Bromodomain and extra-terminal (BET) family proteins are key regulators of gene expression in cancer. Herein, we utilize BRD4 profiling to identify critical pathways involved in pathogenesis of chronic lymphocytic leukemia (CLL). BRD4 is overexpressed in CLL and is enriched proximal to genes upregulated or de novo expressed in CLL with known functions in disease pathogenesis and progression. These genes, including key members of the B-cell receptor (BCR) signaling pathway, provide a rationale for this therapeutic approach to identify new targets in alternative types of cancer. Additionally, we describe PLX51107, a structurally distinct BET inhibitor with novel in vitro and in vivo pharmacologic properties that emulates or exceeds the efficacy of BCR signaling agents in preclinical models of CLL. Herein, the discovery of the involvement of BRD4 in the core CLL transcriptional program provides a compelling rationale for clinical investigation of PLX51107 as epigenetic therapy in CLL and application of BRD4 profiling in other cancers.
Significance: To date, functional studies of BRD4 in CLL are lacking. Through integrated genomic, functional, and pharmacologic analyses, we uncover the existence of BRD4-regulated core CLL transcriptional programs and present preclinical proof-of-concept studies validating BET inhibition as an epigenetic approach to target BCR signaling in CLL. Cancer Discov; 8(4); 458–77. ©2018 AACR.
This article is highlighted in the In This Issue feature, p. 371
Chronic lymphocytic leukemia (CLL) is characterized by increased B-cell receptor (BCR) signaling, defective apoptosis, and disrupted immune effector function (1–3). Ibrutinib, a BCR pathway inhibitor that targets the Bruton tyrosine kinase (BTK), interferes with BCR signaling to impair microenvironment-derived survival signals and reverse disease-mediated immune suppression (4, 5). In patients with CLL, ibrutinib demonstrates dramatic and protracted efficacy with a tolerable toxicity profile (5–8) and thus has transformed the therapeutic landscape of this disease. The limitations of ibrutinib include a low frequency of complete response and frequent relapse in patients with high-risk disease including those with del(17)(p13.1) (9). Relapse often results from BTK mutations that preclude ibrutinib binding and allow reactivation of BCR signaling. This underscores the importance of BCR signaling and attractiveness of novel therapies that can simultaneously target both BCR signaling and multiple other genes involved in CLL pathogenesis. Additionally, 5% to 10% of patients with CLL transform to a rapidly growing aggressive lymphoma, a syndrome known as Richter transformation (RT), which is resistant to aggressive immunochemotherapy and associated with a short overall survival (OS; ∼6 months). RT is the most common type of progression now observed in patients with CLL responding to targeted therapies such as ibrutinib and venetoclax (10, 11).
To date, the molecular mechanisms driving the establishment and progression of CLL remain to be fully elucidated. Aberrant expression of antiapoptotic (e.g., BCL2, TRAIL, and MCL1; refs. 12–14), prosurvival (e.g., IL4R and IL21R; refs. 15, 16), and cell-trafficking proteins (e.g., CCR7 and CXCR4; refs. 17, 18) has been associated with aggressive disease behavior. Additionally, gene expression profiling and pathway analysis has established BCR signaling as a central driver of CLL (19). Furthermore, next-generation sequencing efforts have identified a relatively small number of recurrently mutated genes of potential pathogenetic relevance that affect cellular functions including DNA damage and cell-cycle control (e.g., TP53 and ATM), Notch signaling, mRNA processing (e.g., SF3B1 and XPO1), B-cell transcription (e.g., IKZF3), and chromatin remodeling (e.g., ARID1A and CHD2; ref. 20). Despite the relatively low mutational load compared with other tumors, CLL presents with substantial clinical and biological heterogeneity. In addition, some patients develop a more aggressive disease course despite the absence of high-risk genetic alterations, suggesting that factors other than genetic lesions could play a crucial role in CLL pathogenesis and outcome. For example, epigenetic mechanisms such as aberrant promoter methylation that lead to abnormal expression of TCL1A (21), ZAP70 (22), microRNAs (miR-155 and miR-21; ref. 23), BCL2 (13), and long noncoding RNAs are implicated in CLL pathogenesis. Therefore, a deeper understanding of the epigenetic mechanisms controlling the aberrant expression of these pathogenic factors could provide a novel axis for targeted therapies to reverse the transcriptional abnormalities seen in CLL. Recent DNA methylation and assay for transposase accessible chromatin with high-throughput sequencing (ATAC-seq) studies in CLL revealed that methylation profiles and genome-wide chromatin accessibility can more precisely differentiate CLL subtypes and predict clinical course compared with gene expression profiles (24). Hence, the biological differences between the major subtypes of CLL are imprinted in the epigenome.
The bromodomain and extra-terminal (BET) subfamily of proteins (BRD2, BRD3, BRD4, and BRDT) are transcriptional regulators that bind to acetylated histones and recruit the positive transcription elongation factor complex (P-TEFb) and RNA Pol II (25, 26) to control gene expression at those regions (27, 28). BRD4 associates with all active promoters and a significant fraction of active enhancers in both normal and transformed cell types (29–31). Despite its role as a general transcriptional regulator, inhibition of BRD4 has shown selective anticancer activity in a variety of preclinical models. This therapeutic window versus normal tissues has been attributed to the high BRD4 load at superenhancer (SE) regions regulating cancer-specific genes (i.e., MYC and BCL6), making the expression of those genes exquisitely sensitive to pharmacologic targeting of BET proteins. Although the protumor role of BET proteins has been shown in many highly proliferative cancers, predominantly using cell lines, no studies have examined the involvement of these epigenetic modifiers in a slowly proliferating disease like CLL. CLL therefore represents an important disease to interrogate the potential role of BET proteins in tumorigenesis and the relevance of therapeutic targeting. Furthermore, if pathways such as BCR signaling, which are validated as therapeutically relevant in CLL, were shown to be highly transcriptionally enhanced by BRD4, it would provide a rationale for utilizing BRD4 profiling to explore for relevant therapeutic targets in other cancers.
A number of first-generation small-molecule BET inhibitors sharing the same benzodiazepine scaffold (i.e., JQ1, iBET762, and OTX015) have demonstrated antitumor activity in different preclinical models of solid and hematologic malignancies by dissociating BET proteins from SE regions with consequent suppression of c-MYC and NFκB activity (32–37). The most mature data are derived with OTX015 in acute leukemia (38) or in lymphoma and multiple myeloma (ref. 39; ClinicalTrials.gov, NCT01713582). In the lymphoma study, hematologic dose-limiting toxicities (DLT) at pharmacologically effective doses necessitated intermittent dosing. Even with intermittent dosing, there were on-target DLTs in addition to non-DLT gastrointestinal toxicities that significantly hampered study adherence. The few clinical responses were observed only at dose levels or schedules in which DLTs occurred. These results suggest that although OTX015 has clinical activity in lymphoid malignancies, its therapeutic index is narrow. Following on the exceptional therapeutic index of ibrutinib in B-cell malignancies, there is a need to develop BET inhibitors with a broader therapeutic index to improve safety while maintaining efficacy.
Given the promise of BET targeting in rapidly proliferative tumors, we applied an integrative functional epigenomic approach to examine the genome-wide chromatin distribution of BRD4 in slowly proliferating primary CLL tumor cells. We demonstrate overexpression of BRD4 and characterize its asymmetric binding to a small set of enhancer regions (SEs) that control nearby genes implicated in CLL pathogenesis and disease progression (i.e., BCR signaling–associated genes, MIR155, MIR21, and IKZF1). To develop BET inhibitors with an improved therapeutic index and pharmaceutical profile over current clinical compounds, we applied scaffold-based and crystallography-guided drug design (40). This effort yielded PLX51107, a potent and selective small-molecule BET inhibitor with a different chemotype from the previously published benzodiazepine-based BET inhibitors. Notably, the BRD4-overloaded SE regions were subsequently shown to have exquisite sensitivity to PLX51107 compared with other enhancers, explaining the selective effect of BET inhibitors on CLL-specific oncogenic pathways. Furthermore, PLX51107 demonstrated potent antitumor activity in preclinical models of CLL and RT with a short terminal half-life that differentiates it from other agents currently under clinical development.
BRD4 Binds Genome-Wide to Transcriptionally Active Genes in CLL
Studies exploring the role of BET proteins as transcriptional coactivators in slowly proliferative cancers such as CLL have yet to be reported. Their contribution to relevant pathogenic driving pathways could provide an avenue to identify new therapies. BRD4 and other bromodomain-containing proteins such as BRD2, BRD3, p300, and CREBBP-binding protein (CPB) are ubiquitously expressed in primary CLL cells (Fig. 1A and Supplementary Fig. S1A). Notably, BRD4 is significantly overexpressed in CLL cells when compared with normal B cells (Supplementary Fig. S1B and S1C) whereas no statistically significant difference in BRD4 expression was observed between naïve and memory B cells (Supplementary Fig. S1D).
Cancer cell lines differ drastically in genomic and transcriptional profile from their originating primary tissue (41, 42), making their relevance to understand the tumor epigenome limited. We used primary CLL B cells derived from four patients (patient details summarized in Supplementary Table S1) to assess chromatin accessibility genome-wide, using ATAC-seq, and to identify regions in the epigenome that are accessible to transcription factor binding. In the same samples we probed for BRD4 expression at the mRNA and protein level (Supplementary Fig. S1B and S1C) and the genome-wide localization of BRD4 and its abundance in enhancer areas using chromatin immunoprecipitation coupled to high-throughput sequencing (ChIP-seq) and examining both the histone H3K27ac chromatin mark and RNA Pol II as measures of downstream transcription. The genome-wide correlation between BRD4 occupancy and H3K27ac is strong, with the majority of H3K27ac occurring within open chromatin regions (Fig. 1B). In contrast, genes not expressed in CLL (e.g., EGFR, Myoglobin, etc.) are not identified in this analysis. Rank ordering of enhancer regions by H3K27ac enrichment reveals that about 94% of all chromatin-bound BRD4 resides in regions marked by H3K27ac, indicating the predominant recruitment of BRD4 at active enhancers in human CLL. Accordingly, BRD4 colocalizes with RNA Pol II at transcriptionally active peak regions, promoters, and SEs across the CLL genome (Fig. 1C). For each patient sample, on average, 300 out of 10,500 enhancer regions (∼3%) show a distinctly larger size and higher BRD4 load compared with typical enhancer and are henceforth defined as SEs based on the previously published definition of these regions (30). These SEs account for ∼30% of all enhancer-bound BRD4 in CLL cells. Next, using published data of BRD4 ChIP-seq in a nontumor B-cell line (43), we have analyzed the genome-wide BRD4 load across enhancers in normal B cells and compared this to CLL cells under basal and CpG-stimulated conditions. Overall in normal B cells BRD4 enhancer regions are only modestly loaded with BRD4 compared with CLL (Supplementary Fig. S1E, P = 0 in all samples). As expected, a number of common SEs between CLL and normal B cells are observed, possibly defining typical B-cell lineage regions (Supplementary Fig. S1F). However, there are numerous of SEs that are unique to CLL (Supplementary Fig. S1G).
The topmost BRD4-enriched SEs in CLL, common among all patient samples, include IL4R, IL21R, CLLU1, GRHPR, and RP11-443B7.1, a novel noncoding gene located in chromosome 1 (Fig. 1D). In addition, common highly ranked BRD4-dense enhancers are near genes with established roles in CLL biology and tumor microenvironment interactions and reported aberrant expression in CLL (e.g., CXCR4, CCR7, IKZF3, IL4R, IL2R, and IL21R) or de novo regulated in CLL (e.g., MIR21).
The BRD4-binding profile of these regions confirms a preferential overload in CLL compared with normal B cells. Functional enrichment analysis (Fig. 1E and Supplementary Table S2) of genes near SEs common among all patient samples revealed significant enrichment in these regions for genes involved in the BCR signaling pathway (e.g., PTPN6, BLK, and BCL2) and immune response (e.g., IL2RA, IL19, IL10, IL4R, IL24, TNF superfamily, CCR7, CXCR5, and B2M). As expected, large BRD4-enriched SEs in CLL and normal B cells were also found adjacent to the IGH loci similar to what has been described in multiple myeloma (35). Examples of select tracks for highly ranked BRD4-dense enhancers are illustrated in Fig. 1F; Supplementary Fig. S1I.
PLX51107 Is a Novel BET Inhibitor with Attractive Pharmaceutical Properties
Scaffold screening using low-affinity binding assays followed by cocrystallography identified a 213-Da micromolar hit (PLX5981) consisting of 3,5-dimethylisoxazole coupled to 7-azaindole (Fig. 2A). The isoxazole moiety mimics the disposition of an N-ϵ-acetyl-lysine whereas the 7-azaindole presents a facile framework to access both the Trp81-Pro82-Phe83 (WPF) shelf and the ZA channel (a specificity loop defined by αZ and αA helices) in BRD4 (32). Structure-guided chemistry exploration, including the replacement of 7-azaindole by 4-azaindole, yielded PLX51107, a potent BET inhibitor (Fig. 2A) with optimized pharmacokinetic properties. PLX51107 interacts with the WPF shelf and the ZA channel with its pyridylethyl (R1) and benzoic acid (R3) side chains, respectively (Fig. 2B). In particular, the benzoic acid traverses the ZA channel and forms a salt bridge with Lys91 revealing a level of structural plasticity in BRD4 that has not been observed with other BET inhibitors (Fig. 2B and C; Supplementary Fig. S2). Although the binding of known BET inhibitors (e.g., OTX015) is confined to the space occupied by a typical extended diacetylated histone peptide, PLX51107 represents a new class of inhibitors that engage additional binding elements in the ZA channel (Fig. 2B). The PLX51107-bound structure of BRD4 resembles the conformation induced by the binding of acetylated Lys310 peptide from RELA protein (Fig. 2B; Supplementary Table S3; ref. 44).
PLX51107 exhibits low nanomolar potency in blocking interactions mediated by the four BET family members. Binding affinity (Kd) measurements using isolated bromodomains (Supplementary Table S4) show that PLX51107 has a modest preference for bromodomain 1 (BD1) versus bromodomain 2 (BD2) within each BET protein (Kd = 1.6, 2.1, 1.7, and 5 nmol/L for BD1 and 5.9, 6.2, 6.1, and 120 nmol/L for BD2 of BRD2, BRD3, BRD4, and BRDT, respectively). Among non-BET proteins, PLX51107 shows significant interactions only with the bromodomains of CBP and EP300 (p300; Kd in the 100 nmol/L range). PLX51107 has a single chiral center and bears the (S)-enantiomer. The (R)-enantiomer is 5- to 10-fold less potent than PLX51107 across all BET bromodomains. Profiling PLX51107 in a leukemia and lymphoma panel reveals broad activity in various malignant cell lines (Supplementary Table S5).
Preclinical in vivo studies in the Ba/F3 (murine IL3-dependent pro–B-cell line) splenomegaly mouse model (45) show that PLX51107 is well tolerated and is active at 10-fold lower doses than OTX015 (Fig. 2D). Approximately 75% inhibition of splenomegaly was achieved with the 2 mg/kg PLX51107 dose, which was associated with an AUC0–24 of 17,700 ng • h/mL (or 40 μmol/L • h). This value defines the threshold exposure required for PLX51107 to be effectivein vivo. The same level of efficacy was achieved with 25 mg/kg OTX015 and the associated steady state AUC0–24 was 11,900 ng • h/mL (or 24 μmol/L • h).
BET proteins have important roles in normal tissues; strong and prolonged suppression of their functions have often-untoward effects in multiple tissues (46). Importantly, the relatively short half-life of PLX51107 (<3 hours in rodents and dogs) may underlie an improved therapeutic index, as epigenetic deviations tend to be long-lived. To support this notion, MV4-11 cells xenografted in SCID mice treated with PLX51107 were monitored for transcriptional changes over time. As shown in Fig. 2E, transcriptional changes of HEXIM1, a strong pharmacodynamic marker for monitoring target engagement of BRD4 inhibitors in tumors (47), substantially outlive the plasma drug levels.
A set of cell culture experiments was performed to investigate the effect of treatment duration on the pharmacologic activity of PLX51107. Short-term (4 hours) treatment of PLX51107 resulted in robust change in PD markers (c-MYC was used as a representative; Supplementary Fig. S2C) but did not induce an immediate apoptotic response (Supplementary Fig. S2D). Induction of apoptosis occurred after prolonged treatment (16 hours or more of continuous exposure). A delay between the exposure to a prodeath stimulus and the onset of effector activation is inherent to the apoptotic program. To take this into consideration, flow cytometry analysis of tumor cells treated with PLX51107 for 4 or 8 hours followed by a washout period of 20 or 16 hours, respectively, was performed. Four hours' treatment did not induce apoptosis but longer (8 hours) treatment did (Supplementary Fig. S3E). Note that PLX51107 and OTX-015 have similar residence time (Supplementary Fig. S3F). These results suggest that short-term target engagement may generate desired pharmacologic effect without inducing nonspecific killing. A BET inhibitor with a short half-life is expected to have reduced risk of toxic effects associated with long-term inhibition of BET family proteins (38, 39).
BET Inhibition Exerts In Vitro Antiproliferative Effects in CLL
CLL has been considered an accumulative disease that derives from an inherent defect in apoptosis rather than increased proliferation rate, based on the observation that the majority of the circulating CLL cells are arrested in the G0–G1 phase. Nevertheless, recent studies have shown that CLL disease development is a dynamic process, wherein microenvironmental cell–cell interactions and accessory signals within pseudofollicular proliferation centers are essential for the expansion and survival of CLL cells (48, 49). These signals have also been shown to promote acquisition of chemoresistance. We set out to determine whether BRD4 inhibition could attenuate CLL cell survival and proliferation when stimulated with CpG oligonucleotides, a well-characterized TLR9 agonist that promotes in vitro B-CLL cell proliferation (50) and mimics the effect of pseudoproliferation centers where CLL expands (6, 51, 52). PLX51107 prevents CpG-induced cell proliferation (Fig. 3A and B) in primary CLL cells in a dose-dependent fashion under both continuous and washout conditions. Similar trends were observed with other commercially available BRD4 inhibitors. However, at higher concentrations the decrease was significantly greater with PLX compared with the other three drugs (P < 0.001 for all comparisons under both continuous and washout conditions). Inhibition of proliferation or toxicity was not observed in CLL cells under basal condition. PLX51107 induces accumulation of p21 (cyclin-dependent kinase inhibitor) and IκBα, reduced levels of c-MYC, and modulation of proapoptotic and antiapoptotic proteins (Supplementary Fig. S3B–S3D).
Inhibition of proliferation with PLX51107 was also observed in CpG-stimulated normal B cells under continuous but not washout condition (Supplementary Fig. S3A).
Immune dysfunction is typically observed in CLL and leads to the increased number of infections and second malignancies seen in this disease. In addition, this decreased T-cell immune surveillance permits the survival of leukemia cells. Treatment agents capable of preserving T cells are therefore desired. As shown in Supplementary Fig. S3E and S3F, PLX51107 had no antiproliferative effect in healthy T cells stimulated with CD3/CD28 agonist antibodies under washout conditions. Furthermore, cytokine production by T cells was also not disrupted by PLX51107. The differential result of short-term (washout) versus long-term (continuous) inhibition of BRD4 by PLX51107 on normal B and T cells suggests that short-term BRD4 engagement may generate the desired pharmacologic effect in CLL without inducing nonspecific killing and provides support for the use of agents with a short terminal half-life to improve the therapeutic index of BET inhibition. We next investigated the effect of PLX51107 on CLL cells in the presence of stromal protection. Although primary CLL cells cocultured on stromal cells (9-15c and Hs27) are protected from spontaneous apoptosis, treatment with PLX51107 induces significant cytotoxicity on CLL cells (Fig. 3C) with no effect on stromal cell viability (Supplementary Fig. S3G). This suggests that BET inhibition can overcome stromal protection that contributes to in vivo CLL cell survival and confers resistance to therapeutics used in this disease. Dose-dependent inhibition of cell growth (P < 0.001 for all comparisons except at 0.15 μmol/L) with accumulation of cells in G0–G1 phase and increased apoptosis (Fig. 3D; Supplementary Fig. S3H and S3I) was also observed in malignant B-cell lines (MEC-1 and OCI-LY1) with lower IC50 for PLX51107 compared with iBET762, JQ1, or OTX015. Collectively, these data demonstrate the sensitivity of proliferating CLL cells to BET inhibition and also the ability of BET inhibition to overcome microenvironment-derived resistance to apoptosis. Similar findings were observed with the highly active BCR-targeting CLL therapeutic ibrutinib preclinically (6).
BET Inhibition Selectively Modulates CLL Driver Genes
To gain mechanistic insight into the consequences of BET inhibition on genome-wide BRD4 binding, we augmented ChIP-seq studies to include CpG-stimulated primary CLL cells with or without PLX51107 treatment. BRD4 occupancy was globally decreased (P < 10−6) by PLX51107 in three out of four samples (Supplementary Fig. S4A). More importantly, regardless of the global picture, BRD4 load was decreased in enhancer regions of all samples (Fig. 4A) with significant loss of the SE regions following PLX51107 treatment (Fig. 4A and B). When compared with normal B cell, enhancer regions of CLL samples were also significantly overloaded (P < 10−6; Fig. 4A and B).
A reduced RNA Pol II occupancy following PLX51107 treatment was observed at gene promoters centered ± 5 kb relative to the transcriptional start site (TSS; Fig. 4Ci) and across gene bodies (Fig. 4Cii) indicating a reduced transcription rate at the nearby genes. We next performed microarrays analysis on CpG-stimulated primary CLL cells with or without PLX51107 treatment to gain functional insight on the transcriptional effects of BET inhibition. Microarray analysis revealed that more than 1,300 genes were significantly differentially expressed (>2-fold change, P < 0.01; 745 upregulated and 616 downregulated genes; Fig. 4D; Supplementary Fig. S4B). Comprehensive Ingenuity Pathway Analysis (IPA) of all the modulated genes (up and down) revealed that BRD4 inhibition affects multiple cellular pathways including BCR, PKC, PI3K/AKT, CD40, IL4, IL2, p53, and death receptor signaling (Fig. 4E; Supplementary Table S6A). Inhibition of BRD4, IFNG, TNF, IL12, CD40LG, and FLT3LG as well as activation of the EGR2, FOXO3, CD38, IL1RN, and IL5 pathways were among the most significant effects predicted by differentially expressed genes (Supplementary Table S6B).
A variety of genes were modulated by BRD4, including CXCR5, CXCR4, IKZF1, IKZF3, TCL1, CD180, CCL3, PI3KCG, CTLA4, STAT6, and BLNK. Validation of microarray data was performed for select genes by quantitative real-time PCR at 4 hours (Fig. 4F; Supplementary Fig. S4C) and/or protein expression analysis at 24 and 48 hours in primary CLL cells (Fig. 4G; Supplementary Fig. S4D) as well as malignant B-cell lines (Supplementary Fig. S4E). Of note, overexpression of de novo upregulation of many of these genes has been well described in CLL compared with normal B cells and is also supported by our own data (Supplementary Fig. S5A and S5B; refs. 53–55). Many of these were also shown to be essential for CLL proliferation and associated with a more aggressive disease phenotype and poor response to therapy (56–58). These support the hypothesis that BRD4 controls the expression of many CLL tumor–specific genes and that BRD4 inhibition targets molecular defects in CLL in addition to B-cell lineage pathways.
An integrative approach was then applied to assess the link between BRD4 genomic load and gene expression. First, overlap between changes in BRD4 SEs and gene expression data was assessed. SEs whose ranking decreased upon BET inhibition in at least three out of the four patient samples were selected, then the expression profiles of 166 genes within 5 kb of these SEs were analyzed. Heat maps of unsupervised clustering of these genes with the leading-edge genes annotated and the top canonical pathways identified by IPA analysis are shown respectively in Fig. 5A and B and Supplementary Table S7A. The top signaling pathways identified by IPA to be differentially modulated by PLX51107 include activation of STAT3, BCL6, and CBFB as well as inhibition of IL1, IL4, and IL21 (Supplementary Tables S7B and S7C). Lymphopoiesis is the top biological function predicted to be decreased upon BRD4 inhibition as supported by changes in expression of ARID5B, B2M, BATF, BCL11A, BCL2, BCL6, CCND2, CD44, EIF2AK4, ETV6, and HLA (Supplementary Table S7C). PLX51107 selectively decreased the transcript abundance of 397 genes regulated by BRD4-loaded SEs, in contrast to 205 control genes contained within the least BRD4-loaded enhancers (lower quartile of enhancer regions) in all samples (Fig. 5C). Examples of select gene tracks are shown in Fig. 5D.
Next, overlap between changes in individual BRD4-binding sites and gene expression was evaluated for annotated genes. First, BRD4 peaks (i) within 5 kb of a gene's TSS, (ii) within 5 kb of an H3K27ac mark, and (iii) with at least 2-fold decrease in peak score upon PLX51107 treatment were selected, revealing 537 common genes with decreased BRD4 peaks across the four patient samples. Microarray data were then utilized to identify genes with decreased expression (2-fold decrease and P < 0.05) upon treatment. As a result, 106 genes that are directly regulated by BRD4 were identified (Fig. 5E). Among the top canonical pathways identified by IPA are IL4/IL15/IL2 signaling and JAK/STAT signaling (Fig. 5F and Supplementary Table S8A). IPA upstream regulator analysis revealed that BET inhibition also resulted in inhibition of the BCR complex, STAT1, VEGFA, IFNL1, IL21, IFNγ, and MYC, and activation of IL1RN (Supplementary Table S8B). Additionally, IPA biological function analysis recognized decrease in cell survival, cell viability, cell migration, leukopoiesis, and lymphocyte migration upon BET inhibition (Supplementary Table S8C). Taken together, these data indicate that BRD4 loading at select CLL enhancers underlies the pathway-specific transcriptional consequences of BET inhibition. Moreover, these data raise the possibility that BET inhibition may have broad activity in in vivo models of CLL similar to what we observed with BTK inhibitors such as ibrutinib preclinically (6).
PLX51107 Demonstrates In Vivo Antitumor Effects in Preclinical Models of CLL and Aggressive Lymphoma
We next evaluated the translational potential of BET inhibition in a mouse model of CLL that has been used extensively to evaluate experimental therapeutics for this disease (59–61). Eμ-TCL1 transgenic mice spontaneously develop a CD5/CD19-positive CLL-like leukemia that expresses BRD4 (Supplementary Fig. S6A) and responds to therapies proven effective in CLL (e.g., ibrutinib; refs. 61, 62). PLX51107 induces a dose-dependent cell growth inhibition of B cells derived from Eμ-TCL1 mice with active disease (Supplementary Fig. S6B), with modulation of BRD4 targets such as Myc and Hexim1, whereas Eμ-TCL1 transgene expression remains (Supplementary Fig. S6C). To evaluate BRD4 in vivo target modulation, mice with de novo leukemia, defined as >60 white blood cell (WBC) count, >60% CD19/CD5/CD45-positive peripheral blood lymphocytes (PBL) and splenomegaly, were treated with PLX51107 via oral gavage (20 mg/kg daily) for 8 days. Compared with vehicle-treated mice, PLX51107 significantly reduced leukemic disease burden in peripheral blood and spleen (Fig. 6A and B; Supplementary Fig. S6D) with modulation of BRD4 targets (Fig. 6C). PLX51107 (20 mg/kg, once daily) was also more effective than OTX015 (50 mg/kg, once daily) in reducing tumor burden (Supplementary Fig. S6E and S6F) and in inducing in vivo CLL target modulation in a 30-day study (Supplementary Fig. S6G).
Next we evaluated the long-term in vivo therapeutic effects of PLX51107 utilizing an adoptive transfer model (61, 63) in which leukemic cells from a Eμ-TCL1 mouse are engrafted into healthy wild-type (WT) recipients of the same background. This produces a homogeneous population with the same pathologic findings as the Eμ-TCL1 transgenic mice and a more rapid disease acquisition (weeks instead of months), although time to disease and disease progression can vary depending on the donor used for the engraftment, mimicking the heterogeneous natural history of CLL. PLX51107 significantly prolonged survival compared with vehicle control (Fig. 6D), with marked decrease in leukemia cells in blood (Fig. 6E) and reduction of spleen size (Fig. 6F). In addition, pathologic examination of spleens, bronchus-associated lymphoid tissue, alveolar septa of the lungs, and blood from the cardiac chambers of PLX51107-treated mice revealed depletion of infiltrating leukemic cells identifiable by hematoxylin and eosin (H&E) staining and generally negative for Ki67, a classic proliferation marker (Fig. 6G). We next tested the effect of PLX51107 or ibrutinib in a cohort of mice with engrafted leukemic cells from a different Eμ-TCL1 donor. PLX51107 significantly prolonged survival compared with ibrutinib (Fig. 6H), with significant reduction of spleen weight and blood leukemia cells (Supplementary Fig. S6H and S6I).
To assess the effects of BET inhibition in a highly penetrant, malignant leukemia/lymphoma phenotype analogous to high-grade lymphoma (37), we utilized the Eμ-Myc/TCL1 adoptive transfer mouse model that we recently introduced as a potential model for RT (59). Ex vivo treatment of Eμ-Myc/TCL1-derived splenocytes with PLX51107 (8 hours) suggests that both Eμ-Myc and Eμ-TCL1 transgenes are not affected by BET inhibition (Supplementary Fig. S7), confirming previous findings with JQ1 in a Eμ-Myc model of lymphoma (64). PLX51107 treatment (70 days) significantly prolonged survival (Fig. 7A) and decreased tumor burden (Fig. 7B and C), with marked reduction in spleen size and lymph node masses (Fig. 7D). Histologically, atypical neoplastic lymphocytes were absent in spleens (Fig. 7E), wherein tissue architecture was preserved. Ki67 staining revealed scattered proliferating cells predominantly in the red pulp corresponding mostly to foci of extramedullary hematopoiesis. Similarly, lymph nodes of PLX51107-treated mice were devoid of neoplastic cell infiltration, and Ki67 expression was largely restricted to germinal centers, with few positive cells in the medulla and paracortical zones. Daily administration of PLX51107 was well tolerated and resulted in physiologically relevant plasma levels (AUC0–24h = 96,700 h • ng/mL or 220 μmol/L • h) in WT B6 mice. Collectively, these data provide support for application of PLX51107 in clinical trials directed at both CLL and RT.
A central role for BET bromodomains in governing the transcriptional landscape of many rapidly proliferating hematopoietic cancers has recently been recognized (65). Here, we utilize CLL as a model of slowly proliferating tumor cells to study the relevance of BET bromodomains in driving the expression of genes of the CLL transcriptional program. BRD4 overexpression has been shown to potentially act as an early driving event in lymphomagenesis and relapse-associated events (66), suggesting a role for BRD4 in leukemia disease progression and resistance. Here, we provide mechanistic evidence of BET bromodomains as transcriptional coactivators at SEs that regulate both important pathways (BCR signaling) and genes (TCL1, MCL1, MIR155, IKZF3, etc.) that are known to contribute to the development and progression of this disease. We hypothesize that for other slowly proliferating neoplasms where distinct driving pathogenic pathways have not been identified, the study of BET SE biology and pharmacologic interrogation (BET profiling) may allow identification of new therapeutic targets.
BET inhibition caused proliferation arrest in primary CLL cells even in the presence of microenvironment protection and also significantly improved survival in two independent in vivo models of aggressive CLL.
Our comprehensive functional epigenetic and biochemical studies reveal insight into BET inhibition of oncogenic pathways and identify previously unrecognized targets of BRD4 in CLL. Similar to what was previously reported in other cancers, we observed the presence of a small subset of genes with a disproportionate BRD4 load at their proximal enhancers that are particularly sensitive to BET inhibition.
As these SEs are found adjacent to genes aberrantly expressed, de novo regulated, or with known oncogenic roles in CLL, we hypothesized that SE analysis in CLL could identify previously unrecognized CLL tumor dependencies. Accordingly, we observed direct interaction of BRD4 at regulatory regions of the MIR21, IL4R, IL21R, IKZF3, IKZF1, mir155, and TCL1A genes. The de novo expression of MIR155 and MIR21 in CLL has been reported and associated with a more aggressive disease phenotype and poor response to therapy (57, 58). CLL cells express significantly higher levels of IL4R compared with normal B cells (67), resulting in an increased STAT6-dependent signaling pathway. TCL1A overexpression is also associated with rapid disease progression and poor response to therapy in CLL (68–70). Furthermore, gain-of-function mutations in TCL1A are found in a small subset of patients with CLL (71); TCL1A remains an undruggable target to date. CCND2 is upregulated in CLL (72) and is essential for BCR-induced proliferation and CD5-positive B-cell development (56). Although dysregulation of all the above genes has been documented in CLL (73), their mechanisms of regulation as well as strategies to modulate their expression remain mostly unknown. Here, we demonstrate that MIR21, IL4R, IL21R, MIR155, and TCL1A are all regulated by SEs controlled by BRD4 and can be targeted by PLX51107 in vitro and in vivo in CLL models. Mutations in IKZF3 have recently been proposed as putative CLL drivers (20, 74), and IKZF3 upregulation has been associated with adverse outcome (75). The observed transcriptional modulation of the B cell–specific transcription factors IKZF1 and IKZF3 by BRD4 suggests a therapeutic benefit of BET inhibitors alone or in combination with immunomodulatory drugs (IMiD) in CLL. To this extent, synergistic cytotoxicity of BRD4 inhibitors and the IMiD lenalidomide has recently been demonstrated in preclinical models of primary effusion lymphoma, an aggressive type of non-Hodgkin lymphoma (76).
Our group has defined mutations of either BTK or PLCG2 as mechanisms of resistance to ibrutinib in CLL (77). The transcriptional profiling of PLX51107-treated CLL cells revealed downregulation of multiple core pathways in CLL including B-cell development, BCR and TLR/MYD88 signaling, and cytokine/chemokine interactions. Given the central role of BCR activation in CLL cell maintenance and expansion, our observation that BRD4 regulates multiple components of the BCR pathway (e.g., PLCG2, ZAP70, and PI3K) not only emphasizes an important role of BRD4 in CLL pathology but supports the clinical investigation of BET inhibitors alone or in combination with BCR-targeting agents (e.g., idelalisib and ibrutinib) to overcome resistance and achieve deeper and more durable responses in relapsed/refractory CLL and other B-cell malignancies. This is of critical importance given the increasing number of patients relapsing on targeted therapies including the BTK inhibitor ibrutinib (7, 78). Moreover, the proximal BCR pathway kinases SYK and LYN were reported to be critical for the activation of mutant PLCG2 protein (79), suggesting that intervention with PLX51107 has the potential to overcome acquired ibrutinib resistance that frequently results from reactivation of BCR signaling. This approach was recently shown to synergistically induce apoptosis in ibrutinib-resistant mantle cell lymphoma (80) and diffuse large B-cell lymphoma (DLBCL; 81). Further studies to confirm the therapeutic benefit of this combination in CLL are warranted.
The tool compound JQ1 has been recently reported to downregulate PD-L1 expression in the Eμ-Myc B-cell lymphoma model (82). Here, we show similar effects of the clinical agent PLX51107 on PD-L1 in primary CLL cells. Furthermore, we observed this effect on additional checkpoint molecules (CD200, LAG3, and CTLA4). The regulation of these factors by BRD4 on CLL cells has not been reported to date and points to the potential role of this agent in reversing tumor immune suppression. Comprehensive analysis for the potential immunomodulatory effects of BRD4 inhibition and their functional relevance in CLL is needed, and our data support the clinical application of PLX51107 to target tumor immune evasion alone or in combination with immune-modulatory agents in CLL.
Collectively, our findings demonstrate that BRD4 is enriched at hallmark genes of CLL biology, and that BET inhibition is capable of downregulating multiple survival pathways in malignant CLL B cells. This mechanism of action provides a strong rationale for human clinical investigation of BET inhibitors to target the neoplastic B cells and their supportive microenvironment in relapsed/refractory CLL and other aggressive B-cell malignancies including RT.
Patient Sample Processing, Cell Lines, and Cell Culture
Blood was obtained from healthy volunteers or patients with CLL following written informed consent under a protocol approved by the Institutional Review Board of The Ohio State University (OSU; Columbus, OH) in accordance with the Declaration of Helsinki. All patients examined had CLL as defined by the 2008 International Workshop on Chronic Lymphocytic Leukemia criteria (51), and cells were isolated and cultured as previously described (83). The CLL leukemic cell line MEC-1 was obtained from DSMZ in 2015. MEC-1 cell identity was confirmed in May 2015 via FISH using the CLL probe panel, i.e., CEP12/13q14/13q34, ATM/p53, BCL6, MYC, IGH/CCND1 (Abbott Molecular), and 6q21 (Kreatech; results for 200 analyzed cells were reported). A large number of vials were cryopreserved. For experiments cells were resuscitated, cultured in RPMI (Invitrogen) supplemented with 10% fetal bovine serum (FBS), and used within 3 to 4 weeks from thawing. The DLBCL cell line OCI-LY1 was obtained from Ontario Cancer Institute (Ontario, Canada) in 2016. OCI-LY1 cell identity was confirmed in May 2016 by exome sequencing, and mutation status was compared with publicly available exome sequencing data sets (84, 85), and a large number of vials were banked. For the experiment, cells were cultured in Iscove's Modified Dulbecco's Medium (Invitrogen) supplemented with 20% FBS and used within 3 to 4 weeks from thawing. The human bone marrow stromal Hs27-EGFP cell line was authenticated and provided by Dr. Torok-Storb (Fred Hutchinson Cancer Research Center; ref. 86) in 2010. The murine marrow-derived mesenchymal 9-15c cell line (82) was obtained from RIKEN cell bank in 2012 (87). A large stock was cryopreserved. For the experiments, cells were resuscitated, maintained in RPMI supplemented with 10% FBS as previously described (88), and used within 3 to 4 weeks from thawing. All cell lines were confirmed to be Mycoplasma negative using the MycoAlert Mycoplasma Detection Kit from Lonza according to the manufacturer's instructions and used within 3 to 4 weeks from thawing. See Supplementary Information for details on sample processing.
Mouse B-cell Isolation
Primary B cells were freshly isolated from mouse (WT B6 or TCL1-Tg) spleens using the EasySep mouse pan B-cell isolation kit (STEMCELL Technologies) according to the manufacturer's instructions. B-cell purity was checked using anti-CD19-PE (BD Pharmingen) staining by flow cytometry, and isolated B cells were used for immunoblot analysis of BRD4 levels.
Reagents and Antibodies
The BRD4 inhibitor PLX51107 was provided by Plexxikon Inc. JQ1 and iBET762 were purchased from Selleck Chemicals. OTX015 was purchased from Advanced ChemBlocks Inc. See Supplementary Information for a detailed list of reagents and antibodies.
Proteins extracted from whole-cell lysates were resolved by SDS–PAGE and transferred on nitrocellulose membrane as previously described (89).
Proliferation Assay and Flow Cytometric Studies
MTS and Annexin V-FITC flow-based assays were performed to determine cell proliferation and death, as previously described (90). See Supplementary Information for details.
Experiments were approved by The OSU Institutional Animal Care and Use Committee. For engraftment studies, C57BL/6 WT mice (The Jackson Laboratory) were engrafted with 1E7 cells by tail-vein injection of splenocytes derived from Eμ-TCL1 or Eμ-Myc/TCL1 mice with active disease. At the onset of leukemia (Eμ-TCL1: ≥ 10% CD19/CD5/CD45-positive circulating cells; Eμ-Myc/TCL1: WBC count ≥ 8 and/or ≥ 5% CD19/CD5/CD45-positive circulating cells), mice were randomized to receive treatments as indicated. Vehicle = 10% N-methyl-2-pyrrolidone plus diluent (40% PEG400, 5% TPGS, 5% Poloxamer 407, and 50% water). Mice were sacrificed when meeting early removal criteria (>20% weight loss, impaired motility, splenomegaly, and evident tumor masses), and tissues were collected for further analysis.
RNA Extraction and Microarray Analysis
Total RNA enriched in small RNA (>200 bp) was isolated from primary CLL cells following drug treatment using the miRNeasy Mini Kit (Qiagen) according to the supplier's protocol. Total RNA concentration was measured by using Nano Drop UV/Vis spectrophotometer (2000c, Thermo Scientific). RNA quality was assessed by capillary electrophoresis with a BioAnalyzer 2100 (Agilent Technologies). Afterward, the RNA was pooled (2 μg/patient sample), creating three pools per treatment group. Prior to hybridization, RNA integrity and comparability were tested by BioAnalyzer and only samples with RNA integrity numbers (RIN) > 8.0 were used in microarray experiments. cDNA synthesis, labeling, and the basic microarray analysis was performed by the Genomics Shared Resource at The OSU Comprehensive Cancer Center. Generation of double-stranded cDNA, preparation and labeling of cRNA, hybridization to GeneChip Human Transcriptome Array 2.0 (HTA 2.0; Affymetrix), and washing were performed according to the standard Affymetrix protocol. The arrays were scanned using a GeneChip Scanner 3000 (Affymetrix). Signal intensities were analyzed by Affymetrix Expression Console software.
Quantitative Real-Time PCR
Real-time PCR was performed using cDNA prepared as described (83) using TaqMan gene expression assays. See Supplementary Information for a detailed list of primers.
ChIP-seq and Data Processing
Primary CLL cells (1 × 107 cells per condition) were treated with vehicle (DMSO), or 1 μmol/L PLX51107 with or without CpG oligonucleotides (3.2 μmol/L) for 4 hours. Cells were fixed with 1% formaldehyde for 15 minutes and quenched with 0.125 mol/L glycine. Chromatin was isolated by the addition of lysis buffer, followed by disruption with a Dounce homogenizer. Lysates were sonicated and the DNA sheared to an average length of 300 to 500 bp. See Supplementary Information for detailed cross-link procedure. Genomic DNA regions of interest were isolated using 4 μg antibody against BRD4, H3K27ac, and RNA Pol II. Complexes were washed, eluted from the beads with SDS buffer, and subjected to RNase and proteinase K treatment. Cross-links were reversed by incubation overnight at 65°C, and ChIP DNA was purified by phenol–chloroform extraction and ethanol precipitation.
Illumina sequencing libraries were prepared from the ChIP and input DNAs by the standard consecutive enzymatic steps of end-polishing, dA addition, and adaptor ligation. After a final PCR amplification step, the resulting DNA libraries were quantified and sequenced on Illumina's NextSeq 500. Sequence reads were aligned to the reference genome, and peak locations were identified and annotated for further analysis. See Supplementary Information for additional details.
ATAC-seq and Data Processing
Accessible chromatin mapping was performed using the ATAC-seq method as previously described (24, 91). See Supplementary Information for details.
Data Integration and Visualization
In-house shell and R scripts (https://www.r-project.org) were used for data integration. To describe genome-wide correlation between BRD4 occupancy, H3K27 acetylation, and chromatin accessibility, number of peaks/hotspots within 5 kb of one another were counted and presented as a Venn diagram. To study functional consequences of BRD4 inhibition with PLX51007 treatment, the data integration process was always centered around annotated genes within 5 kb of BRD4 peaks, H3K27ac mark, and expressed in microarray data. The Integrative Genomics Viewer (http://www.broadinstitute.org/igv/) was used for visualization. Annotation files were downloaded from University of California, Santa Cruz.
DAVID functional annotation tool was used to identify enriched gene ontology terms in a given gene list (92). IPA (Qiagen Bioinformatics) software was used for functional annotation of differentially expressed genes regulated by BRD4 inhibition. Fold change information from microarray gene expression data was also provided as an input to IPA. Significantly overrepresented canonical pathways were reported. Upstream transcriptional regulators that could explain the observed changes in gene expression were identified by IPA upstream regulator analysis with a z-score cutoff of +2 for activation and −2 for inhibition. Overrepresented diseases and biological functions with a predicted activation state were identified by IPA diseases/biofunctions analysis with a z-score cutoff of 2 for increase and −2 for decrease.
All analyses were performed using SAS/STAT software, version 9.2 (SAS Institute, Inc.). See Supplementary materials and methods for details.
All raw sequence data are available in the Gene Expression Omnibus via accession number GSE109411.
Disclosure of Potential Conflicts of Interest
No potential conflicts of interest were disclosed.
Conception and design: L. Alinari, W. Spevak, S. Shi, P. Ibrahim, C. Zhang, G. Bollag, J.C. Byrd, R. Lapalombella
Development of methodology: H.G. Ozer, B. Powell, J.S. Blachly, D. Sampath, Y. Ma, W. Spevak, H. Carias, G. Tsang, H. Rezaei, M. Nespi, J.C. Byrd
Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): D. El-Gamal, B. Powell, Z.A. Hing, B. Harrington, S. Mitchell, N.R. Grieselhuber, T.-H. Lai, M. Cannon, L. Beaver, V.M. Goettl, J.A. Woyach, Y. Zhang, P. Severson, R. Shellooe, G. Tsang, K. Dong, A. Marimuthu, B. Matusow, G. Habets, J.C. Byrd
Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): H.G. Ozer, D. El-Gamal, B. Powell, J.S. Blachly, B. Harrington, L. Alinari, L. Brinton, E. Baskin, A.M. Lehman, L. Yu, J. Zhang, Y. Ma, Y. Zhang, P. Severson, R. Shellooe, H. Carias, K. Dong, A. Marimuthu, C. Tantoy, L. Sanftner, H. Rezaei, M. Nespi, B. Matusow, G. Habets, C. Zhang, E.A. Mathé, J.C. Byrd, R. Lapalombella
Writing, review, and/or revision of the manuscript: H.G. Ozer, D. El-Gamal, B. Powell, J.S. Blachly, B. Harrington, N.R. Grieselhuber, L. Alinari, R.A. Baiocchi, D.M. Lucas, J.A. Woyach, D. Sampath, A.M. Lehman, L. Yu, Y. Ma, L. Sanftner, H. Rezaei, C. Zhang, E.A. Mathé, G. Bollag, J.C. Byrd, R. Lapalombella
Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): D. El-Gamal, B. Powell, K. Williams, R.A. Baiocchi, D.M. Lucas, Y. Ma, W. Spevak, R. Shellooe, G. Tsang, T. Ewing, J. Walters, H. Rezaei, M. Nespi, B. Matusow, G. Habets, J.C. Byrd
Study supervision: C. Zhang, J.C. Byrd, R. Lapalombella
Other (scientific input): D. Sampath
We are grateful to the patients and healthy volunteers who provided blood for the above studies and to the OSU Comprehensive Cancer Center Leukemia Tissue Bank (supported by NCIP30 CA016058) for sample procurement. X-ray diffraction data were collected at beamline 8.3.1 at the Advanced Light Source (Lawrence Berkeley National Laboratory), beamline 08ID-1 at Canadian Light Source, and beamline 7.1 at the Stanford Synchrotron Radiation Lightsource (a directorate of the SLAC National Accelerator Laboratory). The authors would also like to acknowledge Alan Flechtner, HTL, and The Ohio State University Veterinary Histology Laboratory for their assistance in all immunohistochemical studies, also supported by NCIP30 CA016058. This work was supported by the National Cancer Institute (K99/R00 CA208017, R01 CA177292, R01 CA206658, R01 CA214046, and R35 CA198183), a Research Scholar grant (129863-RSG-16-158-01-CDD) from the American Cancer Society, NCIP30 (CA 016058), The OSU Comprehensive Cancer Center using Pelotonia funds, and further research support to the Byrd Laboratory from the Four Winds Foundation, the D. Warren Brown Foundation, the Connie Brown CLL Research Fund, and the Sullivan CLL Research Foundation. This study makes use of RNA-sequencing data generated by the Blueprint Consortium. A full list of the investigators who contributed to the generation of the data is available from www.blueprint-epigenome.eu. Funding for the project was provided to the Blueprint Consortium by the European Union's Seventh Framework Programme (FP7/2007-2013) under grant agreement no. 282510 BLUEPRINT.
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.