Overcoming Acquired Epigenetic Resistance to BTK Inhibitors

Abstract

The use of Bruton tyrosine kinase (BTK) inhibitors to block B-cell receptor (BCR)-dependent NF-κB activation in lymphoid malignancies has been a major clinical advance, yet acquired therapeutic resistance is a recurring problem. We modeled the development of resistance to the BTK inhibitor ibrutinib in the activated B-cell (ABC) subtype of diffuse large B-cell lymphoma, which relies on chronic active BCR signaling for survival. The primary mode of resistance was epigenetic, driven in part by the transcription factor TCF4. The resultant phenotypic shift altered BCR signaling such that the GTPase RAC2 substituted for BTK in the activation of phospholipase Cγ2, thereby sustaining NF-κB activity. The interaction of RAC2 with phospholipase Cγ2 was also increased in chronic lymphocytic leukemia cells from patients with persistent or progressive disease on BTK inhibitor treatment. We identified clinically available drugs that can treat epigenetic ibrutinib resistance, suggesting combination therapeutic strategies.

Significance: In diffuse large B-cell lymphoma, we show that primary resistance to BTK inhibitors is due to epigenetic rather than genetic changes that circumvent the BTK blockade. We also observed this resistance mechanism in chronic lymphocytic leukemia, suggesting that epigenetic alterations may contribute more to BTK inhibitor resistance than currently thought.See related commentary by Pasqualucci, p. 555. This article is highlighted in the In This Issue feature, p. 549.

Figure 1.

Epigenetic ibrutinib resistance in ABC DLBCL lines. A, Schematic of the creation of ibrutinib-resistant (IR) pools (pink) and selection of IR single-cell clones (blue) from ibrutinib-sensitive ABC DLBCL parental cell lines. B, Ibrutinib-sensitive ABC DLBCL cells or IR pools and clones maintained in drug or grown without drug for 1 week were marked by transduction with a GFP-expressing retrovirus (GFP⁺), mixed with drug-sensitive (GFP⁻) parental cells, and challenged with 10 nmol/L ibrutinib or vehicle (DMSO). Competitive outgrowth was measured by using the log₂ ratio of the percentage of GFP⁺ IR cells in ibrutinib versus DMSO is plotted over 12 days in culture. C, Parental HBL1 cells, an IR clone with a PLCG2 resistance mutation (R665W), and three independent IR pools were grown in the absence of ibrutinib for the indicated days, in duplicate wells, and challenged with 10 nmol/L ibrutinib or DMSO for 4 days, and then live cells were counted by flow cytometry (calcein⁺, EtBr⁻). Cell numbers are normalized to no drug (DMSO) controls (error bars, SEM). D, Log₂ of the average expression of a signature of genes downregulated by acute ibrutinib treatment (log₂ < −0.4, 24 hours; Supplementary Table S1C), plotted for two ABC DLBCL cell lines as well for IR clones and IR pools of those lines, normalized to ibrutinib (IB)-sensitive parent cells treated with vehicle (DMSO). Significance was determined by ANOVA testing. E, Live cells were counted by flow cytometry from cultures of two ABC DLBCL cell lines, as well as IR clones and IR pools of those lines, from duplicate wells, with and without serial treatment with “epigenetic” drugs 5-azacytidine (AZAC; 250 nmol/L, 5 days, AZAC) followed by the HDAC inhibitor givinostat (HDACi; 50 nmol/L, 5 days). Cell numbers are normalized to DMSO-treated controls (error bars, SEM). See also Supplementary Fig. S1 and Supplementary Table S1.

Figure 2.

Tracking the evolution of ibrutinib-resistance phenotypes. A, Creating single, barcoded cells to track the evolution of ibrutinib resistance. B, One million HBL1 ABC DLBCL cells were transduced (multiplicity of infection of 1:3) such that each cell received one unique barcode. Cells were expanded, and a sample of cells was taken to establish the clonal (barcode) distribution of cells at the start of the experiment. Duplicate cultures were then challenged with increasing concentrations of ibrutinib (0.5 nmol/L week 1 to 10 nmol/L week 5) or cultured with DMSO vehicle alone. One million cells were harvested at weeks 1 and 5. DNA from each sample was subjected to high-throughput sequencing for identification and counting of barcodes (Supplementary Table S2A). Relative depletion or enrichment of each barcode (clone) was calculated, compared with the starting population, and binned with the number of barcodes/bin plotted versus the relative change in clonal frequency (ibrutinib/DMSO). The designated resistance phenotype (sensitive, persistent, resistant) is depicted above the graph. Error bars, SD of counts in each bin for the duplicate samples. C, A similar barcode experiment was carried out on two independently derived, expanded populations of 300 HBL1 cells (A and B), each with unique expressed barcodes (EBC) that could be detected by both next-generation DNA sequencing and scRNA-seq at week 5 following DMSO or ibrutinib (10 nmol/L) treatment versus initial starting populations. Clones with corelated barcodes, DNA versus RNA clonal frequency, were designated as persistent or resistant phenotypes as shown. D, scRNA-seq t-SNE plots of the cells with barcodes from C under control (DMSO) or ibrutinib (10 nmol/L) challenge conditions. E, Summary of gene-expression signature enrichment (Supplementary Table S2B–S2D) from scRNA-seq genes up in both persistent and resistant cells selected with ibrutinib versus DMSO-treated samples; signatures of genes higher in persistent cells than resistant cells selected in ibrutinib and vice versa. Significance of the enrichment is shown with a description of the signature. GCB, germinal center B-cell. F, scRNA-seq t-SNE plots of parental ABC DLBCL lines and three independently derived IR pools of each line (Supplementary Table S2E). G, Overlaps in genes upregulated in IR cells. Top, genes commonly upregulated in scRNA-seq of IR pools versus parental lines (Supplementary Table S2E and S2F). Bottom, overlap in genes upregulated in ≥ 2 of 3 cell line IR pools (F) versus genes upregulated in persistent and resistant cells in ibrutinib versus DMSO samples from the EBC experiment (D). H, The functionally annotated STRING depiction of the intersection genes from the bottom part of G showing genes upregulated in ≥ 2 of 3 cell line IR pools from scRNA-seq and genes higher in persisters and resistors in ibrutinib versus DMSO from the EBC experiment. Unlinked proteins and long intervening noncoding RNAs were removed from this depiction. See also Supplementary Fig. S2 and Supplementary Table S2.

Figure 3.

Epigenetic retuning of oncogenic signaling in ibrutinib resistance. A, Schematic of samples, and subsequent treatments, upon which ATAC-seq was performed in the HBL1 ABC DLBCL line to identify regions of open chromatin. Normalized (to the parental HBL1 control) ATAC-seq peak counts near important ABC DLBCL genes are shown for both IR pools (pink, average of three independent pools) and an IR mutant (PLCG2 R665W) clone (blue), with the peak start site mapped to the human genome (hg19) and the extent of the open region indicated (kb; Supplementary Table S3B). AZAC/HDACi, 5-azacytidine/HDAC inhibitor; UTR, untranslated region. B, Differences in regions of open chromatin near critical ABC DLBCL signaling and survival pathway genes. Chromatin regions more accessible on average in IR pools than in parental HBL1 cells or the PLCG2 resistant mutant (PLCG2 R665W) are indicated (yellow, left side, log₂ difference ≥ 0.3) with a decrease in accessibility of the same region in IR pools when treated with AZAC/HDACi (log₂ change ≤ −0.3; Supplementary Table S3B) indicated in blue (right). C, Percent of regions of open chromatin from the HBL1 parent line or IR pools that have at least one binding motif for the indicated factor(s) based on RSAT (Regulatory Sequence Analysis Tools) analysis (top ENCODE motifs e-value <0.01; Supplementary Table S3C). D, ATAC-seq regions of open chromatin near selected genes that are more accessible on average in IR pools (pink) than the parental line (black) or genetic resistant mutant (dark blue). Also shown: open chromatin determined by DNAse hypersensitivity (black bar, denoting peak region), along with TCF4 transcription factor binding in the parental line (aqua, ChIP-seq), H3K27Ac histone ChIP-seq from HBL1 (68), and fold increase in peak size comparing the average peak signal from IR pools to the parent line (Supplementary Table S3B). E, Measuring the viability of parental ABC DLBCL cells (HBL1 and TMD8) or IR pools, as assessed by flow cytometry, after transduction with retroviruses bearing doxycycline-inducible short hairpin RNAs (shRNA), specific for TCF4 or a control transcript, and a coexpressed GFP marker, over a time course of induction. These represent biological repeats on the parental lines and independently derived IR pools performed at different times. Error bars, SD from the mean. See also Supplementary Table S3.

Figure 4.

Altered dependencies in IR ABC DLBCL. A, Plot of the average difference in CRISPR screen score [(log₂ ΔCSS); Supplementary Table S4A] between IR pools and parental lines from CRISPR dropout screens for all genes in the “Brunello” sgRNA library. Negative ΔCSS value = relative depletion versus parent (more toxicity with gene loss); positive ΔCSS value = relative enrichment versus parent (less toxicity with gene loss). Genes involved in ABC DLBCL pathobiology are shaded red. B, Summary of CRISPR “dropout” screens using the “Brunello” library mapped on to essential pathways controlling ABC DLBCL survival, comparing ibrutinib-sensitive parental lines to IR pools. Average CSS (log₂ scale) was calculated for each gene targeted by guide RNAs in the “Brunello” library (Supplementary Table S4A) and compared between populations. Blue/yellow shading, left side, relative depletion or outgrowth of cells with sgRNA targeting that gene in IR pools. Purple/orange shading, right side, relative essentiality of a gene in the IR pools as compared with the ibrutinib-sensitive parent cell line. See also Supplementary Table S4.

Figure 5.

RAC2 as a mediator of epigenetic ibrutinib resistance. A, Viability of ABC DLBCL parent lines (HBL1 and TMD8) and IR pools after the knockdown of RAC2 as assessed by flow cytometry of cells transduced with retroviruses bearing doxycycline-inducible short hairpin RNAs (shRNA; control transcript or RAC2 targeting) that also constitutively express a GFP marker compared with untransduced cells in the same cultures. These represent triplicate biological repeats on the parental lines and independently derived IR pools performed at different times. Error bars, SD from the mean. B, RAC2 mRNA expression as determined by scRNA-seq (Fig. 2F) for three independently derived IR pools of each ABC DLBCL line, plotted as the percentage of cells in each population having higher RAC2 expression than the parental line average. C, RAC2 protein expression (mean fluorescent index, MFI) as measured by intracellular flow cytometry in parent lines and three independent IR pools, normalized to the parental line RAC2 MFI. D, RAC2 protein expression (MFI) as measured by intracellular flow cytometry in HBL1 parent and IR pools treated with 5-azacytidine plus HDAC inhibitor (AZAC/HDACi) for 4 days. MFI is normalized to the DMSO (control)-treated parental line. E, The expression of critical ABC DLBCL genes is shown from TMD8 parent cells and an IR pool expressing a doxycycline-inducible shRNA targeting RAC2 compared with cells expressing a control shRNA. Note that shRNA induction is partially leaky, resulting in a decrease in RAC2 mRNA at day 0. F, Signature enrichment within the set of genes downregulated upon shRNA knockdown of RAC2 in an IR pool (Supplementary Table S5A and S5B).

Figure 6.

RAC2 protein interactions are a marker of epigenetic ibrutinib resistance. A, Enrichment of RAC2-interacting proteins in ABC DLBCL cells after RAC2–BioID2 expression followed by quantitative mass spectrometry of streptavidin-captured proteins as analyzed by STRING (Supplementary Table S6A and S6B). B, PLA (red puncta) for interaction between RAC2 and PLCG2 (top) or RAC2 and IgM (bottom) in HBL1 parental cells and an IR pool. Nuclei are stained blue (DAPI), and cell membranes are green (wheat germ agluttinin Alexa488). C, Quantitation of PLA puncta (RAC2/PLCG2 and RAC2/IgM) in IR pools (pink box) normalized to parent lines. Puncta values from HBL1 bearing a short hairpin RNA against RAC2 are also shown as a PLA specificity control. D, Quantitation of RAC2/PLCG2 puncta per IgM⁺ cell for each CLL patient category [n = 4/category, from patients pretreatment, during treatment with acalabrutinib although still manifesting lymphocytosis (persistent disease), or on ibrutinib at the time of progressive disease] and for matched pre/on-treatment samples from the same patients, with representative images from these CLL patient samples (showing one of four cases in each category): RAC2/PLCG2 PLA in red; nuclei in blue. Error bars, SEM; P values = Student t test (Supplementary Table S6C and S6D). See also Supplementary Fig. S3 and Supplementary Table S6.

Figure 7.

Targeting epigenetic resistance to BTK inhibitor treatment in ABC DLBCL. A, Summary of the log₂ maximum drug response [dose (μmol/L) at which a maximal inhibitory response is achieved; Supplementary Table S7A and S7B] differences between IR pools and parental ABC DLBCL lines for several classes of drugs, with the number of drugs in each class also shown. Positive numbers indicate resistance in IR pools; negative numbers indicate increased sensitivity in IR pools. Values were averaged across parents and IR pools for individual drugs and further averaged across drug classes that were represented by more than one drug. Error bars, SEM. Imid, immunomodulatory. B, Ibrutinib-sensitive HBL parent and IR pool cells were treated with DMSO, RAC inhibitor EHT1864 (μmol/L), and/or ibrutinib (nmol/L) for 4 days, and live cells (calcein⁺, EtBr⁻) were enumerated by flow cytometry from duplicate wells. Error bars, SD. C, Ibrutinib-sensitive HBL parent, the PLCG2 R665W genetic resistance mutant, and independent IR pool cells were treated with DMSO, RAC inhibitor EHT1864 (8.5 μmol/L), or ibrutinib (10 nmol/L) overnight, and PLCG activity was assessed by ELISA in duplicate wells. Each cell type was normalized to its own untreated DMSO control. Error bars, SD. D, Monitoring the growth of xenografts, by tumor volume, of ABC DLBCL parent lines (HBL1 and DLBCL2) or derived IR pools in NOD/SCID mice (n = 5/treatment) treated with vehicle (DMSO), the BCL2 inhibitor venetoclax, or the combination of both drugs. E, Monitoring the growth of xenografts, by tumor volume, of ABC DLBCL parent lines (HBL1 and DLBCL2) or derived IR pools in NOD/SCID mice (n = 5/treatment) treated with vehicle (DMSO), the BTK inhibitor ibrutinib, the RAC inhibitor EHT1864, or both drugs in combination. See also Supplementary Table S7.