T-cell lymphoid malignancies (TCLM) are in need of novel and more effective therapies. The histone deacetylase (HDAC) inhibitor romidepsin and the synthetic cytotoxic retinoid fenretinide both have achieved durable clinical responses in T-cell lymphomas as single agents. We investigated the potential for using these two agents in combination in TCLMs. We demonstrated cytotoxic synergy between romidepsin and fenretinide in 15 TCLM cell lines at clinically achievable concentrations that lacked cytotoxicity for nonmalignant cells (fibroblasts and blood mononuclear cells). In vivo, romidepsin + fenretinide + ketoconazole (enhances fenretinide exposures by inhibiting fenretinide metabolism) showed greater activity in subcutaneous and disseminated TCLM xenograft models than single-agent romidepsin or fenretinide + ketoconazole. Fenretinide + romidepsin caused a reactive oxygen species (ROS)–dependent increase in proapoptotic proteins (Bim, tBid, Bax, and Bak), apoptosis, and inhibition of HDAC enzymatic activity, which achieved a synergistic increase in histone acetylation. The synergistic cytotoxicity, apoptosis, and histone acetylation of fenretinide + romidepsin were abrogated by antioxidants (vitamins C or E). Romidepsin + fenretinide activated p38 and JNK via ROS, and knockdown of p38 and JNK1 significantly decreased the synergistic cytotoxicity. Romidepsin + fenretinide also showed synergistic cytotoxicity for B-lymphoid malignancy cell lines, but did not increase ROS, acetylation of histones, activation of p38 + JNK, or cytotoxicity in nonmalignant cells. Romidepsin + fenretinide achieved synergistic activity in preclinical models of TCLMs, but not in nonmalignant cells, via a novel molecular mechanism. These data support conducting clinical trials of romidepsin + fenretinide in relapsed and refractory TCLMs. Mol Cancer Ther; 16(4); 649–61. ©2017 AACR.
T-cell lymphomas and leukemias (TCLM) are aggressive neoplasms that often have suboptimal clinical outcomes. The 5-year overall survival rate is <35% for peripheral T-cell lymphoma (PTCL) not otherwise specified (1), ∼24% for disseminated cutaneous T-cell lymphoma (CTCL; Sézary syndrome; ref. 2) and for T-cell acute lymphoblastic leukemia (T-ALL) in adults is ∼ 50% (3). Though the T-ALL survival rate in children is about 80%, 15% to 20% of T-ALL children will relapse, and the long-term event-free survival rate for early relapse is only 15% to 25% (4). Moreover, during or after intensive treatment, many T-ALL patients develop serious acute and chronic therapy–related complications (5). Although TCLMs comprise a diverse group of different clinical entities, they all share common biological features (6), suggesting the potential for identifying drugs or drug combinations that will be active across the spectrum of TCLMs by attacking biological mechanisms shared among TCLMs.
Histone deacetylases (HDAC) remove acetyl groups on histones, leading to a closed chromatin configuration and transcriptional repression (7). Analysis of both cell lines and primary tissue samples showed overexpression of HDACs in lymphoid malignancies (8). Several structurally unique HDAC inhibitors have been developed (7), and in addition to targeting histones, they also target several nonhistone effector molecules (9). HDAC inhibitors vorinostat, romidepsin (ROM), and belinostat have shown clinical activity as single agents against TCLMs and are FDA approved as a second-line therapy for the treatment of CTCL and/or PTCL (9). As TCLM patients often develop progressive disease during or after HDAC inhibitor therapy (10, 11), drug combinations that may overcome resistance to HDAC inhibitors are needed.
Fenretinide (4-HPR) is a synthetic retinoid that induces cytotoxicity independent of the retinoid receptor pathway through increases of reactive oxygen species (ROS; refs. 12, 13) and dihydroceramides (14, 15). ROS generated by fenretinide has been reported to trigger cell death via activation of the stress-activated MAP kinases p38 and JNK (12, 13). Preclinical studies have shown that fenretinide is cytotoxic to a variety of cancer cell lines, including leukemias and lymphomas (12, 16). The initial formulation of fenretinide was a corn-oil capsule that had low bioavailability (1–3 μmol/L) and limited clinical activity (17, 18). An oral powder fenretinide formulation (LXS fenretinide; ∼10 μmol/L) uses a lipid matrix (Lym-X-Sorb) to increase bioavailability via gut absorption to the lymphatic system (similar to a chylomicron) improved fenretinide bioavailability in mice (19) and patients (20). Further improvement in fenretinide exposures (∼30 μmol/L) in mice and children was achieved by inhibiting fenretinide metabolism with coadministration of the CYP3A4 inhibitor ketoconazole (KETO) with oral LXS fenretinide (21, 22). Even higher (and tolerable) fenretinide exposures (∼40–50 μmol/L) have been achieved by continuous infusion of an intravenous emulsion fenretinide formulation (23). In a phase I trial of intravenous emulsion fenretinide for hematologic malignancies, 4 out of 11 (36%) patients with TCLM showed complete or partial responses (3 of these responding patients had failed prior romidepsin treatment), and 5 patients showed stable disease (23). Fenretinide intravenous emulsion is currently being evaluated in an ongoing phase IIa clinical trial for relapsed/refractory PTCL patients who have failed prior systemic therapies (NCT02495415).
Although romidepsin and fenretinide are two different classes of chemotherapeutic agents with different known mechanisms of action, both have achieved durable responses as single agents in clinical trials of PTCL and CTCL. Moreover, TCLM patients who failed prior romidepsin therapy have responded to fenretinide. Therefore, we investigated the potential for using these two agents in combination in TCLMs.
Materials and Methods
Chemicals and drugs
Fenretinide (4-HPR; for in vitro) and Lym-X-Sorb (LXS) fenretinide oral powder (for in vivo) were obtained from the National Cancer Institute (Bethesda, MD), romidepsin (ROM) was obtained from Celgene, and ketoconazole from TEVA Pharmaceuticals. Ascorbic acid (vitamin C, VIT C), α-tocopherol (vitamin E, VIT E), n-acetylcysteine (NAC), sodium thiosulfate (STS), p38 inhibitor PD169316 and 2′,7′-dichlorodihydrofluorescein diacetate (DCFDA) were from Sigma Aldrich, and the JNK inhibitor SP600125 was from Tocris Biosciences. DMSO was obtained from WAK-Chemie Medical GmbHand (Siemensstr, Germany), and ethanol from EMD Millipore.
Cell culture and PDX
We assembled a panel of 23 cell lines (Supplementary Table S1), which included 15 T-cell leukemia/lymphoma cell lines (CCRF-CEM, COG-LL-317h, COG-LL-329h, COG-LL-332h, COG-LL-357h, DERL-2, HH, HUT-78 JURKAT, KARPAS-299, MOLT-3, MOLT-4, My-La-CD4+, SUP-T1, and TX-LY-172) and 8 B-lineage lymphoid malignancy cell lines [COG-LL-319h, TX-LY-245h, KMS-12-PE, NALM-6, OPM-2, RAMOS(RA1), RS4;11, and Toledo]. The cell lines were obtained from Children's Oncology Group Cell Culture and Xenograft Repository (www.COGcell.org; COG-LL-317h, COG-LL-329h, COG-LL-332h, COG-LL-357h, and COG-LL-319h), Texas Cancer Cell Repository (www.txccr.org; TX-LY-172 and TX-LY-245h), American Type Culture Collection [www.atcc.org; CCRF-CEM, HH, HUT-78, JURKAT, MOLT-3, MOLT-4, SUP-T1, RAMOS(RA1), RS4;11, and Toledo], European Collection of Cell Cultures (www.phe-culturecollections.org.uk; My-La-CD4+), and Leibniz Institute DSMZ-German Collection of Microorganisms and Cell Cultures (www.dsmz.de; DERL-2, KARPAS-299, KMS-12-PE, NALM-6, and OPM-2). Cell lines were cultured in a complete medium of Iscove's Modified Dulbecco's Medium (Thermo Scientific) supplemented with 20% FBS (Gibco – Life Technologies), 4 mmol/L l-glutamine (Corning Cellgro), insulin and transferrin (20 μg/mL each), and selenous acid (20 ng/mL; ITS Culture Supplement; BD Biosciences). Cell lines and patient-derived xenograft (PDX; TX-LY-183x, obtained from Texas Cancer Cell Repository; Supplementary Table S1) identities were verified via short-tandem repeat (STR) profiling (ref. 24; validated against a searchable online database at www.COGcell.org.), before starting and after finishing the experiments, and were confirmed to be free of mycoplasma (25). Cells were cultured and treated in a 37°C humidified incubator gassed with 5% CO2 and 90% N2, to achieve bone marrow level hypoxia of 5% O2 (26).
DIMSCAN cytotoxicity assay
To replicate romidepsin exposures in patients (27), we used romidepsin washout dosing for the in vitro cytotoxicity assays. Cells were incubated with romidepsin (0–10 nmol/L; dissolved in DMSO) for 4 hours, subsequently washed and fresh complete medium was added with fenretinide (0–10 μmol/L; dissolved in 100% ethanol) for 96 hours in a fixed ratio of concentrations (28). In the antioxidant-treated plates (vitamin C was dissolved in sterile water, and vitamin E was dissolved in 100% ethanol), cells were pretreated for 1 hour with antioxidants before the addition of fenretinide and/or romidepsin. Controls were treated with the appropriate drug vehicles (final DMSO or ethanol content was ∼50% of the highest concentration of romidepsin and fenretinide tested). DIMSCAN assay was performed as previously described (29–31).
Animal protocols were approved by the TTUHSC Laboratory Animal Resources Center and Institutional Animal Care and Use Committee. Six- to 8-week-old athymic (nu/nu) mice (The Jackson Laboratory) were injected subcutaneously with 10 million cells of COG-LL-317h (cell line) or TX-LY-183x (PDX). Tumor size and mouse weight were measured twice weekly; tumor volumes were calculated as 0.5 * height * width * length (32). Event-free survival (EFS) = Time from initiating treatment until the end point (tumor volume ≥ 1,500 mm3; ref. 33). In case of weight loss or dehydration, mice were given 0.5 mL of 5% sterile glucose water. Mice with progressively growing tumors between 100 and 300 mm3 were randomized into four groups (5 mice per group). Fenretinide-LXS powder formulation was administered by gavage (180 mg/kg/day, slurried in water, two divided doses daily, five days per week), ketoconazole (38 mg/kg/day, dissolved in water, administered by gavage for five days per week) and romidepsin (1.25 mg/kg, dissolved in 1-N-2-methylpyrrolidone, diluted in 5% glucose to a final concentration of pyrrolidone to 2.5%) was administered intravenously (on days 1, 4, 8, 11, and 15, in a 21-day cycle, for 2 cycles). Controls were treated with the appropriate drug vehicles. The study was terminated after 80 days of initiating treatment, and all surviving mice were sacrificed.
Bioluminescent mouse xenografts
Two million cells of COG-LL-317h Luc expressing the luciferase gene transduced with a CD80 reporter gene (>95% CD80 positivity, assessed by flow cytometry) were injected intravenously into 6- to 8-week-old female NOD scidgamma (NSG) mice (Charles River Laboratories; ref. 34). Initial imaging of mice was done with the IVIS Lumina XR (Perkin-Elmer), after injecting d-luciferin-K+ (Perkin-Elmer, 150 mg/kg). Mice were randomized into groups of 4 mice and were treated for 5 days with LXS fenretinide (180 mg/kg/day, two divided doses) + ketoconazole (38 mg/kg/day; both by gavage), and intravenous romidepsin (1.25 mg/kg, on days 1 and 4). Controls were treated with the appropriate drug vehicles. Mice were reimaged posttreatment on day 6. Luminescence was compared with initial imaging to determine leukemia progression for each mouse in each treatment group. Tumor burden was represented as mean total flux (p/sec) ± SEM and was calculated using Living Image Software version 4.3.1 (Perkin-Elmer).
Analysis of TUNEL, ROS, and immunoblotting
Cells were incubated with romidepsin (10 nmol/L for COG-LL-317h and fibroblasts, and 5 nmol/L for TX-LY-172h) for 4 hours, then washed, and fresh complete medium was added with fenretinide (5 μmol/L for COG-LL-317h and 10 μmol/L for TX-LY-172h and fibroblasts). In the antioxidant treated plates, cells were pretreated for 1 hour with antioxidants before the addition of fenretinide and/or romidepsin. Controls were treated with the appropriate drug vehicles (final DMSO or ethanol content was ∼ 50% of the highest concentration of romidepsin and fenretinide tested). Measuring ROS (with 10 μmol/L DCFDA) and apoptotic DNA fragmentation (TUNEL; APO-DIRECT kit, BD Biosciences) was carried out by flow cytometry as previously described (12, 25). Immunoblotting with specific antibodies (Cell Signaling Technology; Supplementary Table S2) was performed as previously described (12, 25).
Analysis of HDAC enzyme activity and histone acetylation
Cells were incubated with romidepsin (5 nmol/L for COG-LL-317h and TX-LY-172h, and 10 nmol/L for fibroblasts) for 4 hours, washed, and fresh complete medium was added with fenretinide (10 μmol/L) for 24 hours and immunoblotting was performed to investigate for acetyl-histone (AH) levels. In the antioxidant treated plates, cells were pretreated for 1 hour with antioxidants before addition of fenretinide and/or romidepsin. Controls were treated with the appropriate drug vehicles (final DMSO or ethanol content was ∼50% of the highest concentration of romidepsin and fenretinide tested). For measuring HDAC enzymatic activity, the nuclear fraction of the cells was extracted using NE-PERTM kit (Thermo Fisher Scientific), 40 μg of nuclear extract were loaded per well, and enzymatic activity was determined with the fluorometric HDAC activity kit from Abcam using SpectraMax M2e (Molecular Devices; ref. 35).
After confirming the leukemia burden in the disseminated bioluminescent model COG-LL-317m Luc at day 6, mice showing the highest tumor burden in each group were treated with the respective treatment on day 8 (romidepsin (1.25 mg/kg), ketoconazole (38 mg/kg/day), and LXS fenretinide (180 mg/kg/day, two divided doses). Eight hours after treatment (reported maximal time for AH in vivo; ref. 36), mice were sacrificed, and enlarged mouse spleens (due to high leukemia content) were removed and analyzed for AH by immunoblotting.
Stable shRNA knockdown experiments
Plasmids carrying shRNA targeting p38α (MAPK14, TRCN0000000509), JNK1 (MAPK8, TRCN0000001055), in the pLKO.1 background were obtained from GE Dharmacon, and nontarget control (eGFP) shRNA from Sigma Aldrich. Lentiviral particles were packaged in 293FT cells (Invitrogen) using pLKO.1 shRNA along with MISSION Lentiviral Packaging Mix (Sigma Aldrich), then transduced into JURKAT, COG-LL-317h, and TX-LY-172h cell lines and further selected with puromycin.
For apoptosis, HDAC enzyme activity, and acetyl-histones, the significance of differences in means was determined by one-way analysis of variance using the Tukey posttest. Significance of in vivo leukemia burden, and ROM + 4-HPR vs ROM + 4-HPR + antioxidants for the DIMSCAN/ROS/HDAC/immunoblotting/apoptosis assays were performed using the Student t test. Combination index (CIN) to assess synergy was calculated with CalcuSyn (Biosoft; refs. 28, 37). Mouse EFS was graphically represented by Kaplan–Meier analysis, and survival curves were compared by the log-rank test. Tests were considered significant at P < 0.05. The data were plotted and analyzed using GraphPad Prism, SigmaPlot, and FlowJo.
Romidepsin + fenretinide synergistically induced cytotoxicity in TCLM cell lines
We determined the cytotoxicity of clinically achievable levels of romidepsin (ROM; 0–10 nmol/L; ref. 27) and fenretinide (4-HPR; 0–10 μmol/L; ref. 23) in 15 TCLM cell lines (Supplementary Table S1) using the DIMSCAN cytotoxicity assay (Fig. 1). Fenretinide as a single agent was highly active against four cell lines inducing ≥2 logs (99%) of cell kill at the maximum concentration (10 μmol/L). Romidepsin as a single agent achieved more modest cytotoxicity in most cell lines, inducing ≥2 logs (99%) of cell kill in only one cell line. Romidepsin + fenretinide at the highest concentration tested achieved synergistic cytotoxicity (CIN < 1) in all the cell lines tested with ≥ 2 logs (99%) of cell kill in 7 of the 15 cell lines (Fig. 1).
Romidepsin + fenretinide + ketoconazole was active against TCLM mouse xenografts
We evaluated the activity of fenretinide combined with romidepsin in vivo against murine subcutaneous xenograft models COG-LL-317m and TX-LY-183x (PDX established from the same patient as the TX-LY-172h cell line). As administration of fenretinide via continuous infusion (which achieves ∼40 μmol/L plasma levels in patients; ref. 23) is not feasible in mice, we used LXS fenretinide oral powder (19) given together with ketoconazole (KETO); KETO increases fenretinide plasma levels in mice from ∼10 μmol/L to ∼20 μmol/L by inhibiting fenretinide metabolism (21). Ketoconazole has been reported to enhance mean plasma concentrations of romidepsin (38), but ketoconazole had no significant effect on the activity of romidepsin against the TX-LY-183x PDX at the tested concentrations (Supplementary Fig. S1A). The EFS of romidepsin + fenretinide + ketoconazole was >2-fold higher than seen with romidepsin + ketoconazole or fenretinide + ketoconazole (P < 0.01) and >3-fold higher compared with controls for both xenograft models (Fig. 2A and B). The combination of romidepsin + fenretinide + ketoconazole was well tolerated as measured by mouse weights over time (Supplementary Fig. S1B and S1C).
We also used a disseminated disease bioluminescent model, the COG-LL-317m Luc (T-ALL xenograft expressing luciferase; ref. 34). Compared with controls, romidepsin + ketoconazole or fenretinide + ketoconazole did not result in a significant reduction in the leukemia burden (P > 0.05). However, romidepsin + fenretinide + ketoconazole significantly (P < 0.05) reduced leukemia burden in luciferase-labeled COG-LL-317m Luc compared with other groups (Figs. 2C and D).
The cytotoxic synergy of romidepsin + fenretinide is mediated by ROS
Generation of ROS is one mechanism of fenretinide cytotoxicity. Increased ROS in response to fenretinide occurred as early as 30 minutes, reaching a maximum at 6 hours in COG-LL-317h and 3 hours in TX-LY-172h (Supplementary Fig. S2A). As shown in Fig. 3A, romidepsin did not induce ROS, and ROS in cells treated with fenretinide + romidepsin was no different than in cells treated only with fenretinide. Vitamins C or E (VIT C or E) abrogated ROS (P < 0.0001) in fenretinide and fenretinide + romidepsin treated cells. Vitamins C or E did not affect romidepsin-induced cytotoxicity, but did significantly decrease both single-agent fenretinide cytotoxicity and the multi-log cytotoxicity and synergy observed with romidepsin + fenretinide (P < 0.0001; CIN > 1). Romidepsin + fenretinide + vitamin C or E induced cytotoxicity was similar to the cells treated with romidepsin only (Fig. 3B).
By contrast, neither ROS nor the cytotoxicity from fenretinide or fenretinide + romidepsin was decreased by the thiol antioxidants sodium thiosulfate (STS) or n-acetylcysteine (NAC; P > 0.05; Supplementary Fig. S2B and S2C).
Romidepsin + fenretinide (mediated by ROS) induced proapoptotic Bcl-2 proteins and apoptosis
A loss of Bcl-2 family proapoptotic protein expression has been previously reported to be involved in romidepsin resistance (39, 40), and fenretinide has been shown to activate Bcl-2 family proapoptotic proteins in T-ALL cell lines (12). In TCLM cell lines, romidepsin, fenretinide, or romidepsin + fenretinide did not alter protein expression of the antiapoptotic proteins Bcl-2, Bcl-xl, Bcl-w, and Mcl-1 (Supplementary Fig. S3).
However, levels of the proapoptotic Bcl-2 family proteins Bim, tBid, Bax, and Bak were increased in response to fenretinide and to a significantly higher extent (P < 0.05) in romidepsin + fenretinide, compared with fenretinide alone (Fig. 4A and B). Consistent with the latter observation, the percentage of apoptotic cells was significantly greater (P < 0.01) with romidepsin + fenretinide compared with fenretinide or romidepsin as single agents (Fig. 4C). The antioxidant vitamin C inhibited the increase in proapoptotic Bcl-2 family proteins and the apoptosis (P < 0.0001) induced by romidepsin + fenretinide (Fig. 4).
ROS mediated enhanced inhibition of HDAC enzymatic activity
Because romidepsin is predominantly a class I HDAC inhibitor (41), we evaluated the effect of romidepsin, fenretinide, and romidepsin + fenretinide on HDAC enzymatic activity in nuclear fractions of cells, where the class 1 HDACs are localized (42). As expected, romidepsin decreased the HDAC enzymatic activity compared with controls (P < 0.05) and (surprisingly) this was also seen with fenretinide (P < 0.05). Combining fenretinide with romidepsin caused a significantly greater inhibition (P < 0.001) of HDAC enzymatic activity compared with romidepsin or fenretinide alone in the TCLM cell lines (Fig. 5A). The antioxidant vitamin C prevented the significant inhibition of HDAC enzymatic activity (P > 0.05) by fenretinide compared with control and romidepsin + fenretinide compared with romidepsin alone. HDAC enzymatic activity with fenretinide + romidepsin was significantly lower compared with fenretinide + romidepsin + vitamin C (P < 0.01) in TCLM cell lines (Fig. 5A).
We did not observe any changes in protein expression of class 1 histone deacetylases HDAC 1, 2, and 3 with romidepsin + fenretinide or with the single agents compared with controls (Supplementary Fig. S4).
HDACs regulate gene expression by altering chromatin structure (43), and both preclinical and clinical studies have shown that romidepsin increased histone acetylation (11, 36, 44). Acetyl-histone 3 and 4 (AH3 and 4) expression were increased relative to controls by romidepsin, but not by fenretinide, while romidepsin + fenretinide significantly increased (>2-fold, P < 0.01) AH3 and AH4 levels relative to romidepsin alone (Figs. 5B and C). The increased AH3 and AH4 expression caused by romidepsin + fenretinide relative to romidepsin alone was prevented by vitamin C (P > 0.05; Figs. 5B and C).
In vivo, romidepsin + fenretinide + ketoconazole increased AH3 and AH4 expression when compared with romidepsin + ketoconazole in the COG-LL-317m Luc xenograft (Supplementary Fig. S5).
ROS-dependent activation of p38 and JNK by romidepsin + fenretinide
The ROS generated by fenretinide was previously shown to trigger cell death by activating p38 and JNK (12, 13). As increased histone acetylation correlates with increased transcription of genes (43, 45), we determined if romidepsin + fenretinide increased p38 and JNK activation relative to fenretinide alone. As shown in Fig. 6A and B, romidepsin + fenretinide increased phospho-MKK4 (p-MKK4), phospho-JNK (p-JNK; P < 0.05), phospho-p38 (p-p38; P < 0.05), and phoshpho-ATF-2 (p-ATF-2, the downstream target of p38 and JNK), compared with fenretinide alone. Romidepsin + fenretinide also increased histone acetylation (AH3 + AH4) compared with romidepsin alone. Vitamin C abrogated the increased activation of p38 and JNK along with the increase in acetylation of histones seen with romidepsin + fenretinide.
ERK-MAPK pathway activation was reported to be involved in romidepsin resistance (40). Fenretinide and romidepsin + fenretinide did not increase p-ERK expression in TCLM cell lines (Supplementary Fig. S6).
JNK1 and p38 knockdowns decreased the cytotoxicity induced by romidepsin + fenretinide
To determine whether p38 and JNK are intermediates responsible for the synergistic cytotoxicity of romidepsin + fenretinide, we used shRNA to knockdown (KD) p38 and JNK1 protein expression in TCLM cell lines (Supplementary Fig. S7A). Because p38 and JNK were simultaneously activated by fenretinide-induced ROS, p38 KD cells were treated with fenretinide, romidepsin, and romidepsin + fenretinide in combination with the JNK inhibitor SP600125 (JNKi). Similarly, JNK1 KD cells were combined with the p38 inhibitor PD169316 (p38i). We observed significantly less cytotoxicity (P < 0.0001) in the p38 KD + JNKi or JNK1 KD + p38i treated with romidepsin + fenretinide, compared with their respective eGFP-transduced controls (Supplementary Fig. S7B). The percentage of apoptotic cells treated with romidepsin + fenretinide was significantly less (P < 0.01) in COG-LL-317h p38 KD + JNKi or JNK1 KD + p38i compared with COG-LL-317h eGFP (Supplementary Fig. S7C).
In contrast to the effect of p38 and JNK1 KD on cytotoxicity and apoptosis of 4-HPR + ROM, ROS and acetyl-histone levels in TCLM cells treated with romidepsin + fenretinide were not significantly altered in p38 and JNK1 KDs (Supplementary Figs. S8A–S8C), which indicate that p38 and JNK are downstream effectors of ROS.
Romidepsin + fenretinide did not affect normal mononuclear blood cells or fibroblasts
Romidepsin, fenretinide, or romidepsin + fenretinide were minimally cytotoxic (<1 log of cell kill and CIN > 1) for normal human fibroblasts and peripheral blood mononuclear cells (PBMC). In fibroblasts, romidepsin + fenretinide did not increase ROS, p-p38, or p-JNK levels, or acetylation of histones compared with romidepsin (P > 0.05; Supplementary Fig. S9).
Romidepsin + fenretinide synergistically induced cytotoxicity in B-lineage lymphoid malignancy cell lines
HDAC inhibitors have shown responses in relapsed B-cell lymphoma patients (46), and the HDAC inhibitor panobinostat is approved for treatment of patients with relapsed multiple myeloma (47). Intravenous emulsion fenretinide has achieved partial responses in relapsed B-cell lymphoma patients (23). Therefore, we also evaluated romidepsin + fenretinide in a panel of multiple myeloma and B-cell lymphoma/leukemia cell lines (Supplementary Table S1). Romidepsin + fenretinide achieved synergistic cytotoxicity (CIN < 1) in all the cell lines tested and induced ≥ 2 logs (99%) of cell kill in four out of eight cell lines at the highest concentrations tested (Supplementary Fig. S10).
Romidepsin, a HDAC inhibitor, and fenretinide, a cytotoxic retinoid for which one mechanism of cytotoxicity is via induction of ROS in cancer cells, both have shown clinical activity against T-cell lymphomas (10, 11, 23). We found that romidepsin + fenretinide were synergistically cytotoxic for all tested T-cell lymphoma and T-cell ALL cell lines and xenografts. The combination of romidepsin + fenretinide was also synergistically cytotoxic for lymphoma, leukemia, and multiple myeloma cell lines (all of B-cell origin). In a ROS-mediated fashion fenretinide + romidepsin synergistically inhibited HDAC enzymatic activity, increased histone acetylation, proapoptotic Bcl-2 family of proteins, and activation of p38 and JNK, which mediated synergistic apoptosis and cytotoxicity in TCLM cell lines.
Ketoconazole is necessary to increase 4-HPR concentrations in vivo (by inhibiting 4-HPR metabolism) but is not necessary for in vitro cytotoxicity studies where 4-HPR concentrations are easily controlled. To exclude any direct effects of ketoconazole on the synergy of romidepsin + fenretinide we determined the cytotoxicity of ketoconazole, ROM + ketoconazole, 4-HPR + ketoconazole, and ROM + 4-HPR + ketoconazole in two TCLM cell lines with high sensitivity (COG-LL-317h) and relatively low sensitivity (KARPAS-299) to ROM + 4-HPR. Single-agent ketoconazole did not induce any significant cytotoxicity compared with controls (we observed < 50% cell killing), nor did ketoconazole enhance the cytotoxicity of ROM, 4-HPR, or ROM + 4-HPR compared with cells treated without addition of ketoconazole (Supplementary Fig. S11).
In the phase I clinical trial of fenretinide in hematologic malignancies, durable clinical responses were observed in T-cell lymphomas when the fenretinide plasma levels were ∼30 to 50 μmol/L (which were well tolerated; ref. 23), while (due to dosing limitations with mice) the maximum steady state plasma concentrations of fenretinide in our xenograft experiments is ∼20 μmol/L (21). In CTCL and PTCL clinical trials, romidepsin is given for 6 cycles (10, 48). In vivo, we administered romidepsin for 2 cycles, unless tumor volume exceeded the endpoint (1,500 mm3). Thus, the range of concentrations used for in vitro studies and the exposure in mice for both romidepsin and fenretinide were less than the exposures achievable at maximum tolerated doses for humans. The romidepsin + fenretinide combination was not cytotoxic for fibroblasts or blood mononuclear cells, was well tolerated by mice, and as the drugs do not have overlapping systemic toxicities in clinical trials (23, 27), it is likely that romidepsin + fenretinide combination would be well tolerated in patients.
Lethal hepatotoxicity was reported in a patient concurrently receiving intravenous fenretinide and ceftriaxone (49). Excluding coadministration of ceftriaxone, which causes biliary sludging in some patients, enables intravenous fenretinide to be safely administered without hepatotoxicity.
Various organelles within the cell can generate ROS. These include mitochondria, the endoplasmic reticulum (ER) and peroxisomes. In addition, various enzymes including oxidases and oxygenases generate ROS (50). The antioxidants vitamin C or E (but not the thiol antioxidants NAC and STS) significantly diminished the ROS and cytotoxicity generated by fenretinide and also romidepsin + fenretinide. Similar results were observed in neuroblastoma preclinical models between fenretinide and the microtubule inhibitor ABT-751 (51). Vitamins C and E are two naturally occurring antioxidants, capable of entering mitochondria, ER, and other compartments (52, 53), while the thiol antioxidants STS and NAC aid in the replenishment of glutathione in the cytoplasm (25, 51, 54, 55), and may be excluded from cellular compartments involved in the generation and/or action of fenretinide-induced ROS, likely mitochondria and/or ER. HDAC inhibitors were shown to synergize with tyrosine kinase and proteasome inhibitors via ROS-mediated activation of MAP kinase pathways, but in those experiments addition of NAC abrogated the ROS, MAPK activation, and synergistic cytotoxicity (56, 57). Our observation that thiol antioxidants did not prevent 4-HPR induced ROS from synergistically enhancing ROM activity suggests that the mechanism we determined is distinct from that reported for HDAC inhibitors mediated synergy with other agents via ROS and MAP kinase pathway activation.
HDACs repress various genes that play major roles in tumor suppression, cell growth, proliferation, differentiation, cell death, and angiogenesis (43). Acetylation neutralizes the positive charge of lysine residues, thus increasing the accessibility of DNA to the transcription machinery (43). Genome-wide analysis of histone modification patterns has shown a correlation between histone acetylation and transcriptional activity (45), and cell lines treated with romidepsin undergo histone acetylation, which is necessary but not sufficient for cell death (58). We observed an increase in histone acetylation, but minimal cytotoxicity in fibroblasts treated with romidepsin. In fibroblasts romidepsin + fenretinide was not cytotoxic, did not increase ROS, acetylation (compared with romidepsin alone), or activate p38 and JNK. However, in TCLM cell lines the increased ROS from romidepsin + fenretinide enhanced AH3 and AH4 expression by inhibiting the HDAC enzymatic activity, leading to increased activation of p38 and JNK, increased apoptosis, and synergistic cytotoxicity. These data, together with demonstration that knockdown of p38 and JNK decreased the apoptosis and cytotoxicity of romidepsin + fenretinide, indicate that fenretinide and romidepsin interact to achieve synergistic cytotoxicity via ROS-mediated inhibition of HDAC activity that leads to p38 and JNK activation specifically in malignant cells. Our mechanistic observations are summarized in Fig. 6C.
Acetylation can occur on many lysine residues of histones (59). For this study, we investigated for AH3 and AH4 acetylation of lysine 9 and 8 residues, respectively. Though we observed that 4-HPR decreased HDAC enzymatic activity, we did not see an increase in AH3 and AH4 compared with controls at the probed lysine sites. This may imply 4-HPR is altering acetylation of histones at other lysine residues or alternatively that the decrease in HDAC enzymatic activity might be secondary to ROS-induced cytotoxicity.
Though knockdown of p38 and JNK1 significantly decreased the cytotoxicity of romidepsin + fenretinide compared to eGFP controls in TCLM cell lines, the romidepsin + fenretinide combination still exhibited multi-log cytotoxicity. This may be due to limited knockdown efficiency and might also imply other mechanisms of synergy. One possibility would be acting on the ERK–MAPK pathway, as activation of that pathway was reported to be involved in romidepsin resistance (40). However, we observed that romidepsin + fenretinide did not activate the ERK–MAPK pathway. Loss of proapoptotic proteins has also been reported to be involved in romidepsin resistance (39). Combining romidepsin with fenretinide significantly increased proapoptotic proteins compared with romidepsin or fenretinide as single agents, suggesting that a fenretinide-mediated increase in proapoptotic proteins represents another potential mechanism of synergy.
It is noteworthy that the TX-LY-172h cell line and TX-LY-183x PDX were established from a patient treated with intravenous emulsion fenretinide that achieved only a minimal subjective response. Similar to the clinical results, we observed no significant response to fenretinide + ketoconazole in TX-LY-183x and less than 1 log of cell kill with fenretinide (10 μmol/L) in TX-LY-172h. However, romidepsin + fenretinide + ketoconazole significantly delayed tumor growth in vivo and romidepsin + fenretinide induced > 5 logs cytotoxicity in vitro at the highest concentrations tested with significant synergy (CIN < 1).
In conclusion, we have demonstrated synergistic cytotoxicity between romidepsin and fenretinide in preclinical models of T-cell lymphoid malignancies via a novel mechanism. Our data support carrying out early-phase clinical trials of romidepsin + fenretinide in relapsed and refractory TCLMs.
Disclosure of Potential Conflicts of Interest
M.H. Kang is a consultant/advisory board member for CerRx, Inc. C.P. Reynolds is Chief Scientific Officer at CerRx, Inc. and has ownership interest in patents owned by CHLA. No potential conflict of interests were disclosed by the other authors.
Conception and design: M.R. Makena, T.H. Nguyen, M.H. Kang, C.P. Reynolds
Development of methodology: M.R. Makena, B. Koneru, C.P. Reynolds
Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): M.R. Makena, T.H. Nguyen, C.P. Reynolds
Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): M.R. Makena, T.H. Nguyen, C.P. Reynolds
Writing, review, and/or revision of the manuscript: M.R. Makena, B. Koneru, T.H. Nguyen, M.H. Kang, C.P. Reynolds
Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): M.R. Makena, C.P. Reynolds
Study supervision: M.R. Makena, C.P. Reynolds
The authors thank Celgene for supplying romidepsin, the Children's Oncology Group (COG) Cell Line and Xenograft Repository (www.COGcell.org) and the Texas Cancer Cell Repository (www.TXCCR.org) for providing cell lines and xenograft models, Dr. Connor Hall for establishing the COG-LL-317m Luc model, Dr. Sung Jen Wei for assistance with shRNA experiments, and the TTUHSC Cancer Center cell culture core staff for assisting with STR and mycoplasma testing.
Institutional support was provided by TTUHSC. COG cell lines and PDXs were supported by Alex's Lemonade Stand Foundation and awarded to C.P. Reynolds.
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.