Introduction

The histone deacetylase (HDAC) inhibitors romidepsin and vorinostat are approved for the treatment of cutaneous T-cell lymphomas.1, 2, 3 These and additional HDAC inhibitors are currently undergoing clinical testing in various neoplasms, including B-cell lymphomas, where they have shown clinical activity.3, 4, 5, 6, 7, 8 As a class, these agents inhibit deacetylation of lysines in multiple proteins, including histones, and enhance transcription of a wide range of genes.9, 10, 11 This altered transcription is widely thought to contribute to the antineoplastic effects of HDAC inhibitors, including in B-cell neoplasms,6, 11 although other mechanisms have also been proposed.12

HDAC inhibitors have also been widely reported to induce apoptosis, which can be triggered by either the mitochondrial or the death receptor pathway in susceptible cells.9 The mitochondrial pathway is regulated by the BCL2 family of proteins, which is comprised of proapoptotic sensors (BH3-only proteins such as BIM, PUMA, NOXA and truncated BID), apoptotic effectors (BAX and BAK), and apoptosis inhibitors (for example, BCL2, BCLXL and MCL1). Interactions between these proteins regulate mitochondrial outer membrane integrity and cell survival. When cells encounter unfavorable conditions, one or more of the BH3-only proteins is typically increased or activated and can then directly bind and activate the apoptotic effectors BAX or BAK.13, 14, 15, 16, 17 Upon activation, BAX and BAK permeabilize the mitochondrial outer membrane, triggering release of cytochrome c and activation of caspase 9, which initiates an intracellular protease cascade that results in the apoptotic phenotype.18, 19 Conversely, anti-apoptotic BCL2 family members such as BCL2, BCLXL and MCL1 bind and neutralize activated BH3-only proteins as well as partially activated BAX and BAK.20, 21, 22 Importantly, increased BCL2 expression, which is seen in lymphomas with t(14;18 translocations),23, 24 or certain lymphoma-associated BCL2 mutations that result in increased affinity of BCL2 proteins for BAK,25 NOXA,26 or PUMA and BIM,27 are associated with increased resistance to apoptosis.

The death receptor pathway, the other major pathway for triggering apoptosis, involves sequential binding of cytokines such as CD95 ligand and tumor necrosis factor-α-related apoptosis inducing ligand to cell surface receptors, recruitment of the adaptor protein Fas-associated death domain to the intracellular domains of these receptors, activation of procaspase 8 by the oligomerized Fas-associated death domain, and subsequent caspase activation. Importantly, inactivating CD95 mutations have been identified in a subset of follicular lymphomas.28

HDAC inhibitors have been reported to induce apoptosis through multiple changes affecting both of these apoptotic pathways. In particular, HDAC inhibitors enhance expression of the TNFRSF10B gene encoding death receptor 5, which binds tumor necrosis factor-α-related apoptosis inducing ligand and activates the death receptor pathway.29, 30, 31, 32 In addition, HDAC inhibitors repress transcription of BCL2 and BCLXL33 as well as upregulate genes encoding the pro-apoptotic BCL2 family members BAX, BAK and BIM in various cell types,34, 35, 36 potentially activating the mitochondrial pathway. Additional studies have demonstrated that BIM expression can also be increased by posttranslational mechanisms37, 38 that involve inhibition RSK2-mediated BIM phosphorylation and subsequent phosphorylation-dependent degradation39, 40 after treatment with MEK or farnesyltransferase inhibitors.41, 42 A separate line of investigation has identified the heat shock protein 90 (HSP90) chaperone as a critical target of HDAC inhibitors in malignant T cells.43 How these various observations are mechanistically linked, and which of these changes accounts for HDAC inhibitor-induced killing of malignant B cells, has been unclear.

Here, we identify the cytotoxic pathway triggered by HDAC inhibitors in B-cell lymphoma and T-cell leukemia cells. In particular, we show for the first time that HDAC inhibitors act on HSP90 to downregulate RASGRP1, a lymphoid-specific guanine nucleotide exchange factor,44 as well as the RAS effector CRAF, leading to diminished mitogen-activated kinase signaling, increased BIM expression and apoptosis. These results not only identify cell type-specific signaling downstream of HSP90 inhibition that potentially contributes to selective cytotoxicity of HDAC inhibitors in malignant T cells, but also show that the same signaling is inhibited during HDAC inhibitor-induced killing of B-cell lymphoma cells.

Materials and methods

Materials

Commercial antibodies (Supplementary Table S1) were obtained as follows: phospho-Thr202/Tyr204-ERK1/2, ERK1/2, phospho-Ser217/Ser221-MEK1/2, MEK1/2, BAX, BIM, BID, MCL1, BCLXL, cleaved caspase 3 and PARP1 from Cell Signaling Technology (Beverly, MA, USA); HSP70 and NOXA from Enzo Life Sciences (Farmington, NY, USA); BAK from Millipore (Billerica, MA, USA); CRAF, PUMA and BCL2 (rabbit polyclonal) from Santa Cruz Biotechnology (Santa Cruz, CA, USA); BCL2 (murine monoclonal) from Dako (Carpenteria, CA, USA); acetylated tubulin and α-tubulin from Sigma-Aldrich (St Louis, MO, USA); acetylated lysine (AcK) antibody from Rockland (Limerick, PA, USA); cytochrome c from BD Biosciences (Franklin Lakes, NJ, USA); and cytochrome oxidase subunit IV (COX IV) from Thermo (Waltham, MA, USA). Anti-HSP90 (clone H9010) was from David Toft (Mayo Clinic, Rochester, MN, USA). Antibody to RASGRP1 was generated and characterized as described.42 Reagents were obtained from the following suppliers: DNA oligonucleotides from Integrated DNA Technologies (Coralville, IA, USA); 3-(4,5-dimethylthiazol-2-yl)-5(3-carboxymethonyphenol)-2-(4-sulfophenyl)-2H-tetrazolium (MTS) from Promega (Madison, WI, USA); romidepsin and 17-allylamino-17-demethoxygeldanamycin (17AAG) were from the Drug Synthesis Branch of the National Cancer Institute and Toronto Research Chemicals (Toronto, ON, Canada); trichostatin A (TSA), etoposide and phenazine methosulfate from Sigma-Aldrich; the broad spectrum caspase inhibitor Q-VD-OPh from SM Biochemicals (Anaheim, CA, USA); OSI-127, navitoclax, venetoclax and panobinostat from Chemietek (Indianapolis, IN, USA); and bortezomib from Enzo Life Sciences. Short oligonucleotides targeting BAX (nucleotides 271–289, GenBank accession number NM_138761), BAK (nucleotides 913–931, NM_001188), BIM (nucleotides 325–343, NM_138621), PUMA (nucleotides 784–802, NM_014417), BIK (nucleotides 65–83, NM_001197), NOXA (nucleotides 272–290, NM_021127) and RASGRP1 (nucleotides 1555–1573, AF_081195) were from Ambion (Austin, TX, USA). All other reagents were obtained as previously described.25

Cell culture

Cell lines obtained as described42, 45 were authenticated by short tandem repeat profiling and confirmed to be mycoplasma free. BCL2 mutation status as reported in the COSMIC database (http://www.sanger.ac.uk/genetics/CGPF/cosmic/) was confirmed by Sanger sequencing.27 Cell lines were propagated at densities of <1 × 106 cells/ml in RPMI 1640 medium containing 100 units/ml penicillin G, 100 μg/ml streptomycin, 2 mM glutamine and 15% (JB-6, I9.2, I2.1, JMR) or 10% (all other cell lines) heat-inactivated fetal bovine serum (medium A). Small interfering RNA (siRNA) was transfected by electroporation using a BTX 830 electroporator delivering a 10 msec pulse at 240 V (WSU) or 280 V (Jurkat cells) as described.42 Jurkat cells stably expressing MEK(DD), a constitutively active form of MEK1 (clones 10 and 12), (ref. 46) were generated by electroporating 1.5 × 107 parental cells with 40 μg plasmid,25 selecting for stable integrants in 5 μg/ml puromycin after 48 h, and cloning the resulting puromycin-resistant cells by limiting dilution.

Detection of apoptosis by flow microfluorimetry

DNA fragmentation and annexin V binding were assayed as described previously.25, 26, 45

Immunoblotting

After treatment with HDAC inhibitors in the presence of 5 μM Q-VD-OPh (to inhibit changes in cellular proteins due to caspase-mediated cleavages) as indicated in the figure legends, cells were washed in serum-free RPMI-HEPES and solubilized in 6 M guanidine hydrochloride containing 250 mM Tris-HCl (pH 8.5 at 21 °C), 10 mM EDTA, 1% (v/v) ß-mercaptoethanol, and 1 mM α-phenylmethylsulfonyl fluoride (freshly added from a 100 mM stock in anhydrous isopropanol). Following reaction with iodoacetamide, cell lysates were dialyzed sequentially into 4 M urea followed by 0.1% (v/v) SDS.47 After lyophilization, aliquots were resuspended in SDS sample buffer at 5 mg protein/ml (assayed by the bicinchoninic acid method), separated by SDS–polyacrylamide gel electrophoresis, transferred to nitrocellulose membranes and probed with various antibodies.48

Cell fractionation

After treatment with diluent or 20 nM romidepsin for 24 h in the presence of 5 μM Q-VD-OPh, cells were sedimented at 200 g for 10 min, washed twice with ice-cold phosphate-buffered saline, and lysed by swelling for 20 min at 4 °C in hypotonic buffer (25 mM HEPES, pH 7.5 at 4 °C, 5 mM MgCl2 and 1 mM EGTA supplemented immediately before use with 1 mM α-phenylmethylsulfonyl fluoride, 10 μg/ml leupeptin and 10 μg/ml pepstatin) followed by Dounce homogenization. After trypan blue staining confirmed that all cells were ruptured, samples were sedimented at 800 g to remove nuclei. The postnuclear supernatant was sedimented at 4000 g to isolate crude mitochondria, which were resuspended in hypotonic buffer with 300 mM sucrose and sedimented at 4000 g a second time. Fractions were prepared for SDS–polyacrylamide gel electrophoresis as described under ‘Immunoblotting.’

Quantitative reverse transcription-PCR

After total RNA was isolated from cells treated with diluent or romidepsin in the presence of 5 μM Q-VD-OPh, quantitative reverse transcriptase-PCR was performed in triplicate employing 50 ng RNA and TaqMan One-Step reverse transcriptase-PCR Master Mix (Applied Biosystems, Carlsbad, CA, USA). Using BIM (Hs00197982_m1) and PUMA (Hs00248075_m1) probe sets, PCR was performed on a ABI Prism 7900HT Real Time System using a program consisting of 48 °C for 30 min, 95 °C for 10 min, then 40 cycles of 95 °C for 15 s and 60 °C for 1 min. Data were analyzed using the following standard equations: ΔCt=Ct(sample)−Ct(endogenous control); ΔΔCt=ΔCt(sample)−ΔCt(untreated); and Fold Change=2-ΔΔCt, where Ct is the PCR cycle where the fluorescence exceeds the threshold.

Xenograft study

After approval by the Institutional Animal Care and Use Committee, SCID/Beige mice (4–5 weeks, Harlan) were implanted subcutaneously with 100 μl of a 1:1 slurry containing Matrigel (BD Bioscience) and 5 × 106 washed, log phase SuDHL-6 cells in the right flank as well as radiofrequency identification chips along the neck. Lymphoma volumes were calculated according to the formula ab2/2, where a is the diameter along the longest dimension and b is the orthogonal diameter. When xenografts reached an average volume of 100 mm3, mice were randomized for euthanasia 0, 24 or 48 h after treatment with a single intravenous dose of romidepsin at 1 mg/kg (dissolved in ethanol and diluted 1:39 in 5% dextrose).49

Samples of lymphoid malignancies

Clinical lymphoma samples from the University of Iowa/Mayo Lymphoma SPORE Biospecimens Core were studied with IRB approval. Mononuclear cells were isolated from excisional lymph node biopsies or splenectomy specimens by mechanical disruption over a wire mesh screen followed by Ficoll-Hypaque centrifugation. All patients provided informed consent for research use of specimens.

MTS assay

Aliquots containing ~2 × 104 cells in 120 μl medium A were incubated at 37 °C with varying concentrations of romidepsin or TSA for 5 days. As a measure of the relative amounts of surviving cells,50 plates were treated with MTS and phenazine methosulfate as instructed by the supplier and incubated for 2–6 h to obtain an absorbance of 0.5–1.0 at 490 nm in control samples.

Statistical analysis

Full dose-response curves were performed in cell lines at least three times independently, with examples presented and bar graphs showing summary data. Error bars represent mean±s.d. of three independent experiments and P-values represent results of unpaired t tests.

Results

Cross-resistance patterns of lymphoid cell lines to chemotherapeutic agents and BH3 mimetics

Building on earlier results showing broad cross-resistance of certain lymphoma lines to multiple agents,51 we examined sensitivity of a panel of 12 lymphoid cell lines (Supplementary Table S2) to a variety of agents. In view of recent results showing that certain BCL2 mutations enhance the ability of BCL2 to inhibit apoptosis,25, 26 we included five lines that harbor BCL2 mutations (identified with black symbols in Figure 1) and seven that do not (gray symbols). Baseline expression of BCL2 family members varied widely across these cell lines (Figure 1a) as discussed below.

Figure 1
figure 1

Drug sensitivity of human lymphoid cell lines. (a) Western blot showing protein levels for pro-apoptotic (BAX, BAK, BIM, BID and NOXA) and anti-apoptotic (BCL2, BCLXL and MCL1) BCL2 family members in human lymphoid cell lines with wild type (lanes 1–7, 13) or mutant BCL2 (lanes 8–12). Whole-cell lysates containing 50 μg of total protein were subjected to SDS–polyacrylamide gel electrophoresis and probed with the indicated antibodies. The polyclonal anti-BCL2 antiserum was employed to assess the possibility that the paucity of signal with monoclonal anti-BCL2 antibody in certain BCL2-mutant lines results from mutation of the recognized epitope. Gray and black bars on top indicate cells lines with wild type (WT) or mutant BCL2, respectively. Histone H3 was used as loading control. (b) Cells were treated with etoposide for 24 h, stained with propidium iodide and subjected to flow microfluorimetry as illustrated in Supplementary Figure S1a. Curves shown are representative of three independent experiments and the gray bar indicates the drug concentration used for the bar graph and box plot in c. Effect of etoposide (2 μM, 24 h) on the number of subdiploid cells. The Jurkat cell line, which is sensitive to all of these treatments, was used as positive control. Error bars indicate mean±s.d. of three independent experiments. *P<0.05.

Full size image

As indicated in Figures 1b and c and Supplementary Figure S1a, these lines exhibited a wide range of sensitivities to the topoisomerase II poison etoposide, with apoptosis ranging from 5 to 50% after treatment with 2 μM for 48 h. There was more apoptosis in cells with wild type BCL2 compared with mutant BCL2 (Figure 1c, right panel; P<0.05). Likewise, the cells exhibited varying sensitivities to ionizing radiation, the antimetabolite cytarabine, the proteasome inhibitor bortezomib, and the mTOR inhibitor OSI-027 (Supplementary Figures S1b–i). With ionizing radiation and cytarabine there was again diminished apoptosis in the BCL2 mutant cell lines (Supplementary Figures S1c and e, P<0.05), but this was not evident with bortezomib or OSI-027 (Supplementary Figures S1g and i).

A somewhat different pattern emerged when sensitivity to BH3 mimetics was examined. In particular, three lines showed striking hypersensitivity to the BCL2 selective BH3 mimetic venetoclax (Figures 2a and b). This sensitivity did not track with BCL2 mutations. Instead, three of the most sensitive lines, Jeko, DoHH2 and RL, expressed the highest levels of BCL2 protein (Figure 1a), Conversely, the majority of lines were resistant to venetoclax despite more modest BCL2 levels. Likewise, the majority of lines were relatively resistant to the BCL2/BCLXL targeted inhibitor navitoclax (Figure 2c and Supplementary Figure S2a), with Jeko, DoHH2 and RL still being the most sensitive. These observations are consistent with previous observations that lymphoid cells with elevated BCL2 express more BIM and are particularly sensitive to BH3 mimetic drugs that inhibit BCL2.52 In contrast, more of the lines responded to obatoclax (Supplementary Figures S2b and c), which reportedly inhibits anti-apoptotic BCL2 family members more broadly53 but also might have off-target effects.54

Figure 2
figure 2

Sensitivity of lymphoid cells to venetoclax and navitoclax. (a) Venetoclax dose–response profile for lymphoid cell lines. Jurkat cells were used as a control. Curves shown are representative of three independent experiments and the gray bar indicates the drug concentration used for the bar graph and box plot in b. Dose–response curves for navitoclax and obatoclax are found in Supplementary Figures S2a and S2b. (b–c) Induction of subdiploid cells at 48 h by treatment with venetoclax (500 nM) or navitoclax (500 nM), respectively. Error bars, mean±s.d. of three independent experiments. n.s., not significant.

Full size image

HDAC inhibitors are cytotoxic in a variety of neoplastic B-cell lines

Based on previous reports that HDAC inhibitors upregulate components of the death receptor pathway, including both ligands and receptors,9, 30, 31 as well as induce other changes that activate the death receptor pathway,55 along with the demonstration that small increases in death receptor expression are sufficient to render death receptor signaling BCL2-independent,56 we examined the effect of three different HDAC inhibitors on the same lymphoid cell lines. Strikingly, all of the lines, including lines that were resistant to conventional agents and BH3 mimetics (for example, WSU and Daudi, Figures 1b and c, Supplementary Figures S1 and S2), were sensitive to the prototypic HDAC inhibitor TSA (Supplementary Figure S3) as well as clinically utilized HDAC inhibitors such as romidepsin (Figures 3a and b) and panobinostat (Figure 3c and Supplementary Figure S3a). Accordingly, further studies focused on the mechanism by which these agents were able to induce cell death even in the face of resistance to other agents.

Figure 3
figure 3

Sensitivity of malignant B cell lines to HDAC inhibitors. (a) Romidepsin dose–response profile after treatment for 48 h. Curves shown are representative of three independent experiments and the gray bar indicates the drug concentration used for the graph and box plot in panel b. Additional dose–response curves for other agents are found in Supplementary Figures S3a and b. (b–c) Effects of romidepsin (10 nM, 72 h) and panobinostat (25 nM, 48 h) on lymphoid cell lines. Jurkat cells were used as a positive control. Error bars, mean±s.d. of three independent experiments. n.s., not significant.

Full size image

Initial mechanistic experiments examined Jurkat cells, a T-cell acute lymphoblastic leukemia line, as a prototypic malignant lymphoid line because of the availability of isogenic lines with disabled death receptor or mitochondrial pathway signaling.26 Jurkat variants that lack procaspase 8 (I9.2) or Fas-associated death domain (I2.1), two critical components of the death receptor pathway (Supplementary Figure S4a), were just as sensitive as parental Jurkat cells to all three HDAC inhibitors (Supplementary Figure S4b). In contrast, Jurkat cells lacking procaspase 9 (JMR) or massively overexpressing BCL2 (JB-6) were resistant, indicating a predominant role for the mitochondrial pathway in romidepsin-, TSA- and panobinostat-induced cell death (Supplementary Figures S4a and S4b).

In further experiments, parental Jurkat cells were treated for 24 h with increasing romidepsin concentrations in the presence of Q-VD-OPh to inhibit caspase-mediated cleavages downstream of mitochondrial pathway activation (Supplementary Figure S4c). Marked upregulation of PUMA and the three most abundant BIM isoforms was observed (Supplementary Figure S4d). Additional experiments demonstrated that the amount of BIM in mitochondria also increased (Supplementary Figure S4e). In contrast, the anti-apoptotic BCL2 family members MCL1, BCL2 and BCLXL were unchanged, as were the apoptotic effectors BAX and BAK (Supplementary Figure S4d). Similar results were observed in the B cell line Nalm6 (Supplementary Figure S4f).

To assess the contribution of the upregulated BH3-only proteins to killing, various proapoptotic BCL2 family members were knocked down with RNA oligonucleotides. Decreased BIM, like knockdown of the mitochondrial outer membrane permeabilizing protein BAK, markedly inhibited romidepsin-induced apoptosis (Supplementary Figures S4g and S5). In contrast, knockdown of other BH3-only proteins, including PUMA, BIK and NOXA, had little effect. Accordingly, further studies examined the cause of the BIM upregulation.

BIM in lymphoid cells is known to be sometimes upregulated by changes in transcription.57 Quantitative reverse transcription-PCR after treatment of Jurkat cells with romidepsin showed a small change in BIM mRNA (Supplementary Figure S6a) that appeared too limited to account entirely for the 5- to 10-fold BIM protein upregulation. Another mechanism of BIM upregulation involves decreased BIM degradation39, 40, 42 as a consequence of diminished signaling through the MAP kinase pathway.39, 41 When examined for potential involvement of this pathway, Jurkat cells exhibited a marked decrease in phosphorylation of ERK1/2 and their upstream kinases MEK1/2 during romidepsin treatment (Figure 4a). Moving further up this signaling pathway, additional analysis (Figure 4a) showed that this decreased signaling reflected HDAC inhibitor-induced diminution of the RAS effector kinase CRAF, which participates in MEK1/2 phosphorylation, and the RAS guanine nucleotide exchange factor RASGRP1, which participates in MAP kinase pathway activation in lymphoid cells.44 Additional investigation demonstrated that these effects reflected romidepsin-induced increases in HSP90 acetylation and accompanying decreases in binding of both CRAF and RASGRP1 to the HSP90 chaperone complex (Figure 4b).

Figure 4
figure 4

HDAC inhibitor diminishes MAP kinase signaling in T-cells. (a) Whole-cell lysates from Jurkat cells treated for 24 h with romidepsin (0, 5, 10, 20 nM) were subjected to SDS–polyacrylamide gel electrophoresis (PAGE) and probed with the indicated antibodies. HSP90 was used as loading control. Because antibodies against acetylated HSP90 gave a high background and recognized multiple bands of varying molecular weight, antibody to acetylated α-tubulin was employed as an indicator of HDAC inhibition in this and subsequent figures. (b) After Jurkat cells were treated with diluent or 20 nM romidepsin for 6 h (a time chosen to optimize HSP90 release of clients without their extensive degradation), immunoprecipitates prepared with anti-HSP90 antibody and subjected to SDS–PAGE and immunoblotting with the indicated antibodies. (c) Immunoblots of whole-cell lysates prepared from Jurkat cells 48 h transfection with control siRNA control or RASGRP1 siRNA. HSP90 served as a loading control.

Full size image

In further experiments, we examined the biochemical impact of RASGRP1 downregulation. After siRNA-mediated RASGRP1 knockdown, signaling through the MAP kinase pathway was markedly diminished and BIM was upregulated (Figure 4c), suggesting that RASGRP1 downregulation contributes to the biochemical changes that are critical for HDAC inhibitor-induced apoptosis. Conversely, Jurkat cells stably transfected with constitutively active MEK1 (MEKDD), which partially protected from ERK dephosphorylation, exhibited diminished BIM induction (Supplementary Figure S6b) and less killing during romidepsin treatment (Supplementary Figure S6c) compared with the cells transfected with empty vector, again indicating the importance of MAP kinase pathway inhibition in the cytotoxicity of HDAC inhibitors.

Romidepsin action in neoplastic B-cell lines

All of the critical HDAC inhibitor-induced changes in MAP kinase signaling and BCL2 family expression were also observed in malignant B cells. When SuDHL-4, SuDHL-6, and WSU cells were treated with romidepsin (Figures 5a–c) or TSA (Figure 5b) in the presence of Q-VD-OPh to again prevent secondary apoptotic effects, downregulation of RASGRP1 and CRAF, inhibition of CRAF→MEK→ERK signaling, and BIM upregulation were observed. This was accompanied by increased mitochondrial BIM accumulation and cytochrome c release (Figure 6a). BIM knockdown, like BAK knockdown, was again protective against cell death, as illustrated by results in WSU cells (Figure 6b). Importantly, similar effects, including RASGRP1 and CRAF downregulation, inhibition of MAP kinase pathway signaling, and BIM upregulation were observed in these lines after treatment with the HSP90 inhibitor 17AAG (Figure 6c), supporting the view that disruption of HSP90/client protein interactions is sufficient to account for these changes.

Figure 5
figure 5

BIM upregulation reflects inhibition of MAP kinase signaling in multiple malignant B-cell lines. (a–c) SuDHL-4 (a), SuDHL-6 (b) or WSU cells (c) were treated for 24 h with romidepsin (0, 5, 10, 20 nM, panels a–c) or TSA (0, 50, 100, 200 nM, panel b) in the presence of 5 μM Q-VD-OPh. Whole-cell lysates were subjected to SDS–polyacrylamide gel electrophoresis and probed with the indicated antibodies. Acetylated tubulin (Ac-Tubulin) served as marker of HDAC inhibition. HSP90 and/or α-tubulin served as loading controls.

Full size image
Figure 6
figure 6

BIM upregulation contributes to HDAC inhibitor-induced killing in malignant B-cell lines. (a) After WSU cells were treated with diluent (0.1% dimethyl sulfoxide, -) or 20 nM romidepsin (+) in the presence of 5 μM Q-VD-OPh for 24 h, the indicated subcellular fractions were isolated and subjected to immunoblotting. GAPDH served as a marker for cytosol and cytochrome c oxidase subunit IV (CoX IV) served as a marker for mitochondria. (b) Twenty-four hours after transfection of the indicated siRNAs along with plasmid encoding EGFP-histone H2B (to mark successfully transfected cells), WSU cells were treated for 24 h with diluent or 20 nM romidepsin before staining with APC-conjugated annexin V and analysis by 2-color flow cytometry. Error bars,±s.d. of three independent experiments. **P<0.01. Inset: Immunoblots of whole-cell lysates prepared from siRNA-treated cells incubated in drug-free medium in parallel with samples harvested for flow cytometry. HSP90 served as a loading control. (c) SuDHL-6 and WSU cells were treated for 48 h with diluent (lanes 1 and 5, respectively) or the HSP90 inhibitor 17AAG at 0.3, 1 or 3 μM (lanes 2–4 and 6–8, respectively) or diluent in the presence of 5 μM Q-VD-OPh. Whole-cell lysates were subjected to SDS–polyacrylamide gel electrophoresis and probed with the indicated antibodies. GAPDH served as a loading control.

Full size image

In further experiments, we examined HDAC inhibitor sensitivity of clinical lymphoma specimens (Table S3) ex vivo. This analysis showed that a wide range of B-cell lymphoma specimens responded to both romidepsin (Figure 7a) and TSA (Supplementary Figure S7) at concentrations similar to those that affected T-cell lymphoma. In particular, the IC50s for romidepsin and TSA were 1–10 nM and 30–700 nM, respectively.

Figure 7
figure 7

Effects of HDAC inhibitors on malignant B cells ex vivo and in vivo. (a) B-cell lymphoma specimens (solid lines) or T-cell lymphoma specimens (dashed lines) were treated with the indicated concentrations of romidepsin for 5 days ex vivo and assayed for relative cell survival using MTS assays. (b, c) Reverse transcriptase-PCR for BIM (b) and immunoblotting for the indicated antigens (c) in SuDHL-6 xenografts harvested from nine mice immediately before, 24 h and 48 h after the first dose of romidepsin (lanes 3–11). SuDHL-6 cells treated for 48 h with diluent (−) or 10 nM romidepsin (+) ex vivo in the presence of 5 μM Q-VD-OPh served as a positive control for romidepsin-induced signaling changes (lanes 1–2).

Full size image

To extend these results, the effects of romidepsin were assessed in SuDHL-6 xenografts in vivo. As was the case in vitro, romidepsin treatment of xenografts caused limited upregulation of BIM mRNA (Figure 7b), but readily detectable increases in BIM protein that again accompanied decreases in the HSP90 client proteins RASGRP1 and CRAF as well as diminished signaling through the MAP kinase pathway (Figure 7c). Thus, the results observed in tissue culture were readily observed 24–48 h after romidepsin treatment in vivo.

Discussion

Collectively, results of the present study suggest that HDAC inhibitors kill malignant B lymphocytes through a series of signaling events outlined in Figure 8. In particular, working upstream from the observation that BIM has a critical role in HDAC inhibitor-induced apoptosis, we have traced BIM upregulation to diminished MAP kinase pathway signaling, which in turn reflects HDAC inhibitor-induced downregulation of RASGRP1 and CRAF that follows release of these proteins upon acetylation of HSP90. Importantly, expression of constitutively active MEK (Supplementary Figure S6b), like downregulation of BIM or BAK (Figure 6b and Supplementary Figure S4g), diminishes HDAC inhibitor-induced apoptosis in malignant T and B cell lines, suggesting the importance of this pathway for HDAC inhibitor-induced killing.

Figure 8
figure 8

Summary of signaling elucidated in the present study. Under normal circumstances, RASGRP1 and CRAF are bound to the HSP90 chaperone complex, which stabilizes them and facilitates signaling through the MEK/ERK pathway to induce phosphorylation-dependent BIM degradation.37 Treatment with romidepsin shifts the equilibrium in favor of HSP90 acetylation, causing release and degradation of RASGRP1 and CRAF, inhibition of MEK/ERK signaling, elevation of BIM and apoptosis. Interventions performed in this paper to dissect the pathway (and the corresponding figures) are indicated.

Full size image

Although inhibition of HSP90(ref. 43) and upregulation of BIM36 have been previously observed during HDAC inhibitor treatment of susceptible T lymphocytes, steps tying these events together have been unclear. Results of the present study extend these prior studies in several important ways. First, the present results identify RASGRP1, a lymphoid cell-specific guanine nucleotide exchange factor that facilitates activation of RAS and its effectors,44 as a novel client protein that is released from HSP90 upon HDAC inhibitor treatment. The present studies not only show that HDAC inhibitors promote release of RASGRP1 from HSP90 at early time points before total cellular levels have decreased (Figure 4b), but also demonstrate that RASGRP1 levels decrease when cells are treated with the well established HSP90 inhibitor 17AAG (Figure 6c). These are hallmarks of an HSP90 client. Second, the present results show that signaling changes downstream of RASGRP1 and CRAF loss (that is, inhibition of MAP kinase signaling and consequent BIM upregulation) are observed in malignant B cells (Figures 5–7) as well as T cells (Supplementary Figure S4). Third, the present observations extend the potential use of HDAC inhibitors to malignant B cells that are resistant to other agents through a variety of mechanisms. Results observed in several cell lines are particularly informative in this regard.

The SuDHL-6 line has low levels of BIM relative to other lymphoid cell lines (Figure 1a) and, consistent with a previous study,51 is resistant to navitoclax (Figure 2) as well as etoposide and cytarabine (Figure 1 and Supplementary Figure S1). Interestingly, however, multiple HDAC inhibitors were able to upregulate BIM and induce apoptosis in this line (Figures 3 and 5b and Supplementary Figure S3), indicating that resistance due to low BIM levels can be pharmacologically reversed.

Likewise, the WSU line has low levels of BAX and BAK relative to other lymphoid lines (Figure 1a), although these proteins can be detected upon prolonged exposure (Figure 5c). Like SuDHL-6 cells, WSU cells are resistant to multiple conventional and experimental anti-lymphoma agents (Figures 1 and 2 and Supplementary Figure S1), but are able to upregulate BIM and undergo BIM-dependent apoptosis after HDAC inhibitor treatment (Figures 3, 5c, 6 and Supplementary Figure S3).

Cell lines such as Jeko, DoHH2 and RL, which express extremely high concentrations of BCL2 (Figure 1a), appeared dependent on this BCL2 overexpression for survival, as evidenced by their heightened sensitivity to the BCL2 antagonist venetoclax (Figure 2b) and diminished sensitivity to agents that activate the mitochondrial apoptotic pathway, including etoposide, bortezomib and OSI-027 (Supplementary Figure S1). Despite high BCL2 levels, these cells were sensitive to romidepsin (Figure 3b).

Consistent with recent studies demonstrating that some lymphoma-associated nonsynonymous nucleotide substitutions in the BCL2 gene enhance BCL2 anti-apoptotic function ex vivo,25, 26, 27 the lymphoid cell lines with BCL2 mutations were somewhat less sensitive to etoposide, ionizing radiation and cytarabine than lymphoma lines without BCL2 mutations (Figures 1b and c, Supplementary Figures S1c and S1e). Importantly, these BCL2-mutant lines were just as sensitive as BCL2 wild-type lines to multiple HDAC inhibitors, including romidepsin and panobinostat (Figure 3 and Supplementary Figure S3).

During HDAC inhibitor treatment of malignant lymphoid cells, BIM mRNA increased modestly, in agreement with previous studies.34, 35, 36 BIM protein, however, increased more than BIM message (Figure 7). Moreover, constitutively active MEKK blunted the increase in BIM protein (Supplementary Figure S6b). Accordingly, HSP90 inhibition and the resulting downregulation of mitogen-activated kinase signaling appear to contribute as least as much as transcriptional activation during HDAC inhibitor-induced killing of the lymphoid cell lines examined in the present study.

HDAC inhibitor-induced killing was observed in a wide range of malignant B cells (Figure 7a). Importantly, however, the IC50s varied over a 10-fold range, suggesting that some lymphomas might be more sensitive than others. Further investigation is required to assess the cause of this heterogeneity. It is possible, for example, that varied sensitivities reflect differences in dependence of lymphoma cells on the HSP90•CRAF and HSP90•RASGRP1 complexes for suppression of BIM levels. If so, pretreatment assessment of signaling through the mitogen-activated protein kinase pathway might help identify lymphomas that are more responsive to HDAC inhibitors.

In summary, we have observed that HDAC inhibitors such as romidepsin induce apoptosis in lymphoma cell lines relatively independent of their sensitivity to other agents. Although romidepsin is currently Food and Drug Administration approved for the treatment of cutaneous T-cell lymphoma,1, 2, 3 it has also been shown to enhance CD20 expression in B cells.58 Moreover, previous preclinical studies have demonstrated its activity in B-cell chronic lymphocytic leukemia55, 59 and B-cell lymphomas60 alone and in combination with rituximab.58 The present results provide a mechanistic explanation for these findings by identifying the lymphoid-specific RAS guanine nucleotide exchange factor RASGRP1 as a new HSP90 client that is released and degraded together with CRAF during treatment with romidepsin, providing at least a partial explanation for the prominent clinical activity of this class of agent in lymphoid malignancies, including both T- and B-cell neoplasms.