A Phase II Trial of Guadecitabine in Children and Adults with SDH-Deficient GIST, Pheochromocytoma, Paraganglioma, and HLRCC-Associated Renal Cell Carcinoma

Loss of the succinate dehydrogenase (SDH) complex or fumarate hydratase leads to the accumulation of the Krebs cycle metabolites succinate and fumarate, inhibiting α-ketoglutarate-dependent dioxygenases (1). SDH deficiency (dSDH) has now been identified in 88% of patients with pediatric gastrointestinal stromal tumors (GIST) and those tumors that do not present KIT or platelet-derived growth factor receptor alpha (PDGFRA) mutations, which are often described as “wild-type” GIST (2). GISTs are the most common sarcomas of the gastrointestinal tract (3–5). These dSDH GIST are generally resistant to targeted therapy with tyrosine kinase inhibitors such as imatinib (6–8), as well as chemotherapy and radiotherapy (9). Thus, novel agents are urgently needed for these deadly tumors. dSDH has also been identified in 10% to 30% of patients with pheochromocytoma and paraganglioma (10–12), and similarly, abnormalities in fumarate hydratase have been detected in 93% of patients with hereditary leiomyomatosis and renal cell cancer–associated renal cell carcinoma (HLRCC-RCC; refs. 13–15).

The specific mechanism or mechanisms of tumorigenesis due to dSDH is not known, and until recently a lack of preclinical models has hindered development of new treatments for these tumors. Prior attempts to target the metabolic derangements present in dSDH GIST with the tyrosine kinase inhibitor vandetanib (7) or the IGF-1R inhibitor linsitinib (8) were successfully completed but did not demonstrate clinical activity. However, α-ketoglutarate-dependent dioxygenases are important regulators of DNA demethylation, and consequently dSDH-mediated increases in succinate and fumarate may lead to global DNA hypermethylation. The potential relevance of DNA hypermethylation in dSDH GIST is supported by global epigenetic dysregulation (16) and decreased levels of the downstream Krebs cycle intermediate 5-hydroxymethylcytosine (17) found in clinical samples. This suggests a potential therapeutic role for agents, which can reverse DNA hypermethylation in dSDH tumors. Guadecitabine (SGI-110), a dinucleotide containing the small molecule inhibitor decitabine, functions as a DNA methyltransferase inhibitor. Guadecitabine has entered phase I and II clinical trials in acute myeloid leukemia/myelodysplastic syndrome (18, 19), hepatocellular carcinoma (20), and ovarian cancer (21, 22). We hypothesized that guadecitabine would have activity in dSDH tumors including GIST, pheochromocytoma/paraganglioma, and HLRCC-RCC, and tested this hypothesis in a small phase II study.

Patients ≥12 years of age with measurable disease of histologically-confirmed recurrent or refractory/unresectable dSDH: (i) GIST, (ii) pheochromocytoma/paraganglioma, or (iii) HLRCC-RCC. For GIST and HLRCC-RCC, previously untreated patients with metastatic and/or unresectable disease were eligible. For pheochromocytoma/paraganglioma, progression following ¹³¹I-meta-iodobenzylguanidine (MIBG) in patients with MIBG avid tumors or cytotoxic chemotherapy (dacarbazine or temozolomide) was required prior to enrollment on this trial unless the patient refused or the investigator deemed being treated on this trial prior to cytotoxic chemotherapy was in the best interest of the patient. Other eligibility criteria included recovery from toxic effects of prior therapy to grade 1; Eastern Cooperative Oncology Group (ECOG) performance status ≤2, or Lansky/Karnofsky score ≥60%; interval from prior therapy ≥4 weeks from prior surgical procedures with complete healing of surgical sites; ≥28 days from a last dose of cytotoxic chemotherapy, mAb, or any investigational agent; ≥28 days from prior biological therapy including biological response modifiers (e.g., cytokines) immunomodulatory agents, vaccines, and differentiating agent; and ≥4 weeks from external beam radiotherapy. Patients were required to have normal organ and marrow function including adequate renal function (age-adjusted normal serum creatinine, or a creatinine clearance ≥60 mL/min/1.73 m²) and adequate liver function (total bilirubin within normal institutional limits, alanine aminotransferase, and aspartate aminotransferase ≤2.5× institutional upper limit of normal). Adequate bone marrow function was required and defined as an absolute neutrophil count ≥1,500/μL and transfusion-independent platelet count of ≥100,000/μL.

This trial conformed to the Declaration of Helsinki and Good Clinical Practice guidelines and was approved by the NCI's Institutional Review Board. Investigators obtained written informed consent from all patients or their legal guardians, indicating their understanding of the investigational nature and risks of this study. Assent was obtained according to institutional guidelines.

The study (NCT03165721) was conducted as a single-site Simon optimal two-stage phase II trial in three strata of patients with one of the three eligible diagnoses with the primary objective to assess clinical activity (complete response or partial response) of guadecitabine to rule out an unacceptably low overall response rate of 5% (p0 = 0.05), in favor of a response rate of 30% (p1 = 0.30). With α = 0.10 and β = 0.10, if ≥1 of 7 patients in stage 1 responded in a given stratum, enrollment would be expanded in that stratum to 21 patients, and if ≥3 of 21 responded, guadecitabine would be considered active in that stratum. Guadecitabine was supplied to the NCI (sponsor: Cancer Therapy Evaluation Program) by Astex Pharmaceuticals, Inc. and was administered under an investigator-held Investigational New Drug application. Guadecitabine was administered subcutaneously at a dose of 45 mg/m²/day for 5 consecutive days on a planned 28-day cycle.

Monitoring for guadecitabine-related toxicity included physical examination with vital sign measurements (including observation in clinic following initial injection of guadecitabine) as well as complete blood counts with differential, serum chemistries including electrolytes, calcium, phosphate, magnesium, creatinine, glucose, blood urea nitrogen, albumin, aspartate aminotransferase, alanine aminotransferase, total bilirubin, total protein, creatine phosphokinase, and urinalysis at baseline and prior to each subsequent cycle. Pregnancy testing was done at baseline and prior to each cycle in female patients of childbearing potential. Adverse events were graded according to the Common Terminology Criteria for Adverse Events Version 5.0.

Response was evaluated using RECIST guideline v1.1 (23) at baseline and prior to cycle 2 and then prior to every other cycle. Assessment of disease was performed using radiologic evaluation which could include computerized tomography and/or magnetic resonance imaging, and/or 2[18F]fluoro-2-deoxy-D-glucose-PET as clinically indicated. A consistent method of disease evaluation was used for each patient throughout the study.

Any toxicity of grade 3 or higher that could not be resolved/reversed within 72 hours with supportive care was considered a TLT, as was any persistent grade 2 toxicity that was intolerable to the patient. Guadecitabine was held for TLT until resolution to grade 1 or baseline, and then could be resumed with up to two dose reductions (approximately 30% each, 30 mg/m²/dose at dose level −1 and 20 mg/m²/dose at dose level −2).

Health-related quality of life was evaluated by patient and parent report (for pediatric patients) using the distress thermometer (DT) and patient-reported outcomes measurement information system (PROMIS) short-form measures (anxiety, depression, fatigue, pain interference, and physical function). The measures were administered to all consenting English- and Spanish-speaking patients with parallel parent proxy instruments administered to parents of patients ages 12 to 17. These instruments were administered at baseline, after cycle 4 and after each subsequent fourth cycle, as well as at the time a patient was taken off treatment.

Decitabine was extracted from heparinized plasma samples using solid-phase extraction, then analyzed using an LC/MS-MS assay with a calibrated range of 5 to 1,000 ng/mL. Noncompartmental analysis was performed using Phoenix WinNonlin, v8.2 (Certara; RRID:SCR_021370), validated per FDA 21 CFR Part 11 regulations, to estimate pharmacokinetic parameters. Specifically, the maximum plasma concentration (C_MAX) was recorded as observed values, as was the time to C_MAX (T_MAX). The area under the plasma concentration versus time curve (AUC) to the last time point was calculated using the linear up/log down trapezoidal rule. If possible, with sufficient terminal data, an elimination rate was calculated, from which half-life (t_1/2), AUC to time infinity, apparent clearance (CL/F), and volume of distribution (V_z/F) were calculated.

Peripheral blood mononuclear cell (PBMC) global methylation was assessed by whole-genome bisulfate sequencing (24) at baseline and on day 14 of cycle 1. Briefly, peripheral blood was collected in ethylenediaminetetraacetic acid tubes, centrifuged, and then the buffy coat was resuspended in phosphate-buffered saline and stored at −20°C until further processing. Subsequently, the buffy coat was resuspended in phosphate-buffered saline for germline DNA extraction using the DNeasy Blood and Tissue Kit (Qiagen). One hundred nanograms of per sample of DNA was fragmented to a target length of 350 base pair and bisulfate converted using EZ DNA Methylation-Gold Kit (Zymo Research). Sequencing libraries were constructed from bisulfate converted DNA using Swift Accel-NGS Methyl-Seq (Swift Biosciences) per the manufacturer's protocol. Libraries were then pooled and sequenced to a goal depth of 30× using 150 base pair paired-end reads on a NovaSeq platform (Illumina) and reads were trimmed to remove low-quality bases and adapters using trimgalore (25) v0.6.6. Bismark was used to align reads to GRCh38, demultiplex, and extract methylation calls (26). Data were smoothed across nearby CpG sites and global methylation was calculated with bsseq v1.26.0 (27). To identify differentially methylated regions, CpGs with less than 5× coverage in all samples were removed to decrease false positives, t statistics were calculated and thresholded using a cutoff of 4.6. Differentially methylated regions were filtered to include at least 5 CpGs and a mean difference of at least 0.2. Statistical analyses were performed using R v.4.0.3.

Serum and urine samples were collected at baseline and on days 7, 14, and 28 of cycle 1 to explore the metabolomic profile of patients with tumors related to Krebs cycle defects. Isotope dilution LC/MS was performed to measure the concentrations of metabolites in several key cellular metabolic pathways. For detailed description, see Supplementary Materials and Methods. In brief, all reference target compounds and isotopic standards were purchased from Sigma-Aldrich, Cambridge Isotope Laboratory, Medical Isotopes, Inc., and CDN Isotopic Inc. (Supplementary Methods Table S1). Urine samples were processed by dilution in water, except for oxidation and phosphorylation metabolites, which were diluted with Tris buffer supplemented with internal standards to optimize pH. Serum samples were extracted in either methanol (central carbon pathway), percholic acid (electron transfer metabolites), or diluted in water (pyruvate and short-chain fatty acids) and the extraction solutions were supplemented with appropriate isotopic standards. Reversed-phase ion-pairing LC/MS² analysis was performed using a Thermo TSQ triple quadrupole mass spectrometers (Thermo Fisher Scientific) coupled to either Shimadzu 20AC-XR or Vanquish (Thermo Fisher Scientific) liquid chromatographic system for the central carbon pathway and electron transfer metabolites respectively. Pyruvate and short-chain fatty acids were analyzed by reveres phase LC/MS² on Thermo TSQ triple quadrupole mass spectrometers (Thermo Fisher Scientific) coupled to Shimadzu or 20AC-XR HPLC system. The mass spectrometers were operated in either negative or positive mode and set to monitor parent-product ion transitions using Selected Reaction Monitoring (Supplementary Methods Table S2). Quantitation of targeted metabolites was carried out using Xcalibur Quan Browser (Thermo Scientific). Calibration curves for each metabolite were constructed by plotting reference compound/isotopic peak area ratios obtained from the calibration standards curve and fitting the data using linear regression with 1/X weighting. The analyte concentrations in samples were then interpolated using the linear function obtained from the calibration curve.

Raw data are available as a supplementary table and upon reasonable request.

Nine adult patients (3 male and 6 female; median age, 36 years; range, 18–58) were enrolled from August 16, 2017, to March 12, 2019. The race/ethnicity of patients on study included: white/not Hispanic or Latino (6), White/Hispanic or Latino (1), Black or African American/not Hispanic or Latino (1), and Asian (1). Table 1 provides a summary of demographic, clinical, and baseline disease characteristics. Supplementary Table S1 discusses the representativeness of study participants. All patients had dSDH tumors as determined by genomic analysis (one wild-type GIST patient was confirmed to be negative for KIT and PDGFRA mutations). Seven patients had dSDH GIST, 1 patient had paraganglioma, and 1 patient had HLRCC-RCC. The median number of lines of prior therapy was 2 (range 0–6). One patient was ECOG 0, whereas 8 patients were ECOG 1. All patients had evidence of metastatic disease.

Two patients developed TLT of severe neutropenia (one grade 4, one grade 3) requiring dose reduction following cycle 1, and subsequently tolerated treatment. One additional patient developed grade 2 lung infection requiring hospitalization that delayed the start of cycle 2 (Table 2). No partial or complete responses were observed, and the best response was stable disease (median number of completed cycles, 7; range, 1–12). Three of seven patients with GIST, the patient with paraganglioma and the patient with HLRCC-RCC were taken off study for progressive disease. Among patients with stable disease, 3 of 7 patients with GIST were taken off study for patient preference. Although not considered a TLT, 1 patient with GIST was taken off study at investigator discretion due to a combination of unexpected toxicity (posterior reversible encephalopathy syndrome, PRES—reversible and not previously reported in trials of guadecitabine, attributed by sponsor as not related to guadecitabine) and interval surgical tumor debulking.

Because there were no partial or complete responses observed in the first 7 patients with GIST, the GIST stratum was terminated after stage 1 on February 24, 2020. Concurrently, the pheochromocytoma/paraganglioma and HLRCC-RCC stratums were terminated due to low accrual.

Six of seven patients with GIST continued therapy for at least five cycles (Table 1) with time to progression greater than 6 months (Table 3). The 3 patients with stable disease who discontinued therapy for patient preference completed at least 12 cycles and their time to progression was at least 10 months. The patient taken off therapy for investigator discretion (patient 4) had surgical debulking. All four patients who had stable disease when they were taken off therapy were censored at that time. For the GIST stratum (n = 7), using the Kaplan–Meier method as implemented in SAS version 9.4, progression-free survival (PFS) censoring at the dates the patients came off treatment and overall survival (OS) at 12 months were 53.6% [95% confidence interval (CI; Fig. 1A), 13.2%–82.5%] and 100% (Fig. 1B), respectively. Median PFS was not reached, and median OS was 38.2 months (CI, not estimable).

The outcomes for the 1 patient in the pheochromocytoma/paraganglioma stratum (patient 3) and the 1 patient in the HLRCC-RCC stratum (patient 8) are presented in Table 3.

All 9 patients completed the baseline PROMIS and DT measures with 7 of 8 completing evaluation at the end of cycle 4 (one patient off treatment prior to end of cycle 4) and 8 of 9 completing the end of therapy evaluation. Although there was a trend for patients to report less pain at baseline compared with end of treatment on both the DT and the PROMIS pain interference measure, the small number of patients and lack of responses in any stratum precluded evaluation of significant changes in the DT or in the t scores over time in any of the PROMIS domains.

All 9 patients had sufficient pharmacokinetic sampling and were thus included in this analysis. The median absolute dose was 80.7 mg, based on a median body surface area of 1.79 m². Following a subcutaneous dose of guadecitabine, decitabine plasma concentrations demonstrated an approximate 30 minute lag time prior to rising to quantifiable levels. Decitabine concentrations peaked on average at 4 hours, consistent with the desired formulation and route of administration. Summarized exposure parameters are presented in Table 4. Only 2 of 9 patients had sufficient data to estimate an elimination rate. Given that t_1/2, V_z/F, and CL/F are derived from measure, these parameters were only reported for the following two patients: t_1/2 = 1.73 hours, 1.35 hours; V_z/F = 1,127 L, 1,147 L; and CL/F = 452 L/hour, 588 L/hour.

Two patients had sufficient samples collected to perform methylation analysis; the other seven patients did not return for collection on day 14. Both patients, patient 2 (GIST, stable disease but taken off treatment for patient preference after 17 cycles) and patient 3 (paraganglioma, progressive disease after 3 cycles) demonstrated significant global demethylation on cycle 1 day 14 compared with baseline (−42.9%, −47.0%, Supplementary Fig. S1). In all, 1,734 differentially methylated regions with at least five CpGs, and a mean difference of greater than 20% were identified between day 14 and baseline white blood cells (Supplementary Table S2).

Metabolites from the glycolytic and pentose phosphate pathway and the tricarboxylic acid cycle, as well as oxidation and phosphorylation metabolites and short chain fatty acids were evaluated in serum and urine at baseline and on days 7, 14, and 28 (±3 days, when available) from all 9 patients. No significant changes in any of these metabolites could be detected consistently across patients, and the lack of responses precluded correlation with clinical activity (Supplementary Figs. S2–S9). Metabolites not displayed in Supplementary figures were below the limit of detection at each measured timepoint.

Although tyrosine kinase inhibitors such as imatinib have been transformative in the treatment of KIT and PDGFRA mutated GIST (28, 29), no active targeted therapies have been identified in the subset of patients with dSDH GIST. Several small studies using various tyrosine kinase inhibitors have not demonstrated responses (7, 30–33), highlighting the need for novel agents for these patients. Given the global hypermethylation observed in patients with dSDH GIST (16), we hypothesized that targeting this pathway, common among dSDH tumors, with guadecitabine could result in therapeutic benefit. Unfortunately, guadecitabine did not achieve the target response rate of 30% in dSDH GIST. No definitive conclusions can be drawn in dSDH pheochromocytoma/paraganglioma or HLRCC-RCC as only one patient was enrolled in each of these strata, and larger studies perhaps through a cooperative group or consortium may be necessary to truly determine the activity of guadecitabine in these rare tumors. Nevertheless, we were able to gain experience with this novel therapy and to evaluate meaningful biologic correlates, which could be used in future trials.

Interestingly, from a pharmacokinetic standpoint the active metabolite, decitabine, achieved higher C_max than was previously reported in the phase I trial of this dose (45 ng/mL compared with 26.3 ng/mL), with similar elimination rates (t_1/2 = 1.54 hours compared with 1.25–1.3 hours; ref. 22). The reason for this may be multifactorial: (i) our trial used guadecitabine as a single-agent rather than in combination with carboplatin; (ii) specificities related to the underlying diseases; or (iii) C_max was only evaluated in a small number of patients in both trials. Our pharmacodynamic analyses suggest that guadecitabine was able to execute its mechanism of action as demonstrated by relative global demethylation in PBMCs on day 14. The lack of responses may be due to: (i) inadequate tumor penetration of guadecitabine; (ii) inability of guadecitabine to perform its demethylating action within the tumor; (iii) while global hypermethylation is seemingly relevant in tumorigenesis for dSDH tumors, reversal may be insufficient to cause regression of an already-established tumor; (iv) while global demethylation appears to occur, it may be that the target genes where hypermethylation drives tumorigenesis (which are unknown) were not demethylated. On-treatment tumor samples would be valuable in determining the action of guadecitabine within the tumor. Furthermore, although we determined that it was feasible to collect serum and urine metabolites over the course of the first month of a small phase II clinical trial in this patient population, we were unable to clearly show that guadecitabine consistently altered the metabolites present in patients with dSDH tumors over 1 month of therapy.

Toxicities of guadecitabine observed in other clinical trials (18–22), predominantly hematologic, were also observed in our trial and required a dose reduction or delay in 3 of 9 patients. An additional patient developed a severe unexpected toxicity of PRES, which was not attributed to guadecitabine, but did contribute to that patient being taken off study. This unique toxicity has not previously been reported in trials of guadecitabine.

This work also illustrates some of the challenges in evaluating new therapies for rare cancers where more indolent behavior can occur. Although several patients had a prolonged period of stable disease with a median duration of therapy of seven cycles [compared, for example, with a median number of four cycles (7) or median duration on treatment of 7.7 months (8) on previous studies], it is difficult to clearly attribute this to guadecitabine, especially given the indolent nature of dSDH GIST and the small study size. A more complete understanding of the natural history of patients with SDH-deficient GIST may help to facilitate the design of future trials. No improvement in quality of life was identified. Our study highlights the difficulty of developing effective therapies in a relatively indolent and rare disease with few relevant preclinical models. Yebra and colleagues reported the derivation of in vitro patient-derived dSDH GIST models and showed a response to the alkylating agent temozolomide in these studies (34). Recently, Flavahan and colleagues identified methylation in insulator regions for fibroblast growth factor 4 (FGF4) and KIT followed by the establishment of a patient-derived xenograft model, which was sensitive to combined fibroblast growth factor receptor (FGFR) and KIT inhibition (35). The lack of activity of guadecitabine does not exclude the possibility that other agents acting on the hypermethylation state of the tumor could be effective. However, such studies highlight the importance of preclinical models, and in this case, suggests that specific methylation of FGF4 and KIT-insulator regions may be more relevant and targetable than global hypermethylation. Recurrent gene duplication of FGF4 has also been reported in “quadruple wild-type GIST” (36), suggesting that FGFR inhibition may be a common targetable pathway for these two different subtypes of GIST. These preclinical studies suggest that FGFR inhibition should be further studied in early-phase clinical trials and highlight the importance of developing clinically relevant preclinical models of dSDH GIST.

In summary, although we were able to demonstrate biologic activity of guadecitabine in terms of global demethylation on PBMCs in a subset of patients, this was insufficient to result in clinical disease regression of dSDH tumors. Larger studies in diseases, which respond to guadecitabine are needed to discover whether global demethylation in PBMCs as well as changes in serum and urine metabolite are relevant biomarkers for clinical outcomes. Several avenues of investigation remain open for new treatment modalities for patients with these difficult to treat dSDH tumors.

This trial conformed to the Declaration of Helsinki and Good Clinical Practice guidelines and was approved by the NCI's Institutional Review Board. Investigators obtained written consent from all patients or their legal guardians indicating their understanding of the investigational nature and risks of this study. Assent was obtained according to institutional guidelines.

All authors consent for publication.

J.A. Ligon reports grants from NIH Intramural Research Program, NCI Center for Cancer Research, and NCI CTEP and other support from Astex Pharmaceuticals during the conduct of the study as well as personal fees from NCI outside the submitted work. F.I. Arnaldez reports other support from MacroGenics, Inc., and AstraZeneca, PLC outside the submitted work. L. Wiener reports other support from NIH Intramural Program during the conduct of the study. R. Srinivasan reports other support from Merck, Novartis, and Genentech Roche outside the submitted work. J.K. Killian reports personal fees from Foundation Medicine outside the submitted work. L.J. Helman reports personal fees from SpringWorks and OncoHeros outside the submitted work as well as ownership of AstraZeneca stock. J. Glod reports other support from Celgene and Apollomics outside the submitted work. No disclosures were reported by the other authors.

J.A. Ligon: Resources, data curation, formal analysis, investigation, writing–original draft, project administration, writing–review and editing. R.T. Sundby: Formal analysis, methodology, writing–review and editing. M.F. Wedekind: Data curation, formal analysis, writing–review and editing. F.I. Arnaldez: Conceptualization, data curation, writing–review and editing. J. Del Rivero: Data curation, writing–review and editing. L. Wiener: Conceptualization, data curation, formal analysis, investigation, writing–review and editing. R. Srinivasan: Conceptualization, data curation, writing–review and editing. M. Spencer: Data curation, writing–review and editing. A. Carbonell: Data curation, writing–review and editing. H. Lei: Formal analysis, writing–review and editing. J. Shern: Formal analysis, supervision, writing–review and editing. S.M. Steinberg: Conceptualization, formal analysis, writing–review and editing. W.D. Figg: Conceptualization, resources, formal analysis, supervision, writing–review and editing. C.J. Peer: Conceptualization, formal analysis, writing–review and editing. S. Zimmerman: Formal analysis, writing-review and editing. J. Moraly: Formal analysis, writing–review and editing. X. Xu: Formal analysis, writing–review and editing. S. Fox: Formal analysis, writing–review and editing. K. Chan: Formal analysis, writing–review and editing. M.I. Barbato: Formal analysis, methodology, writing–review and editing. T. Andresson: Formal analysis, supervision, writing–review and editing. N. Taylor: Formal analysis, supervision, writing–review and editing. K. Pacak: Conceptualization, resources, supervision, investigation, writing–review and editing. J.K. Killian: Conceptualization, formal analysis, writing–review and editing. E. Dombi: Data curation, formal analysis, writing–review and editing. W.M. Linehan: Conceptualization, supervision, writing–review and editing. M. Miettinen: Conceptualization, data curation, supervision, writing–review and editing. R. Piekarz: Conceptualization, resources, data curation, supervision, funding acquisition, project administration, writing–review and editing. L.J. Helman: Conceptualization, supervision, writing–review and editing. P. Meltzer: Conceptualization, writing–review and editing. B. Widemann: Conceptualization, resources, data curation, supervision, investigation, project administration, writing–review and editing. J. Glod: Conceptualization, resources, data curation, formal analysis, supervision, investigation, methodology, writing–original draft, project administration, writing–review and editing.

This research was supported by the Intramural Research Program of the National Institutes of Health; the Center for Cancer Research, National Cancer Institute (NCI); and the NCI Cancer Therapy Evaluation Program (CTEP). Astex Pharmaceuticals, Inc., provided guadecitabine for this study.

The publication costs of this article were defrayed in part by the payment of publication fees. Therefore, and solely to indicate this fact, this article is hereby marked “advertisement” in accordance with 18 USC section 1734.