Essential role of autophagy in protecting neonatal haematopoietic stem cells from oxidative stress in a p62-independent manner

Autophagy is a cellular degradation system contributing to homeostasis of tissue stem cells including haematopoietic stem cells (HSCs). It plays pleiotropic roles in HSC characteristics throughout life, but its stage-specific roles in HSC self-renewal are unclear. To investigate the effects of Atg5 deletion on stage-specific HSC functions, we compared the repopulating capacity of HSCs in Atg5f/f;Vavi-cre mice from postnatal day (P) 0–7 weeks of age. Interestingly, Atg5 deficiency led to no remarkable abnormality in the HSC self-renewal capacity at P0, but significant defects at P7, followed by severe defects. Induction of Atg5 deletion at P5 by tamoxifen administration to Atg5f/f;Rosa26-Cre-ERT2 mice resulted in normal haematopoiesis, including the HSC population, until around 1 year, suggesting that Atg5 in the early neonatal period was critical for haematopoiesis in adults. Mitochondrial oxidative stress was increased by Atg5 loss in neonatal HSC/progenitor cells. Although p62 had accumulated in immature bone marrow cells of Atg5f/f;Vavi-cre mice, p62 deletion did not restore defective HSC functions, indicating that Atg5-dependent haematopoietic regulation in the developmental period was independent of p62. This study proposes a critical role of autophagy in HSC protection against harsh environments in the early neonatal stage, which is essential for healthy long-term haematopoiesis.

As a result, we demonstrated that autophagy plays a critical role in protecting neonatal HSCs from oxidative stress in a p62-independent manner.

Results
Progressive haematopoietic failure by Atg5 deletion during developmental period. To investigate the roles of autophagy in development of haematopoiesis in mice, we analysed the effects of Atg5 deletion on haematopoiesis using the Vav-Cre-loxP system, in which Cre recombinase-mediated loxP-flanked gene deletion in haematopoietic tissue starts from around embryonic day (E) 11.5 34 . The Atg5 deletion efficiency was sufficiently high using this system (Supplementary Figure S1). Consistent with previous reports 29,30 , Atg5 f⁄f ;Vav mice showed severe haematopoietic failure at 7 weeks of age, and most Atg5-deficient mice died presumably because of further progression of haematopoietic failure ( Supplementary Fig. S2). To evaluate the effect of Atg5 on HSC development, we investigated the long-term HSC population (SLAM LSK; CD48 − CD150 + lineage − Sca-1 + c-Kit + ) in P7, 3-week-old and 7-week-old mice. In bone marrow (BM) of 7-week-old mice, SLAM LSK cells were reduced dramatically by loss of the Atg5 gene ( Fig. 1) as shown previously 29,30 . This abnormality was not remarkable at earlier stages (Fig. 1). Therefore, as suggested in the previous reports, these data indicate that autophagy may play more critical roles in the later period rather than earlier developmental stages, such as the neonatal period.
Defect in the self-renewal capacity of HSCs among BM cells of Atg5 f/f ;Vav mice at P7, but not P0. To evaluate the effect of Atg5 deficiency on HSC functions, we performed competitive reconstitution assays using BM cells of Atg5 f/f ;Vav mice from P0 to 7 weeks of age. When we transplanted BM from P0 to 7-week-old mice, all lineages derived from Atg5-deficient mice were reduced significantly in PB, which reduced further with age ( Fig. 2a, Supplementary Fig. S3). In BM, regeneration of Atg5-deficient BM cells at 16 weeks after transplantation was reduced dramatically in 3-and 7-week-old BM cell recipients, indicating that Atg5deficient mice lost their self-renewal capacity at 3 weeks of age (Fig. 2b). However, interestingly, its repopulation in BM was almost comparable to the control, when we transplanted Atg5-deficient BM cells at P0 (Fig. 2b). These data suggest that although production of mature cells in PB may be impaired by Atg5, development of immature cells was unaffected at P0. Hence, we analysed donor-derived HSC/progenitor cells (Lineage − Sca1 + c-Kit + ; LSK cells) at 16 weeks after transplantation to assess repopulation of HSCs, that is, their self-renewal capacity. We found that the frequency of donor-derived LSK cells was comparable to the control in transplantation of Atg5-deficient BM cells at P0, indicating that the self-renewal capacity of Atg5-deficient cells was normal at P0 (Fig. 2b). In transplantation of BM cells from P7 mice, we found a mild but significant reduction in repopulation of Atg5-deficient LSK cells, as well as myeloid, B, and T cells, indicating a reduction in the self-renewal capacity of HSCs (Fig. 2b). Thus, HSC dysfunction due to Atg5 loss was detected at P7, but not P0, followed by progressive dysfunction of HSCs in later stages.
No remarkable abnormality in long-term haematopoiesis, including HSC functions, by Atg5 deletion after P5. Next, we investigated the effect of Atg5 loss on haematopoiesis, when it was induced at an earlier time after birth. For this purpose, we used two systems, Atg5 f/f ;Mx1-Cre (Atg5 f/f ; Mx1), and Atg5 f/f ;Rosa26-Cre-ER T2 (Atg5 f/f ;Rosa) or Atg5 f/− ;Rosa26-Cre-ER T2 (Atg5 f/− ;Rosa), in which Atg5 gene deletion is induced by poly(I:C) and tamoxifen treatment, respectively. As the standard method, we started administration of poly(I:C) at 4 weeks of age and analysed haematopoiesis in adults. In our experiments, the Atg5 deficiency caused a mild but insignificant increase of LSK cells and LT-HSCs associated with increased myeloid cells and decreased B cells in PB, which is consistent with a previous report (Supplementary Fig. S4) 33 . These data indicated that deletion of Atg5 gene in later developmental periods did not remarkably affect the HSC population. To delete the Atg5 gene at the earliest timing after birth, we attempted to administrate tamoxifen at P5, followed by long-term observation of haematopoiesis. Although some mouse neonates died because of injury due to the intraperitoneal injection, most survived and appeared healthy. Analysis of PB showed normal haematopoiesis of Atg5-deficient mice. In some mice, we analysed HSCs and progenitors in BM after long-term observation (e.g., 8-12 months of age). The HSC and progenitor populations were comparable to the control (Fig. 3). To check the efficiency of gene deletion, we performed colony formation assays, followed by genotyping of individual colonies. We found that about 70% of colonies on average showed deletion of both alleles, indicating that the majority of LT-HSCs were deficient for Atg5 ( Supplementary Fig. S5). Although there was the possibility that partial deletion of the Atg5 gene may fail to result in remarkable phenotypes, these data support that Atg5 was dispensable to develop and maintain HSCs when the Atg5 gene was deleted at around 1 week after birth.
Increased mitochondrial oxidative stress by Atg5 deletion is associated with cell death at the neonatal stage. The above data prompted us to focus our analysis on HSCs at P7. First, to monitor autophagic activity in Atg5 f/f ;Vav and Control (Atg5 f/w ;Vav and Atg5 f/w ) mice, we used the GFP-LC3 system with FACS analysis 35,36 . If autophagy occurs, the GFP intensity decreases because it is digested and/or attenuated by the acidic condition in lysosomes in addition to LC3 protein, whereas GFP-LC3 accumulates upon autophagy inactivation. We compared the GFP fluorescence intensity in HSCs/progenitors (LSK) in GFP-LC3;Atg5 f/f ;Vav and GFP-LC3;Atg5 f/w ;Vav or GFP-LC3;Atg5 f/w mice. We found that Atg5 deletion caused a clear increase of GFP intensity, which indicated that autophagy was impaired by Atg5 deletion (Fig. 4a). Although, haematopoietic populations of Atg5-deficient mice appeared to be normal as shown in Fig. 1, we performed gene expression pro-    Supplementary Fig. S6). Additionally, we found elevated mitochondrial reactive oxygen species (ROS) detected by MitoSox staining in LSK cells of Atg5 f⁄f ;Vav mice at P7 (Fig. 4c). Increased mitochondrial ROS was also observed in LSK cells of 3-weekold Atg5 f⁄f ;Vav mice. We did not find an apparent increase of mitochondrial mass detected by MitoTracker staining or constant alteration of the mitochondrial membrane potential detected by TMRM staining caused by Atg5 deficiency at both P7 and 3 weeks of age. While p62 mediates selective degradation of autophagic cargo, it also has a role in the stress response through regulation of the Keap1-Nrf2 system that has been recognised as one of the major cellular defence mechanisms against oxidative and electrophilic stresses. It has been reported that p62  www.nature.com/scientificreports/ accumulation leads to activation of Nrf2, which induces many anti-oxidative genes 37 . It has also been reported that Nrf2 induces gene expression involved in anabolic pathways such as the pentose phosphate pathway 12,38 . As expected, we confirmed that p62 protein had accumulated in immature BM cells (lineage − c-Kit + ) of Atg5 f⁄f ;Vav   Fig. S7). Consistently, we found that Nrf2 target molecules, including G6pd2 and Taldo1 at around P7 (Fig. 4d) and NAD(P)H dehydrogenase quinone 1 (Nqo1) and glutathione S-transferase mu1 (Gstm1) at 7 weeks of age (Fig. 4e), were upregulated in LSK cells of Atg5-deficient mice. Furthermore, dead cells (7-AAD + ) were increased by Atg5 (Fig. 4f). These data suggested that mitochondrial oxidative stress caused HSC damage in the neonatal stage, leading to a severe abnormality of haematopoiesis in adults as shown in Supplementary  Fig. S2.
No restoration of haematopoietic abnormality in Atg5-deficient mice by p62 deletion. Finally, we investigated whether p62 was involved in the phenotype of Atg5 deletion by the Vav-Cre-loxP system. We confirmed sufficiently high efficiency of p62 deletion using this system (Supplementary Figure S1). p62 f/f ;Vav-Cre (p62 f/f ;Vav) mice appeared healthy and did not show any haematopoietic abnormalities such as anaemia or an altered distribution of lineage cells in PB (Fig. 5a,b). In BM, HSC/progenitor populations were almost comparable to the control, except for mildly reduced CMP in p62 f/f ;Vav mice at 7 weeks of age (Fig. 5c,d). Myeloid cells and lymphocytes in BM and spleen were normal (Supplementary Fig. S8). Reconstitution assays showed that the functions of p62-deficient HSCs, including self-renewal, were not impaired ( Supplementary Fig. S8). Thus, p62 loss alone barely affected haematopoiesis. Hence, we assessed the effects of loss of p62 on haematopoiesis in Atg5 f⁄f ;Vav mice at 7 weeks of age. As shown in Supplementary Fig. S2, Atg5 f⁄f ;Vav mice at 7 weeks of age showed severe haematopoietic dysfunctions such as severe anaemia and leukocytopenia associated with an abnormality in balance of peripheral cell components, such as increased ratios of myeloid/B cells, compared with control (Atg5 w/w ;Vav) mice (Fig. 5a,b). In Atg5 f⁄f ;Vav mouse BM, in addition to reduced LT-HSCs 29,30 , significant increases of myeloid-biased multipotent progenitor (MPP) 2 and MPP3 were observed (Fig. 5c). Consistent with previous reports, granulocyte-macrophage progenitors (GMPs) were reduced significantly 30 , and common myeloid progenitors (CMPs)/megakaryocyte-erythrocyte progenitors (MEPs) were reduced similarly. Additionally, we found that mature B cells in BM were reduced drastically in Atg5 f⁄f ;Vav mice ( Fig. 5d). In Atg5 f/f ;p62 f/f ;Vav mice, most haematopoietic parameters in PB, BM, and spleen appeared to be comparable with Atg5 f/f ;Vav mice ( Fig. 5a-c, Supplementary Fig. S8), indicating that loss of p62 did not restore the defective haematopoiesis due to Atg5 deficiency. Interestingly, we found that Atg5 f/f ;p62 f/f ;Vav mice consistently exhibited more severe phenotypes, particularly in the LT-HSC population as well as proB/preB and immature B cells, suggesting that Atg5 and p62 partly support HSC functions in a collaborative manner. To investigate the roles of p62 in defective phenotypes of Atg5 deficiency, we performed reconstitution assays using Atg5 f⁄f ;p62 f⁄f ;Vav mice. While the repopulating capacity in all lineage was reduced significantly by Atg5 loss both in PB and BM of recipients, p62 loss did not restore it (Fig. 6a,b). Because of severe haematopoietic defects, Atg5 f⁄f ;Vav mice survived for 15 weeks on average, and lethality of Atg5/p62-deficient mice was also comparable to that of Atg5-deficient mice (Fig. 6c). Thus, we concluded that Atg5-dependent haematopoietic regulation during the developmental period was independent of p62. Additionally, we evaluated the effects of p62 deletion in Atg5-deficient mice that lost Atg5 after birth by poly (I:C) administration using the Mx1-Cre-loxP system. Consistent with a previous report 33 , Atg5 f/f ;Mx1 mice showed a slight increase of GMPs, but it was largely comparable with the control for the HSC/progenitor population (Fig. 7a). In lineage cells, a slight increase of splenic myeloid cells and altered erythroid development were observed (Supplementary Fig. S9). Apart from Atg5 f/f ;Vav mice, Atg5 f/f ;Mx1 mice also showed a reduction of mature B cells (Fig. 7b). These alterations were not restored by p62 deletion (Fig. 7a,b, Supplementary Fig. S9 and S10). In summary, haematopoietic regulation by Atg5 during both developmental and adult periods was independent of p62.

Discussion
In mice, primitive haematopoiesis begins in blood islands of the yolk sac at E7, followed by the switch from primitive to definitive haematopoiesis at E10 and 11. At the beginning of the perinatal period, the liver is still a source of extra-medullary haematopoiesis, and then the BM eventually takes over 39 . During the perinatal period, foetal and neonatal development and maturation are deeply affected by environmental conditions such as oxidative stress, nutritional stress, and microbial infection. Although it has been demonstrated that autophagy plays a critical role in development and aging of HSCs, the importance of autophagy in perinatal HSCs, especially during the neonatal period, has not been revealed. Based on this study, we propose a critical role of autophagy in protection of HSCs against harsh environments in the early neonatal stage, which is essential for healthy long-term haematopoiesis. The perinatal period is accompanied by dramatic environmental changes in both oxidative stress and nutritional conditions, which may affect autophagic activity. Because the transplacental nutrient supply is suddenly interrupted at birth, neonates have stress due to starving until milk feeding. Foetal life evolves under a hypoxic condition, but it provides comfortable environments for development of organs. During foetal-to-neonatal transition, asphyxia is characterised by periods of severe hypoxia that may evolve to ischemic organ impairment. In another aspect of oxygen, foetal-to-neonatal transition may cause "relative hyperoxia". In both cases, oxidative stress occurs in neonatal tissues. Although it is unclear which factors activate autophagy, it is assumed that autophagy is important for survival of HSCs against specific harsh neonatal environments. To understand the neonatal period better, we analyzed public RNA-seq data from wild type neonatal (P7) HSCs (CD48 -CD150 + Lineage -Sca-1 + c-Kit -) and adult HSCs deposited in GEO (GSE128762) 40 . We found that expression of many genes involved in glucose and glutamine metabolisms, mitochondria, and redox regulation varied significantly between at P7 and adult mice. The expression levels of most of these genes are significantly lower in neonatal HSCs compared with those in adult HSCs (Supplementary Fig. S11). It has been reported that the metabolism and redox regulation are closely related to the regulation and maintenance of HSCs in adults [41][42][43] . Mitochondria, a dynamic organelle in which many metabolic processes occur, also have critical roles in maintenance of HSCs 44 . The lower expression of these genes in HSCs during neonatal period suggest that the protective www.nature.com/scientificreports/ system of HSCs against metabolic stress may be somewhat different among the developmental stages, and that survival of neonatal HSCs may be highly dependent on autophagy. Additional analysis is needed to understand the stage-specific roles of autophagy. Further understanding of the roles of autophagy in neonates may uncover novel pathophysiological insights of disease that caused by autophagy-related abnormalities in the neonatal period. One such insight may be the mitochondrial behaviour of HSCs. Accumulating evidence has shown that HSCs with increased and active mitochondria have loss of haematopoietic functions such as the reconstitution ability [45][46][47] . In our study, alteration of the TMRM level was not apparent at P7 and 3 weeks of age, although an increased mitochondrial mass, TMRM level, and mitochondrial ROS were observed at 7 weeks ( Supplementary Fig. S12) consistent with previous data 30 . Although we do not have clear evidence of the roles of oxidative stress in defective phenotypes of HSCs in Atg5 f/f ;Vav mice, we assume that the functional change of mitochondria may occur at the perinatal stage. Further analysis will be needed to understand the molecular mechanism. www.nature.com/scientificreports/ In this study, we focused on the role of p62 in autophagy-dependent anti-oxidative effects on HSCs. p62 is a signalling hub that interacts with various partners. Previous functional analyses of p62 knockout (p62 −⁄− ) mice indicated that the main phenotype of the mice is obesity 48 . p62 −⁄− mice do not show haematopoietic failure for up to 6 months after birth or transplantation 49 . Consistently, our analysis of p62 f/f ;Vav mice showed that the haematopoietic stem cells were normal and their reconstitution ability were maintained (Fig. 5c and Supplementary Fig. S8). Because it has been reported that p62 is required for amino acid sensing via the mTOR pathway 50 , p62 might play a critical role in the protective response to abnormalities in amino acid levels. Additionally, we analysed Atg5/p62-deficient HSCs. p62 ablation suppresses the hepatic phenotype, but not affect the neuronal defect caused by autophagy [10][11][12] . This study clarified that p62 was not the cause of the severe functional decline of   Fig. S7 and Fig. 4d,e), the gene induction was not remarkable compared with Atg5-deficient hepatic cells 12 . Because autophagy has pleiotropic roles in haematopoiesis, we also analysed other haematopoietic cells in addition to HSCs. For example, we found severe impairment of B cell development in Atg5 f/f ;Vav mice, which was consistent with the phenotype of B cell-specific Atg5-deficient mice 23 , but p62 loss did not rescue it. Atg5 f/f ;Mx1 mice also showed impairment of B cell development, but the phenotype was much more milder than that of Atg5 f/f ;Vav mice. p62 loss did not rescue impaired B cell development in Atg5 f/f ;Mx1 mice. In some cases, concomitant deletion of p62 worsens the phenotype of Atg5 loss, suggesting that p62-mediated pathways collaborate with autophagy for proper HSC development. One possibility is that p62 may have a protective role mediating mTOR signalling under autophagy-deficient conditions. It has been reported that p62 interacts with mTORC1 components Raptor and Rags, and mediates amino acid sensing 50 . mTORC1 has an essential role in HSC functions in vivo 51,52 . Because p62 deficiency did not cause apparent haematopoietic abnormalities except for a mild reduction of CMP, autophagy and p62 may act synergistically for HSC functions, at least in part. Further analysis is needed to understand how HSCs normally develop in the harsh environment during the neonatal period.
In conclusion, we demonstrated that autophagy plays a critical role in protecting neonatal HSCs from oxidative stress in a p62-independent manner. Our findings provide novel insights regarding a critical role of autophagy in protection of HSCs against harsh environments in the early neonatal stage, which is essential for healthy long-term haematopoiesis in adults.

Materials and methods
Mice. Atg5 flox/flox (Atg5 f/f ) mice 53 59 , p62 (Abnova) 60 , LC3 (Nanotools) 61 , and β-actin (Sigma-Aldrich) 62 . Immunocomplexes were then labelled with a HRP-conjugated anti-mouse, or with anti-rabbit antibodies, and visualised using ECL Prime Western Blotting Detection Reagent (GE Healthcare) with an ImageQuant LAS 4000 (GE Healthcare). To detect different proteins in the same membrane, conjugated antibodies were removed using stripping solution (62.5 mM Tris-HCl, pH = 6.8, 2% SDS, and 0.7% β-mercaptoethanol) by gently shaking for 50 min at 50 °C. After washing with PBST, blocking and antibody reactions were carried out. Microarray and gene set analyses. BM cells of Atg5 f/f ;Vav and Atg5 w/w ;Vav mice were collected and LSK cells were stained as described above. LSK cells (> 1 × 10 4 cells) were sorted directly into TRIzol (Invitrogen). Subsequent sample preparation and analysis were carried out by Hokkaido System Science. Details are shown in a previous report 62 . Normalised expression data were examined by Gene Set Enrichment Analysis (GSEA) v4.0.1 software (Broad Institute) using the GO gene set. The number of permutations was set to 1000. Statistical analysis. Unless indicated otherwise, data are shown as the mean ± SD. Dots on bar graphs indicate values of individual mice. Statistical differences between groups were determined by the Student's unpaired t-test using Prism 5 software (GraphPad). *P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001. In survival assays, statistical differences between groups were determined by the Mantel-Cox test.