Combination decitabine, arsenic trioxide, and ascorbic acid for the treatment of myelodysplastic syndrome and acute myeloid leukemia: A phase I study

Abstract

We conducted a phase I study to assess the safety and efficacy of decitabine in combination with arsenic trioxide and ascorbic acid for patients with myelodysplastic syndrome (MDS) and acute myeloid leukemia (AML). Patients received fixed doses of decitabine (20 mg/m2/day intravenous (IV) × 5 days every 28 days) and ascorbic acid (1,000 mg IV following each dose of arsenic trioxide). Arsenic trioxide was escalated in three dose cohorts (0.1 mg/kg/day × 5 days followed by 0.1 mg/kg IV weekly; 0.2 mg/kg/day × 5 days followed by 0.2 mg/kg IV weekly; and 0.3 mg/kg/day × 5 days followed by 0.3 mg/kg IV weekly). We enrolled 13 patients and identified 0.2 mg/kg as the maximum tolerated dose of arsenic in combination with decitabine and ascorbic acid. We observed one morphologic complete remission with incomplete blood count recovery (CRi) and five patients with stable disease (SD) after four cycles of therapy. We evaluated the effect of this therapy on angiogenesis in vivo. Unexpectedly, we observed increased microvessel density following two cycles with no consistent effect on angiogenic mRNA expression. These results identify an appropriate dose of combination decitabine, arsenic trioxide and ascorbic acid for Phase II studies. DNA hypomethylating agents have activity in both myelodysplastic syndrome (MDS) and acute myeloid leukemia (AML), although response rates have been modest (in AML CR rates are ∼ 26%) [1-3]. Arsenic trioxide has response rates in MDS of 19%–27% [4, 5]. The mechanism of action remains unclear and current models include antiangiogenic activity, modulation of apoptosis via increased oxidative stress, and induction of differentiation [6-13]. Ascorbic acid increases the apoptotic effect of arsenic trioxide on AML cells in vitro and can be safely combined in vivo with CR/CRi observed in 2 of 11 AML [6, 14]. We therefore undertook to assess the feasibility of combining decitabine, arsenic trioxide, and ascorbic acid. We selected a well characterized dosing schedule of decitabine and an arsenic schedule similar to Vey et al., but with a weekly maintenance schedule [2, 4, 15, 16]. Ascorbic acid was administrated intravenously to avoid decreased gastrointestinal absorption by arsenic [17]. Thirteen patients were enrolled into three cohorts (Table I). Eight were male and five were female. Median age was 67 (range, 24–77). Five had MDS and 8 had AML. Seven had cytogenetic abnormalities including: −7, +8, +19, inv [3], and complex cytogenetics. Eleven had an ECOG performance score of 0 or 1. Cytopenias and transfusion dependency were common (10 had an ANC <1,500 mm3, nine were RBC transfusion dependent, and six were platelet transfusion dependent). Prior therapy was common (five had received prior salvage chemotherapy for AML, six had received hypomethylating agents, three had previously undergone an allogenic stem cell transplant, and only five had received no prior therapy). In general, the combination was well tolerated, with the expected hematologic and infectious complications that make treatment of this patient population challenging. Dose limiting toxicities (DLTs) were assessed only during Cycle 1. One DLT, Grade 3 QTc prolongation, occurred in Cohort 2, and this cohort was expanded to six patients (Table II). This patient had a baseline QTc of 460 msec. On Day 7, her QTc was 508 msec. After correcting electrolytes and delaying therapy for two days, her QTc improved to 467 msec and she continued therapy until developing progressive disease (PD) during Cycle 2. There were no further DLTs in this cohort. In Cohort 3, two patients developed sepsis with hypotension and hypoxia. Both patients were admitted to the intensive care unit and both died. Both had prolonged disease-related neutropenia prior to enrollment on this trial. Given the severity of these outcomes, both were deemed DLTs. In total, four patients discontinued therapy due to toxicities during any cycle (Cohort 2: 1 discontinued therapy due to elevated ALT; Cohort 3: 2 discontinued therapy due to death from infectious complications and 1 due to death from PD). Other common toxicities observed were hyperglycemia, hyponatriemia, and QTc elongation. We did not observe any symptomatic arrhythmias in patients with prolonged QTc, consistent with other observations of patients treated with arsenic trioxide [18, 19]. Across all cohorts, patients remained on study for an average of 67 days (Table I). Patients in Cohorts 1 and 2 completed an average of 2.9 cycles and remained on study an average of 80.8 days. Patients in Cohort 3 remained on study for 35 days and only one completed Cycle 2. Patients with MDS remained on study for an average of 85 days and none withdrew due to toxicity (one discontinued therapy due to travel difficulties, two completed four cycles, and two discontinued therapy due to PD after two cycles). One patient achieved complete remission with incomplete blood count recovery (CRi), five had SD, four had PD, and three did not complete Cycle 2 and were not evaluated for disease response. This outcome is similar to the results of a small, Phase I study of patients treated with combination arsenic trioxide and ascorbic acid (1/11 CR, 1/11 CRi) and a Phase II study with single agent arsenic trioxide (0/11 CR) [6, 20]. However, this response rate was lower than has been reported for single agent decitabine [1-3]. The majority of our patients were heavily pretreated and this may have contributed to lower response rates. Prior studies have suggested that arsenic trioxide has antiangiogenic properties and there is emerging data suggesting that angiogenesis may play an important role in MDS and AML [12, 13, 21]. We assessed the microvessel density in core bone marrow biopsies before therapy and after two cycles. Four patients had adequate samples for evaluation. Unexpectedly, we observed an increase in microvessel density in all four patients (Fig. 1A). We identified an additional cohort of five patients who were treated with decitabine alone at Washington University on a separate clinical protocol. We noted no change in microvessel density in four of these patients and an increase in microvessel density in only one patient (Fig. 1B). Microvessel density did not correlate with disease response in either cohort. Thus, the increase in microvessel density cannot be attributed to an effect of disease progression or decitabine therapy alone. Microvessel density. A: Microvessel density in patients before and after two cycles of decitabine, arsenic trioxide, and ascorbic acid. Each line represents the results from one patient and the average result from two independent pathologists. Represented are patients 4, 6, 7, and 9. B: Microvessel density in patients before and after two cycles of decitabine. One of these patients had progressive disease, two had stable disease, and two had a partial response with incomplete count recovery. C: Expression of hematopoietic and angiogenic transcripts in total bone marrow aspirate cells prior to therapy. Total RNA was generated from total bone marrow aspirate cells collected from seven patients prior to initiating therapy. Using the nCounter system, we quantitated the absolute expression of each transcript and normalized this to OAZ1 expression. Three patients had MDS (Patients 3, 4, and 7) and four had AML (Patients 5, 6, 9, and 11). Bars represent the median values ± standard deviation. We used NanoString nCounter technology to measure angiogenic expression patterns before therapy and after two cycles of decitabine, arsenic trioxide, and ascorbic acid. Adequate bone marrow aspirate samples were available from seven patients. We validated this approach by comparing the percentage of myeloid and erythroid cells observed in the clinical bone marrow aspirate differential cell counts at each of these time points and the expression of the following transcripts: CTSG (Cathepsin G: R2 = 0.3, P = 0.04), ELANE (neutrophil elastase: R2 = 0.37, P = 0.02), and GYPA (Ter119: R2 = 0.37, P = 0.02) (Supporting Information Fig. 1A–C). None of the 12 angiogenic transcripts assessed (ANGPT1 (Ang-1), ANGPT3 (Ang-2), FGF2, TGF, TEK (Tie2), TNFA, VEGFA, VEGFB, VEGFC, VEGFD, VEFGR1, VEGFR2) were consistently up-regulated or down-regulated in all seven patients tested (Supporting Information Figs. 2 and 3>). VEGFC was significantly increased in three of seven patients (P < 0.01 for each patient), but the effect was modest (average fold change 2.1, range, 1.9–2.4); TNF was decreased in four patients (P = < 0.01 for each patient), and again the effect was modest (average fold change 0.6, range, 0.4–0.78) (Supporting Information Figs. 2F and 3C). In contrast to Keith et al., [13] none of the 12 angiogenic transcripts assessed were consistently elevated or consistently decreased in patients with MDS relative to those with AML. In contrast to Stifter et al., [22] we did not observe a correlation between TNF expression and erythroid hematopoiesis (Supporting Information Fig. 1D). However, their comparison was between hemoglobin levels and TNF expression, while we have compared TNF expression with bone marrow erythroid precursor numbers. Four of our seven patients were red blood cell transfusion dependent and so the comparison of hemoglobin levels would not have been meaningful. However, TNF levels did not correlate with transfusion requirements. NanoString nCounter technology has the advantage of digitally quantitating the number of transcripts within a pool of total RNA. This allows accurate comparison of the relative expression levels of two different transcripts. We observed that angiogenic transcripts are generally expressed at very low levels, with only TGFB, VEGFA, TNF, and ANGPT1 expressed at comparable levels to CD34, KIT, POU2F1, GATA1, or FLT3 (Fig. 1C). Our expression profiling did not illuminate an obvious candidate transcript that might explain the increased microvessel density. It is possible that this effect is related to transcriptional changes not assessed with this candidate gene strategy, or that arsenic trioxide alters the expression of one of these transcripts in bone marrow stromal cells (which would be underrepresented, if not absent, during bone marrow aspirate collection). It is also possible that greater transcriptional changes might have been seen if there had been more clinical responses. However, the patient with the best clinical response (Patient 6, who achieved CRi) displayed no statistical up-regulation or down-regulation in 11 of the 12 angiogenic transcripts that were measured (TGF was statistically down regulated, but only 0.6-fold) and expression of none of the 12 transcripts correlated with changes in the percentage of bone marrow blasts. Response rates to single agent decitabine have been modest, with CR rates of ∼ 26% [1-3]. Some success has been seen when combining decitabine with valproic acid or with gemtuzumab, and combining azacitidine with lenalidomide, although this work has been limited to single-arm, Phase I studies [23-25]. The combination of decitabine, arsenic trioxide, and ascorbic acid appears to be reasonably well tolerated, although response rates were low in this heavily pretreated population. This study has now identified a dose that is safe for further exploration in a Phase II study and in comparison with other combinations of targeted chemotherapy and hypomethylating agents. However, it is possible that a higher dose of arsenic may be tolerated if the accrual were limited to de novo patients. All patients signed an informed consent form for a Phase I trial approved by the Institutional Review Board at Washington University School of Medicine. Eligible MDS patients required therapy due to blast counts (>5%) or cytopenias, and had not received more than four cycles of a hypomethylating agent. AML patients were enrolled if they were either older than 60 years old or had relapsed disease following salvage chemotherapy and had not received more than four cycles of a hypomethylating agent. Additional inclusion criterion were: ECOG performance status <3; QTc <460 milliseconds; serum potassium >4.0 mEq/L; serum magnesium >1.8 mg/dL; adequate end organ function (creatinine and total bilirubin ≤1.5x institutional upper limit of normal, AST and ALT ≤2x institutional upper limit of normal), absence of intercurrent illness such as radiotherapy within 14 days of enrollment, hepatic tumors, HIV, uncontrolled heart failure, unstable angina pectoris, cardiac arrhythmia, or psychiatric illness that precluded appropriate informed consent. Patients were excluded for a history of glucose-6-phosphatedehydrogenase (G6PD) deficiency, pregnancy, or the inability to utilize appropriate contraception while being treated with chemotherapy. Transfusions and prophylactic antibiotics were administered according to institutional standards. Erythropoietin was not administered during the study, and G-CSF was only administered in the setting of neutropenic sepsis. The study design was a standard 3 + 3 algorithm to evaluate dose tolerability. The arsenic trioxide dose was determined using a modified Fibonacci dose escalation, with a minimum of three evaluable patients at each dose level, defined as patients who completed Cycle 1 and did not require therapy discontinuation for infection or progressive disease. The maximum tolerated dose (MTD) was defined as the dose level immediately below the dose level at which one patient of a cohort of three patients (or two of a cohort of six patients) experience dose-limiting toxicity during the first cycle. This protocol design was approved by the IRB at Washington University School of Medicine and registered at clinicaltrials.gov (NCT00671697). Cytopenias alone were not considered a DLT. Hematologic DLT was defined as persistent bone marrow aplasia with ≤20% cellularity, which persisted for >8 weeks despite dose modifications. Nonhematologic DLT was defined as any Grade 3 or Grade 4 nonhematologic toxicity that occurred during the first cycle with the specific exceptions of nausea, vomiting, anorexia, weight loss, infections, or electrolyte abnormalities attributable to any other cause. A DLT was also considered for a delay of treatment of 14 or more days (consecutive or cumulative) during the first cycle not attributable to disease progression. We used this approach to avoid unnecessary halting of dose escalation for toxicities that are considered acceptable in Phase 1 trials of drugs in aggressive hematologic malignancies [26]. Hematological and non-hematological toxicities were graded according to the revised National Cancer Institute (NCI) Common Terminology Criteria for Adverse Events (CTCAE) version 3.0, published December 12, 2003 and available at: http://ctep.cancer.gov/reporting/ctc.html. Because of concern for QTc prolongation with arsenic trioxide, all patients were screened for a list of medications that have been associated with prolonged QTc. Bone marrow core and aspirate were reviewed by the Division of Hematopathology at Washington University School of Medicine and interpreted according to the World Health Organization Classification. Estimation of microvessel density was done by two independent observers (AH, JMK) as previously described [27]. In brief, immunohistochemistry was performed with a prediluted monoclonal antibody (anti-CD34) on four-micron thick sections of formalin-fixed, paraffin-embedded tissue from bone marrow core biopsies using an automated slide preparation system and Enhanced DAB Detection Kit (Benchmark XT, Ventana, Tucson, AZ). Positive and negative control slides were performed. Observers were blinded to patient name and treatment phase. Total bone marrow aspirate cells were cryopreserved in 10% DMSO following lysis of red blood cells in ACK lysis buffer (0.15 M NH4Cl, 10 mM KHC03, 0.1 mM Na2EDTA). Cryovials were quickly thawed at 37°, washed in phosphate buffered saline with 50 μM β-mercaptoethanol and 10% fetal bovine serum, and lysed in Trizol reagent (Invitrogen, Carlsbad, CA). Total RNA was prepared per manufacturers recommendations and qualitatively assessed using an Experion Bioanalyzer (Hercules, CA). The nCounter Analysis System (NanoString Technologies) was used as reported previously and per manufacturer's recommendations [28, 29]. In brief, each sample was hybridized in triplicate with 250 ng of total RNA in each reaction. All genes and controls were assayed simultaneously in multiplexed reactions. To account for slight differences in hybridization and purification efficiency, the raw data were normalized to the standard curve generated via the nCounter system spike-in controls present in all reactions and the data was normalized to OAZ1 expression, although similar results were seen when normalizing to GAPDH [30]. Linear regression and goodness of fit were performed in Prism (Graphpad Software, San Diego, CA version 5.04). Paired t-test was performed using Excel (Microsoft, Seattle, WA version 2007). Additional Supporting Information may be found in the online version of this article. Please note: The publisher is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article. John S. Welch*, Jeffery M. Klco , Feng Gao , Elizabeth Procknow*, Geoffery L. Uy*, Keith E. Stockerl-Goldstein*, Camille N. Abboud*, Peter Westervelt*, John F. DiPersio*, Anjum Hassan , Amanda F. Cashen*, Ravi Vij [e-mail]*, * Division of Oncology, Department of Medicine, Washington University School of Medicine, St. Louis, Missouri, Pathology and Immunology, Washington University School of Medicine, St. Louis, Missouri, Division of Biostatistics, Washington University School of Medicine, St. Louis, Missouri.