NY-ESO-1 Vaccination in Combination with Decitabine Induces Antigen-Specific T-lymphocyte Responses in Patients with Myelodysplastic Syndrome

Myelodysplastic syndromes (MDS) are hematologic disorders with an estimated overall incidence between 5 and 13 cases per 100,000 people annually in the United States, and a substantially higher incidence in those over the age of 65 years (1). They are characterized by ineffective hematopoiesis with progressive cytopenias and a variable risk of transformation to acute myeloid leukemia (AML; ref. 1). For patients with higher risk disease, the median overall survival is between 0.4 and 1.2 years (2, 3). Nonintensive therapy with azanucleosides (azacitidine and decitabine) has demonstrated a survival advantage in patients with MDS (4, 5). Unfortunately, responses to these therapies are relatively short-lived, and patients whose disease progresses while on therapy have a poor prognosis (6, 7). Although allogeneic hematopoietic cell transplantation (aHCT) is potentially curative, this approach is unsuitable for most patients due to their age and comorbidity (8). Despite this barrier, the relative clinical effectiveness of aHCT acts as a proof of concept for immunotherapeutic approaches in the treatment of MDS and provides a rationale for developing alternative immunotherapeutic strategies.

The mechanism of clinical action for azanucleoside therapy remains a matter of debate, and there is a growing literature supporting enhanced or altered immunologic milieu as a significant contributor to response (9, 10). The cross-talk between the tumor and immune systems is comprised of a complex series of mechanisms, many of which can be epigenetically regulated (11). One such mechanism that can be exploited by azanucleosides is their effect on expression and presentation of tumor antigens that are recognized by the adaptive immune system (11). Recent studies have demonstrated that azanucleosides induce expression of endogenous retroviral genes and activate type I or type III IFN responses (12, 13). Azanucleosides have also been shown to enhance the expression of genes involved in the antigen presentation machinery (10, 14). In addition, several groups, including ours, have demonstrated that azanucleosides can induce expression of a class of immunogenic antigens termed “cancer testis antigens” (CTA; refs. 14, 15).

CTAs are a family of more than 130 X-linked and non–X-linked genes that are expressed in the embryonic ovary and the adult testis. In all other normal tissues, expression of CTA family genes is low due to epigenetic silencing of regulatory elements (16, 17). CTAs are aberrantly expressed in nonhematologic cancers, including lung cancer, melanoma, and ovarian cancer (16, 18). The immunogenicity of these antigens prompted development of vaccine and engineered T-cell strategies targeting CTAs in different cancer types (16). A majority of myeloid cancers do not express CTAs due to promoter hypermethylation. Studies from our group and others demonstrate that MDS/AML samples from patients receiving azanucleosides exhibit induced expression of CTA family members (14, 15). Goodyear and colleagues demonstrated that the combination of azacitidine and the histone deacetylase inhibitor valproic acid resulted in CTA-specific T-lymphocyte responses in patients with MDS/AML (19). These T-lymphocyte responses have been correlated with therapeutic response (19, 20).

The NY-ESO-1 CTA is of particular interest in cancer immunotherapy due to its immunogenicity, restricted tissue expression, and safety profile as an immune target in a large variety of solid tumors (18, 21–23). We and others have shown that azanucleosides induce expression of NY-ESO-1 protein in AML cell lines and AML xenografts (15, 24). We further demonstrated that induction of NY-ESO-1 expression occurs in circulating AML blasts isolated from patients treated with decitabine as a standard of care (14). Induction of NY-ESO-1 expression was sufficient to activate a cytotoxic response from HLA-compatible NY-ESO-1–specific T lymphocytes. On the basis of these data, we hypothesized that vaccination against NY-ESO-1 in patients with MDS would activate an antigen-specific immune response against the malignant myeloid compartment in patients who demonstrate decitabine-induced expression of NY-ESO-1.

To test this hypothesis, we designed a phase I study in which 9 MDS patients were enrolled. Our group has previously demonstrated the safety and feasibility of such an approach in a phase I study combining NY-ESO-1 vaccination (CDX-1401 + poly-ICLC) with decitabine (and doxil) in patients with platinum-refractory ovarian cancer (21, 23). This approach is similarly safe in patients with MDS, with toxicities chiefly related to the underlying myeloid malignancy and the chemotherapy decitabine. We show that (i) a majority of patients develop NY-ESO-1–specific CD4⁺ and CD8⁺ T-lymphocyte responses; (ii) these NY-ESO-1–specific T-lymphocytes can recognize autologous myeloid cells from patients undergoing decitabine therapy; and (iii) antigen-specific humoral and adaptive immune responses to vaccination were associated with detectable numbers of CD141^HI conventional dendritic cells (cDC), a subtype of antigen-presenting cell (APC). CD141^HI cDCs have a recognized role in antitumor immune responses and express the antigen uptake receptor for CDX-1401 (23, 25).

These data demonstrate the feasibility of vaccination against an azanucleoside-induced antigen in a nonimmunogenic myeloid cancer and provide an avenue for targeted immunotherapy in myeloid malignancy. Critically, as azanucleosides are the standard of care in MDS, this approach offers the opportunity for rapid translational development of combination immune adjuvant therapy.

This was an open-label, nonrandomized, single-center, phase I dose–de-escalation study of NY-ESO-1 vaccine administered in combination with standard dose decitabine 20 mg/m²/day in subjects with MDS or low blast count AML (26). Planned study treatments included five vaccinations and four cycles of decitabine; the study ended after cycle 4, day 29 (Fig. 1A). The primary endpoint of this study was safety. Secondary endpoints were (i) evaluation of NY-ESO-1–specific cellular and humoral immune responses and (ii) determination of combination treatment on peripheral blood myeloid cells for NY-ESO-1 target gene expression, NY-ESO1 protein expression, NY-ESO-1 promoter methylation, and global DNA methylation. A modified 3 + 3 design with a 3-patient expansion cohort at the maximum administered dose (MAD) was used. All 9 patients were accrued and treated at dose level 1 (CDX-1401 at 1 mg; poly-ICLC at 2 mg); this dose was chosen based upon data from a previously completed study in patients with solid tumors (23). The dose-limiting toxicity (DLT) window was from cycle 1, day -14 to cycle 2, day 1 (see Fig. 1A); related ≥ grade 3 nonhematologic toxicities were considered dose limiting. If 0 or 1 of the first 3 patients had a DLT, then 3 more patients were to be enrolled at this dose. As ≤1 of the first 6 patients had a DLT, dose level 1 was declared the MAD, and 3 additional patients were enrolled to an expansion cohort to inform correlative endpoints. Provisions were made for dose reduction of vaccine (dose level -1; CDX-1401 at 0.5 mg; poly-ICLC at 2 mg), but were not required. Nine patients were enrolled and treated on the study (NCT01834248; ClinicalTrials.gov), which was conducted in accordance with the Declaration of Helsinki and approved by the Roswell Park Cancer Institute (RPCI) Internal Review Board. All patients provided written informed consent. Clinical characteristics are described in Supplementary Table S1.

Peripheral blood was obtained pretreatment, twice weekly, and at end of study (EOS). Bone marrow was collected pretreatment and at the EOS. For extraction of DNA and RNA, CD11b⁺ myeloid cells were isolated from peripheral blood buffy coats using CD11b MicroBeads as per manufacturer's instructions (Miltenyi Biotec). CD11b⁺ cells used as APCs in T-lymphocyte recognition assays were isolated from peripheral blood mononuclear cells (PBMC) following Ficoll centrifugation. CD11b⁺ cells were stained with anti-CD14 and anti-CD15 (Supplementary Table S2). Live cells were determined by staining cells with LIVE/DEAD Fixable Blue Dead Cell Stain Kit (Thermo Fisher Scientific). Cells were analyzed using an LSR II flow cytometer (Becton Dickinson). All raw flow cytometry data were analyzed using FlowJo v.10.2 software (TreeStar).

DNA sequencing of genes commonly mutated in myeloid malignancy was performed on 100 ng of gDNA from bone marrow aspirate samples using the ThunderBolts Myeloid Panel (RainDance Technologies), which covers 49 gene regions using 548 amplicons. Libraries from 16 samples were pooled and sequenced as 2 × 300 bp on a MiSeq using reagent kit v3 (Illumina). Results were analyzed with NextGENe version 2.4.2.1 (SoftGenetics) aligning to GRCh37, making variant calls after limiting to regions of at least 500 read depth coverage, removing known panel artifacts, and restricting to variants with likely missense, in-frame, frameshift, and nonsense functional consequence.

AllPrep DNA/RNA and QIAamp DNA Blood Mini Kits (QIAGEN) were used to isolate genomic DNA from CD11b⁺ cells and plasma, respectively, and sodium bisulfite conversion was performed using the EZ DNA Methylation Kit (Zymo Research). Methylation of the NY-ESO-1 promoter and the LINE-1 repetitive elements was determined by sodium bisulfite pyrosequencing as described previously (27). Primer sequences are shown in Supplementary Table S3.

RNA and cDNA were prepared from CD11b⁺ cells and NY-ESO-1 reverse transcriptase quantitative PCR (qRT-PCR), and RT nested PCR was performed as described previously (14). TaqMan probes and primer sequences are shown in Supplementary Table S3.

ELISA was used to measure anti–NY-ESO-1 antibody titers in patient sera collected pretreatment and at the EOS as described previously (21, 28).

Enzyme-linked immunospot (ELISPOT) analysis was performed on CD4⁺ and CD8⁺ T lymphocytes isolated from PBMCs harvested from patients prior to start of treatment and at EOS as described previously (21). Responses were scored positive when spot numbers in the presence of NY-ESO-1 peptide-pulsed target cells were >25 spots/50,000 cells and were at least 2 times more than the spot count of peptide unpulsed target cells. The average number of spots against unpulsed cells was 21.

Induction of a cytotoxic response in NY-ESO-1–specific CD8⁺ T lymphocytes was performed as described previously (14). Antibodies used in experiments are listed in Supplementary Table S2. NY-ESO-1–specific CD8⁺ T lymphocytes were identified using an NY-ESO-1 tetramer (Ludwig Institute for Cancer Research, Zurich, Switzerland).

To generate NY-ESO-1 p94-104–specific HLA-B35–restricted CD8⁺ T cells, CD8⁺ T lymphocytes isolated from the PBMCs of patient 9 were stimulated with NY-ESO-1 p94-104 peptide-pulsed autologous CD4⁻CD8⁻ cells. After 14 days of culture, the frequency of tetramer⁺ CD8⁺ T cells was 7.2%. Tetramer⁺ CD8⁺ T cells expressed TCR Vb4 that was analyzed using IOTest Beta Mark TCR V beta Repertoire Kit (Beckman Coulter). For enrichment of the tetramer⁺ CD8⁺ T cells, CD8⁺ T cells were labeled with PE-anti-TCR Vb4 antibody, stained with UltraPure anti-PE MACS beads, and sorted by MS column (Miltenyi Biotec). The cells were expanded with PHA in the presence of 30 Gy γ-irradiated normal donor PBMCs and cytokines (IL2 and IL7). After sorting and expansion, tetramer⁺ CD8⁺ cells were enriched to 21.5%.

PBMCs or bone marrow from healthy age-matched donors and patients were stained for 30 minutes with a cocktail of primary antibodies and secondary reagents shown in Supplementary Table S2 and analyzed as described.

We performed a phase I trial of NY-ESO-1 vaccine [DEC205mAb–NY-ESO-1 fusion protein (CDX-1401) with poly-ICLC adjuvant; Celldex Therapeutics] in combination with standard-dose decitabine (20 mg/m²/day × 5 days; Fig. 1A; refs. 5, 23). Decitabine was selected on the basis of our prior work demonstrating a more robust induction of NY-ESO-1 expression compared with azacitidine in preclinical models and patient-derived samples (14, 24). CDX-1401 is a fusion protein consisting of a fully human mAb (HuMab) of IgG1 (kappa) isotype with specificity for the dendritic cell (DC) receptor DEC-205, genetically linked to the full-length NY-ESO-1 tumor antigen (Ag) protein (23). The poly-ICLC adjuvant (Hiltonol; Oncovir) is an experimental viral mimic and broad activator of innate immunity through activation of Toll-like receptor 3 (TLR3; ref. 23).

Eligible patients had intermediate/high-risk MDS by revised International Prognostic Scoring System (IPSS) or low blast count (<30%) AML, were ≥18 years old, had Eastern Cooperative Oncology Group (ECOG) Performance Status ≤2, and had adequate hepatic and renal function (26). No prior azanucleoside exposure was allowed, although prior growth factors therapy was permitted. Patients with uncontrolled medical illness, known HIV positivity, autoimmunity, or recent corticosteroid/radiotherapy were excluded. Nine patients were enrolled and treated on study (6 to the safety cohort and an additional 3 to the expansion cohort).

Patients underwent a baseline bone marrow biopsy with cytogenetics at the time of screening (Supplementary Table S1). A diagnosis was rendered by one of four treating pathologists at RPCI. Baseline transfusion requirements in the 3 months prior to enrollment on study were collected as well as baseline chemistries and complete blood counts for calculation of revised IPSS scores (2, 3). Following enrollment, patients received vaccination on day 14, comprised of 1 mg of CDX-1401 via intracutaneous injection (a mixture of subcutaneous and intradermal administration) with 2 mg of poly-ICLC given subcutaneously within a 5 × 5 cm area on the extremities or the abdomen. Patients then received decitabine 20 mg/m²/day on days 1 to 5 of every cycle, with revaccination on day 15 of each cycle (Fig. 1A). A total of five vaccinations and four cycles of decitabine therapy were planned, and 7 of 9 patients reached EOS; all 9 treated patients received the same therapy.

The most frequent adverse events were deemed related to decitabine or the underlying hematologic malignancy and included cytopenias (predominantly grades 3/4), elevated liver enzymes (grade 3), fatigue (grade 2), edema (grade 2/3), and diarrhea (grade 1/2; Supplementary Table S4). A majority of patients treated on study developed localized skin reactions to the vaccine. These were progressively more prominent with each vaccination and occurred 24 to 48 hours following injection. Two patients did not complete four cycles of therapy due to serious adverse events. Neither of these events was deemed related to protocol therapy. One patient with a history of myocardial infarction developed an in-stent restenosis and recurrent myocardial infarction during the second cycle of therapy (patient 1; Supplementary Table S1). Patient 1 required urgent cardiac catheterization and elected to discontinue vaccine therapy after cycle 1. The patient continued to receive decitabine as standard of care, achieving disease response that allowed her to proceed to allogeneic bone marrow transplant. A second patient (patient 3; Supplementary Table S1) died on protocol cycle 4, day 11 from a terminal intracranial hemorrhage while hospitalized for acute renal failure in the context of sepsis. The patient had received prophylactic low molecular weight heparin per standard hospital policy. Autopsy revealed low-grade MDS without increased blasts. Thus, samples are not available for later time points for these patients. There were no DLTs during the DLT window (Fig. 1A). Data on toxicity during the DLT window are summarized in Supplementary Table S4. These data support the safety of vaccination in combination with decitabine in accordance with our earlier ovarian cancer study (21).

Response assessments using modified International Working Group (IWG) criteria (Supplementary Table S1) were performed based upon the EOS bone marrow biopsy and peripheral blood counts (29). Three patients demonstrated a complete response (CR) to study therapy, 1 patient demonstrated hematologic improvement (HI) for both platelets (-P) and neutrophils (-N), 2 patients had HI-P, and 2 patients has stable disease (SD). One patient had progressive disease at the time of final disease assessment. Molecular profiling revealed TP53 mutations in 3 of 9 patients treated on study (Supplementary Tables S1 and S5).

We determined the effects of decitabine/vaccine combination treatment on global and NY-ESO-1 promoter methylation using DNA isolated from serially collected CD11b⁺ myeloid cells. CD11b⁺ cells were collected from patients' buffy coats and were predominantly comprised of CD14⁺ monocytes and CD15⁺ granulocytes (Supplementary Fig. S1). Methylation of LINE-1–repetitive elements was used as a surrogate for genome-wide methylation. Decitabine therapy resulted in LINE-1 hypomethylation in CD11b⁺ cells compared with samples obtained at diagnosis (Fig. 1B). The methylation nadir occurred between days 8 and 15 of each decitabine cycle. Hypomethylation of LINE-1 was also detected in cell-free DNA isolated from patient plasma. The NY-ESO-1 promoter showed a similar pattern of hypomethylation across all patients (Fig. 1C) in both CD11b⁺ and cell-free DNA. Changes in LINE-1 and NY-ESO-1 methylation were tightly correlated for 8 patients (range of r values = 0.91–0.99, P < 0.01, Spearman correlation).

We then determined NY-ESO-1 expression in circulating CD11b⁺ myeloid cells (Fig. 1D; Supplementary Fig. S2). Using RT-qPCR across all patients, we observed a trend toward increased NY-ESO-1 expression that coincided with the methylation nadir (Fig. 1D). As observed previously, induction of NY-ESO-1 expression was varied among the patients (Supplementary Fig. S2; ref. 13). When examined by nested endpoint PCR, 7 of 9 patients showed induction of NY-ESO-1 during the first cycle of decitabine treatment compared with diagnosis (Supplementary Fig. S2). Patient 6 exhibited baseline expression of NY-ESO-1, but this was not observed throughout treatment (Supplementary Fig. S2). Patients 4 and 5 demonstrated NY-ESO-1 expression during only the first decitabine cycle (Supplementary Fig. S2). In contrast, patients 2, 7, and 9 exhibited induction of NY-ESO-1 expression during multiple cycles, including the first and fourth (final) cycles. These data agree with our previous results in patients with ovarian cancer and AML, which demonstrated that patients receiving decitabine therapy develop hypomethylation of the NY-ESO-1 promoter and induce NY-ESO-1 expression in ovarian cancer cells and circulating AML blasts (14, 21).

To test whether NY-ESO-1 vaccination resulted in the expansion of NY-ESO-1–specific T lymphocytes, we performed ELISPOT assays. Patient T cells isolated at diagnosis and at EOS were stimulated using overlapping NY-ESO-1 peptide pools that spanned the full-length protein. Six of 7 and 4 of 7 patients, respectively, had CD4⁺ T lymphocytes and CD8⁺ T lymphocytes that were responsive to NY-ESO-1 peptides (Table 1).

At diagnosis, patients 1, 2, and 9 showed CD4⁺ lymphocytes that responded to NY-ESO-1 peptides at a level above background (Tables 1 and 2). Patient 1 was negative for NY-ESO-1 expression, and this response at diagnosis may indicate a nonspecific reaction. Although patients 2 and 9 exhibited an NY-ESO-1–responsive CD4⁺ population at diagnosis, the frequency and epitope recognition of these cells increased following vaccination. Patients 5, 6, and 7 exhibited relatively lower frequencies of NY-ESO-1–responsive CD4⁺ lymphocytes directed against a single epitope. As observed in the CD4⁺ response, patient 9 exhibited the highest frequency of NY-ESO-1–responsive CD8⁺ lymphocytes that responded to multiple epitopes at EOS. We observed no significant differences in the frequency of immunosuppressive regulatory T cells (Tregs; CD4⁺, CD25⁺, and FOXP3⁺) in the peripheral blood at diagnosis compared with EOS for any of our study patients (Supplementary Fig. S3). These data indicate that NY-ESO-1 vaccination in combination with decitabine treatment can produce an adaptive immune response in patients with MDS.

Previously, we showed that expression of NY-ESO-1 in circulating blasts from AML patients receiving decitabine was sufficient to activate a cytotoxic T-lymphocyte (CTL) response in an NY-ESO-1–specific CD8⁺ T-cell clone (14). We expanded upon this finding to test whether circulating myeloid cells (presumably from the malignant clone; ref. 30) that express NY-ESO-1 could induce an NY-ESO-1–specific cytotoxic response from T lymphocytes isolated from the same patient.

To validate that the induced level of NY-ESO-1 expression in our patients' myeloid cells was sufficient to activate a cytotoxic response, we cocultured unselected PBMCs isolated from patient 9 with an HLA-compatible NY-ESO-1–specific CD8⁺ T-lymphocyte clone (HLA-B35; ref. 14). Unselected PBMCs isolated at either cycle 1, day 15 (C1D15) or cycle 2, day 15 (C2D15) of decitabine therapy resulted in IFNγ production and upregulation of cell-surface CD107 in clonal T lymphocytes (Fig. 2A). In contrast, unselected PBMCs isolated at diagnosis (prior to decitabine) did not activate a cytotoxic response.

We then enriched HLA-B35⁺ NY-ESO-1–specific T lymphocytes from patient 9 at EOS. These T lymphocytes are comprised of both NY-ESO-1 antigen–specific T cells and polyclonal T cells that do not recognize NY-ESO-1. We tested whether these NY-ESO-1–specific enriched T lymphocytes responded to serially collected autologous CD11b⁺ myeloid cells (30). CD11b⁺ myeloid cells were isolated following Ficoll centrifugation and were comprised predominantly of CD14⁺ monocytes (Supplementary Fig. S1). CD11b⁺ cells from diagnostic samples were unable to activate a cytotoxic response in NY-ESO-1 tetramer⁺ CD8⁺ lymphocytes (Fig. 2B). In comparison, CD11b⁺ myeloid cells isolated at C1D15, C2D15, and EOS were able to induce a cytotoxic response in NY-ESO-1 tetramer⁺ CD8⁺ lymphocytes (Fig. 2B). These time points coincided with expression of NY-ESO-1 (Supplementary Fig. S2). There were no cytotoxic responses in tetramer-negative CD8⁺ lymphocytes (Fig. 2C), demonstrating the specificity of this response for NY-ESO-1 expression. Together, these data indicate that NY-ESO-1 vaccination of patients with MDS resulted in generation of NY-ESO-1–specific CD8⁺ lymphocytes that recognize autologous malignant myeloid cells induced to express NY-ESO-1 by decitabine.

NY-ESO-1–specific humoral immune responses were determined. All patients were seronegative for NY-ESO-1–specific antibodies at diagnosis; patients 2 and 9 became seropositive at EOS (Tables 1 and 2). These patients also exhibited the highest frequencies of NY-ESO-1–specific CD4⁺ and CD8⁺ T lymphocytes at EOS. This result is in contrast to our previous ovarian cancer study in which a majority of patients developed NY-ESO-1–specific antibodies following vaccination (21). As NY-ESO-1 is an intracellular protein, the presence of antibodies is a marker of vaccine response rather than a definitive source of antitumor immune recognition (31).

Patients with myeloid malignancies are known to have poor humoral responses to vaccination despite relatively preserved T-cell immunity (32). At the time of diagnosis, the average percentage of B lymphocytes was 1.91% ± 1.88% of total nucleated compared with a range reported for healthy donors of 6.46% ± 4.76% (33). Five of the 9 patients were evaluated for allogeneic transplant and tested for pretransplant vaccination titers against common viral pathogens (Supplementary Table S6). A majority of the tested patients showed immunity against childhood vaccines including mumps (4/5), polio (5/5), rubella (3/5), rubeola (4/5), and tetanus (5/5). Immunity against influenza A and B was observed in 4 of 5 and 2 of 5 patients, respectively; all patients on the study had received yearly influenza vaccination at least once in the prior 2 years. These titers suggest that memory B-cell function is intact in our MDS patients.

Successful NY-ESO-1 vaccination (with generation of adaptive and serologic responses) requires the presence of DEC-205⁺ APCs, including DCs, which take up the peptide and process the antigen for presentation. On the basis of studies demonstrating decreased numbers of DCs in patients with MDS (34, 35), we determined whether the number of DCs at diagnosis was associated with an antigen-specific response. The frequency of DCs in the peripheral blood was used as a surrogate for the presence of DCs at the site of vaccine administration.

We focused on a population of cDCs that expresses the cell-surface marker CD141 (36–38). CD141^Hi cDCs express high levels of DEC-205 and TLR3. In comparison, CD1c⁺ cDCs express relatively lower levels of DEC-205 and TLR3, and plasmacytoid DCs express low levels of DEC-205 and do not express TLR3 (39, 40). As the adjuvant for our NY-ESO-1 vaccine is a TLR3 agonist, we hypothesized that the presence of a CD141^Hi DC population would be an important modifier of the response to vaccination. There were significantly fewer CD141^Hi DCs (as a percentage of CD45⁺ cells) in the peripheral blood of patients with MDS compared with healthy age-matched controls (Fig. 3A; Supplementary Fig. S4). Only 3 of the 8 patients exhibited a frequency of CD141^Hi DCs that was greater than 0.001% (Fig. 3B). Patients 2 and 9 showed the highest frequency of CD141^Hi cDCs. These patients exhibited the highest response to vaccination as measured by both antibody titers and NY-ESO-1–specific CD4⁺ and CD8⁺ lymphocytes (Fig. 3B).

As DCs are often derived from the malignant clones in patients with MDS, reduced numbers of CD141^HI cDCs could indicate a potential defect in cDC differentiation (34). There were no differences in the frequency of CD1c⁺ cDCs in patients with MDS compared with healthy age-matched controls (Fig. 3C), suggesting adequate differentiation of this population. We also determined whether differences in DEC-205 expression were associated with vaccine response. There was no apparent difference in DEC-205 expression between CD141^Hi cDC populations in patients with MDS (n = 3 detectable) versus healthy controls (n = 8), although in 5 of 8 tested patients, the number of CD141^Hi cDCs was too low for analysis (Fig. 3D). DEC-205 expression in CD1c⁺ cDCs in patients with MDS was significantly higher than in healthy controls. The presence of CD141^Hi cDCs indicates the potential for a robust response to NY-ESO-1 vaccination, as observed in patients 2 and 9, and highlights a potential immunologic defect in patients with MDS.

We determined whether the frequency of cDCs was altered during treatment. We compared diagnostic bone marrow samples to EOS samples for patients 5, 7, 8, and 9 (Fig. 4A). Patients 5 and 8 did not exhibit an increase in CD141^Hi cDC frequency in the bone marrow at EOS (Fig. 4B). Patients 7 and 9 showed an increased frequency of CD141^Hi cDCs at EOS. In contrast to patients 5 and 8, patients 7 and 9 had a prolonged clinical response to decitabine (Supplementary Table S1). Notably, both patients 7 and 9 exhibited a double positive CD141^Hi/CD1c⁺ population at EOS. These CD141^Hi/CD1c⁺ cells expressed similar levels of CLEC9A, DEC-205, and HLA-DR compared with CD141^Hi, CD1c⁺ cells (Supplementary Fig. S5). The frequency of CD1c⁺ cDCs also increased for 3 of 4 patients (Fig. 4C).

Immunotherapeutic approaches that utilize aberrant tumor-specific expression of CTAs have shown clinical promise due to the immunogenicity of these antigens and their relative low-level expression in normal tissues (16, 18, 23, 28). In myeloid malignancies such as MDS or AML, CTAs are not expressed in the malignant compartment due to promoter methylation (14, 15). We and others have demonstrated that standard-of-care regimens with hypomethylating agents induce expression of CTAs at a level sufficient to activate an antigen-specific CTL response (14). Here, we report for the first time in a phase I trial that vaccination against the NY-ESO-1 CTA is safe and induces an antigen-specific immune response in MDS patients receiving decitabine. Thus, our study demonstrates the feasibility of this approach and highlights specific features of the immunologic milieu in patients with MDS that might be manipulated in future studies.

In agreement with previous reports, decitabine monotherapy was sufficient to induce hypomethylation of the NY-ESO-1 promoter and induce gene expression in CD11b⁺ myeloid cells in the majority of patients (15, 17). This molecular response is analogous to results from our prior study of induced NY-ESO-1 expression observed in peripheral AML blasts following clinical decitabine (14). Both studies showed variance in the kinetics of gene expression across individual patients, although the timing of NY-ESO-1 expression was not associated with the magnitude of response. It is unlikely that differences in NY-ESO-1 expression are entirely due to pharmacodynamic effects, as all patients showed a similar pattern of global hypomethylation following decitabine. Cell-free DNA and DNA isolated from CD11b⁺ myeloid cells exhibited identical patterns of global and gene-specific hypomethylation during treatment, suggesting that changes in DNA methylation in response to hypomethylating agents could be assessed using DNA isolated from plasma rather than the cellular elements.

Our observation that DEC-205 expression is high in cDCs from patients with MDS suggests that the variances in NY-ESO-1–specific humoral and adaptive immune responses following vaccination were not due to inadequate receptor expression. Patients with the highest frequency of CD141^Hi cDCs showed the strongest response to vaccination. CD141^Hi cDCs express higher levels of TLR3 than other DC populations, supporting the hypothesis that this population would be preferentially activated by the poly-ICLC adjuvant, a TLR3 agonist (39, 40). Both quantity and quality of CD141^Hi cDCs may be important. Data showing that CD141^Hi cDCs were lower in patients with MDS compared with healthy controls are similar to those reported by Dickinson and colleagues (41). Previous studies have indicated that DC function is decreased in patients with MDS, which may explain why patient 6 did not develop NY-ESO-1 antibodies despite detectable numbers of these cells pretreatment (42). Although we did not observe any difference in the number of CD1c⁺ cDCs in patients with MDS compared with healthy controls, the contribution of these cells to the vaccine response in our MDS patients remains unclear.

Our observation that CD141^Hi cDC frequencies can increase during the course of treatment suggests that the optimal approach for some patients may involve vaccination after several cycles of treatment to increase the size of the appropriate APC population. The biological significance of the double positive CD141^Hi/CD1c⁺ population observed following treatment is unclear. This population is present in healthy volunteers receiving Flt3L and represents an expanding cDC population (43, 44). It is possible that administration of poly-ICLC increased Flt3L levels and, thus, raises the question of whether patients receiving our combination therapy have increased Flt3L signaling (45). Additional studies will be required to determine the ability of these cDCs to function as APCs.

In this study, we have demonstrated the feasibility of activating an endogenous immune response against an azanucleoside-induced target. The common clinical use of azanucleosides in this patient population encourages such an approach. Observed responses were less robust than those seen in solid tumor patients with endogenous gene expression receiving the same vaccine. In addition to our observations that specific DC populations are decreased in patients with MDS, there are several reports on the increased numbers of immunosuppressive cells such as Tregs and myeloid-derived suppressor cells in patients with MDS (46, 47). Thus, there are multiple immunologic mechanisms that could dictate response to vaccination, and elucidation of these mechanisms will require a larger study. Azanucleosides have also been demonstrated to increase expression of immune checkpoint proteins in both myeloid blasts as well as T lymphocytes of patients with MDS and AML (48, 49). The inclusion of immune checkpoint inhibitors such as nivolumab could block this effect to further enhance the response to vaccination. This study highlights the potential for future combinations aimed at enhancing this type of response.

E.A. Griffiths reports receiving commercial research grants from Astex Pharmaceuticals and Genentech, speakers bureau honoraria from Alexion Pharmaceuticals, and is a consultant/advisory board member for Alexion Pharmaceuticals, Celgene, Otsuka, Inc., and Pfizer. C.S. Hourigan reports receiving other commercial research support from Merck and Sellas, and is a consultant/advisory board member for Janssen. No potential conflicts of interest were disclosed by the other authors.

Conception and design: E.A. Griffiths, P. Srivastava, A.R. Karpf, M.J. Nemeth

Development of methodology: E.A. Griffiths, P. Srivastava

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): E.A. Griffiths, P. Srivastava, J. Matsuzaki, Z. Brumberger, E.S. Wang, J. Kocent, G.W. Roloff, H.Y. Wong, K. Odunsi, C.S. Hourigan

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): E.A. Griffiths, P. Srivastava, Z. Brumberger, A. Miller, H.Y. Wong, C.S. Hourigan, M.J. Nemeth

Writing, review, and/or revision of the manuscript: E.A. Griffiths, P. Srivastava, J. Matsuzaki, Z. Brumberger, E.S. Wang, A. Miller, A.R. Karpf, C.S. Hourigan, M.J. Nemeth

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): J. Kocent, B.E. Paluch, L.G. Lutgen-Dunckley, B.L. Martens

Study supervision: E.A. Griffiths, M.J. Nemeth

This work was supported by the Louis M. Sklarow Foundation (to E.A. Griffiths), Alliance Developmental Awards from the Alliance Foundation (to E.A. Griffiths and M.J. Nemeth), NCI Cancer Center Support Grant CA016056, Institutional National Research Service Award 5T32CA009072-39 (to B.E. Paluch), the Rapaport Family Foundation (to E.A. Griffiths and M.J. Nemeth), institutional funds provided by RPCI (to M.J. Nemeth and E.A. Griffiths), by the NIH Medical Research Scholars Program, and by the Doris Duke Charitable Foundation (grant #2014194; to G.W. Roloff), and by funds from the Intramural Research Program of the National Heart, Lung, and Blood Institute of the NIH (to G.W. Roloff, H.Y. Wong, and C.S. Hourigan).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.