PCI-24781 (abexinostat), a novel histone deacetylase inhibitor, induces reactive oxygen species-dependent apoptosis and is synergistic with bortezomib in neuroblastoma

In this study, we investigated the cytotoxic effects of a broad-spectrumhistone deacetylase(HDAC) inhibitor, PCI-24781, alone and in combination with the proteasome inhibitor bortezomib in neuroblastoma cell lines. The combination was shown to induce synergistic cytotoxity involving the formation of reactive oxygen species. The cleavage of caspase-3 and PARP, as determined by western blotting, indicated that cell death was primarily due to apoptosis. Xenograft mouse models indicatedincreased survival among animalstreated with this combination. The Notch signaling pathway and MYCN gene expression were quantified by reverse transcription-polymerase chain reaction (PCR) in cells treated with PCI-24781 and bortezomib, alone and in combination. Notch pathway expression increased in response to an HDAC inhibitor. NFKB1and MYCN were both significantly down regulated. Our results suggest that PCI-24781 and bortezomib are synergistic in neuroblastoma cell lines and may be a new therapeutic strategy for this disease.


Introduction
Neuroblastoma (NB) is the most common extracranial solid tumor in childhood, accounting for 12% of pediatric cancer fatalities annually in the United States [1]. Despite advances in therapy, cure rates for Stage 4 high-risk neuroblastoma continue to be poor, with a five-year survival rate of only 30-40% in spite of intensive multimodality treatment [2]. To date, no effective therapies exist for relapsed high-risk neuroblastoma,which remains uniformly fatal. The majority of cases of high risk neuroblastoma are initially sensitive to chemotherapy, but upon relapse they are inevitably highly chemoresistant [2,3]. Novel treatments are therefore urgently needed [3].
Epigenetic regulation of gene transcription via histone deacetylation is an important mechanism in cellular homeostasis and has been implicated in tumorigenesis. Specifically, acetylation of the e-amino group of lysine residues within the histone proteins maintains an open, transcriptionally active chromatin configuration [4,5]. Histone deacetylases (HDACs) are a family of enzymes that remove these acetyl residues, allowing chromatin folding and resulting in inhibition of transcription. Additionally, HDACs have been found to modify the lysine residues of impor-tant non-histone proteins, including tumor suppressor p53 [6]. Recent studies have shown that disruption of acetylationdeacetylation homeostasis may lead to the silencing of tumor suppressor pathways and may contribute to the development and progression of numerous malignancies [6][7][8].
HDAC inhibitors are a diverse class of compounds targeting those proteins that modulate transcription by removing acetyl groups from lysine residues in the histone tails of chromatin and by deacetylation of non-histone proteins. These compounds are actively being studied in numerous cancers [8][9][10][11][12]. A potentially important effect of HDAC inhibitors is their induction of reactive oxygen species (ROS) including superoxide, leading to increased oxidative stress and apoptosis [13][14][15][16]. HDAC inhibition produces growth inhibitory effects in NB cell lines in vitro [17,18]. First-generation HDAC inhibitors have had some clinical success to date, most notably vorinostat (suberoylanilide hydroxamic acid [SAHA]), which is approved by the FDA for use in cutaneous T-cell lymphoma [19]. A phase I study of vorinostat with or without 13-cis-retinoic acid in refractory pediatric cancers (including NB) was recently completed and showed that SAHA was well-tolerated in children, the subset analysis for NB for safety was not reported [20]. Despite their promise as novel anti-cancer agents,many first-generation HDAC inhibitors, including SAHA, have suffered from nonoptimal pharmacokinetics and dosage regimens, leading to poor tolerance at therapeutic drug concentrations. This has led to a substantial effort to identify new, second-generation HDAC inhibitors having greater potency and specificity [21].
PCI-24781 (abexinostat; Pharmacyclics, Inc.) is a novel, second-generation phenyl hydroxamic acid-based, orally bioavailable HDAC inhibitor that has previously been shown to have activity in vitro and in vivo against a broad array of cancers, including hematopoietic malignancies and bone and soft-tissue sarcomas [13,[22][23][24]. It has also shown good tolerability and activity in Phase I and II clinical trials against lymphoma, as well as against solid tumors in Phase-I trials [25]. Additionally, it acts as a potent radiosensitizing agent [26] and is synergistic with cytotoxic chemotherapy, such as doxorubicin [27] in preclinical models.
Bortezomib is a proteasome inhibitor currently used for first-line therapy of multiple myeloma and in relapsed mantle cell lymphoma [28]. In both of these tumors, increased generation of ROS has been associated with cell death [29][30][31]. Bortezomib was shown to induce apoptosis of NB cells as a single agent both in vitro and in vivo [32,33].
Recently, Bhalla et al., showed that the combination of PCI-24781 plus bortezomib was synergistic and induced ROSdependent apoptosis in Hodgkin and non-Hodgkin lymphoma cells [13]. To our knowledge, however, there have been no reports of combining proteasome and HDAC in hibitors in NB. In this study, we show that the potent HDAC inhibitor PCI-24781 is both cytotoxic to NB cells as a single agent at nanomolar concentrations and is synergistic in vitro and in vivo with the proteasome inhibitor bortezomib. We also show that this combination therapy induces apoptosis, up-regulates Notch pathway signaling, and inhibits MYCN expression. These effects are abrogated by pretreatment with the antioxidant N-acetylcysteine, suggesting that ROS-mediated apoptosis induction may bea major mechanism of action.

Human neuroblastoma cell lines
The following NB cell lines were used in all experiments: SMS-KCNR, SKNBE(2)C, LA1-55n, and LAN2. The SMS-KCNR cell line was obtained from the lab of Dr. John Maris (Children's Hospital of Philadelphia, Philadelphia, PA). SKNBE(2)C, LA1-55n, and LAN2 cell lines were a gift from the labs of Drs. Robert Ross and Barbara Spengler (Fordham University, Bronx NY). All cell lines were authenticated using the Short Tandem Repeat method by the DNA Diagnostics Center (Fairfield, OH). All cell lines were cultured at 37°C under humidified atmospheric oxygen and 5% CO 2 in RPMI 1640 (Mediatech, Manassas, VA) supplemented with 10% (v/v) FBS (Invitrogen, Carlsbad, CA), 100 units/mL penicillin, and 100 µg/mL streptomycin (Mediatech) as previously described [34][35][36].

Drugs and antibodies
The second-generation HDAC inhibitor PCI-24781 was provided by Pharmacyclics, Inc. (Sunnyvale, CA). Bortezomib and the HDAC inhibitor suberoylanilidehydroxamic acid (SAHA/ vorinostat) were obtained from LC Laboratories (Woburn, MA). Sodium butyrate and valproic acid were from Sigma Aldrich (St. Louis, MO). Rabbit-anti-human cleaved-caspase-3 and mouse-anti-human PARP antibodies were purchased from Cell Signaling Technology (Danvers, MA).

Cell viability assays
Neuroblastoma cell lines were treated with various concentrations of bortezomib and PCI-24781(1-1000nM) alone or in combination for 48 h. Cells were also treated with the HDAC inhibitors SAHA, sodium butyrate, and valproic acid. Cell viability was assessed using Calcein AM (AnaSpec, Freemont, CA). IC 50 values for each compound or combination were calculated using Prism 5 software (GraphPad Inc., La Jolla, CA). Cell viability assays were repeated in the presence of 15mMN-acetyl-L-cysteine (Sigma Aldrich, St. Louis, MO).
Measurements of fractional survival, s , at concentrations of bortezomib b , and PCI-24781 p, were interpreted with the aid of the models= exp(β 10 b + β 01 p + β 20 b 2 + β 02 p 2 + β 11 bp)+ ε, where β is the vector of model parameters and ε is the random error. Model parameters were estimated using the R software program (www.R-project.org).

Western blotting
Cells were treated for 48 h with HDAC inhibitors and/or bortezomib in 100-mm plates. Cells were collected by scraping, protein was extracted, and electrophoresis performed on a 15% SDS-polyacrylamide gel and blotted onto polyvinylidene fluoride (PVDF) membranes. The blots were probed with rabbit anti-human antibody to cleaved caspase 3 and β-actin, as well as mouse anti-human PARP antibody. Protein bands were visualized using infrared dye-conjugated anti-rabbit secondary antibodies (LI-COR Biosciences, Lincoln, NE) and photographed using an Odyssey Infrared Imaging System (LI-COR Biosciences).

mRNA preparation and gene expression profiling
SMS-KCNR cells were treated in duplicate with 4nM bortezomib, 125nM PCI-24781, or the combination for 24 h. RNA was extracted using the RNeasy micro kit (Qiagen, Valencia, CA) following the manufacturer's instructions and eluted in Riboblock RNase inhibitor (Formentas). RNA quality was verified with all calculated RIN scores above 9 (Quality control pass is RIN>6.5). Five micrograms of RNA per array was subsequently hybridized to Affymetrix GeneChip Human U133 Plus 2.0 arrays per the manufacturer's protocols. Arrays were analyzed by the Vermont Genetics Network Microarray Facility using Affymetrix GCOS software, according to the manufacturer's instructions. All other calculations were performed using R/ BioConductor tools available from http://www.R-project.org. Probe set by sample matrix expression statistics were calculated using the Robust Multichip Average method [37,38]. Quality statistics were calculated using the Simpleaffy and affyQCReport packages [39].

Quantitative RT-PCR
Notch pathway and MYCN gene expression levels were measured using an Applied Biosystems 7500 Real-Time PCR system (Applied Biosystems, Foster City, CA) using dual-labeled fluorescent probes, with TaqMan One-Step RT-PCR Master Mix reagents (Applied Biosystems). Briefly, 50 or 100ng of mRNA per reaction was used. Reaction conditions were 48°C for 30 min, 95°C for 10 min, followed by 95°C for 15 s and 60°C for 1 min, for a total of 40 cycles. Gene expression for NOTCH1-4, JAG1 and 2, HES1, and MYCN was measured using TaqMan Assay-on-Demand predesigned probe sets (Applied Biosystems). The expression level of 18S ribosomal RNA was used as an endogenous control to normalize expression results, and fold changes between samples were calculated using the 2 -ΔΔCt method according to the manufacturer's instructions. Statistical analysis was performed using InStat Software (Graphpad).

In vivo assays
The Animal Care and Use Committee at the University of Vermont approved all animal studies. Six-week-old female nude mice (NCRNU-F, CrTac:NCr-Foxn1 nu ) (Taconic Farms, Germantown, NY) were injected with 8x10 6 SMS-KCNR cells suspended in 0.2mL of Matrigel subcutaneously into the left flank. Beginning 10 d post-injection, 6 mice per group were treated with daily intraperitoneal doses of vehicle control, 0.5 mg/kg bortezomib, 12.5 mg/kg PCI-24781, or a combination of bortezomib plus PCI-27481 at 0.5 and 12.5 mg/kg, respectively. Tumors were measured twice weekly and tumor volume was calculated using the ellipsoid formula [length x width x height x 0.52]. Treatment was continued until tumors exceeded 3.0 cm 3 , at which point the mice were euthanized. Tumors were collected and weighed. The experiment was performed in duplicate.

PCI-24781 and bortezomib exhibited single-agent cytotoxicity against NB cell lines
Four neuroblastoma cell lines SKNBE(2)C, SMS-KCNR, LA1-55n, and LAN2, were treated with various concentrations of four different HDAC inhibitors. Three of the inhibitors were first-generation drugs being tested in neuroblastoma clinical studies (SAHA, sodium butyrate, and valproic acid), and the fourth was the second-generation HDAC inhibitor PCI-24781 (Figure 1a). PCI-24781 exhibited the greatest cytotoxicity against the NB cell lines, with IC 50 values ranging from 83.5nM for SMS-KCNR to 196.4nM for SKNBE(2)C. The IC 50 values for the other HDAC inhibitors were for SAHA, 5.8-123.3 mM; for sodium butyrate, 3.8-10.4 mM; and for valproic acid, 2.0-17 mM. Bortezomib also exhibited cytotoxicity against the SMS-KCNR, LA1-55n and SKNBE(2)Ccell lines (Figure 1b). The IC 50 of bortezomib for these cell lines was estimated to be 2.4nM for SMS-KCNR, 3.6nM for LA1-55n, and 3.5nM for SKNBE(2)C, but 6.31nM for LAN2 cells.

PCI-24781 and bortezomib induced caspase-3-mediated apoptosis in NB cell lines
Treatment of neuroblastoma cell lines (SMS-KCNR and LA1-55n) with PCI-24781 and bortezomib agents led to induction of apoptosis as measured by the presence of cleaved PARP and of cleaved caspase-3 by western blotting (Figure 2). The amount of cleaved caspase-3 was greater after treatment with the combination of agents than with either agent alone.

N-acetyl-cysteine pretreatment blocked cytotoxicity of PCI-24781 and bortezomib
PCI-24781 induces the formation of reactive oxygen species in neuroblastoma cells. To confirm that the effect of these drugs involved the formation of ROS, cells were pretreated with the antioxidant N-acetyl-cysteine (NAC). Cells were thentreated with either PCI-24781 (100nM), bortezomib (6nM), or both. In the absence of NAC, a decrease in cell viability was seen for the combination relative to the single agents (Figure 3). With 15mMNAC pretreatment, the cytotoxicity of the combination was greatly decreased, indicating that the cytotoxicity induced by these agents involves ROS generation (p<0.005 based on 9 experiments representing a range of PCI-24781 and bortezomib concentrations, data not shown).

PCI-24781 and bortezomib up-regulated NOTCH pathway genes and suppressedMYCN
SMS-KCNR cells were treated with PCI-24781 (125nM), bortezomib (4nM), or the combination. After24 h of treatment, microarray analysis revealed significant up-regulation of multiple various Notch signaling pathway genes (  sample groups were identified based on at least 1.2 fold-change (in either direction) and adjusted p-value < 0.05 (Figure 4a, Table 1). Treatment with PCI-24781 alone resulted in up-regulation of JAG1 at 6 h and up-regulation of NOTCH2 at 24 h. These findings were confirmedusing independent biological replicates by real-time qPCR (Figure 4b).
Of importance in neuroblastoma, MYCN was suppressed with either bortezomib or PCI-24781 as well as with the combination at 24 h relative to controls.

PCI-24781 and bortezomib in combination increased survival in a mouse xenograft models
Nude [nu/nu] mice (n = 6) were injected with SMS-KCNR cells subcutaneously in the left flank. Once the tumors were palpable, the mice were treated once daily with vehicle, 0.5 mg/kg bortezomib, 12.5 mg/kg PCI-24781, or a combination of both. The tumors were measured twice weekly. Mice were euthanized when a maximum tumor size of 3 cm 3 was reached.  There were no signs of drug-related toxicities as noted by change in body weight or behavior. Neuroblastoma xenograft mice models showed an increase in time to survival from 55 days in the control group to 80 days in mice treated with the PCI-24781 plus bortezomib combination (Figure 5; p = 0.001).

Discussion
In this study, the HDAC inhibitor PCI-24781 had significant activity in vitro and in vivo as a single agent against several NB cell lines. Specifically, PCI-24781 showed dose-dependent cytotoxicity with an IC 50 of <200nM, which is well within previously reported clinically achievable range [40]. This dose range was shown to have no cytotoxic effect on human normal foreskin fibroblast cell line . This is in contrast to the other HDAC inhibitors tested (SAHA, sodium butyrate, and valproic acid),which had IC 50 values in the micromolar to millimolar ranges. PCI-24781 induced apoptosis in all the NB cell lines tested, as shown by cleavage of caspase-3 and PARP. This effect was notably increased when combined with bortezomib. While PCI-24781 showed significant single-agent cytotoxicity, its synergy with bortezomib was striking. At doses that caused no effect as single agents, the combination of PCI-24781 plusbortezomib resulted in cell viability of 5% or less in all the cell lines tested. Treatment of mouse xenografts with this drug combination led to a doubling of mean survival time from 55 to 80 days. Importantly, the animals showed no evidence of systemic toxicity as reported by Bhalla et al., as well [42], suggesting that this combination could be well tolerated in relapsed neuroblastoma patients, all of whom have been heavily pretreated with cytotoxic chemotherapy.
Gene expression profiling of NB cells treated with PCI-24781 plusbortezomib identified significant down-regulation of numerous genes, including MYCN and NFKB1. MYCN is a critically important proto-oncogene in NB, with nearly 1/3 of cases having MYCN amplification and associated poor outcomes [2]. A report by Nawrocki et al., [43] showed that Myc regulates the sensitivity of multiple myeloma cells to bortezomib and SAHA-induced cell death. Interestingly, they found that the sensitivity of multiple myeloma to treatment with the combination of bortezomib and SAHA was increased in patients with overexpression of c-Myc. Our finding that treatment with PCI-24781 significantly down-regulated the expression of this critical neuroblastoma oncogene suggests that MYCN-amplified patients may particularly benefit from this therapy.
The mechanism by which HDAC inhibitors and bortezomib cause cell death remains controversial. Our results suggest that the observed cytotoxicity is ROS-dependent, as the effects of these compounds (especially in combination) were significantly decreased when cells were pretreated with the antioxidant NAC. Both PCI-24781 and bortezomib have been shown independently to cause ROS generation; additionally, PCI-24781 was previously shown to suppress NF-KB expression, a critically important regulator of the cellular oxidative stress response [13]. Our findings confirm these observations. It seems plausible that the combination of these two compounds leads to both increased ROS generation and suppression of important oxidative stress responses, leading to increased apoptosis.
Interestingly, our gene expression studies identified the Notch signaling pathway as a significantly up-regulated pathway in our cell lines after treatment with PCI-24781 plus bortezomib. While it is well established that Notch genes function as oncogenes in hematologic malignancies, most notable T-cell acute lymphoblastic leukemia, its specific role remains controversial in solid tumors [44]. Notch signaling has been shown to be oncogenic in breast and colon cancers among others [45,46], but it acts as a tumor suppressor in skin and pancreatic cancer models [47,48]. Two previous studies of HDAC inhibitors in neuroblastoma showed up-regulation of the Notch pathway in response to HDAC inhibitor treatment [49,50]. They also showed that this treatment up-regulatedgenes known to be involved in cellular differentiation and growth arrest, leading them to hypothesize that increased Notch signaling was exerting a tumor suppressor effect. On the other hand, more recent studies in neuroblastoma have suggested that increased levels of Notch pathway gene expression correlate with advanced stages and worse clinical outcome, and that blockade of Notch signaling inhibits neuroblastoma growth in vitro and in vivo [51,52]. These seemingly contradictory roles of Notch signaling suggest that its effects are likely highly dependent on cellular context and on interactions with other signaling pathways and the tumor microenvironment. This has clearly been shown to be true for many other solid tumors [44]. Additional studies are needed to draw conclusions regarding the role of Notch signaling in neuroblastoma.
In summary, our results show that the HDAC inhibitor PCI-24781, when given as a single agent, exhibited significant cytotoxicity and induced apoptosis in a panel of NB cell lines. The combination of PCI-24781 and bortezomib was strongly synergistic in cell lines tested and significantly prolonged survivalin xenograft models compared to either agent alone. This anti-NB activity appears to be ROS-dependent, as shown by prevention of apoptosis and cell death with NAC pretreatment. Enhanced generation of ROS combined with simultaneous inhibition of oxidative stress response genes may represent an important mechanism underlying NB cell death caused by this drug combination. Our finding of significantly decreased expression of NF-KB, one of the primary regulators of oxidative stress response in cells, supports this hypothesis. Finally, we showed that treatment with this combination led to significant up-regulation of many members of the Notch signaling pathway, suggesting a novel pathway of action of these compounds in neuroblastoma that deserves further study. These data suggest that PCI-24781 has a potential for