九州大学学術情報リポジトリKyushu University Institutional

Hematopoietic stem cells (HSCs) are maintained in a hypoxic niche to limit oxidative stress. Although iron elicits oxidative stress, the importance of iron homeostasis in HSCs has been unknown. Here we show that iron regulation by the F-box protein FBXL5 is required for HSC self-renewal. Conditional deletion of Fbxl5 in mouse HSCs results in cellular iron overload and a reduced cell number. Bone marrow transplantation reveals that FBXL5-deﬁcient HSCs are unable to reconstitute the hematopoietic system of irradiated recipients as a result of stem cell exhaustion. Transcriptomic analysis shows abnormal activation of oxidative stress responses and the cell cycle in FBXL5-deﬁcient mouse HSCs as well as downregulation of FBXL5 expression in HSCs of patients with myelodysplastic syndrome. Suppression of iron regulatory protein 2 (IRP2) accumulation in FBXL5-deﬁcient mouse HSCs restores stem cell function, implicating IRP2 as a potential therapeutic target for human hematopoietic diseases associated with FBXL5 downregulation.

H ematopoietic stem cells (HSCs) are the most undifferentiated cells in the mammalian hematopoietic system, which they maintain throughout life. At steady state, HSCs are quiescent and reside in their hypoxic niche. They expend energy mostly via anaerobic metabolism by maintaining a high rate of glycolysis. These characteristics promote HSC maintenance by limiting the production of reactive oxygen species (ROS) 1 , to which HSCs are highly vulnerable compared with other hematopoietic cells 2 . Homeostasis of cellular iron, which is a major elicitor of ROS production, is thus likely to be strictly regulated in HSCs in order for them to maintain their stemness.
Iron is essential for fundamental metabolic processes in cells and organisms, and it is incorporated into many proteins in the form of cofactors such as heme and iron-sulfur clusters. Iron also readily participates in the Fenton reaction, however, resulting in uncontrolled production of the hydroxyl radical, which is the most harmful of ROS in vivo and damages lipid membranes, proteins and DNA. It is therefore important that cellular iron levels are subject to regulation 3 . We previously showed that iron homeostasis in vivo is regulated predominantly by F-box and leucine-rich repeat protein 5 (FBXL5) and iron regulatory protein 2 (IRP2) 4 . IRP2 functions as an RNA binding protein to regulate the translation and stability of mRNAs that encode proteins required for cellular iron homeostasis. IRP2 thereby increases the size of the available iron pool under iron-limiting conditions. In contrast, under iron-replete conditions, FBXL5, which is the substrate recognition component of the SCF FBXL5 E3 ubiquitin ligase, mediates ubiquitylation and degradation of IRP2. Whereas FBXL5 is unstable under iron-deficient conditions, direct binding of iron to its hemerythrin domain stabilizes the protein, with this iron-sensing ability allowing FBXL5 to control the abundance of IRP2 in an iron-dependent manner 5,6 . Disruption of the Fbxl5 gene in mice results in the failure of cells to sense increased cellular iron availability, which leads to constitutive accumulation of IRP2 and misexpression of its target genes. FBXL5-null mice die during embryogenesis as a result of overwhelming oxidative stress, indicating the vital role of FBXL5 in cellular iron homeostasis during embryogenesis 4 .
A substantial proportion of iron in the adult body is present in the liver and hematopoietic system. Excess iron in the liver is clinically important given that cirrhosis and hepatocellular carcinoma often develop in individuals with systemic ironoverload disorders 7 . Conditional FBXL5 deficiency in mouse liver was found to result in iron accumulation and mitochondrial dysfunction in hepatocytes, leading to the development of steatohepatitis 4 . In contrast, hematopoiesis is sensitive to iron deficiency, with an insufficiency of available iron in the body being readily reflected as iron-deficiency anaemia 8 .
Iron overload in the haematopoietic system is also clinically important, however. Systemic iron overload is thus frequently associated with hematologic diseases such as myelodysplastic syndrome (MDS), a clonal HSC disorder characterized by hematopoietic failure as a result of ineffective hematopoiesis [9][10][11] . Such iron overload is a consequence of the inevitability of frequent blood transfusions and suppression of hepcidin production as a result of ineffective erythropoiesis 12 . Clinical evidence suggests that systemic iron overload has a suppressive effect on hematopoiesis in individuals with MDS or aplastic anaemia, and that iron-chelation therapy often improves this situation [13][14][15] . These observations thus imply that hematopoietic failure promotes systemic iron overload, which in turn exacerbates hematopoietic failure, with the two conditions forming a vicious cycle. Oxidative stress was found to be increased in bone marrow (BM) cells of patients with iron overload, and the impaired hematopoietic function of these individuals was partially rescued by treatment with an antioxidant or iron chelator, suggestive of the initial presence of ROS-induced cellular injury 16 . However, the molecular mechanisms underlying hematopoietic suppression by systemic iron overload in patients as well as the cell-autonomous effect of cellular iron overload on HSC stemness have remained largely unknown.
Here, we show that cellular iron homeostasis governed by the FBXL5-IRP2 axis is integral to the maintenance of HSCs. Ablation of FBXL5 specifically in the hematopoietic system of mice resulted in cellular iron overload in HSCs and impaired their ability to repopulate BM. FBXL5 was also found to be indispensable for the resistance of HSCs to stress induced by myelotoxic agents. FBXL5-deficient HSCs manifested oxidative stress, increased exit from quiescence and eventual exhaustion. Of note, FBXL5 expression was shown to be downregulated in HSCs of some MDS patients, suggesting that disruption of cellular iron homeostasis contributes to hematopoietic failure in such individuals by compromising HSC function. Suppression of IRP2 activity in FBXL5-deficient HSCs restored stem cell function, implicating IRP2 as a potential novel therapeutic target in stem cell diseases such as MDS that are associated with cellular iron overload.

Results
Cellular iron homeostasis is essential for HSC maintenance. We first examined the expression of Fbxl5 in various hematopoietic cell lineages of wild-type mice by reverse transcription (RT) and real-time polymerase chain reaction (rtPCR) analysis. FBXL5 mRNA was detected in many cell lineages including HSCs ( Supplementary Fig. 1a). Among differentiated cells, FBXL5 mRNA was most abundant in myeloid (Gr1 þ Mac1 þ ) cells and least abundant in the erythroid (Ter119 þ ) lineage. These findings are largely consistent with the results of previous microarray 17 ( Supplementary Fig. 1b) and RNA-sequencing 18,19 (Supplementary Fig. 1c) analyses.
To explore the role of cellular iron homeostasis in the maintenance and function of HSCs, we generated mice in which deletion of Fbxl5 is inducible in the hematopoietic system. Crossing of Fbxl5 F/F mice (which harbour floxed alleles of Fbxl5) with Mx1-Cre transgenic mice (which express Cre recombinase under the control of the Mx1 gene promoter) followed by intraperitoneal injection of the resulting offspring (Mx1-Cre/Fbxl5 F/F mice) with poly(I)-poly(C) [poly(I:C)] gives rise to the Fbxl5 D/D genotype in HSCs. Mx1-Cre/Fbxl5 þ / þ mice injected with poly(I:C) were examined as controls. Deletion of Fbxl5 alleles resulted in increased expression of transferrin receptor 1 (TfR1, also known as CD71) at the cell surface for both hematopoietic progenitors (c-Kit þ Sca-1 þ Lin -, or KSL, cells) and HSCs (CD150 þ CD48 -KSL cells) (Fig. 1a). To examine the intracellular abundance of iron in hematopoietic progenitors, we loaded hematopoietic cells with the iron-sensitive fluorophore calcein-AM, the fluorescence of which is quenched on binding to ferrous iron (Fe 2 þ ) 20 . The intensity of calcein fluorescence was significantly lower in Fbxl5 D/D KSL cells than in control KSL cells (Fig. 1b). Exposure of the cells to the cell-permeable Fe 2 þ chelator 2,2 0 -bipyridyl abolished the difference in calcein fluorescence intensity between the two genotypes, confirming that the abundance of Fe 2 þ was increased in the Fbxl5 D/D KSL cells. These data collectively suggested that FBXL5 governs cellular iron homeostasis in HSCs.
We next examined the role of FBXL5 in the maintenance of HSCs. Flow cytometric analysis revealed a slight but significant reduction in the frequency of KSL cells or HSCs in BM of Fbxl5 D/D mice 4 weeks after the last poly(I:C) injection, whereas this reduction was more prominent at 20 weeks (Fig. 1c-e). The frequency of annexin V þ apoptotic cells did not differ between control and Fbxl5 D/D HSCs at 4 weeks (Fig. 1f), indicating that FBXL5 loss impaired HSC maintenance without a significant effect on their survival. These results thus suggested that FBXL5 is required for the maintenance of HSCs.
We further evaluated whether the decrease in the number of Fbxl5 D/D hematopoietic progenitors was exacerbated under iron-overload conditions. To avoid a deleterious effect of iron overload on the liver 4 , we transplanted BM cells from either Mx1-Cre/Fbxl5 F/F or Mx1-Cre/Fbxl5 þ / þ mice not treated with poly(I:C) into lethally irradiated recipients and then injected these animals with poly(I:C). The recipients were fed a high-iron diet for 4 weeks after the last poly(I:C) injection. The frequency of Fbxl5 D/D KSL cells or HSCs in BM was found to be markedly reduced compared with that of the corresponding control cells   To examine whether the reduced frequency of Fbxl5 D/D HSCs in poly(I:C)-treated Mx1-Cre/Fbxl5 F/F mice compromised hematopoietic capacity in vivo, we evaluated the survival rate of the mice after repeated injection with 5-fluorouracil (5-FU) at 10-day intervals to eliminate differentiated hematopoietic cells and induce activation and proliferation of HSCs. The survival rate of the Fbxl5 D/D mice began to decline rapidly at B15 days after the first 5-FU injection, whereas no control mice died for at least 30 days after the first treatment (Fig. 2), suggesting that stress-induced hematopoiesis is defective in the FBXL5-deficient animals.
We also tested the in vitro colony formation capacity of KSL cells in a serial replating assay. Whereas FBXL5 ablation did not affect such capacity at the first plating, the number of clonogenic progenitor cells was markedly reduced for Fbxl5 D/D KSL cells at the second and subsequent platings (Fig. 3a), suggesting that this impairment is induced in a cell-autonomous manner. Even at the first plating, iron overload induced by the presence of ferric ammonium citrate (FAC) reduced the colony formation capacity of Fbxl5 D/D KSL cells compared with control cells (Fig. 3b). In contrast, Fbxl5 D/D and control KSL cells showed a similar colony formation capacity at the second plating in the presence of the ferric iron (Fe 3 þ ) chelator deferoxamine (DFO). These observations suggested that FBXL5 deficiency results in a deterioration in HSC function due to cellular iron overload.
Effect of cellular iron overload on differentiated cells. We next checked whether ablation of FBXL5 in the hematopoietic system affects systemic iron homeostasis and the maintenance of differentiated hematopoietic cells. Measurement of serum iron parameters revealed that the serum iron concentration and transferrin saturation were not altered in Fbxl5 D/D mice (Fig. 4a), suggesting that the loss of FBXL5 in the hematopoietic system had no substantial effect on systemic iron homeostasis. Haematologic parameters of Fbxl5 D/D mice at 4 weeks after poly(I:C) injection were also similar to those of control mice, with the exception of a slight decrease in the number of platelets (Fig. 4b). We further examined the effect of FBXL5 ablation on the differentiation of hematopoietic cells. The frequency of differentiated cells (Gr1 þ Mac1 þ myeloid cells, B220 þ B cells or CD3 þ T cells) in peripheral blood (PB), BM or the spleen of Fbxl5 D/D mice was similar to that in control animals, with the exception of a small increase in the frequency of myeloid cells and decrease in the frequency of B cells in BM ( Fig. 4c-e). In addition, erythropoietic parameters of Fbxl5 D/D mice did not differ from those of control mice at 20 weeks after poly(I:C) injection (Fig. 4f). The frequency of Ter119 þ (erythroid) cells in BM was also not affected by FBXL5 ablation (Fig. 4g). Together, these observations suggested that cellular iron overload induced by FBXL5 ablation had only a small effect on the maintenance of differentiated cells.

Cellular iron overload impairs HSC self-renewal capacity.
Given that FBXL5 deficiency results in a decline in the number of HSCs, we postulated that FBXL5 is essential for the self-renewal capacity of these cells. To examine this possibility, we first assessed the repopulation capacity of Fbxl5 D/D BM cells in a noncompetitive setting. Most lethally irradiated mice transplanted with BM cells (1 Â 10 6 ) from Fbxl5 D/D mice died by B20 days after BM transfer, whereas those transplanted with control BM cells survived (Fig. 5a), suggesting that the repopulation capacity of Fbxl5 D/D BM cells was significantly impaired. To examine the long-term repopulation capacity of Fbxl5 D/D HSCs, we performed a competitive reconstitution assay in which Fbxl5 D/D or control BM cells were transplanted into lethally irradiated C57BL/6 congenic recipient mice together with competitor cells. Flow cytometric analysis of the resulting chimerism in PB of the recipients until 16 weeks after the BM transfer revealed that the long-term repopulation capacity of Fbxl5 D/D HSCs was indeed markedly compromised (Fig. 5b). We further confirmed that Fbxl5 D/D hematopoietic cells manifested almost no reconstitution capacity after a second BM transfer (Fig. 5c). The number of KSL cells derived from Fbxl5 D/D donor cells was greatly reduced in BM of the initial recipient mice at 16 weeks after BM transfer (Fig. 5d). To characterize the repopulation defect in Fbxl5 D/D HSCs, we examined the homing capacity of Fbxl5 D/D hematopoietic progenitor cells after transplantation. Fbxl5 D/D or control KSL cells were sorted, labelled with carboxyfluorescein succinimidyl ester (CFSE) and transplanted into lethally irradiated recipients, and the recipient BM was analysed 16 h after transplantation. The homing capacity of the transplanted CFSE þ cells for BM was similar for the two genotypes ( Fig. 5e), excluding the possibility that a homing defect is responsible for the repopulation defect of Fbxl5 D/D HSCs. To evaluate the stem cell capacity of Fbxl5 D/D HSCs excluding homing and engraftment, we transplanted BM cells from either Mx1-Cre/Fbxl5 F/F or Mx1-Cre/Fbxl5 þ / þ mice not treated with poly(I:C) into lethally irradiated recipients together with competitor cells. Four weeks after BM transfer, we confirmed that donor cells were reconstituted in the recipient BM and then injected the recipient mice with poly(I:C). Fbxl5 D/D HSCs gradually lost long-term repopulation capacity (Fig. 5f), showing that such capacity after homing and engraftment was impaired in Fbxl5 D/D HSCs. The differentiation of Fbxl5 D/D HSCs appeared normal in this setting of competitive repopulation (Fig. 5g). These results collectively indicated that cellular iron homeostasis is essential for the self-renewal capacity of HSCs.
Role of IRP2 in the effect of FBXL5 on HSC stemness. Given that impaired degradation of IRP2 is primarily responsible for the embryonic mortality of Fbxl5 -/mice 4 , we hypothesized that the defective stem cell capacity of Fbxl5 D/D HSCs might also be due to IRP2 accumulation. We therefore evaluated whether suppression of IRP2 is required for the repopulation capacity of HSCs with the    60 use of a competitive reconstitution assay. Both the repopulation and differentiation capacities of Irp2 -/-HSCs were similar to those of control (Irp2 þ / þ ) HSCs (Fig. 6a-c). We next prepared Mx1-Cre/Fbxl5 F/F /Irp2 -/and Mx1-Cre/Fbxl5 þ / þ /Irp2 þ / þ mice to be able to analyse Fbxl5 D/D /Irp2 -/-HSCs after poly (I:C) injection. The competitive reconstitution assay revealed that the long-term repopulation capacity of Fbxl5 D/D /Irp2 -/-HSCs did not differ significantly from that of control HSCs after the first or second BM transfer (Fig. 6d-  ablation on the repopulation capacity of HSCs. FBXL5 thus protects HSC stemness through suppression of IRP2.

Cellular iron overload disrupts redox regulation in HSCs.
To investigate further the mechanism by which cellular iron overload impairs HSC function, we profiled gene expression in  Data 1 and 2). A complete list of these genes has been deposited in the Gene Expression Omnibus (GEO) database under the accession number GSE93649. As expected, expression of genes related to cellular iron metabolism such as Hbb, Slc48a1, Ftl1, Lcn2 and Abcb6 was upregulated. The upregulated genes also included many genes important for redox regulation, such as Mt1, Mt2, Hmox1, Gstm2, Slc7a11, Gclm, Gsta4, Cat, Txn1, Nqo1 and Sod1.
Ingenuity pathway analysis (IPA) revealed that the differentially expressed genes were most highly associated with the NRF2mediated oxidative stress response (Fig. 7a). Gene set enrichment analysis (GSEA) also confirmed that the antioxidant defense system is activated in Fbxl5 D/D HSCs (Fig. 7b,c). Changes in the expression of several genes related to oxidative stress responses were validated by RT and rtPCR analysis (Fig. 7d). On the basis of these results, we concluded that FBXL5 ablation evokes oxidative stress in HSCs. Oxidative stress in HSCs is also known to give rise to phosphorylation (activation) of p38 mitogen-activated protein kinase (MAPK), which results in aberrant cell proliferation and exhaustion 1,2 . Intracellular flow cytometric analysis revealed that the frequency of cells positive for phosphorylated p38 MAPK was greater among Fbxl5 D/D HSCs than among control HSCs (Fig. 7e). The mean fluorescence intensity (MFI) of phospho-p38 for Fbxl5 D/D HSCs also tended to be greater than that for control HSCs, although this difference did not achieve statistical significance (P ¼ 0.062). Given that both extrinsic   21 , we sought to examine the influence of only intrinsic factors on p38 phosphorylation in HSCs. To this end, we transplanted BM cells from Mx1-Cre/Fbxl5 F/F or Mx1-Cre/Fbxl5 þ / þ mice into lethally irradiated recipients and then injected these animals with poly(I:C). The frequency of Fbxl5 D/D HSCs in BM of the recipients was significantly reduced compared with that of control HSCs (Fig. 7f), as was the case for Mx1-Cre/Fbxl5 F/F mice treated with poly(I:C) (Fig. 1e). The frequency of cells positive for phosphorylated p38 MAPK as well as the MFI for phospho-p38 were also significantly  NQO1  GCLC  ABCC1  CAT  PRDX1  FES  MPO  SRXN1  NDUFB4  GLRX  PFKP  PRDX2  ERCC2  LSP1  GPX4  EGLN2  NDUFS2  TXNRD1  SOD1  CDKN2D  SOD2  PRDX6  NDUFA6  HHEX  PPP2R4  STK25  SBNO2  ATOX1  OXSR1  GPX3  GLRX2  HMOX2  PDLIM1  TXNRD2  GSR  MGST1  MBP  increased in Fbxl5 D/D HSCs compared with control HSCs in the recipient mice (Fig. 7g). These results thus also confirmed that Fbxl5 D/D HSCs are exposed to intense oxidative stress.
Cellular iron homeostasis is essential for HSC dormancy. GSEA also showed that the gene expression profile of Fbxl5 D/D HSCs was shifted toward proliferation compared with that of control HSCs (Fig. 8a). Given that the loss of dormancy in HSCs leads to their exhaustion, we hypothesized that FBXL5 ablation might promote exit from the dormant state in HSCs. We therefore evaluated the cell cycle kinetics of Fbxl5 D/D HSCs by intracellular staining of the proliferation marker Ki-67 and analysis of DNA ploidy by staining with Hoechst 33342. The frequency of cells in the dormant state (Ki-67fraction) was reduced for Fbxl5 D/D HSCs compared with control cells (Fig. 8b). A similar difference in the frequency of dormant HSCs was also observed in poly(I:C)treated recipients of transplanted BM cells from Mx1-Cre/Fbxl5 F/ F or Mx1-Cre/Fbxl5 þ / þ mice (Fig. 8c). GSEA also revealed loss of an HSC-specific gene signature in Fbxl5 D/D HSCs (Fig. 8d). The top 30 downregulated genes in our microarray analysis (Supplementary Data 2) include the HSC-specific gene Necdin 22 , whose downregulation in Fbxl5 D/D HSCs was confirmed by RT and rtPCR analysis (Fig. 8e). We also confirmed the downregulation of the HSC-specific gene p57 (ref. 23), whereas the expression of other such genes including Hoxb5 (ref. 24) and Hif1a (ref. 25) was not affected (Fig. 8e). These results suggested that cellular iron overload induces exit of HSCs from the dormant state and loss of an HSC-specific gene expression signature.

Disrupted iron homeostasis in HSCs is associated with MDS.
The results of our mouse experiments together indicated that cellular iron homeostasis governed by FBXL5 plays an essential role in the maintenance and function of HSCs. We finally examined whether FBXL5 deficiency might be associated with human hematopoietic diseases. Impaired HSC function can result in the development of MDS, a clonal HSC disorder characterized by ineffective hematopoiesis [9][10][11] . Expression of FBXL5 was shown to be differentially downregulated in CD34 þ CD38 -CD90 þ HSCs from MDS patients with deletion of chromosome 5q relative to those from healthy control subjects 26 . Analysis of a published set of microarray data revealed that FBXL5 expression was also significantly downregulated in Lin -CD34 þ CD38 -CD90 þ CD45RA -HSCs from eight MDS patients without deletion of chromosome 5q compared with 11 age-matched healthy control samples 27 (Fig. 9a). It is of note that the expression of both TFR1 and DMT1, which is upregulated by IRP2 and therefore represents an index of IRP2 activity, was also increased in the MDS patients (Fig. 9b,c). To determine whether FBXL5 expression is also downregulated in more differentiated hematopoietic progenitor cells in MDS patients, we evaluated a published set of microarray data for CD34 þ hematopoietic progenitor cells from 183 MDS patients with various cytogenetic abnormalities and 17 healthy control subjects 28 . Expression of FBXL5 was significantly downregulated in the CD34 þ cells of patients with refractory anaemia with ringed sideroblasts (RARS), a subgroup of MDS characterized by iron deposition and apoptosis in hematopoietic progenitor cells (Fig. 9d). Consistent with this finding, TFR1 expression was also significantly upregulated in the CD34 þ cells from the RARS patients (Fig. 9e). These findings implicate cellular iron overload due to FBXL5 downregulation in the pathogenesis of human hematopoietic failure. Given that suppression of aberrant IRP2 activity cancelled the deleterious effect of FBXL5 ablation on the repopulation capacity of HSCs (Fig. 6), IRP2 is a potential therapeutic target for cellular iron overload in HSCs due to FBXL5 downregulation, including that in patients with MDS. Consistent with this notion, a published microarray data set revealed that an increased IRP2 mRNA abundance in CD34 þ hematopoietic progenitor cells was related to reduced survival in MDS patients without deletion of chromosome 5q (ref. 29) (Fig. 9f).

Discussion
We have here discovered a previously unrecognized role for cellular iron homeostasis in the maintenance of HSCs with the use of mouse models of conditional Fbxl5 deletion. Mechanistically, cellular iron homeostasis in HSCs regulates oxidative stress, quiescence and self-renewal capacity. Analysis of public data sets revealed that downregulation of FBXL5 expression was associated with MDS, a disease characterized by BM failure. Suppression of IRP2 activity in FBXL5-deficient HSCs restored stem cell function, implicating IRP2 as a potential therapeutic target for cellular iron overload in HSCs with FBXL5 deficiency (Fig. 10). FBXL5 is a master regulator of cellular iron metabolism by virtue of its role as the substrate recognition component of the SCF FBXL5 E3 ubiquitin ligase for IRP2 degradation 5,6 . Other proteins targeted by FBXL5 for proteasomal degradation include p150 Glued , cortactin and single-stranded DNA binding protein 1 (SSB1) [30][31][32] . FBXL5 has also been shown to interact with Snail1 (refs 33,34) and CBP/p300-interacting transactivator 2 (CITED2) 35 , leading to their degradation. CITED2 was shown to control the proliferation of mouse embryonic fibroblasts by promoting expression of the Polycomb group genes Bmi1 and Mel18 (ref. 36) as well as to selectively maintain adult HSC function at least in part through regulation of p16 and p53 (ref. 37). If CITED2 accumulates in FBXL5-deficient HSCs, it might also promote their proliferation and exhaustion. However, FBXL5-deficient mice die during embryogenesis and their mortality is prevented by additional ablation of IRP2, suggesting that impaired IRP2 degradation is primarily responsible for the embryonic death 4 . Our data now provide evidence that IRP2 is also the major target of SCF FBXL5 in HSCs, given that the defect in repopulation capacity of FBXL5-deficient HSCs was rescued by additional ablation of IRP2. These lines of evidence also suggest that the contribution of CITED2 to the phenotype of FBXL5-deficient HSCs is limited. We therefore conclude that FBXL5 plays an essential role in the maintenance of HSCs through suppression of IRP2 activity.
Our present study shows that disruption of cellular iron homeostasis by FBXL5 ablation in HSCs resulted in cellular iron overload, oxidative stress responses, exit from dormancy and eventual exhaustion in these cells. Increased ROS levels promote the proliferation and differentiation of HSCs, primarily via modulation of p38 MAPK and the forkhead box O (FOXO) family of transcription factors 1 . Many mutations that result in aberrantly high ROS levels in HSCs also lead to impairment of quiescence and self-renewal potential as a result of enhanced differentiation 38 . We now show that FBXL5 is also an essential ROS regulator in HSCs and plays a key role in the maintenance of stemness. The expression of many genes that contribute to oxidative stress responses was found to be upregulated in FBXL5deficient HSCs: The most upregulated ROS-related genes included those for metallothionein (MT) 1 and MT2, which are small, cysteine-rich, and heavy metal-binding proteins that participate in an array of protective stress responses 39 . These proteins thus protect cells from exposure to oxidants and electrophiles, which react readily with sulfhydryl groups. Moreover, they play a key role in regulation of cellular zinc levels by binding and releasing zinc. The marked upregulation of MT gene expression in FBXL5-deficient HSCs might thus also modify cellular zinc metabolism. Another such upregulated gene was Slc7a11, which encodes a component of system Xc -. System Xccontributes to the maintenance of redox homeostasis by importing cystine for synthesis of the major cellular antioxidant glutathione. Inhibition of system Xcby erastin in cancer cells triggers ferroptosis, a recently recognized form of iron-dependent cell death 40 . Upregulation of Slc7a11 expression might therefore represent a mechanism to protect HSCs with cellular iron overload from ferroptosis. Indeed, FBXL5 ablation was shown to promote erastin-induced ferroptosis in cultured cells 40 . These changes in gene expression in FBXL5-deficient HSCs are thus indicative of iron-mediated cellular damage and disruption of redox homeostasis.
In contrast to the deleterious effect of iron overload evoked by FBXL5 loss on HSC function, FBXL5 deficiency did not substantially affect differentiated hematopoietic cells. Although cross talk between systemic iron homeostasis and erythropoiesis is well established 41 , detailed analysis of erythropoiesis in FBXL5deficient mice indicated that FBXL5 has a limited role in erythropoiesis, consistent with the finding that the amount of FBXL5 mRNA is smallest in the erythroid lineage among differentiated haematopoietic cells. In general, erythroid cells require large amounts of iron to sustain haemoglobin synthesis 41 , suggesting that the importance of FBXL5 as a brake on iron uptake might be rather limited in erythropoiesis.
Systemic iron overload is sometimes a complication of hematopoietic failure such as that associated with MDS as a result of the required frequent blood transfusions and the suppression of hepcidin production due to ineffective erythropoiesis 12 . Systemic iron overload in turn has a suppressive effect on hematopoiesis in patients with hematopoietic failure, with iron-chelation therapy having been found to be beneficial in these patients [13][14][15] . However, iron overload has little effect on hematopoiesis in patients with hereditary hemochromatosis, which is also a major cause of systemic iron overload due to hepcidin deficiency. These various observations suggest that the hematopoietic system in some patients with BM failure is intrinsically vulnerable to iron. By analysing public data sets, we found that FBXL5 expression was significantly downregulated in HSCs in a subset of patients with MDS. Reduced expression of FBXL5 might promote loss of quiescence and exhaustion in HSCs of such patients. Similar to the Fbxl5 D/D HSCs examined in the present study, HSCs of MDS patients with low FBXL5 expression might also be vulnerable to systemic iron overload. The expression of FBXL5 was also found to be significantly downregulated in CD34 þ progenitor cells of patients with RARS, a subtype of MDS characterized by iron deposition in hematopoietic progenitor cells. Downregulation of FBXL5 expression in RARS is of interest given that RARS progenitor cells are loaded with excess iron in mitochondria and are vulnerable to ROS-induced apoptosis 9 . FBXL5 deficiency might thus exacerbate the cellular and mitochondrial iron overload in RARS progenitor cells, contributing to disease pathogenesis. Our findings raise the possibility that FBXL5 plays a key role in the pathogenesis of BM failure syndromes including MDS, a possibility that warrants further investigation.
Our finding that suppression of aberrant IRP2 activity rescued the defect in repopulation capacity of HSCs induced by FBXL5 ablation suggests that targeting of IRP2 is effective for mitigation of cellular iron overload and is therefore a candidate for therapeutic application. This notion is further supported by the finding that an increased abundance of IRP2 mRNA in CD34 þ hematopoietic progenitor cells was related to reduced clinical survival in MDS patients without deletion of chromosome 5q. Given that complete loss of IRP2 gives rise to microcytic anaemia 42,43 , however, an appropriate level of IRP2 suppression would be needed. Another limitation of such a therapeutic strategy is that an IRP2 inhibitor has not yet been developed. Given that IRP2 is an mRNA binding protein, inhibition of such binding with antisense oligonucleotides is a possible approach. Despite these limitations, inhibition of IRP2 is a potentially novel approach to the treatment of hematopoietic failure associated with FBXL5 downregulation and cellular iron overload.

Methods
Mice. Generation of Fbxl5 F/F mice was described previously 4 . These mice were crossed with Mx1-Cre transgenic mice 44 or Irp2 À / À mice 45 to generate Mx1-Cre/Fbxl5 F/F and Mx1-Cre/Fbxl5 F/F /Irp2 À / À mice. All of these mice were backcrossed with C57BL/6 mice for more than six generations. Expression of Cre recombinase in mice harbouring the Mx1-Cre transgene was induced by intraperitoneal injection of poly(I:C) (R&D Systems, Minneapolis, MN) at a dose of 20 mg kg À 1 on seven alternate days beginning at 8 weeks of age. Mx1-Cre/Fbxl5 F/F mice and Mx1-Cre/Fbxl5 F/F /Irp2 À / À mice were analysed 4 weeks after the last poly(I:C) injection unless indicated otherwise. Irp2 À / À mice were analysed at 14 weeks of age. C57BL/6-Ly5.1 congenic mice were obtained from The Jackson  Figure 10 | Cellular iron homeostasis is essential for the maintenance of HSCs. (a) Limitation of the production of ROS is essential for the maintenance of HSCs. Iron is a major elicitor of oxidative stress, with the result that cellular iron homeostasis is strictly regulated in HSCs. Cellular iron homeostasis is governed by the FBXL5-IRP2 axis in these cells. (b) FBXL5 deficiency in HSCs results in cellular iron overload as a result of IRP2 overactivity. Cellular iron overload promotes oxidative stress, which leads to dysfunction and exhaustion of HSCs. Downregulation of FBXL5 expression in HSCs is apparent in human hematopoietic failure such as that associated with MDS. (c) Suppression of aberrant IRP2 activity attenuates the deleterious effect of FBXL5 ablation on the stem cell capacity of HSCs, suggesting that IRP2 is a potential therapeutic target for cellular iron overload in HSCs associated with FBXL5 downregulation.
Laboratory (Bar Harbor, ME). For some experiments, mice were injected intraperitoneally with 5-FU (Sigma, St Louis, MO) at a dose of 150 mg kg À 1 or fed a high-iron diet formulated by supplementation of CA-1 (containing 0.03% ferric citrate; CLEA, Tokyo, Japan) with 2% ferric citrate. All mouse experiments were approved by the Animal Ethics Committee of Kyushu University.
Evaluation of intracellular iron status. BM mononuclear cells were stained with 0.25 mM calcein-AM (Molecular Probes) in PBS for 5 min at 37°C and then washed with PBS. The calcein-loaded cells were further stained with antibodies to surface markers for 30 min on ice, washed, incubated for 20 min on ice in PBS with or without 100 mM 2,2 0 -bipyridyl (Sigma), and analysed with the FACSAria instrument.
Detection of apoptosis. For assay of apoptosis, BM cells stained with antibodies to cell surface markers were further stained for 15 min at room temperature with annexin V and propidium iodide with the use of an Annexin V-FITC Apoptosis Detection Kit (BD Biosciences).