Invest New Drugs DOI 10.1007/s10637-014-0066-9

PHASE II STUDIES

Phase II trial of vorinostat in advanced melanoma N. B. Haas & I. Quirt & S. Hotte & E. McWhirter & R. Polintan & S. Litwin & P. D. Adams & T. McBryan & L. Wang & L. P. Martin & M. vonMehren & R. K. Alpaugh & J. Zweibel & A. Oza

Received: 3 December 2013 / Accepted: 10 January 2014 # Springer Science+Business Media New York 2014

Summary Introduction Vorinostat is a small molecule inhibitor of class I and II histone deacetylases with preclinical activity in melanoma. Methods We evaluated 32 patients with advanced primary cutaneous or ocular melanoma in a multiinstitutional setting (PMH Phase II Consortium) with continuous daily oral vorinostat 400 mg. The primary endpoint was response rate by RECIST, with time to progression as a secondary endpoint. The study was designed to distinguish a response rate of 20 % from a RR of 5 % and to distinguish a 2 month median progression-free survival (PFS), from one of 3.1 months. The study proceeded to stage 2 following 2 of 16 responses.. We also assessed VEGF, FGF levels, P52 polymorphisms and chromatin-associated proteins as potential biomarkers. Results Therapy was associated with significant side effects, including fatigue, nausea, lymphopenia, and hyperglycemia. Eleven patients experienced at least one grade 3 or higher adverse event. There were two confirmed PRs in patients with cutaneous melanoma. Sixteen patients had stable disease and 14 patients had progressive disease for best response. In addition, two patients with cutaneous melanoma N. B. Haas (*) University of Pennsylvania, Philadelphia, PA, USA e-mail: [email protected] I. Quirt : R. Polintan : L. Wang : A. Oza Princess Margaret Cancer Centre, Toronto, ON, Canada S. Hotte : E. McWhirter Juravinski Cancer Centre, Hamilton, ON, Canada S. Litwin : L. P. Martin : M. vonMehren : R. K. Alpaugh Fox Chase Cancer Center, Philadelphia, PA, USA P. D. Adams : T. McBryan University of Glasgow, Glasgow, Scotland, UK J. Zweibel Clinical Trials Evaluation Program, Bethesda, MD, USA

scored as stable disease had early unconfirmed partial responses with subsequent progression. Patients with stable disease or partial response (n=18) had a median progression free survival of 5 months. (range 2–12 months). Conclusions Vorinostat demonstrated some early responses and a high proportion of patients with stable disease, but did not meet its primary endpoint of response. Different schedules of this agent with BRAF mutation status and markers of histone acetylation could be explored in melanoma. Keywords Histone deacetylase inhibitor . Melanoma . Angiogenesis . Chromatin associated proteins

Introduction Histone deacetylase (HDAC) inhibitors (HDACis) Histones are proteins that direct chromatin structure and function. The dynamic equilibrium between histone acetylation and deacetylation is regulated by histone acetyltransferases (HATs) and HDACs. The action of HDACs on nucleosomal histones is thought to lead to tight coiling of chromatin and silencing of expression of various genes, including those implicated in the regulation of cell survival, proliferation, differentiation, and apoptosis [1]. The effects of HDACis are not limited to histone deacetylation; they also act as members of a protein complex to recruit transcription factors to the promoter region of genes, including those of tumor suppressors, and they affect the acetylation status of specific cell cycle regulatory proteins [2]. Because, epigenetic regulation of cell function is broad, the effects of HDACis on cell function are expected to be broad. Vorinostat (suberoylanilide hydroxamic acid) is a small molecule inhibitor of histone deacetylase (HDAC). Vorinostat targets most human Class 1 (related to the yeast transcriptional regulator Rpd3) and Class 2 (similar

Invest New Drugs

to the yeast Hda1) HDAC enzymes [3, 4]. Based on its preclinical activity, vorinostat was investigated in a multicenter phase II trial of advanced melanoma. HDAC inhibitors in melanoma Promising preclinical and early clinical data of various HDAC inhibitors led to the exploration of vorinistat in advanced melanoma. The HDACis, apicin and suberic bishydroxamate, and depsipeptide demonstrated anti-invasive, anti-angiogenic and apoptotic activity in cutaneous and uveal melanoma cell lines, respectively [5, 6] Another HDACi, sodium butyrate, potentiated the effect of retinoic acid in murine and human melanoma cell lines [7]. In vivo studies of HDACis in combination with gene therapy using herpes simplex virusthymidine kinase gene (HSV-tk) followed by administration of ganciclovir (GCV) prolonged survival times in a mouse melanoma model [8]. In addition, Trichostatin A used in combination with depletion (by knockout) of Enhancer of Zest Homologue 2 (EZH2), expressed in most melanoma cell lines, synergistically activated p21/CDKN1A expression in melanoma xenografts, leading to cell senescence [9]. Finally, the activity of HDACis (LAQ824, vorinostat and doxorubicin, vorinostat alone) demonstrated in several phase I trials, either as monotherapy or in combination, supported investigating vorinostat in melanoma [10–13]. Potential biomarkers of vorinostat Chromatin repair The anti-tumor activity of vorinostat is thought to depend, in part, on its preferential killing of tumor cells, compared to normal cells. Therefore, with a view to development of molecular biomarkers, and based on previous cell-based studies [14, 15], we proposed a novel mechanism to account for the preferential killing of tumor cells by HDACi: that inhibition of HDACs by HDACi causes histone hyperacetylation, a form of “chromatin damage”. We sought to explore the idea that normal cells are able to “repair” this damage, and so are relatively resistant to the effects of HDACi, while tumor cells are deficient in these repair mechanisms, and so are more sensitive to killing by these drugs. As preliminary data, we defined a number of gene products likely to participate in this chromatin repair response, so-called chromatin repair genes [14], including chromatin-associated proteins (HP1a, b and g), chromatin regulators (HIRA) and histone variants (histone H3.3 and macroH2A). We hypothesized that monitoring the chromatin repair response in normal tissue of patients treated with vorinostat would give a read-out of a pharmacologically active dose of the drug. A previous intravenous vorinostat phase I trial [13] demonstrated an increase in immunohistochemical staining using an anti-acetylated-H3 antibody in post

therapy tumor biopsies of patients receiving intravenous vorinostat and appeared to be a valuable marker in determining inhibition of target enzymes by vorinostat in vivo. We wanted to investigate this further and thus we measured histone acetylation in normal skin biopsies as a comparison to HP1 and macroH2A as targets. We determined the level of expression and subcellular localization of chromatin repair proteins, namely HP1γ, macroH2A1.1 and macroH2A1.1, by IHC in FFPE and frozen blocks of normal skin, both pretreatment and on-treatment. p53 polymorphisms Clinical evidence also indicated that response following some therapies for melanoma is influenced by a naturally occurring polymorphism at residue 72, with cancers expressing Arg 72 having lower response rates than those expressing Pro 72 [16]. Anti-apoptotic mechanisms related to these two structurally and functionally distinct p53 alleles resulting from the substitution of an arginine (R) to a proline (P) at codon 72, might influence melanoma progression. Shen and colleagues also demonstrated increased frequency of cutaneous melanoma in populations with the p53 72R allele [17]. In vitro cell lines containing inducible versions of alleles encoding the 72R and 72P variants, and cells with wild-type endogenous p53, the 72R variant induced apoptosis markedly better than did the 72P variant [18]. Additionally, the response following treatment for some cancers appeared to be influenced by this polymorphism, with cancers expressing 72R mutants having lower response rates than those expressing 72P mutants [16]. Thus, we hypothesized melanoma progression and response to vorinostat therapy might be influenced by inter-patient variation of these alleles, and this was evaluated for participants in this trial. Angiogenesis Anti-angiogenesis mechanisms are important in treatment of melanoma due to a variety of observations: 1. VEGF expression is increased in malignant melanoma cell lines in comparison with normal melanocytes [19, 20]. 2. basic fibroblast growth factor (bFGF) is expressed by most melanoma cells but not by normal melanocytes [21]. 3. Low expression of serum bFGF in patients receiving high dose interferon was associated with higher recurrence-free survival and may be indicative that angiogenesis inhibition plays a role in survival in melanoma [22]. 4. Several agents with known anti-angiogenic mechanisms of action (TNP470, thalidomide and small molecule inhibitors targeting VEGF, PDGF, or FGF) have demonstrated activity against melanoma in small

Invest New Drugs

phase I and II trials [23]. Based on these observations and a potential anti-angiogenic role of vorinostat [23–25], we hypothesized that (1) patients with elevated serum bFGF levels would have more aggressive melanoma, which might respond to vorinostat; and (2) levels of serum VEGF and bFGF would decrease in response to vorinostat therapy. We therefore assessed the effects of vorinostat on serum levels of VEGF and bFGF, with inhibition of serum VEGF and b-FGF levels of greater than 15 % considered significant [22].

Methods Clinical This was a CTEP-sponsored phase II open label multi center trial led by the Princess Margaret Phase II Consortium with the participation of the Fox Chase Cancer Center, Princess Margaret Hospital and Juravinski Cancer Center in Hamilton. The primary objective was to determine the activity of vorinostat in advanced melanoma using RECISTv1.0 criteria [26]. Secondary objectives were to determine whether vorinostat delayed time to progression in advanced melanoma and to identify potential biomarkers indicative of activity. Eligible patients had histologically confirmed cutaneous, ocular, mucosal or unknown origin melanoma that was metastatic or unresectable and evaluable by RECIST. Patients were allowed to receive up to one prior therapy, including vaccine, for advanced disease and prior adjuvant interferon or vaccine for stage II or stage III disease. Prior radiation therapy was also allowed. Patients had to be at least 4 weeks from prior therapy to be eligible (or 6 weeks if the last regimen included BCNU or mitomycin C). Patients could not take valproic acid (an HDAC inhibitor) for at least 2 weeks prior to enrollment. Systemic steroids were prohibited. Patients had to have a life expectancy of at least 3 months, an ECOG performance status of at least 2 and be at least 18 years of age. Patients had normal organ function (leukocytes ≥3,000/mcL;absolute neutrophil count ≥1,500/mcL; platelets ≥100,000/mcL; total bilirubin within normal institutional limits; AST(SGOT)/ALT(SGPT) ≤2.5 X institutional upper limit of normal; creatinine within normal institutional limits or creatinine clearance ≥60 mL/ min/1.73 m2 for patients with creatinine levels above institutional normal); and patients with childbearing potential or sexually active, had to agree to use an accepted and effective method of contraception prior to study entry and for the duration of the study. Patients registered on or after October 01, 2007 had to have a paraffin block of tumor tissue available for future studies. In addition, pregnant patients, HIV-positive patients receiving combination antiretroviral therapy, and patients with uncontrolled medical illnesses that would limit

compliance with study requirements were excluded from participation. Vorinostat was administered at an oral dose of 400 mg for 28 consecutive days per cycle. Vorinostat capsules were taken at the same time daily (±1 h), with food, and recorded in a diary provided to each patient, with a follow-up pill count at the end of each cycle. Radiologic assessment of response by computed tomography or magnetic resonance imaging was measured according to the RECIST every 8 weeks (every 2 cycles). In addition, a bone scan was completed at baseline and repeated every 16 weeks if initially abnormal due to metastatic disease. Computed tomography or magnetic resonance imaging of brain was done pre-study, if patient was suspected of having brain metastasis, and then repeated as clinically indicated. Toxicities were defined by the NCI – CT CAD version 3.0. Response rate by RECIST was the primary endpoint, with time to progression as a secondary endpoint. The study was designed to distinguish a RR of 20 % from 5 % (90 % power and 7 % Type I error) and to distinguish a 2 month median progression-free survival (PFS), from one of 3.1 months with 80 % power and 5 % type I error. The study was designed to proceed to a second stage following at least one of 16 responses in the first stage. The new treatment would be of interest if the proportion of patients progression-free at 3.1 months was at least 30 %. Laboratory correlatives Three correlative objectives were analyzed as outlined in the introduction: 1. To determine the utility of HP1 and/or macroH2A nuclear foci as biomarkers of response to vorinostat; 2. To determine if there was a relationship between the presence of 72R or 72P variant p53 polymorphisms in patients and response to therapy; 3. To determine if vorinostat inhibited markers of angiogenesis using serum VEGF and bFGF as surrogate markers. Collection of materials for correlative studies Pre-treatment and on-treatment punch biopsies of normal skin were collected from all patients prior to treatment and at day 15 (±1 day) of treatment cycle 1, respectively. Pre-treatment and on treatment blood samples for preparation of peripheral blood mononuclear cells (PBMCs) and angiogenic markers (VEGF and FGF) were collected on day 1, cycle 1 prior to treatment and day 15 (±1 day), cycle 1. Day 1, cycle1 blood was also collected for analysis of genomic DNA for P53 codon 72 mutational analysis. Additionally two patients had tumor biopsies (both formalin fixed and frozen) pre and post vorinostat administration sent to Quest Diagnostics (MERCK). Most but not all patients had an original paraffin tumor block from diagnosis available.

Invest New Drugs

Chromatin repair analysis correlative studies in normal skin Using standard institutional procedures, a punch biopsy of skin was obtained using a 3–4 mm punch instrument. The specimen was bisected perpendicular to the skin surface and half of the sample was placed in 10 % formalin, and transported to the Protocol Support Laboratory at Fox Chase Cancer Center. The remaining one-half of the bisected specimen was placed immediately in OCT freezing media in a block mold and placed immediately in −70 °C and then shipped on dry ice overnight to the Protocol Support Laboratory at the Fox Chase Cancer Center. Normal tissue from the punch biopsy was analyzed at the Beatson Institute in Glasgow, Scotland by Dr. Peter Adams. The level of expression of known chromatin repair proteins (namely HP1γ, macroH2A1.1 and macroH2A1.1) was analyzed by immunohistochemistry (IHC). As negative controls, tissue was also stained with irrelevant species and classmatched antibodies (mouse monoclonal 9E10 (Santa Cruz) and in-house raised and purified anti-mouse rabbit polyclonal (Sigma). For IHC, mouse monoclonal to HP1γ(Millipore Mab3450), and in-house raised and purified rabbit polyclonal antibodies to macroH2A1.1 and 1.2 were used. Sections from FFPE tissue blocks were stained according to a standard protocol with antigen retrieval in pH6 sodium citrate buffer. Quantitation was performed on the Clarient ACISII Chromavision quantitative IHC microscope and associated software. The IHC results from the skin specimens (macroH2a 1.1 and 1.2, and HP1), were analyzed by a variety of statistical methods. Briefly, the data was analyzed using a repeated measures anova test within PASW18, Mauchlys test of Sphericity was applied to test for equal variances; as a result the Greenhouse-geisser correction was applied to the resulting model, and visualized by box plots, to determine if there was intrapatient variability over time (pre and post treatment with vorinostat), or interpatient variability according to outcome (SD or DP). Fisher’s exact test was applied to the classification HP1 or macroH2A nuclear localization foci localization (+ or −) and response (+ or −) to address if these were useful biomarkers for response to vorinostat. The hypothesis, based on previous in vitro studies [14], was that clinically beneficial doses of vorinostat trigger formation of HP1 and/or macroH2A nuclear foci. The null hypothesis was that patient response and chromatin repair are unrelated versus the alternative that expression of HP1 or mutation of HIRA was correlated with patient response. This was done with defined power of 81.3 % and type I error of 2.51 %. P53 polymorphism as indicator of response to therapy A blood specimen was collected in a 9 ml EDTA tube from patients on Day 1, cycle 1 prior to treatment. The plastic 9 mL

EDTA tube was frozen at −70 °C and shipped on dry ice to the Protocol Support Laboratory at the Fox Chase Cancer Center. Genomic DNA was extracted from 3 ml of anticoagulant whole blood using the Wizard® Genomic DNA Purification Kit (Promega Corporation) according to manufacturer’s directions. The quality and quantity of DNA was verified using spectrophotometry. Extracted DNA was stored at 4 °C until further analysis. Genotyping was done at FCCC in the Biomarker and Genotyping Facility. This laboratory was equipped to carry out high-throughput analyses of single nucleotide polymorphisms (SNPs). The SNP of interest was analyzed using TaqMan assays in a 96 well plate format. The p53 codon 72 Arg/Pro polymorphism was analyzed using a fluorogenic 5′ nuclease allelic discrimination assay with primers and TaqMan probes as described by the NCI SNP500 Cancer Database (http://snp500cancer.nci.nih.gov/home.cfm). Primers and probes were synthesized by Applied Biosystems (Foster City, CA). The analysis was performed using an ABI7900 Sequence Detection System (Applied Biosystems) according to manufacturer’s instructions. Control DNA samples with known p53 codon 72 genotypes, confirmed by standard sequencing, were included in each run. In addition, a no template control was included to assess contamination. SNP assignment was achieved automatically with the SDS software (Applied Biosystems) using a proprietary algorithm. The analysis typically yields three major clusters corresponding to homozygous or heterozygous genotype. After automatic SNP scoring and quality assessment, the operator can choose to inspect raw data and manually edit the results. Raw data, original results, and edited results were all stored on a Hewlett Packard Vectra hard drive. Genotyping data was exported to Excel spreadsheets for further analysis. P53 allelic variations were also assessed by Fisher’s exact test to the classification presence or absence of polymorphism and presence or absence of response and progression greater than or less than 2 months. Markers of angiogenesis The pre-treatment serum VEGF and bFGF blood specimen were collected from patients on Day 1, cycle 1 prior to treatment. Samples were also obtained in patients 3–4 h post ingestion of vorinostat day 1, day 8 and day 15 (±1 day) of therapy. Serum was frozen at −70 °C and assayed using the Luminex® Assay, a fluorescent bead based immunodetection system similar to ELISA (Life Technologies™ Grand Island, NY) which allowed for multiplexing the VEGF and b-FGF assays. Values were extrapolated from a standard curve. Samples were batched until the end of study and assayed in the Godwin laboratory at the Fox Chase Cancer Center, Philadelphia, PA. The null hypothesis was that VEGF or b-FGF levels after treatment would be the same as at baseline (H0: p[baseline level > post treatment level] = 1/2, versus H1: p[baseline level > post treatment level] = 1/2 + delta where delta > 0.)

Invest New Drugs Table 1 Demographics

Table 2 Reported serious adverse events of patients receiving vorinostat cycle 1

Age

Median (range)

61(38–89)

Gender PS Prior regimens Prior Therapy

Female/Male 0/1/2 0/1/2 Adjuvant CT Systemic CT RT Median (range) Lung Liver Nodes Abdomen Pelvis Other PMH Fox Chase Hamilton Median (range)

12/20 17/141 18/13/1 0 14 10 4 (1–10)/2 (0−6) 15/17 12/8 13/11 6/4 2/0 7/8 11 13 8 4 (1–12)

PR SD PD PD Withdrawal Non compliance

118 2 16 14 30 1 1

Target/Non-target lesions

Hospital

Cycles/Patient Total cycles Best objective response

Off-treatment reason

Results Toxicity One hundred and eighteen cycles of treatment were given to 32 patients with advanced cutaneous or uveal melanoma (Table 1). Grade 3–5 adverse events are listed in Table 2. Therapy was moderately toxic: Eleven [52 %] patients experienced at least one grade 3 or higher adverse event. Moreover, most patients required at least one dose reduction or interruption in therapy due to fatigue (18 [86 %] patients) or nausea (15[71 %]). Other common adverse effects were lymphopenia (13 [62 %]) and hyperglycemia (14 [67 %] patients). Response All of the planned 32 patients were registered by 3 institutions and accrued to this study from September 2005 to July 2008. A waterfall plot of response is shown in Fig. 1. There were 2 partial responses in patients with cutaneous melanoma origin: One remained on study for 7 cycles and the other for 5 cycles. One patient had metastases to the lung and a retina, and the second patient had lung and bone metastases. Sixteen patients had stable disease. Of these, ten had

Grade

#pts/grade 3, 4, 5

Worst grade

# Cycles grade 3, 4, 5

Fatigue Anemia Thrombosis Hyperglycemia Hypophosphatemia INR increased Thrombocytopenia ↑ Alkaline phosphatase Lymphopenia

5 2 3 3 3 1 3 2 3

4 3 4 3 3 3 4 3 3

6 6 6 5 5 4 4 3 3

Vascular disorder ALT increased Anorexia Back pain Dyspnea Neutropenia Hyponatremia Syncope ↑ Bilirubin ↑ Creatinine

1 2 1 2 1 2 2 2 1 1

3 3 3 4 4 4 3 3 3 3

3 2 2 2 2 2 2 2 1 1

cutaneous melanoma, five had uveal melanoma primaries and one was of unknown primary origin. Two additional patients with cutaneous melanoma, scored as stable disease, had brief but dramatic responses (33-50 % shrinkage) which lasted about two months, but not long enough to be confirmed responses as per protocol definition. Both of these patients had lung and bone metastases. The patients with stable disease or partial response (n=18) had a median PFS of 5 months (range 2–8 months). Finally, 14 patients developed

Fig. 1 Waterfall plot for target lesions in all patients

a

Legend for above: Blue = MH2A1.1 Green = MH2A1.2 Yellow = HP1

Log Fold Change (post vs pre)

0.2

0.4

0.6

0.8

Progression-free Estimate 95% CI

0.0

Progression-free proportion

1.0

Invest New Drugs

0

2

4

6

8

10 Outcome

months

progressive disease for best response. The median PFS overall was 4 months (range 1–8) (95 % CI: 1.8–4 month) and thus the secondary activity criteria was met. This is shown in Fig. 2. Correlative results Chromatin repair analysis Two samples from each of 31 patients, pre-treatment (Day 1) and post-treatment (Day 15), were analyzed for chromatin associated proteins. Patient 32 had a pretreatment punch biopsy that was inadequate for testing. MacroH2A1.2 IHC expression appeared to increase after exposure to vorinostat in patients with stable disease, but to decrease in patients with progressive disease. An example is shown in Fig. 3. However, the change in expression of MacroH2A1.2 was not

b 12.00 Fold change in Histone Acetylation

Fig. 2 Progression-free survival of patients on continuous oral vorinostat

Legend: Blue = H3 Green = H4

10.00

8.00

6.00

4.00

2.00

0.00 PD

SD Outcome

Fig. 4 Box Plots The change in expression of MacroH2A1.2 was not statistically significant as confirmed by analysis of log fold changes in antibody expression in a multivariate general linear model (MacroH2A1.1: p=0.685, MacroH2A1.2: p=0.120, HP1: p=0.847) a Log Fold Change (post vs pre) b H3 and H4 acetylation were similarly analyzed and although there was a slight trend in H3 acetylation in predicting stable disease from progressive disease it did not achieve statistical significance

statistically significant as confirmed by analysis of log fold changes in antibody expression in a multivariate general linear

Table 3 MacroH2A multivariate model pairwise comparisons. MacroH2A1.1, MacroH2A1.2 and HP1 were compared with SPSS within a multivariate general linear model between PD and SD outcomes. The results of each comparison indicated p>0.05 in each case Dependent Variable

Note possible foci, as described before in cell culture. [14]

Fig. 3 IHC expression of macroH2A1.2 1000x in patient 13 punch skin biopsy post-vorinostat therapy

(I) outcome (J) outcome Mean Std. Difference Error (I-J)

MacroH2A1.1 PD MacroH2A1.2 PD HP1 PD

SD SD SD

0.149 −0.748 −0.077

Based on estimated marginal means a

Adjustment for multiple comparisons: Bonferroni

Sig.a

0.364 0.685 0.466 0.120 0.395 0.847

Invest New Drugs Table 4 H3 univariate model. H3 acetylation fold changes were compared with SPSS within a univariate general linear model between PD and SD outcomes. The results of this comparison indicated p>0.05 Tests of between-subjects effects Dependent Variable:value_mean Source Type III Sum of Squares Corrected Model 4.434a Intercept 201.139 Outcome 4.434 Error 45.151 Total 248.501 Corrected Total 49.585 a

df 1 1 1 22 24 23

Mean Square 4.434 201.139 4.434 2.052

20 11 31

1/2 1/0 2/2

8 7 16

9 4 14

Not significant P53 (codon 72) G proline, C arginine, PRc confirmed PR, PRu unconfirmed PR

associated with response (p=0.029, Fishers exact test), but the published cut-point value, 363.8 was not significant for our data.

With 32 evaluable patients the null hypothesis was distinguished from H1 with delta=0.24 with 81.3 % power and 2.51 % type I error. VEGF C1D1 (>203 pg) pre-treatment was significantly Table 5 H4 univariate model. H4 acetylation fold changes were compared with SPSS within a univariate general linear model between PD and SD outcomes. The results of this comparison indicated p>0.05

R Squared=0.080 (Adjusted R Squared=0.038)

PD

0.156 0.000 0.156

VEGF or FGF levels as predictors of response to therapy

a

SD

2.160 98.005 2.160

The Fisher’s exact test was applied to the classification polymorphism (+ or -) and response (+ or -) and progression > or