MBTD1 preserves adult hematopoietic stem cell pool size and function

To investigate the role of MBTD1 in adult hematopoiesis, we first examined Mbtd1 expression patterns at various stages of hematopoietic differentiation in adult BM. RNAs extracted from cells in long-term HSC (LT-HSC), short-term HSC (ST-HSC), MPP, and lineage (Lin)⁻, as well as Lin⁺ fractions were subjected to quantitative real-time PCR (qRT-PCR) (surface markers of hematopoietic HSPC fractions are described in ref. 19 and summarized in SI Appendix, Table S1). Mbtd1 was preferentially expressed in the LT-HSC, ST-HSC, MPP, and Lin⁻ fractions but not in the Lin⁺ fraction (SI Appendix, Fig. S1A). Similarly, data from a public database showed that Mbtd1 is highly expressed in human and mouse BM HSCs compared with progenitor cells (20) (SI Appendix, Fig. S1B). These results suggest that MBTD1 may function in HSPCs in the adult BM as it does in the FL (9).

To clarify the role of MBTD1 in adult hematopoiesis, we generated Mbtd1^flox/flox mice (SI Appendix, Fig. S2 A and B) and crossed them with interferon (IFN)-inducible Mx1-Cre⁺ (Mx1⁺) transgenic mice (21) to create Mbtd1^flox/flox/Mx1⁺ mice (Upper two panels of SI Appendix, Fig. S3A and Left two lanes of SI Appendix, Fig. S3B). Administration of pIpC, a strong and transient IFN inducer, efficiently deleted the floxed region (SI Appendix, Fig. S2C) and almost completely abolished MBTD1 protein in the hematopoietic tissues of Mbtd1^flox/flox/Mx1⁺ mice (SI Appendix, Fig. S2D) [hereafter, pIpC-treated Mbtd1^flox/flox/Mx1⁻ and Mbtd1^flox/flox/Mx1⁺ mice are referred to as control (Ctrl) and Mbtd1 cKO (cKO) mice, respectively].

The hematopoietic parameters of Ctrl and cKO mice were analyzed. At steady state, there was no significant change in the number of mononuclear cells in the thymus, spleen, and BM between Ctrl and cKO mice (SI Appendix, Fig. S4A). In addition, white blood cell (WBC) count, hemoglobin (Hb) concentration, and platelet (Plt) number in peripheral blood (PB) were comparable between the two groups (SI Appendix, Fig. S4B). Next, we investigated cell numbers of various fractions in the hematopoietic hierarchy. The cell numbers in the LSK fraction and those in the LT-HSC and ST-HSC fractions were significantly increased in the cKO BM compared with those in the Ctrl BM (Fig. 1A). Furthermore, accumulation of common myeloid progenitor (CMP) and granulocyte/macrophage progenitor (GMP) cells was observed in the cKO BM (Fig. 1B). We also performed CFU-C (colony-forming units in culture) and HPP-CFC (high proliferative potential colony-forming cell) assays, which assess the proliferative capacity of HSPCs, and found no apparent differences in both assays (SI Appendix, Fig. S4C). To analyze the early effect of MBTD1 deficiency on hematopoiesis, we examined HSPCs and differentiated cells of Mbtd1^flox/flox/Mx1⁻ and Mbtd1^flox/flox/Mx1⁺ mice at one week after pIpC administration. As shown in SI Appendix, Fig. S5, no changes were detected in HSPC and differentiated cell fractions, except a slight decrease in CLPs and a slight increase in MEPs. Thus, the increase in HSPCs, including LT-HSCs, ST-HSCs, CMPs, and GMPs, would be a late event which may compensate for a possible defect of cKO HSPCs. These results together indicate that although Mbtd1 deficiency does not affect parameters in the hematopoietic tissues and PB, it alters the pool size of HSPCs at steady state.

One of the primary functions of HSC is to regenerate hematopoiesis following stress. We assessed the tolerance of cKO HSCs after hematological stresses by performing 5-fluorouracil (5-FU) treatment (22) and BM transplant (BMT) assays (23). We first analyzed the response to the 5-FU treatment, which induces apoptosis of cycling hematopoietic cells and switches HSCs from a dormant to a proliferative state. The results showed that the cKO PB exhibited a significant increase in WBC count and platelet number during the recovery phase (16 d after 5-FU, Left panel of Fig. 2A and SI Appendix, Fig. S6A). The proliferated WBCs were composed mostly of immature and mature granulocytes and partly of lymphocytes (Right panel of Fig. 2A and SI Appendix, Fig. S6A). We also analyzed HSPC fractions of 5-FU-treated mice after recovery (33 d after 5-FU). As shown in SI Appendix, Fig. S6B, the analysis revealed that the cell numbers of LSK, ST-HSC, and GMP fractions were significantly increased in cKO mice compared with Ctrl mice, indicating that the expansion of HSPCs in 5-FU-treated cKO mice still remained when PB parameters returned to normal.

Next, we performed a BMT assay, another stress-loading model for HSCs. Ctrl and cKO LT-HSCs (Ly5.2⁺) were transplanted into lethally irradiated Ly5.1⁺ congenic recipient mice with Ly5.1⁺ competitor cells. The analysis of the donor-derived (Ly5.2⁺) chimerism in the PB showed that recipients transplanted with cKO LT-HSCs exhibited a marked decrease in donor-derived cell percentages compared with those observed in mice transplanted with Ctrl LT-HSCs (Left panel of Fig. 2B). Differentiation status in the PB was almost similar between the two groups, except a decrease in T cells (Right panel of Fig. 2B), indicating an overall repopulating defect of cKO HSCs. The in vivo homing assays showed that the percentages of BM-homed cells in total, LSK, and Lin⁻ fractions per transplanted cells in the same fractions were comparable between recipients transplanted with Ctrl and cKO BM cells (SI Appendix, Fig. S7), indicating that the reduced repopulating activity of cKO HSCs was not due to impaired homing ability but rather to other cell-intrinsic defect(s). These results together demonstrate that in adult mice, cKO HSCs exhibit abnormal behaviors under stressful conditions, such as hyperresponsiveness during the recovery process from 5-FU-induced BM suppression and a markedly reduced repopulation ability following BMT.

To clarify the molecular mechanism(s) underlying altered functional properties of cKO HSCs, we searched for genes whose expression levels were affected by Mbtd1 deficiency. RNAs extracted from LT-HSC, ST-HSC, and MPP fractions of Ctrl and cKO BMs were subjected to qRT-PCR array analysis (Fluidigm) (24) (genes used for Fluidigm are summarized in SI Appendix, Table S2). Since the phenotypes observed in cKO mice were thought to be induced by abnormalities at the LT-HSC level, we focused on the LT-HSC fraction. We found that the expression levels of three genes, FoxO3a, Egr4, and Tcf3, were significantly reduced in cKO LT-HSC fraction compared with Ctrl LT-HSC fraction (SI Appendix, Fig. S8A), and qRT-PCR confirmed significant downregulation of these genes in cKO LT-HSCs compared with Ctrl LT-HSCs (Left panel in Fig. 3A and SI Appendix, Fig. S8B). We focused on FoxO3a, since previous studies, including ours, have demonstrated that FoxO transcriptional factors play a pivotal role in HSC function through the maintenance of cell cycle quiescence (25, 26), and the behaviors of cKO mice under stressful conditions, such as 5-FU stimulation and BMT (Fig. 2 A and B), closely resembled those reported in FoxO3a KO mice (22, 26).

To confirm the contribution of FOXO3a to the phenotypes of cKO mice, we examined the expression of FOXO3a protein in HSCs of Ctrl and cKO mice. As shown in the Middle panel of Fig. 3A, intracellular FACS demonstrated that the expression of FOXO3a protein was significantly reduced in cKO LT-HSCs compared with Ctrl LT-HSCs. Previous studies showed that nuclear localization of FOXO3a is essential for HSC quiescence (26, 27). Thus, we examined subcellular localization of FOXO3a by intracellular staining and measured the percentages of cytoplasmic and nuclear staining patterns in LT-HSC and ST-HSC fractions. As shown in the Right panel of Fig. 3A, the percentage of LT-HSCs localizing FOXO3a to the nucleus was reduced in cKO LT-HSCs compared with Ctrl LT-HSCs, whereas the percentage of the two staining patterns were almost comparable in the ST-HSC fraction. These results collectively indicate that the phenotypes of cKO mice were primarily due to reduction of FOXO3a expression in LT-HSCs.

FOXO3a maintains HSC quiescence by up-regulating the expression of cell cycle checkpoint genes, p57^Kip2 (p57), p21^Cip1 (p21), and p27^Kip1 (p27) (26), and autophagy-related genes including, Atg4b, Bbc3, Bnip3, Gabarapl2, Prkaa2, and Sesn1 (28). Therefore, we quantified the expression levels of these genes between Ctrl and cKO LT-HSCs. As for cycle checkpoint genes, the expression levels of p57 and p21 were significantly down-regulated, whereas that of p27 was significantly up-regulated in cKO LT-HSCs compared with Ctrl LT-HSCs (Fig. 3B). In contrast, none of the autophagy-related genes was down-regulated in cKO LT-HSCs compared with Ctrl LT-HSCs (SI Appendix, Fig. S9).

To further investigate whether the alteration in these gene expression profiles affected cell cycle status, we performed Pyronin Y staining and 5-Bromo-2′-deoxyuridine (BrdU) incorporation assay, which allow the identification of cells in the G0 resting phase and the proliferative stage, respectively (9). A significant decrease in Pyronin Y-negative cells and a significant increase in BrdU-positive cells were detected in the LT-HSC fraction of cKO mice, which indicates an enhanced cell cycle entry of cKO LT-HSCs (Fig. 3 C and D). Cell cycle analysis using Ki67 and Hoechst 33342 also showed that cKO HSCs were more active in cell cycle at steady state and after 5-FU administration (left two bars in each phase of SI Appendix, Fig. S10 A and B). Next, we examined the relationship between the increased stem/progenitor cell pool and the cell cycle in cKO HSCs. In the early phase after the pIpC administration (1 wk later) when the number of HSCs or progenitor cells did not increase (SI Appendix, Fig. S5), the enhanced cell cycling of cKO HSCs was already evident (left two bars in each phase of SI Appendix, Fig. S10C), suggesting that the first event was a change in cell cycle state of HSCs. FOXO3a deficiency enhances ERK and AKT signaling in stressed HSCs (22). Phosphorylation of ERK in HSCs was higher at steady state by Mbtd1 deficiency, and there was no difference in the increase after 5-FU administration (Left panel of SI Appendix, Fig. S11). In contrast, phosphorylation of AKT was not different at steady state, but was significantly increased after 5-FU administration by Mbtd1 deficiency (Right panel of SI Appendix, Fig. S11).

To investigate whether MBTD1 directly regulates FoxO3a transcription in HSCs, the binding ability of MBTD1 to the FoxO3a gene was assessed by a chromatin immunoprecipitation (ChIP) assay using LSK cells. We designed PCR primers that encompass three regions in the promoter (P1-P3) and a region in the 1st intron (P4) of the FoxO3a gene (Left panel of Fig. 3E). Protein-DNA complexes immunoprecipitated by control IgG or the MBTD1 antibody (9) were amplified with the primers corresponding to the P1-P4 regions. Three regions (P1, P2, and P3), located upstream of the putative transcription start site (TSS), were specifically immunoprecipitated by the MBTD1 antibody, whereas region P4, located downstream of the TSS, was not (Right panel of Fig. 3E), indicating that MBTD1 was directly recruited to the promoter region of the FoxO3a gene. These results demonstrate that MBTD1 contributes to HSC quiescence, at least partly, by regulating the expression of FOXO3a and downstream Cip/Kip family CDK inhibitors.

Given the evidence of FOXO3a as a direct downstream effector of MBTD1, we designed a line of experiments for the functional assessment of the MBTD1-FOXO3a axis in hematopoiesis. To this end, we used FoxO3a^+/TM mutant mice, in which an floxed Neo resistance cassette coupled with a constitutive-active mutant of FoxO3a cDNA (FoxO3a^TM) were knocked into the FoxO3a locus (26) (SI Appendix, Fig. S12A). The knockin allele functions as a FoxO3a-null allele at steady state; however, after Cre expression, the floxed Neo resistance cassette is excised and FoxO3a^TM mRNA is expressed under the regulation of endogenous FoxO3a (SI Appendix, Fig. S12B). To analyze the effect of FOXO3a^TM on normal hematopoiesis, we compared cell numbers in HSPC and differentiated cell fractions and also analyzed cell cycle of HSCs between Mbtd1^+/+/Mx1⁺/FoxO3a^+/+ (FoxO3a^+/+) and Mbtd1^+/+/Mx1⁺/FoxO3a^+/TM (FoxO3a^+/TM) mice at 4 wk after pIpC treatment. As shown in SI Appendix, Fig. S12C, no changes were observed in all fractions. In addition, as shown in SI Appendix, Fig. S12D, the frequency of LT-HSCs in the G0, G1, and S/G2/M phases was comparable between FoxO3a^+/+ and FoxO3a^+/TM groups. These results indicate that expression of FOXO3a^TM alone does not affect hematopoiesis, including proliferation and differentiation of HSCs.

We then crossed Mbtd1^flox/flox/Mx1⁺ mice with FoxO3a^+/TM mutant mice to generate Mbtd1^flox/flox/Mx1⁺/FoxO3a^+/TM mice (SI Appendix, Fig. S3A and the 3rd lane of SI Appendix, Fig. S3B). After pIpC administration, Mbtd1^flox/flox/Mx1⁺/FoxO3a^+/TM mice simultaneously lost Mbtd1 and acquired constitutively active FoxO3a, FoxO3a^TM [hereafter, pIpC-treated Mbtd1^flox/flox/Mx1⁺/FoxO3a^+/TM mice are referred to as FoxO3a^TM-rescued (Rescue) mice]. In this model, we examined the cell numbers in HSPC fractions. We found that the increased cell numbers in the LT-HSC, ST-HSC, CMP, and GMP fractions of cKO mice were successfully restored in Rescue mice expressing FoxO3a^TM (Fig. 4A). To investigate self-renewal and multilineage differentiation capacity, Ctrl, cKO, and Rescue LT-HSCs were transplanted into lethally irradiated recipient mice. In the 1st transplant, PB chimerism of Rescue LT-HSCs was significantly higher than that of cKO LT-HSCs (Left panel of Fig. 4B). However, in the 2nd transplant, PB chimerism of Rescue cells declined and was almost comparable with that of cKO cells, which was associated with lower chimerism in BM cell fractions of the cKO and Rescue groups than the Ctrl group at the end of the 2nd BMT (4 mo, Right panel of Fig. 4B). These results collectively indicate that the MBTD1-FOXO3a axis functions as a determinant of the pool size of HSPCs and partially contributes to the long-term repopulating ability of HSCs.

The aforementioned results demonstrated that FOXO3a^TM successfully rescued abnormal HSPC pool size induced by Mbtd1 deficiency. Next, we analyzed the underlying mechanisms associated with the cell cycle quiescence and metabolic homeostasis of LT-HSCs. We first examined the cell cycle kinetics of Ctrl, cKO, and Rescue LT-HSCs using multicolor flow cytometry. Consistent with the Mbtd1 deficiency-dependent increase in the LT-HSC pool (Figs. 1A and 4A), we observed a specific activation of cell cycle, characterized by a reduction in the Pyronin Y-negative G0 fraction in cKO CD34⁻ LSK cells, which was recovered in Rescue CD34⁻ LSK cells to the Ctrl level (Left panel of Fig. 5A, see also Fig. 3C). In contrast, the Pyronin Y-negative G0 fraction in CD34⁺ LSK cells was comparable among the three groups (Right panel of Fig. 5A, see also Fig. 3C). Furthermore, short-term BrdU labeling indicated that the increase in cycling cells in the cKO LT-HSC fraction was returned to the Ctrl level by FoxO3a-rescue (Fig. 5B, see also Fig. 3D). Consistently, the decrease in the G0 phase fraction of cKO LT-HSCs in steady state and at early phase after pIpC administration was restored in the Rescue group (SI Appendix, Fig. S10 A and C). In addition, among the alteration of cell cycle regulators in cKO LT-HSCs (Fig. 3B), the decrease in p57 and p21 expression levels in cKO LT-HSCs was corrected in Rescue LT-HSCs, whereas, the increase in p27 expression was further enhanced (Fig. 5C).

In addition to cell cycle quiescence, the metabolic property associated with ATP production is another important feature of HSCs (6). We detected significantly reduced levels of ATP in cKO LT-HSCs compared with those of Ctrl LT-HSCs, which not recovered in Rescue LT-HSCs (Left panel of Fig. 5D). We examined the expression levels of glycolytic genes between Ctrl and cKO LT-HSCs and found that the expression levels of Hexikinase 2 (Hk2), Enolase 2 (Eno2), Enolase 3 (Eno3), and Pyruvate kinase liver and red blood cell (Pklr) were significantly decreased in cKO cells (SI Appendix, Fig. S13). In parallel, intracellular lactate dehydrogenase (LDH) activity, which reflects the cellular glycolytic capacity, was significantly reduced in cKO LT-HSCs compared with that of Ctrl LT-HSCs, which was not recovered in Rescue LT-HSCs (Right panel of Fig. 5D). These data collectively indicate that the MBTD1-FOXO3a axis specifically regulates cell cycle quiescence in LT-HSCs and suggest that MBTD1 maintains the glycolytic activity of HSCs in a FOXO3a-independent manner.

Given our in vivo studies suggesting that FOXO3a restoration only partially ameliorates HSC properties, MBTD1-binding proteins were investigated to elucidate the MBTD1-regulated pathways. A murine HSPC cell line, EML (29), was used to identify proteins that directly bind to MBTD1. Whole-cell extracts of EML cells were immunoprecipitated using the anti-MBTD1 antibody (9), the immunoprecipitates were digested, and the cleaved fragments were directly analyzed by a direct nanoflow liquid chromatography-tandem mass spectrometry system (Fig. 6A). As a result of four repeated assays, the identification of MBTD1 in all experiments and MBTD1-associated TIP60 complex proteins, such as EP400, EPC1, TRPPAP, MRGBP, RUVBL1, and RUVBL2 (17) in more than two independent experiments verified the accuracy of this system (yellow circle in Fig. 6B and SI Appendix, Table S3). Of note, among repeatedly identified proteins, we found molecules that have been reported to be involved in HSC or other stem cell functions, which included proteins belonging to TIP60 complex (EP400, TRPPAP, RUVBL1, and RUVBL2) (30–33), ribosomal proteins (RPL17, RPL18, and RPS19) (34), heat shock protein (HSPA9) (35), Calmodulin (CaM) (36, 37), and an mRNA regulator (DDX6) (38) (blue circle in Fig. 6B and SI Appendix, Table S3). It has been shown that sufficient ribosomal function is crucial for suppressing protein synthesis for HSC maintenance (34). In addition, HSPA9 has been suggested to play an essential role in regulating mitochondrial homeostasis in HSPCs (35). These findings indicate that protein synthesis and mitochondrial reactive oxygen species (ROS) production may be altered by MBTD1 deficiency.

Fine-tuning of protein synthesis rate by ribosomal protein is critical for HSC homeostasis (34). Given that MBTD1 interacted with RPL17, RPL18, and RPS19 (Fig. 6B), protein synthesis rates of isolated LT-HSCs were analyzed using O-propargyl-puromycin (OP-Puro). Because Mbtd1-deficiency changes cell cycle progression in LT-HSCs (Fig. 3 C and D and SI Appendix, Fig. S10), the protein synthesis rate of G0/G1 cells and S/G2/M phases were separately analyzed. OP-Puro uptake in cKO LT-HSCs was significantly increased in both phases compared with that observed in Ctrl LT-HSCs; however, the OP-Puro uptake observed in Rescue LT-HSCs was significantly suppressed and returned to the Ctrl level (Fig. 6C). In contrast, mitochondrial ROS levels in LT-HSCs were comparable among the three groups (SI Appendix, Fig. S14). The physical interaction network of MBTD1-binding proteins (SI Appendix, Table S3) predicted by STRING (https://string-db.org/) (39) and clusters of above-mentioned protein groups are summarized in Fig. 6D.

The detailed procedures for the construction of the targeting vector and generation of Mbtd1^flox/flox mice are described in SI Appendix. Mbtd1^flox/flox mice were crossed with Mx1-Cre⁺ (Mx1⁺) transgenic mice (21) and FoxO3a^+/TM mutant mice (26) to generate Mbtd1^flox/flox/Mx1⁺ and Mbtd1^flox/flox/Mx1⁺/FoxO3a^+/TM mice. pIpC was injected interperitoneally to induce Cre expression as previously described (47). All mice were housed according to the guidelines of the Institute of Laboratory Animal Science, Hiroshima University (permission no. 25-7) and Tokyo Women’s Medical University (permission no. AE21-023).

Mouse survival curves were constructed using the Kaplan–Meier method and compared by the log-rank test using the GraphPad Prism software. Other statistical analyses were performed using Student’s t test unless otherwise stated.

A complete and detailed description of the methods is provided in SI Appendix, Supplemental Materials and Methods.

K.T., P.W.H., T.U., Y.S., M.I., T.S., and H.H. designed research; K.T., P.W.H., T.U., Y.S., M.I., M.K., K.S., H.K., M.H., S.W., Z.-i.H., N.Y., A.N.-I., F.A., T.H., T.N., and H.H. performed research; N.M. contributed new reagents/analytic tools; K.T., P.W.H., T.U., Y.S., M.I., M.K., K.S., H.K., M.H., S.W., Z.-i.H., N.Y., A.N.-I., F.A., N.M., T.H., T.N., T.S., and H.H. analyzed data; and K.T. and H.H. wrote the paper.

The authors declare no competing interest.