Megakaryopoiesis impairment through acute innate immune signaling activation by azacitidine

Myelodysplastic syndrome (MDS) and acute myeloid leukemia (AML) are hallmarked by ineffective hematopoiesis, generation of dysfunctional myeloid cells, and single or multi-lineage cytopenia, including thrombocytopenia (Will et al., 2012).

The cytosine analog and DNA hypomethylating agent (HMA) azacitidine (AZA) is one of the current therapies for MDS and is used in AML for individuals who are ineligible for stem cell transplantation or high-dose myeloablative chemotherapy (Fenaux et al., 2009, National Comprehensive Cancer Network, 2017). While this cytosine analog effectively reverses disease-associated DNA methylation and alleviates dysregulated myeloid differentiation, it is not curative (Christman et al., 1983). Its clinical application is largely limited by its bone marrow (BM)–suppressive effects, including neutropenia and thrombocytopenia (Kantarjian et al., 2007; Aparicio and Weber, 2002). How AZA elicits these inhibitory effects has remained incompletely unsolved, hindering its effective use as an anti-leukemic therapy, particularly in MDS/AML.

Thrombocytopenia, platelet (PLT) counts of <100 × 10⁹/liter, occurs in 40–65% of patients with MDS (Neukirchen et al., 2009). It can increase the risk of bleeding and hampers the effective clinical management of patients as it necessitates the discontinuation of anti-leukemic treatment (Webb and Anderson, 1999). Low PLT counts are an independent prognostic factor and have been included in clinical classification systems (de Swart et al., 2015). The capacity to generate PLTs during primary disease and early after anti-leukemic therapy appears to be reflective of patients’ non-leukemic cell reserve and BM function and is highly associated with successful therapeutic intervention (van den Bosch et al., 2004). Thus, supporting efficient megakaryopoiesis and PLT generation are highly important clinical priorities in the management and treatment of MDS/AML. PLT transfusions have long been the primary therapeutic option for clinically significant thrombocytopenia (McCullough, 2000); albeit, they are only effective short-term as they increase the risk of alloimmunization (Reckhaus et al., 2018; Webb and Anderson, 1999).

Steady-state PLT production from megakaryocytes is governed primarily by thrombopoietin (TPO) and its receptor (TPO-R; encoded by the myeloproliferative leukemia virus oncogene [cMPL]; Yu and Cantor, 2012). TPO ensures adequate megakaryocyte numbers (de Sauvage et al., 1996; Kaushansky, 2005). Highly potent TPO-R agonists (TPO-RA), eltrombopag (EP; Erickson-Miller et al., 2009) and romiplostim, have been clinically explored to counteract disease and HMA treatment-associated thrombocytopenia (Kuter, 2009; Wang et al., 2004). Despite several encouraging phase I/II trials in MDS and AML testing EP as a single agent or along with conventional therapy (Olnes et al., 2012; Svensson et al., 2014), the TPO-R mimetic failed to enhance PLT production in patients during early stages of AZA therapy (Dickinson et al., 2018).

No study has clarified the molecular pathogenesis of HMA-induced thrombocytopenia yet, but this will be essential to develop effective therapeutic strategies in mitigating AZA-induced thrombocytopenia. Our findings delineate the molecular and functional consequences of AZA in megakaryocytic (Mk) progenitors and provide a preclinical rationale for evaluating p38 MAPK inhibitors for increasing PLT in patients in the early stages of AZA therapy.

To determine the effects of AZA on megakaryopoiesis, we characterized ex vivo Mk progenitor cell growth and differentiation of mononuclear cells (MNC) derived from healthy individuals or patients with MDS or AML (MDS/AML; Table S1) in the presence of a clinically relevant (Roulois et al., 2015) but non-cytotoxic dose of AZA (0.3 µM; Fig. 1 A and Fig. S1 A). Compared with vehicle controls, AZA treatment specifically inhibited immature and mature colony-forming Mk progenitor cell growth of healthy (Fig. 1 B) or MDS/AML MNC (Fig. 1 C) upon TPO-R stimulation with TPO-RA EP, or recombinant human TPO irrespective of TPO-RA dosage (Fig. S1, B–F). AZA also impaired the expansion of MNC in megakaryopoiesis-inducing cultures (Fig. 1, D and E; and Fig. S1, G–I) and decreased megakaryocyte-specific integrin CD41 (Haas et al., 2015) expression compared with vehicle-treated controls (Fig. 1 F), which was not attributable to the reduced presentation of Mk cell surface markers at the single cell level (Fig. S1, I–L). We found a 20% decrease in cells with clear morphological signs of Mk differentiation (increased cell size, multi-lobulated nucleus, and a lower cytosol/nucleus ratio) in AZA-treated cultures compared with controls (Fig. 1 G). Assessment of DNA content (increasing upon Mk cell maturation; Mattia et al., 2002) revealed a 30% increase in the number of diploid cells and a concomitant decrease in polyploid cells exposed to AZA compared with controls (Fig. 1 H and Fig. S1 M). We observed a decrease in the number of PLTs produced upon AZA treatment (Fig. 1 I and Fig. S1, N and O). Decitabine (DEC) showed similar megakaryopoiesis inhibitory effects (Fig. S1, P and Q). These data support that AZA inhibits Mk progenitor expansion, differentiation, and PLT production upon TPO-R stimulation, which may be common to chemical cytidine analogs.

Characterization of gene expression changes in FACS-sorted primary MPL presenting cells upon acute AZA exposure (16 h; Fig. 2 A and Table S1) under megakaryopoiesis-promoting conditions revealed a substantial number of genes to be differentially expressed (Fig. 2, B–D; and Table S2) compared with mock-treated controls. We found a prevalence of myeloid gene expression signatures at the expense of Mk-erythroid progenitor cell programs (Fig. 2 E and Table S3) and activation of IFN-I responsive genes (Fig. 2, F and G; and Table S4).

Additionally, analysis of differentially expressed genes in paired BM-MNC (BM-MNC) specimens from patients before (pre-AZA) and following long-term treatment with AZA (post-AZA) in vivo (Fig. 2 H and Table S5) also showed a strong enrichment of transcripts associated with the negative regulation of Mk differentiation and activation of IFNα response genes (Fig. 2, I and J; and Tables. S6 and S7) in post-AZA specimen compared with their treatment naive counterparts, a finding which was recapitulated in an unrelated published gene expression set (Fig. 2 K).

These molecular alterations strongly suggested that AZA suppresses Mk differentiation programs in TPO-R presenting stem and progenitor cells and further indicated that AZA and DEC may activate innate immune signaling more rapidly than previously appreciated (Chiappinelli et al., 2015; Li et al., 2014; Roulois et al., 2015; Stone et al., 2017).

RNA hypomethylation following AZA treatment has been described previously (Cheng et al., 2018; Schaefer et al., 2009). As the loss of methylation can increase the formation of immunogenic double-stranded RNA (dsRNA) species (Lee et al., 2019; Schaefer et al., 2009), we tested AZA’s effects on total RNA cytosine methylation and dsRNA abundance. Following ex vivo stimulation of megakaryopoiesis using TPO-RA, AZA led to a rapid and transient decrease in global RNA 5-methylcytosine (5 mC) levels (Fig. 3, A–C and Fig. S2 A). Concomitantly, AZA exposure resulted in increased dsRNA levels within only 1 h in healthy and MDS/AML MNC (Fig. 3, D–F), as well as purified immature CD34⁺ hematopoietic stem and progenitor cells (HSPCs; Fig. S2, B–E). DEC treatment decreased total RNA 5-methyl cytosine marks (Fig. S2 F) and rapidly induced dsRNA accumulation in a dose-dependent manner (Fig. S2, G and H). This demonstrates that AZA triggers RNA hypomethylation independent of its ability to integrate into RNA.

RNA sensors, such as endosomal membrane-associated toll-like receptor 3 (TLR3) and the cytosolic MDA5/MAVS complex, activate innate immune pathways upon binding of immunogenic RNA species (Chiappinelli et al., 2015; Gantier and Williams, 2007). Consistently, we found increased levels of TLR3 in total MNC as well as purified HSPC within 1 h of AZA exposure (Fig. 3, G and H). MDA5 and MAVS mRNA were also found to increase upon treatment with AZA (Fig. S2, I and J). Upon a 24-h exposure to AZA and compared with mock controls, we observed increased MDA5 protein levels (Fig. 3 I) which returned to baseline by 4 d after treatment (Fig. S2 K). AZA rapidly elicited activation of the IFN-I response hallmarked by IFNα secretion (Fig. S2 L) and increased IFNβ levels within only 1 h of AZA (Fig. 3 J) or DEC (Fig. S2 M) treatment of MNC. Notably, purified CD34⁺ cMPL⁺ stem and progenitor cell populations harboring extensive Mk colony forming capacity showed the strongest increase of IFNβ-expressing cells upon AZA exposure compared with CD34⁻ cMPL⁺ MNC, which also harbors the lowest capacity for Mk cell generation (Fig. 3, K and L). To impair IFNβ production downstream of TLR3 activation, we chose BX795 (Clark et al., 2009), a chemical inhibitor of TANK-binding Kinase I and inhibitor of nuclear factor kappa B kinase subunit epsilon (TBK1/IKKε) and key mediator of TLR3 triggered IFN-I release (Fitzgerald et al., 2003), which resulted in a partial, statistically non-significant rescue (P = 0.055) of MDS/AML patient–derived Mk progenitor cells in AZA containing colony assays (Fig. 3 M).

Reactivation of endogenous retroviral (ERV) elements upon DNA demethylation can trigger viral mimicry within several days of HMA exposure (Chiappinelli et al., 2015; Li et al., 2014; Roulois et al., 2015; Stone et al., 2017). We probed whether the AZA-induced rapid IFN production coincided with ERV reactivation and DNA demethylation following short-term AZA treatment of MNCs and found no evidence for a robust reactivation of these immunogenic sequences upon acute exposure to AZA (Fig. S2, N–R). In line, assessment of DNA 5mC levels showed no changes at this early time point, albeit significant global DNA demethylation was detectable after 4 d of AZA treatment (Fig. S2 S), consistent with previous studies (Cheng et al., 2018). These data strongly suggest that AZA elicits a rapid activation of IFN production following the accumulation of hypomethylated immunogenic RNA species in MNC cells, including stem and Mk progenitor cells. This response appears to precede detectable global DNA hypomethylation and ERV reactivation, albeit the experimental strategy chosen does not exclude functionally relevant changes in DNA cytosine methylation at specific loci.

We next tested whether cells with Mk potential would be able to elicit an IFN-I response upon binding of IFN-I cytokines to the IFNAR (Fig. 4 A). IFNAR is a heteromeric cell surface receptor composed of two subunits, IFNα/β R1 and IFNα R2 (Abramovich et al., 1994; Colamonici et al., 1994). Assessment of cell surface expression of R1 and R2 on healthy MNC revealed similar R1-positive cell numbers but a larger fraction of R2 positive cells within TPO-R/cMPL expressing MNC compared with the pool of total MNC (Fig. S3, A and B). Quantification of IFNAR expressing cells within stem and progenitor cell populations uncovered more IFNAR⁺ cells within the CD34⁺cMPL⁺ cell population compared with CD34⁻cMPL⁺ cells (Fig. S3, C and D). We further found that a greater number of CD34⁺cMPL⁺ cells co-expressed both cytokine receptor subunits on their cell surface compared with CD34⁻ cMPL⁺ cells (Fig. 4, B and C). These results strongly suggested that a substantial fraction of Mk progenitors (Fig. 3 L) can activate IFNAR upon AZA exposure and suggests their increased sensitivity to activation of this innate immune signaling pathway.

IFNAR signaling activates receptor-associated protein tyrosine kinases Janus kinase 1 (JAK1) and tyrosine kinase 2 (TYK2), and these, in turn, phosphorylate cytoplasmic transcription factors signal transducers and activators of transcription (STAT) 1 and 2a (Bromberg et al., 1996; Li et al., 1997; Stark and Darnell, 2012). In line, we found a moderate but significant increase in STAT1 phosphorylation in CD34⁺ cMPL⁺ cells upon treatment with AZA (Fig. 4, D and E; and Fig. S3 E), as well as STAT1-mediated gene target expression activation in healthy (Fig. 4 F) and patient-derived MNC (Fig. 4, G and H). We also detected increased expression of STAT1 target suppressor of cytokine signaling 1 (SOCS1) at the mRNA (Fig. S3 F) and protein levels after AZA treatment (Fig. 4 I). Furthermore, compared with paired pre-treatment specimens, the expression of a cluster of various IFN-I–stimulated genes (ISG; Fig. S3 G) and SOCS1 (Fig. S3 H) was elevated in patients under longer-term AZA therapy. Together, these results demonstrate that AZA treatment can trigger a rapid and persistent activation of IFNAR downstream signaling in healthy and leukemic MNC, including at the stem and progenitor cell levels.

MPL-expressing cells expressing high levels of functional IFNAR (Fig. 4 C) may activate downstream signaling in response to IFN-I activation. As SOCS proteins are known to inhibit cytokine receptor signaling (Piganis et al., 2011) through interaction with associated intracellular Janus kinases (Endo et al., 1997), we predicted that AZA-mediated IFN-I/STAT1/SOCS1 induction inhibited TPO-R signaling. To test this, we assessed TPO-R signaling in healthy or MDS/AML-derived MNC in the presence and absence of AZA focusing on canonical signaling components of the TPO-R, which critically relies on the activation of JAK2 (illustrated in Fig. S4 A; Endo et al., 1997; Zhang et al., 2018). AZA exposure of MDS/AML patient–derived (Fig. 5, A–D) or healthy MNC (Fig. 5, E–L) prevented the increase in phosphorylation of STAT3/5, AKT, and ERK upon TPO-R stimulation (Fig. S4, B and C; and Fig. 5, A–L). The addition of exogenous IFNα phenocopied this effect (Fig. S4, D–G). Conversely, pharmacological inhibition of IFNAR using dercernotinib, a highly specific JAK3 inhibitor (JAK3i; Farmer et al., 2015) previously shown to curtail AZA-triggered IFNAR activation (Roulois et al., 2015), or an IFNα/β-blocking B18R peptide (Alcami et al., 2000; Symons et al., 1995) led to a significant increase and restoration of pSTAT3/5, pAKT, and pERK levels in AZA-exposed Mk cells upon TPO-R stimulation using TPO (Fig. 5, E–H); while, except for pERK, no statistically significant alterations were seen with EP (Fig. 5, I–L). We conclude that AZA-mediated IFNAR activation can attenuate the stimulation of TPO-R signaling.

To test whether the inhibitory effects of AZA on megakaryopoiesis were attributable to IFNAR activation, we quantified Mk progenitor cell frequencies in healthy or MDS/AML MNC specimens by colony assay in the presence of AZA alone or in combination with IFN-I inhibitors. We found that pharmacologic inhibition of IFN-I signaling normalized immature as well as mature Mk colony-initiating cell numbers in AZA-exposed healthy (Fig. 6 A and Fig. S5 A) and patient-derived cell cultures (Fig. 6 B). IFN-I signaling inhibition also led to a reversal of increased SOCS1 mRNA (Fig. 6 C) and protein levels (Fig. 6, D and E) after AZA treatment. Notably, SOCS1 expression returned to baseline levels by day four after AZA exposure (Fig. S5, B and C).

We next tested whether the return of SOCS1 to low baseline levels after treatment with AZA allowed for effective TPO-R mediated stimulation of megakaryopoiesis, for which we quantified stem and progenitor cell proliferation and differentiation in megakaryopoiesis stimulating cultures upon sequential treatment with AZA followed by TPO-RA (Fig. 6 F). This revealed the restoration of immature and mature Mk colony-forming cell numbers in sequentially treated cultures compared with cell cultures simultaneously exposed to AZA and TPO-RA (Fig. 6 G). The sequential treatment regimen also increased the number of CD41 expressing cells by 30% (Fig. 6 H and Fig. S5 D) compared with cultures simultaneously exposed to AZA and the TPO-RA. Moreover, RNAi-mediated downregulation of SOCS1 (Fig. S5 E) rescued AZA-mediated inhibition of STAT5 activation (Fig. S5 F).

Lastly, IFN-I signaling is critical in the context of anti-leukemic therapies (Benjamin et al., 2007). Therefore, curtailing its activation through the use of direct IFN signaling inhibitors may blunt crucial immune-mediated leukemic cell eradiation pathways. To circumvent this potential caveat, we focused on P38 mitogen-activated protein kinase (p38 MAPK, p38) which is involved in the execution of IFNα-mediated antiviral responses (Bachegowda et al., 2016; Mayer et al., 2001). We found higher levels of phosphorylated p38 in AZA-treated cells compared with controls (Fig. 6 I) and observed that pharmacological inhibition of p38 counteracted the inhibitory effects of AZA on megakaryocyte progenitor growth stimulation (Fig. 6 J). Taken together, these results support that mitigating the activation of IFN-I signaling via the use of p38 inhibitors which have been tested for therapy of MDS (Bachegowda et al., 2016; Garcia-Manero et al., 2015) may allow for effective enhancement of Mk progenitor cell growth and differentiation in patients with MDS/AML undergoing therapy with AZA.

Our observations of AZA-induced innate immune signaling in MNC and its inhibitory effects on megakaryopoiesis (Fig. S5 G) are in line with well-documented anti-Mk/anti-thrombocytic effects of IFNα (Haas et al., 2015; Wang et al., 2000), particularly in patients with essential thrombocytosis (Mazur et al., 1986). We found that increased IFN-I dependent gene expression in MNC from longitudinally monitored patients is consistent with past reports on AZA and DEC-mediated activation of IFN-I response following DNA hypomethylation, the reactivation of ERV sequences upon long-term exposure (Chiappinelli et al., 2015; Roulois et al., 2015).

Moreover, our study uncovered a novel mode of IFN-I signaling activation in several MNC populations which appears to precede datable global DNA hypomethylation and ERV reactivation; we also found it to be acute and transient. In line with previous studies reporting AZA-mediated inhibition of RNA methylation, particularly affecting tRNAs (Schaefer et al., 2009; Schaefer et al., 2010), we detected rapid RNA cytosine demethylation in concert with dsRNA accumulation and activation of IFN-I anti-viral signaling upon AZA in MNC and purified MPL⁺ stem and progenitor cells. Our data strongly suggest that DEC elicits a very similar response. Thus, acute RNA demethylation upon AZA treatment is unlikely to stem from direct RNA intercalation, but rather a consequence of an upstream event, e.g., alterations in methyl group metabolism. Whether HMA and other anti-leukemic regimens eliciting thrombocytopenia share a molecular mode of action remains to be determined.

The observed functional impairment of specifically Mk progenitor cell expansion and differentiation after AZA exposure occurs in both healthy control and MDS/AML patient–derived primary MNC supports that AZA affects patient-derived stem and progenitors capable of Mk lineage differentiation. To unequivocally test this prediction, differential effects of AZA on healthy residual, preleukemic, and leukemic cell clone growth and their respective Mk lineage potential will need to be assessed. Albeit, we found rapid IFN-I signaling induction in various MNC subpopulations, and its impact at the functional level appears to be cell context-specific (exerting inhibitory effects in Mk-potent, but not other myeloid progenitor cells). We attributed this effect to premature TPO-R signaling attenuation, which was observed following AZA exposure of MNC stimulated with endogenous TPO or EP. Pharmacologic mitigation of IFN-I signaling activation restored TPO-R signaling in cells stimulated with TPO, but to a lesser extent in cells treated with EP. This finding may stem from the differences in eliciting cMPL downstream signaling due to differences in binding the receptor (TPO binds the extracellular ligand binding domain triggering strong STAT5/3 and MAPK activation; while EP binds the juxtamembrane domain eliciting strong STAT5 and delayed STAT3 and slightly differential MAPK activation; Erickson-Miller et al., 2009; Will et al., 2009).

Contrary to its inhibitory effects in Mk lineage-committed progenitor cells, IFN-I can activate quiescent stem cells and increase Mk lineage output (Essers et al., 2009; Haas et al., 2015). It is tempting to speculate that AZA/IFN-I triggers the expansion of non-leukemic cells, and Mk differentiation competent HSPC may compensate for the inhibitory effects of AZA/IFN-I on lineage committed Mk progenitor cells, thus alleviating ineffective megakaryopoiesis seen during early phases of AZA therapy. Further studies into the dynamics of leukemic and residual healthy stem cell generation and differentiation under AZA therapy need to be carried out to clarify this relevant aspect.

Lastly, our data show that mitigation of AZA-mediated inhibitory effects on megakaryopoiesis is possible via inhibition of IFN-I activation. Activated by IFN-I signaling, p38 MAPK activation is among myelosuppressive signals in human hematopoiesis; inhibition of this pathway showed pre-clinical (Bachegowda et al., 2016) and clinical efficacy (Garcia-Manero et al., 2015) in the management of MDS/AML improving erythroid and myeloid blood cell generation. Our data demonstrate that low-dose p38 inhibitor treatment can restore megakaryopoiesis and strongly suggests that clinically available p38 inhibitors, such as ARRY-614 (Garcia-Manero et al., 2015), may offer a suitable strategy to alleviate AZA-mediated thrombocytopenia.

The purpose of the study was to define the molecular and functional consequences of AZA treatment on megakaryopoiesis. Mk differentiation potential of primary hematopoietic cells derived from human cord blood (CB), peripheral blood (PB), or BM was analyzed in the absence or presence of AZA. IFN-I–signaling induction was rescued using JAK3 inhibitor dercernotinib (VX509), IFNα-blocking peptide B18R, genetic inhibition of SOCS1, or inhibitors of TBK1 (BX795) or p38 (SB203580).

EP (provided by Novartis) was dissolved in ddH₂O at 10 mg/ml. Working stock dilutions of 0.5 or 0.3 mg/ml EP were prepared and filtered.

AZA (cat. #A72012; Sigma-Aldrich) was reconstituted in 1× diluted Dulbecco’s PBS (cat. #14200075; Thermo Fisher Scientific) at 40 mM and stored at −20°C in single-use aliquots. DEC (cat. #A3656; Sigma-Aldrich) was reconstituted in DMSO (cat. #D128-500; Fisher Chemical) at 50 mg/ml and stored at −20°C in single-use aliquots.

Recombinant human cytokines were all purchased from Gemini. IFNα (rhIFNα; cat. #300-206P), stem cell factor (rhSCF; cat. #300-07), TPO (rhTPO; cat. #300-188P), IL-3 (rhIL3; cat. #AF-200-03), and IL-6 (rhIL-6; cat. #AF-200-06) were reconstituted and stored according to the manufacturer’s recommendation. Recombinant vaccinia virus B18R protein (cat. #ab190383) was reconstituted at 0.2 mg/ml and stored in single-use aliquots at −80°C.

Decernotinib (cat. #S7541; Selleckchem) was reconstituted at a concentration of 50 mM in DMSO (cat. #BP231-100; Thermo Fisher Scientific). Stocks were stored as single-use aliquots at −20°C.

SB 203580 (cat. #559398; Millipore Sigma) was stored as single-use aliquots at −20°C. Prior to use, the stock was pre-diluted with medium to 500 µM and used immediately.

BX795 (cat. #S1274; Selleckchem) was resuspended in DMSO at a concentration of 10 mM and stored as single-use aliquots at −80°C. Before use, stocks were diluted with medium to a working stock concentration of 100 µM (100×) and used immediately.

Human BM or PB specimens were collected at Montefiore Medical Center and the Albert Einstein College of Medicine (Table S1). Work conducted by this observational study was approved by the Albert Einstein College of Medicine’s Institutional Review Board (#2016-6770). Healthy control BM and CB specimens were purchased from Lonza and the New York Blood Bank, respectively. MDS/AML patient samples were longitudinally collected before and after AZA treatment using a protocol approved by an ethics committee at Helsinki University Hospital (303/13/03/01/2011) and were processed at Institute for Molecular Medicine Finland (FIMM; Helsinki, Finland; Table S5). Cells were maintained in recovery cultures in StemSpan SFEM (cat. #09650; StemCell Technologies) supplemented with 50 ng/rhSCF and 100 μg/ml Primocin (cat. #ANTPM1; Invivogen) at 37°C in 5% CO₂ in a fully humidified atmosphere for up to 24 h before use. All treatments were performed in cultures containing cells at a density of 0.5 − 1 × 10⁶/ml. Please note that because of considerable baseline inter-sample variability, we resorted to the statistical analysis of mock-control normalized data for the specimen tested. To ensure data robustness, we have used biological repeats (as opposed to technical repeats), which result in data representations without variances in the controls.

MNCs were isolated using low-density Ficoll-Paque Plus (1.077 g/ml, cat. #17-0300; GE Healthcare Life Sciences) gradient centrifugation at 400 g for 30 min at room temperature (RT). Before gradient centrifugation, PB and CB specimens were depleted of red blood cells for 30 min at RT using 5% (w/v) dextran (MW 266000; Sigma-Aldrich) in 0.9% NaCl₂.

CD34⁺ HSPCs were enriched from MNC by immunomagnetic bead sorting using the CD34 MicroBead Kit, human (cat. #130-046-702; Miltenyi Biotec) according to the manufacturer’s protocol.

UKE1 cells were grown in RPMI 1640 with glutamine (Corning) supplemented with 5% FBS (cat. #900-208; Gemini BioProducts), 1% penicillin/streptomycin (cat. #30-002-Cl; Corning), 1% HEPES buffer (cat. #25-060-CI; Corning), 1 µM hydrocortisone (cat. #H4001-1G; Sigma-Aldrich) at 37°C in 5% CO₂ in a fully humidified atmosphere. UKE1 cells were assured to be free of Mycoplasma contaminations.

MNC were grown in StemSpan SFEM supplemented with TPO-RA (rhTPO or EP, at concentrations as indicated) and 100 μg/ml Primocin at 37°C in 5% CO₂ in a fully humidified atmosphere for 1–24 h, as indicated.

CD34⁺ cells were grown in StemSpan SFEM supplemented with 5 ng/ml rhIL3, 50 ng/ml rhSCF, 20 ng/ml rhTPO, 20 ng/ml rhIL-6, and 100 μg/ml Primocin at 37°C in 5% CO₂ in a fully humidified atmosphere for a total of 14 d. The culture medium was refreshed every 2–3 d; treatments were refreshed after 7 d of culture.

Assay setup, cell culture, and colony scoring were performed according to the manufacturer’s protocol (cat. #04961; StemCell Technologies). Fixed and stained colony assays were imaged using a Cytation 5 inverted microscope (BioTek) or an EVOS FL Auto (Thermo Fisher Scientific) in scanning mode. Colony morphology was assessed, and colony numbers were enumerated manually. Assays containing at least 10 clearly visible and identifiable Mk CFU were used for data analysis.

For the first 4 d, CD34⁺ cells were subjected to liquid cultures at a density of 8 × 10⁴ cells/ml at 37°C in 5% CO₂ in a fully humidified atmosphere in Mk differentiation cultures supplemented with either 5 µg/ml EP (group A; EP) or 5 µg/ml EP and 0.3 µM AZA (group B; EP + AZA) or 0.3 µM AZA (group C). After this initial liquid culture phase, half of the cells were continued in liquid cultures with refreshed medium and treatment supplements for an additional 10 days, and the other half was subjected to Mk progenitor enumeration in MegaCult assays. To group C culture (liquid and MegaCult), EP (5 µg/ml) was also added (AZA → D4 EP).

Cells were cytospun upon culture completion. Cytospins were prepared using a Cytofuge (Statspin). Cell morphology was examined after May-Grϋnwald Giemsa staining (Cat. #660; Millipore Sigma) and documented using a Cytation 5 (BioTek) inverted microscope.

Cells were stained with the following antibodies: anti–human cMPL PE (BD Pharmingen), anti–human CD34 PB (Biolegend), and anti–human CD41a fluorescein isothiocyanate (eBioscience). Details pertaining to the antibodies used are listed in Table S8.

Upon a 16-h Mk differentiation culture in the presence or absence of AZA (0.3 µM), MNC were stained with anti-human cMPL PE (BD Pharmingen). Details pertaining to the antibodies used are described in Table S8. cMPL⁺ cells were sorted using a MoFlo ASTRIOS (BD Biosciences) cell sorter. Sorted cells were spun down at 300 g at RT and lysed using RLT lysis buffer for RNA extraction and downstream analysis.

Polyploidization was analyzed by propidium iodide (PI) and CD41a co-staining and cells were analyzed using an LSRII analyzer (BD Biosciences) as before with minor modifications.(Mattia et al., 2002) Details pertaining to the antibodies used are described in Table S8.

Flow cytometry data were collected on a Special Order ARIA II or LSR II flow cytometer (both from BD Biosciences) using FACSDiva software (BD Biosciences) or high throughput iQue Screener PLUS (Intellicyt). FACS data were analyzed using FlowJo software v10 (Treestar). All details for antibodies used in this study are listed in Table S8. Detailed procedures can be found in the supplemental methods section.

Phosphorylation of the signal transducer and activator of transcription (STAT) 5 and 3 was determined after serum starvation in cytokine-free StemSpan SFEM for 1 h, as previously described (Karjalainen et al., 2017; Will et al., 2009). Cells were stained with Alexa 647-anti-phospho-Stat5 (pY694), PE- CF594-anti-phospho-Stat3 (pY705), BV421-anti-phospho-Akt (pS473), and PE-anti-phospho-Erk1/2 (pT202/pY204; all antibodies from BD Biosciences) using 1:10–1:20 dilutions.

MNC or CD34⁺ cells were fixed with 3.7% paraformaldehyde for 15 min at RT, washed twice with 1× PBS at RT, and permeabilized with 0.2% T-PBS for 30 min at RT. The cells were washed twice in 1× PBS and the protein-bound nucleic acids were released by incubation of the permeabilized cells with proteinase K for 1 h at 37°C. Additionally, antibody specificity was tested using RNase III treatment for 1 h at 37°C. Cells were washed and stained with dsRNA-specific antibodies on ice for 30 min. Cells were washed twice with 1× PBS, stained with a rabbit anti-mouse AF488 secondary antibody (cat. #31584; 1:1,500; Life Technologies) for 30 min at RT, washed, and analyzed immediately.

Cells were stained with anti-cMPL and anti-CD34 antibodies in culture for 15 min followed by immediate fixation with pre-warmed BD Cytofix buffer (BD Biosciences) for 12 min at 37°C. After washing, cells were permeabilized in cold BD Perm Buffer III for 30 min on ice, washed, and stained with the anti-SOCS1 antibody for 40 min at RT. After washing, cells were stained with a rabbit anti-mouse AF488 secondary antibody (cat. #31584; Life Technologies; 1:500) for 40–60 min at RT, washed, and analyzed immediately.

Cells were stained with anti-cMPL and anti-CD34 antibodies, washed, and fixed in 40% methanol at RT for 15 min. Cells were then washed and permeabilized using 0.1% T-PBS for 15 min at RT. After washing, cells were treated with 1% BSA and 22.52 mg/ml glycine in 0.1% PBS-Tween20 for 30 min at RT, washed, and stained with anti-MDA5 or anti-TLR3 antibodies in 1% BSA/1× PBS at RT for 40 min. After washing, cells are stained with a rabbit anti-mouse AF488 secondary antibody (cat. #31584; Life Technologies; 1:1,500) for 40–60 min at RT, washed, and analyzed immediately.

MNC were treated as indicated for 30 min at 37°C. Monensin (Golgi Stop; BD Biosciences) was added to the cultures, and the cells were once again incubated for 30 min. Cells were washed and stained with anti-CD34 PB and anti-cMPL PE for 15 min on ice light protected. Cells were then fixed with BD Cytofix/Cytoperm buffer (BD Biosciences) for 20 min on ice, washed twice, and permeabilized in BD Perm/Wash buffer (BD Biosciences) for 15 min at RT. Cells were stained with anti-IFNβ FITC antibody (1:100) for 30 min on ice and light protected, subsequently washed, and analyzed immediately.

MNC were fixed using BD cytofix/cytoperm (BD Biosciences) for 20 min on ice. Cells were washed twice with 1× PBS and permeabilized with BD perm/wash buffer (BD Biosciences) for 15 min at RT. Nucleic acid bound to proteins is released by proteinase K treatment for 1 h at 37°C. After washing twice, cells are incubated for up to 1 h at 37°C with Turbo DNase I (cat. # AM1907; Thermo Fisher Scientific) for the digestion of DNA (determination of 5mC in RNA) or with RNase A (cat. # 19101; Qiagen) for digestion of RNA (determination of 5mC in DNA). Cells were washed twice with BD perm/wash buffer and stained with an anti-5mC primary antibody (Abnova) for 30 min on ice. After washing, cells were stained with anti-rabbit Alexa Fluor 488 (Life technologies) secondary antibody (1:500) in 1× PBS at RT for 30 min, washed, and analyzed.

Cells were cytospun onto poly-L-lysine coated slides (cat. #67–762; Thermo Fisher Scientific) and fixed in 3.7% paraformaldehyde (cat. #BM1585; Thermo Fisher Scientific), permeabilized with 1× PBS containing 0.2% Triton X-100 (cat. #MTX15681; Thermo Fisher Scientific) for 5 min at RT, and washed with 1× PBS at RT three times for 5 min each time. Cells were then incubated with 20 μg/ml proteinase K (cat. #P9107; New England Biolabs), in 50 mM Tris-HCl, pH 8.0, and 5 mM CaCl₂ for 1 h at 37°C. For RNase III digestion, the slides were washed twice and incubated in RNase III (cat# AM2290; Invitrogen) at least 30 min after proteinase K treatments. Primary antibodies against dsRNA (9D5 or J2; Table S8) were incubated overnight at 4°C (diluted 1:100 in 1× PBS), washed three times at RT for 5 min each time in PBS before incubation with a rabbit anti-mouse AF488 secondary antibody (cat. #31584; Life Technologies) for 30 min, and counterstained with 1 μg/ml DAPI (cat. #10236276001; Millipore Sigma) in 1× PBS for 3 min at RT. Slides were washed once with 1× PBS containing 0.05% Triton X-100 for 5 min each time at RT, twice with 1× PBS for 5 min each time, and once with distilled H₂O for 1 min. Cells were mounted with Prolong Gold antifade gel mount (cat. #P36930; Thermo Fisher Scientific) and imaged on a Cytation 5 inverted fluorescent microscope (BioTek).

MNCs from healthy donors or patients were serum-starved in plain STEM SPAN for 1–2 h and then subjected to treatment for 1 h. IFNα secreted into the cell culture supernatant of primary cells was measured using the Verikine-HS Human IFNα ELISA kit (cat. #41115; PBL Assay Science) according to the manufacturer’s instructions.

Cells were lysed with 50 mM Tris⋅HCl (pH 7.6) containing 1% NP40, 30 mM NaCl, 1 mM EDTA, 150 mM NaCl, and 1× proteinase inhibitor cocktail (cat. # 4693159001; Millipore Sigma) or RIPA buffer supplemented with EDTA-free protease inhibitor cocktail (11873580001; Sigma-Aldrich) and 1 mM PMSF. Protein concentrations were determined using Protein Assay Kit (5000002; BioRad). Lysates were resolved on 10% NUPAGE Bis-Tris gels (Invitrogen), transferred to polyvinylidene fluoride membranes, and incubated with the relevant antibodies. Stained membrane signals were visualized by using Hyglo HRP detection kit (cat. # E2500; Denville Scientific) and exposed to HyBlot CL film (cat. # E3012; Denville Scientific) by LICOR Odyssey Fc (LI-COR Biosciences). Protein levels were quantified using ImageJ (https://imagej.nih.gov/ij/). Details on antibodies used are listed in Table S8.

Total RNA was extracted from cells using the RNeasy Mini kit (Qiagen) according to the manufacturer’s protocol. RNA concentrations were assessed using a Nanodrop 1000 spectrophotometer (Thermo Fisher Scientific). RNA integrity was assessed using a 2100 Bioanalyzer (Agilent) when needed. RNA was reverse transcribed using the SuperScript II (cat. # 18064014; Thermo Fisher Scientific) along with random primers (cat. # 48190011; Thermo Fisher Scientific) according to the manufacturer’s recommendations.

Quantitative RT-PCR assays were performed in technical triplicates using the SYBR Green chemistry (cat. #4309155; Thermo Fisher Scientific) on a ViiA 7 Real-Time PCR System (Thermo Fisher Scientific) under standard cycling conditions with melting curve analysis. Expression values were efficiency corrected using the Pfaffl method (Pfaffl, 2001). Relative expression values of target genes were normalized to GAPDH expression. Detailed information on the primers utilized is listed in Table S9.

RNA isolated from FACS-sorted MPL⁺ MNC after treatment for 16 h was amplified and cRNA converted to sense-strand DNA using the Clariom D Pico Assay (cat. #902924; Thermo Fisher Scientific). The product was hybridized using the GeneChip Hybridization Kit (cat. # 900454; Thermo Fisher Scientific) onto Clariom D Assays, human (cat. # 902922; Thermo Fisher Scientific). The array image was generated using a GeneArray Scanner 3000 7 G (Thermo Fisher Scientific). Following RMA normalization, differentially expressed genes (fold change 1.2 and P value <0.05) were evaluated using the transcriptome analysis console software (Thermo Fisher Scientific). Ingenuity pathway analysis (IPA; Ingenuity Systems), EnrichR (Kuleshov et al., 2016), and GSEA (Broad Institute) software packages were used to further evaluate the differentially expressed genes and dysregulated pathways.

TEs can be embedded within gene bodies, correlating with adjacent gene expression (Jordan et al., 2003; Wang et al., 2013). To quantify TE expression, we identified TE-associated genes (based on their locus information) in the reference genome and subsequently quantified differential gene expression between AZA and control-treated cells. We also quantified and compared the number of TE in 679 differentially expressed genes in MPL⁺ MNC exposed to AZA (vs. mock controls) and 100× randomly sampled transcripts.

RNA was depleted of ribosomal RNA (Ribo-Zero rRNA Removal Kit; Epicentre), purified (RNeasy Clean-up Kit; Qiagen), and reverse transcribed to double-stranded cDNA (SuperScript Double-Stranded cDNA Synthesis Kit; Thermo Fisher Scientific). RNA-seq library preparation was performed according to manufacturer’s instructions using Illumina compatible ScriptSeq/TruSeq, size-selected, and purified (QIAquick Gel Extraction kit; Qiagen). Transcriptomes were sequenced with the Illumina HiSeq2000 platform using the TruSeq SBS Kit v3-HS reagent kit for paired-end sequencing with 100 bp read length. Reads were corrected with Trimmomatic and after filtering, aligned to the human genome (GRCH38) using STAR and EnsEMBL v82 gene models. Picard was used for sorting aligned reads and marking PCR duplicates. Gene expression was assessed by RNA-seq of BM MNCs extracted from AML patients (N = 7) before and after exposure to AZA treatment. Description of patients and patient sample characteristics, including exposure time to treatment and sampling time after last exposure, is available in Table S5. Differential expression analysis was performed using edgeR (McCarthy et al., 2012) and limma Bioconductor packages (Ritchie et al., 2015). The gene list was subsequently ranked according to the log fold-change (logFC) parameter. The ranked gene list was run in the GSEAPreranked module of GenePattern against MsigDB collections of pathways for the identification of significantly enriched pathways.

The statistical significance of the difference between two groups of paired and unpaired samples was assessed by two-tailed Student’s t test, Kolmogorov–Smirnov test, or Welch’s t test as indicated in the figure legends. Prism 8.3.0 (GraphPad Software) or Microsoft Excel was used for statistical analyses and/or plotting of graphs. Differential expression analysis was performed with transcriptome analysis console software.

Fig. S1 shows AZA dose finding, effects of AZA on myeloid progenitors, and details on ex vivo megakaryopoiesis in the presence of AZA. Table S1 lists patient information on the primary cell specimen (Einstein cohort). Table S2 contains a list of differentially expressed genes in MPL expressing MNC after AZA exposure. Table S3 shows the list of significantly enriched gene signatures in MPL⁺ MNC after AZA treatment. Table S4 holds the list of upregulated genes after AZA treatment also enriched in GSE40666. Fig. S2 contains information on dsRNA, 5mC (RNA and DNA) quantification in stem and progenitor cells in megakaryopoiesis, inducing cultures in the presence of AZA or DEC, as well as details on innate immune activation. Fig. S3 depicts details on IFNAR presentation and downstream signaling upon ex vivo or in vivo exposure of HSCP to AZA. Table S5 contains the details for longitudinally sampled AZA treated MDS/AML patients (FIMM cohort). Table S6 contains a list of genes enriched and positively correlated post-AZA treatment in GO negative regulation of Mk differentiation. Table S7 lists genes enriched and positively correlated post-AZA treatment in Hallmark—IFNα response. Fig. S4 holds information on phosphoflow-based strategy to quantify TPO-R signaling in HSPC and MNC exposed to AZA or recombinant IFNα. Fig. S5 shows the model and experimental details on RNAi-mediated SOCS1 knock-down in primary HPSC and contains the proposed mechanistic model supported by the study. Table S8 lists details for antibodies, and Table S9 contains information on primers used in the study.

All data associated with this study are present in the paper or the supplemental material. Gene expression microarray data for this study have been deposited in the Gene Expression Omnibus database with accession number GSE144410.

We thank Dr. D. Sun from the Einstein Stem Cell Isolation and Xenotransplantation Facility (funded through New York Stem Cell Science grant C029154) and C. Prophete for assistance with flow cytometry and D. Reynolds and W. Tran from the Einstein Genomics Core Facility for help with the microarray experiments. We thank Dr. J. Chen for helpful suggestions. We also thank P. Schultes from the Department of Cell Biology at Albert Einstein College of Medicine for expert technical assistance. We would also like to thank Drs. J. Bussel and K. Gritsman, as well as the team members of the Will and Heckman groups for very helpful discussions and suggestions.

This work was supported by Novartis Pharmaceuticals, the National Institutes of Health grants K12CA132783 (to U.C. Okoye-Okafor), CA230756, and DK105134 (to B. Will), and Cancer Center Support Grant P30CA013330 (pilot project to B. Will).

Author contributions: B. Will designed the study; U.C. Okoye-Okafor and B. Will designed the experiments; U.C. Okoye-Okafor, K.K. Javarappa, D. Tsallos, J. Saad, D. Yang, C. Zhang, L. Benard, V.J. Thiruthuvanathan, S. Cole, S. Ruiz, M. Tatiparthy, G. Choudhary, and S. DeFronzo performed the experiments; U.C. Okoye-Okafor, C. Zhang, and B.A. Bartholdy performed computational analysis; U.C. Okoye-Okafor assembled the data; U.C. Okoye-Okafor, K.K. Javarappa, D. Tsallos, J. Saad, D. Yhang, C. Zhang, A. Verma, A. Shastri, C. Pallaud, P.M. Ramos, C.A. Heckman, and B. Will analyzed and interpreted the data; A. Verma and A. Shastri provided patient samples; C. Pallaud, and P.M. Ramos provided materials; U.C. Okoye-Okafor and B. Will wrote the manuscript; and all authors gave final approval of the manuscript and agreed to be accountable for all aspects of the work.