A Therapeutic Trial of Decitabine and Vorinostat in Combination with Chemotherapy for Relapsed/Refractory Acute Lymphoblastic Leukemia (ALL)

Abstract

DNA hypermethylation and histone deacetylation are pathways of leukemia resistance. We investigated the tolerability and efficacy of decitabine and vorinostat plus chemotherapy in relapse/refractory acute lymphoblastic leukemia (ALL). Decitabine (15mg/m² iv) and vorinostat (230mg/m² PO div BID) were given days 1-4 followed by vincristine, prednisone, PEG-asparaginase and doxorubicin. Genome wide methylation profiles were performed in 8 matched patient bone marrow (BM) samples taken at day 0 and day 5 (post-decitabine). The median age was 16 (range, 3–54) years. All patients had a prior BM relapse, with five relapsing after allogeneic transplant. The most common non-hematological toxicities possibly related to decitabine or vorinostat were infection with neutropenia (grade 3; n=4) and fever/neutropenia (grade 3, n=4; grade 4, n=1). Of the 13 eligible patients, four achieved complete remission without platelet recovery (CRp), two partial response (PR), one stable disease (SD), one progressive disease (PD), two deaths on study and three patients who did not have end of therapy disease evaluations for an overall response rate of 46.2% (CRp + PR). Following decitabine, significant genome-wide hypo-methylation was observed. Comparison of clinical responders with non-responders identified methylation profiles of clinical and biological relevance. Decitabine and vorinostat followed by re-Induction chemotherapy was tolerable and demonstrated clinical benefit in relapsed patients with ALL. Methylation differences were identified between responders and non-responders indicating inter-patient variation, which could impact clinical outcome. This study was registered at www.clinicaltrials.gov as NCT00882206.

Introduction

Treatment outcomes for children and adults with relapsed or refractory acute lymphoblastic leukemia (ALL) continue to be poor with current salvage therapies often failing in eliminating resistant disease.[1; 2] Although a significant proportion (45-90%) of patients in first relapse can successfully achieve a second complete remission (CR2), most will die of progressive disease.[3; 4; 5; 6] As resistance to chemotherapy is the primary reason for treatment failure, there is an urgent need for clinical trials that test new therapies to may overcome chemotherapy resistance.

One approach to improving clinical outcomes in the setting of relapse might be to apply agents that target mechanisms of drug resistance. In hematological malignancies, histone deacetylation and DNA hypermethylation both play critical roles in gene regulation.[7] These discoveries prompted the design of drug therapies targeting these pathways. Collectively, these drugs are referred to as “epigenetic modifying agents” and include histone deacetylase inhibitors (HDACi) and DNA methyltransferase inhibitors (DNMTi). These drugs act on proteins involved in the wrapping of DNA around histones and the methylation of proximal promoter regions of genes, respectively. Both processes play a critical role in regulating gene expression and frequently these genes are involved in chemotherapy resistance.[8] Decitabine is a cytidine anti-metabolite analogue that acts as a demethylating agent by incorporating itself into cellular DNA and covalently trapping DNA methyltransferase as a protein-DNA adduct.[9] This demethylation process leads to transcription of previously silenced (inactivated) genes. Reactivation of these genes therefore has the potential to suppress cell growth and induce apoptosis.[8] Vorinostat's main mechanism of action is to restore histone acetylation by inhibiting deacetylase.[10] This action may also further reactivate aberrantly silenced genes and allow gene transcription to proceed.

The combination of decitabine and vorinostat have previously been shown to have synergistic in vitro effects in altering signaling pathways that drive neoplastic cell initiation and growth,[11; 12] including B-ALL,[8] and may therefore be more effective in combination than as single agents. Considering these agents are well tolerated in human clinical studies,[13; 14] we hypothesized that decitabine and vorinostat could alter the aberrant gene expression of leukemic blasts, therefore allowing the leukemia cells to become primed for cytotoxic cell killing with subsequent chemotherapy. We performed a pilot study investigating the combination of decitabine and vorinostat followed by standard re-induction chemotherapy (vincristine, prednisone, doxorubicin, PEG-asparaginase) testing the feasibility, tolerability and efficacy in patients with relapse or refractory ALL. Additionally we performed exploratory correlative studies evaluating baseline and post-decitabine genome-wide methylation profiles using the illumina 450K methylation array comparing day 0 and day 5 bone marrow and the decitabine induced methylation profiles between non-responders and responders to better understand the mechanism of decitabine mediated differences on clinical outcome.

Patients and Methods

The Cancer Experimental Therapeutics Initiative (CETI) Program and the Clinical Trials Office at the University of Minnesota supported this clinical trial, which was registered at www.clinicaltrials.gov as NCT00882206. All patients and/or their parents or guardians signed informed consent to participate in this University of Minnesota institutional review board approved therapeutic trial in accordance to the Declaration of Helsinki. Eligible patients were between 0 and 60 years of age and had relapsed or refractory ALL with >5% leukemia blasts in the bone marrow at time of study enrollment. Refractory disease was defined as failure to achieve initial remission after two attempts of standard induction therapy. Patients with central nervous system (CNS) involvement were eligible. Performance status of ≥50% for Karnofsky and Lansky scores were required for ages ≥16 years and <16 years, respectively. Patients who relapsed after allogeneic hematopoietic cell transplantation (HCT) were eligible if at least 3 months had passed and they had no evidence of graft-versus-host-disease (GVHD). Patients were excluded if they were pregnant or lactating, had uncontrolled or progressive infection or if they were receiving valproic acid which has been associated with encephalopathy when used in combination with decitabine.[15] Patients with known allergies to any of the study agents with the exception of PEG-asparaginase, who could be treated with Erwinia L-asparaginase, were excluded.

Fourteen patients were enrolled on this single institution trial between 2009 and 2012. One patient was removed from study and excluded from the analysis due to ineligibility (<5% leukemia blasts in the pre-study bone marrow) and another patient was excluded from the analysis after study removal by the treating physician on day 5 of therapy when the patient was found to have CNS leukemic involvement, but was included in the intention-to-treat analysis for a total of 13 patients. This patient did not have CSF evaluation at study entry as it was only required in symptomatic patients and this patient did not have a prior history of CNS disease or symptoms at study entry. Patient characteristics for the 13 eligible patients are described in Table 1.

Table I. Patient Characteristics^#.

Patient	Age (years)	ALL Subtype	Cytogenetics	Time to Relapse	Prior Salvage Therapy	Prior HCT	Study Response
1	24	B-ALL	Complex	LMR, 2nd relapse	A	No	CRp/MRD-
2	16	B-ALL	9p del	PIF/Refractory*	A, B, C	No	PR
3	37	B-ALL	t(9;22); Ph+	Primary Refractory*	D	No	PR
4	10	B-ALL	Normal	LMR, 1st relapse	-	No	CRp/MRD-
5	3	B-ALL	t(17;19)	EMR, 1st relapse, Refractory*	E, F, G	No	PD
6	16	B-ALL	Complex	EMR, 1st relapse	F	No	CRp/MRD-
7	54	B-ALL	t(4;15)	HCT in CR1	-	Yes, relapsed 6 months post-HCT	Off study
8	12	T-ALL	t(10;11)	EMR, 1st relapse	A, H, B	No	SD
9	18	B-ALL	t(1;19)	LMR, 1st relapse	-	Yes, relapsed day 140	Off study
10	11	B-ALL	t(4;11q23)	EMR, 1st relapse	-	Yes, relapsed day 71	PD
11	27	B-ALL	9p del	HCT in CR1	A, B	Yes, relapsed 9 months post-HCT	CRp/MRD-
12	5	B-ALL	Complex	EMR, 1st relapse	I, J	Yes, relapsed day 36	Toxic death
13	16	B-ALL	9p del	PIF, HCT in CR1	-	Yes, relapsed day 130	PD

^#characteristics of the 13 eligible patients. HCT, allogeneic hematopoietic cell transplantation; LMR, late marrow relapse (≥36 months from diagnosis); EMR, early marrow relapse (<36 months from diagnosis); del, deletion; PIF, Primary Induction Failure; A, vincristine/prednisone/PEG-asparaginase/ daunorubicin (VPLD); B, ifosfamide/etoposide; C, methotrexate/cytarabine; D, Hyper-CVAD with rituximab; E, epratuzumab with VPLD; F, cyclophosphamide/etoposide; G, clofarabine/cytarabine; H, Nelarabine/methotrexate/ PEG-asparaginase/ mercaptopurine; I, clofarabine/ cyclophosphamide/ etoposide; J, topotecan/ vinorelbine/ thiotepa/ gemcitabine/ dexamethasone; CRp, complete remission without platelet recovery; PR partial response; SD, stable disease; PD, progressive disease; MRD-, minimal residual disease negative (<0.01% by flow cytometry);

^*Failed ≥2 attempts of induction therapy.

Treatment

Table 2 outlines the treatment schema. A single course of treatment was permitted which included decitabine (15mg/m² given intravenously (iv) over 1-hour) and vorinostat (230mg/m² divided twice-daily orally) days 1 through 4. The doses of decitabine (15 mg/m²) and vorinostat (230mg/m²) used in this study were chosen based on prior phase I studies identifying the maximum tolerated dose (MTD) for these agents as well as the optimal epigenetic effects reported at these dose levels.[15; 16; 17; 18] Starting on day 5, patients received doxorubicin (60mg/m²), vincristine (1.5mg/m²), prednisone (40mg/m² divided BID), PEG-asparaginase (2,500 IU/m²) and intrathecal cytarabine (70mg) or methotrexate (15mg). Erwinia was given to patients with an allergy to PEG-asparaginase and patients with Philadelphia chromosome-positive (Ph+) ALL were to receive imatinib days 5 through 33 in addition to the protocol therapy.

Table II. Treatment Schema.

Day 1-4	Day 5	Day 12	Day 19	Day 26	Day 33
Decitabine	Prednisone -------------------------------------------------------------------------------
Vorinostat	Vincristine	Vincristine	Vincristine	Vincristine
	Doxorubicin
	PEG-Asp	PEG-Asp	PEG-Asp	PEG-Asp
	IT Ara-C	IT MTX	IT MTX*	IT MTX*	IT MTX

IT Ara-C, intrathecal cytarabine; IT MTX, intrathecal methotrexate;

^*central nervous system (CNS) leukemia positive patients only;

PEG-Asp, Peg-asparaginase

Methylation Analyses

Patient samples of peripheral blood and bone marrow were collected for methylation analyses on days 0 (baseline), 5 (post-epigenetic therapy/pre chemotherapy) and 33 (end of study). The genomic DNA was isolated from bone marrow specimens using the QIAmp DNA mini kit as per the manufacturer's instructions (Qiagen, Valencia, CA). Bone marrow samples were diluted with phosphate-buffered saline and mononuclear cells were isolated by Ficoll-Hypaque density-gradient centrifugation. Genomic DNA (500 ng) was bisulfite converted and assessed for DNA methylation using Infinium Hyman Methylation 450 Bead Chips (Illumina Inc.). Quantitative methylation of 485,577 CpG sites was performed covering 99% of the reference sequence genes, 96% of CpG islands and 92% of CpG island shores (detailed information available at www.illumina.com). Additionally, long interspersed nuclear element (LINE-1) methylation was quantified using quantitative bisulfite pyrosequencing as previously described.[19; 20]

Toxicity and Response Evaluation

The National Cancer Institute's Common Terminology Criteria for Adverse Events (CTCAE) version 3.0 was used for toxicity evaluation. Adverse events (AE) included any symptom, sign, illness or experience, regardless of causality, that developed or worsened in severity during the course of the study. A serious adverse event (SAE) was any AE, occurring at any dose and regardless of causality that either was life threatening, required inpatient hospitalization / prolonged an existing hospitalization, was determined an important medical event by the treating physician or resulted in death. Patients who exhibited signs of disease progression or experienced unacceptable toxicity were removed from study. There were no dose delays or dose reductions of study drugs for hematologic toxicities during re-induction therapy (Day 1 through Day 33); however, prolonged hematopoietic recovery or bone marrow aplasia during the first 42 days were included in the study stopping rules.

Bone marrow evaluations for response determination were performed on day 33 of the study. If the patient's bone marrow was hypocellular and without evidence of normal tri-lineage hematopoiesis, then a bone marrow procedure was repeated on day 42 of the study. The criteria used to determine treatment response included complete remission (CR), defined as <5% leukemic blasts in the bone marrow with adequate hematologic recovery (ANC>1000/mm³ and platelet count >100,000/mm³); complete remission without platelet recovery (CRp), defined as <5% leukemic blasts in the bone marrow with an ANC>1000/mm³ and platelet count between 50,000/mm³ and 100,000/mm³; partial response (PR), defined as a greater than 35% reduction in the bone marrow leukemia blast percentage; stable disease (SD) defined as no significant change in blast count and progressive disease (PD) defined as worsening of blast count.

Statistical Methods

The enrollment goal planned for this study was 16 patients. This sample size was based on the primary outcome of a complete remission (CR or CRp) rate at day 33. The lower limit of the 95% confidence interval (CI) for the overall response rate was pre-determined to be above 30% for the trial to be considered successful. The expected toxic death rate for this study was 10%; however, stopping rules were established if the rate exceeded 15%. Patients were recruited from all age groups up to 60 years. Statistical analysis for this study included a simple proportion and a 95% exact binomial CI for the day 33 CR rate. A secondary outcome of minimal residual disease (MRD) was also evaluated with a proportion and 95% CI. Time to progression-free survival was estimated using the Kaplan-Meier method. A cumulative incidence curve was created for progression with death included as a competing risk.

The minfi[19] package in Bioconductor was used to process Infinium HumanMethlyation450 data in original idat files. To match the distribution of Type I and Type II probes in a sample, SWAN[20], subset-quartile normalization within array, was applied to the array data. Beta values [Beta = M /(M+U+100)], where M is a methylated signal and U is an unmethylated signal, and M values were extracted using the minfi package. The beta value represents degree of methylation for the CpG site where a higher beta value indicates higher methylation levels. The beta values were logit-transformed into M-values [M= logit(beta)] to be more accurately approximated by the normal distribution for subsequent statistical analysis. The paired t-test was applied to the change in M-values for all subjects to identify loci with a change in average methylation status irrespective of response. Additionally, the two-sample t-test was applied to M-values to identify loci with differential baseline methylation status, day 5 methylation status, or change in methylation status according to response. Results with p< 0.0005 were considered statistically significant and the method of Pounds and Cheng[21] was used to estimate the false discovery rate (reported as q-value) at this p-value threshold and estimate the proportion of the genome with a change in methylation status during decitabine treatment.

Results

Thirteen eligible patients enrolled, 12 with B-Cell ALL (B-ALL) and one with T-Cell ALL (T-ALL). One of the 12 B-ALL patients had Ph+ ALL and had failed prior Induction attempts which included the tyrosine kinase inhibitor imatinib. Twelve patients completed the full 4 days of decitabine and vorinostat treatment and were evaluable for toxicity assessment with eight of these patients completing full protocol therapy and end of course evaluation. The one patient who did not complete the 4 days of decitabine and vorinostat died of progressive disease on day 4 of study. Reasons for the remaining four patients not completing protocol therapy were progressive disease (n=1), toxic death (n=1), family decision to stop protocol therapy secondary to treatment related toxicities (n=1) and the fourth patient was removed from study at day 5 of therapy when found to have CNS leukemic involvement and the treating physician wanting to change their treatment approach. Of the eight patients who completed protocol therapy and were evaluable for toxicity and response, four achieved CRp as well as MRD negativity by multi-parameter flow cytometry (50%, CI: 15.7% to 84.3%). Two patients achieved a PR (25%), while one patient had stable disease and another had progressive disease. Results based on intention-to-treat include four patients with CR out of the total eligible of 13 patients, giving a complete remission rate of 30.8% (95% CI 9.1% to 61.4%). Including the two PR patients, the overall response rate was 46.2% (95% CI 19.2% to 74.9%). When limiting the study results to the eight patients that were evaluable for response at day 33, the complete remission rate becomes 50% (n=4/8) (95% CI 15.7% to 84.3%) and the overall response rate (CR+PR) 75% (n=6/8) (95% CI 34.9% to 96.8%). The lower end of this last confidence interval remains above 30%, which was considered the lower limit of treatment response in this study. The progression-free survival at day 60 for the study cohort (n=13) was 45% (95% CI 17% - 73%) (Figure 1).

Figure 1.

Progression-free Survival; The 60-day PFS for the study cohort (n=13) was 45% (95% CI 17% - 73%).

Of the eight evaluable patients who completed protocol therapy, five proceeded to HCT (4 in CR2 and 1 CR3). Three of these patients died of transplant related causes without evidence of leukemia while the remaining two patients remain alive with no evidence of disease.

Toxicity Results

Eleven of the 12 evaluable patients (92%) enrolled on study completed the 4-day course of decitabine and vorinostat and were evaluable for toxicity. See Table 3 for detailed grade ≥3 toxicities with attributions of at least possibly related to decitabine and/or vorinostat. Grade 3 or 4 toxicities attributed to the chemotherapy regimen only included grade 4 dyspnea (n=1), grade 4 hypercholesterolemia (n=1), grade 4 hypertriglyceridemia (n=1), grade 4 lipase (n=1), and grade 4 hypoxia/ acute respiratory distress (n=1). There was a single toxic death occurring on study attributed to the chemotherapy regimen which included a grade 5 hemorrhage/ bleeding (n=1) with a second patient experiencing grade 5 hypoxia/ acute respiratory distress who died on day 4 of study attributed to disease progression (n=1). There were an additional 14 SAE grade ≥3 which were at least possibly attributed to decitabine or vorinostat on this study (Table 3).

Table III. Toxicities Possibly, Probably or Definitely Related to Decitabine and/or Vorinostat.

Adverse Events
Toxicity – Grade 3	# of Events	% of Total Events
Fever with neutropenia	3	27.3
ALT	1	9
Hyperglycemia	1	9
Hyponatremia	2	18.2
Hypokalemia	1	9
Low albumin	1	9
Pain – abdomen	1	9
Neuro – Right hemiparesis	1	9
Dehydration	1	9
Fatigue	1	9
Nausea – intermittent	1	9
Toxicity – Grade 4		%
Neutropenia	2	18.2
Neutropenic fever	1	9
Severe Adverse Events
Toxicity – Grade 3	# of Events	% of Total Events
Fever with neutropenia	2	18.2
Infection (Blood) with neutropenia	3	27.3
Infection (Abdomen - NOS) with neutropenia	1	9
Pain – Abdomen NOS	1	9
Hyponatremia	1	9
Toxicity – Grade 4		%
Hypercholesteremia	1	9
Hyperglycemia	1	9
Hypertriglyceridemia	1	9
Pancreatitis	1	9
Infection (Blood) with neutropenia	2	18.2

All toxicities were reported according to CTCAE version 3.0

Comparison of DNA methylation profiles between responders and non-responders

As expected, we observed significant demethylation across the entire genome during the treatment with decitabine. Overall, results of the LINE-1 methylation determined by pyrosequencing demonstrated a similar trend with methylation ranging from 81-87% on day 0 (mean 84.98% ± 1.48%; n=10), 74-83% on day 5 (mean 79.82% ± 3.04%; n=10) and 83-89% on day 33 (mean 86.1% ± 1.87%; n=6) (Figure 2A). Although we observed inter-patient variation with the extent of demethylation due to decitabine (Supplementary Figure 1), there was no correlation between decitabine induced demethylation and clinical response observed. Genome-wide methylation profiling using illumina 450K bead chip identified 17,150 probes with significant change in methylation at the p<0.0005 significance level (q = 0.008). A total of 17,137 (99.9%) of these probes showed some degree of demethylation during decitabine treatment. We estimate that at least 45% of the probes on the Illumina 450k platform experienced a change in average methylation during decitabine treatment (Supplementary Figure 2).

Figure 2.

Baseline and decitabine induced methylation signatures: 2A) in LINE-1 Methylation indicating global DNA methylation decrease following 4 days of treatment with decitabine in 10 patients. Day 0 indicates LINE-1 methylation pretreatment, day 5 reflects LINE-1 methylation after 4 days of decitabine treatment and day 33 indicates LINE-1 methylation levels at day 33 of study therapy. 2B): Heat map representation of genes/probes with significant differences in methylation at day 0 between responders and non-responders. 2C): Heat map representation of genes/probes with significant differences in methylation at day 5 between responders and non-responders. 2D): Heat map representation of genes/probes with significant differences in methylation for the change in methylation (day 5-day 0) between responders and non-responders. For the Heat maps only probes annotating to known genes are included in each row and are ordered by the p value (smallest p value at the top or bottom, and largest p value in the middle); each column represents a patient.

We next evaluated day 0 and day 5 methylation profiles and the change in methylation post-decitabine among responders (n=4) and non-responders (n=4). Responders were patients who achieved a CR/CRp whereas non-responders achieved PR, SD or PD. Methylation profile differences at baseline (day 0) between responders and non-responders identified 158 probes (annotating to 101 genes) at a p-value cutoff of 0.0005 (q = 0.61). Of these 158 probes, 28 probes (17.7%), annotating to 18 genes, had a lower mean methylation at day 0 in non-responders compared to responders and 130 probes (82.3%), annotating to 83 genes, had higher (base line day 0) mean methylation in non-responders as compared to responders (Supplementary table 1 shows the complete list of probes). The heat map showing beta values for methylation for probes demonstrating a significant difference between patients achieving a CR versus not is shown in Figure 2B. Further, ingenuity pathway analysis (IPA) mapped these genes to 7 networks (Table 4), with the top networks being Cell Cycle, DNA Replication/Recombination/Repair, and Tissue Morphology and included the genes ATXN1, MAPK10 (Map kinase), IGFBP3 (insulin-like growth factor binding protein 3), NFKBPIL2, ETV6 (ETS variant 6), MAD1L1 (MAD1 mitotic arrest deficient-like 1), RGS10 (regulator of G protein signaling 10) and TRRAP (transformation/transcription domain-associated protein).

Table IV. Ingenuity Pathway analysis mapped the genes that showed significant change (p<0.0005) in methylation profiles at day0 and day 5 and change in methylation (day5-day0) between responders and non-responders.

Responder vs. Non-responder DAY 0
	Molecules in Network	Score	Focus Molecules	Top Diseases and Functions
1	Actin,Akt,ATXN1,caspase,CD3,CR1L,cytochrome, ETV6, FBLN5,FBN1,GPHN,Hdac,HIPK2, HISTONE,Histone h3,HMGN3,IGFBP3,MAD1L1,MAPK10,N-cor,NFAT (complex),NPIPA1 (includes others),NRXN1,NXPH1, PAG1, PRDM16,PSMB9,RGS10,RNA polymerase II,SFRP2, SIX5,SLC4A1,SMOC2,TCR,TRRAP	48	23	Cell Cycle, DNA Replication, Recombination, and Repair, Tissue Morphology
2	ADIPOR1,Alp,AMPK,BHLHE40,CABIN1,CAST,Cg,DNAS E1,ERK1/2,FSTL1,G-Actin,IFN Beta,IFNGR1,Ige,IgG1, Igm, Immunoglobulin, JAM3,KLF6,KRT72,LIMD1,LST1, MARK2,NPLOC4,PDGF BB,PFKFB3,PLCG1,PLEKHA1, PRKCH,RAPGEF2,RGS3,Sos,Tgf b,TNFRSF19,Ubiquitin	44	21	Cellular Assembly and Organization, Cellular Development, Embryonic Development
3	ADGB,ADNP2,BOD1L1,BUD31,C19orf53,CACHD1,CADM 3,CADM4,CCDC93,FAM86A,GLTSCR1L,KBTBD8,KIAA01 01,LRP10,MEA1,MOB1B,MROH1,MRPS17,NFYC,NIN,PL BD2,PRTFDC1,RAI14,RBM34,RNF166,ROR1,SF3B5,SIM C1,TONSL,TPM4,TTBK2,UBC,ZC3HAV1L,ZCCHC3,ZFHX 3	31	16	Cellular Development, Embryonic Development, Hereditary Disorder
4	ABHD11,ADCK2,ASH1L,ATP5I,ATP6V1D,BAZ2B,BCAT2, BTN3A2,CALB2,CAPG,CCP110,CEP97,CEP104,CUEDC1 ,DENND3,DNASE2,ENPP4,GABRB3,GPNMB,GSTO1,HD AC8,HIST1H2AD,IVD,MECP2,MFAP3L,NR2F6,PLA2G16, SLAMF8,SMG6,STARD3NL,TMX4,UBC,ZNF226,ZNF502, ZNF492/ZNF98	24	13	Amino Acid Metabolism, Small Molecule Biochemistry, Cellular Assembly and Organization
5	ABCA4,ALB,APP,ATG7,DNAJC1,EEF1G,ESR1,FAM71E2 ,FBLN7,FOXK1,GABARAPL2,GNAQ,Gpcr,GPR62,GPR75, GPR97,GPR111,GPR112,GPR114,GPR144,GPR152,GPR 157,GPR162,GPR171,GPR174,GPRC5D,HTR1E,IRGQ,IT GB1,MAP1LC3B2,PNMAL1,TAAR8,TLN2,UGGT2,VN1R2	20	12	Cellular Assembly and Organization, Cellular Compromise, Developmental Disorder
6	ASIC2,beta- estradiol,C1orf109,CDC123,CDH1,CLYBL,CMSS1, COA4,CPA2,CWC25, DNPH1, DSCR3,CE2,FXYD5,HNF4A, IL1A,IQSEC1, KIAA1244,LARS2,MAL2,PPP1CA,PPP1R11, PPP1R14D,RAP2C,RCAN3,SH2D4A,SLC39A6,SSR2,ST OML1,TMEM258,TTC25,YPEL3,ZBTB11,ZCCHC9,ZNF39 8	15	9	Cell Death and Survival, Connective Tissue Development and Function, Endocrine System Development and Function
7	AC1/8,Ap1,ATP5G2,CALM1 (includes others),Calmodulin, CLDN1, CORO6,CPNE5, CTXN1,ERK,GALNT9,IL1, Insulin, Interferon alpha,Jnk,LGI3,LRRTM1,LRRTM3,MAP9,Mapk,mir- 515,MYO9B,NECAB1,NFkB (complex),P38 MAPK,PHOSPHO1, PI3K (complex),Pkc(s),PPEF1,Ras,SLC2A12,SPA17, TMEM74,Vegf,YPEL2	9	6	Gastrointestinal Disease, Immunological Disease, Cardiac Arrythmia

Responder vs. Non-responder Change in methylation (DAY5-DAY0)
	Molecules in Network	Score	Focus Molecules	Top Diseases and Functions
1	ACTL6A,ACTR5,ACTR8,ARID2,BCAM,BCL7B,BRD7,CD1 B,CHD7,CRYL1,ELAVL3,ERICH1,FAM118B,FARP2,FRY L,INO80C,KHDRBS2,KIAA1524,LAMA5,MAGEA4,MAN1A 2,MAST1,MCRS1,MRPL44,NFRKB,PBRM1,PELI2,PSMD1 0,SLC12A8,UBC,UQCRB,UVSSA,ZBTB17,ZNF235,ZNF31 9	36	15	Cell-To-Cell Signaling and Interaction, Renal and Urological System Development and Function, Tissue Development
2	ADAMTS1,ADAMTS4,ADAMTS5,ADAMTS9,AHR,CCL8,C CL21,CDC42EP1,CGREF1,chondroitin sulfate B,CXCR3,D-glucose,EFNA1,EIF2S1,ERK,F7,FBLN2, GLIS3, HAPLN1,HOXB3,IL5,LTBP1,Metalloprotease,NFkB (complex),NREP,PLXNA2,PTPRF,RBPMS,SLC1A5,SLC2 2A7,SLC2A12,TFG,TFPI,TLR6,VCAN	24	11	Cell Morphology, Connective Tissue Disorders, Hematological System Development and Function
3	ADAMTS1,AIM1L,APP,ARHGAP9,CHFR,chondroitin sulfate B,CPNE8,CRYL1,DHX36,ELN,FADS3,FBLN1, GBP4,Gbp8,GLB1,GPR110,GPR123,HAPLN1,HEYL,IFNG ,KRT23,L3MBTL4,MAGEA4,MMEL1,NVL,PDZD9,PI4K2B, PNPLA6,RARRES1,SLC25A13,SUCNR1,TMEM87A,UBC, VCAN,WARS2	24	11	Neurological Disease, Cell-To- Cell Signaling and Interaction, Inflammatory Disease

Post-decitabine (day 5) DNA methylation levels were compared between non-responders and responders at a p-value of 0.0005 (q=1). Thirty-three probes annotating to 23 genes were differentially methylated between the two groups. Of the 35 probes, 21 (60%) had lower mean methylation and 14 (40%) had higher mean methylation in non-responders than in responders (Figure 2C; Supplementary Table 1). A heat map showing differences in methylation between responders and non-responders is shown in Figure 2C. Since we did not identify more probes as significant than would be expected by chance, we did not explore their relevance further using IPA.

Since the degree of decitabine-mediated hypomethylation can vary between patients, we evaluated treatment-mediated changes in the methylation profile (M value change: day 5-day 0) and identified 63 probes (p-value of <0.0005, q = 0.80). List of probes with significant differences between responders and non-responders are included in supplementary table 2. Surprisingly, of these 63 probes 60 (95.2%) had significant hypomethylation in non-responders as compared to responders (Figure 2D). Some of the more interesting genes included DHX36 (member of DEAH-box family of RNA-dependent NTPases), solute carrier transporters (SLC12A8, SLC22A7, SLC25A13 and SLC2A12), HOXB3, EIF2S1 (Eukaryotic translation initiation factor2, subunit1 alpha), and PDCD4 (programmed cell death 4). IPA analysis of the genes with a differential change in methylation profiling post-decitabine treatment mapped these to 3 networks, namely (summarized in second part of Table 4): Network 1: Cell-to-Cell signaling and interaction, Renal and Urological System Development and Function and Tissue Development; Network 2: Cell Morphology, Connective Tissue Disorders, and Hematological System Development and Function and Network 3: Neurological Disease, Cell–to-Cell Signaling and Interaction and Inflammatory Disease.

Discussion

Although we have witnessed dramatic improvements in the outcome of children and young adults with ALL, approximately 15% of patients will fail frontline treatment.[1] Despite aggressive salvage regimens, sustained remission can be achieved in very few and disease recurrence remains the largest obstacle.[22] Therefore, novel approaches to relapsed leukemia remain paramount. As epigenetic alterations involving pathways of treatment resistance are prevalent in cancer,[23; 24] attempts reverse the resistance profile of leukemic blasts using epigenetic modifying therapies[8; 25] would be one approach. We pursued this strategy in a phase II study combining decitabine and vorinostat followed by chemotherapy in patients with relapse/refractory ALL (NCT00882206).

Prior to this clinical trial, there had been limited investigation using decitabine and/or vorinostat in patients with relapsed ALL[17; 26; 27; 28] and no clinical trial using the combination of these two agents followed by re-induction chemotherapy. Garcia-Manero et al. reported results of a phase I study using decitabine alone (dose escalation up to 60mg/m² daily × 5 days) or in combination (15 mg/m² daily × 5 days of decitabine) with Hyper-CVAD in relapsed or refractory adult ALL.[29] There were no dose limiting toxicities and a complete responses was observed in 4/12 (30%) patients who received decitabine as a single agent and in 3/9 (30%) patients receiving decitabine concurrently with Hyper-CVAD therapy. A second phase I study investigating decitabine and vorinostat in adult patients with relapsed, refractory or poor prognosis leukemia (n=31), using sequential dosing of decitabine (10 to 25 mg/m² daily × 5) followed by vorinostat (300 mg × 14 days in the first cohort and 600 mg × 14 days in all subsequent cohorts), but no additional chemotherapy was delivered.[18] The median age was 62 (range, 22-82) years. Toxicities included syncope, neutropenia with fever, diarrhea, fatigue, renal failure, rash, nausea, thrombosis, and angioedema. Of 30 evaluable patients, one achieved a CR lasting 5.5 weeks, four had partial responses with reductions in bone marrow blast percentage, four had stable disease, fourteen patients had no response or disease progression, and seven were too early for response evaluation. The results of our phase II study are encouraging with an overall response rate of 46.2% in the intention-to-treat analysis and considerably better in patients who completed therapy, having a complete remission rate of 50%, overall response rate of 75% and an acceptable toxic death rate of 15% for this heavily pre-treated population. Although there is overlap in the chemotherapy agents between our study and the decitabine/Hyper-CVAD phase I trial listed above, the inclusion of PEG-asparaginase along with combining vorinostat with decitabine for synergy were likely responsible for the difference in outcome between the two studies. Among the 5 patients in our study who went on to receive HCT after completing protocol therapy, 3 died of transplant related causes which appears greater than what is expected but the small number limits any ability to associate these deaths to the prior epigenetic therapy.

We performed correlative analyses including genome-wide DNA methylation in matched day 0 and day 5 samples from 8 patients (4 responders and 4 non-responders). Although the sample size in this study was small we observed significant gene hypomethylation post-decitabine treatment, as well as significant differences in the methylation profiles between responders and non-responders at days 0 and 5. Identification of genes mapping to biologically relevant networks (e.g. cell-to-cell signaling and interaction, cell morphology, connective tissue disorders, and hematological system development and function) indicated that differences in methylation profiles between non-responders and responders might influence the observed differences in outcome. We observed significant differences in overall methylation at day 0 between the 2 groups, as well as a greater number of genes with significantly lower methylation in patients who achieved a CR versus patients who did not. One of the more interesting genes identified to be hypomethylated in patients who achieved a CR was ETV6 (oncogene encoding an ETS family transcription factor). ETV6 is located on chromosome 12q13 and when translocated with chromosome 21, resulting in the TEL-AML fusion, it becomes the most common chromosomal rearrangement reported in pediatric B-ALL. Although none of the patients in the current study had TEL-AML [t(12:21)] rearrangements, this translocation can be seen in 25-30% of pediatric ALL and has been associated with a better prognosis.[30] As ETV6 is required for hematopoiesis, it is possible that higher expression of ETV6 due to hypomethylation might contribute to the observed differences between responders and non-responders. One of the other biologically relevant genes we identified included APAF1 (apoptotic peptidase activating factor 1) (supplementary Table1). This gene had significantly lower methylation at day 5 (post decitabine) among responders versus non-responders with a p-value (p=0.00503) marginally above significance threshold. APAF1 codes for a cytoplasmic protein that triggers apoptosis and down-regulation of APAF1 has been associated with relapse in AML.[31; 32]

We also compared the results from this study (day 0, day 5 and change in methylation differences from day 5 to day 0 in responders vs. non-responders at a p<0.001) with the recently published epigenetic study in ALL patients, which reported subtype and disease specific alterations in methylation.[33] Overall we observed an overlap of 32 genes that were differentially methylated in B-ALL and were significantly different between responders and non-responders for the decitabine induced methylation changes (day 5 to day 0). Some of the biologically relevant genes included those involved in apoptosis and cell death (DNAJB6, ARHGEF7, MAP3K5, SQSTM1 and SIRT1), as well as genes associated with negative regulation of transcription and gene expression (DNAJB6, BTAF1, SIRT and UB2E1). For genes identified as differentially methylated at day 0 and day 5, 16 and 8 genes overlapped respectively with 4 being common (RICH2, EPP4, FSTL1 And GPNH) (Supplementary Table 3).

The overall clinical responses seen on this trial are quite encouraging, particularly given how heavily pre-treated the patients were, with five patients previously failing HCT. Not only were four patients able to achieve a CR, but they achieved MRD negativity (<0.01%). The expected MRD-negativity rate for patients treated with this chemotherapy backbone (VPLD) who successfully achieve a CR, without decitabine and vorinostat, is around 38%.[34] The fact that 100% of the patients on this study who achieved a CR also became MRD-negative is encouraging and supports the hypothesis that epigenetic therapy can reverse chemotherapy resistance. This approach of “re-programming” the resistance profile of leukemia blasts using epigenetic modifying agents appears promising but requires further study in larger trials of both pediatric and adult patients. If epigenetic modifying agents are able to increase the sensitivity of resistant leukemia blasts to chemotherapy, then it may be possible to reduce the doses (or frequency) of more traditional cytotoxic chemotherapy agents in future relapsed ALL trials and achieve similar or greater response rates.

In summary, we report the results of a phase II clinical trial investigating decitabine and vorinostat in combination with chemotherapy, which shows encouraging results with regard to toxicity and remission induction in relapsed/refractory ALL. Based on the results of this study, a pediatric trial entitled “A pilot study of decitabine and vorinostat with chemotherapy for relapsed ALL (NCT01483690; R21CA161688-01)” is investigating a similar chemotherapy platform with an extended schedule of decitabine and vorinostat for relapse/refractory ALL through the Therapeutic Advances in Childhood Leukemia and Lymphoma (TACL) Consortium. To the best of our knowledge this is the first report of exploratory analysis of genome-wide DNA methylation profiles determined at baseline and post-decitabine treatment in matched samples from patients with relapsed ALL on a clinical trial. Further we report the baseline and drug induced changes in DNA methylation profiles differentiating responders from non-responders. These results point to the fact that inter-patient variation in response to DNA modifying agents such as decitabine can influence the treatment outcome and warrants further validation studies in a larger trial cohort.

Supplementary Material

Supp FigureS1-S2

Supplementary Figure 1: Methylation changes in each patient. P-values of comparison are between day 5 and day 0 for each patient with significant p-values (p<0.05) indicated in bold. The clinical response for each patient is listed below each set of bars.

Supplementary Figure 2: Manhattan Plot of p-values of equality for beta value between day 0 and day 5 of decitabine and vorinostat in patients. X-axis represents genomic location and Y-axis is Log₁₀ values of differences between responders and non-responders.