Ineffective hematopoiesis is a fundamental process leading to the pathogenesis of myelodysplastic syndromes (MDS). However, the pathobiological mediators of ineffective hematopoiesis in MDS remain unclear. Here, we demonstrated that overwhelming mitochondrial fragmentation in mutant hematopoietic stem cells and progenitors (HSC/P) triggers ineffective hematopoiesis in MDS. Mouse modeling of CBL exon deletion with RUNX1 mutants, previously unreported comutations in patients with MDS, recapitulated not only clinically relevant MDS phenotypes but also a distinct MDS-related gene signature. Mechanistically, dynamin-related protein 1 (DRP1)–dependent excessive mitochondrial fragmentation in HSC/Ps led to excessive reactive oxygen species production, induced inflammatory signaling activation, and promoted subsequent dysplasia formation and impairment of granulopoiesis. Mitochondrial fragmentation was generally observed in patients with MDS. Pharmacologic inhibition of DRP1 attenuated mitochondrial fragmentation and rescued ineffective hematopoiesis phenotypes in mice with MDS. These findings provide mechanistic insights into ineffective hematopoiesis and indicate that dysregulated mitochondrial dynamics could be a therapeutic target for bone marrow failure in MDS.
We demonstrated that excessive mitochondrial fragmentation is a fundamental pathobiological phenomenon that could trigger dysplasia formation and ineffective hematopoiesis in MDS. Our findings provide mechanistic insights into ineffective hematopoiesis and suggest dysregulated mitochondrial dynamics as a therapeutic target for treating MDS.
This article is highlighted in the In This Issue feature, p. 1
Myelodysplastic syndromes (MDS) comprise a group of clonal hematopoietic stem cell (HSC) disorders characterized by ineffective hematopoiesis, in which clonal expansion coexists with accelerated cell death (1). Chronic noninfectious inflammation caused by dysregulated inflammatory and innate immune signaling is a characteristic hallmark of MDS pathogenesis (2, 3). MDS HSCs and progenitors (HSC/P) exhibit clonal expansion advantages as well as defects in multilineage differentiation (4–7). However, despite differentiation defects, mutant MDS HSC/Ps retain differentiation potential and constantly produce mature mutant cells, which are a major component of MDS clones in the bone marrow (BM; refs. 1, 8). This is the fundamental difference between MDS and acute myeloid leukemia (AML), in which almost all cells in the mutant clones remain as immature myeloblasts due to arrested differentiation. However, only a limited number of in vivo murine MDS models, which represent the BM environment of ineffective hematopoiesis, are currently available for evaluating such pathogeneses (5, 9, 10). Therefore, the mechanisms underlying ineffective hematopoiesis in MDS clones remain largely unclear.
Recent advances in genomics have revealed the mutational profiles of many hematologic malignancies (11). Among a variety of genetic events, mutations in RUNX1 constitute one of the major subclonal mutations in myeloid malignancies (12, 13). RUNX1 mutations, in particular, are frequently detected in patients with higher-risk MDS (14, 15). In these cases, RUNX1 mutations are significantly associated with mutations in several other genes, such as ASXL1, EZH2, and STAG2 (11). However, it is clinically and experimentally evident that the RUNX1 mutant alone is not sufficient to cause disease (5, 16). These findings clearly indicate the critical role of synergistic interplay between collaborating genetic events and RUNX1 mutations in disease development. Although the current sequencing approach is a powerful tool for detecting single-nucleotide variants and minimal deletions/insertions, its capacity for detecting some genetic events, such as tandem duplications and exon deletion, is limited. Point mutations in CBL are frequently detected in patients with myeloid malignancies, including MDS (17–19). Most point mutations in CBL are clustered in exons 8 and 9 (20). Although it is less frequent, a statistically significant co-occurrence of RUNX1 and CBL point mutations has been reported in patients with MDS (11). Given that deletion of exon 8/9 in CBL was reported in a few patients with AML (21), the frequency of CBL gene alteration in MDS may be underestimated.
Mitochondria are dynamic organelles that play critical roles in ATP generation through oxidative phosphorylation (OXPHOS). Mitochondria constantly undergo dynamic fission and fusion to maintain their function and integrity (22). It is well documented that the impairment of mitochondrial plasticity due to aberrant fission or fusion contributes to the pathogenesis of various stress conditions and diseases, including cancers (23). In this study, we identified a previously unreported CBL exon 8/9 deletion (CBLΔE8/9) comutation in patients with MDS with RUNX1 mutations. We found that synergistic interplay between CBLΔE8/9 and RUNX1 mutations could induce phenotypically and genetically relevant characteristic MDS-ineffective hematopoiesis in BM within a short period of time and that dysregulated mitochondrial dynamics serve as a fundamental mediator of MDS-ineffective hematopoiesis.
CBLΔE8/9 Collaborates with RUNX1 Mutation to Develop Bona Fide MDS
Although most gene mutations in patients with MDS have been identified via genome sequencing (11), the frequency of several types of mutations has been underestimated. Therefore, in addition to our routine targeted sequencing that covers 68 recurrently mutated genes in MDS, we evaluated the incidence of CBLΔE8/9 in patients with MDS carrying RUNX1 mutations (Fig. 1A; Supplementary Fig. S1A and S1B; Supplementary Tables S1 and S2). Substantiating the findings of previous studies, we found several known comutations, such as mutations in ASXL1, EZH2, SRSF2, and STAG2, in patients carrying RUNX1 mutations (Fig. 1A; ref. 11). CBL point mutations were also detected at the expected frequency (Fig. 1A). Remarkably, we identified CBLΔE8/9, a previously unreported comutation, in 3 of 15 patients carrying the RUNX1 mutation. Among these three patients, two exhibited other mutations in PTPN11 or GATA2 that were recently reported to be significantly associated with higher risk MDS or secondary AML (sAML; ref. 24). Indeed, the disease subtype of these two patients was MDS with excess blasts (higher-risk MDS). On the other hand, the disease subtype of the patient with only RUNX1 and CBLΔE8/9 mutations was MDS with multilineage dysplasia without excess blasts (lower-risk MDS; Fig. 1A). These findings led us to hypothesize that cooperation between CBLΔE8/9 and RUNX1 mutations may lead to MDS pathogenesis.
To test this hypothesis, we retrovirally introduced both the CBLΔE8/9 mutant and the RUNX1-S291fs (RUNX1S291fs) mutant (Supplementary Fig. S2), which is an MDS patient–derived RUNX1 mutant (15), into wild-type (WT) murine HSC/Ps and performed an in vivo BM transplantation (BMT) assay. We also generated mice transplanted with single RUNX1S291fs mutant–transduced cells or single CBLΔE8/9 mutant–transduced cells. Empty vector–transduced cells served as control. Consistent with our previous reports (25, 26), RUNX1S291fs mice developed an MDS-like phenotype with long latency (median survival, 249 days; Fig. 1B–D). CBLΔE8/9 mice developed leukocytosis and myeloproliferative neoplasms but not MDS (median survival, 87 days; Fig. 1B–D). Notably, CBLΔE8/9/RUNX1S291fs mice became moribund within 2 months following BMT. The median survival of CBLΔE8/9/RUNX1S291fs mice was significantly shorter than that of either individual mutant alone (Fig. 1B). The moribund CBLΔE8/9/RUNX1S291fs mice showed leukocytopenia, macrocytic anemia, and thrombocytopenia (Fig. 1C). Despite the myeloid differentiation bias induced by CBLΔE8/9 and RUNX1 mutants, the absolute number of neutrophils in peripheral blood (PB) was comparable in the control and CBLΔE8/9/RUNX1S291fs mice (Supplementary Fig. S3). All CBLΔE8/9/RUNX1S291fs mice showed multilineage dysplasia in the BM (Fig. 1D and E; Supplementary Fig. S4). More than 10% of dysplastic cells were detected in each lineage (Fig. 1F; Supplementary Fig. S4). On the other hand, the frequency of BM blasts was less than 10% (Fig. 1G; Supplementary Fig. S4). In addition, CBLΔE8/9/RUNX1S291fs mice developed mild BM reticulin fibrosis (Supplementary Fig. S5A), which is observed in approximately 20% of the patients with MDS. Ring sideroblast formation, which occurs due to abnormal iron accumulation in mitochondria, was not observed in this model (Supplementary Fig. S5B). As CBL encodes an E3 ubiquitin ligase that negatively regulates receptor tyrosine kinase signaling (20), CBL mutations are associated with aberrant activation of RAS/MAPK signaling pathways and RASopathies (27). In addition to CBL mutations, RUNX1 is frequently comutated with other RASopathy-associated genes, such as PTPN11, KRAS, and NRAS (11, 28). However, coexistence of these RASopathy-associated gene mutations (but not CBL mutations) and RUNX1 mutations is significantly enriched in cases of sAML (24, 28). Consistent with this clinical evidence, PTPN11 mutant in conjunction with RUNX1 mutant did not develop MDS pathogenesis in mice (Supplementary Fig. S6A–S6C). Overall, these results suggest that the synergistic effect of CBLΔE8/9 and RUNX1 mutants induces major MDS features in mice in a short period of time.
To further characterize the myelodysplasia phenotypes generated by interplay between CBLΔE8/9 and RUNX1 mutants, we focused on CBLΔE8/9/RUNX1S291fs mice and subjected their BM cells to flow cytometric analysis. When donor BM cells were initially transplanted, they contained not only double-transduced CBLΔE8/9/RUNX1S291fs cells but also single-transduced clones and nontransduced WT cells (Fig. 1H). Among these cells, WT cells were dominant while double-transduced CBLΔE8/9/RUNX1S291fs cells were minor constituents (Fig. 1H). When CBLΔE8/9/RUNX1S291fs mice became moribund, WT and single-mutant clones were outcompeted by the CBLΔE8/9/RUNX1S291fs clones (Fig. 1H). This suggested that the synergistic effect of CBLΔE8/9 and RUNX1 mutants conferred a clonal advantage to HSC/Ps. Although the frequency of immature c-Kit+ cells increased, it was less than 20% in the BM of diseased mice (Supplementary Fig. S7). Lineage marker− Sca1+ c-Kit+ (LSK) cells as well as common myeloid progenitors and granulocyte-macrophage progenitors (GMP) within lineage marker− Sca1− c-Kit+ cells were significantly increased in CBLΔE8/9/RUNX1S291fs mice (Fig. 1I and J). Among LSK cells, increase in CD150+ CD48+ hematopoietic progenitor cell 2 was prominent (Fig. 1I and J), suggesting the expansion of progenitors with myeloid skewing. This is consistent with observations in patients with MDS (29). As most cells in the CBLΔE8/9/RUNX1S291fs clone were not immature cells, we also analyzed the mature myeloid population in the BM. Based on the expression levels of c-Kit and Ly6G, a specific marker of murine neutrophils, granulopoiesis in the BM can be categorized into five individual stages showing distinct morphologic features and granule protein expression (30). As observed in the control mice (Fig. 1K and L), most granulocytes belonged to subpopulation V, which comprises band cells and segmented granulocytes (30). In contrast, late-stage BM granulopoiesis in CBLΔE8/9/RUNX1S291fs mice was impaired (Fig. 1K and L). Although subpopulation V was significantly decreased, the frequency of subpopulation IV, which comprises myelocytes and metamyelocytes, was significantly increased in CBLΔE8/9/RUNX1S291fs mice compared with that in control mice (Fig. 1K and L). Concordant with the impaired BM granulopoiesis, the total BM cells significantly decreased in CBLΔE8/9/RUNX1S291fs mice (Fig. 1M). These results suggest the occurrence of ineffective hematopoiesis, the fundamental hallmark of MDS pathogenesis, in the BM of CBLΔE8/9/RUNX1S291fs mice.
CBLΔE8/9/RUNX1S291fs HSC/Ps Exhibit Distinct Gene Signature Related to Lower Risk MDS and Inflammatory Signaling Activation
To study the gene expression profile of HSC/Ps derived from the CBLΔE8/9/RUNX1S291fs mice, we performed RNA sequencing (RNA-seq) analysis of mutant or empty vector(s)–transduced c-Kit+ BM cells obtained from individual groups of mice. Principal component analysis revealed that the gene expression pattern of HSC/Ps derived from CBLΔE8/9/RUNX1S291fs mice was completely distinct from those of control mice as well as mice with one individual mutant alone (Fig. 2A). Although exogenous human CBL- and RUNX1-mutant genes were prominently overexpressed compared with murine homologous genes in the CBLΔE8/9/RUNX1S291fs double-transduced HSC/Ps, the expression level of each mutant protein was approximately threefold that of the endogenous WT protein (Supplementary Fig. S8A–S8D). To characterize the gene signature induced by the combination of CBLΔE8/9 and RUNX1 mutants, we generated a gene set using genes exhibiting threefold significant (P < 0.05) upregulation in CBLΔE8/9/RUNX1S291fs cells compared with those of control cells (Supplementary Table S3). We then tested this gene set in a data set of human MDS CD34+ BM cells (GSE19429; ref. 31) via gene set enrichment analysis (GSEA). This large MDS cohort included patients without excess blasts (n = 55; a relatively lower risk group), patients with excess blasts (n = 80; a relatively higher risk group), and healthy donors (n = 17). GSEA revealed that the CBLΔE8/9/RUNX1S291fs–induced genes were significantly upregulated in the MDS cohort, particularly the lower-risk MDS cohort (Fig. 2B–D). Recent analyses of large MDS and sAML cohorts by Makishima and colleagues (24) linked several distinct genetic alterations with disease progression steps. They have revealed that mutations in FLT3, IDH1, IDH2, NPM1, NRAS, PTPN11, and WT1 (defined as “type 1 mutation”) are associated with sAML transformation. Mutations in several other genes, including ASXL1, GATA2, KRAS, RUNX1, STAG2, TET2, TP53, and ZRSR2 (defined as “type 2 mutation”), were enriched in higher-risk MDS. Moreover, mutations in SF3B1 were significantly associated with lower-risk MDS. Patients with type 1 or type 2 mutations exhibited inferior survival compared with patients without both type 1 and type 2 mutations. To determine whether the CBLΔE8/9/RUNX1S291fs–induced gene signature in MDS is associated with specific genetic subtypes, we tested the leading-edge CBLΔE8/9/RUNX1S291fs–induced genes (Fig. 2B) in the genome sequencing/gene expression data set of human MDS CD34+ BM cells (GSE58831; Fig. 2E). Although the MDS-related CBLΔE8/9/RUNX1S291fs genes were enriched in the MDS cohort regardless of genetic alterations in comparison with the healthy donors (Fig. 2F), these genes were significantly enriched in genetic subgroups without both type 1 and type 2 mutations (lower-risk MDS) in patients with MDS (Fig. 2F). Network enrichment analysis of Gene Ontology (GO) terms and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway analysis indicated that CBLΔE8/9/RUNX1S291fs–induced genes were significantly enriched in pathways associated with the inflammatory immune response (Fig. 2G and H; Supplementary Tables S4 and S5). Although it was less significant, some of these pathways were upregulated in mice with a single mutant alone (Supplementary Fig. S9). We note that dysregulated innate immune response and inflammation are central pathobiological mediators of MDS, particularly in the lower-risk group (2, 3). Consistent with transcriptional analysis of mutant HSC/Ps, the levels of several inflammatory cytokines and chemokines, such as CXCL10, G-CSF, Il1RAP, and MCP1, were increased in the plasma from CBLΔE8/9/RUNX1S291fs mice (Fig. 2I). These cytokines and chemokines have been reported to be significantly elevated in the plasma from patients with MDS (32, 33). Overall, these results suggest that CBLΔE8/9/RUNX1S291fs mice exhibited a gene expression profile specific to MDS pathogenesis.
Identification of Mitochondrial Fragmentation in Murine and Human MDS Clones
Given that neither mice with the CBLΔE8/9 mutant alone nor the RUNX1S291fs mutant alone developed bona fide lower-risk MDS, a specific gene signature in the CBLΔE8/9/RUNX1S291fs mice could be responsible for the MDS-ineffective hematopoiesis. To identify distinct marker genes, we used the MarkerFinder algorithm in AltAnalyze (v2.1.0; ref. 34), followed by gene set cluster enrichment (Fig. 3A). Among the pathways enriched in the CBLΔE8/9/RUNX1S291fs cluster, we identified the OXPHOS pathway as the top upregulated pathway (Fig. 3A). Genes in the OXPHOS pathway were significantly upregulated in HSC/Ps of CBLΔE8/9/RUNX1S291fs mice compared with those in control mice (Fig. 3B). Notably, these genes in the OXPHOS pathway were significantly upregulated in the lower-risk MDS cohort compared with the higher-risk MDS cohort (Fig. 3C). Considering that transcriptome analysis indicated distinct genes associated with the OXPHOS pathway in CBLΔE8/9/RUNX1S291fs HSC/Ps, we performed a Seahorse functional assay using an equal number of c-Kit+ BM cells from CBLΔE8/9/RUNX1S291fs mice and control mice (Supplementary Fig. S10A). Mitochondrial basal respiration and ATP production were significantly increased in CBLΔE8/9/RUNX1S291fs HSC/Ps (Supplementary Fig. S10B). However, OXPHOS activity did not reflect a significant alteration of OXPHOS genes in HSC/Ps from the CBLΔE8/9/RUNX1S291fs mice. To investigate the discrepancy between mitochondrial gene expression and mitochondrial output measured via the Seahorse assay, we evaluated mitochondrial content in HSC/Ps using immunofluorescence staining of translocase of outer membrane 20 (TOM20), an outer mitochondrial membrane protein. We found enhanced mitochondrial fragmentation in CBLΔE8/9/RUNX1S291fs cells compared with those in control cells and cells with one individual mutant alone (Fig. 3D and E; Supplementary Fig. S11A). We confirmed excessive mitochondrial fragmentation in CBLΔE8/9/RUNX1S291fs cells by transmission electron microscopy (Fig. 3F and G). Mitochondrial fragmentation facilitates the production of reactive oxygen species (ROS; ref. 35). Therefore, we measured the level of cellular ROS in HSC/Ps obtained from CBLΔE8/9/RUNX1S291fs mice as well as mice with one individual mutant alone. ROS levels in the CBLΔE8/9/RUNX1S291fs cells were significantly higher than those in control cells and cells with one individual mutant alone (Fig. 3H–L; Supplementary Fig. S11B).
Given that CBLΔE8/9/RUNX1S291fs mice consistently exhibited relevant MDS phenotypes and MDS-related gene signature, we hypothesized that mitochondrial fragmentation may be a fundamental cell biological phenomenon in MDS pathogenesis. To assess this hypothesis, we purified BM mononuclear cells (BMMNC) from patients with MDS (n = 24) and analyzed the mitochondrial structure by confocal fluorescence microscopy (Supplementary Table S6). BMMNCs were also isolated from healthy donors (n = 8) and patients with clonal hematopoiesis of indeterminate potential (CHIP; n = 3), idiopathic cytopenia of undetermined significance (ICUS; n = 1), clonal cytopenia of unknown significance (CCUS; n = 4), AML (n = 10), and other diseases or conditions (aplastic anemia, multiple myeloma, malignant lymphoma, and macrocytic anemia; n = 5; Supplementary Table S6). Consistent with our observations in mice with MDS, frequency of cells with fragmented mitochondria was significantly increased in patients with CCUS and lower- risk MDS compared with that in healthy donors and other patients regardless of the mutational profile (Fig. 4A–C). We note that mitochondrial fragmentation was more prominent in patients with CCUS, who exhibit cytopenia and genetic alteration but no noticeable dysplasia based on the current diagnostic criteria, compared with that in patients with higher-risk MDS (MDS with excess blast; Fig. 4A and B). Moreover, excessive mitochondrial fragmentation was confirmed in CD34+ BM HSC/Ps from patients with MDS (Fig. 4D and E). Taken together, these results suggest that mitochondrial fragmentation is a common cell pathobiological phenomenon in MDS.
Dynamin-Related Protein 1 Activation, at Least Partially through mTOR Signaling, Leads to Mitochondrial Fragmentation
Because mitochondrial depolarization is known to cause mitochondrial fragmentation (36), we tested whether loss of membrane potential contributes to mitochondrial fragmentation. Mitochondrial membrane potential of HSC/Ps was not altered in the CBLΔE8/9/RUNX1S291fs cells compared with control cells (Supplementary Fig. S12), suggesting that other mechanisms may be responsible for excessive mitochondrial fragmentation. Enhanced mitochondrial fission has been reported to drive mitochondrial fragmentation in many cancers (23, 37). Dynamin-related protein 1 (DRP1) is a well-known master regulator of mitochondrial fission (38). Posttranslational regulation processes, such as phosphorylation, are critical for DRP1 activity (39). Thus, to evaluate whether the process of mitochondrial fission is activated in the HSC/Ps of CBLΔE8/9/RUNX1S291fs mice, we performed immunoblot analysis of the DRP1 protein. The expression level of DRP1 protein in c-Kit+ BM cells obtained from CBLΔE8/9/RUNX1S291fs mice was elevated compared with that in cells from control mice (Fig. 5A; Supplementary Fig. S13). Importantly, the level of phosphorylation at serine residue 616 (S616), a key serine for DRP1 activation (39), was increased in the CBLΔE8/9/RUNX1S291fs cells (Fig. 5A; Supplementary Fig. S13). On the other hand, the level of phosphorylation at S637, a key serine for DRP1 inactivation (39), was not prominently altered (Fig. 5A; Supplementary Fig. S13). Expression level of dynamin 2 protein, which could also drive mitochondrial division (40), was also elevated in the CBLΔE8/9/RUNX1S291fs cells (Supplementary Fig. S13). Consistent with the observation in the CBLΔE8/9/RUNX1S291fs mice, the expression level of total DRP1 protein was increased in BMMNCs from patients with MDS and CCUS (Fig. 5B; Supplementary Fig. S14). The increase in the level of phosphorylation at S616 was observed in some patients (Fig. 5B; Supplementary Fig. S14). Furthermore, we confirmed increased expression levels of DRP1 protein and DNM1L, which encodes DRP1 protein, in CD34+ BM cells obtained from patients with MDS (Fig. 5C–E). Previous studies have identified several crucial signaling pathways responsible for regulation of DRP1 activity, such as mTOR, ERK1/2, and Ca2+/calmodulin-dependent kinase signaling (41–43). To assess mechanisms of DRP1 activation in the MDS clone, we performed enrichment analysis of known DRP1 regulatory pathways using differentially expressed genes from RNA-seq data of MDS HSC/Ps (Fig. 5F). Among the known major DRP1 regulatory signaling pathways, we found that CBLΔE8/9/RUNX1S291fs–induced genes were significantly enriched in the mTOR signaling pathway, a critical regulator of phosphorylation at S616 of DRP1 (refs. 41–43; Fig. 5F and G). We confirmed that mTOR-dependent phosphorylation of S6 and eukaryotic translation initiation factor 4E-binding protein 1 (4EBP1) proteins was increased in HSC/Ps from CBLΔE8/9/RUNX1S291fs mice compared with that in control cells (Fig. 5H). Genes in the mTOR signaling pathway were also significantly upregulated in the human MDS cohort compared with the healthy donor cohort (Fig. 5I). To determine whether inhibition of mTOR signaling could reduce the expression level of pS616-DRP1 in MDS cells, we isolated mutant HSC/Ps from CBLΔE8/9/RUNX1S291fs and control mice and treated them with the mTOR inhibitor rapamycin. The expression levels of both pS616-DRP1 and total DRP1 proteins in the CBLΔE8/9/RUNX1S291fs cells were reduced but still higher than those in control cells, as determined via mTOR signaling inhibition (Fig. 5J). Considered together, these data suggest that enhanced DRP1-mediated fission drives mitochondrial fragmentation in the MDS clone, at least partially, through mTOR signaling.
Inhibition of Aberrant Mitochondrial Fission Rescues Ineffective Hematopoiesis in MDS
Excessive mitochondrial fragmentation could drive cellular ROS production, inflammation, and several types of cell death (44). Thus, we hypothesized that enhanced fission–mediated mitochondrial fragmentation in mutant HSC/Ps may function as a trigger of ineffective hematopoiesis in MDS BM. To test this hypothesis, we treated HSC/Ps obtained from the CBLΔE8/9/RUNX1S291fs and CD34+ BM cells obtained from patients with MDS with mdivi-1, a widely used inhibitor of DRP1 activity (Fig. 6A; ref. 45). The mdivi-1 treatment promoted mitochondrial aggregation in the CBLΔE8/9/RUNX1S291fs HSC/Ps (Fig. 6B–D). In contrast, the inhibition of ROS with N-acetyl-l-cysteine (NAC) did not rescue the excessive fragmentation in the CBLΔE8/9/RUNX1S291fs HSC/Ps (Fig. 6B–D). This indicates that the increased ROS level observed in CBLΔE8/9/RUNX1S291fs mice was a consequence of excessive mitochondrial fragmentation. Moreover, mdivi-1 treatment promoted mitochondrial aggregation (Fig. 6E and F) and decreased the frequency of cells with fragmented mitochondria in CD34+ BM cells obtained from patients with MDS (Fig. 6G).
To evaluate the effect of mdivi-1 in vivo, we first treated WT mice and determined its effect on normal hematopoiesis (Supplementary Fig. S15A). Body weight and PB cell counts were comparable between vehicle- and mdivi-1–treated groups (Supplementary Fig. S15B and S15C). Flow cytometric analysis revealed that mdivi-1 treatment did not affect the proportion of each cell lineage in PB, HSC/Ps, and BM granulopoiesis of WT mice (Supplementary Fig. S15D–S15H). We then treated mice with MDS with mdivi-1 (Fig. 7A). Mitochondrial fragmentation in HSC/Ps f mdivi-1–treated CBLΔE8/9/RUNX1S291fs mice was significantly decreased compared with that in vehicle control (Fig. 7B). After mdivi-1 treatment, leukocytopenia phenotype in PB from the CBLΔE8/9/RUNX1S291fs mice was rescued (Fig. 7C and D). Mdivi-1 treatment attenuated the clonal dominance of the mutant HSC/Ps (Fig. 7E and F). Among mutant HSC/Ps, the frequency of the mutant GMP population significantly decreased in BM of the mdivi-1–treated group (Fig. 7G). The impaired BM granulopoiesis phenotype (Fig. 7H and I) and dysplasia formation in myeloid lineage of CBLΔE8/9/RUNX1S291fs mice (Fig. 7J and K) were rescued in the mdivi-1–treated group. Inhibition of enhanced mitochondrial fission reduced the level of cellular ROS (but not mitochondrial ROS) in CBLΔE8/9/RUNX1S291fs HSC/Ps (Fig. 7L–O), suggesting that excessive mitochondrial fragmentation could lead to an increase in cellular ROS levels in this model. The levels of several inflammatory cytokines and chemokines in the plasma remarkably decreased upon mdivi-1 treatment (Fig. 7P).
To determine the role of the increased ROS levels in the development of MDS phenotypes, we treated CBLΔE8/9/RUNX1S291fs mice with NAC. Although the level of cellular ROS in HSC/Ps of NAC-treated mice decreased to the level of that in mdivi-1–treated mice (Fig. 7L–O), NAC treatment did not restore leukocytopenia, dysplasia formation, and impaired BM granulopoiesis phenotypes in mice with MDS (Supplementary Fig. S16A–S16F). The levels of plasma inflammatory cytokines and chemokines in mice with MDS were partially decreased after NAC treatment (Supplementary Fig. S16G and S16H). These results indicate the contribution of other critical mechanisms besides ROS-induced signaling in the development of mitochondrial fragmentation–derived MDS pathogenesis.
To better characterize the effect of mdivi-1 treatment in this model, we performed RNA-seq analysis of mutant HSC/Ps and BM granulocytes obtained from CBLΔE8/9/RUNX1S291fs mice with or without mdivi-1 treatment. Gene expression profiling in mdivi-1–treated CBLΔE8/9/RUNX1S291fs HSC/Ps and mature granulocytes showed that genes downregulated by mdivi-1 were significantly enriched in the IFN-related inflammatory signaling pathway, which is one of the most significantly dysregulated gene pathways in patients with MDS (Supplementary Fig. S17; Supplementary Table S7; ref. 31). Other inflammatory signaling pathways, which were upregulated in CBLΔE8/9/RUNX1S291fs HSC/Ps compared with control HSC/Ps (Fig. 2H), were also downregulated in mdivi-1–treated CBLΔE8/9/RUNX1S291fs HSC/Ps (Supplementary Fig. S18). On the basis of the analysis of the gene expression profile of WT control mice, we identified distinct gene clusters (cluster 2 in HSC/Ps and cluster 4 in Ly6G+ BM myeloid cells) that were prominently upregulated in vehicle-treated CBLΔE8/9/RUNX1S291fs mice and downregulated to normal levels in mdivi-1–treated CBLΔE8/9/RUNX1S291fs mice (Fig. 7Q and R; Supplementary Tables S8 and S9). Gene cluster 2 in HSC/Ps and cluster 4 in Ly6G+ BM myeloid cells were significantly enriched for several inflammatory pathways related to the signaling of IFN stimulation, interleukin, cytokine, Toll-like receptor, and canonical/noncanonical NFκB (Fig. 7S and T; Supplementary Fig. S19; Supplementary Tables S10 and S11). Recently, critical roles of the noncanonical NFκB pathway in clonal advantage of MDS HSC/Ps and disease progression have been reported (7). Mdivi-1 treatment significantly downregulated the noncanonical NFκB pathway (CD40/CD40L signaling) in mutant HSC/Ps of CBLΔE8/9/RUNX1S291fs mice (Supplementary Figs. S19 and S20). OXPHOS gene signature was not altered in HSC/Ps of mice with MDS after mdivi-1 treatment (Supplementary Fig. S20). Concordant with the phenotypic alteration of myeloid progenitors and blood counts in PB, the expression levels of genes related to erythroid and megakaryocyte development were comparable among CBLΔE8/9/RUNX1S291fs mice with or without mdivi-1 treatment (Fig. 7Q and R).
Finally, to assess in vivo therapeutic efficacy of targeting mitochondrial dynamics in MDS, we observed the survival of CBLΔE8/9/RUNX1S291fs mice with or without mdivi-1 treatment. Inhibition of excessive mitochondrial fission significantly prolonged the survival of CBLΔE8/9/RUNX1S291fs mice (Fig. 7U and V; Supplementary Fig. S21). Together, these results suggest that enhanced fission-mediated mitochondrial fragmentation in mutant HSC/Ps causes subsequent inflammatory signaling activation and impairment of granulopoiesis, leading to ineffective hematopoiesis in MDS (Supplementary Fig. S22).
We identified a previously unreported CBLΔE8/9 comutation in patients with MDS with RUNX1 mutations. By focusing on this comutation, we established a useful in vivo preclinical model of MDS BM failure and identified DRP1-dependent mitochondrial fragmentation as a critical pathobiological mediator of, and thus a potential therapeutic target for, ineffective hematopoiesis in MDS. Dysregulated mitochondrial dynamics due to fission–fusion imbalance have been identified in a variety of diseases, including cancers (23, 35). Our data indicate that mitochondrial fragmentation causes abundant cellular ROS production, activates inflammatory signaling in the MDS clone, confers clonal dominance to mutant HSC/Ps, and promotes subsequent impairment of myeloid maturation within the mutant clone. As electron leak from the mitochondrial electron transport chain produces ROS, mitochondria are considered to contribute to the processes of aging and cell death (46); overwhelming mitochondrial fragmentation, in particular, could release large amounts of ROS into the cytoplasmic space (35). Besides mitochondria, many enzymatic reactions, including NADPH oxidases, as well as other organelles (endoplasmic reticulum and peroxisome), also contribute to cellular ROS production in response to cytokines, growth factors, and various cellular stresses (47). Growing evidence highlights that ROS, as signaling molecules, are a critical trigger for activating both innate and adaptive immune signaling (3, 47, 48). Inflammation and cellular stress or damage further facilitate cellular ROS production (49). However, NAC-mediated ROS inhibition exhibited a limited effect on the levels of plasma inflammatory cytokines/chemokines in mice with MDS. Several other MDS phenotypes, which were rescued by the inhibition of mitochondrial division with mdivi-1, were not restored upon ROS inhibition. Thus, other mechanisms in addition to ROS-induced signaling may be critical for aberrant mitochondrial dynamics–derived MDS pathogenesis. We found that IFN-related signaling signature, such as genes related to IFN stimulation- and retinoic acid–inducible gene I (RIG-I)–like receptor pathways, was downregulated in HSC/Ps and myeloid cells of mice with MDS after mdivi-1 treatment. It has been well documented that cytosolic DNA, particularly mitochondrial DNA (mtDNA) released due to dysregulated mitochondrial dynamics, could stimulate several innate immune pathways, such as cyclic GMP–AMP synthase (cGAS)/stimulator of interferon genes (STING) DNA sensing and Nod-like receptor (NLR)–inflammasome pathways, which could subsequently lead to activation of IFN-stimulated genes (ISG) and excessive inflammatory programmed cell death, such as pyroptosis and necroptosis (50, 51). Moreover, cytoplasmic DNA is released into the circulation during cellular injury and further accelerates systemic inflammatory responses (50, 52). In cancer cells including MDS HSC/Ps, the accumulation of mtDNA in the cytoplasm and an increase in circulating mtDNA levels have been reported (50, 53). Mitochondrial double-strand RNA also stimulates RIG-I–like receptors/mitochondria antiviral signaling protein pathway, leading to ISG activation (50, 54, 55). Such noninfectious inflammation caused by dysregulated innate immunity is well documented as a hallmark of MDS pathogenesis (2, 3). As excessive mitochondrial fragmentation was observed in patients with MDS regardless of the mutational profile, these DNA/RNA-sensing pathways may be responsible for the inflammatory signaling activation as a consequence of mitochondrial fragmentation in MDS. Furthermore, mitochondrial fission is important for the mitophagy process that could eliminate impaired mitochondria (23). Mitophagy plays a critical role in maintaining self-renewal activity of normal HSCs and leukemic HSCs (56, 57). However, unlike AML pathogenesis, the competitive advantage of MDS HSC/Ps compared with normal HSC/Ps does not primarily rely on increased self-renewal potential but on other mechanisms, including noncanonical NFκB signaling activation (7). Thus, the impact of mitophagy on clonal expansion and fitness of MDS HSC/Ps is still unknown. Further elucidation of downstream mechanisms of mitochondrial fragmentation that contribute to inflammatory signaling activation, clonal advantage, and impaired myeloid differentiation of the MDS clone will enhance the understanding of the complex MDS pathobiology.
Mechanistically, we found that mTOR signaling is activated in CBLΔE8/9/RUNX1S291fs HSC/Ps and the MDS cohort and is associated with increased DRP1 expression and phosphorylation at S616, one of the key phosphorylation sites for DRP1 regulation. However, our data also indicated the involvement of other factors in DRP1 regulation. Indeed, DRP1 expression and its activity are regulated by several posttranslational modifications, such as phosphorylation, ubiquitination, and sumoylation, which also involve a variety of signaling pathways (23). Transcriptional regulation of DRP1 has also been reported (58). Future studies are required to elucidate the mechanisms underlying DRP1 activation with respect to MDS pathogenesis. As increased expression and activity of DRP1 were observed in mice and patients with MDS, we used the compound mdivi-1, which is an extensively studied and widely used inhibitor of DRP1 activity, to inhibit mitochondrial division. Although several recent studies showed that mdivi-1 does not affect DRP1 activity but mitochondrial metabolism (such as complex I activity and ROS production; ref. 59), we observed increased mitochondrial aggregation and reduced frequency of cells with fragmented mitochondria in mdivi-1–treated HSC/Ps of mice and patients with MDS. OXPHOS gene signature and mitochondrial ROS production were not altered in HSC/Ps of mice with MDS after mdivi-1 treatment. Thus, consistent with the findings of previous and recent studies (42, 45, 60, 61), our results suggest that mdivi-1 primarily affects mitochondrial morphology. Furthermore, it might depend on cell context or conditions. Nonetheless, considering that excessive mitochondrial fragmentation was observed in patients with MDS regardless of the mutational profile and that restoration of mitochondrial fragmentation by mdivi-1 rescued dysplastic and ineffective hematopoiesis phenotypes in mice with MDS, targeting aberrant mitochondrial dynamics may be a potential therapeutic option in MDS.
Although many genetic alterations have been identified in patients with MDS during the past decade (11), myeloid dysplasia is a fundamental common feature of MDS regardless of the mutational profile (1, 62, 63). Our data showed that inhibition of mitochondrial fragmentation significantly attenuated inflammatory signaling activation in both MDS HSC/Ps and mature myeloid cells, leading to improved myeloid dysplasia phenotype. This indicates that activation of inflammatory pathways, such as IFN-related signaling, may be the underlying mechanism that dictates myeloid dysplasia formation. Interestingly, our analysis of mitochondrial structure in human BMMNCs and CD34+ cells, using confocal fluorescence microscopy, revealed that excessive mitochondrial fragmentation is observed in patients with not only MDS but also CCUS. Because of the identification of clonal hematopoiesis in aging healthy individuals (64), a variety of potential pre-MDS conditions have been defined, such as ICUS, CHIP, and CCUS (63). According to the current criteria for diagnosing dysplasia, patients with CCUS do not exhibit dysplasia but exhibit cytopenias and MDS-related gene alterations (63). However, considering that the existence of ringed sideroblasts due to mitochondrial iron accumulation is currently one of the important morphologic criteria for MDS diagnosis, excessive mitochondrial fragmentation may also be included in the diagnostic criteria of dysplasia.
Despite multiple life-threatening conditions, such as compromised states due to leukocytopenia (8), the available effective therapeutic strategies for patients with MDS are limited. Our data indicate that modulation of enhanced mitochondrial fission may provide a useful therapeutic option in patients with MDS who are compromised due to leukocytopenia triggered by ineffective hematopoiesis. However, macrocytic anemia and the thrombocytopenia phenotype were not rescued by mdivi-1 treatment, indicating the involvement of other mechanisms for this phenotype. Inhibition of NLR inflammasome activity reportedly rescued the erythroid differentiation capacity of MDS HSC/Ps (65). In addition to the mutant-derived cell-intrinsic program, innate immune cells, such as macrophages, also contribute to the impairment of MDS erythropoiesis via cell-extrinsic mechanisms (5, 66). Contribution of innate immunity as well as adoptive immunity to thrombocytopenia has also been well documented. Although mdivi-1 treatment strongly attenuated IFN-related inflammatory signaling in the MDS cl one in this study, an inhibitory effect of mdivi-1 treatment on innate immune signaling may not be sufficient. Thus, direct inhibition of innate immune signaling activation or targeting of innate immune cells may be effective for complex anemia phenotypes in MDS.
In summary, our data indicate that dysregulated mitochondrial dynamics in MDS HSC/Ps cause abundant cytoplasmic ROS production and constitutively activate inflammatory signaling, leading to progenitor expansion and impaired granulopoiesis in mutant clones. Our study provides new insights into the cell biological mechanism of ineffective hematopoiesis in MDS and clues for targeting BM failure caused by ineffective hematopoiesis.
Patients and BM Samples
Patients with MDS (n = 37), AML (n = 12), CHIP/ICUS (n = 4), CCUS (n = 4), and other diseases or conditions (aplastic anemia, n = 1; multiple myeloma, n = 1; malignant lymphoma, n = 2; and macrocytic anemia, n = 1) were enrolled in this study. The study was approved by the institutional review board, and all patients provided written informed consent. Mononuclear cells were isolated from diagnostic BM aspirates and healthy donors (n = 4). Some BMMNCs obtained from healthy donors (n = 4) were purchased from Lonza. These cells were isolated after obtaining permission for their use in research applications by informed consent or legal authorization. Genomic DNA was extracted from mononuclear cells using a Gentra Puregene Blood Kit (Qiagen) in accordance with the manufacturer's instructions. CD34+ BM cells were sorted using cell sorters (SH800, SONY; FACSAriaIII, BD Biosciences) from mononuclear cells.
Targeted sequencing was performed using AmpliSeq for Illumina Myeloid Panel (Illumina) and a custom-designed panel to detect mutations in 68 genes (Supplementary Table S1). As a template, 10 ng DNA was used to amplify the target genes. AmpliSeq Library Plus for Illumina (Illumina) was used to generate libraries. The libraries were analyzed using a MiniSeq High Output Reagent Kit (300 cycles) on a MiniSeq (Illumina) platform.
Detection of Variants
The fastq files generated herein were cleaned using Trimmomatic, and the results were aligned to the human reference genome, hg19, with Burrows–Wheeler Alignment. Gene variants were detected using HaplotypeCaller (for high-frequency variants) and Mutect2 (for low-frequency variants), which is included in GATK. Gene variants obtained from HaplotypeCaller were filtered using the following parameters: quality/depth, mapping quality, and strand bias to exclude error-suspected variants. The variants from Mutect2 were filtered via GATK FilterMutectCalls. Variants were annotated with information from Refseq and Exac databases in VariantStudio 3.0 software (Illumina). Variants with frequencies greater than 1% in the regional population were excluded. Finally, mutations were manually checked by researchers experienced in hematologic malignancies.
Detection of CBL Exon Deletion
A large deletion of CBL was detected via genomic PCR with a forward primer in exon 7 (5′-GTCGAGATATCTAGACCCAG-3′) and a reverse primer in exon 10 (5′-GCGCCCTGTGCACAGGATGT-3′; Supplementary Fig. S1A). The following parameters were used for PCR amplification using PrimeSTAR GXL DNA Polymerase (TAKARA): 98°C for 3 minutes, followed by 35 cycles at 98°C for 10 seconds, 70°C for 15 seconds, and 68°C for 30 seconds.
C57BL/6 mice (7-week-old, female and male) were purchased from Tokyo Laboratory Animals Science. All animal studies were approved and conducted according to the Institutional Animal Care Protocol and Guidelines of the Tokyo University of Pharmacy and Life Sciences.
Immunofluorescence Staining and Confocal Microscopy
Cell suspensions of BMMNCs from patients and healthy donors, FACS-sorted (SH800, SONY; FACSAriaIII, BD Biosciences) c-Kit+ BM cells from CBLΔE8/9/RUNX1S291fs or control mice, and FACS-sorted CD34+ BM cells from patients or healthy donors were mixed with Smear Gell (GenoStaff) and spread on slides. Cells were then fixed with 4% paraformaldehyde in PBS, after which 15 mmol/L glycine in PBS was added. Following PBS washing, cells were stained with the following antibodies: anti-Tom20 (sc17764, Santa Cruz Biotechnology), anti-Drp1 (611112, BD Biosciences), Alexa Fluor 647–conjugated anti-mouse IgG (H+L; a21235, Invitrogen), and Alexa Fluor 594–conjugated anti-mouse IgG (H+L; a32742, Invitrogen). Cells were then washed three times with PBS, mounted with Hoechst, and examined using an Olympus FV1000-IX81/Olympus FV3000-IX83 confocal fluorescence microscope (Olympus) and a BZ-X700 fluorescence microscope (Keyence). Mitochondria were classified as fragmented when most of the mitochondria of the cell were spherical. Integrated density of DRP1 and area occupied by mitochondria in each cell were measured using ImageJ (NIH, Bethesda, MD) threshold analysis. To assess mitochondrial membrane potential, FACS-sorted c-Kit+ BM cells were stained with MitoBright LT Red (Dojindo) or MT-1 dye (Dojindo) prior to spreading on slides.
Transmission Electron Microscopy
Cell suspensions of FACS-sorted (SH800, SONY) c-Kit+ BM cells from CBLΔE8/9/RUNX1S291fs or control mice were fixed with 2.5% glutaraldehyde at 4°C for overnight, then fixed with 2% osmium tetroxide at 4°C for 2 hours. Samples were dehydrated with ethanol (30%–100%) and embedded in Epon812. Using an ultra-microtome, samples were cut into ultrathin sections (80–90 nm) and stained with uranyl acetate and lead citrate. Images were acquired using a Hitachi H-7600 transmission electron microscope (Hitachi) operating at 100 kV.
Flow cytometric analysis was performed using FACSCanto (BD Biosciences), FACSCelesta (BD Biosciences), and CytoFLEX (Beckman Coulter). Murine PB and BM cells were incubated with the following anti-mouse antibodies: CD11b (M1/70), c-Kit (2B8), Sca1 (D7), CD48 (HM48-1), CD150 (TC15-12F12.2), CD115 (AFS98), and Ly6G (1A8; BioLegend). In analyses of HSC/Ps, biotin-conjugated anti-mouse CD3e (145-2C11), CD4 (RM4-5), CD8a (53-6.7), CD19 (6D5), B220 (RA3-6B2), Ter119 (Ter-119), CD11b (M1/70), and Gr-1 (RB6-8C5) antibodies (BioLegend) were used followed by staining with streptavidin-PerCP/Cy5.5 for lineage exclusion. In analyses of myeloid differentiation, biotin-conjugated CD3e (145-2C11), CD4 (RM4-5), CD8a (53-6.7), CD19 (6D5), B220 (RA3-6B2), and Ter119 (Ter-119) antibodies (BioLegend) were used, followed by staining with streptavidin-PerCP/Cy5.5 for exclusion of lymphoid and erythroid lineages. NGFR protein was detected by anti-human CD271 (ME20.4) antibody (BioLegend). CellROX Deep Red Reagent (Thermo Fisher Scientific) and MitoSOX Red mitochondrial superoxide indicator (Thermo Fisher Scientific) were used according to the manufacturer's protocol for cellular ROS and mitochondrial ROS detection, respectively. PBS containing 2% FBS and 2 mmol/L EDTA was used as a FACS buffer. Data were analyzed using FlowJo software (BD Biosciences).
Whole-cell extracts were prepared by sonicating cells in a SDS sample buffer containing 0.2 mmol/L Na3VO4 (New England Biolabs), 1 mmol/L phenylmethylsulfonylfluoride (Sigma-Aldrich), and 10% 2-mercaptoethanol and proteinase inhibitors (Roche Diagnosis). After boiling at 95°C for 10 minutes, samples were separated via SDS-PAGE and electro-transferred to polyvinylidene fluoride membranes (Millipore). The following primary antibodies were used: anti-DRP1 (611112, BD Biosciences), anti–pDRP1-S616 (4494, Cell Signaling Technology), anti–pDRP1-S637 (4867, Cell Signaling Technology), anti–phospho-S6 ribosomal protein (Ser240/244; 2215, Cell Signaling Technology), anti–phospho-4E-BP1 (Thr37/46; 2855, Cell Signaling Technology), anti–Dynamin 2 (ab3457, abcam), anti–FLAG M2 (F3165, Sigma-Aldrich), anti–c-CBL (8447, Cell Signaling Technology), anti-RUNX1 (N-terminal; R0406, Sigma-Aldrich), and anti–β-actin (sc47778, Santa Cruz Biotechnology). Horseradish peroxidase antibodies conjugated to rabbit IgG (7074, Cell Signaling Technology) or mouse IgG (7076, Cell Signaling Technology) were used as a secondary antibody. Antibodies were diluted in Can Get Signal solutions (Toyobo). ECL Prime Western Blotting Detection Reagent (GE Healthcare) was used for ECL detection.
Plat-E packaging cells were transfected with retrovirus vectors (pMYs-IRES-eGFP, pMYs-RUNX1-S291fs-IRES-hNGFR, pMYs-RUNX1-S291fs-IRES-eGFP, pMYs-CBL exon 8/9 deletion-IRES-eGFP, pMYs-PTPN11-E76K-IRES-hNGFR) using FuGENE HD (Promega). The medium was changed 24 hours following transfection. The supernatant containing retroviruses was collected 48 hours after transfection and concentrated using polyethylene glycol solution.
Mouse BMT Experiment
C57BL/6 donor mice (female or male) were euthanized via cervical dislocation under anesthesia 4 days after administering 150 mg/kg 5-fluorouracil (Kyowa Kirin) intraperitoneally. BM cells were isolated from the femurs and tibias of donor mice immediately following euthanasia. After stimulation with 50 ng/mL mouse stem cell factor (SCF), mouse FMS-like tyrosine kinase 3 ligand, mouse IL6, and mouse thrombopoietin (TPO; PeproTech), cells were transduced with the retrovirus for 60 hours using RetroNectin (Takara)–coated dishes. Next, 3–10 × 105 of the nonsorted cells (including 1–2.5 × 105 of CBLΔE8/9/RUNX1S291fs double-transduced cells) were injected through the tail vein into sublethally irradiated WT recipient mice.
RNA-seq and Data Analysis
BM cells were obtained from CBLΔE8/9/RUNX1S291fs mice at 1 to 2 months after BMT before becoming moribund or mice with empty/single mutant alone at 2 to 4 months after BMT. BM c-Kit+ cells were separated using an EasySep mouse c-Kit Positive Selection Kit (STEMCELL Technologies). Next, mutant or empty vector(s)–transduced cells within the c-Kit+ cells were FACS sorted. For Ly6G+ cells, total BM cells were immunostained, and B220−, CD3−, Ter119−, c-Kit−, Ly6G+, and CBLΔE8/9/RUNX1S291fs double-transduced BM cells were FACS sorted. Total RNA was isolated using the RNeasy Mini Kit (Qiagen). RNA degradation and contamination were monitored on 1% agarose gels. RNA purity was checked using a NanoPhotometer spectrophotometer (IMPLEN). RNA integrity and quantitation were assessed using the RNA Nano 6000 Assay Kit and the Bioanalyzer 2100 system (Agilent Technologies). A total amount of 1 μg RNA per sample was used as input material for RNA sample preparations. Sequencing libraries were generated using the NEBNext Ultra RNA Library Prep Kit for Illumina (New England Biolabs) following the manufacturer's recommendations, and index codes were added to attribute sequences to each sample. Briefly, mRNA was purified from total RNA using poly-T oligo–attached magnetic beads. Fragmentation was carried out using divalent cations under elevated temperature in NEBNext First Strand Synthesis Reaction Buffer (5×; New England Biolabs). First-strand cDNA was synthesized using the random hexamer primer and M-MuLV Reverse Transcriptase (RNase H−). Second-strand cDNA synthesis was subsequently performed using DNA Polymerase I and RNase H. Remaining overhangs were converted into blunt ends via exonuclease/polymerase activities. After adenylation of the 3′ ends of DNA fragments, a NEBNext Adaptor (New England Biolabs) with a hairpin loop structure was ligated to prepare for hybridization. To preferentially select cDNA fragments of 150 to 200 bp in length, the library fragments were purified with the AMPure XP system (Beckman Coulter). Then, 3 μL USER Enzyme (New England Biolabs) was used with size-selected, adaptor-ligated cDNA at 37°C for 15 minutes followed by 5 minutes at 95°C before PCR. PCR was performed with Phusion High-Fidelity DNA polymerase, Universal PCR primers, and Index Primer. Finally, PCR products were purified (AMPure XP system) and library quality was assessed on an Agilent Bioanalyzer 2100 system. Clustering of index-coded samples was performed on a cBot Cluster Generation System using the PE Cluster Kit cBot-HS (Illumina) according to the manufacturer's instructions. Following cluster generation, library preparations were sequenced on an Illumina platform and paired-end reads were generated. Reads contained in FASTQ files were mapped to modified mice reference sequence (mm10) that contained the human CBL (NM_005188.4) and RUNX1 (NM_001001890.3) sequence using STAR (67). Mapped reads were counted by featureCounts (68) and calculated to TPM with R. To confirm the mutation and deletion in transduced genes, the Integrative Genomics Viewer (69) was used. For the analyses of RNA-seq data, FASTQ files were processed in AltAnalyze (version 2.1.0; ref. 70). Transcript per million estimates were calculated in AltAnalyze via a built-in call to Kallisto for read pseudoalignment and transcript quantification (Ensembl 72, mm10). Differentially expressed genes (DEG) were selected in AltAnalyze (fold change >2.0 or 1.5, P < 0.05). DEG log2 folds relative to the row mean were used for generating heat maps and hierarchical clustering in AltAnalyze. MarkerFinder analysis was performed using the GO-Elite algorithm in AltAnalyze. The data have been deposited in Gene Expression Omnibus (GSE151694, https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE151694 and GSE154308, https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE154308).
BM c-Kit+ cells were separated using an EasySep Mouse c-Kit Positive Selection Kit (STEMCELL Technologies) and plated in Cell-Tak (Corning)–coated Seahorse XF microplates (Agilent) at 5.5 × 105 cells/well. The rates of oxygen consumption in c-Kit+ BM cells were measured using a Seahorse XF under basal conditions and after injection of 1.5 μmol/L oligomycin (Agilent), 1.0 μmol/L FCCP (Agilent), 0.5 μmol/L Antimycin A, and 0.5 μmol/L Rotenone (Agilent).
Complete Blood Counts
Blood counts were measured using pocH-100iV Diff (Sysmex).
Hematoxylin and Eosin Staining and Reticulin Staining
Collected tibia were fixed using 10% Formalin Neutral Buffer Solution (Fujifilm Wako Pure Chemical Corporation). Specimens were decalcified and paraffin-embedded for hematoxylin and eosin and reticulin staining. These stainings were performed at Genostaff.
Cytospin slides of mice BM cells were fixed with 2.5% glutaraldehyde (Fujifilm Wako Pure Chemical Corporation) in PBS, after which an equivalent amount of 2% HCl solution and potassium ferrocyanide solution (Scy-tek) were added. Following water washing, Nuclear Fast Red (Scy-tek) was added. Morphologic images were obtained using the Eclipse Ni (Nikon) and USB cam system (HOZAN).
Cytospin slides of mice BM cells were stained using Diff-Quik (SYSMEX). Morphologic images were obtained using the Eclipse Ni (Nikon) and USB cam system (HOZAN).
Cytokine Array Analysis
PB was collected from CBLΔE8/9/RUNX1S291fs mice or control mice under anesthesia. Collected whole blood was immediately centrifuged, and the plasma fraction was aliquoted. The expression level of multiple cytokines and chemokines in plasma was determined using Proteome Profiler Mouse Cytokine Array Kit, Panel A (R&D Systems).
To generate the CBLΔE8/9/RUNX1S291fs-induced gene set, 201 genes demonstrating more than threefold upregulation (P < 0.05) in CBLΔE8/9/RUNX1S291fs c-Kit+ cells, compared with control cells, were used. The CBLΔE8/9/RUNX1S291fs-induced gene set was then tested for enrichment in the human MDS cohort (GSE19429) using GSEA (v4.0.3; ref. 71). The following parameters were used: 1,000 gene set permutations, Signal2Noise ranking metric, and descending-order real mode for gene sorting. The leading-edge CBLΔE8/9/RUNX1S291fs–induced genes in MDS CD34+ cells (Fig. 2B) were tested in the genome sequencing/gene expression data set of human MDS CD34+ BM cells (GSE58831). The following parameters were used: 1,000 gene set permutations, Signal2Noise ranking metric, and descending-order real mode for gene sorting. Gene sets of oxidative phosphorylation and mTORC1 in hallmark gene sets of the Molecular Signatures Database (MSigDB; ref. 71) were tested for enrichment in the data set of murine c-Kit+ cells obtained from CBLΔE8/9/RUNX1S291fs and control mice. The following parameters were used: 1,000 gene set permutations, Diff_of_Classes ranking metric, and descending-order real mode for gene sorting. The oxidative phosphorylation and mTORC1 gene sets were also tested in the human MDS cohort. The following parameters were used: 1,000 gene set permutations, Signal2Noise ranking metric, and descending-order real mode for gene sorting. The CD40 signaling and oxidative phosphorylation gene sets were tested for enrichment in the data sets of c-Kit+ BM cells obtained from CBLΔE8/9/RUNX1S291fs with or without mdivi-1 treatment. The following parameters were used: 1,000 gene set permutations, Diff_of_Classes ranking metric, and descending-order real mode for gene sorting.
Pathway and Network Enrichment Analysis
Pathway enrichment analyses of differentially expressed genes were performed using ToppFun in the ToppGene Suite (72) and EnrichR (ref. 73; Figs. 2H, 5F, 7S, and 7T; Supplementary Figs. S9 and S17–S19). The following parameters for ToppGene analysis were used: Reactome pathway, P < 0.05, 15 < gene size < 500; KEGG pathway, P < 0.05, 15 < gene size < 500. Pathway enrichment analysis of marker genes was performed using KEGG pathways and the biomarker gene set of the GO-Elite algorithm in AltAnalyze. GO term analysis was performed using BiNGO (74) in Cytoscape (v3.7.2; ref. 75). Network analysis was performed using EnrichmentMap (76) with AutoAnnotate (ref. 77; Fig. 2G).
Mdivi-1 and NAC Treatment
From day 14 after BMT, CBLΔE8/9/RUNX1S291fs mice were injected with 2.5 mg/kg mdivi-1 (Selleck), 200 mg/kg NAC (Nacalai-tesque), or DMSO vehicle intraperitoneally daily for 5 to 6 weeks (5 days a week). WT mice were also treated with mdivi-1 or DMSO vehicle daily for 2 weeks (5 days a week). For in vitro treatment, FACS-sorted c-Kit+ BM cells from CBLΔE8/9/RUNX1S291fs mice were cultured in Iscove modified Dulbecco medium (IMDM) containing 20% FBS, 50 ng/mL mouse SCF, and 50 ng/mL mouse TPO (PeproTech). FACS-sorted CD34+ BM cells from patients' samples were cultured in IMDM containing 20% FBS, 100 ng/mL human SCF, and 100 ng/mL human TPO (PeproTech). These cells were treated with 10 nmol/L mdivi-1 (Selleck), 3 mmol/L NAC (Nacalai-tesque), 50 nmol/L rapamycin (Selleck), or DMSO vehicle for 24 hours.
GraphPad Prism (v8; GraphPad) was used. Survival of mice was analyzed using the log-rank test. Statistical analyses for comparison of two independent, normally distributed samples were compared using the two-tailed Student t test. For multiple pairwise comparisons, the data were analyzed by one-way ANOVA followed by Tukey multiple comparison test. Statistical significance was set at P < 0.05.
The authors declare no competing interests.
Y. Aoyagi: Conceptualization, investigation. Y. Hayashi: Conceptualization, data curation, supervision, investigation, visualization, writing–original draft, project administration, writing–review and editing. Y. Harada: Resources, investigation, writing–original draft. K. Choi: Investigation. N. Matsunuma: Investigation. D. Sadato: Resources, investigation. Y. Maemoto: Investigation. A. Ito: Investigation. S. Yanagi: Conceptualization, investigation. D.T. Starczynowski: Conceptualization, writing–original draft. H. Harada: Conceptualization, resources, writing–original draft, project administration, writing–review and editing.
We thank Dr. Seishi Ogawa (Kyoto University, Kyoto, Japan) for plasmid, thoughtful discussions, and insight. We thank Dr. Masafumi Onodera (National Center for Child Health and Development, Tokyo, Japan) and Dr. Susumu Goyama (University of Tokyo, Tokyo, Japan) for plasmids. We thank Dr. Akihiko Goto (Tokyo Medical University, Tokyo, Japan) and Dr. Noriko Doki (Tokyo Metropolitan Cancer and Infectious Diseases Center, Komagome Hospital, Tokyo, Japan) for BM samples from patients. We thank Dr. Yuta Fujikawa and Dr. Hiroki Kobayashi (Tokyo University of Pharmacy and Life Sciences) for insightful suggestions and feedback. This study was supported by the JSPS KAKENHI (Grants JP20K17411, to Y. Hayashi; JP16K09831 and JP19K08824, to H. Harada and Y. Harada) and Clinical Research Fund of Tokyo Metropolitan Government (R010302001, to Y. Harada).
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.