduced colony formation capacity and induced apoptosis in MLL fusion cell lines and primary patient samples (50). Exposure to ABL tyrosine kinase inhibitors (TKI) strongly induces BCL6 expression in BCR-ABL⁺ B-ALLs. In this context, BCL6 functioned as a survival feedback mechanism that enables leukemia cells to resist TKI treatment (51). Combined treatment of primary human BCR-ABL⁺ B-ALL with TKI and RI-BPI yielded massively synergistic antileukemia effects in vitro and in vivo (51). Chronic myeloid leukemias (CML) are also driven by the BCR–ABL fusion protein. Similar to B-ALLs, exposure to TKI induced BCL6 in CML cells (52). CD34⁺ CML leukemic stem cells (LSC) upregulate BCL6 in response to TKI in a FOXO-dependent manner. It has been appreciated that CML LSCs are less responsive to TKI than bulk leukemia cells, and, in many patients, the disease is not eradicated by TKI therapy alone (52). It is notable that CML LSCs require BCL6 for their ability to form colonies, and initiate leukemia in mice. Treatment of patient-derived CML cells with RI-BPI selectively induced cell-cycle arrest of CD34⁺ LSCs, causing depletion of CD34⁺ CD38⁻ LSCs, whereas CD34⁻ subpopulations remained intact (52). Hence, BCL6-targeted therapy could potentially eradicate CML LSCs and reduce the need for long-term TKI therapy.

Angioimmunoblastic T-cell lymphomas (AITL) manifest the phenotypic hallmarks of TFH cells. BCL6 is required for the formation of TFH cells (53). AITLs strongly express BCL6, and, hence, it has been proposed that BCL6 might be a therapeutic target in this disease (53). Up to 50% of breast tumors and many breast cancer cell lines contain amplification of the BCL6 locus (54–56). Functional studies suggest that BCL6 contributes to the development of breast cancer (55–58). BCL6 is important for tumor maintenance as treatment of breast cancer cell lines with RI-BPI or 79-6 reduced cell viability (54). BCL6 may have cell context-specific functions in breast cancer cells, as its target genes are only partially overlapping with those in B cells (54). BCL6 was shown to be required for survival and proliferation of non–small cell lung cancer (NSCLC) cells, in part due to repression of genes involved in the DNA damage response. Inhibition of BCL6 by FX1 kills NSCLC cells in vitro and in vivo and exerts synergistic effect with cisplatin in xenograft models (59).

Given the complexity and heterogeneity of tumor cells, it is unlikely that single agents will cure disease. Hence, rational combination of BCL6 inhibitors with other drugs is the best way to achieve maximal antitumor effect. Blocking the BCL6 BTB domain is not expected to induce major biological side effects, so it should be straightforward to combine BCL6 inhibitors with other therapeutic agents (36, 38, 41). For example, the fact that BCL6 and EZH2 cooperate and require each other to repress transcription led to the notion of combining EZH2 and BCL6 inhibitors (29). As might be expected, this combination yielded enhanced reactivation of BCL6/EZH2 target genes, with corresponding cooperative effects against DLBCL cells in vitro and in vivo as well as against primary human DLBCL specimens ex vivo (29). The result was observed in EZH2-mutant and wild-type DLBCL cells and is summarized in Fig. 4A (29).

The positive feedback loop between BCL6, Hsp90, and EP300 mentioned earlier was the basis to combine BCL6 inhibitor with drugs against either Hsp90 or histone deacetylase (HDAC) inhibitors (18). This is because inhibition of Hsp90 impairs BCL6 repressor functions and reduces BCL6 mRNA and protein levels (Fig. 4B). This approach yields enhanced derepression of BCL6 target genes, such as EP300, resulting in reduced Hsp90 activity through EP300-mediated acetylation, increased TP53 expression and function, and restored DNA damage and proliferation checkpoints (18). Accordingly, the combination of RI-BPI with Hsp90 or HDAC inhibitor was highly synergistic in vitro and yielded more potent antilymphoma effects in vivo (18). In a similar vein, enhanced checkpoint reactivation through combination of BCL6 inhibitor and TP53 activator also yielded enhanced antitumor effects (60). Indeed, as BCL6 reactivates DNA damage checkpoints, it is not surprising that combinations of chemotherapy drugs plus BCL6 inhibitor yield enhanced activity against lymphomas and leukemias (41). The feedback mechanism where upregulation of BCL6 can protect BCR-ABL⁺ neoplasms from TKIs supports the testing of BCL6 inhibitor combination therapy in B-ALL and CML (51, 52).

Other combination therapies could target feedback mechanisms that enable cells to escape killing by BCL6 inhibitors. For example, BCL6 directly binds and represses BCL2 and BCL2L1 (BCL-XL; Fig. 4C; refs. 10, 11). Inhibition of BCL6 induces the expression of BCL2 and BCL-XL, which can help lymphoma cells survive exposure to BCL6 inhibitors, a phenomenon that has been called “oncogene switching.” The combination of BCL6 inhibitors with BH3 mimetics overcomes this effect and yields synergistic activity (61). BCL6 represses multiple NFκB-related genes, which are induced after inhibiting BCL6. The effects of NFκB are complex, but proof-of-principle experiments combining proteasome inhibitors with BCL6 inhibitors showed highly synergistic actions. BCL6 also represses STAT3 (62, 63), an oncogenic driver in tumors including breast cancer, which could be induced by BCL6 inhibitors. Suppression of STAT activation through Jak2 or STATs inhibition caused additive loss of breast cancer cell viability, perhaps in part by suppressing additional STAT protumorigenic effects (54, 63). As the functions of BCL6 in various tumor types are better understood, it is likely that other such mechanism-based opportunities will be developed.

In summary, BCL6 is a suitable therapeutic target that plays a broad and critical role in many cancers. Targeting the BCL6 BTB domain is an effective means of disrupting its functions without inducing significant off-target effects against normal tissues. Mechanism-based combination therapy of BCL6 inhibitors yields dramatic antitumor effects. Hence, we propose that development of clinical grade BCL6 compounds is a mission worth vigorously pursuing.

A.D. MacKerell Jr is a cofounder and chief scientific officer of SilcsBio LLC and is a consultant/advisory board member for BioVia. A.M. Melnick reports receiving other commercial research support from Eli Lilly, GlaxoSmithKline, Janssen, and Roche and is a consultant/advisory board member for Boehringer Ingelheim, Eli Lilly, Epizyme, and Roche. M.G. Cardenas, F. Xue, A. D. MacKerell Jr, and A. M. Melnick are listed as coinventors on a patent on BCL6 inhibitors owned by Weill Cornell Medicine. No potential conflicts of interest were disclosed by the other authors.

This work was supported by the Leukemia & Lymphoma Society Therapy Acceleration Program.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.