Evaluation of noncytotoxic DNMT1-depleting therapy in patients with myelodysplastic syndromes

Abstract

Background: Mutational inactivation in cancer of key apoptotic pathway components, such as TP53/p53, undermines cytotoxic therapies that aim to increase apoptosis. Accordingly, TP53 mutations are reproducibly associated with poor treatment outcomes. Moreover, cytotoxic treatments destroy normal stem cells with intact p53 systems, a problem especially for myeloid neoplasms, as these cells reverse the low blood counts that cause morbidity and death. Preclinical studies suggest that noncytotoxic concentrations of the DNA methyltransferase 1 (DNMT1) inhibitor decitabine produce p53-independent cell-cycle exits by reversing aberrant epigenetic repression of proliferation-terminating (MYC-antagonizing) differentiation genes in cancer cells.

Methods: In this clinical trial, patients with myelodysplastic syndrome (n=25) received reduced decitabine dosages (0.1-0.2 mg/kg/day compared with the FDA-approved 20-45 mg/m2/day dosage, a 75%-90% reduction) to avoid cytotoxicity. These well-tolerated doses were frequently administered 1-3 days per week, instead of pulse cycled for 3 to 5 days over a 4- to 6-week period, to increase the probability that cancer S-phase entries would coincide with drug exposure, which is required for S-phase-dependent DNMT1 depletion.

Results: The median subject age was 73 years (range, 46-85 years), 9 subjects had relapsed disease or were refractory to 5-azacytidine and/or lenalidomide, and 3 had received intensive chemoradiation to treat other cancers. Adverse events were related to neutropenia present at baseline: neutropenic fever (13 of 25 subjects) and septic death (1 of 25 subjects). Blood count improvements meeting the International Working Group criteria for response occurred in 11 of 25 (44%) subjects and were highly durable. Treatment-induced freedom from transfusion lasted a median of 1,025 days (range, 186-1,152 days; 3 ongoing), and 20% of subjects were treated for more than 3 years. Mutations and/or deletions of key apoptosis genes were frequent (present in 55% of responders and in 36% of nonresponders). Noncytotoxic DNMT1 depletion was confirmed by serial BM γ-H2AX (DNA repair/damage marker) and DNMT1 analyses. MYC master oncoprotein levels were markedly decreased.

Conclusion: Decitabine regimens can be redesigned to minimize cytotoxicity and increase exposure time for DNMT1 depletion, to safely and effectively circumvent mutational apoptotic defects.

Trial registration: Clinicaltrials.gov NCT01165996.

Funding: NIH (R01CA138858, CA043703); Department of Defense (PR081404); Clinical and Translational Science Award (CTSA) (UL1RR024989); and the Leukemia and Lymphoma Society (Translational Research Program).

Figures

Figure 7. Increase in frequency of drug administration, though at a lower dose, salvaged response, consistent with the importance of exposure time with an S-phase–dependent mode of therapy.

Blood count profiles over a 52-week period in a study subject. Green arrow denotes BM evaluation at nadir after the initial induction phase; red arrow denotes BM evaluation of relapsing disease. Cellularity denotes BM cellularity. Each diamond mark in the bottom panel indicates a decitabine administration, with the dose indicated on the y axis.

Figure 6. Functional hematopoietic capacity.

(A) Shorter ATLs in MDS subjects compared with those observed in age-categorized normal controls and especially in nonresponders compared with responders. Shorter telomere length implies greater replications by fewer hematopoietic clones. Telomere length was measured by qRT-PCR. The standard curve is shown in Supplemental Figure 10. Error bars represent the median ± IQR. P values were determined by the Mann-Whitney U test. (B) Pretreatment BM cellularity was especially decreased in subjects with short telomere lengths, including in a young subject with a previous diagnosis of aplastic anemia and cytogenetic response without hematologic response on this protocol. Orange pyramids denote responders; purple diamonds denote nonresponders. (C) Pretreatment ANCs were significantly decreased in subjects previously treated with 5-azacytidine (Aza) and/or lenalidomide (Len). Error bars represent the median ± IQR. P values were determined by the Mann-Whitney U test. (D) Lower pretreatment neutrophil counts in nonresponders resulted in fewer decitabine administrations during the induction period (weeks 0–6). Error bars represent the IQR. P values were determined by the Mann-Whitney U test.

Figure 5. Noncytotoxic DNMT1 depletion and decrease in MYC.

(A) γH2AX (DNA damage/repair/apoptosis marker) was measured by flow cytometric analysis of BM (an example is shown here). Data for all subjects are provided in the Supplemental material. (B) γH2AX expression in BM cells before treatment (week 0) and at week 6. R, responders; NR, nonresponders. P values were determined by a paired 1-tailed t test. (C) Objective quantification of immunohistochemical data. Decalcified and formalin-fixed, paraffin-embedded sections (4 μm thick) of BM biopsies from different time points were immunostained on the same slide. Normal BM and positive and negative controls were concurrently stained. ImageIQ software was used to segment the image, and positive nuclei were objectively quantified in cellular segments. Raw data from software quantification of positive nuclei are provided in Supplemental Table 2. (D and E) Change in DNMT1 and change in MYC between baseline and week 6. Week 6 corresponded to the nadir (suppression of clonal hematopoiesis). A mean of 3 to 25 tissue segments was quantified per time point per biopsy. Additional graphical representation is provided in Supplemental Figure 7.

Figure 4. Serial blood counts, myeloblasts, and cytogenetics results in responders and nonresponders.

(A) Profile of blood counts between baseline and week 52. Left: Responders (those with CR and/or HI). Right: Nonresponders (those with stable disease or progressive disease). Protected data from the Oncore database. Values only for subjects actively receiving drug (Figure 3B) (even some nonresponders continued to receive drug because of stable disease or other treatment benefits). Mean values are indicated with a purple line, and 95% CI is indicated by gray shading. Platelet and ANC values × 109/l. Hemoglobin (Hgb) values in g/dl. Values were clipped if they exceeded the depicted y axis scales. (B) Change in BM myeloblast percentages between weeks 0 and 12. (C) Change in the percentage of abnormal metaphases between weeks 0 and week 12 (abnormal metaphases are listed in Table 1).

Figure 3. Treatment schema and dosages actually administered to individual subjects.

(A) Protocol treatment schema. (B) Decitabine dose and frequency of administration for individual subjects. After an initial 4-week induction phase, decitabine dose and schedule modifications were protocol mandated to adapt to known inter- and intraindividual differences in decitabine metabolism, cancer S-phase fraction, cancer burden, blood counts, and hematopoietic reserve. (C) CONSORT diagram for enrollment and analysis.

Figure 2. Chemical properties of decitabine and S-phase–dependent mechanism-of-action considerations that influenced regimen design.

(A) The sugar moiety of the cytidine analog cytarabine is unnatural and thus terminates DNA chain elongation. This is the intended molecular pharmacodynamic effect. Decreasing the cytarabine dose to improve safety decreases this intended effect (reduces efficacy). (B) The sugar in decitabine is natural, enabling DNA incorporation without terminating chain elongation. This shifts the DNA damage curve to the right compared with cytarabine (10, 14, 33, 34), even though the decitabine-DNMT1 DNA-protein cross-link requires repair by homologous recombination (–37). The DNA repair delays cell-cycle progression; however, concentrations can be found that deplete DNMT1 without cytotoxicity (green zone), the intent in this clinical trial (unlike with cytarabine, a lower dose need not imply lower efficacy). (C) Since DNMT1 depletion is S-phase dependent, for an equivalent tumor burden, a lower S-phase fraction (green in pie chart) may require more drug exposure time (tlo-S) (e.g., more frequent administration) to treat the same fraction (f) of disease. That is, counterintuitively, less aggressive disease may require more frequent drug administration. More intuitive is that a similar S-phase fraction, but a higher total tumor burden (e.g., 1012 versus 1011 cancer cells, i.e., more advanced disease), may require greater exposure time to reduce the tumor burden to a level that permits functional hematopoiesis. (D) These considerations guided protocol regimen design. Top: FDA-approved regimen of decitabine 45 mg/m2/day i.v. on days 1–3 every 6 weeks. Middle: FDA-approved regimen of decitabine 20 mg/m2/day i.v. on days 1–5 every 4 weeks. Bottom: Protocol regimen 0.1–0.2 mg/kg/day (~3.5–7 mg/m2/day) s.c. 1–3 days per week (shown is 0.2 mg/kg 2 days/week). All treatment regimens may be continued indefinitely.

Figure 1. Master transcription factor expression in NHSCs versus MDS and AML stem cells.

HLF drives stem cell fate, and CEBPA, PU.1, and GATA1 drive myeloid lineage fates, as demonstrated in lineage conversion and murine knockout studies (25, 47). Surface phenotypes used for flow purification reproducibly identify cell fractions with the capacity to reconstitute long-term myelopoiesis in immunocompromised mice (29, 46) (GEO GSE55689 and GSE24006). These phenotypes were DAPI–Lin–CD34+CD38–/loCD90+CD45RA– for NHSCs and MDS stem cells (MDS SCs) and DAPI–Lin–CD34+CD38-/loCD90– for AML SCs (AML engrafting cells differ from NHSCs and MDS SCs in being CD90– and can also be CD38+, a fraction not analyzed here) (29, 46). MDS and AML cases were representative of the morphologic and genetic spectrum of disease, and in MDS, included low- and intermediate-risk cases (29, 46). MDS cases with TP53 mutation or deletion are indicated (blue and red circles, respectively). Gene expression is shown relative to the mean expression of the same gene in simultaneously analyzed NHSCs. Error bars represent the median ± interquartile range (IQR). P values were determined by the Mann-Whitney U test.