C/EBPβ is a critical mediator of IFN-α–induced exhaustion of chronic myeloid leukemia stem cells

The BCR-ABL fusion protein, resulting from a reciprocal translocation between chromosome 9 and 22, causes chronic myeloid leukemia (CML) via its tyrosine kinase activity.^1-3  CML arises from the hematopoietic stem cell (HSC) compartment. In its chronic phase (CP), CML is characterized by silent expansion of myeloid cells, eventually progressing to life-threatening blast crisis. The development of ABL tyrosine kinase inhibitors (TKIs) has drastically improved the prognosis of patients with CML.^4,5  However, it remains to be determined whether CML can be cured using TKIs alone. Several clinical studies revealed that approximately one-half of patients that maintain remission for a certain duration following TKI treatment eventually suffer relapse after cessation of the regimen,^6-8  indicative of the persistence of CML stem cells. Indeed, accumulating evidence has revealed that CML stem cells survive in the bone marrow (BM) microenvironment independently of BCR-ABL signaling and acquire mutations that promote disease progression.^9-13  Therefore, eradication of CML stem cells would greatly benefit patients with CML-CP.

CCAAT/enhancer binding protein β (C/EBPβ) is a leucine-zipper transcription factor that plays critical roles in granulopoiesis, especially under stress conditions such as infection or cytokine stimulation.^14-18  In response to such external stimuli, C/EBPβ promotes both proliferation and differentiation of hematopoietic stem/progenitor cells (HSPCs) to supply granulocytes on demand.¹⁹  Previously, we showed that BCR-ABL hijacks the stress-induced pathway of granulopoiesis by upregulating C/EBPβ in HSPCs via activation of STAT5.²⁰  C/EBPβ contributes to myeloid expansion by accelerating differentiation, thereby facilitating exhaustion of CML stem cells.²⁰  These findings suggest that CML stem cells are susceptible to differentiation induced by C/EBPβ, and that upregulation of C/EBPβ activity via BCR-ABL–independent signals represents a promising therapeutic strategy for eradicating CML stem cells.

The effects of interferons on CML stem cells have been investigated in multiple studies.^21-24  In particular, interferon-α (IFN-α), a type I interferon, induces hematological and cytogenetic responses in patients with CML-CP, and has long been used for the treatment of this disease.^25-27  The efficacy of IFN-α has recently been reevaluated in several clinical studies.^28-33  IFN-α has multiple biological functions and exerts both direct^34-36  and indirect^37-39  effects on CML cells, including immunomodulation, but its effects on CML stem cells have not yet been elucidated. Previous studies^40-42  demonstrated that IFN-α binds to its receptor on normal HSCs and accelerates their cycling, differentiation, and exhaustion. Given that CML stem cells share many features with normal HSCs, IFN-α may also act directly on CML stem cells. In addition, IFN-α is a proinflammatory cytokine that induces C/EBPβ expression/activity in mature myeloid cells.^43,44  Accordingly, we hypothesized that IFN-α induces myeloid differentiation and exhaustion of CML stem cells through upregulation of C/EBPβ. In this study, we investigated the C/EBPβ-mediated effect of IFN-α on CML stem cells.

Mononuclear cells were obtained from BM or peripheral blood from 5 patients with CML at the time of diagnosis and stored in liquid nitrogen (supplemental Table 1). This study protocol was approved by the institutional review board of Kyoto University (Kyoto, Japan), and patients provided their consent for sample use and data analysis before this study in accordance with the Declaration of Helsinki.

The guide RNA (gRNA) targeting STAT5 binding sites in the Cebpb enhancer was designed using the CRISPRdirect Web site (https://crispr.dbcls.jp), and the synthesized oligonucleotides were inserted into the gRNA cloning vector (supplemental Figures 3B and 4A). The HindIII-NotI fragment containing the U6 promoter and gRNA sequences was cloned into pQCGFPMCS. For genome editing-mediated suppression, EML cells were retrovirally transduced with pdCas9 (catalytically dead Cas9)–humanized and the pQCGFP vector expressing gRNA specific for the target region (supplemental Figure 3B). Puromycin-resistant GFP⁺ cells were subjected to further analyses. For genome editing-mediated deletion of STAT5-binding sites in the enhancer region, EML cells were transfected with pCAG–human codon optimized Cas9 and derivatives of the specific gRNA cloning vector using Nucleofector 2b (Lonza, Basel, Switzerland). Single-cell clones lacking STAT5-binding sites were identified and subjected to further analyses.

Flow cytometric analysis and cell sorting were performed using FACSCanto II and FACSAria III instruments (BD Biosciences, San Jose, CA), respectively. Fluorescent-conjugated antibodies used in this study are listed in supplemental Table 2. Dead cells, identified as propidium iodide⁺, were excluded from analysis. Data were analyzed using the FlowJo software (Tree Star, Ashland, OR).

Quantitative reverse-transcription polymerase chain reaction (PCR) was performed as previously described.²⁰  The primers and probes used in this study are listed in supplemental Table 3.

Cell suspensions in phosphate-buffered saline were mixed with an equal volume of 2× Laemmli sample buffer (Bio-Rad, Hercules, CA) and 1/100 volume of Phosphatase Inhibitor Cocktail Solution I (Wako, Osaka, Japan) and then incubated at 100°C for 15 minutes. Samples were separated by sodium dodecyl sulfate-polyacrylamide gel electrophoresis and transferred to polyvinylidene fluoride membrane. Primary and secondary antibodies were diluted with Can Get Signal Immunoreaction Enhancer Solution (Toyobo, Osaka, Japan). The primary antibodies used in this study are listed in supplemental Table 4. Immunoreactive proteins were detected using horseradish peroxidase–conjugated anti-rabbit immunoglobulin G (NA934V, GE Healthcare, Little Chalfont, United Kingdom) and visualized using ECL Prime Western Blotting Detection Reagent (GE Healthcare). Images were obtained using myECL Imager (Thermo Fisher Scientific, Yokohama, Japan). ImageJ software was used to quantify the intensity of each band.⁴⁵ 

ChIP was performed as previously described.⁴⁶  In brief, EML cells were fixed with 1% formaldehyde in culture medium; fixation was then stopped by addition of 1.5 M glycine. Fixed cells were lysed with buffer containing sodium dodecyl sulfate and sonicated in a Bioruptor (Cosmo Bio, Tokyo, Japan). The sheared chromatin samples were subjected to immunoprecipitation using protein G magnetic beads (Dynabeads Protein G, Thermo Fisher Scientific). Antibodies used for ChIP are listed in supplemental Table 5. Immunoprecipitated chromatin was purified using the QIAquick PCR Purification Kit (Qiagen, Valencia, CA).

Immunoprecipitated chromatin was analyzed using THUNDERBIRD SYBR qPCR Mix (Toyobo). Primers used in ChIP-PCR are listed in supplemental Table 6.

Next-generation sequencing (seq) libraries were prepared from purified immunoprecipitated chromatin using the TruSeq ChIP Sample Prep Kit (Illumina, San Diego, CA). Library DNA was subjected to cluster generation on a flow cell using the TruSeq SR Cluster Kit v3 (Illumina) and sequenced on a HiScanSQ System in the NGS Core Facility at Kyoto Prefectural University of Medicine using the TruSeq SBS Kit v3 (Illumina). Base calls were performed using CASAVA, version 1.8.2, and ChIP-seq reads were aligned to the mm10 genome assembly using Bowtie2.⁴⁷  Peak calling was performed using MACS2,⁴⁸  and significantly enriched regions were identified using whole-cell lysate as input (false discovery rate = 0.05).

Statistical significance was determined using the 2-tailed Student t test. Survival of mice was analyzed using the log-rank test. P < .05 was considered statistically significant.

Information regarding mice, cell lines, plasmids, retrovirus infection, colony-forming assay, and BM transplantation can be found in supplemental Materials and methods.

IFN-α activates various downstream signaling events,^49,50  but it remains unclear whether it can upregulate C/EBPβ in CML cells. Hence, we assessed the role of IFN-α signaling in regulation of C/EBPβ in the mouse HSPC cell line EML transduced with either empty or BCR-ABL–expression retroviral vector (Figure 1A). In EML cells transduced with the empty vector, IFN-α treatment resulted in rapid phosphorylation of STAT1, a major downstream component of the IFN signaling pathway, as well as STAT5 (Figure 1B, lanes 1-3; supplemental Figure 1A). IFN-α also induced upregulation of C/EBPβ messenger RNA (mRNA) and protein expression (Figure 1B-D, lanes 1-3). Consistent with our previous report,²⁰  BCR-ABL itself phosphorylated STAT5 and upregulated C/EBPβ at both the mRNA and protein levels (Figure 1B-D, lanes 1 and 7; supplemental Figure 1A). Notably, IFN-α treatment further increased STAT5 phosphorylation and expression of C/EBPβ in BCR-ABL–expressing cells (Figure 1B-D, lanes 7-9; supplemental Figure 1A). IFN-α treatment promoted STAT1 phosphorylation at comparable levels in BCR-ABL–expressing and empty vector control cells, indicating that the STAT1-mediated pathway downstream of IFN-α remained intact in cells expressing BCR-ABL (Figure 1B-D, lanes 1-3 and 7-9; supplemental Figure 1A). Treatment with 1 μM imatinib suppressed STAT5 phosphorylation and C/EBPβ expression in BCR-ABL–expressing cells (Figure 1B-D, lanes 7 and 10; supplemental Figure 1A). However, even in the presence of imatinib, IFN-α induced STAT5 phosphorylation and upregulation of C/EBPβ (Fig. 1B-D, lanes 4-6 and 10-12; supplemental Figure 1A). In sharp contrast, expression of C/EBPα, another member of the C/EBP family that is essential for steady-state granulopoiesis,^51,52  was not increased by IFN-α treatment (supplemental Figure 2A). In summary, IFN-α activated both STAT1 and STAT5 and upregulated C/EBPβ expression irrespective of the presence of BCR-ABL, suggesting that STAT1 and STAT5 are involved in IFN-α–induced transactivation of C/EBPβ in CML cells.

The results described here also revealed that IFN-α and BCR-ABL share STAT5 as a component of the signal transduction machinery that upregulates C/EBPβ. We then performed ChIP-seq analysis to identify STAT5 binding sites that regulate Cebpb expression. As shown in Figure 2A, we identified a region ∼75 kb downstream of Cebpb where enrichment of STAT5 was observed only in the presence of BCR-ABL. Importantly, this ∼200 bp region contains 2 tandemly aligned copies of the IFN-γ–activated site (GAS), a consensus binding site for STAT5 that is highly conserved among species, including humans (supplemental Figure 3A). ChIP-PCR confirmed that STAT5 was recruited to this region in a BCR-ABL–dependent manner, whereas STAT5 was not recruited to the Cebpb promoter region irrespective of the presence of BCR-ABL (Figure 2B). In addition, H3K27Ac, a histone mark of active enhancers, was highly enriched in this distal region of Cebpb gene in the presence of BCR-ABL (Figure 2C).

These findings prompted us to evaluate the functional importance of this novel candidate enhancer in the regulation of Cebpb. To this end, we took advantage of dCas9, a nuclease-deficient Cas9 mutant, which has been used in combination with specific gRNAs to suppress promoter or enhancer activity by interfering with the binding of native transcription factors.^53,54  When dCas9 and gRNA-targeting STAT5 binding sites (supplemental Figure 3B) were retrovirally transduced into EML cells, dCas9 was specifically recruited to the enhancer (supplemental Figure 3C) and BCR-ABL–dependent upregulation of Cebpb was significantly suppressed (Figure 2D). Accordingly, expression of dCas9 and gRNA suppressed BCR-ABL–induced myeloid differentiation of EML cells approximately to one-half (Figure 2E). Furthermore, by introducing wild-type (WT) Cas9 and 2 gRNAs flanking the tandem GAS, we established EML cells in which the tandem GAS was heterozygously or homozygously deleted (supplemental Figure 4A-E). In cells homozygously lacking the tandem GAS, BCR-ABL–induced upregulation of Cebpb expression was significantly impaired (supplemental Figure 4D). Collectively, these data clearly show that this distal region is an enhancer required for sufficient Cebpb upregulation and myeloid differentiation induced by BCR-ABL/STAT5 signaling.

As shown in Figure 1 and supplemental Figure 1, IFN-α–induced upregulation of Cebpb was accompanied by activation of both STAT1 and STAT5, which potentially share consensus binding sites. Hence, we investigated the involvement of the distal enhancer of Cebpb in IFN-α–induced upregulation of Cebpb in EML cells. Although BCR-ABL alone did not recruit STAT1, IFN-α treatment strongly recruited STAT1 to this enhancer irrespective of the presence of BCR-ABL (Figure 2F, top). Moreover, both BCR-ABL and IFN-α treatment separately promoted recruitment of STAT5 to the enhancer, and the combination of BCR-ABL and IFN-α exerted an additive effect (Figure 2F, bottom). Importantly, the binding of STAT1 and STAT5 to the novel enhancer was conserved in the human CML cell line K562 (supplemental Figure 4F). In addition, in EML cells expressing dCas9 and gRNA targeting the enhancer, IFN-α-induced upregulation of Cebpb was significantly reduced, whereas basal expression was not affected (Figure 2G). In EML cells homozygously lacking the tandem GAS within the enhancer, upregulation of Cebpb in response to IFN-α was also abrogated (supplemental Figure 4E). To determine the respective importance of STAT1 and STAT5, we used dominant negative (dn) mutants of STAT1 and STAT5 in EML cells stably expressing BCR-ABL. dnSTAT5, which lacks the transactivation domain while maintaining DNA binding,⁵⁵  completely inhibited IFN-α–mediated upregulation of Cebpb (Figure 2H). Because dnSTAT5 can inhibit binding of both STAT5 and STAT1 to the enhancer, these results suggest that both STAT1 and STAT5 are functionally involved in this process. Moreover, dnSTAT1, which inhibits dimerization of STAT1,⁵⁶  significantly repressed IFN-α–mediated upregulation of Cebpb mRNA, suggesting the significant contribution of STAT1. In addition, IFN-α–induced upregulation of Cebpb in normal HSC in vivo was blunted in the absence of Stat5 (supplemental Figure 5A), suggesting that STAT5 may play the predominant role in this regulatory process. Collectively, these results indicate that the novel enhancer contributes to both BCR-ABL– and IFN-α–induced upregulation of Cebpb by recruiting activated STAT1 and STAT5.

Previously, we showed that C/EBPβ induces differentiation of leukemic stem cells (LSCs) and accelerates their exhaustion, suggesting that further activation of C/EBPβ represents a promising therapeutic strategy for eradicating these cells.²⁰  Hence, we sought to evaluate the effect of IFN-α on the colony-replating ability of CML stem cells. For this purpose, we retrovirally transduced lineage⁻ (lin⁻) c-kit⁺ Sca-1⁺ BM cells from WT or C/EBPβ knockout (KO) mice with BCR-ABL, and then subjected the transduced cells to a serial replating assay to evaluate LSC activity in vitro in the presence or absence of IFN-α (Figure 3A). WT cells could be replated up to 3 times in the absence of IFN-α, but IFN-α significantly diminished their replating ability (Figure 3B). This effect of IFN-α was abolished in C/EBPβ KO cells (Figure 3B). Colony-forming cells derived from WT mice exhibited a macrophage-like morphology, and to a greater extent in the presence of IFN-α, whereas C/EBPβ-deficient colony-forming cells maintained their immature appearance even in the presence of IFN-α (Figure 3C; supplemental Figure 6A). In accordance with the morphological features, flow cytometric analysis revealed that IFN-α induced differentiation of WT-LSCs toward CD11b⁺ myeloid cells, whereas myeloid differentiation was abrogated in C/EBPβ-deficient LSCs (Figure 3D). Accordingly, transduction of C/EBPβ in addition to BCR-ABL resulted in acceleration of myeloid differentiation and exhaustion, recapitulating the effects of IFN-α (supplemental Figure 7A-C). These in vitro data show that IFN-α induces differentiation and exhaustion of LSCs, and that these effects are largely dependent on C/EBPβ. Given that IFN-α induces exhaustion of myeloproliferative neoplasms (MPNs)-initiating clones,^57,58  we evaluated the effects of IFN-α on mouse BM cells transduced with MPL W515K/L, a causative mutation for MPN. IFN-α significantly impaired their colony-replating ability and enhanced myeloid differentiation under cytokine-free conditions, but these effects of IFN-α were significantly attenuated in the absence of C/EBPβ (supplemental Figure 8A-C). These results indicate that IFN-α can exhaust not only CML LSCs, but also MPN stem cells, in a C/EBPβ-dependent manner.

Next, we assessed the effects of IFN-α on CML stem cells in vivo. For these experiments, WT or C/EBPβ-deficient BM cells transduced with BCR-ABL were transplanted into lethally irradiated recipient mice. On day 9, these mice were intraperitoneally injected with either vehicle or polyinosinic:polycytidylic (poly I:C) to induce IFN-α production in vivo (Figure 4A). We confirmed that the dose of poly I:C used in this experiment upregulated Stat1 and Cebpb expression, but downregulated Cebpa expression in c-kit⁺ BM cells 12 hours after injection (supplemental Figure 9A). On day 17 after transplantation, BM cells were harvested from the primary recipients and analyzed by flow cytometry (Figure 4A-B). Because Sca-1 expression on HSCs is strongly affected by IFN-α,^41,59  and by BCR-ABL expression, we defined CML stem cells as CD150⁺ c-kit⁺ lin⁻. Notably, poly I:C treatment dramatically decreased the abundance of phenotypically defined CML stem cells (GFP⁺ cells within CD150⁺ c-kit⁺ lin⁻ BM cells) among primary recipients of WT cells, whereas the IFN-α–mediated reduction in the abundance of CML stem cells induced by poly I:C was abrogated in recipients of C/EBPβ-deficient cells (Figure 4B-C).

To confirm the effects of IFN-α on functional aspects of CML stem cells, we transplanted GFP⁺ cells purified from BM of primary recipients into lethally irradiated secondary recipient mice. The secondary recipients of WT cells, which have been harvested from the primary recipients without poly I:C treatment, died of leukemia. Poly I:C treatment of primary recipients significantly prolonged the survival of secondary recipients of WT cells (Figure 4D). Consistent with our previous observation, survival was significantly shorter in secondary recipients of C/EBPβ-deficient cells than in recipients of WT cells, and poly I:C treatment failed to prolong the survival of recipients of C/EBPβ-deficient cells (Figure 4D). Collectively, these data indicate that IFN-α requires C/EBPβ to induce exhaustion of CML stem cells in vivo. Although our in vitro data showed that IFN-α also promotes differentiation and exhaustion of normal HSCs in a C/EBPβ-dependent manner (supplemental Figure 10A-B), the emergence of GFP⁻ cells within the CD150⁺ c-kit⁺ lin⁻ population in poly I:C-treated recipients of WT cells indicated that nonleukemic HSCs were less sensitive to IFN-α than CML stem cells in vivo (Figure 4B).

To investigate the clinical relevance of the findings obtained in the mouse model, we evaluated the effect of IFN-α on CML stem/progenitor cells derived from patients (supplemental Table 1). For this purpose, we purified lin⁻ CD34⁺ cells from BM or peripheral blood of 5 patients with untreated CML at diagnosis (4 CP and 1 accelerated phase). When these purified cells were treated with IFN-α in vitro, CEBPB was significantly upregulated in a time- and dose-dependent manner. By contrast, expression of CEBPA was not altered by IFN-α stimulation (Figure 5A).

Furthermore, when lin⁻ CD34⁺ CML cells were subjected to a colony-forming assay, IFN-α treatment significantly decreased the total colony number in a dose-dependent manner in cells from all 5 patients (Figure 5B). In addition, the frequency of myeloid colonies was elevated in response to IFN-α, whereas the frequencies of erythroid and mixed-lineage colonies were reduced (Figure 5C; supplemental Figure 11A; supplemental Table 7). Interestingly, IFN-α exerted a weaker effect on colony formation in the UPN3 sample, which also exhibited weaker upregulation of CEBPB, suggesting a correlation between the CEBPB expression level and the effect of IFN-α. Flow cytometric analysis of colony-forming cells showed that IFN-α significantly increased the number of CD66b-expressing myeloid cells (Figure 5D). Although second-round colony formation was successful in UPN2 only, the effect of IFN-α on exhaustion and differentiation of lin⁻ CD34⁺ CML cells was more obvious at the second round of plating (supplemental Figure 11B-C). Collectively, these data indicate that IFN-α promotes exhaustion of patient-derived CML stem/progenitor cells, accompanied by upregulation of CEBPB and skewing of differentiation toward myeloid lineages.

In this study, we showed that IFN-α upregulates C/EBPβ, a transcription factor required for emergency granulopoiesis. In addition, we identified a novel 3′ distal enhancer located 75 kb downstream of the Cebpb locus as a critical element for upregulation of Cebpb by the BCR-ABL–STAT5 axis. IFN-α recruits both STAT1 and STAT5 to this 3′ distal enhancer and induces differentiation and exhaustion of CML stem cells through further upregulation of Cebpb.

In CML cells, C/EBPβ is upregulated by BCR-ABL through activation of STAT5.²⁰  Here, we demonstrated that IFN-α treatment further increases the expression of C/EBPβ at both the mRNA and protein levels, and we confirmed these findings in lin⁻ CD34⁺ CML cells obtained from patients. IFN-α treatment resulted in upregulation of CEBPB, but not CEBPA. These findings are consistent with the changes observed during stress granulopoiesis.¹⁴  STAT1 is a major signaling molecule acting downstream of type I IFN receptors, and other STATs have been implicated as components of type I IFN signaling as well.^60,61  We found that IFN-α activated not only STAT1 but also STAT5, the latter of which is also a downstream target of BCR-ABL, and that IFN-α and BCR-ABL exerted additive effects on STAT5 activation. These findings indicate that IFN-α significantly enhances BCR-ABL–induced Cebpb expression by activating STAT1 and further augmenting the existing activation of STAT5. This may explain why CML stem cells are more sensitive to IFN-α than normal HSCs.

To obtain further insights into the regulation of Cebpb, we sought to identify the genomic elements responsible for upregulation of Cebpb by STAT5, which acts downstream of both BCR-ABL and IFN-α. The 5′ proximal promoter region is responsible for upregulation of Cebpb by STAT3 in response to granulocyte colony-stimulating factor signaling and other stimuli.¹⁶  In reporter assays, we could not identify STAT5-binding sites in the proximal promoter region (up to ∼10 kb relative to the transcription start site). However, ChIP-seq analysis revealed a genomic region ∼75 kb downstream of Cebpb that contains 2 tandemly aligned GAS elements. This region was bound by STAT5 in the presence of BCR-ABL or IFN-α, as well as by STAT1 in response to IFN-α. CRISPR-Cas9–mediated suppression or deletion of this element significantly impaired upregulation of Cebpb by BCR-ABL or IFN-α, suggesting that this region is an enhancer that promotes expression of Cebpb in response to these stimuli. Accordingly, the H3K27Ac histone modification, a mark of active enhancers, was enriched in this region in the presence of BCR-ABL. A recent study reported retroviral integration “hot spots” around this region in BM cells transduced with Evi-1.⁶²  In these cells, Cebpb is upregulated and collaborates with Evi-1 to induce myeloid leukemia, suggesting the existence of a Cebpb enhancer in the region. The region identified in this study is the first example of a distal enhancer element involved in regulation of Cebpb expression. We are currently investigating the importance of this enhancer element under physiological conditions, including stress hematopoiesis in vivo.

A recent series of “Stop TKI” clinical trials revealed that approximately one-half of patients maintained treatment-free remission after discontinuation of TKI therapy, whereas the remainder molecularly relapsed,^6-8  indicating that CML stem cells had persisted. These findings emphasize the necessity of targeting LSCs to achieve a complete cure of CML. IFN-α may induce cytogenetic responses in CML-CP patients; therefore, if the molecular mechanisms that mediate the effects of IFN-α differ from those involved in the response to TKI therapy, a combination of these therapeutic approaches should be more effective than either alone. Indeed, the efficacy of IFN-α has recently been reevaluated,^63,64  and several clinical trials have shown that combination treatment with IFN-α significantly improves the therapeutic effects of TKIs.^28-33  In some Stop TKI trials, a history of IFN-α therapy was associated with a higher incidence of treatment-free remission.^7,65  In addition, the efficacy of IFN-α in treatment of MPN has recently been revealed.^66-68  In this study, we showed that IFN-α induces differentiation and exhaustion of CML and MPN stem cells. These IFN-α–mediated changes were significantly abrogated in the absence of Cebpb, suggesting that C/EBPβ is a critical mediator of the effect of IFN-α. Although IFN-α induces differentiation and exhaustion of normal HSPCs, our in vivo results in addition to ChIP data suggest that the mechanisms that upregulate C/EBPβ independently of IFN-α, such as BCR-ABL, might sensitize the HSPCs to IFN-α. We recently identified c-myc as a downstream target of a C/EBPβ isoform under stress conditions in HSPCs (A.S., A.Y., and H.H., unpublished data, 7 February 2018), and previous work has shown that c-Myc could drive differentiation and exhaustion of LSC.^12,69  We are currently identifying the roles and targets of C/EBPβ isoforms in CML stem cells. Although our in vitro data suggest that IFN-α directly acts on CML stem cells to induce C/EBPβ, the possible involvement of the BM niche should be taken into account as well.

Collectively, our data demonstrate that IFN-α upregulates C/EBPβ in CML stem cells by recruiting STAT1 and STAT5 to the novel 3′ distal enhancer of Cebpb. Upregulation of C/EBPβ is responsible for IFN-α–induced differentiation and exhaustion of CML stem cells. The proposed mechanism of action provides insight into the efficacy of IFN-α against CML stem cells.

The authors thank Toshio Kitamura (The University of Tokyo) for providing Plat-E cells, pMX-IRES-EGFP, and pMX-STAT5Δ749-IRES-EGFP vectors; Schickwann Tsai (University of Utah) for EML cells; Anthony Green (Cambridge Institute for Medical Research) for pLKO.3G STAT1DN; Gang Huang (Cincinnati Children’s Hospital Medical Center) for pMSCV-IRES-Thy1.1; Naoko Watanabe-Okochi and Mineo Kurokawa (The University of Tokyo) for pGCDNsam-IRES-Kusabira Orange; Keiko Okuda (Kyoto Prefectural University of Medicine) for MIG-BCR-ABL; and Lothar Hennighausen (National Institutes of Health, National Institute of Diabetes and Digestive and Kidney Diseases) for Stat5^{floxed/floxed} mice. They also thank Shizue Taniichi at Kyoto University, Masahiro Kawahara at Shiga Medical University, Akihiko Yokoyama at Kyoto University and the National Cancer Center, and all members of the Maekawa Laboratory for their advice on our experiments, as well as Yoko Nakagawa for her excellent technical support in our laboratory.

This work was partly supported by KAKENHI Grants-in-Aid for Scientific Research from the Japan Society for the Promotion of Science (25461415 and 18K08354 [H.H.], 25430149 [A.Y.], and 25112706 and 16K07171 [T.M.]); an Extramural Collaborative Research Grant of the Cancer Research Institute, Kanazawa University (H.H.); the Highway Program for Realization of Regenerative Medicine (JP17bm0504008) (H.H.); the Acceleration of Transformative Research for Medical Innovation from the Japan Agency for Medical Research and Development (H.H.); and a grant from the Kyoto University Research Development Program “Ishizue” (H.H.). D.G.T. was supported by a STaR Investigator Award, an RCE Core grant, Tier 3 RNA Biology Center grant MOE2014-T3-1-006 from the NRF and MOE, Singapore, and the National Institutes of Health, National Cancer Institute (CA66996).

Contribution: A.Y. designed and performed the experiments, analyzed the data, and wrote the manuscript; H.H. supervised the project, designed and performed the experiments, analyzed the data, and wrote the manuscript; A.S., N.K., Y.H., and Y.M. performed the experiments; R.S., H.A., F.S., and M.N. performed the experiments and analyzed the chromatin immunoprecipitation-sequencing data; and S.K., D.G.T., K.T., and T.M. supervised the project and wrote the manuscript.

Conflict-of-interest disclosure: H.H. received research funding from Kyowa Hakko Kirin and Novartis Pharma. S.K. received honoraria from Bristol-Myers Squibb K.K., Pfizer, and Otsuka Pharmaceutical, Novartis, and Sumitomo Dainippon Pharma. T.M. received research funding from Bristol-Meyers K.K. The remaining authors declare no competing financial interests.

Correspondence: Hideyo Hirai, Department of Transfusion Medicine and Cell Therapy, Kyoto University Hospital, 54 Shogoin-Kawahara-cho, Sakyo-ku, Kyoto 606-8507, Japan; e-mail: hhirai@kuhp.kyoto-u.ac.jp.