GATA-3 Dependent Gene Transcription is Impaired upon HDAC Inhibition

Abstract

Purpose:

Many peripheral and cutaneous T-cell lymphoma subtypes are poorly responsive to conventional chemotherapeutic agents and associated with dismal outcomes. The zinc finger transcription factor GATA-3 and the transcriptional program it instigates are oncogenic and highly expressed in various T-cell neoplasms. Post-translational acetylation regulates GATA-3 DNA binding and target gene expression. Given the widespread use of histone deacetylase inhibitors (HDACi) in relapsed/refractory cutaneous T-cell lymphomas (CTCL), we sought to examine the extent to which these agents attenuate the transcriptional landscape in these lymphomas.

Experimental Design:

Integrated GATA-3 ChIP-seq and RNA-seq analyses were performed in complementary cell line models and primary CTCL specimens treated with clinically available HDACi.

Results:

We observed that exposure to clinically available HDACi led to significant transcriptional reprogramming and increased GATA-3 acetylation. HDACi dependent GATA-3 acetylation significantly impaired both its ability to bind DNA and transcriptionally regulate its target genes, thus leading to significant transcriptional reprogramming in HDACi treated CTCL.

Conclusions:

Beyond shedding new light on the mechanism of action associated with HDACi in CTCL, these findings have significant implications for their use, both as single agents and in combination with other novel agents, in GATA-3 driven lymphoproliferative neoplasms.

Introduction

GATA-3 is a lymphoid-lineage, zinc-finger transcription factor that plays a pivotal role in the development, differentiation, and homeostatic survival of thymic T-cell progenitors and subsets of their mature progeny (1,2). Diverse T-cell lymphoproliferative neoplasms, derived from conventional (non-malignant) T cells expressing GATA-3, highly express GATA-3 and a GATA-3 dependent transcriptional program, and are generally associated with poor outcomes. For example, transgenic GATA-3 expression in T-cell progenitors promotes T-cell acute lymphoblastic leukemia (T-ALL) development (3), and GATA-3 is preferentially expressed in a subset of human T-ALL (4–6). Among mature (post-thymic) T-cell lymphomas, GATA-3 is highly expressed in a clinically, molecularly, and genetically distinct subset of peripheral T-cell lymphomas, not otherwise specified (PTCL, NOS) (7–9). In advanced-stage cutaneous T-cell lymphoma (CTCL), particularly those with clinically aggressive and genetically complex large cell transformation (LCT), GATA-3 is highly expressed and its gene targets characterized (6). Upon binding gene enhancers and promoters, albeit in a context dependent manner (6), GATA-3 regulates gene transcription, including those with cell-autonomous (e.g. oncogenes, including c-Myc) and non-cell-autonomous (e.g. type 2 cytokines) functions in T-cell lymphomas (6,10). Loss-of-function studies, including those performed in genetically diverse GEM models, demonstrate that GATA-3, and the transcriptional program it instigates, is a bona-fide proto-oncogene in these lymphomas (6,11), and its expression associated with dismal outcomes (6,7,12,13). Among GATA-3 expressing PTCL, NOS, for example, a significant difference in survival was not observed between patients treated with first-line anthracycline-based regimens and those receiving palliative (usually hospice) care alone. Clearly, improved therapeutic strategies for these lymphomas are needed, and those impairing GATA-3 dependent transcriptional programming may have significant therapeutic implications.

Like many transcription factors, GATA-3 lacks an obvious small molecule binding domain suitable for targeting. Therefore, we (6), and others (14), have comprehensively characterized the GATA-3 interactome using proximity-dependent biotin identification, in hopes that this approach may unveil novel therapeutic vulnerabilities. The histone acetyltransferase p300, among other transcriptional co-activators, repressors, and chromatin remodeling proteins, was identified as a GATA-3 binding protein. Orthogonal approaches further demonstrated that p300 acetylates at least two lysine residues adjacent to the c-terminal zinc finger (K358 and K377). Molecular dynamic simulations and multiple site-directed mutagenesis studies demonstrated that K358 acetylation is required for optimal GATA-3 DNA binding, while K377 acetylation impaired DNA binding (6). Therefore, GATA-3 acetylation is dynamically regulated, and its acetylation status, at least in part, controls its DNA binding affinity and target gene expression. This is notable, as GATA family members bind histone deacetylases (15–18), and histone deacetylase inhibitors (HDACi), including belinostat, romidepsin and vorinostat, are routinely utilized in both relapsed/refractory PTCL and advanced-stage CTCL (19–21), leading to significant alterations in chromatin accessibility and gene expression in both malignant and non-malignant T cells (22–24). Therefore, we sought to examine the extent to which HDAC inhibition may alter the GATA-3 dependent transcriptome in CTCL.

Materials and Methods

Cell lines, primary patient specimens, and xenografts

HEK293T (CRL-3216) cells were cultured in DMEM medium containing 10% FBS (Sigma, F4135), 1% penicillin/streptomycin (Corning, 30-002-CI) and 10 mM HEPES (Corning, 25-060-CI). H9 (HTB-176), MyLa, MOLT4 (CRL-1582), SUP-T1 (CRL-1942), MAC1, Karpas 299 cells were cultured in RPMI1640 medium containing 10% FBS, 1% penicillin/streptomycin and 10 mM HEPES. Karpas 299 and MAC1 cells were kindly provided by Dr. Megan Lim, and other cell lines were purchased from American Type Culture Collection (ATCC). Karpas 299 overexpressing GFP-tagged GATA-3 cells were maintained in the presence of 0.5 μg/ml puromycin. All cell lines were mycoplasma free and independently authenticated by short tandem repeat (STR) profiling, performed by ATCC (data not shown), and immunophenotyping (data not shown). Primary malignant T cells were isolated from patients with Sezary syndrome with significant leukemic involvement (≥85% of lymphocytes immunophenotypically aberrant), as determined by concurrent clinical flow cytometric analysis, as previously described (6,11). Sorted cells were cultured with anti-CD3/CD28 Dynabeads (Thermo Fisher Scientific, 11132D) at a 1:1 ratio, where indicated. These studies were conducted in accordance with US federal regulations and the Declaration of Helsinki, and with institutional review board (IRB) approvals and written informed consent was acquired from all patients prior to the study. Belinostat (S1085) and SAHA (Vorinostat) were purchased from Selleckchem. Romidepsin (HY-15149) was purchased from MedChemExpress. CPI-818 was kindly provided by Corvus Pharmaceuticals, Inc. All drugs were dissolved in DMSO and prepared as ≥1000× stock solution. Final concentration of DMSO did not exceed 0.1%. Micromolar range concentrations of romidepsin and belinostat are clinically achievable and maintained for 3–6 hours (25–27). Therefore, unless indicated otherwise, we utilized 1 uM romidepsin or belinostat for 6 hours, as previously suggested by the Bates group (28), for the studies performed here. For the xenograft studies, the flanks of NSG mice (Jackson Laboratory, Strain #005557) were shaved an injected with 5 ×10⁶ H9 cells, as previously described (11). When tumors were well engrafted (≈10 mm in diameter), mice were randomized to treatment with belinostat (generously provided by Acrotech Biopharma, Inc) or vehicle control (n=3/group). Belinostat was administered intravenously daily (x4 days) at a dose (300 mg/kg) that is equivalent to the currently approved dose in humans (1000 mg/m²). Six hours after the last dose (on day 4), mice were euthanized and tumors explants for downstream qRT-PCR and ChIP analyses. The mouse study underwent review and approval by the Institutional Animal Care and Use Committee (IACUC) at the University of Michigan.

Immunoprecipitation (IP)

IP was performed as previously described (6). In brief, cells were treated or non-treated with 1 μM HDACi (Beli, Romi, SAHA) for 6 hours and then collected with IP buffer (Thermo Fisher Scientific, 87788) containing protease inhibitor (Thermo Fisher Scientific, 78410) and benzonase (Sigma, E1014). Cells were lysed at 4°C for 2 hours. After centrifuging at 13,000 r.p.m. for 10 min at 4°C, the supernatant was pre-cleared with Dynabeads Protein G (Invitrogen, 10004D) and then incubated with 2–4 μg of antibodies-conjugated Dynabeads Protein G at 4°C overnight with gentle rotation. The samples were washed 2 times with PBS-T and boiled in the elusion buffer (0.1 M Glycine, pH 2.8) and SDS gel sample buffer at 70°C for 10 min, followed by immunoblot with indicated antibodies.

Immunoblot (IB)

Cells were treated with 1 μM HDACi for 6 hours and 12 hours. Primary patient specimen was treated with 1 μM Belinostat for 16 hours in the presence of anti-CD3/CD28 Dynabeads (Thermo Fisher Scientific, 11132D). Immunoblot was performed as previously described (6). In brief, cells were lysed with RIPA buffer (Thermo Fisher Scientific, 89900) containing protease inhibitor. The protein concentration was quantified by Pierce BCA protein assay (Thermo Fisher Scientific). 30–40 μg samples or IP sample were running using 4%−12% NuPAGE Bis-Tris protein gels and transferred to PVDF membranes (Bio-Rad, 1620177). The membrane was then blocked with 5% nonfat milk and incubated with the following primary antibodies: anti-GATA-3 (Invitrogen, MA1–028), anti-GATA-3 (Boster, A00593–1), anti-GAPDH (CST, 2118), anti-H3K18Ac (CST, 13998), anti-ITK (CST, 2380), anti-c-MYC (CST, 13987), anti-HDAC1 (CST, 5356 or 34589 ), anti-HDAC2 (CST, 5113 or 2545), anti-GFP (Santa Cruz, sc-9996), anti-histone3 (CST, 9715), anti-LaminB2 (CST, 13823). Following day, membrane was washed three time with PBS-T and incubated with horseradish peroxidase (HRP) conjugated anti-mouse (CST, 7076s) or anti-rabbit (CST, 7074s) IgG antibodies were used for 1 hour. Blots were developed using SuperSignal West Femto (Thermo Fisher Scientific, 34096).

Quantitative RT-PCR

5–7×10⁵ cell in 3 ml fresh medium in a 6-well plate were treated with 1 μM HDACi for 6 hours. 1×10⁶ primary patient specimen was incubated with ImmunoCult Human CD3/CD28 T Cell Activator (STEMCELL Technologies, 10971) overnight in 2 ml fresh medium in a 12-well plate and then treated with 1 μM HDACi for 6 hours. RNA was extracted using RNeasy Mini Kit (Qiagen) and 1 μg RNA was converted to cDNA in a 20 μl reaction by reverse transcription using QuantiTect reverse transcription kit (Qiagen). Then 20 μl cDNAs were diluted in 50 μl RNases-free ddH₂O. One μl cDNAs were then quantified by Radiant^™ Green Hi-ROX qPCR Kit (ASI, QS2020) on CFX96^™ Real-time system (Bio-Rad) according to the manufacturer’s instruction. The hypoxanthine phosphoribosyl transferase 1 (HPRT1) was chosen as reference gene. Relative gene expression was quantified by the 2^−ΔΔCq method. The primers used are summarized in Supplementary Table 1. Data were normalized to HPRT1 values, followed by fold change comparison with vehicle control DMSO.

ChIP-qPCR

6 ×10⁶ cells H9 cells or 10×10⁶ primary patient specimens (Pt.#1,3,5) pre-incubated with ImmunoCult Human CD3/CD28 T Cell Activator were treated with 1 μM HDACi (Beli, Romi) for 6 hours. ChIP-qPCR was performed as previously described.⁷ Briefly, cells were subjected to 1.0 % formaldehyde cross-linking for 10 min at room temperature and quenched by 125 mM glycine. Cells were sonicated to obtain chromatin fragments in the 300~500 bp range. Cross-linked chromatin was subsequently immunoprecipitated with GATA-3 antibodies (CST, 5852) or normal lgG (CST, 66362) as control at 4℃ for overnight. Following day, 30 μl of Dynabeads Protein G was added and incubated at 4℃ for 4 hours. Then complex was washed 3 times with 0.5 ml 1 x ChIP buffer and 1 time with 0.5 ml 1x ChIP buffer containing 350 mM NaCl. Elution and de-crosslinking were performed overnight in 150 μl elution buffer (300 mM NaCl, 5 mM DTT and 0.1% SDS in TE buffer, pH8.0) at 65 °C, and RNA and proteins were digested by adding RNase A and proteinase K, respectively. Eluted samples were purified by QIAquick^™ PCR Purification Kit (Qiagen, 28104). The purified DNA from precipitated chromatin was quantified by real time PCR amplification.

GATA-3 DNA-binding assay

HEK293T cells were transfected with GFP alone or GFP-tagged GATA-3. After 48 hours post-transfection, cells were treated with 1 μM HDACi (Beli, Romi) for 6 hours. DNA-binding assay was performed as previously describe (6). In brief, nuclear extract was prepared freshly according to manufacturer’s instruction (Thermo Fisher Scientific, 78835). The nuclear extract was pre-cleared with Dynabeads Protein G for 1 hour at 4°C and then quantified by Pierce BCA protein assay. For the binding reaction, 10 nM DNA oligo probe, 100 μg of nuclear extract were added to the PBS buffer containing proteinase inhibitor in a final assay volume of 200 μl. The binding assay was carried out at 4°C for 1 hour, and then 1 μg GFP antibody (Santa Cruz, sc-9996) was added for overnight. Following day, 8 μl Dynabeads Protein G was added to the assay mixture for 2 hours at 4°C. Non-specific bindings were washed twice with 200 μl PBS-T (0.05% Tween-20 ) at 4°C. GATA-3 protein and probe complex were eluted in 35 μl buffer (0.1M Glycine, pH 2.8) and heated at 70°C for 10 min. GATA-3 DNA binding abilities were measured using real-time PCR, which was performed on the Bio-Rad C1000 Thermal Cycler PCR instrument with 2× SYBR Green (Alkali Scientific Inc., QS2001). Primer sequences used in this study were as follows: 5’-AGAATGTAGCCCTGGACTTC-3’ and 5’-TTCTCCCGCTCGCTATCA-3’. Ct value was normalized to Ct value from the sample transfected with GFP-tagged GATA-3 without probe. The relative GATA-3 DNA binding abilities were expressed as a fold change of GFP alone vs. the GFP-tagged GATA-3. Three independent experiments were performed.

Flow Cytometry

H9 cells were treated with 0.01, 0.1 and 1 μM HDACi for 8 hours. Cells were stained with 1 μg isotype IgG control (Thermo Fisher Scientific, 12-4714-82) or CCR4 antibody (BD Biosciences, 561110) for 15 min in dark at room temperature. Flow cytometry data was acquired using Attune Flow Cytometers (Thermo Fisher Scientific) and the data were analyzed by FlowJo vX.0.6 (Tree Star, Inc., Ashland, OR). Delta mean fluorescent intensity (ΔMFI) of CCR4 expression was calculated by MFI _CCR4 − MFI _{isotype control} and data were presented as ΔMFI. Difference between DMSO and HDACi was analyzed by unpaired Student’s t test.

Dwell time determination of EGFP-tagged GATA-3 in Karpas 299

5–10×10⁴ cell were untreated or treated with 1 μM Beli for 5.5 hours. Then cells were immobilized on CD45 surface (in imaging buffer containing HDACi) for 30 min. GATA-3 imaging and dwell time were determined as previously described (6).

Cell viability

Primary patient specimens (Pt.#7–12) were incubated with anti-CD3/CD28 Dynabeads (Thermo Fisher Scientific, 11132D) beads (1:1 ratio) and treated with DMSO or 1 μM CPI-818 and/or 1 nM Romidepsin for additional 48 hours. Cell viability was determined using CellTiter-Glo (Promega # G7572) as previously described (6). Cell viability was calculated as (mean luminescence with beads under each treatment / mean luminescence without beads) × 100%.

RNA sequencing (RNA-seq)

H9 cells or primary patient specimens (Pt.#3–5, incubated with ImmunoCult Human CD3/CD28 T Cell Activator) were treated with 1 μM HDACi (Beli, Romi) for 6 hours. RNA-seq was performed, as previously described (6). Briefly, sample libraries were prepared using the NEBNext Ultra II Directional RNA Library Prep Kit for Illumina (NEB #E7760L), Final libraries were checked for quality and quantity by TapeStation (Agilent) and Qubit (Thermo Fishser Scientific). This pool was subjected to 151bp paired-end sequencing according to the manufacturer’s protocol (Illumina NovaSeq). Bcl2fastq2 Conversion Software (Illumina) was used to generate de-multiplexed Fastq files. This data is available in Gene Expression Omnibus (accession: GSE219065).

Chromatin immunoprecipitation (ChIP) sequencing

ChIP was performed was performed as previously described (6). Briefly, cells were treated with 1 μM HDACi (Beli, Romi) for 6 hours before fixing using formaldehyde (1% final concentration) for 10 min and stopping with 125 mM glycine addition for 5 min at room temperature. After washing, cells were lysed and digested with Micrococcal Nuclease for 20 min at 37 °C. Nuclei were then sonicated with two pulses of 40 seconds each and 30 s incubation on wet ice at setting 4 using an ultrasonic cell disruptor (Microson) with an 18-inch probe. Chromatin was clarified (16,000× g, 4°C, 10 min) and incubated with GATA-3 and IgG antibody overnight. Chromatin was then incubated with 30 μL Dynabeads Protein G (Invitrogen #10004D) for 4 hours and washed. Then chromatin was washed and eluted overnight at 65 °C, and RNA and proteins were digested by adding RNase A and proteinase K, respectively. Eluted samples were purified by QIAquick^™ PCR Purification Kit (Qiagen #28104). For ChIP-seq library preparation, NEBNext ChIP-Seq Library Prep Master Mix Set for Illumina (NEB #E6240) was used according to the manual’s instructions. Library quality was determined by TapeStation (Aglient) prior to being sequenced on the Illumina NovaSeq-6000 platform.

Bioinformatics

Raw reads were quality checked with FastQC (v0.11.8) and adapter trimmed with Cutadapt (v2.3). Trimmed reads were mapped to the reference genome GRCh38 (ENSEMBL) using STAR (v2.7.8a) and assigned count estimates to genes using RSEM (v1.3.3) with default parameters. Differentially expressed genes were determined using expected counts by DESeq2 package (v1.32), p-value <−0.05 and fold-change >=1.5 or <=0.67 were considered statistically significant. Heatmap were generated by pheatmap (v1.0.12) using ‘WARD’ clustering method. Pathway enrichment analysis was performed by Metascape (v3.0) with default parameters. GATA-3 target gene set enrichment was analyzed by GSEA (v4.1.0), using Kolmogorov-Smirnov test and in-house gene targets database. Pathways were considered significant if FDR <= 0.25. For motif enrichment analysis, overlap gene lists of belinostat and romidepsin regulated genes in H9 cells and primary patients were collected and analyzed by RcisTarget (v3.16) against a collection of transcription factor binding database. Normalized enrichment score (NES) was calculated under 0.005 receiver operating characteristic (ROC) threshold.

ChIP-seq analysis were performed as previously described. Bigwig files, Heatmaps and profile plots for various ChIP-seq were created by Deeptools (v2.0) and visualized by WashU epigenome browser.

Statistical analysis

GraphPad Prism 9.0 was used for data analysis. Data were presented as mean ± s.e.m. Unless indicated otherwise, data was analyzed by the paired or unpaired Student’s t-test, as appropriate. P values less than 0.05 were considered statistically significant.

Data and Materials Availability

The raw data for this study are available from the corresponding author upon reasonable request. The sequencing data is available in Gene Expression Omnibus (accession: GSE219065).

Results

HDAC inhibition significantly alters the GATA-3 dependent transcriptome in CTCL

In order to investigate early transcriptional changes upon HDACi treatment, a well characterized CTCL cell line (H9) and primary Sezary Syndrome (SS) specimens (n=3) were transcriptionally profiled. Enrichment analysis was performed using the differentially expressed genes we identified (Supp. Table 2), and a convergence on relevant T-cell growth and survival pathways was observed (Fig. 1A). Transcription factor (TF) binding motif analysis revealed that Sp1 binding sites, a ubiquitous TF in many cancers that is functionally regulated and post-translationally acetylated upon HDACi treatment (29,30), were enriched among differentially expressed genes (Fig. 1B). GATA-3 target genes were differentially expressed upon HDACi treat (Fig. 1A), and binding sites for GATA-3, and TFs it transcriptionally regulates (e.g. IKZF1, TCF7, LEF1), were similarly enriched among HDACi responsive genes (Fig. 1B). Among the 5,510–6,811 transcripts that were differentially expressed upon either belinostat or romidepsin treatment, respectively, ≈6–7% of transcripts were GATA-3 targets. Therefore, supervised hierarchical clustering, restricted to previously identified GATA-3 target genes in CTCL (6), was performed (Fig. 1C). GATA-3 dependent transcripts were significantly altered upon treatment with belinostat or romidepsin, (Fig. 1C). In H9 cells, 854 GATA-3 dependent target genes were previously identified (6), and significant transcriptional differences were observed in 45% (n=384) and 51% (n=435) of target genes upon treatment with belinostat or romidepsin, respectively (Fig. 1D, Supp. Table 3). Gene set enrichment analysis (GSEA) demonstrated that HDACi exposure similarly inhibited the GATA-3 transcriptional program in primary SS specimens (Fig. 1E).

Figure 1. HDAC inhibition transcriptionally reprograms malignant T cells in CTCL.

A, H9 cells and primary patient specimens (n=3, patient #3–5) were treated with Belinostat (Beli; 1 μM), Romidepsin (Romi; 1 μM) or vehicle control (DMSO) for 6 hours and RNA-seq performed. Pathway analysis was performed using differentially expressed transcripts and ranked by their FDR value and Z-score. Selected pathways, and the number of GATA-3 target genes included, are indicated at left. B, Transcription factor (TF) binding motif enrichment analysis was performed on HDACi regulated genes from H9 cells and primary patient specimens. Normalized enrichment score (NES) > 0 (dotted line) indicates significant TF binding motif enrichment. The highest NES was observed for Sp1, as indicated, and GATA-3 is highlighted in green. GATA-3 dependent TF’s are listed at right and those shared between H9 cells and primary patient specimens highlighted. C, Unsupervised hierarchical clustering was performed utilizing GATA-3 target genes. The treatment group, sample type, the extent to which a specific transcript was differentially expressed upon Beli or Romi (or both) exposure, and whether GATA-3 is a transcriptional activator or repressor at these loci are indicated in the legends at right. D, GATA-3 target genes in H9 cells (n=854) and genes differentially expressed upon Beli (n=5510) or Romi (n=6811) treatment are shown in Venn diagrams. Expression of 45% (n=383) and 51% (n=434) of GATA-3 target genes were significantly altered by Beli and Romi treatment, respectively, 74% (n=349) of which were shared. E, GSEA for GATA-3 target genes, showing negative enrichment of GATA-3 target genes upon HDAC inhibitor treatment in primary patient specimens (patient #3–5).

In order to validate these findings, we selected previously validated genes that are transcriptionally activated by GATA-3 and are biologically and therapeutically relevant (i.e. GATA-3, ITK, IKZF3, CCR4, and c-Myc). In CTCL cell lines and independent primary specimens, exposure to either romidepsin or belinostat was associated with a significant decrease in these GATA-3 dependent transcripts (Fig. 2A), and in additional transcripts that have been validated more recently (Supp. Fig. 1). Similar results were observed in vivo using cell line xenografts after 4 consecutive days of belinostat treatment (Fig. 2A).Conversely, increased transcription of genes that are transcriptionally repressed by GATA-3 was observed upon HDACi treatment in vitro and in vivo (Fig. 2B). Similar changes in GATA-3 target gene expression were observed using previously published datasets including paired primary CTCL specimens treated with romidepsin (Supp. Fig. 2) (31,32). As many GATA-3 target genes are biologically and therapeutically relevant, protein level expression was examined upon HDACi exposure, and a significant reduction in GATA-3 target gene expression, including ITK (Fig. 2C) and CCR4 (Fig. 2D), was observed. The observation that HDACi significantly decreased CCR4 expression is consistent with previous observations (33), and may have significant implications for the use of mogamulizumab in combination with HDACi in these patients. Collectively, these data provide new insight, demonstrating that HDACi significantly impair oncogenic transcriptional programs driven by GATA-3.

Figure 2. HDACi alters GATA-3 target genes expression.

A, Expression of representative target genes that are transcriptionally activated by GATA-3 (including GATA-3, ITK, IKZF3, CCR4, c-MYC) were examined by qRT-PCR upon HDAC inhibitor treatment in H9 (●), MAC1 (■), MOLT4 (▲), SUP-T1 (◆), Pt.#1 (), Pt.#2 (), Pt.#3 (), Pt.#4 () and Pt.#5 () cells. Cells were treated with the HDACi indicated for 6 hours. Data were normalized to HPRT1, followed by fold change comparison with vehicle control. Similar experiments were performed in NSG mice with established H9 xenografts treated intravenously with vehicle control (n=3) or belinostat (n=3). Target gene expression was similarly examined by qRT-PCR in tumors explanted 6 hours after the last dose of belinostat. B, Expression of representative target genes that are transcriptionally repressed by GATA-3 were similarly examined by qRT-PCR. The genes included (in B) were not GATA-3 targets in all cells examined. Therefore, data for those cells (and specific target) were not examined. Target gene expression was similarly examined in H9 xenografts. C, Immunoblot analysis of GATA-3 target gene products (GATA-3, ITK, IKZF3, c-MYC) was performed in representative cell lines and a primary patient specimen (Pt. #6) upon HDAC inhibitor treatment, as indicated. D, CCR4 expression was determined by flow cytometry (representative example shown at left), and the data summarized at right. ns: not significant. *P < 0.05. **P < 0.01. ***P < 0.001.

HDAC inhibition increases GATA-3 acetylation

We next examined the association between GATA-3 and class I HDACs in CTCL cell lines. HDAC1/HDAC2 and GATA-3 binding was observed (Fig. 3A), suggesting that HDAC inhibition may lead to GATA-3 hyperacetylation. Treatment with the three clinically available HDACi led to a significant increase in GATA-3 acetylation (Fig. 3B). We had previously demonstrated that acetylation at two lysines (K358, K377) dynamically regulated GATA-3 DNA binding, such that K358 acetylation enhanced DNA binding, whereas K377 acetylation impaired DNA binding (6). Therefore, site directed mutagenesis was performed, replacing lysine with arginine (K-R) at both amino acid residues, mimicking unacetylated lysine at these mutated sites. In contrast, gain-of-function mutants were generated by replacing lysine with glutamine (K-Q). Orthogonal strategies demonstrated that K358 acetylation is dominant, as DNA binding was significantly impaired in the K-R double mutant, but significantly increased in the K-Q mutant (6). Therefore, we utilized unmutated GATA-3 and the K-R/K-Q double mutants, overexpressed in Karpas 299 cells (which do not express endogenous GATA-3), to examine the extent to which HDAC inhibition leads to GATA-3 hyperacetylation at alternative sites. While acetylation was globally diminished in the K-R mutant (Fig. 3C), a significant increase in GATA-3 acetylation was observed upon belinostat treatment (Fig. 3C). Similar results were obtained using the K-Q double mutant overexpressed in HEK293T cells, further demonstrating enhanced GATA-3 acetylation at non-K358/377 residues in the presence of HDACi (Fig. 3D).

Figure 3. HDACi leads to increased GATA-3 acetylation.

A, GATA-3 co-immunoprecipitation (IP) was performed in cell lines and cell lysates immunoblotted (IB) for GATA-3, HDAC1 and HDAC2, as indicated. B, IP was performed using anti-acetylated lysine (AcK) in cells treated for 6 hours with vehicle control or the HDAC inhibitors Belinostat (Beli), Romidepsin (Romi), or Vorinostat (SAHA), and cell lysates immunoblotted (IB) with GATA-3 or H3K18Ac (as a positive control). C, IP/IB was similarly performed in Karpas 299 cells overexpressing GFP-tagged, wild-type (WT), K358R/K377R (K-R), or K358Q/K377Q (K-Q) GATA-3 following Beli treatment. D, IP/IB was similarly performed in HEK293T cells transiently transfected with GFP-tagged wild-type (WT) or K358Q/K377Q (K-Q) GATA-3 following Beli treatment.

HDACi mediated GATA-3 hyperacetylation impairs DNA binding

As GATA-3 acetylation regulates its ability to bind DNA and regulate transcription (6), we examined GATA-3 DNA binding using a previously described stem-loop DNA probe containing a GATA consensus binding motif and sequences for PCR amplification (6). HDACi abolished probe binding to GATA-3 in this assay (Fig. 4A), suggesting that HDACi-mediated GATA-3 acetylation impairs DNA binding. Recognizing the potential limitations of an in silico approach, Karpas 299 cells overexpressing GFP-tagged, K-Q (K358Q/K377Q) GATA-3 were treated with DMSO or belinostat and single-molecule imaging performed. Belinostat treatment led to a qualitative reduction in prolonged (>4 seconds) dwell times, and a significant reduction in dwell times overall (Fig. 4B). Genome wide GATA-3 binding was then examined in an unbiased manner by ChIP-seq in H9 cells, and a significant reduction in GATA-3 binding peaks was observed (Fig. 4C–E, Supp. Fig. 3).

Figure 4. HDACi impairs GATA-3 DNA binding.

A, GATA-3 DNA binding assay was performed in HEK293T cells transiently transfected with GFP alone or GFP-tagged GATA-3. GFP-GATA-3 transfected cells were treated with 1 μM HDACi for 6 hours, as indicated. Nuclear lysates were obtained, and IB for GFP (indicated by asterisk), and H3K18Ac performed. Relative GATA-3 binding to negative control and GATA-3 binding-site containing DNA probes was determined by qPCR, and the data from 3 independent replicates summarized. B, Distribution of GATA-3 dwell times (>1 second) measured by single-molecule microscopy in Karpas 299 cells expressing a mutated, gain-of-function (K358Q/K377Q) GATA-3 following treatment with vehicle control (n=113 nuclei, black) or Beli (n=79 nuclei, red). GATA-3 dwell times presumed as DNA binding events were calculated; dwell times <1 second were considered transient events and were excluded from the analysis. A Welch’s unpaired T test was used to determine significance. C, Average profiles of ChIP-seq peaks at GATA-3 binding peaks (n = 2849) in untreated and treated (Belinostat and Romidepsin) H9 cells, presented as reads per genomic content (1x normalization) to visualize the location of the peaks. D, Average ChIP-seq metagene occupancy heatmaps on binding peaks of GATA-3 target genes (n=854) in H9 cells treated with Belinostat or Romidepsin for 6 hours. E, Representative ChIP-seq peaks are shown for ITK, CCR4, IKZF3 and KLF6) in H9 cells treated with Belinostat or Romidepsin for 6 hours. ns: not significant. *P < 0.05. ****P<0.0001.

In order to further validate these findings, GATA-3 binding to specific and previously validated target genes was examined. H9 and SUP-T1 cells (Fig. 5A) and primary SS specimens (Fig. 5B) were treated with DMSO, romidepsin, or belinostat, and GATA-3 binding to selected gene targets was examined by ChIP-qPCR. Impaired DNA binding to selected target genes was similarly observed following treatment in a xenograft model (Fig. 5C). A significant reduction in GATA-3 DNA binding at these loci was observed following HDAC inhibition. These findings, particularly when viewed in light of the gene expression changes observed in HDACi treated cells (Fig. 1), demonstrate that GATA-3 hyperacetylation upon HDAC inhibition significantly impairs GATA-3 DNA binding and GATA-3 dependent gene transcription.

Figure 5. HDACi impairs GATA-3 binding at its target genes, including ITK.

A-B, GATA-3 ChIP and binding enrichment at selected GATA-3 target genes was performed in H9 (left), SUP-T1 (right) (A) and primary patient specimens (patient #1,3,5, in B) following HDACi treatment (1 μM) or vehicle control, as indicated. Data were analyzed using 2-tailed unpaired (A) and 1-tailed paired (B) Student’s t test. C, GATA-3 ChIP and binding enrichment at selected GATA-3 target genes was similarly performed in H9 xenografts following treatment with vehicle control or belinostat. D, CRISPR/cas9-mediated GATA-3 knockout (G3 KO) was achieved in SUP-T1 cells, as previously described. G3 KO cells were transduced with GFP or GFP-GATA-3, and cells treated with Beli, Romi, or vehicle control, as before, and ITK (left) and c-MYC (right) transcripts quantified by qRT-PCR. E, Cell viability in primary patient specimens (patient #7–12) treated with ITK inhibitor CPI-818 (1 μM) or/with Romi (1 nM) for 48 hours in the presence of anti-CD3/CD28 beads. Each symbol represents one patient specimen. ns: not significant. *P < 0.05. **P< 0.01. ***P<0.001.

As selected GATA-3 gene targets are therapeutically targetable (e.g. ITK, CCR4), these findings have implications for the development of future combinatorial strategies, which we sought to examine further. The Tec family kinase IL-2 inducible T-cell kinase (ITK) plays a pivotal role in T-cell receptor driven T-cell lymphomagenesis (11), and has emerged as an attractive therapeutic target in GATA-3 associated T-cell neoplasms (6). A first-in-class, selective and irreversible ITK inhibitor (CPI-818) is under investigation in T-cell lymphomas (ClinicalTrials.gov Identifier: NCT03952078). Therefore, we sought to examine the extent to which alterations in its expression upon HDACi are GATA-3 dependent. A gene editing approach was adopted to knockout GATA-3 in SUPT-1 cells, as previously described (6). Belinostat decreased ITK expression in control cells, as predicted (Fig. 5D). However, no significant change was observed upon HDACi treatment in GATA-3 deficient cells, whereas restoration of GATA-3 expression in these cells restored responsiveness to HDAC inhibition (Fig. 5D). Therefore, HDACi-dependent effects on ITK expression may be entirely explained by altered GATA-3 acetylation. Similarly designed experiments were performed examining c-myc, but in contrast to ITK, c-myc transcripts were reduced upon belinostat treatment, irrespective of GATA-3 status (Fig. 5D), suggesting additional HDACi-dependent (and GATA-3 independent) modes of transcriptional regulation at the c-Myc locus. To investigate vertical pathway inhibition as a novel combinatorial strategy with these agents, primary SS specimens were treated for 48 hours in the presence of CPI-818 and/or romidepsin, and cell viability determined (Fig. 5E). As primary malignant T cells undergo spontaneous apoptosis upon ex vivo culture, viability was maintained utilizing CD3/CD28 activation beads, as previously described (11), and viability normalized to cell viability in the absence of bead stimulation. While single-agent romidepsin or CPI-818 impaired T-cell survival and expansion by ≈50%, a significant and additive reduction in viability was observed in cells treated with the combination (Fig. 5E), suggesting that vertical pathway inhibition (Fig. 6) incorporating HDAC and ITK inhibitors is mechanistically rational.

Figure 6. ITK and HDAC inhibitors impair expression of GATA-3 and its target genes.

The ITK selective inhibitor CPI-818 inhibits T-cell receptor (TCR) dependent NF- κB activation, which is required for GATA-3 upregulation following TCR engagement. GATA-3 is post-translationally acetylated by p300 at multiple lysines, including K458, K377, and other lysine(s) not yet identified. HDACi lead to GATA-3 hyperacetylation and impair its ability to bind DNA at its gene targets. Consequently, HDACi impair the expression of those genes (including GATA-3 itself and ITK) for which GATA-3 is a transcriptional activator (indicated by red down arrow) and increases the expression of those genes (indicated by green up arrow) for which it is a transcriptional repressor. Therefore, ITK and HDAC inhibitors may be rationally combined in GATA-3 driven T-cell lymphomas.

Discussion

Histone acetylation and chromatin accessibility are regulated by opposing histone acetyltransferases and deacetylases, both of which are rational therapeutic targets due to their dysregulation in human cancers. While the first HDACi approved for human use (vorinostat) was granted in CTCL well over a decade ago, the mechanisms of action (and resistance) to these agents are pleiotropic and poorly understood. HDACi significantly accentuate existing patterns of DNA accessibility and alter gene transcription in malignant T cells, and the extent of these transcriptional alterations may be associated with sensitivity to HDACi (22). However, the transcriptional alterations observed in non-malignant T cells (and likely other constituents of the host immune response and tumor microenvironment) are also profound, suggesting that the host immune response also contributes to the mechanism of action associated with HDACi (22). Further adding to this complexity, HDAC regulate the post-translational acetylation of non-histone proteins with a role in T-cell lymphomagenesis, including those like p53 (34), NFkB (35), c-Myc (36), and STAT transcription factors (37). Post-translational acetylation of these transcription factors regulates their stability and affinity for other proteins and/or DNA.

Members of the GATA family of transcription factors are post-translationally acetylated, including GATA-1 (38–40), GATA-2 (41), GATA-4 (16,42), and GATA-5 (43). Seven lysines are acetylated in GATA-1, six of which are conserved in GATA-3, prompting Yamagata et. al. to remove specific GATA-3 lysines by site-directed mutagenesis, and a subset of the GATA-3 mutants generated were hypoacetylated (43). These earlier findings are certainly consistent with our own, as we demonstrated that acetylation at K358 and K377 have discordant effects on GATA-3 DNA binding and target gene expression (6). While we demonstrated that HDACi leads to increased GATA-3 acetylation at alternative (non-K358/K377) site(s), we were unable to identify the specific site(s) acetylated in the presence of HDACi, which is a limitation of the work presented here. This is possibly due to poor representation of six lysines in our mass spectrometry data (6). Nonetheless, the orthogonal approaches adopted here demonstrate that HDAC inhibition leads to GATA-3 hyperacetylation, diminished DNA binding, and significant transcriptional reprogramming in GATA-3 driven T-cell lymphomas.

As p300 and class I HDACs bind GATA-3, and its acetylation state regulates its DNA binding affinity and target gene expression (6), the extent to which HDACi functionally impact GATA-3, while certainly a reasonable question, is also a clinically significant one, as HDACi are routinely utilized in both cutaneous and peripheral T-cell lymphomas, many of which highly express GATA-3 and its target genes (6–8,11). Herein, we demonstrated that clinically available HDACi significantly alter the transcriptome in malignant T cells, effecting pathways required for cell growth and survival, many of which include a significant number of GATA-3 target genes. Transcription factor motif analysis further demonstrated enrichment for GATA-3 binding motifs, as well as binding motifs for a number of other transcription factors that are also GATA-3 target genes [e.g. IKZF1, TCF7, LEF1 (6)], among HDACi responsive genes. Therefore, HDACi significantly impair GATA-3 dependent transcription, by directly impairing its DNA binding and gene expression, but also indirectly via secondary effects on the transcription factors it regulates, some of which are oncogenic in their own right [e.g. IKZF1 (44)].

While challenging, given the pleiotropic affects associated with HDACi, GATA-3 expression itself, or dynamic changes in GATA-3 dependent gene expression upon HDACi treatment, may warrant future consideration as predictive biomarkers. However, the work presented here has more immediate clinical implications, which we attempted to highlight. That is, HDACi dependent effects on GATA-3 target genes that are therapeutically targetable have potential implications for combinatorial strategies. For example, the Tec family kinase ITK is both a therapeutic vulnerability and GATA-3 target gene in T-cell lymphomas (6,11). The data presented here suggests that a first-in-class selective ITK inhibitor (CPI-818) may be rationally combined with HDACi in these lymphomas (Fig. 6). This approach, including recapitulation of our findings here with single-agent HDACi, may warrant future study in GEM and PDX models. GATA-3 was previously shown to confer resistance to various novel and chemotherapeutic agents (11), and synergy between HDACi and other novel agents has been described pre-clinically utilizing cell line models that are GATA-3 dependent (45,46). These observations may suggest that the synergy observed between HDACi and other agents may be explained, at least in part, by HDACi mediated impairment of a GATA-3 dependent transcriptional program that is associated with chemotherapy resistance. Conversely, and consistent with a prior report (33), we observed decreased CCR4 expression upon HDACi exposure, suggesting that concurrent treatment with mogamulizumab may be suboptimal. However, we would anticipate that cessation of HDACi would restore homeostasis, restoring CCR4 expression to basal levels, thus supporting sequential use of these agents. A sequential approach may be supported by the observation that prior treatment with an HDACi did not appear to compromise the clinical efficacy of mogamulizumab (47). If so, disease classification and GATA-3 expression are relevant considerations when analyzing clinical trial datasets obtained in mixed populations of patients treated with HDACi-containing multiagent chemotherapy regimens (48).

GATA-3 and its target genes identify clinically and transcriptionally distinct T-cell lymphomas, including CTCL, that are resistant to conventional chemotherapeutic agents and generally associated with dismal outcomes (6–9). Therefore, GATA-3 is a rational, albeit challenging, therapeutic target in these T-cell lymphomas. We have previously explored “upstream” strategies to target GATA-3 by exploiting transcription factors that regulate its expression (7,49), and have more recently explored “downstream” strategies by targeting its target genes, including ITK (6). To the best of our knowledge, the work described here is the first – and a clinically approved – strategy to therapeutically target GATA-3 directly, as HDACi significantly increase GATA-3 acetylation, and by doing so, impair its ability to bind DNA and regulate transcription of its gene targets. Consequently, these findings have significant implications for the interpretation of existing clinical trial datasets and for the development of future combinatorial strategies utilizing an HDACi-containing backbone in GATA-3 driven T-cell lymphomas.

Translational Relevance.

The first clinical approval for any histone deacetylase inhibitor (HDACi) was obtained in cutaneous T-cell lymphomas (CTCL), as HDACi are associated with a clinically significant response rate, some of which are durable. Despite widespread use their mechanism of action, while certainly pleiotropic, is incompletely understood. GATA-3 is a proto-oncogene in various T-cell lymphomas, including CTCL, and is functionally regulated upon post-translational acetylation. Herein, we demonstrate that HDAC inhibition increases GATA-3 acetylation and impairs its ability to bind DNA and transcriptionally regulate its target genes. Therefore, these findings have significant therapeutic implications for the rational use of HDACi in GATA-3 driven T-cell lymphomas.