Little is known about the regulation of hypoxia on erythroid differentiation. In view of this, first of all we determine the effect of hypoxia on erythroid differentiation using K562 cells and human UCB-derived HSPCs, respectively. K562 cells were induced to differentiate into erythroid cells with hemin for 1, 2, and 3 days. The color of the pellets was first observed to preliminarily assess the ability to differentiate into erythroid cells. With the time of differentiation, the pellets gradually turned red, and compared with normoxia, hypoxia darkened the red pellets (Figure 1A), indicating that more erythroid cells were generated under hypoxia conditions. Real-time PCR and western blot analysis of the expression of CD235a and γ-globin representing the erythroid differentiation showed that hypoxia significantly increased their mRNA (CD235a and HBG) and protein expression levels (Figures 1B,C), suggesting that hypoxia can promote erythroid differentiation. K562 cells from day 1, 2, 3 of differentiation were subsequently stained with benzidine to detect hemoglobin biosynthesis that means differentiation into erythroid cells. The benzidine staining illustrated that the percentage of benzidine-positive cells increased with time, and there were more benzidine-positive cells in hypoxia than normoxia conditions at the corresponding time points (Figure 1D). Flow cytometry analysis of the expression of CD71 and CD235a, which are specific erythroid surface antigens and mainly expressed on relatively early and late erythroid cells, provided further evidence that hypoxia is more conducive to the erythroid differentiation than normoxia (Figure 1E).

In accordance with the phenotypic changes in K562 cells, in HSPCs derived from human umbilical cord blood, we observed a more powerful role of hypoxia in promoting erythroid differentiation. HSPCs were induced to differentiate into erythroid cells using the established culture system with minor modifications (Zheng et al., 2014), and the erythroid differentiation was monitored on day 4, 7, and 11. The erythroid differentiation was assessed by the expression of specific erythrocyte marker molecules. The mRNA levels of CD235a, HBG and HBB increased with the time of differentiation, and notably, their mRNA levels were distinctly higher in hypoxia than normoxia conditions (Figure 1F). Correspondingly, the protein levels of CD235a, γ-globin and β-globin also increased with the time of differentiation, and hypoxia resulted in higher expression levels than normoxia at each time point, especially on the 7th and 11th day of differentiation (Figure 1G). Similarly, the co-expression of CD235a and CD71 analyzed by flow cytometry showed that the percentage of differentiated erythroid cells was increased by hypoxia as expected (Figure 1H). In addition, using May-Grunwald Giemsa staining we found that during erythroid development, hypoxia strongly promoted the maturation of erythroid cells (Figure 1I). Together, the above evidence we provide demonstrates that hypoxia is beneficial to accelerate the process of erythroid differentiation.

Autophagy is a conserved cellular catabolism mechanism by which the degraded metabolites rapidly provide the basic building blocks needed by a cell for anabolic processes, such as cell growth, proliferation and differentiation (Levine and Klionsky, 2004; Dikic and Elazar, 2018; Allen and Baehrecke, 2020). Given that autophagy can quickly supply blocks/, we wonder whether autophagy is required for the erythroid differentiation accelerated by hypoxia. First, in K562-differentiated erythroid cells, we tested the expression of autophagy receptor protein sequestosome 1 (SQSTM1/p62) and autophagy marker microtubule-associated protein 1 light chain 3 (MAP1LC3B/LC3) via western blot analysis, and found a seemingly paradoxical phenomenon, that is, hypoxia simultaneously reduced the expression of p62 and LC3-II (Figure 2A). Generally, when the protein level of p62 decreases due to autophagy, the LC3-II protein level or the flux of LC3-II/LC3-I increases. To identify whether autophagy is enhanced by hypoxia, we used the late stage autophagy inhibitor bafilomycin A1 (Baf A1) to treat K562-differentiated erythroid cells under normoxia and hypoxia, and then observed that p62 and LC3-II proteins were markedly accumulated by Baf A1 and the accumulation became more with the time of differentiation, especially under hypoxia conditions (Figure 2B), indicating that autophagy increased with the differentiation process, and hypoxia actually enhanced autophagic flux. For ease of detection, the decrease in p62 protein level is used to indicate autophagy in K562 cells in the following experiments. During autophagy, LC3 is processed to form LC3-II, which is then recruited to the outer and inner membranes of the autophagosome. Therefore, cells undergoing autophagy can be identified by visualizing fluorescently labeled LC3 puncta. Likewise, by means of immunofluorescence staining of endogenous LC3 expression and counting the number of puncta, we noticed that the LC3 puncta were not obviously increased by hypoxia exposure, but after Baf A1 treatment, there were more LC3 puncta in hypoxia than normoxia conditions (Figure 2C). Furthermore, we used transmission electron microscopy (TEM) to compare the differences in autophagosomes between normoxia and hypoxia after blocking autophagy with Baf A1 in K562-differentiated erythroid cells. As illustrated in Figure 2D, we observed that more autophagosomes appeared under hypoxia than normoxia, indicating that hypoxia accelerates autophagic flux.

During the erythroid differentiation of HSPCs, we found that hypoxia resulted in an increase in LC3-II and a decrease in p62 through examining protein levels via western blot analysis (Figure 2E). This clearly demonstrates that autophagy increases with the erythroid differentiation and hypoxia augments autophagy in this process. Immunofluorescence analysis of LC3 expression showed that hypoxia stimulated more LC3 puncta formation than normoxia (Figure 2F). At the same time, we used TEM to compare the dynamic changes of autophagosomes in HSPCs-differentiated erythrocytes under hypoxia and normoxia. We were pleasantly surprised to find that the number of autophagosomes was obviously more in hypoxia than normoxia, and increased with differentiation time in both normoxic and hypoxic conditions (Figure 2G). The above results suggest that hypoxia can enhance autophagy which increases with erythroid differentiation.

Considering that hypoxia can accelerate erythroid differentiation and enhance autophagy, we would like to know whether autophagy plays a role in erythroid differentiation promoted by hypoxia. We first treated K562-differentiated erythroid cells with Baf A1 for 3 days, and observed that the color of pellets was lightened by Baf A1 in both normoxia and hypoxia (Figure 3A), which means that blocking autophagy hampers the differentiation into erythroid cells. Simultaneously, the mRNA levels of CD235a and HBG were evidently decreased by Baf A1 at each indicated time point of differentiation, whether under normoxia or hypoxia (Figure 3B). The γ-globin protein levels were obviously reduced on the first, second and third day of differentiation into erythroid cells in both normoxia and hypoxia conditions, after autophagy was inhibited with Baf A1 as indicated by the accumulation of p62 and LC3-II (Figure 3C). In addition, flow cytometry analysis showed that the percentage of CD71⁺/CD235a⁺ was sharply reduced by Baf A1 treatment for 3 days under these two oxygen conditions (Figure 3D). The benzidine staining provided another evidence that inhibition of autophagy with Baf A1 resulted in a remarkable decrease in the percentage of benzidine-positive cells (Figure 3E). The above data consistently illustrate that blocking the autophagy process leads to serious obstacles in erythroid differentiation, and attenuates the accelerating effect of hypoxia on erythroid differentiation. Further, the expression of autophagy-related gene 5 (ATG5) or autophagy-related gene 7 (ATG7), which is involved in the initiation of autophagy, was silenced with siRNA to detect the perturbation of erythroid differentiation. Similarly, we found that knockdown of ATG5 or ATG7 significantly obstructed the differentiation into erythroid cells. In this case, the role of hypoxia in promoting erythroid differentiation was largely abated, which was manifested by a decrease in the protein and mRNA levels of CD235a and γ-globin under normoxia, as well as a decrease to their respective normoxia levels under hypoxia (Figures 3F,G).

In primary HSPCs, we first evaluated the effect of blocking autophagy with Baf A1 on the expression of β-globin and γ-globin in protein levels on the seventh day of erythroid differentiation. As expected, we discovered that the protein expression of β-globin and γ-globin was quite obviously decreased, especially under hypoxia, when autophagy was blocked as shown by the accumulation of p62 and LC3-II (Figure 3H). At the same time, we also examined the mRNA expression of CD235a, HBG and HBB after autophagy was blocked with Baf A1, and noted that their mRNA expression levels were substantially reduced without exception, particularly in hypoxia conditions (Figure 3I), potently supporting the results at the protein level. Flow cytometry analysis of the expression of CD235a and CD71 under the same treatment conditions as above showed that the autophagy inhibitor Baf A1 decreased the population of CD71⁺/CD235a⁺ in differentiated erythroid cells, and led to a greater reduction under hypoxia (Figure 3J). By May-Grunwald Giemsa staining to assess the process of erythroid differentiation, we provided another piece of evidence that Baf A1 deferred erythroid differentiation while inhibiting autophagy, and attenuated the acceleration of erythroid differentiation caused by hypoxia (Figure 3K). Collectively, our data show that autophagy plays a pivotal role in erythroid differentiation promoted by hypoxia.

To reveal the upstream regulators that promote autophagy in erythroid differentiation under hypoxia, we performed RNA-seq analysis after K562 or CD34⁺ cells were exposed to hypoxia during erythroid differentiation. We identified several signaling pathways by gene set enrichment analysis (GSEA) via comparing hypoxia with normoxia at day 3 or Day 7 of erythroid differentiation in K562 or HSPCs, respectively (Figure 4A). The results showed that the mRNAs upregulated in the hypoxia group were significantly enriched in the HIF-1, hypoxia and glycolysis pathways, simultaneously the mRNAs downregulated in the hypoxia group were enriched in the mTOR signaling pathway, and the pathways involved in HIF-1, hypoxia and glycolysis were negatively associated with the mTOR signaling pathway, both in K562 and HSPCs. The mTOR signaling that induces autophagy is directly controlled by the key switch, mammalian target of rapamycin complex 1 (mTORC1). It has been reported that mTORC1 is necessary for mitochondrial biosynthesis in erythroid differentiation (Liu et al., 2017). However, whether mTORC1 is involved in the induction of autophagy in erythroid differentiation, especially under hypoxia, is still unclear. In this work, we examined the dynamic changes in mTORC1 signaling, autophagy activity, and erythroid differentiation under normoxia and hypoxia via western blot analysis. In K562 cells, we found that the activities of mTOR, ribosomal S6 kinase 1 (S6K), and ULK1 decreased with differentiation time, which was indicated by the decrease in their respective phosphorylation levels, suggesting that mTORC1 signaling was gradually suppressed with the progress of erythroid differentiation. In the meantime, p62 and γ-globin protein levels decreased and increased with time, respectively, illustrating that the activity of autophagy increased with the differentiation into erythroid cells. Most notably, hypoxia caused these trends in advance (Figure 4B), which may be the reason why hypoxia accelerates erythroid differentiation. In HSPCs, we observed more clear changes in mTORC1 signaling and autophagy activity in the process of erythroid differentiation, and more effective impacts of hypoxia on them (Figure 4C).

To determine the role of mTORC1 signaling in regulating autophagy and thereby controlling erythroid differentiation, we used the mTORC1 inhibitor rapamycin (Rapa) to treat K562 cells and HSPCs during their differentiation into erythroid cells under normoxia conditions. As expected, we were excited to discover that when mTORC1 signaling was inhibited, autophagy activity and erythroid differentiation were enhanced (Figures 4D,E), which mimics the effect of hypoxia to some extent. Furthermore, the mRNA levels of CD235a, HBG or HBB and the percentage of CD71⁺/CD235a⁺ in HSPCs were all augmented in a dose-dependent manner by Rapa (Figures 4F,G), suggesting that the inhibition of mTORC1 activity is conducive to promoting the process of erythroid differentiation. On the contrary, increasing mTORC1 activity by silencing TSC2, which functions as a GTPase-activating protein (GAP) to inhibit the activation of mTORC1 mediated by accelerating GTP hydrolysis of Rheb (Yang et al., 2021), during the differentiation of these two types of cells into erythrocytes under hypoxia, resulted in a decrease in autophagy activity and erythroid differentiation (Figures 4H,I). Altogether, we provide strong evidence that mTORC1 is required for the control of autophagy, and the gradually decreased mTORC1 activity by hypoxia is conducive to a progressive increase in autophagy during erythroid differentiation.

We next determined how mTORC1 is regulated by hypoxia during erythroid differentiation. Based on GSEA and heatmap analysis, we also found that genes related to the HIF-1 pathway were upregulated under hypoxia (Figure 4A, Supplementary Figure S6). Among them, REDD1 is well-known to play an essential role in suppression of mTORC1 activity in response to hypoxia (Brugarolas et al., 2004; Horak et al., 2010; Hoppe-Seyler et al., 2017; Sun and Yue, 2019; Gao et al., 2020; Shang et al., 2020). However, it has not yet been elucidated whether REDD1 induces autophagy by inhibiting mTORC1 signaling, thereby accelerating erythroid differentiation. First, to investigate the relationship between REDD1 and erythroid differentiation, we examined the changes in protein and mRNA levels of REDD1 as K562 or HSPCs were differentiated into erythrocytes under normoxia and hypoxia conditions, respectively. We noted that under normoxia, the expression levels of REDD1 protein and mRNA increased with erythroid differentiation, and compared with normoxia, hypoxia stimulated a further increase in REDD1 protein and mRNA levels, which was highly consistent with the changes in γ-globin or β-globin (Figures 5A,B), indicating that a higher level of REDD1 may be conducive to the acceleration of erythroid differentiation. Next, in order to gain insight into the role of REDD1 in regulating erythroid differentiation through the mTORC1-autophagy pathway, we silenced the expression of REDD1 with siRNA under hypoxia, and then detected the effects of REDD1 knockdown on mTORC1 signaling, autophagy and erythroid differentiation. Western blot analysis showed that the silencing of REDD1 in K562 and HSPCs reduced the protein expression of CD235a, γ-globin or β-globin while activating mTORC1 activity and decreasing autophagy activity, manifested as an increase in the phosphorylation level of mTOR, S6K, ULK1 and an increase in p62 or a decrease in LC3-II (Figures 5C,D). In addition, we further confirmed the role of REDD1 in hypoxia-promoted erythroid differentiation by silencing REDD1 in both K562 and HSPCs. We discovered that in the process of erythroid differentiation under hypoxia conditions, knockdown of REDD1 in K562 and HSPC significantly reduced the percentage of CD71⁺/CD235a⁺, as measured by flow cytometry analysis (Figure 5E). Meanwhile, the benzidine staining in hemin-induced K562 cells showed that silencing REDD1 led to an obvious decrease in the percentage of benzidine-positive cells (Figure 5F). Likewise, the knockdown of REDD1 during the differentiation of HSPCs into erythrocytes caused a serious delay in erythroid development, which was assessed by May-Grunwald Giemsa staining (Figure 5G). Overall, these data provide sufficient evidence that REDD1 plays a crucial role in erythroid differentiation, especially in the case of hypoxia accelerating erythroid differentiation, which may be achieved by suppressing the activity of mTORC1 to enhance autophagy.

Human umbilical cord blood (UCB) was obtained from donors undergoing normal full-term deliveries after informed consent in accordance with procedures approved by the Institutional Research Ethics Committee of our institute. Primary human CD34⁺ cells were obtained from magnetically-sorted mononuclear samples of UCB as previously described (Zheng et al., 2014) and were frozen down after isolation. Briefly, Mononuclear cells were prepared by HISTOPAQUE (Sigma Aldrich, 10771). CD34⁺ cells were enriched using human CD34 MicroBead Kit (Miltenyi Biotec, 130-046-702). The CD34⁺ cells were maintained, expanded and differentiated as previously described with minor modifications (Zheng et al., 2014). Briefly, cells were thawed and washed into IMDM (Sigma Aldrich, 16529) with 20% FBS and 2 mM l-Glutamine (Sigma Aldrich, G7513), and then seeded for expansion in StemSpan SFEM Medium (STEMCELL Technologies, 09650) supplied with 50 ng/ml rhSCF (PeproTech, 300-07), 100 ng/ml rhTPO (PeproTech, 300-18), 100 ng/ml rhFlt3-Ligand (PeproTech, 300-19), 50 ng/ml rhIL-3 (PeproTech, 200-03) and 100 ng/ml rhIL-6 (PeproTech, 200-06). Cells were incubated at 37°C with 5% CO₂ and maintained in the expansion medium at a density of 0.3 × 10⁶ to 1 × 10⁶ cells per milliliter with media changes every 2–3 day. Ex vivo erythroid maturation of human CD34⁺ cells was performed using a two-phase liquid culture system. On day 7, cells were centrifuged and transferred into EDM (erythroid differentiation medium) comprising IMDM medium, 3 U/ml rhEPO (PeproTech, 100-64), 2 mM l-Glutamine, 2% PS, 330 μg/ml human holo-Transferrin (Sigma Aldrich, T4132), 10 μg/ml human insulin (Sigma Aldrich, I9278), 2 U/ml heparin (STEMCELL Technologies, 07980) and 5% inactivated human serum (Sigma Aldrich, H3667) supplemented with 100 ng/ml rhSCF, 5ng/ml rhIL-3 and 1 μM dexamethasone (Sigma Aldrich, D4902). On day 14, cells were centrifuged and transferred into EDM supplemented with 100 ng/ml rhSCF. Cells were maintained in differentiation medium for indicated times at a density of 0.3–1 × 10⁶ cells/ml by supplementing cultures every few days with fresh media. K562 cells were maintained in RPMI 1640 Medium containing 10% fetal bovine serum (FBS), and were induced to differentiate into erythroid cells with the supplement of Hemin (30 μmol/L).

For hypoxia treatment, cells were incubated in an O₂/N₂/CO₂ incubator (Thermo Fisher Scientific, Marietta, OH, United States; 3131) and exposed to 3% O₂ for the indicated time periods. For reagent treatment, DMSO or particular chemicals were added to the differentiation medium, which were used as the control group and the experimental group, respectively. Baf A1 (HY-100558) and Rapa (HY-10219) were purchased from MedChemExpress (Monmouth Junction).

Transfection was performed 30 min after cells were plated at a density of 3×10⁵ cells/ml. Negative control (si-NC) and siRNA targeting the indicated mRNA (Supplementary Table S1) were purchased from RiboBio (Guangzhou, China). Cells were transiently transfected with si-NC or si-ATG5/si-ATG7/REDD1/TSC2 at a final concentration of 100 nM using Lipofectamine 2000 (Thermo Fisher Scientific, 11668-019) according to the manufacturer’s instructions. For K562 cells, the relevant analysis was performed on the third day after transfection that was conducted on the first day of differentiation; For HSPCs, the relevant analysis was performed on the seventh day after transfection that was conducted on the fourth day of differentiation.

For staining, single-cell suspensions were incubated on ice with indicated antibodies in PBS. Cells were run on FACSAria II Cell Sorter (BD Biosciences) and analyzed using Flow Jo VX. FITC Mouse Anti-human CD71 (BD Biosciences, 555536) and anti-human CD235a APC (Thermo Fisher Scientific, 17-9987-41) were used for cell surface staining.

Benzidine staining assay was performed as previously described (Fibach and Prus, 2005). Briefly, benzidine stock solution containing 0.5 M glacial acetic acid and 0.4% benzidine dihydrochloride (Sigma Aldrich, B3383) was mixed with PBS and 33% H₂O₂ at a 25:25:1 (v/v/v) ratio to prepare working solution. Cells were resuspended in PBS and incubated with benzidine working solution at a 100:1 (v/v) ratio at room temperature. After 3–5 min, the cells were examined using a light microscope (Nikon, ECLIPSE, Ti2-U). The benzidine-positive cells were blue.

Giemsa stain assay was performed as previously described (Ma et al., 2020). Briefly, the cells were collected by centrifugation and resuspended in PBS at different times of differentiation. Then cells were cytocentrifuged onto glass slides and stained with Wright Giemsa Stain Kit (Solarbio, Beijing, China; G1021) at room temperature. Stained cells from each group were examined under a light microscope (Nikon, ECLIPSE, Ci-L).

For Immunofluorescence analysis, cells were centrifuged onto glass slides and fixed with 4% paraformaldehyde (PFA) in PBS for 10 min. Then, cells were incubated with 0.3% Triton X-100 in PBS for 10 min and blocked with 2% bovine serum albumin in PBS for 1 h, followed by overnight incubation at 4°C with primary antibodies (Supplementary Table S2). On the second day, cells were washed 3 times in PBS, 5 min each time, and corresponding secondary antibodies (Supplementary Table S2) were added and incubated for 1 h at room temperature, followed by incubation with Hoechst 33342 (1:1000; Solarbio, C0031) for 5 min at room temperature and rinsing 3 times before mounting on glass slides with VECTASHIELD antifade mounting medium (Wolcavi, H-1200). Images were captured by an inverted confocal microscope (UltraVIEW VOX, PerkinElmer, Waltham, Massachusetts, United States) using CY3/DAPI filters at 60X objective, and fluorescence quantification was performed using ImageJ.

For electron microscopy evaluation, cells were collected and washed with PBS after cells were exposed to normoxia or hypoxia for indicated times, and then fixed with 3% glutaraldehyde in 0.1 M PBS at 4°C for 2 h. The subsequent steps were performed as previously described (Verheije et al., 2008). TEM analysis was performed by using a transmission electron microscope (H-7650; HITACHI, Tokyo, Japan) at 80 kV.

Total RNA was extracted from cells using RNApure Tissue/Cell Kit (DNase Ⅰ) (CWBIO, CW0560S). Next, 500 ng of total RNA was reverse transcribed into cDNA using the MonScript RTⅢ All-in-One Mix (BioPro, MR05001) according to the manufacturer’s instructions. RT-qPCR was performed with KAPA SYBR^® FAST (Roche, KK4601) under the following conditions: 5 min 20 s at 95°C, followed by 40 cycles (5 s at 95°C and 30 s at 60°C). Primers for qPCR were synthesized by Tianyi Huiyuan (Beijing, China) and listed in Supplementary Table S3. Each reaction was repeated in at least three independent experiments. The 2^−ΔΔCt method to calculate the relative expression levels (fold change) was used. Amplification of the housekeeping gene ACTB was used to normalize the amount of cDNA.

Total protein was extracted from cells with RIPA Lysis Buffer (CWBIO, CW2333) containing a protease inhibitor cocktail (Roche, 04,693,132,001). Equal amounts of protein were separated on 8%–12% SDS-PAGE gels and transferred to PVDF membranes (Millipore, ISEQ00010), which were blocked with 5% skimmed milk for 1 h and incubated with primary antibodies (Supplementary Table S2) overnight at 4°C. This was followed by incubation with appropriate secondary antibodies (Supplementary Table S2) for 2 h at room temperature. Next, the protein bands were visualized using SuperSignal West Pico PLUS (Thermo Fisher Scientific, 34580). β-actin was used as the loading control. The quantitation of western blot is shown in supplementary figures (Supplementary Figure S1-S5).

Data were represented as mean ± standard error of the mean of at least three independent experiments. Statistical significance differences were carried out using the unpaired two-sided Student’s t test, one-way analysis of variance (ANOVA) followed by Dunnet’s post hoc test and two-way analysis of variance (ANOVA) followed by sidak’s post hoc test, using the Microsoft Excel or GraphPad Prism 8.0.2. The levels of significance were indicated by ns, not significant; *p < 0.05, **p < 0.01, and ***p < 0.001. p < 0.05 was considered statistically significant.