Exposure to microbial products followed by loss of Tet2 promotes myelodysplastic syndrome via remodeling HSCs

Myelodysplastic syndrome (MDS) is a clonal disorder that originates from hematopoietic stem cells (HSCs), and is characterized by cytopenia, dysplastic blood cells, and a predisposition to acute myeloid leukemia (Nimer, 2008; Malcovati et al., 2013). Recent genome sequencing studies revealed that clonal hematopoiesis (CH) originating from HSCs in aged individuals often harbored mutations in the epigenetic modifiers frequently observed in MDS, acute myeloid leukemia, and myeloproliferative neoplasms (MPN), such as the DMNT3A, TET2, and ASXL1 genes (Jaiswal et al., 2014; Genovese et al., 2014). Age-related CH is associated with an increased risk of the transformation of MDS, indicating a pre-malignant state (Sperling et al., 2017; Jaiswal and Ebert, 2019). In murine models, a driver mutation in the CH gene induces the development of MDS at a very long latency (Moran-Crusio et al., 2011; Guryanova et al., 2016; Yang et al., 2018), while concurrent mutations in transcription factors (TF) and epigenetic modifiers result in the progressive diseases of MDS under specific pathogen–free conditions (Abdel-Wahab et al., 2013; Muto et al., 2013; Zhang et al., 2016). In patients, the development of MDS appears to require the clonal evolution of malignant HSC, which is largely driven by natural selection to give a competitive advantage to CH clones over normal HSCs residing in the bone marrow (BM) due to additional genetic mutation(s) in stem cells and environmental alterations including infectious diseases (Sallman and List, 2019; Lyne et al., 2021). However, the mechanisms by which a genetic driver mutation and environmental factors interact to induce epigenetic and transcriptional alterations in HSCs, thereby driving the pathogenesis of MDS currently remain unclear.

Aberrant activation of innate immune and inflammatory pathways in MDS HSPCs has been implicated as a driver of the pathogenesis of MDS in murine models and humans (Pietras, 2017; Sallman and List, 2019). The overexpression of and mutations in genes involved in innate immune signaling were reported in patients with 5q- MDS, in which the haploinsufficiency of microRNA 146a located on chromosome 5q induced the overexpression of TRAF6, a Toll-like receptor (Tlr) mediator, leading to the activation of the downstream NF-κB and MAPK pathways (Starczynowski et al., 2010; Muto et al., 2020). A large population-based clinical registry study indicated that a previous history of bacterial and viral infections at more than 3 yr before the diagnosis of MDS significantly increased the risk of MDS (Kristinsson et al., 2011). Therefore, the following possibilities for the clonal evolution of MDS stem cells have been proposed. Multiple and/or continuous infections promote the natural selection of pre-existing mutated HSCs in aged humans showing CH, resulting in the development of MDS, as recently demonstrated in an analysis of a Mycobacterium avium–mediated chronic infection model using Dnmt3a knockout mice (Hormaechea-Agulla et al., 2021). Furthermore, prior infection stress modulates normal (non-mutated) HSCs and facilitates the development of MDS upon the acquisition of genetic driver mutation(s).

We herein investigated the effects of which activation of innate immune signaling due to an infection stimulation and a driver mutation in the Tet2 gene were induced in HSCs on cellular functions, transcriptomes, genomes, and epigenomes to promote the development of MDS. A prior combined stimulation with LPS and polyinosinic:polycytidylic acid (pIpC), which are biochemicals that mimic bacterial and viral infections, compromised normal hematopoiesis, as previously reported (Essers et al., 2009; Baldridge et al., 2010; Takizawa et al., 2017), while the deletion of the Tet2 gene mitigated the compromised repopulation capacity of HSPCs stimulated with LPS and pIpC. Furthermore, the combined stimulation with LPS/pIpC in WT (non-mutated background) mice followed by the deletion of the Tet2 gene promoted the development of MDS more than in naive Tet2-deleted mice (Moran-Crusio et al., 2011; Quivoron et al., 2011; Shide et al., 2012), whereas a stimulation with LPS/pIpC at the same dosage in Tet2-deleted mice did not. The stimulation with LPS/pIpC promoted the activation of innate immune signaling of Tlr-Trif-Polo-like kinase (Plk) and Elf1, an Ets TF that was shown to regulate an anti-viral reprogram at downstream of Plk in innate immune cells but also impede normal erythropoiesis (Seifert et al., 2019; Calero-Nieto et al., 2010). We demonstrated that the activation of Myd88/Trif-Plk signaling was required for the stimulation-enhanced chromatin accessibility and the activation of Elf1-driven transcriptome in HSCs that were augmented by the deletion of Tet2 gene, resulting in the development of MDS. Moreover, the stimulation-induced Elf1-targets signature was significantly enriched in MDS HSPCs in humans, implying the oncogenic function of the Elf1 gene in MDS.

To establish whether prior infection stress facilitated the development of MDS, we selected the in vivo administration of LPS and pIpC, which mimic bacteria and virus infections, to induce Tlr4- and Tlr3-dependent signaling, respectively. We then attempted to clarify the mechanisms by which the stimulation with LPS and pIpC, mimicking systemic infection-induced stress, modulated WT HSCs in order to drive the development of MDS upon the deletion of the Tet2 gene. To achieve this, we intraperitoneally injected 10 μg LPS and 125 μg pIpC (six times each, every other day for 1 mo) or PBS as a control into Tet2^wt/wt;Cre-ERT2 and Tet2^flox/flox;Cre-ERT2 mice, from which 5 × 10⁶ BM cells were transplanted into lethally irradiated recipient mice 2 wk after the last injection of LPS/pIpC (Fig. 1 A). Flow cytometry revealed that HSCs expressed the Tlr4 protein, which was consistent with previous findings (Nagai et al., 2006), and Tlr3 protein levels were lower than those in monocytes, but higher than those in Tlr3 null cells (Fig. S1, A and B). We found that purified WT HSCs responded to the LPS and/or pIpC stimulation under in vitro conditions to activate the expression of canonical NF-κB target genes and IFN-stimulated genes (Fig. S1 C), which indicated that HSCs directly sensed and responded to both LPS and pIpC. While LPS and pIpC were shown to transiently elevate the expression levels of HSC surface markers (e.g., Sca-1; Essers et al., 2009), LPS/pIpC-treated mice showed similar BM cell counts and normalized frequencies of phenotypic HSCs and multipotent progenitors (MPPs) 1–3 in BM to PBS-treated mice after 2 wk, but lower lymphoid-primed MPP4 (Fig. S2; Challen et al., 2021). 1 mo after transplantation, we deleted the Tet2 gene specific for blood cells in recipient mice and observed blood phenotypes for 12 mo (Fig. 1 A). Recipient mice transplanted with PBS-Cre-ERT2, LPS (6×)/pIpC (6×)-Cre-ERT2, PBS-Tet2^Δ/Δ, and LPS (6×)/pIpC (6×)-Tet2^Δ/Δ BM cells were referred to as WT, LPS/pIpC-WT, Tet2^Δ/Δ, and LPS/pIpC-Tet2^Δ/Δ mice, respectively.

Complete blood counts (CBC) did not significantly change in WT or LPS/pIpC-WT mice 13 mo after transplantation, and survival times were similar, whereas LPS/pIpC-Tet2^Δ/Δ mice started to develop anemia and had a larger mean platelet volume 6 mo after transplantation (Fig. 1 B). Moribund LPS/pIpC-Tet2^Δ/Δ MDS mice showed anemia varying hemoglobin levels (Fig. 1 B), but also significantly impeded erythroid differentiation at the CD71⁺Ter119⁺ stage in BM (Fig. 1 C). Moribund LPS/pIpC-Tet2^Δ/Δ mice significantly impaired B-lymphopoiesis in peripheral blood (PB; Fig. 1 D), and had dysplastic neutrophils, anisocytosis, and giant platelets in PB and dysplastic megakaryocytes in BM (Fig. 1, E and F), which are characteristic of MDS (Nimer, 2008; Malcovati et al., 2013). Based on these results, the criteria for murine MDS and MDS/MPN subcategories (Zhou et al., 2015) indicated that the majority of LPS/pIpC-Tet2^Δ/Δ mice died due to MDS (13 out of 20 mice, 65%) and MDS/MPN showing proliferative phenotypes (2 out of 20 mice, 10%); however, 9 out of 15 Tet2^Δ/Δ mice survived, 5 of which developed MDS within 13 mo (Fig. 1 G). The prior LPS/pIpC stimulation led to shorter survival in LPS/pIpC-Tet2^Δ/Δ mice than in naive Tet2^Δ/Δ mice (median survival after the Tet2 deletion, 307.5 d versus undetermined; Fig. 1 H). At 12 mo after transplantation, CBC revealed lower hemoglobin levels in alive LPS/pIpC-Tet2^Δ/Δ mice than in alive Tet2^Δ/Δ mice (Fig. S3), indicating the continuous development of MDS in LPS/pIpC-Tet2^Δ/Δ mice. Therefore, the prior combined stimulation had the ability to promote the development of Tet2-deficient MDS over a long period.

We next investigated whether the combined stimulation with LPS and pIpC in Tet2-deleted (mutated background) cells promoted the development of MDS, which is the reverse order to LPS/pIpC followed by the Tet2 deletion. To achieve this, we deleted the Tet2 gene and then injected LPS and pIpC (at the same dosage in Fig. 1 A, six times each) or PBS for 1 mo into Tet2^Δ/Δ mice, from which BM cells after the 2-wk wait were transplanted into irradiated mice, referred to as Tet2^Δ/Δ-LPS/pIpC mice (Fig. 1 I). During a 13 mo observation period, 8 out of 20 Tet2^Δ/Δ-LPS/pIpC mice developed anemia and had smaller platelet counts in PB, presenting the dysplastic and proliferative features of MDS/MPN and MDS at a lower penetrance (Fig. 1 J and Fig. S3). Tet2^Δ/Δ -LPS/pIpC mice showed similar survival to control Tet2^Δ/Δ-PBS mice (median survivals were undetermined; Fig. 1 K), indicating that the stimulation with LPS/pIpC at the same dosage in Tet2-deficient cells failed to promote the development of MDS, at least in this experimental setting. Direct comparisons of data between Fig. 1, H and K revealed significantly shorter survival in mice stimulated with LPS/pIpC prior to the deletion of Tet2. Overall, the prior combined stimulation and the deletion of the Tet2 gene facilitated the development of MDS.

Based on the progression of MDS in LPS/pIpC-Tet2^Δ/Δ mice, we assessed the effects of different dosages of the LPS/pIpC stimulation on the fitness advantage of mutant cells in BM by performing a competitive transplantation assay using Tet2^wt/wt;Cre-ERT2 and Tet2^flox/flox;Cre-ERT2 mice injected with two dosages (6× or 12×) of LPS/pIpC followed by a 2-wk wait (Fig. 2 A). HSC potential was compromised by the exposure to LPS or pIpC (Essers et al., 2009; Takizawa et al., 2017). Flow cytometry revealed that LPS (6×)/pIpC (6×)-WT cells severely compromised the production of mature cells in PB (Fig. 2 B) and of CD150^-CD34⁺n B (Fig. 2 C), which was more prominent in LPS (12×)/pIpC (12×)-WT cells, including the CD150⁺CD34⁻LSK long-term (LT)-HSC and CD150^midCD34⁺LSK short-term HSC fractions (Fig. 2 D). In the absence of the LPS/pIpC stimulation, Tet2^Δ/Δ cells showed a slightly higher repopulation capacity of myeloid cells in PB than WT cells. The deletion of the Tet2 gene mitigated LPS/pIpC-dependent reductions in the competitive repopulation of MPPs and GMPs in LPS (6×)/pIpC (6×)-WT cells in BM (Fig. 2 C), but not in mature myeloid cells in PB (Fig. 2 B). In comparisons with WT cells, the reduced competitive repopulation was partially cancelled in HSCs, MPPs, and GMPs in LPS (12×)/pIpC (12×)-Tet2^Δ/Δ cells (Fig. 2 D). Therefore, the deletion of the Tet2 gene mitigated the impaired repopulating capacity of HSPCs and myeloid progenitor cells in the BM due to the stimulation with LPS/pIpC in a dose-dependent manner; however, the impaired differentiation of myeloid cells persisted in the PB.

To elucidate the mechanisms underlying the clonal expansion and impaired differentiation of LPS/pIpC-Tet2^Δ/Δ MDS HSCs, we performed sequential RNA sequencing (RNA-seq) analyses of CD150⁺CD34⁻LSK HSCs isolated from mice with four genotypes 4 mo at a pre-disease stage and 12 mo after transplantation, including two LPS/pIpC-Tet2^Δ/Δ MDS mice. A principal component analysis using the transcriptomes of all genes showed that the four genotypes at 4 mo were located closer than those at 12 mo, revealing larger variabilities in Tet2^Δ/Δ HSCs and LPS/pIpC-Tet2^Δ/Δ MDS HSCs (Fig. 3 A) than in WT HSCs. Because there were more changes in the transcriptome in LPS/pIpC-Tet2^Δ/Δ MDS HSCs than in Tet2^Δ/Δ HSCs at 12 mo (Fig. 3 B), we investigated differentially expressed genes in these HSCs. A hierarchical clustering map clarified that LPS/pIpC-Tet2^Δ/Δ HSCs shared differentially expressed genes with LPS/pIpC-WT and Tet2^Δ/Δ HSCs 4 mo after transplantation (cluster 4 in Fig. 3 C), which were not expressed at higher levels than those in WT HSCs 12 mo after transplantation. A gene ontology (GO) analysis indicated that genes in cluster 4 were involved in infectious diseases and TNF signaling pathway (Fig. 3 D and Data S1). A gene set enrichment analysis (GSEA) consistently showed that in comparisons with WT HSCs, LPS/pIpC-WT, and LPS/pIpC-Tet2^Δ/Δ HSCs at 4 mo had significantly positive enrichments in hallmark inflammatory response pathways (e.g., TNFα-NF-κB; Fig. 3 E), which were dampened at 12 mo in LPS/pIpC-WT HSCs or LPS/pIpC-Tet2^Δ/Δ MDS HSCs. However, we found that alterations in the transcriptome in Tet2^Δ/Δ HSCs at 12 mo were markedly enriched in clusters 1 (LPS/pIpC and Tet2 loss synergistic) and 3 (Tet2 loss dependent; Fig. 3 C). The GO analysis indicated that cluster 1 genes highly activated in LPS/pIpC-Tet2^Δ/Δ were involved in viral carcinogenesis, cell cycles, and innate immunity, while cluster 3 genes activated in both LPS/pIpC-Tet2^Δ/Δ and Tet2^Δ/Δ HSCs were involved in chemokine/adhesion signaling pathways (Fig. 3 D). GSEA revealed that the canonical target genes of MYC and ELF1 TFs were positively enriched in LPS/pIpC-Tet2^Δ/Δ MDS HSCs, and this was accompanied by the up-regulation of proliferative stem cell signatures and dysregulation of myeloid cell signatures at 12 mo, compared to Tet2^Δ/Δ HSCs (Fig. 3 E), which was consistent with progressive phenotypes of MDS in LPS/pIpC-Tet2^Δ/Δ mice.

Elf1 TF attracted our interest because of its role in anti-viral response in innate immune cells downstream of TLR signaling (Chevrier et al., 2011; Seifert et al., 2019), which may alter the epigenome in HSPCs upon infections (Foster et al., 2007; de Laval et al., 2020). We attempted to determine Elf1-binding genes in HSPCs in response to the LPS/pIpC stimulation. 2 d after a single LPS/pIpC in vivo injection, ChIP sequencing (ChIP-seq) indicated that Elf1 directly bound to the promoters of more genes (Fig. 3 F), including epigenetic modifiers such as Kmt2b/Mll4, Phf23 and Kat7/Hbo1 in lineage⁻c-Kit⁺ blood cells (459 versus 48 genes in naive cells; Fig. 3 G and Data S2). By utilizing this LPS/pIpC-induced Elf1-binding genes set, we found that LPS/pIpC-Tet2^Δ/Δ MDS HSCs up-regulated expression of more these Elf1-binding genes than other genotypes (Fig. 3 H), which was also confirmed by GSEA (Fig. 3 I). We attempted to clarify whether human and murine MDS HSPCs enhanced the expression of LPS/pIpC-induced Elf1-binding genes using published datasets. GSEA revealed the positive enrichment of Elf1-binding genes in non-5q-MDS HSPCs in both humans and mice (Fig. 3 J), indicating that the activation of Elf1-binding genes could drive the pathogenesis of MDS. Overall, the stimulation with LPS/pIpC increased the binding of Elf1 to promoter regions, in which most of genes were not activated in WT HSCs at a later period; however, the activation of Elf1-binding genes was augmented by the deletion of Tet2 gene, leading to the development of MDS.

Since the prior stimulation and the deletion of Tet2 augmented changes in the transcriptome of HSCs and facilitated the development of MDS, we investigated the emergence of mutations in mice with the Tet2 deletion and/or LPS/pIpC stimulation. We performed whole-exome sequencing of genomic DNA (gDNA) purified from the c-Kit⁺ BM cells of mice 10–12 mo after transplantation and the tail DNA of donor mice, from which BM cells were transplanted into recipients. Consistent with the accumulation of genomic mutations due to the loss of the Tet2 gene (Pan et al., 2017), Tet2^Δ/Δ cells harbored more mutations, such as single-nucleotide variants, insertions, and deletions, than WT mice and LPS/pIpC-Tet2^Δ/Δ mice (Fig. 4 A), which were frequently located at the coding exon and intron regions (Fig. 4 B and Data S3). The mean variant allele frequency (VAF) of mutated genes was higher in Tet2^Δ/Δ cells than in WT cells (Fig. 4 C). In contrast, LPS/pIpC-WT and LPS/pIpC-Tet2^Δ/Δ mice both showed similar or lower counts of mutations and lower VAF than WT mice (Fig. 4, A and C). An analysis of the impact of mutations that may cause cancer (Lange et al., 2020) revealed that LPS/pIpC-WT and LPS/pIpC-Tet2^Δ/Δ cells did not increase impacting mutations more than those in WT mice, while Tet2^Δ/Δ cells markedly increased nonsynonymous mutations (Fig. 4 D) and high/moderate impact mutations (Fig. 4 E), indicating that the LPS/pIpC stimulation did not accumulate mutations in genes in blood cells after transplantation. Therefore, Tet2^Δ/Δ mice accumulated mutations and developed MDS, whereas the LPS/pIpC stimulation did not increase mutations after the deletion of Tet2. Additional mutations did not appear to contribute to the development of MDS in LPS/pIpC-Tet2^Δ/Δ mice.

Based on fewer mutations in LPS/pIpC-Tet2^Δ/Δ mice, we examined the effects of the prior LPS/pIpC stimulation and the deletion of Tet2 on the epigenome in HSCs to elucidate the mechanisms by which LPS/pIpC-Tet2^Δ/Δ HSCs activated the transcription of Elf1-binding genes and contributed to the development of MDS. Since Tet2 functions as a demethylase for DNA, we initially performed reduced representation bisulfite sequencing (RRBS) on LSK (Lineage⁻Sca-1⁺c-Kit⁺) HSPCs isolated from mice 1 yr after transplantation. RRBS revealed similar numbers of hyper- and hypomethylated regions in both PBS- and LPS/pIpC-treated Tet2^Δ/Δ HSPCs to those in WT HSPCs (Fig. 5 A). When we analyzed DNA methylation levels at stimulation-induced Elf1-binding regions in these HSPCs, methylation levels were lower in LPS/pIpC-WT HSPCs than in WT HSPCs, while Tet2^Δ/Δ and LPS/pIpC-Tet2^Δ/Δ HSPCs both showed similar levels of DNA methylation to that in WT cells 1 yr after transplantation (Fig. 5 B). These methylation levels were higher than those in HSCs from young (3 mo old) WT and Tet2^Δ/Δ mice (Fig. 5 B), indicating aging-dependent increases in DNA methylation (Beerman et al., 2013; Horvath, 2013). LPS/pIpC-Tet2^Δ/Δ HSPCs showed more hypermethylated regions than LPS/pIpC-WT HSPCs (Fig. 5 A); however, we did not detect the general effects of DNA methylation levels at promoters on the differential expression of the corresponding genes between LPS/pIpC-Tet2^Δ/Δ HSCs and LPS/pIpC-WT HSCs (Fig. 5 C), indicating a complex relationship between DNA methylation and transcription levels and the contribution of another mechanism to the activation of Elf1 target genes.

Because chromatin accessibility is a hallmark of active regulatory elements that regulate gene expression programs in cancer (Corces et al., 2018), we examined alterations in chromatin accessibility in WT CD150⁺CD34^-LSK HSCs 2 wk after the LPS (6×)/pIpC (6×) in vivo stimulation by performing assay for transposase-accessible chromatin using sequencing (ATAC-seq). The LPS/pIpC stimulation induced 10-fold larger regions of opened chromatins than closed chromatins in HSCs (15801 versus 1,560 out of 64,042 peaks; Fig. 6 A), indicating the potential of the LPS/pIpC stimulation to introduce chromatin remodeling. ATAC-seq peaks were enriched at promoter and intron regions in these HSCs (Fig. 6 B), harboring distinct histone modifications that were identified at promoter, enhancer, and repressive chromatin in HSCs using published datasets (Fig. S4). Since the accessibility of the TF-binding motif is a strong marker of cellular fate and function, in part, due to its pioneering activity (Zaret and Carroll, 2011), we performed motif enrichment analyses of opened chromatins in HSCs isolated from WT, Tet2^Δ/Δ, and Myd88^−/−;Trif^−/−;Tet2^Δ/Δ (TKO) mice 4 mo after LPS (6×)/pIpC (6×) injections and transplantation (Fig. 6 C). Myd88 and Trif are adaptor proteins downstream of TLRs that activate their signaling (Kawai and Akira, 2006). In contrast to the robust opened chromatin counts in HSCs 2 wk after the LPS/pIpC injection (Fig. 6 A), LPS/pIpC-WT HSCs after 4 mo showed smaller and similar peaks between opened and closed chromatins (286 versus 249 out of 79,069 peaks). LPS/pIpC-Tet2^Δ/Δ HSCs maintained opened chromatin more than closed chromatin (554 versus 180 peaks), which was decreased in LPS/pIpC-TKO HSCs (170 versus 83 peaks; Fig. 6 C). 2 wk after the LPS (6×)/pIpC (6×) stimulation, HSCs showed enrichments in the binding motifs of Ctcf, Ets (class I), and Elf TFs in opened chromatins (Fig. 6 D and Data S4). 4 mo after the stimulation and transplantation, WT and Tet2^Δ/Δ HSCs both showed the highest enrichment of the Elf-binding motif in opened chromatins, which was still significant, but lower in TKO HSCs (Fig. 6 E). LPS/pIpC-Tet2^Δ/Δ HSCs showed the higher enrichment of the Elf-binding motif in opened chromatins than WT and LPS/pIpC-WT HSCs, while Tet2^Δ/Δ HSCs showed the significant, but the lower enrichment of the Elf-binding motif than WT HSCs (Fig. S5), indicating that the LPS/pIpC stimulation and deletion of Tet2 both promoted the enrichment of the Elf-binding motif in opened chromatins in HSCs. To investigate regional changes in chromatin accessibility, a hierarchical clustering map for ATAC peaks revealed that WT HSCs 2 wk after the LPS/pIpC stimulation showed opened regions forming cluster 1 (acute and transient) and cluster 4 (persistent), which were shared with LPS/pIpC-Tet2^Δ/Δ HSCs at 4 mo after the stimulation (Fig. 6 F), showing increased intensity of ATAC peaks over those of LPS/pIpC-WT and naive Tet2^Δ/Δ HSCs (Fig. 6 G). Thus, the LPS/pIpC stimulation remodeled chromatin in HSCs, in which open chromatins enriched with the Elf-binding motif were maintained and accumulated by the deletion of Tet2 in a region-dependent manner.

Based on the enrichment of the Elf-binding motif in opened chromatins in LPS/pIpC-stimulated HSCs, we investigated whether Elf TFs regulated other TF genes. We used a previously described method to analyze TF networks and positions within 13 three-node network motifs, which was shown to identify TFs that were critical for cellular identity and function (Neph et al., 2012; Stergachis et al., 2014). By utilizing ATAC peaks in WT HSCs 2 wk after the stimulation to depict the binding motifs of TFs and identify TF-to-TF networks, we found the different utilization of three-node network motifs between control and LPS/pIpC-stimulated HSCs (Fig. 6 H). Among the 13 motifs, LPS/pIpC-stimulated HSCs enriched feed forward/mutual loop circuit motifs (no. 5 and 11). We attempted to identify the TFs positioned in these network motifs enriched in LPS/pIpC-stimulated HSCs (Fig. 6 I). Stem cell signature TFs, such as Runx1, Gata2, and Hoxa9, preferentially localized to the driver positions in these networks in HSCs. While Spi1 and Fli1, other ETS family TFs, maintained their driver positions regardless of the stimulation with LPS/pIpC, LPS/pIpC-stimulated HSCs showed the enrichment of Elf TFs at the driver positions, whereas control HSCs did not. In contrast, inflammation responders, such as Tp53, Srf, and Irf5, localized to the passenger positions (Fig. 6 I). These results support the specific role of Elf TFs to activate TF networks in response to the stimulation and regulate other TF genes. Overall, the stimulation with LPS/pIpC opened the chromatin enriched with the Elf-binding motif and activated the Elf-driven TF networks in HSCs.

To elucidate the mechanisms by which the LPS/pIpC stimulation activated the transcriptional function of Elf1, we initially assessed Elf1 and Elf4 protein expression levels in lineage⁻c-Kit⁺ HSPCs in mice 2 d after the injection of 10 μg LPS and 125 μg pIpC. We found a marked increase in Elf1 expression at the protein, but not mRNA level following an injection of both LPS and pIpC (Fig. 7, A and B), but a mild or moderate increase in Elf1 protein levels after a single treatment with LPS or pIpC (Fig. 7 C). Elf4 protein levels were not changed after the LPS/pIpC injection (Fig. 7 A). Under in vitro conditions, the treatment with LPS and pIpC increased Elf1 and Elf4 protein levels in WT HSPCs after 24 h, and the induction of Elf1 was completely inhibited in Myd88^−/−;Trif^−/− HSPCs and Trif^−/− HSPCs (Fig. 7 D). Therefore, HSPCs directly sensed both LPS and pIpC and the stimulation-induced elevation in Elf1 protein levels appeared to primarily depend on Trif. Based on the increased expression of Elf1 by LPS/pIpC, we attempted to clarify whether normal aging increases Elf1 protein expression levels in HSPCs due to environmental stress under a specific pathogen–free, but not germ-free, condition. As expected, Elf1 protein expression levels in lineage⁻c-Kit⁺ HSPCs were higher in 2-yr-old mice than in young mice (2 mo old; Fig. 7 E). Notably, naive Tet2^Δ/Δ HSPCs showed low Elf1 protein expression levels, similar to those in naive WT HSCPs in young mice (Fig. 7 F), indicating that the deletion of the Tet2 gene was not sufficient to increase Elf1 protein expression. Therefore, the stimulation with LPS/pIpC and aging-related environmental stress up-regulated the expression of the Elf1 protein in HSPCs in a Trif-dependent manner.

TANK-binding kinase 1 (Tbk1) and Plk are essential components at downstream of the Trif signaling and activate transcription of antiviral genes (Chevrier et al., 2011; Liu et al., 2015). To understand how the LPS/pIpC stimulation increased the Elf1 protein via the Myd88/Trif signaling, we screened an inhibitory effect on expression level of Elf1 protein by chemical inhibitors such as Plk inhibitors (BI6727 and BI2536), a Tbk1 inhibitor (BX795) and a Tbk1/Jak inhibitor (momelotinib) in an in vitro condition. We found that LPS/pIpC-induced increases in Elf1 protein levels were markedly attenuated by Plk inhibitors, but not by Tbk1 inhibitors, in a dose-dependent manner (Fig. 8 A), indicating that Plks were critical for the LPS/pIpC-induced up-regulation of the Elf1 protein.

To investigate the role of Plks in the increased chromatin accessibility, ATAC-seq of HSCs was performed under the same experimental setting. While the in vitro treatment of LPS/pIpC induced more opened chromatin regions than closed chromatin regions in HSCs (5,997 versus 572 out of 91,192 peaks), consistent with the data of the in vivo stimulation in Fig. 6, a 1-h pre-treatment with 1 μM BI6727 led to a balanced change between opened and closed chromatin induced by the 24-h incubation of LPS/pIpC in HSCs (3,102 versus 2,056; Fig. 8 B) and also decreased the enrichment of the Elf-binding motif (Fig. 8 C and Data S5). ATAC peak levels within Elf1-binding regions were increased in LPS/pIpC-treated HSCs, but reduced by the pre-treatment with BI6727 (Fig. 8 D), indicating that HSCs directly sensed LPS and pIpC to enhance chromatin accessibility, which was widely suppressed by the inhibition of Plks. Overall, the LPS/pIpC stimulation required the Plk function to increase the expression of Elf1 protein and simultaneously enhance chromatin accessibility at Elf1-binding regions in HSCs.

Based on the increased expression of Elf1 and enhanced chromatin accessibility via Trif-Plk signaling (Fig. 6 A and Fig. 8 B), we attempted to clarify whether the Elf1 gene was required for the LPS/pIpC stimulation-induced remodeling of chromatin in HSCs. We knocked down Elf1 expression in HSCs in an in vitro culture and performed ATAC-seq in WT HSCs 24 h after the treatment with LPS and pIpC (Fig. 9 A). We confirmed that Elf1 expression levels were 70% lower in shRNA-directed Elf1-transduced cells than in control shRNA-directed luciferase cells (Fig. 9 B). The stimulation with LPS/pIpC increased chromatin accessibility over that in control PBS-treated cells (6,186 open versus 1,168 closed peaks; Fig. 9 C) and showed the enrichment of the Elf-binding motif in open chromatin (Fig. 9 D). We found that the knockdown of Elf1 markedly reduced chromatin accessibility by the stimulation with LPS/pIpC (2,363 open versus 3,108 closed peaks; Fig. 9 C). To assess a regional change in chromatin accessibility, a hierarchical clustering map revealed that LPS/pIpC stimulation-induced open regions were clearly abolished by the knockdown of Elf1 expression (Fig. 9 E). Therefore, the Elf1 gene was required for the stimulation with LPS/pIpC to remodel chromatin in HSCs, suggesting that the activation of Elf1 function is important for the emergence of MDS-initiating clones.

Because the expression levels of Elf1-target genes were enhanced in various genotypes of MDS (Fig. 3, I and J), we finally attempted to clarify whether the Elf1 gene was required for the aberrant hematopoiesis in MDS at a later period after the stimulation. Since Elf1 is expressed in HSPCs, but decreased in erythroid cells, in which the ectopic expression of Elf1 impaired normal erythropoiesis (Calero-Nieto et al., 2010), we correlated the opened Elf-binding motif peaks in LPS/pIpC-Tet2^Δ/Δ HSCs relative to WT HSCs with gene expression profiles in hematopoietic subsets using a published dataset (Seita et al., 2012). The analysis revealed that 35% of genes were highly expressed in HSPCs, while 65% were highly expressed in mature differentiated cells, including 5.7% in erythroid cells (Fig. 9 F), indicating potential roles for Elf1 in LPS/pIpC-Tet2^Δ/Δ HSCs to modulate their differentiation. To investigate the requirement of the Elf1 gene for impaired differentiation, we knocked down Elf1 expression in WT, LPS/pIpC-WT, Tet2^Δ/Δ, and LPS/pIpC-Tet2^Δ/Δ HSCs isolated from mice with these genotypes 4 mo after transplantation using shRNA-directed Elf1 vectors and assessed the erythroid differentiation of Elf1-knockdown cells under in vitro and in vivo conditions (Fig. 9, G and H). WT cells showed a mild elevation in Ter119⁺ erythroid cell counts after the knockdown of Elf1 under both conditions. We found that LPS/pIpC-Tet2^Δ/Δ HSCs cells increased Ter119⁺ erythroid cell counts by the knockdown of Elf1 under these culture conditions (Fig. 9 G). Furthermore, the knockdown of Elf1 expression markedly enhanced erythropoiesis in LPS/pIpC-Tet2^Δ/Δ mice 4 mo after transplantation (Fig. 9 H). Therefore, the Elf1 gene impeded both normal and malignant erythropoiesis, while the impaired erythropoiesis of LPS/pIpC-Tet2^Δ/Δ HSCs was highly dependent on the Elf1 gene.

To assess the proliferative capacity of LPS/pIpC-Tet2^Δ/Δ cells showing stem cell proliferative signatures in the transcriptome (Fig. 3 E), we performed an in vitro colony formation assay. Although we did not examine the colony formation units of LPS/pIpC-WT HSPCs due to their paucity and impaired in vitro proliferation, we found that the knockdown of Elf1 reduced myeloid colony formation units significantly more in LPS/pIpC-Tet2^Δ/Δ HSPCs than in WT and Tet2^Δ/Δ HSPCs (Fig. 9 I). Overall, the expression of Elf1 was required for the enhanced clonogenicity and impaired erythropoiesis of LPS/pIpC-Tet2^Δ/Δ HSPCs, indicating that the Elf1 gene is critical for aberrant hematopoiesis in MDS.

The present results indicated that a prior stimulation with LPS/pIpC, which partly mimicked normal aging with a low-grade inflammatory condition accumulating the Elf1 protein, drove the development of MDS, in which HSCs gained a driver mutation of CH. This was supported by the clinical finding showing that a previous history of multiple infections more than 3 yr prior to the onset of diseases correlated with an increased risk of MDS in humans (Kristinsson et al., 2011). Furthermore, a recent study indicated that atherosclerosis promoted the proliferation of HSCs, resulting in somatic mutations and CH (Heyde et al., 2021), which are considered to cause arteriosclerosis and fatal events in the heart and brain in humans, forming a positive feedback loop to drive transformation (Pietras, 2017; Jaiswal and Ebert, 2019). These inflammation stresses reduced normal stem cell clones due to their differentiation and favored the expansion of CH such as Tet2-deficient stem cell suppressing aberrant activation of inflammatory pathway genes. Therefore, CH does not necessarily precede infection and inflammatory conditions, which may, in turn, play an initiating role in the development of MDS.

HSCs have been proposed to conserve the epigenetic memory of a previous infection stress to regulate the differentiation (Khan et al., 2020; de Laval et al., 2020). A single injection of LPS induced opened chromatin in a Cebpb TF-dependent manner in WT HSCs, which efficiently responded to a secondary injection by enhancing myeloid differentiation and resistance to the infection (de Laval et al., 2020). Under our experimental settings, the prior stimulation with LPS and pIpC did not change the proportion of phenotypic HSCs in BM after 2 wk from that in naive cells, but significantly compromised hematopoiesis in the long term, which is consistent with previous findings on the effects of inflammatory stress on HSCs using various infection stimuli (Essers et al., 2009; Baldridge et al., 2010; Takizawa et al., 2017). In contrast, the subsequent deletion of Tet2 in HSCs maintained LPS/pIpC stimulation-enhanced chromatin accessibility forming an Elf-driven TF network architecture and led to an increase in expression of Elf1-target genes. Increased chromatin accessibility enriched at Elf1-binding regions was inhibited by the deletion of the Myd88/Trif genes or the inhibition of the kinase activity of Plks downstream of the Tlr-Trif pathway. In addition, the Elf1 gene was required for stimulation-enhanced chromatin accessibility, forming the Tlr3/4-Trif-Plks-Elf1 axis in HSCs. Elf family TFs are critical for development, differentiation, stem cell regulation, and anti-viral responses. Among the members of this family, Elf1 and Elf4 are expressed in HSPCs and play important roles in normal hematopoiesis and the anti-viral function of innate immune cells (Choi et al., 2011; Seifert et al., 2019; You et al., 2013). Elf1 was previously shown to reprogram innate immune cells downstream of Tlr-Myd88/Trif and induce the production of anti-viral cytokines, including IFN, in the second challenge with a virus infection (Seifert et al., 2019). Notably, we observed that the Elf1-target gene signature was also positively enriched in non-5q-MDS HSPCs in both humans and mice. These findings also support the epigenetic reprogramming and the transcriptional alteration in MDS HSCs being mediated by the activation of Tlr3/4-Trif-Plk signaling and the overexpression of Elf1. Among the target genes of Elf1, Elf1 appeared to directly activate the expression of epigenetic modifiers, such as Mll4/Kmt2b and Hbo1/Kat7. HBO1 is a histone acetyltransferase of the MYST family, the overexpression of which by the NUP98-HBO1 fusion was identified as a promoter of chronic myelomonocytic leukemia in a mouse model showing aberrant histone acetylation at the Hoxa9 gene (Hayashi et al., 2019). Recently, Hbo1 was shown to promote expression of stem-cell signature genes, including Hoxa9, to maintain normal HSCs (Yang et al., 2022). Therefore, the up-regulated expression of these modifier genes may contribute to the remodeling of the epigenome in MDS HSCs and persistently activate the expression of Elf1-target genes, which warrants further study.

The in vitro LPS/pIpC stimulation slightly up-regulated the expression of Elf4, which has been shown to mediate anti-viral responses in vitro and in vivo downstream of STING-MAVS-TBK1 signaling (You et al., 2013); however, the chemical inhibition of Tbk1 did not prevent LPS/pIpC-induced increases in Elf4 protein nor Elf1 protein levels in HSPCs, at least, under our in vitro condition. Elf4 promotes the cell cycle in HSCs, at least in part, due to the direct activation of Mdm2 in order to suppress the expression of Tp53 (Liu et al., 2009; Sashida et al., 2009), suggesting a tumor-promoting role in MDS. Further studies are needed to clarify the roles of Elf4 in driving the proliferation of MDS stem cells and the altered function of their progenies, including innate immune cells.

We demonstrated that prior infection stresses and the deletion of the Tet2 gene remodeled chromatin accessibility in HSCs via the Tlr-Tirf-Plks-Elf1 axis, leading to the development of MDS, while the loss of Tet2 alone was not sufficient to increase Elf1 protein expression or enrich the Elf-binding motif in open chromatin in HSCs to a similar degree to that in LPS/pIpC-Tet2^Δ/Δ HSCs, indicating critical cooperation between two distinct insults to remodel the epigenome and drive the pathogenesis of MDS. The loss of Tet2 induced small changes in differential DNA methylation and led to both the gain and loss of DNA methylation in a region-specific manner (Zhang et al., 2016). Tet2-deficient HSC promoted DNA methylation of accessible erythroid TF-binding sites, which was implicated in impeding erythropoiesis (Izzo et al., 2020); however, no significant changes were observed in DNA methylation levels in stimulation-induced Elf1-binding promoter regions between Tet2^Δ/Δ HSPCs and WT HSPCs. In addition, DNA hyper-methylation did not prevent the binding of Elf1 to chromatins in HSPCs and monocytes in vivo (Hogart et al., 2012; Tulstrup et al., 2021), indicating that Tet2 loss–induced alterations in DNA methylation may not account for the impaired erythropoiesis that was mediated by the Elf1 TF. A recent study revealed that Tet2-deficient tumor-infiltrating CD8⁺ T lymphocytes showed the significant enrichment of the Elf-binding motif in opened chromatin and enhanced anti-tumor activity partly due to the secretion of IFN-γ in a melanoma model (Lee et al., 2021), supporting a regional cooperation between the binding of Elf1 and the loss of Tet2 in HSCs to open the chromatin following the infection stresses in the present study. Further studies are needed to clarify how interactions between the chromatin remodeling and differential DNA methylation expanded MDS-initiating clones.

The limitation of this work is the transplantation of the treated BM blood cells, which completely erase the effect of the microenvironment in the BM by the LPS/pIpC stimulation on the development of MDS. As we found that aged mice up-regulated the expression of the Elf1 protein in HSPCs, the aging-induced effect of the BM may contribute to the constitutive elevation of the Elf1 protein in HSPCs and drive transformation of CH clones. However, in this study, we focused on the cell intrinsic effects of the stimulation with LPS/pIpC on HSPCs to drive the development of MDS. By using transplantation-independent mouse models, further studies are needed to obtain molecular insights into the effect of the BM including the microenvironment by aging and infection stresses to facilitate the development of MDS, which will be more physiologically corresponding to what happen in patient.

In summary, we demonstrated that prior infection stress followed by the loss of Tet2 remodeled the epigenome and transcriptome in HSCs via the Tlr3/4-Myd88/Trif-Plk-Elf1 axis, which drove the development of MDS. Prior infection stress influenced the phenotypes, transcriptomes, and epigenomes of stem cells, which promoted the development of MDS. The present study also proposes the Trif-Plk-Elf1 axis as a potential target for therapeutic interventions for MDS. Further studies on the impact of the in vivo inhibition of this axis on MDS are needed.

All mice were of the C57BL/6 background. Tet2 conditional knockout (Tet2^flox/flox) mice were crossed with Rosa26:Cre-ERT2 mice (TaconicArtemis GmbH) to generate conditional deletion mice (Moran-Crusio et al., 2011). 2 mg of tamoxifen (T5648; Sigma-Aldrich) was administered via an intraperitoneal injection for 5 consecutive days twice to completely delete the Tet2 gene (Yokomizo-Nakano et al., 2020). C57BL/6 mice congenic for the Ly5 locus (CD45.1) were purchased from Sankyo-Lab Service (Tokyo). Tlr3^−/− knockout mice and Myd88^−/−;Trif^−/− double knockout mice were purchased from Oriental Bioscience (Kyoto). All experiments using these mice were performed in accordance with our institutional guidelines for the use of laboratory animals and approved by the Review Board for Animal Experiments of Kumamoto University (Kumamoto, Japan). All mouse experiments were performed without randomization and blinding.

Tet2^wt/wt;Cre-ERT2 and Tet2^flox/flox;Cre-ERT2 mice were intraperitoneally injected with 10 μg LPS (InvivoGen) and 125 μg pIpC (InvivoGen) or PBS. 2 wk after the last injection of LPS and pIpC, BM cells were harvested and transplanted into lethally irradiated 2–3-mo-old WT male mice (CD45.1⁺).

shRNA directed against Elf1 was expressed using the lentivirus vector, pCS-H1-shRNA-EF-1-EGFP. Target sequences were designated as follows: shElf1 #1; 5′-GAA​GAG​CCT​AAT​GAC​ATG​AT-3′, shElf1 #2; 5′-GTG​TTT​GAA​TTT​GCA​AGT​AAC-3′, shElf1 #3; 5′-GCC​GAA​GAT​GAT​TTG​AAT​GAA-3′. A virus supernatant (VSV-G pseudo-type lentiviral supernatant) was prepared by transfecting 293T cells with an empty control or the shElf1 lentivirus vector plasmid using the calcium phosphate transfection method. The virus supernatant was concentrated by centrifugation at 6,000 ×g for 16 h. The final titers of lentiviral supernatants were 30,000–50,000 infectious U/μl, as assessed by transducing serially diluted viral supernatants into the human Jurkat cell line (Yokomizo-Nakano et al., 2020). In total, 350 purified HSCs were cultured in serum-free SF-03 medium (Iwai North America) supplemented with 1% FBS, 1% L-glutamine, penicillin, streptomycin solution (GPS; Sigma-Aldrich), 50 μM 2-mercaptoethanol (2-ME), 100 ng/ml mouse stem cell factor (SCF; PeproTech), and 100 ng/ml human thrombopoietin (TPO; PeproTech). After 24 h, cells were transduced with the indicated virus at a multiplicity of infection of 2,000 in the presence of 1 μg/ml RetroNectin (Takara) and 10 μg/ml protamine sulfate, and then transplanted into lethally irradiated WT recipient mice (CD45.1⁺).

In total, 30,000 c-Kit⁺ BM cells were incubated in IMDM (Sigma-Aldrich) supplemented with 10% FBS, 1% GPS, 100 μM 2-ME, 50 ng/ml mouse SCF, 50 ng/ml human TPO, 50 ng/ml mouse IL-6 (PeproTech), and 10 ng/ml mouse IL-3. After 24 h, cells were transduced with the indicated virus at a multiplicity of infection of 20 in the presence of 1 μg/ml RetroNectin and 10 μg/ml protamine sulfate for 24 h. On the day after infection, half of the medium was replaced with EDM-I (IMDM supplemented with 15% FBS, 1% BSA, 1% GPS, 100 μM 2-ME, and 50 U/ml human erythropoietin). After 24 h, half of the medium was replaced with EDM-II (IMDM supplemented with 20% FBS, 1% GPS, and 100 μM 2-ME) and cultured for a further 5 d. All in vitro experiments were performed without randomization or blinding.

In total, 350 purified HSCs were harvested and incubated in serum-free SF-03 medium supplemented with 1% FBS, 1% GPS, 50 μM 2-ME, 100 ng/ml mouse SCF, and 100 ng/ml human TPO. After 24 h, cells were transduced with the indicated virus at a multiplicity of infection of 2,000 in the presence of 1 μg/ml RetroNectin and 10 μg/ml protamine sulfate for 24 h. On the day after infection, half of the medium was replaced with SF-03 medium supplemented with 0.1% BSA, 1% GPS, 50 μM 2-ME, 50 ng/ml mouse SCF, and 50 ng/ml human TPO. A colony formation assay was performed using Methocult M3234 (Stem Cell Technologies) supplemented with 10 ng/ml mouse SCF, 10 ng/ml human TPO, 10 ng/ml mouse IL-3, and 2 U/ml human erythropoietin with purified Lineage⁻Sca-1⁺c-Kit⁺ cells. All in vitro experiments were conducted without randomization or blinding.

The Jurkat cell line was obtained from RIKEN (Japan) and cultured in RPMI 1640 containing 10% FBS. The 293T cell line was cultured in DMEM containing 10% FBS. Mycoplasma contamination was tested in all cell lines by performing PCR.

Flow cytometry and cell sorting were performed using the following anti-murine antibodies purchased from BioLegend or eBioscience (clone and catalog numbers): CD45.2 (104, 109820), CD45.1 (A20, 110730), Gr-1 (RB6-8C5, 108404), CD11b/Mac1 (M1/70, 101208), Ter119 (116204), IL-7Rα (A7R34, 121104), B220 (RA3-6B2, 103212), CD4 (L3T4, 100526), CD8α (53-6.7, 100714), c-Kit (2B8, 105812), Sca-1 (D7, 108114), CD34 (MEC14.7, 11-0341-85), CD150 (TC15-12F12.2, 115924), CD71 (R27217, 113808), CD135 (A2F10, 135306), TLR4 (MTS510, 117609), TLR3 (11F8, 141905), and FcγRII-III (93, 101308). The lineage mixture solution contained biotin-conjugated anti-Gr-1, B220, CD4, CD8α, Ter119, and IL-7Rα antibodies. All flow cytometric analyses and cell sorting were performed on FACSAriaII or FACSCantoII (BD).

ATAC-seq libraries were prepared using the Omni-ATAC protocol (Corces et al., 2017). Briefly, HSCs were lysed and then incubated with transposase solution containing 1 μl of Tagment DNA Enzyme (Illumina) at 37°C for 30 min with shaking at 1,000 rpm. Transposed DNA was purified using a MinElute PCR Purification kit (QIAGEN) and then amplified using a NEBNext High Fidelity 2× PCR Master Mix (New England Biolabs) with indexed primers. The prepared libraries were sequenced on NextSeq500 (Illumina) with 38 pb paired-end reads. ATAC-seq data were deposited in the DNA Data Bank of Japan (DDBJ) under the accession numbers DRA012111, DRA012112, DRA012113, and DRA015387.

Lineage⁻c-Kit⁺ cells were fixed by 1.0% paraformaldehyde at 37°C for 5 min and then lysed. Lysed cells were sonicated 15 times at a 50% amplitude for 10 s. Samples were incubated with an anti-Elf1 antibody (Santa Cruz Biotechnology, C-4) conjugated by Dynabeads protein A/G at 4°C overnight. Inputs and immunoprecipitates were incubated at 65°C for 4 h for reverse cross-linking, and DNA was purified using the MinElute PCR Purification Kit (Qiagen). ChIP-seq libraries were generated using the ThruPLEX DNA-seq kit (Rubicon Genomics), and then sequenced on NextSeq500 (Illumina). Bowtie2 (version 2.2.6; default parameters) was used to map the reads to the reference genome (UCSC/mm9). Peak calling and motif analyses were performed using HOMER (version 4.9). ChIP-seq data were deposited under the accession number DRA012114.

Total RNA was extracted using ISOGEN (Nippon Gene), and cDNA was synthesized using the SMARTer Pico PCR cDNA Synthesis kit (Clontech). ds-cDNA was fragmented, and cDNA libraries were generated using the KAPA HyperPlus Library Preparation kit (KAPA Biosystems) and FastGene Adaptor kit (Fastgene). RamDA-seq was performed as previously reported (Hayashi et al., 2018). Briefly, first strand cDNA was synthesized from 100-cell lysate using PrimeScript RT reagent kit (TAKARA Bio, Inc.) with first not-so random primers, and then the second strand cDNA was synthesized using Klenow Fragment (3′-5′ exo-; New England Biolabs, Inc.) and second not-so random primers. Sequencing libraries were generated using Nextera XT DNA Library Prep kit (Illumina, Inc.). The cDNA libraries were sequenced on NextSeq500 (Illumina). Kallisto (version 0.43.1) was used for read counts and the calculation of transcripts per million. EdgeR was used for statistical analyses. RNA-seq data were deposited under the accession number DRA012115 and DRA012116. RamDA-seq data were deposited under the accession number DRA013483.

RRBS was performed as described in our previous study (Sashida et al., 2014). After bisulfate conversion using the Imprint DNA Modification Kit (Sigma-Aldrich) followed by the agarose gel excision of fragments from 150 to 400 bp, bisulfite-converted DNA was amplified by PfuTurbo Cx Hotstart polymerase (Agilent) using NEBNext Multiplex Oligos for Illumina (New England Biolabs). The libraries were purified using AMPure XP beads (Beckman Coulter) and were sequenced on NextSeq500 (Illumina). The quantification of methylation levels and statistical analyses were performed by methylKit (v1.16.1). RRBS data were deposited under the accession number DRA015388.

gDNA was extracted from c-Kit⁺ BM cells and tail tissues. The gDNA library was prepared using a SureSelect Mouse All Exon kit (Agilent Technologies) and sequenced via 2 × 150 paired-end sequencing using a NovaSeq 6000 System (Illumina), performed in Macrogen in Korea. Sequence reads were aligned with the Burrows-Wheeler Aligner (v0.7.17) algorithm to the GRCm38.p6 mouse assembly (Li and Durbin, 2009). Single nucleotide variants and indels were identified using MuTect2 (gatk-4.1.0.0; Cibulskis et al., 2013). Whole-exome sequencing data were deposited in the DDBJ under the accession number DRA015410.

Transcriptional network analysis followed previously published methods with minor modification (Neph et al., 2012; Stergachis et al., 2014). TF and motif data for network analysis were got from the JASPAR database through UCSC Genome Browser. For network analysis we picked up the 307 TF which were expressed in HSC (transcripts per million > 1) from 841 TF. A directed edge was drawn from a gene node to another when a first gene’s motif was detected within ATAC-seq peaks within 5 kb of the second gene’s TSS, indicating regulatory potential. Top 5% highly enriched motifs compared with background were picked up as node. A z-score for network analysis was calculated with three-node network motifs using 1,000 randomized networks by mfinder software (Milo et al., 2002).

q-PCR was performed on LightCycler 480 (Roche) using Luna Universal qPCR Master Mix (New England Biolabs). Expression levels were normalized to those of Gapdh or B2m/β2 microglobulin. Primers for PCR are listed in Table S1.

Whole cell lysates were used for Western blotting. Briefly, Lineage⁻c-Kit⁺ BM cells were isolated from mice and lysed in modified RIPA buffer (50 mmol/liter Tris, pH 7.5, 150 mmol/liter NaCl, 1% NP40, 0.25% sodium deoxycholate, and 1 mmol/liter EDTA). In in vitro cultures, c-Kit⁺ BM cells were seeded at 400,000 cells per well and pre-incubated with the indicated concentration of chemical inhibitors for 1 h before exposure to 0.1 μg/ml LPS and 10 μg/ml pIpC for 24 h. The following antibodies were used for Western blotting: Elf1 (Santa Cruz Biotechnology, C-4), Elf4 (Santa Cruz Biotechnology, E-11), and actin (Santa Cruz Biotechnology, C4). Uncropped immunoblot images are available in source data files.

All statistical tests were performed using GraphPad Prism version 9 (GraphPad Software). The significance of differences was measured by an unpaired two-tailed Student’s t test or the Mann–Whitney non-parametric test. A P value <0.05 was considered to be significant. No statistical methods were used to predetermine sample sizes for animal studies.

Fig. S1 provides expression levels of Tlr3 and Tlr4 in HSCs that directly respond to LPS and pIpC in an in vitro condition. Fig. S2 shows BM cell counts and cell numbers of HSC and MPP in the BM 2 wk after the LPS/pIpC injection. Fig. S3 shows CBC and disease subtypes identified in Tet2^Δ/Δ-PBS mice and Tet2^Δ/Δ-LPS/pIpC mice. Fig. S4 provides heatmaps showing the levels of described histone modifications and ATAC peak identified in HSCs in open chromatin regions in LPS/pIpC-stimulated HSCs. Fig. S5 shows rank of TF-motif enrichments of open chromatin in Tet2^Δ/ΔHSCs 4 mo after the stimulation with LPS/pIpC. Table S1 lists primers for q-PCR. Data S1 shows GO enrichment analyses of up-regulated genes in five clusters. Data S2 lists Elf1-target genes 2 d after the LPS/pIpC stimulation. Data S3 lists mutations in c-Kit⁺ BM cells isolated from WT, LPS/pIpC-WT, Tet2^Δ/Δ, and LPS/pIpC-Tet2^Δ/Δ mice 10–12 mo after transplantation. Data S4 shows motif enrichment analyses of ATAC peaks in HSCs 2 wk or 4 mo after the LPS (×6)/pIpC (×6) stimulation. Data S5 shows motif enrichment analyses of ATAC peaks in LPS/pIpC- or BI6727 with LPS/pIpC-treated HSCs. Data S6 shows detailed information and links of sequencing data deposited in the DDBJ.

Sequencing data that support the results of the present study were deposited in the DDBJ under the accession numbers DRA012111, DRA012112, DRA012113, and DRA015387 for ATAC-seq; DRA012114 for ChIP-seq; DRA012115 and DRA012116 for RNA-seq; DRA013483 for RamDA-seqencing; DRA015388 for RRBS; and DRA015410 for whole-exome sequencing. Detailed information and links for these data deposited in the DDBJ are shown in Data S6.

The authors thank Ms. Otowa Takahashi, Dr. Takako Ideue, Ms. Ryoko Koitabashi, Mr. Akifumi Kiyota, Mr. Shinji Kudoh, and Dr. Takaaki Ito for their technical help.

This work was supported in part by a grant from the Takeda Science Foundation (to G. Sashida), the Japanese Society of Hematology (to G. Sashida), The Chemo-Sero-Therapeutic Research Institute (to H. Takizawa), International Joint Usage/Research Center, the Institute of Medical Science, the University of Tokyo (to G. Sashida), Grants-in-Aid for Scientific Research (16KT0113, 18H02842, 19K22640, 21H02952 [to G. Sashida], 19K08842 [to T. Yokomizo-Nakano], and 21K06870 [to G. Huang]) from the Ministry of Education, Culture, Sports, Science and Technology of Japan, and Core-to-Core Program Advanced Research Networks “Integrative approach for normal and leukemic stem cells” from the Japan Society for the Promotion of Science of Japan.

Author contributions: T. Yokomizo-Nakano performed experiments and analyzed data jointly with A. Hamashima, S. Kubota, J. Bai, S. Sorin, Y. Sun, and M. Morii; A. Hamashima and M. Iimori assisted with animal experiments; S. Kubota, K. Kikuchi, D. Kurotaki, and H. Matsui assisted with RRBS analysis; H. Matsui and A. Kanai assisted with whole-exome sequencing analysis; A. Iwama, G. Huang and H. Takizawa provided advices, reagents, and grants; T. Yokomizo-Nakano and G. Sashida wrote the paper; G. Sashida conceived and designed the study, and supervised the project.