Integrated Human and Murine Clinical Study Establishes Clinical Efficacy of Ruxolitinib in Chronic Myelomonocytic Leukemia

This article is featured in Highlights of This Issue, p. 6067

Rare cancers, defined by the International Rare Cancers Initiative as those having an incidence of less than 6 per 100,000, cumulatively account for 25%–30% of all cancer diagnoses and the majority of adult hematologic malignancies (1). Although their incidence is rapidly increasing secondary to molecular subtyping of common cancers (2), preclinical studies and clinical trials aimed specifically at therapeutic development represent a far lower proportion of cancer research (3). Unsurprisingly, rare cancers display adverse outcomes and limited therapeutic options compared with their more common counterparts (2). Moreover, even if an attractive therapeutic strategy is preclinically identified, adequately powered phase III clinical trials are challenging to complete and often never performed because they are deemed intractable. Therefore, novel developmental strategies are required to improve outcomes and expand treatment options for the increasing population of patients afflicted by rare cancers.

Chronic myelomonocytic leukemia (CMML) is a rare clonal myeloid neoplasm characterized by monocytosis, ineffective hematopoiesis, and a propensity for transformation to acute myeloid leukemia (AML; refs. 4, 5). Initially described in 1937, CMML is now formally recognized as a myelodysplastic syndrome/myeloproliferative neoplasm (MDS/MPN) overlap syndrome by the World Health Organization (WHO; refs. 6–10). Although the hypomethylating agents (HMA) azacitidine and decitabine are approved by the Food and Drug Administration for the treatment of CMML, their approval was based on studies that enrolled predominantly patients with MDS rather than CMML (11–16). Few CMML-specific studies have tested novel therapies, and only two phase III clinical studies have been reported in this disease to date, in part, due to incomplete understanding of pathobiology (17, 18).

We have demonstrated that CMML is molecularly characterized by hematopoietic stem and progenitor cell hypersensitivity to granulocyte-macrophage colony-stimulating factor (GM-CSF), representing a convergent molecular pathway with potential for therapeutic intervention (19–21). We have also demonstrated that CMML cells can robustly xenograft in immunocompromised mice that secrete human GM-CSF, IL3, and SCF and recapitulate many features of the human disease (22). As the sentinel kinase required for GM-CSF signaling, JAK2 represents a rational therapeutic target in CMML and pharmacologic inhibition has demonstrated activity in preclinical study (21, 23). We therefore hypothesized that ruxolitinib, a potent JAK1/2 inhibitor approved for the treatment of several MPNs, may be a novel therapeutic in CMML.

To test this, we conducted a phase I and II multi-institution clinical study testing ruxolitinib in CMML. We have previously reported the results of the phase I portion of the study (24), but herein we report the final study results, including the addition of 30 patients treated in phase II with a protracted duration of follow-up in all study participants. In addition to performing laboratory correlative studies aimed at identifying molecular signatures of response, we generated patient-derived xenograft (PDX) models from subjects enrolled on this clinical trial and used these PDX models to recapitulate clinical responses and confirm disease modification by the study drug. This integrated murine and human approach was designed to more rigorously confirm the activity of ruxolitinib and support its future development in this rare cancer.

This multi-institution study consisted of a combined phase I and II clinical trial, as detailed in Supplementary Fig. S1. The phase I portion followed a dose escalation, “rolling 6” design (25). Patients were allocated to starting doses of 10 mg/day to 40 mg/day, divided in two equal doses, escalated by 10 mg/day for each cohort according to a predetermined algorithm. The phase II study was a Simon's two-stage design with a total of 10 patients in the first stage and 20 patients accrued in the second if more than one clinical response was observed. The dose of ruxolitinib in phase II was 40 mg/day divided in two equal doses. Treatment was continued until progression or unacceptable toxicity, and in patients with stable disease was discontinued at the discretion of the treating physician. The study was conducted in accordance with the Declaration of Helsinki and was approved by ethical boards at each institution. Written informed consent was obtained from all subjects prior to enrollment. The study was registered at clinicaltrials.gov (NCT01776723).

Inclusion criteria included a confirmed diagnosis of CMML using the WHO 2016 classification and the capacity to provide adequate bone marrow specimens for histopathologic and cytogenetic analyses. Previous therapy for CMML was allowed, but not required, presuming a 28-day washout period. Laboratory values that excluded patients from the study included a platelet count of <35,000 cells/dL, an absolute neutrophil count (ANC) <250 cells/dL, a serum creatinine >2.0 mg/dL, and a serum total bilirubin >1.5 times the upper limit of laboratory normal. Concurrent use of myeloid growth factors was prohibited.

Response to therapy was assessed using the proposed International Working Group (IWG) response criteria for MDS/MPNs (Supplementary Table S1; ref. 26). To evaluate responses, peripheral blood counts were assessed weekly during the first four cycles, with spleen size assessed every four weeks by physical exam and bone marrow aspirate and biopsies performed after cycles two and four. For patients continuing therapy beyond four cycles, spleen size, and peripheral blood counts were assessed monthly, and bone marrow aspirate and biopsies were repeated every four cycles. Symptoms were evaluated daily in a subset of patients who were asked to complete the MPN Symptom Assessment Form Total Symptom Score (MPN-SAF TSS), which was utilized in exploratory analyses (27). However, because symptoms were not assessed in all patients, symptom responses were not included in the response definition for this study.

Demographic and clinical variables for study patients were summarized using descriptive statistics. The Fisher exact test and χ² test were used to analyze categorical variables and the Mann–Whitney, Wilcoxon matched-pairs, and Kruskal–Wallis (with Dunn test used to adjust for multiple comparisons) tests were used for quantitative variables. The Kaplan–Meier method was used to calculate overall survival (OS) curves and the log-rank test used to determine if curves were significantly different. Landmark analyses were performed when comparing survival between response groups, with a 4-month landmark utilized. P values were two-sided and P ≤ 0.05 were considered statistically significant.

Somatic mutations were identified prior to treatment using a 49-gene targeted next-generation sequencing panel (RainDance Technologies; cat. #20-07218) sequenced on a NextSeq instrument (Illumina, Inc.). Library preparation and variant calling strategy were as previously described (24).

Plasma was collected from the supernatant of ficolled blood mononuclear cells, and a custom 46-plex assay (Luminex Corporation) was used to profile baseline cytokine levels as previously described (28).

NOD-scid IL2Rgnull-3/GM/SF (NSGS) mice were bred under pathogen-free conditions (JAX Stock #013062). Two to five million bone marrow mononuclear cells (BMNC) from unique screening patient samples were transplanted via tail vein into 4–6 sublethally irradiated, 6-week-old female NSGS mice (22). The number of mice per cohort was determined based on the number of BMNCs available. Animals were housed in accordance with institutional standards set by Moffitt Cancer Center and the University of South Florida (USF). All procedures were approved by the Institutional Animal Care and Use Committee of USF (protocol IS00004320).

Seven to 14 days after transplant, mice were randomized 1:1 to receive vehicle (5% N,N-dimethylacetamide solution in 0.5% methylcellulose) or pharmacokinetically equivalent doses of ruxolitinib (60 mg/kg) orally twice daily. Because peripheral blood engraftment is a heterogeneous and late event in CMML PDX models and cannot, therefore, be used to establish engraftment onset in real time, engraftment was estimated using previously engrafted comparable models and via magnetic resonance imaging (MRI) of interval splenomegaly (29). Treatment continued until mice became moribund and were euthanized. The primary endpoint of the study was survival, with secondary response assessments to include leukemic cell engraftment in target organs, spleen weight, spleen volume, and hematologic parameters. Spleen volume was evaluated by axial T1-weighted MRI, which was performed seven days after transplant and when the first cohort mouse became moribund. Human CD45-selected splenocytes and whole bone marrow were collected for further analysis.

Leukemic engraftment was determined by FACS flow cytometry using an LSR II (BD Sciences). The percentage of human CD45 was used to define engraftment in each target organ (BD cat. #564047, RRID:AB_2744403). All antibodies were obtained from BD or BioLegend. For all flow analyses, FlowJo v8 was used (RRID:SCR_008520). All gating was done to exclude non-singlets and dead cells.

A total of 50 patients were enrolled between March 2013 and October 2016, with baseline characteristics summarized in Table 1. The first 20 patients were enrolled to the phase I study portion and the remaining 30 to the phase II portion. The median age of all patients was 69.5 years (range, 42–86 years), and 62% of patients were male. A total of 66% of CMML patients were lower risk (low or intermediate-1 categories) by the Global MD Anderson Scoring System (30); 56% of patients were classified as CMML-0, 32% as CMML-1, and 12% as CMML-2 per the WHO classification. Approximately half of patients enrolled had palpable splenomegaly and 68% were classified as MPN-CMML by the FAB classification (7). Thirty percent of patients on study had previously been treated with an HMA. Following study treatment, two patients proceeded to allogeneic hematopoietic cell transplantation.

Treatment-emergent adverse events, defined as those deemed at least possibly related to study drug by investigators, were similar to the phase I report (24). Grade 3/4 anemia was observed in five patients, and three patients sustained grade 3/4 thrombocytopenia. No other treatment-emergent grade 3/4 toxicities were recorded more than twice (Supplementary Table S2). The full list of nontreatment and treatment-emergent toxicities is available in Supplementary Table S3. The majority of patients who discontinued ruxolitinib did so secondary to disease progression (31 of 50 patients). Seven patients discontinued therapy secondary to adverse events or intolerance and three because of lack of response (Supplementary Table S4). A total of seven patients (23.3%) in the phase II portion of the study required dose reductions, predominantly secondary to hematologic toxicities. There were three deaths on therapy; however, none were attributed to ruxolitinib therapy. Eight patients (16%) progressed to AML at a median of 5.1 months from ruxolitinib initiation and 29.8 months from diagnosis. Five remained on ruxolitinib at the time of transformation.

As of August 25, 2020, the mean duration of therapy (DOT) was 365 days (range, 6–1981 days). A total of 38% (19 of 50) of patients achieved a clinical response by MDS/MPN IWG response criteria. Five hematologic responses (two erythroid and three platelet) were noted in addition to one complete remission, two partial remissions, three optimal marrow responses, and one partial marrow response, as summarized in Table 2. Eight patients were transfusion dependent at baseline with one achieving transfusion independence >8 weeks and three demonstrating a ≥50% reduction in transfusion burden (one spleen response, one spleen and platelet response, and one partial marrow response). Among 23 patients with baseline splenomegaly, 10 (43.5%) achieved a spleen response, with three such patients also experiencing hematologic improvement. An additional 44% (22 of 50) of patients achieved stable disease as best response, defined as not meeting criteria for response or progressive disease. Response rates (P = 0.77) and proportions of stable disease (P = 0.25) were not different between phase I and II. Similarly, no differences in response rates were observed based on WHO subtype (P = 0.41), FAB subtype (P = 0.76), prognostic risk group (P = 0.55 using Global MDACC Score, P = 0.35 using IPSS-R), or prior exposure to HMA (P = 0.35).

The median time to best response was 3.7 months, with duration on therapy and timing of individual responses depicted in Fig. 1. Among responding patients, the median duration of response was 7.7 months. A significant difference in treatment duration (P = 0.0002) was noted when accounting for treatment response, with a median DOT of 375 days in responders, 152 days in patients with stable disease, and 69 days in patients with progressive disease. Notably, 14 of 22 patients (63.6%) with stable disease at best response remained on therapy for over three months.

Because the MDS/MPN IWG response criteria and its incorporation of the MPN-SAF TSS in CMML response were established only after accrual began, symptom improvements were not considered in the overall response rates and the MPN-SAF TSS was not captured in most patients (26, 27). However, in 11 patients in whom the MPN-SAF was available and a score ≥20 (specified by MDS/MPN IWG criteria) was noted at baseline, a statistically significant reduction in TSS was observed at day 30 (median baseline TSS 30 vs. 21 at day 30, P = 0.01; Fig. 2A), and seven patients (four of whom did not achieve a clinical response by study criteria) demonstrated a ≥50% decrease in TSS at any point during follow-up. Notably, a decrease in TSS was seen across response categories, with median percent decreases at day 30 of 41.2%, 38.5%, and 23.1% observed in patients with clinical response, stable disease, and progressive disease, respectively (Supplementary Fig. S2).

The median OS of ruxolitinib-treated patients from the time of study entry was 24.7 months with a median follow-up of 52 months (Supplementary Fig. S3). A landmark analysis was performed using a 4-month landmark to evaluate the relationship between OS and response to therapy. A median OS of 28.8 months was observed in responding patients, compared with 22.3 months in patients with stable disease, and 22.3 months in patients with progressive disease (P = 0.58; Fig. 2B). The median OS of patients who had received prior treatment with an HMA was 23.7 months compared with 31.2 months in those who had not received prior HMA therapy (HR 1.55, P = 0.20; Fig. 2C).

In related myeloid neoplasms, and consistent with the observed time to best response in this study, clinical responses to ruxolitinib largely plateau after approximately three months on treatment (31–34). Because the median duration on treatment was greater than 3 months (152 days) in patients with stable disease, formal symptom assessment was not available in most patients, and duration on treatment has been used as a metric of clinical benefit in other ruxolitinib studies, we evaluated the clinical impact of DOT on CMML outcomes. Indeed, DOT was associated with an OS improvement that was most pronounced in patients with stable disease. Patients with stable disease and a treatment duration ≥3 months demonstrated improved OS compared with those with a treatment duration <3 months (median OS 32.8 vs. 7.6 months, HR 0.25, P = 0.002). Further, patients with a clinical response or stable disease and a duration on therapy lasting ≥3 months (31 patients, 62%) demonstrated a greater percent decrease in MPN-SAF TSS at day 30 (43.2% vs. 23.1%, P = 0.05; Fig. 2D) and improved OS (median OS 32.8 months vs. 7 months, HR 0.36, P < 0.001; Fig. 2E) compared with patients who did not meet these parameters.

Previous clinical studies in other myeloid neoplasms have not demonstrated a significant impact of ruxolitinib therapy on mutant allele burden in the majority of patients, but have demonstrated a substantial reduction of inflammatory cytokine secretion secondary to the JAK1 inhibitory properties of ruxolitinib (24, 35). These studies motivated us to determine whether specific somatic mutations or cytokine secretion patterns were associated with ruxolitinib response in CMML. We tested 49 recurrently mutated genes in 49 of 50 patients with an identical assay in a central laboratory. The most common mutations were TET2 (61%), SRSF2 (47%), ASXL1 (20%, frameshift or nonsense only), and NRAS (20%) consistent with known mutation frequencies in CMML (Fig. 3). Defining response as clinical benefit or better by the MDS/MPN IWG response criteria, no individual somatic mutation or functional class of mutations, such as epigenetic (i.e., TET2, IDH2, ASXL1, and EZH2), splicing (i.e., SRSF2, SF3B1, and ZRSR2), or signaling (N-/K-RAS, CBL, PTPN11), correlated with response. Similarly, no comutation pattern was statistically associated with response, though no responses were observed in SRSF2/ASXL1-comutated patients (observed in six nonresponders, P = 0.07).

We profiled 46 cytokine and chemokine peripheral blood plasma levels at screening in 48 study participants, selected based on our previous work exploring the impact of cytokines and chemokines on clinical and genetic features in CMML (28). When comparing baseline cytokine secretion patterns in those CMML patients achieving clinical benefit or better, none of those evaluated was found to correlate with likelihood of response after adjusting for multiple comparisons (Supplementary Fig. S4).

We have previously demonstrated that PDX models recapitulate the clinical, genetic, and pathologic features of human CMML (22). However, whether PDX models can recapitulate clinical responses to therapy from an individual patient is unknown across cancer. To address this and validate the clinical efficacy of ruxolitinib seen in this study, we xenografted BMNCs collected at the time of screening for 13 patients (Fig. 4A; Supplementary Table S5). Patient samples used were selected based on the availability of sufficient cryopreserved screening BMNCs (>5.0 × 10⁷ cells).

A total of 49 CMML PDX mice were generated, and the mean DOT was 134 days (range, 7–390 days). The median OS of all PDX mice, irrespective of the patient-specific response, was 150 days in the ruxolitinib cohort versus 86 days for the vehicle cohort (HR 0.81, P = 0.60; Supplementary Fig. S5A). Leukemic engraftment was assessed by bone marrow hCD45 content (median 15.6% in ruxolitinib-treated mice vs. 33.5% in vehicle-treated mice, P = 0.19; Supplementary Fig. S5B), and change in spleen volume was assessed by MRI (32.2 mm³ vs. 57.9 mm³, P = 0.11; Supplementary Fig. S5C). PDX mice were then stratified based on clinical response observed in the corresponding patient from which they were derived to assess whether pharmacologic activity in PDX mice paralleled that observed in humans. Indeed, mice derived from responding patients demonstrated a significant improvement in calculated median OS with ruxolitinib therapy when compared with intrapatient vehicle (+11 days, P = 0.006), which was not observed in mice derived from nonresponding patients (−4.15 days, P = 0.88; Fig. 4B.i). A greater reduction in leukemic engraftment was also observed in ruxolitinib-treated versus vehicle-treated mice derived from responders (median bone marrow hCD45⁺ 13.9% vs. 36%, P = 0.054) compared with that observed in PDXs derived from nonresponding patients (16% vs. 25.5%, P = 0.87; Fig. 4C.i). Both groups exhibited a decrease in spleen volume with ruxolitinib therapy; however, these findings were more consistently observed in mice derived from responding patients (−14.35 mm³, P = 0.006) than in mice derived from nonresponders (−27.7 mm³, P = 0.01; Fig. 4D.iand E). Additional metrics such as blood counts and spleen weight did not statistically associate by treatment groups (Supplementary Fig. S6).

In support of observations made in the human trial (Fig. 2D and E), PDX mice derived from pretreatment patient samples that experienced clinical response or stable disease and ultimately remained on therapy for ≥3 months demonstrated a significant improvement in OS (median +11 days compared with the average of intrapatient vehicle, P = 0.02; Fig. 4B.ii), a reduction in leukemic engraftment (median hCD45⁺ 13.9% vs. 36%, P = 0.04; Fig. 4C.ii), and a reduction in spleen volume (median reduction −14.4 mm³, P = 0.0007; Fig. 4D.ii) when treated with ruxolitinib compared with vehicle. In contrast, these findings did not extend to mice derived from patients who did not meet these parameters except for a reduction in spleen volume, which was not uniformly observed across all ruxolitinib-treated mice (Fig. 4B–D.ii). These data suggest that the observed survival differences based on DOT are due to a therapeutic effect and not simply a result of selection bias elicited by variable disease biology.

Interestingly, vehicle-treated PDXs derived from patients with a clinical response experienced a significant decrease in median survival (52 vs. 185 days, P = 0.04) and trend toward increased leukemic engraftment (36% vs. 25.5%, P = 0.37) when compared with vehicle-treated PDXs derived from patients who did not achieve response (Fig. 4F). Using a receiver-operating characteristic curve analysis, survival of vehicle-treated mice was found to be predictive of the corresponding patient demonstrating a clinical response on study (P = 0.04), with a cutoff of <56 days demonstrating the greatest predictive value (sensitivity 72.7%, specificity 92.9%; likelihood ratio 10.2; Supplementary Fig. S7). Given that our murine models secrete the human cytokines IL3, GM-CSF, and SCF, and that ruxolitinib results in profound reduction of intracellular cytokine signaling, this observation suggests that untreated CMML PDX models may identify a subset of CMML patients who are both cytokine sensitive and responsive to ruxolitinib (35, 36).

Through integration of clinical trial outcomes with therapeutic PDX modeling from patients on study, we establish that “first passage” PDX modeling can recapitulate clinical trial response in CMML patients treated with ruxolitinib. Several studies have associated response of PDX models with reported response rates seen in unrelated human clinical trials (37, 38). However, to our knowledge, this is the first report demonstrating that PDX models can recapitulate the clinical response seen in an individual patient from which the model originated and who received the same therapeutic intervention on protocol. This observation needs to be validated with other agents, but provides evidence for use of PDX modeling as a novel clinical correlative aimed at corroborating clinical response and biological activity in single-arm human studies of rare cancers. We believe these PDX models, and their capacity for randomization to placebo, can facilitate “go-no-go” decisions required to move agents forward in human interventional trials. In addition, the phenotypic features of vehicle-treated PDX models were associated with response to ruxolitinib, suggesting that our models may provide insights into clinical responses without therapeutic challenge and functionally identify a cytokine-sensitive CMML subset.

The current study demonstrates that ruxolitinib displays clinical activity and may represent a useful therapeutic agent for patients with CMML. Although there appeared to be an enrollment bias toward MPN-CMML patients, responses were observed across disease subgroups. Importantly, prior exposure to HMA therapy did not significantly affect treatment response, with clinical benefit observed in both treatment-naïve and HMA-exposed patients alike. The median OS of patients who had received HMA therapy prior to enrollment was 23.7 months, considerably longer than what has historically been reported in this population (39–41). This is of particular clinical relevance given the reported therapy-refractory nature of post-HMA treatment CMML and paucity of active agents in this clinical setting.

Improvements in constitutional symptoms have been demonstrated to be a major clinical benefit associated with ruxolitinib therapy in the treatment of MPNs (31, 42). Although they were not captured in all patients in the current study, improvements in symptom scores were seen in the majority of assessable patients, independent of clinical response. Although included in the IWG MDS/MPN response criteria, symptom improvement was not used to define response in the current study, suggesting clinical benefit may extend beyond the reported overall response rate. Similarly, protocol-directed imaging to assess spleen volume was not performed, making it difficult to determine the true magnitude of benefit. In the PDX experiment, reductions in spleen volume were demonstrated across response groups, consistent with clinical studies in myelofibrosis in which the vast majority of patients experienced some degree of reduction in spleen size (31, 42).

Further supporting the clinical activity of ruxolitinib in CMML, we observed that nearly two thirds of patients with stable disease remained on therapy for over 3 months. When incorporated along with responding patients with an equivalent treatment duration, this group of patients demonstrated a greater improvement in symptom scores and improved OS. While recognizing the potential bias in such analyses, these findings were replicated in our randomized PDX studies, supporting their clinical relevance. Further, vehicle-treated mice derived from patients with clinical response actually demonstrated shorter OS than those from nonresponders, indicating that these findings are not due to favorable disease biology alone.

This study also highlights several differences between the safety and efficacy of ruxolitinib in CMML compared with other MPNs. Although improvements in splenomegaly and disease-related symptoms were observed, as seen in primary myelofibrosis and polycythemia vera (31, 43), hematologic improvements and bone marrow responses were more common in CMML. Furthermore, ruxolitinib-induced myelosuppression was uncommon despite enrolling patients with pretreatment anemia and thrombocytopenia. Given these differences, our study suggests that the quality and nature of responses in CMML may be aligned more closely with those observed in CSF3R-mutant CNL (44, 45) than those reported in primary myelofibrosis and polycythemia vera (31, 43). The reason for this finding remains unclear, but likely represents the unique biology of CMML and a differential role of activated JAK–STAT signaling in disease pathogenesis in comparison with MPNs. The nature of ruxolitinib responses in CMML patients will continue to be assessed in an ongoing phase II expansion study.

In addition to evaluating efficacy in PDX and human clinical study, we report two biologically informative assays as correlates to this study. We performed mutational profiling of virtually all cases and identified no associations with individual gene mutation or functional class of mutation and response to ruxolitinib. Additionally, we performed broad cytokine profiling at time of screening and found that none of the assessed cytokines was predictive of response. Although these analyses, particularly those evaluating individual mutations, are limited by sample size, this further highlights the importance of our PDX experiments as a novel correlative assay.

As with many rare cancers, the therapeutic armamentarium in CMML is limited and disease-specific therapies are required to improve outcomes. The HMAs remain the standard of care at this time, particularly in patients with cytopenias or elevated bone marrow blasts, based on their established activity. Recent work has demonstrated that the addition of venetoclax to HMAs may further improve outcomes in patients with CMML, yet limitations remain and this therapy may not be appropriate in all settings (46). In this study, we demonstrate that ruxolitinib is efficacious in a subset of patients and represents a novel therapy that may be beneficial in HMA-refractory patients and those with splenomegaly or symptomatic proliferative features.

Although a larger, randomized study would be ideally performed to confirm the disease-modifying activity of ruxolitinib, this is challenging and often intractable in rare cancers like CMML. Therefore, this integrated human and murine study provides evidence for ruxolitinib as a viable therapeutic in CMML and proof of principle that this approach may be more broadly used in rare cancers in which randomized clinical trials and therapeutic development have been historically challenging.

H. Newman reports grants from Incyte Corporation, NCI-R37, and Evans Foundation during the conduct of the study. A.E. Dezern has received honoraria from Taiho, AbbVie, and Novartis. D.P. Steensma reports other support from Novartis outside the submitted work. G.J. Roboz reports grants from MDS Clinical Research Consortium during the conduct of the study; personal fees from AbbVie, Actinium, Agios, Astex, Amgen, Astellas, AstraZeneca, Bayer, Blueprint Medicines, BMS, Celgene, Daiichi, GSK, Helsinn, Janssen, Jasper, Jazz, MEI, Mesoblast, Novartis, Otsuka, Pfizer, Roche/Genentech, Sandoz, and Takeda outside the submitted work. A. Gerds reports personal fees from Incyte and other support from Incyte during the conduct of the study; personal fees and other support from Pharmessentia, Celgene/BMS, Kartos Therapeutics, CTI Biopharma, Constellation, and Sierra Oncology and personal fees from Pfizer, Premedior, Novartis, and AbbVie outside the submitted work. D.A. Sallman reports personal fees from AbbVie, Agios, Aprea, BMS, Incyte, Intellia, Kite, Magenta, Novartis, Shattuck Labs, Syndax, and Takeda and grants from Celgene outside the submitted work. C. Letson reports grants from Incyte Corporation, Evan's Foundation, and NCI during the conduct of the study. M.E. Balasis reports grants from Evans Foundation, Incyte Corporation, and NCI during the conduct of the study. J.E. Lancet reports personal fees from Jazz Pharmaceuticals, Agios, AbbVie, Guidepoint, Elevate Bio, BerGen Bio, Bristol Myers Squibb, and Takeda Pharmaceuticals outside the submitted work. A.F. List reports personal fees from Precision Biosciences, Halia Therapeutics, and CTI Biopharma and other support Aileron and CBMG outside the submitted work. M.A. Sekeres reports personal fees from BMS, Novartis, and Pfizer outside the submitted work. R.S. Komrokji reports personal fees from JAZZ, BMS, AbbVie, Novartis, Acceleron, and Geron outside the submitted work. E. Padron reports grants from Incyte during the conduct of the study; grants from Kura and grants from BMS outside the submitted work. No disclosures were reported by the other authors.

A.M. Hunter: Data curation, formal analysis, investigation, writing–original draft, writing–review and editing. H. Newman: Formal analysis, investigation, writing–original draft, writing–review and editing. A.E. Dezern: Supervision, investigation, writing–review and editing. D.P. Steensma: Supervision, investigation, writing–review and editing. S. Niyongere: Investigation, writing–review and editing. G.J. Roboz: Supervision, investigation, writing–review and editing. Q. Mo: Formal analysis, writing–review and editing. O. Chan: Formal analysis, writing–review and editing. A. Gerds: Investigation, writing–review and editing. D.A. Sallman: Supervision, investigation, writing–review and editing. W. Dominguez-Viqueira: Formal analysis, writing–review and editing. C. Letson: Formal analysis, investigation, writing–review and editing. M.E. Balasis: Formal analysis, investigation, writing–review and editing. M. Ball: Formal analysis, investigation, writing–review and editing. T. Kruer: Formal analysis, investigation, writing–review and editing. H. Zhang: Formal analysis, investigation, writing–review and editing. J.E. Lancet: Supervision, investigation, writing–review and editing. A.F. List: Supervision, investigation, methodology, writing–review and editing. M.A. Sekeres: Supervision, investigation, writing–review and editing. R.S. Komrokji: Conceptualization, data curation, formal analysis, supervision, investigation, methodology, project administration, writing–review and editing. E. Padron: Conceptualization, resources, data curation, formal analysis, supervision, funding acquisition, investigation, methodology, writing–original draft, project administration, writing–review and editing.

This study was performed on behalf of the MDS Clinical Research Consortium and was supported in part by the Small Animal Imaging Lab (SAIL) Core, Flow Cytometry Core, and the Molecular Genomics Core at the H. Lee Moffitt Cancer Center and Research Institute (P30-CA076292), the NCI (R37-CA234021), Incyte Corporation, and the Edward P. Evans Foundation. We thank Dr. Omar Abdel-Wahab for his thoughtful review and edits to this manuscript.

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.