Skip to main content
eLife logoLink to eLife
. 2019 Apr 29;8:e46976. doi: 10.7554/eLife.46976

HRI coordinates translation necessary for protein homeostasis and mitochondrial function in erythropoiesis

Shuping Zhang 1, Alejandra Macias-Garcia 1, Jacob C Ulirsch 2,3,4,5, Jason Velazquez 1, Vincent L Butty 6, Stuart S Levine 6, Vijay G Sankaran 2,3,4, Jane-Jane Chen 1,
Editors: Jian Xu7, David Ron8
PMCID: PMC6533081  PMID: 31033440

Abstract

Iron and heme play central roles in the production of red blood cells, but the underlying mechanisms remain incompletely understood. Heme-regulated eIF2α kinase (HRI) controls translation by phosphorylating eIF2α. Here, we investigate the global impact of iron, heme, and HRI on protein translation in vivo in murine primary erythroblasts using ribosome profiling. We validate the known role of HRI-mediated translational stimulation of integratedstressresponse mRNAs during iron deficiency in vivo. Moreover, we find that the translation of mRNAs encoding cytosolic and mitochondrial ribosomal proteins is substantially repressed by HRI during iron deficiency, causing a decrease in cytosolic and mitochondrial protein synthesis. The absence of HRI during iron deficiency elicits a prominent cytoplasmic unfolded protein response and impairs mitochondrial respiration. Importantly, ATF4 target genes are activated during iron deficiency to maintain mitochondrial function and to enable erythroid differentiation. We further identify GRB10 as a previously unappreciated regulator of terminal erythropoiesis.

Research organism: Mouse

eLife digest

Red blood cells use a molecule called hemoglobin to transport oxygen around the body. To make hemoglobin, cells require iron to build a component called heme. If an individual does not get enough iron in their diet, the body cannot produce enough red blood cells, or the cells lack hemoglobin. This condition is known as iron deficiency anemia, and it affects around one-third of the world’s population.

Researchers did not know exactly how iron levels control red blood cell production, though several proteins had been identified to play important roles. Heme forms in the cell's mitochondria: the compartments in the cell that supply it with energy. When heme levels in a developing red blood cell are low, a protein called HRI reduces the production of many proteins, most importantly the proteins that make up hemoglobin. HRI also boosts the production of a protein called ATF4, which switches on a set of genes that help both the cell and its mitochondria to adapt to the lack of heme. In turn, HRI and ATF4 reduce the activity of a signaling pathway called mTORC1, which controls the production of proteins that help cells to grow and divide.

To understand in more detail how iron and heme regulate the production of new red blood cells, Zhang et al. looked at immature red blood cells from the livers of developing mice. Some of the mice lacked the gene that produces HRI, and some experienced iron deficiency. Comparing gene activity in the different mice revealed that in the developing blood cells of iron-deficient mice, HRI largely reduces the rate of protein production in both the mitochondria and the wider cell. At the same time, the increased activity of ATF4 allows the mitochondria to carry on releasing energy and the cells to continue developing. Zhang et al. also found that a protein that inhibits the mTORC1 signaling pathway needs to be active for the new red blood cells to mature.

Overall, the results suggest that drugs that activate HRI or block the activity of the mTORC1 pathway could help to treat anemia. The next step is to test the effects that such drugs have in anemic mice and cells from anemic people.

Introduction

Iron-deficiency anemia is estimated to affect one-third of the global population (Camaschella, 2019). Iron and heme are key components of hemoglobin, the primary oxygen transport molecule, and their cellular levels impact globin synthesis and red blood cell production. Specifically, globin is transcriptionally regulated by BACH1 (BTB Domain and CNC homolog 1) (Igarashi and Sun, 2006) and is regulated at the level of protein translation by HRI (heme-regulated eIF2α kinase) (Chen, 2007), both of which are heme-sensing proteins.

During heme deficiency that is induced by dietary iron deficiency (ID) in mice, HRI is activated and phosphorylates the α subunit of eukaryotic initiation factor eIF2 (eIF2α) to inhibit translation of α- and β-globin mRNAs, thereby preventing proteotoxicity resulting from heme-free globin chains (Han et al., 2001). Meanwhile, phosphorylated eIF2α (eIF2αP) selectively enhances the translation of activating transcription factor 4 (Atf4) mRNA (Suragani et al., 2012; Chen, 2014). This coordinated translational repression of general protein synthesis with the specific translational enhancement of Atf4 mRNA by eIF2αP is termed the integrated stress response (ISR) (Harding et al., 2003). The ISR is a universal response to several types of cellular stress (Chen, 2014; Pakos-Zebrucka et al., 2016) and it is initiated by members of the eIF2α kinase family. Besides HRI, mammalian cells have three additional eIF2α kinases, which are expressed in distinct tissues and which combat specific physiological stress. PKR responds to viral infection (Kaufman, 2000), GCN2 senses nutrient starvations (Hinnebusch, 1996), and PERK is activated by endoplasmic reticulum (ER) stress (Ron and Harding, 2000). All four eIF2α kinases respond to oxidative and environmental stresses.

In the erythroid lineage, HRI expression increases during differentiation, with higher expression occurring in hemoglobinized erythroblasts (Liu et al., 2008a). Starting at the basophilic erythroblast stage, HRI is the predominant eIF2α kinase and is expressed at levels that are two orders of magnitude greater than those of the other three eIF2α kinases (Kingsley et al., 2013). At these stages, HRI is responsible for over 90% of eIF2α phosphorylation (Liu et al., 2008b). HRI-ISR is necessary for effective erythropoiesis during ID and acts by reducing oxidative stress and promoting erythroid differentiation (Suragani et al., 2012; Zhang et al., 2018). Furthermore, HRI-ISR represses the mTORC1 signaling that is activated by the elevated erythropoietin (Epo) levels occurring during ID specifically in the erythroid lineage (Zhang et al., 2018). Thus, HRI coordinates two key translation-regulation pathways, eIF2αP and mTORC1, during ID. However, the exact molecular mechanisms through which iron and heme regulate erythropoiesis are incompletely understood.

Mitochondria not only are the energy powerhouses of the cell, but also are necessary for amino acid metabolism, nucleotide production, and the biosynthesis of heme and iron-sulfur clusters (Shpilka and Haynes, 2018). Translational regulation of mitochondrial biogenesis by mTORC1 is particularly important for erythropoiesis because of the high demand of heme for hemoglobin production and oxidative stress (Liu et al., 2017). However, the roles of HRI and eIF2αP in mitochondrial biogenesis and function are still unknown.

Transcriptional regulation during erythropoiesis has been studied extensively (Kerenyi and Orkin, 2010; An et al., 2014), but much less is known about the translational control of this process (Mills et al., 2016; Khajuria et al., 2018). Ribosome profiling (Ribo-seq) has emerged as a powerful tool that can be used to interrogate translation genome-wide (Ingolia et al., 2009). Here, we performed Ribo-seq and mRNA-seq in primary basophilic erythroblasts to investigate how in vivo translation is regulated by iron, heme, and HRI in order to gain a global understanding of the molecular mechanisms that govern erythropoiesis. We hypothesized that by globally surveying the landscape of translational and concomitant transcriptional changes that occur in the context of HRI deficiency (comparing the changes seen in iron replete (+Fe) or iron deficiency (–Fe) conditions), we could gain important insights into the mechanisms through which iron and heme regulate the process of erythropoiesis. Our results demonstrate that heme and HRI mediate the translation of both cytosolic and mitochondrial ribosomal protein mRNAs. Furthermore, HRI–ATF4 mediated gene expression is essential during ID to prevent the accumulation of unfolded proteins in the cytoplasm, to maintain mitochondrial oxidative phosphorylation, and to enable erythroid differentiation in developing basophilic erythroblasts.

Results

Overview of Ribo-seq and mRNA-seq data

Beginning at the basophilic erythroblast stage, erythropoiesis is finely regulated by iron and heme levels (Chiabrando et al., 2014; Muckenthaler et al., 2017). Thus, basophilic erythroblasts (hereafter referred to as EBs for simplicity) from Wt +Fe, Wt –Fe, Hri–/– +Fe, and Hri–/– –Fe fetal livers (FLs) were used as sources to generate Ribo-seq and mRNA-seq libraries for genome-wide analysis of transcriptional and translational changes (Figure 1A).

Figure 1. Overview of Ribo-seq and mRNA-seq data.

Figure 1.

(A) Illustration of experimental designs. Basophilic erythroblasts (EBs) (S3) from E14.5 FLs of Wt +Fe, Wt –Fe, Hri–/– +Fe and Hri–/– –Fe embryos were sorted and subjected to Ribo-seq and mRNA-seq library preparations. (B) Distribution of the mapped reads from Ribo-seq and mRNA-seq from one replica. (C) A representative plot of the triplet periodicity of Ribo-seq from Wt –Fe EBs. Arrows indicate the start and stop codons. (D) Gene coverages of Ribo-seq and mRNA-seq data in the mouse genome (UCSC, mm10). (E) Scatter plot and correlation analysis of log2-transformed TPM (transcript per million) of Ribo-seq and mRNA-seq data from Wt +Fe EBs. E, Erythroblast; RBC, red blood cell; Retic, Reticulocyte.

After applying standard protocols of quality control and preprocessing to remove rRNA and tRNA, we obtained 9.7–41.1 (median 26.2) million reads of Ribo-seq and 29.4–66.6 (median 42.5) million reads of mRNA-seq for subsequent mappings (Supplementary file 1a). As expected, most of the reads from both Ribo-seq (84–88%) and mRNA-seq (72–74%) were mapped to protein coding sequence (CDS), whereas some reads mapped to the 5′ and 3′ UTR or to other regions, mostly introns and the regions around transcription start sites (Figure 1B).

Our Ribo-seq data displayed excellent triplet periodicity, CDS enrichment, and limited 3′ UTR reads, validating the high quality of the ribosome-protected fragments (RPFs) that we obtained (Figure 1B–C). The proportions of reads mapping to the 5′ UTRs of Ribo-seq data were similar to those from mRNA-seq (Figure 1B), in agreement with reports of pervasive translation outside of annotated CDSs (Ingolia, 2014). Overall, 62.9% of the expressed genes in mRNA-seq were detected in Ribo-seq, and therefore appeared to be actively translated in EBs (Figure 1D). As shown in Figure 1D, 99% of the mRNAs detected in Ribo-seq data (with reads greater than 25 in at least one of the conditions) were present in the mRNA-seq data, demonstrating the high quality of our Ribo-seq data and the correlation of RPFs and mRNAs (Figure 1E).

Upregulation of the in vivo translation of ISR mRNAs in WT EBs compared to HRI–/– EBs

A total of 317 mRNAs were significantly differentially translated between Wt and Hri–/– EBs in the +Fe condition (Figure 2A and Supplementary file 1b), supporting the role of HRI during normal fetal erythropoiesis. Twenty-five differentially translated mRNAs were common under both +Fe and –Fe conditions; these included the well-characterized ISR mRNAs, Atf4, Ppp1r15a, and Ddit3 (Figure 2A–B and Figure 2—figure supplement 1A) as illustrated in Figure 8. Atf4 mRNA was the most differentially translated mRNA between Wt and Hri–/– cells during ID (8.8-fold increase in translational efficiency (TE)), followed by the Ddit3 (8.1-fold) and Ppp1r15a (3.7-fold) mRNAs (Figure 2B and Supplementary file 1b). Each of these mRNAs contains upstream open reading frames (uORFs) in their 5′ UTR, the use of which is upregulated by eIF2αP in cell lines under endoplasmic reticulum (ER) tostress or amino-acid starvation (Pavitt and Ron, 2012). We observed HRI-dependent regulation of the translation of these mRNAs involving ribosome occupancies in uORFs in vivo (Figure 2C–D and Figure 2—figure supplement 1B–C). We also verified that changes in TE corresponded to concordant changes in ATF4 protein levels in EBs (Figure 2E).

Figure 2. Differentially translated mRNAs in Hri–/– EBs and iron deficiency.

(A) The mRNAs that are significantly differentially translated between Hri–/– and Wt EBs in +Fe or –Fe conditions. (B) Volcano plot of mRNAs that are differentially translated between Hri/– –Fe and Wt –Fe EBs. Red dots on the positive end of the X-axis indicate significantly differentially translated mRNAs that were upregulated in Hri–/– –Fe EBs, whereas red dots on the negative end of the X-axis indicate significantly differentially translated mRNAs that were upregulated in Wt –Fe EBs. Green and gray dots indicate mRNAs that have no significant difference in translation. TE, translational efficiency. (C) Ribosome occupancies, as visualized using Integrative Genomics Viewer (IGV), of Atf4 mRNA, with an enlarged view shown in (D). (E) ATF4 protein expression in E14.5 FL cells. Ratios of ATF4/β-actin expression are indicated.

Figure 2.

Figure 2—figure supplement 1. Translational regulation of ISR mRNAs by HRI.

Figure 2—figure supplement 1.

(A) Increased translation of ISR mRNAs in Wt +Fe EBs compared to Hri–/– +Fe EBs. Red dots on the positive end of the X-axis indicate significantly differentially translated mRNAs that are upregulated in Hri–/– +Fe EBs compared to Wt +Fe EBs, whereas red dots on the negative end of the X-axis indicate mRNAs that are upregulated in Wt +Fe EBs compared to Hri–/– +Fe EBs. Green and gray dots indicate mRNAs that have no significant differences in translation. Genes labeled in blue indicate ISR mRNAs. TE, translational efficiency. (B–C) IGV-illustration of ribosome occupancies of the Ppp1r15a and Ddit3 mRNAs. Mapped reads of mRNA-seq data from Wt +Fe EBs only are shown because no significant change in mRNA levels was observed among four samples.

As Eif2ak1 (HRI) and Atf4 are among the most highly expressed and efficiently translated mRNAs in Wt EBs (top 3%, Supplementary file 1c), we investigated whether the translation of Atf4 mRNA is regulated by HRI in vivo. Atf4 mRNA has two well-characterized uORFs in its 5′ UTR. uORF1 encodes three amino acids and is translated regardless of stress and eIF2αP levels (Pavitt and Ron, 2012). We observed that uORF1 had exceptionally high ribosome occupancy (5.7-fold and 21.3-fold higher than eIF2s1 (eIF2α) and Rps6 initiating AUGs, respectively (Supplementary file 1d)). However, uORF2 and the canonical ORF of Atf4 were poorly translated in Hri–/-– compared to Wt EBs under both +Fe and –Fe conditions (Figure 2C–D). Together, these data support the idea that Atf4 mRNA is primed for translation by HRI in developing EBs, and that HRI is a master translational regulator of key ISR mRNAs in vivo in primary EBs, especially during ID.

Upregulation of translation of cytosolic and mitochondrial ribosomal proteins in HRI–/– –Fe EBs compared to Wt –Fe EBs

Gene Ontology (GO) analysis of the differentially translated mRNAs revealed that the translation of components of ribosomes and mitochondria is most significantly upregulated in HRI- and iron-deficient states (Figure 3A). Gene Sets Enrichment Analysis (GSEA) of differentially translated mRNAs further showed that translation of mRNAs that are involved in ribosome synthesis, mTORC1 signaling, and translation initiation and elongation were most highly elevated in Hri–/– –Fe EBs compared to Wt –Fe EBs (Figure 3B–C). Increased 43S translation initiation complex formation in Hri–/– –Fe cells (Figure 3B) is consistent with the known function of HRI–eIF2αP in ternary and 43S complexes formation (Chen, 2007). Furthermore, we observed that Eef1a1 and Ccnd3, which have been shown to be regulated by mTORC1 signaling (Thoreen et al., 2012), were also more efficiently translated in Hri–/– –Fe EBs (Figure 2B). Notably, 56 ribosomal protein (RP) mRNAs were more highly translated in Hri–/– –Fe EBs than in Wt-Fe EBs (Figure 3D, Figure 3—figure supplement 1A–B and Supplementary file 2a). Whereas 38 of these RP mRNAs were known to be regulated by mTORC1 via 5′ terminal oligopyrimidine (TOP) motifs in their 5′ UTR (Thoreen et al., 2012), translation of the other 18 RP mRNAs was regulated by HRI, but not mTORC1. There was less overlap for non-RP targets between mTORC1- and HRI-mediated translation. Among the 20 non-RP mRNAs, translation of 12 mRNA was regulated by HRI but not by mTORC1 (Figure 3D and Supplementary file 2a).

Figure 3. Analyses of the mRNAs that are differentially translated between Wt and Hri–/– EBs.

(A) GO analysis of the most highly translated mRNAs in Hri-/– –Fe EBs compared to Wt –Fe EBs. (B) Increased 43S initiation complex, ribosomal protein (RP) synthesis and mitochondrial pathways in Hri–/– –Fe EBs compared to Wt –Fe EBs by Gene Set Enrichment Analysis. (C) Increased mTORC1 signaling pathway in Hri–/– –Fe EBs compared to Wt –Fe EBs. (D) Venn diagrams comparing mRNAs that are differentially translated in Hri–/– –Fe EBs and Wt –Fe EBs, and that are known mTORC1 targets in the categories of RPs, non-RPs and mitochondrial proteins. (E) Protein synthesis (total (left) and in mitochondria (right)) of erythroid cells from the bone marrow (BM) of Wt –Fe and Hri–/– –Fe mice. Cells were treated with dimethyl sulfoxide (DMSO) or cycloheximide (CHX) for 15 min before the addition of puromycin for 15–60 min as indicated. Protein synthesis was determined by the nascent polypeptides covalently linked by puromycin using anti-puromycin antibody. Equal numbers of nucleated cells were loaded to each lane. The numbers of cells loaded for cytoplasmic protein synthesis were 30% of those loaded for mitochondrial protein synthesis. The results shown are from the same exposure time for developing the Western blot. Puromycin signals from polypeptides of the entire lane were quantified for protein synthesis activity and the protein synthesis in the first lane was define as 1. –Puro indicates cells without puromycin treatment, which were used as a negative control for Western signals from anti-puromycin antibody. FDR, false discovery rate; NES, normalized enrichment score.

Figure 3.

Figure 3—figure supplement 1. Differential translation of cytoplasmic and mitochondrial RP mRNAs between Wt and Hri–/– EBs.

Figure 3—figure supplement 1.

(A) The heatmaps of significantly differentially translated 5’ TOP/TOP-like mRNAs, the mTORC1 translational targets. Positive values of Log2(foldchange of TE) indicate upregulated translation in Hri–/– –Fe EBs as compared to Wt –Fe EBs or in Hri–/– +Fe EBs as compared to Wt +Fe EBs. (B) IGV-illustration of ribosome occupancies of two representative cytoplasmic RP mRNAs, Rplp1 and Rps21, and (C) of the mitochondrial RP mRNAs, Mrpl4 and Mrpl53.
Figure 3—figure supplement 2. Inhibition of mitochondrial protein synthesis by chloramphenicol.

Figure 3—figure supplement 2.

Protein Synthesis assays were performed in DMSO-treated, CHL-treated, CHX-treated and CHX + CHL-treated erythroid cells from the BM of Wt –Fe and Hri–/– –Fe mice. The DMSO-treated control is used to measure total protein synthesis in the cells, cytoplasm and mitochondria. In the presence of CHX, a cytoplasmic protein synthesis inhibitor, only mitochondrial protein synthesis was measured. In the presence of chloramphenicol (CHL), a mitochondrial protein synthesis inhibitor, only cytoplasmic protein synthesis was measured. In the presence of both CHX and CHL, protein synthesis in both the cytoplasm and mitochondria was inhibited. Cells were treated with DMSO and the inhibitors for 15 min before addition of puromycin for 15–60 min as indicated. Protein synthesis was determined by the nascent polypeptides covalently linked by puromycin using anti-puromycin antibody. Equal numbers of nucleated cells were loaded to each lane. The numbers of cells loaded for cytoplasmic protein synthesis were 30% of those loaded for mitochondrial protein synthesis. The results shown are from the same exposure time for developing the Western blots. –Puro indicates cells without puromycin treatment, which were used as a negative control for Western signals from the anti-puromycin antibody.
Figure 3—figure supplement 3. Inhibition of mitochondrial protein synthesis by INK128.

Figure 3—figure supplement 3.

(A) Protein synthesis in the CHX-treated and CHX + INK128-treated erythroid cells from the BM of Wt –Fe and Hri–/– –Fe mice. In the presence of CHX, only mitochondrial protein synthesis was measured. In the presence of both CHX and INK128, a mTORC1 inhibitor, mTORC1-regulated mitochondrial protein synthesis was inhibited. Cells were treated with inhibitors for 15 min before the addition of puromycin for 15–60 min as indicated. Measurement of protein synthesis was as described in Figure 3—figure supplement 2. (B) Inhibition of mTORC1 activities as measured by pS6 and p4EBP1 levels. Signal from Wt –Fe with CHX treatment and 15 min treatment of puromycin was defined as 1. Equal numbers of nucleated cells were loaded into each lane. The results shown are from the same exposure time for developing the Western blot.

The second set of most preferentially translated mRNAs were mitochondrial proteins (Figure 3A–B). The translation of 163 mRNAs of nuclear-encoded mitochondrial proteins was upregulated in Hri–/– –Fe EBs compared to Wt –Fe EBs. These include mitochondrial ribosomal proteins, oxidative phosphorylation proteins (comprising complexes I through V), and matrix proteins (Figure 3—figure supplement 1C and Supplementary file 2b). Interestingly, this set of mRNAs encoding mitochondrial proteins that are differentially translated in HRI deficiency was distinct from those encoding mitochondrial proteins that are upregulated by mTORC1/eIF4EBPs-mediated translation (Morita et al., 2013) (Figure 3D). Overall, these genome-wide results indicate roles for both HRI–eIF2αP and mTORC1 signaling in the global translation of ribosomal and mitochondrial proteins in developing EBs during ID.

Increased cytoplasmic and mitochondrial protein synthesis but reduced mitochondrial respiratory activity in HRI–/– –Fe EBs

To examine the impact of increased translation of ribosomal proteins (both cytosolic and mitochondrial), protein synthesis in the cytoplasm and mitochondria were measured in the primary erythroid cells isolated from bone marrows of both Wt –Fe and Hri–/– –Fe mice. We took advantage of the fact that cycloheximide (CHX) inhibits only cytoplasmic and not mitochondrial protein synthesis. As shown in Figure 3E, the rate of total protein synthesis in Hri–/– –Fe erythroid cells at 15–60 mins was greater than that in Wt –Fe cells. Interestingly, the rate of protein synthesis in cells under CHX treatment was also greater in Hri–/– –Fe erythroid cells than in Wt-Fe cells. Furthermore, protein synthesis under CHX treatment was inhibited by chloramphenicol (Figure 3—figure supplement 2), a selective inhibitor of mitochondrial translation (Richter et al., 2013). These results demonstrate that HRI is necessary to inhibit both cytoplasmic and mitochondrial protein synthesis in iron-deficient erythroid cells. To discern the contributions of HRI–eIF2αP and mTORC1 activity in the regulation of mitochondrial protein synthesis, INK128, a mTOR inhibitor, was employed in mitochondrial protein synthesis assays (Figure 3—figure supplement 3). INK128 treatment reduced mitochondrial protein synthesis by about 50% with a near complete inhibition of pS6 and p4EBP1. These results show that the HRI–eIF2αP pathway contributes directly to the regulation of mitochondrial protein synthesis during ID, in addition to repressing mTORC1 activity.

We then asked whether the increased protein synthesis in Hri–/– –Fe erythroid cells affects the activity and mass of mitochondria in these cells when compared with Wt –Fe cells. As iron and heme are cofactors for several enzymes in the respiratory electron transport chain, the oxygen consumption rates of primary erythroid cells from bone marrows of +Fe and –Fe mice were measured using the Seahorse assays (Figure 4). Wt erythroid cells were able to maintain their oxygen consumption rate in ID (Figure 4A). There were no significant differences between Wt –Fe and Wt +Fe cells in either the basal or maximal respiration rates (Figure 4B). By contrast, Hri–/– erythroid cells displayed decreased basal and maximal respiration under both +Fe and –Fe conditions, as compared to Wt erythroid cells (Figure 4A–B). The similar extent to which respiration was reduced in Hri–/– +Fe cells and Hri–/– –Fe cells was somewhat unexpected because the phenotype of Hri–/–+Fe in vivo is milder than that of Hri–/– –Fe cells. This may be due in part to the stress experienced by Hri–/– +Fe cells during the Seahorse assays. Nonetheless, the significant difference between Wt -–Fe and Hri–/– –Fe cells demonstrates that HRI is necessary to maintain mitochondrial respiration under –Fe conditions. This impaired mitochondrial respiration in Hri–/– erythroid cells was not due to the reduced mitochondrial mass in these cells (Figure 4C). The mitochondrial mass in primary erythroid cells was similar for Wt +Fe and Hri–/– +Fe mice, but was increased in Hri–/– –Fe cells as compared to Wt –Fe cells, especially during the later stages of terminal erythroid differentiation (Figure 4C, populations II and III). There was no significant difference in mitochondrial DNA contents between Wt and Hri–/– erythroid cells in either +Fe or –Fe conditions (Figure 4—figure supplement 1).

Figure 4. Decreased mitochondrial respiration in HRI deficiency.

(A) A representative result for the oxygen consumption rate (OCR) of Wt and Hri–/– erythroid cells. Erythroid cells were isolated from the BM of Wt and Hri-–/– mice in +Fe and –Fe conditions. Five technical replicas were performed for each condition. (B) Quantitative analysis of basal and maximal OCR from four separate experiments. The numbers of mice used in each condition are indicated. OCR of Wt –Fe erythroid cells was defined as 1. Data were presented as mean ± SE. **p<0.01, ***p<0.001. (C) Representative plots of mitochondrial mass of erythroid cells at different differentiation stages obtained from mitotracker flow cytometry analysis. A diagram of gating for the differentiation stages is shown on the left.

Figure 4.

Figure 4—figure supplement 1. The relative mitochondrial DNA contents of erythroid cells.

Figure 4—figure supplement 1.

The ratios of mitochondrial DNA (mtDNA) versus genomic DNA (gDNA) were determined by qPCR using the ΔΔCt method for erythroid cells isolated from bone marrow samples. The level in Wt +Fe was defined as 1 (n = 3). Data are presented as mean ± SE.

Cytoplasmic unfolded protein response in Hri–/– –Fe cells

We next investigated the transcriptional impact of ID and the role of HRI in mediating the cellular response to the ID-induced stress response. Analysis of mRNA-seq data revealed that substantially more genes displayed significant differential expression between Wt –Fe and Wt +Fe EBs than between Hri–/– –Fe and Hri–/– +Fe EBs (232 vs 37, Figure 5A and Supplementary file 1e), demonstrating the near-absolute requirement for HRI in regulating the transcriptional response to ID. GO analysis revealed that the most upregulated process in Hri–/– -Fe EBs when compared to Wt –Fe EBs is the process of protein folding and refolding, which involves many chaperone genes from the Hsp70 (Hspa1a, Bag3, Hspa1b, Hspa8, HspA4L), Hsp90 (Hsp90ab1, Hsp90aa1, Chordc1, Ahsa1), Hsp40 (Dnaja4, Dnajb1), and Hsp110 (Hsph1) families (Figure 5A–B and Figure 6B). These are cytosolic chaperones that are upregulated specifically in response to cytoplasmic unfolded proteins. Chaperones involved in neither ER (Hspa5, Hsp90b1) nor mitochondrial (Hspa9, Hspd1) unfolded-protein responses (UPR) were not significantly increased. This comprehensive upregulation of cytoplasmic chaperones indicates the response to cytoplasmic unfolded proteins, such as heme-free globins and nuclear-encoded mitochondrial proteins, in Hri–/– –Fe EBs. Cytoplasmic chaperones were not induced in the presence of HRI during ID (Figure 5A and Figure 6A), consistent with the essential role of HRI in inhibiting protein synthesis in iron/heme deficiency shown in Figure 3E.

Figure 5. Increased expression of cytoplasmic chaperones, accumulation of unfolded protein inclusions and impaired FL differentiation in Hri–/– erythroid cells.

Figure 5.

(A) Venn diagrams depicting the numbers of mRNAs that are significantly differentially expressed between –Fe and +Fe conditions in Wt or Hri–/– EBs. The table on the right lists mRNAs that were expressed at higher level only in Hri–/––Fe EBs. Results were obtained from three biological replicas. (B) GO terms related to protein folding which were most highly upregulated in Hri–/– –Fe EBs compared to Wt –Fe EBs from the Gene ontology analysis of significantly differentially expressed mRNAs. (C–D) Ex vivo erythroid differentiation from HRI-, eIF2αP-, and ATF4-deficient FL erythroid progenitors. The representative images are of cytospin slides stained with May-Grunwald/Giemsa staining. Cells at 20, 30 and 42 hr of erythroid differentiation are shown. AA, universal eIF2α Ala51/Ala51 knockin resulting in complete ablation of eIF2αP. Arrow in (C) indicates globin inclusions and arrowhead in (D) indicates myeloid cells. Numbers of FL differentiation performed were: n = 6 for Wt and Hri–/–; n = 4 for AA and n = 3 for Atf4–/. Some of the cell images in (C) were cropped from the same slides and compiled together for illustrative purposes.

Figure 6. Analyses of differentially expressed mRNAs in iron and HRI deficiencies.

(A) Volcano plots of mRNAs that are differentially expressed between Wt –Fe and Wt +Fe or (B) between Hri–/– –Fe and Wt –Fe EBs. Red dots represent the significantly differentially expressed mRNAs. Green and gray dots indicate mRNAs whose expression does not differ significantly. ATF4 target genes are labeled in blue. (C) GO terms for amino acid metabolism that are most highly upregulated in Wt –Fe EBs compared to Hri–/– –Fe EBs.

Figure 6.

Figure 6—figure supplement 1. Analyses of differentially expressed ISR-target mRNAs.

Figure 6—figure supplement 1.

(A) Heatmaps of ISR-target genes that have a significant difference in expression levels between Hri–/– –Fe and Wt –Fe EBs or between Wt –Fe and Wt +Fe EBs. (B) Expression of Shmt2 in E14.5 FL cells. The expression level in Wt –Fe was defined as 1 (n = 3). Data were presented as mean ± SE. **p<0.01. (C) The heatmaps of ISR-target genes that are differentially expressed between Hri–/– –Fe and Wt –Fe EBs or between Hri–/– +Fe and Wt +Fe EBs. (D) Epo-stimulated ISR-target genes that are differentially expressed between Hri–/– -Fe and Wt –Fe EBs or between Wt –Fe and Wt +Fe EBs.

We have shown previously that HRI-ISR is activated in an ex vivo FL differentiation system and that erythroid differentiation of Hri–/– progenitors is impaired even under +Fe conditions (Suragani et al., 2012). Here, we showed that there was no significant effect of HRI, eIF2αP, or ATF4 deficiencies on growth and viability during expansion and during up to 20 hr of erythroid differentiation (Figure 5C–D). However, eIF2α Ala51/Ala51 (AA) erythroid precursors, which were devoid of eIF2αP, accumulated significant numbers of globin inclusions between 20 hr and 30 hr of erythroid differentiation, resulting in cell death and fragments of cell debris (Figure 5C). A similar observation was found in Hri–/– erythroid precursors (Figure 5C), but the effect was less severe in these cells because eIF2αP is completely absent from AA erythroid precursors, whereas Hri–/ cells still have a low level of eIF2αP.

Together, the presence of aggregated protein inclusions in ex vivo differentiation and the induction of cytoplasmic UPR in Hri–/– EBs underscore the primary function of HRI–eIF2αP in inhibiting translation, and thereby coordinating protein homeostasis with available iron and heme concentrations to mitigate proteotoxicity during FL erythropoiesis.

HRI-ATF4 mediated mRNA expression is most highly upregulated in ID

As shown in Figure 5D, Atf4–/ FL erythroid cells did not suffer from proteotoxicity due to the presence of functional HRI-eIF2αP in inhibiting globin mRNA translation (Zhang et al., 2018). However, the differentiation of Atf4–/erythroid precursors at 30 and 42 hr was impaired (Figure 5D). These results demonstrate that the enhanced translation of Atf4 mRNA is also required for erythroid differentiation.

The majority of highly induced genes in Wt –Fe EBs compared to Wt +Fe EBs were ATF4 target genes (Figure 6A) (Pakos-Zebrucka et al., 2016). However, these genes were not upregulated in Hri–/– –Fe EBs (Figure 6B, Figure 6—figure supplement 1A), indicating that HRI is required to activate ISR in ID (Figure 2). The GO terms for the biological processes that were most highly differentially expressed in Wt –Fe EBs as compared to Hri–/– –Fe EB were amino acid metabolism and biosynthesis (Figure 6C). The corresponding upregulated ISR genes are involved in serine-glycine biosynthesis (Phgdh, Psat1, Psph, Shmt2), one-carbon metabolism (Shmt2, Mthfd2), proline-glutamine synthesis (Pycr1, Aldh18a1, Asns) and glutathione metabolism (Chac1) (Figure 6B, Figure 6—figure supplement 1B and Supplementary file 1e). Furthermore, the expression levels of these genes were lower in Hri–/– EBs than in Wt EBs under the +Fe condition (Figure 6—figure supplement 1C), suggesting that HRI finetunes the ISR during iron replete erythropoiesis and thus has an important role even under normal conditions. Interestingly, some of these ATF4 target genes, most notably Atf5, Trib3 and Chac1, are also Epo-stimulated genes (Figure 6—figure supplement 1D) (Singh et al., 2012), consistent with the interaction of the HRI-ISR pathway and Epo signaling (Zhang et al., 2018).

Grb10 is required for late-stage erythroid differentiation

Having mapped the genome-wide translational and transcriptional responses to ID, we set out to characterize the function of one of the most highly induced ATF4 target genes, growth factor receptor-bound protein 10 (Grb10), in erythropoiesis. Grb10 has been reported to be part of a feedback mechanism that inhibits the mediation of mTORC1 signaling by growth factors such as insulin (Plasschaert and Bartolomei, 2015) and stem cell factor (SCF) (Yan et al., 2016). We employed ex vivo FL differentiation to interrogate the function of Grb10 in erythropoiesis.

First, we validated that Grb10 and Atf5, but not Atf4, expression was increased in Wt –Fe EBs, but not in Hri–/– –Fe EBs (Figure 7—figure supplement 1A). In addition, Grb10 expression in Wt EBs was increased during ex vivo erythroid differentiation from 20 to 30 hr, and was greatly reduced in Hri–/–, AA and Atf4–/ erythroblasts (Figure 7—figure supplement 1B). We prepared eight shRNA recombinant retroviruses, all of which were able to knockdown Grb10 RNA expression by more than 80% during the expansion phase. Two of these retroviruses, shRNA_G3 and shRNA_G7, were able to maintain persistent knockdown of Grb10 RNA during differentiation (Figure 7A, Figure 7—figure supplements 23). Reduction of Grb10 expression by shRNAs (Figure 7D and Figure 7—figure supplement 3A) increased the numbers of differentiating erythroblasts (Figure 7B). This increase was mostly observed between 16 and 26 hr of differentiation (Figure 7B), a period during which Epo concentrations were increased and SCF was withdrawn (Figure 7A). We also observed increased expression of cyclin D3, a known target of mTORC1 signaling (Figure 7D). Importantly, terminal erythroid differentiation was inhibited in Grb10 knockdown cells, as indicated by an accumulation of polychromatic erythroblasts and a decrease in orthochromatic erythroblasts (Figure 7C). Thus, the increased proliferation but decreased terminal differentiation upon reduction of Grb10 expression (Figure 7) recapitulates the hallmarks of ineffective erythropoiesis observed in Hri–/– mice in ID (Han et al., 2001; Suragani et al., 2012; Zhang et al., 2018).

Figure 7. Knockdown of Grb10 increases proliferation and inhibits differentiation of FL erythroblasts.

(A) Designs of shRNA knockdown experiments using LinTer119CD71 FL erythroid progenitors. (B) Proliferation of control and Grb10 knockdown GFP+ cells. Three biological replicas were performed. Data are presented as means ± SEs. *p<0.05, **p<0.01. (C) Representative cell morphology of GFP control and Grb10 knockdown cells. (D) GRB10 and Cyclin D3 protein levels of GFP control and Grb10 knockdown cells during differentiation. Arrows indicate the doublets of GRB10. The green fluorescent protein (GFP) control was infected with retrovirus expressing GFP only; G3 and G7 were infected with retroviruses expressing shRNA_G3 and shRNA_G7, respectively.

Figure 7.

Figure 7—figure supplement 1. Expression of Atf4, Atf5 and Grb10 mRNAs in HRI-ISR defective erythroid cells.

Figure 7—figure supplement 1.

(A) Atf4, Atf5 and Grb10 expression in sorted basophilic EBs as illustrated in Figure 1A. The expression level in Wt +Fe EBs was defined as 1 (n = 3). (B) Grb10 expression at 20 hr and 30 hr of ex vivo differentiation of Wt, Hri–/–, AA and Atf4 FL erythroid progenitors. The expression level in Wt EBs at 20 hr was defined as 1. For each time point, expression levels in Hri–/–, AA and Atf4 EBs were compared to those in Wt EBs (n = 3). Data are presented as means ± SEs. *p<0.05, **p<0.01, ***p<0.001.
Figure 7—figure supplement 2. Enrichment of LinTer119CD71 erythroid progenitors from Wt +Fe E13.5 FLs for ex vivo experiments.

Figure 7—figure supplement 2.

Figure 7—figure supplement 3. Knockdown efficiency of Grb10 in FL cells during ex vivo differentiation.

Figure 7—figure supplement 3.

(A) Knockdown efficiency of Grb10 RNA by shRNA_G3 and shRNA_G7. Grb10 expression in the GFP control at 0 hr of differentiation is defined as 1. (B) Percentages of GFP+ cells during the expansion and differentiation phases. The GFP control was infected with retrovirus expressing GFP only. G3 and G7 were infected with retroviruses expressing shRNA_G3 and shRNA_G7, respectively. Data are presented as means ± SEs. ***p<0.001.

These results strengthen our global observation of an interaction between HRI-ISR and the Epo signaling pathway by showing that the induction of Grb10 by HRI-ISR serves as an important feedback mechanism for Epo signaling, resulting in reduced proliferation and thus promoting erythroid differentiation during stress erythropoiesis (Figure 8).

Figure 8. Proposed model of the regulation of erythropoiesis by heme and the HRI-ISR pathway during ID.

Diet-induced systemic ID results in heme deficiency, which activates HRI to phosphorylate its substrate eIF2α. The primary function of eIF2αP is to inhibit general protein synthesis in both the cytoplasm and mitochondria via decreased translation of ribosomal protein mRNAs from both cellular compartments. In the absence of HRI, continued protein synthesis results in the accumulation of unfolded proteins in both the cytoplasm and mitochondria, leading to proteotoxicity and mitochonfrial dysfunction. Second, eIF2αP also enhances the translation of Atf4 mRNA via regulation by uORFs. Atf4 mRNA is the mRNA that shows the greatest differences in translation levels between Wt and Hri–/– EBs. The ATF4 protein induces the expression of three genetic pathways to activate mitochondrial UPR (UPRmt), reprogram mitochondrial metabolism and reduce oxidative stress, all of which enables adaptation to ID and erythroid differentiation. Strikingly, the expression of ATF4 target genes is most highly activated in ID and requires HRI. Thus, assessment of the global genome-wide gene expression of primary EBs in vivo reveals that HRI-ISR contributes most significantly to adaptation to ID. We have shown previously that HRI-ISR suppressed mTORC1 signaling, which is activated by elevated Epo levels in ID (Zhang et al., 2018). Here, we provide evidence that GRB10 that is induced by ATF4 may be one of the molecules involved in the repression of mTORC1 signaling.

Figure 8.

Figure 8—figure supplement 1. Expression of Atf4 mRNA in hematopoietic cells.

Figure 8—figure supplement 1.

The plot of Atf4 expression levels in hematopoietic cells of murine bone marrow was obtained from the BloodSpot database (Bagger et al., 2016). The dataset used for ploting the tree was Mouse Normal RNA-seq, which was derived from the work of Lara-Astiaso et al. (2014). Expression level is log2-transformed and visualized by the size and color of the nodes.

Discussion

Although the specific role of HRI in the translational regulation of globin and Atf4 mRNAs in erythroid cells has been appreciated, an understanding of the global impact of HRI-mediated translational regulation on erythropoiesis is lacking. Here, we report a global, unbiased assessment of iron, heme, and HRI-mediated translational and transcriptional alterations in primary EBs in vivo. We show that Eif2ak1 and Atf4 mRNAs are abundantly expressed in EBs and are poised to respond to ID during terminal erythroid differentiation. HRI is a major regulator of gene expression in erythropoiesis: only limited changes of mRNA expression are observed in Hri–/– EBs during ID. Our global analysis supports a distinct model of heme and HRI-regulated translational control involving several separate and interacting pathways that allow EBs to adapt to systemic ID (Figure 8). Activation of HRI in ID elicits three distinct pathways of translation. First and foremost, inhibition of protein synthesis prevents the accumulation of unfolded proteins and maintains protein homeostasis. Second, induction of Atf4 mRNA translation increases the expression of genes encoding mitochondrial UPR, redox homeostasis and metabolic reprogramming in order to maintain mitochondrial respiration and erythroid differentiation. Third, repression of Epo-mTORC1 signaling inhibits protein synthesis and enables erythroid differentiation.

The present study identifies a previously unappreciated role of HRI in the translational regulation of cytosolic and mitochondrial ribosomal proteins during erythropoiesis. We observed, at a global level, additional evidence of mTORC1's involvement in translational regulation in Hri–/– –Fe EBs, further supporting the role of HRI-ISR in repressing Epo–mTORC1 signaling to mitigate ineffective erythropoiesis during ID (Zhang et al., 2018). Importantly, we find that HRI–eIF2αP also controls the translation of 18 cytosolic RP mRNAs, which are not known targets of mTORC1. Consistent with our results, mTORC1-independent translation of RP mRNAs requiring GCN2-eIF2αP has been reported recently in several cancer cell lines (Li et al., 2018). Interestingly, the translation of many mitochondrial RPs and oxidative phosphorylation complexes is also increased in Hri–/– –Fe EBs. Furthermore, HRI directly inhibits mitochondrial protein synthesis in addition to repressing mTORC1 activity. Thus, both the HRI–eIF2αP and the mTORC1 pathways contribute to the increased mitochondrial protein synthesis in Hri–/– –Fe EBs.

Increased protein synthesis in Hri–/– –Fe EBs requires additional protein folding capacity for heme-free globins and increased import into mitochondria of mitochondrial protein precursors synthesized in the cytoplasm. Indeed, Hri–/– –Fe EBs develop a prominent cytoplasmic UPR, as suggested here by the increased expression of cytoplasmic chaperones in these cells. These results are consistent with and help to identify the mechanisms underlying our previous reports of insoluble inclusions and protein precipitates accompanied by increased levels of Hsp70 and Hsp90 in erythroid cells of Hri–/– and eAA mice in ID (Han et al., 2001; Zhang et al., 2018). Nevertheless, activation of cytoplasmic UPR alone is not sufficient to compensate for the loss of HRI.

Translational regulation of mitochondrial biogenesis by mTORC1 is particularly important for erythropoiesis because of the high demand for heme for hemoglobin production and the prevention of oxidative stress (Liu et al., 2017). Most mitochondrial proteins are encoded by nuclear genes and synthesized by cytosolic ribosomes as precursors that are imported into mitochondria. Mitochondrial DNA encodes only 13 proteins, which are all components of the respiratory chain and oxidative phosphorylation (Calvo and Mootha, 2010). Therefore, the nuclear and mitochondrial synthesis of proteins that are involved in respiratory complexes needs to be coordinated to avoid excessive uncomplexed subunits (Priesnitz and Becker, 2018). In mitochondrial disorders, mitochondrial UPR (UPRmt) is activated to coordinate with nuclear transcription in order to enable restoration of mitochondrial function (Shpilka and Haynes, 2018). Interestingly, ISR is also activated in mitochondrial dysfunction and mediates UPRmt (Fiorese et al., 2016; Quirós et al., 2017; Dogan et al., 2014). However, the eIF2α kinase responsible for UPRmt remains unknown. Our observations that Hri–/– erythroid cells have impaired mitochondrial function and no significant increase in the expression of mitochondrial chaperones suggest that HRI may be the eIF2α kinase responsible for the activation of UPRmt and might thereby contribute to the proper maintenance of the mitochondrial function of erythroid cells under ID.

While inhibition of protein synthesis by the activation of HRI is the first line of defense in reducing proteotoxicity in ID, the second arm of the HRI-ISR, inducing ATF4 target gene expression as a consequence of enhanced translation of Atf4 mRNA, is the most highly activated transcriptome pathway in ID. These ATF4 target genes are largely unstudied factors with yet unknown functions in erythropoiesis. They are involved in redox and amino acid metabolism, particularly one-carbon metabolism. Interestingly, these ATF4-ISR genes are also upregulated under mitochondrial stress (Quirós et al., 2017; Bao et al., 2016) in order to maintain mitochondrial redox homeostasis. The deficiency of induction of ATF4-ISR gene expression is likely to be responsible for elevated oxidative stress in erythroid cells from iron-deficient Hri–/–, eAA and Atf4–/– mice (Suragani et al., 2012; Zhang et al., 2018), as well as the mitochondrial dysfunction reported here.

Hyporesponsiveness to Epo therapy is commonly encountered in ID (Goodnough, 2007). Our recent studies provide evidence that HRI-ISR constitutes one feedback mechanism of erythroid homeostasis in ID (Zhang et al., 2018). We show here that Grb10 is probably one of the ATF4 target genes involved in inhibiting Epo–mTORC1 signaling in ID. The increased proliferation of differentiating erythroblasts upon knockdown of Grb10 is consistent with the growth suppressor function of Grb10 (Plasschaert and Bartolomei, 2015). In hematopoiesis, Grb10 regulates the self-renewal and regeneration of hematopoietic stem cells (HSCs). Grb10-deficient HSCs exhibit increased proliferation that is dependent on the SCF-AKT/mTORC1 pathway (Yan et al., 2016), consistent with our Grb10 knockdown results. Grb10 has also been shown to be a GATA2 target gene and to play a role in the transition from BFU-E to CFU-E (Mehta et al., 2017). Our results support an additional role of ATF4-induced Grb10 expression in terminal erythropoiesis in which GATA2 expression is low.

Notably, Atf4 mRNA is most highly expressed in Ter119+CD71+ erythroblasts among the 16 differentiation stages of murine bone marrow hematopoietic cells (Figure 8—figure supplement 1) (Lara-Astiaso et al., 2014). Altogether, our findings underscore the role of HRI-mediated translation of Atf4 mRNA in terminal erythroid differentiation. In summary, our genome-wide study reveals the prominent and coordinated contribution of heme- and HRI-mediated translational regulation in maintaining protein homeostasis, mitochondrial activity and erythroid differentiation during iron-restricted erythropoiesis.

Materials and methods

Animals and diet-induced iron deficiency

Mice were maintained at the Massachusetts Institute of Technology (MIT) animal facility, and all experiments were carried out using protocols approved by the Division of Comparative Medicine, MIT. Hri–/–, Atf4–/ and universal eIF2α Ala51/+ heterozygote knockin mice were as described previously (Han et al., 2001; Masuoka and Townes, 2002; Scheuner et al., 2001). Diet-induced iron deficiency in mice was achieved as previously described (Han et al., 2001; Zhang et al., 2018). Iron-sufficient and -deficient mice (8–12 weeks old) were mated for E13.5 or E14.5 fetal livers as sources of erythroid precursors. Under these iron-deficient conditions, embryos were pale and anemic with decreased hematocrits in embryonic blood (Liu et al., 2008b).

Isolation of EBs, library preparations and DNA sequencing

EBs from E14.5 FLs of Wt and Hri–/– mice maintained under +Fe or –Fe conditions were sorted using anti-Ter119 and anti-CD71 antibodies by flow cytometry using FACS Aria (BD Biosciences, San Jose, CA) (Figure 1A). In order to have sufficient EBs for Ribo-seq library, FLs from embryos of the same mother were pooled then sorted for EBs as one biological replica. Two (5 million cells each) and three biological replicas (1 million cells each) for each condition from separate mothers were collected for preparation of Ribo-seq and mRNA-seq libraries, respectively. The third replica of Ribo-seq using 3 million cells was unsuccessful, probably because of lower cell numbers. All procedures for labeling and washing of cells for sorting were carried out at 4°C. Cells were sorted into tubes with 20% fetal bovine serum (FBS, Atlanta Biologicals, Norcross, GA) to help to preserve cell integrity.

To preserve polyribosomes, sorted EBs were washed twice with cold phosphate-buffered saline (PBS) and then treated with cycloheximde (100 μg/mL) at 37°C for 5 min, followed by isolation of RPFs as previously described (Guo et al., 2010; Ingolia et al., 2011) using the ARTseq-Ribosome Profiling Kit (Illumina, San Diego, CA). Briefly, cells were lysed in cold polysome buffer on ice for 10 min. To generate RPFs, lysates were digested with Rnase 1. Monosomes were then purified by S-400 gel filtration spin columns (GE Healthcare, Chicago, IL). RNAs in monosomes were extracted and purified. Size selection of 28–34 nucleotide RPFs was carried out using denaturing polyacrylamide gel electrophoresis. Ribosomal RNA contaminants in the RPF preparations were removed using an Ribo-Zero rRNA Removal Kit (Illumina, San Diego, CA). RPFs were then purified by 15% urea-polyacrylamide gel electrophoresis. Libraries of RPFs were prepared as described previously (Ingolia et al., 2009; Ingolia et al., 2011) for Illumina next generation DNA sequencing. RPFs were dephosphorylated and linkers were ligated using T4 RNA ligase. RNA samples were then reverse transcribed, circularized and PCR amplified for 12 cycles. PCR products were subjected to gel purification before sequencing. The results of Ribo-seq presented in Figure 2 were from the second set of experiments. For the first set of Ribo-seq experiments, EBs were treated with harringtonine (2 μg/ml) at 37°C for 2 min followed by cycloheximide treatment as described above. However, harringtonine treatment at this condition did not work well for EBs and consequently ribosome occupancy was observed throughout many mRNAs. For the second set of Ribo-seq experiments, harringtonine was not used. Nonetheless, increased translation of Atf4, Pppr115a and Ddit3 mRNAs was also observed in both sets of Ribo-seq.

Total RNAs were extracted using an RNeasy Plus kit (Qiagen, Germantown, MD) and polyA+ mRNAs were isolated using an Oligotex mRNA kit (Qiagen, Germantown, MD). mRNA-seq cDNA libraries were prepared by the MIT BioMicro Center. cDNA libraries of RPFs and mRNA were sequenced on an HiSeq 2000 platform (Illumina, San Diego, CA) at the MIT BioMicro Center. After standard preprocessing and quality control analysis, reads were mapped to mouse genome mm10 (UCSC), beforedownstream analyses were performed.

Genome-wide data analysis

Raw data (fastq files) were trimmed by Cutadapt to remove adapters and reads with a base quality score of less than 10. Reads with a length of less than 26 or 9 nucleotides were also discarded for Ribo-seq and mRNA-seq samples, respectively. For Ribo-seq samples, reads containing rRNA and tRNA sequences were further removed using Bowtie2. After quality-control analysis using FastQC, reads were mapped to mouse genome mm10 (UCSC) using STAR aligner with maxima of two mismatches and eight multiple loci. The quality of the Ribo-seq data was examined by the triplet periodicity using RibORF (Ji et al., 2015).

For the visualization of ribosome occupancies, all mapped reads (bam files) that overlap each bin (bin size = 1) were first calculated and normalized to effective mouse genome size to get a 1x depth of coverage (RPGC) using bamCoverage of deepTools (Reimand et al., 2016; Ji et al., 2015). Then, ribosome occupancies were visualized using the Integrative Genomics Viewer (Broad Institute of MIT and Harvard). As there was no significant change in mRNA levels among four conditions, only mapped reads of Wt + Fe EBs are shown for mRNA-seq data in Figure 2C–D, Figure 2—figure supplement 1B–C and Figure 3—figure supplement 1B–C.

The uniquely mapped reads were counted using HTseq for gene coverage analysis, transcript per million (TPM) calculation, translational efficiency (TE) calculation and analysis of differentially expressed mRNAs (DEG). Genes with greater than 25 uniquely mapped reads in at least one of the conditions were used for gene coverage analysis and TPM calculation.

Generally, RPFs are piled up around the start codons and stop codons because of the slower kinetics of translation initiation and termination. The use of cycloheximide to freeze polysomes during the preparation of RPFs also enhances the pile up of reads near start codons. We therefore removed the first 15 and the last 5 codons from reads counting for the TE calculation of Ribo-seq data, in which TE is defined as the ratio of RPF counts to mRNA counts. TE was calculated using the xtail package of R/Bioconductor, with parameter ‘bins = 10000’, in order to identify genes that are differentially translated between different conditions (Xiao et al., 2016). Genes with reads greater than 50 in at least one of the conditions were included in each comparison and were used for TE calculation (Supplementary file 1b). Genes with adjusted p-value<0.05 together with log2(foldchange of TE)>1.5 or <−1.5 were considered to be significantly differentially translated (Figure 2A), whereas those with adjusted p-value<0.05 together with log2(foldchange of TE)>1 or < −1 were considered to be significantly differentially translated (Figure 2B, Figure 3A, Figure 2—figure supplement 1A and Figure 3—figure supplement 1A).

For Gene Set Enrichment Analysis (GSEA), pre-ranked gene lists were obtained using the formula –Log10(adjusted p-value) × Log2(foldchange of TE). Mouse gene symbols were converted into human symbols using the biomaRt package of R/Bioconductor. GSEA was conducted using the GSEA tool from the Broad Institute with the c2.all.v6.0.symbols.gmt database, which includes 4738 gene sets. Gene sets shown in Figure 3B–C were REACTOME_FORMATION_OF_THE_TERNARY_COMPLEX_AND_SUBSEQUENTLY_THE_43S_COMPL, KEGG_RIBOSOME, WONG_MITOCHONDRIA_GENE_MODULE, KEGG_OXIDATIVE_PHOSPHORYLATION and BILANGES_SERUM_AND_RAPAMYCIN_SENSITIVE_GENES.

Differentially expressed genes from mRNA-seq data were determined using the DESeq2 package of R/Bioconductor, and genes with the mean of reads for all conditions greater than 100 were used for further analysis (Supplementary file 1e). Genes with adjusted p-value<0.05 were considered as significantly differentially (Figures 56). Lists of 5′ TOP/TOP-like genes were obtained from Thoreen et al. (2012), whereas the integrated stress response (ISR) gene list was derived from Palam et al. (2015).

Gene ontology analysis was performed using g:Profiler for the TE and DEG datasets (Reimand et al., 2016). The START App was employed to generate the volcano plots, in which red dots indicate significantly differentially translated or expressed genes, while green and gray dots indicate non-significantly changed genes (Nelson et al., 2017; Reimand et al., 2016). Heatmaps were plotted using GENE-E tool from the Broad Institute.

All Ribo-seq data and mRNA-seq described in this paper are available at the Gene Expression Ominibus (http://www.ncbi.nlm.nih.gov/geo/) under accession GSE119365.

RT-qPCR and western blot analyses

Gene expression was performed by RT-qPCR and western blot analyses as previously described (Zhang et al., 2018). Primers are listed in Supplementary file 3a. Antibodies used in western blots are described in Supplementary file 3b. Gapdh was used as the internal control for RT-qPCR. β-actin or eIF2α was used as a loading control for western blots.

Measurement of protein synthesis

The incorporation of puromycin into nascent polypeptide chains was used as a measure of protein synthesis as described previously (Schmidt et al., 2009; Zhang et al., 2018). Briefly, erythroid cells from bone marrow (BM) samples were isolated as previously described (Zhang et al., 2018). Isolated cells were further subjected to dead-cell removal using the MACS Dead Cell Removal Kit according to the manufacturer's protocol (Miltenyi Biotech). Then, cells were first treated with DMSO or cycloheximide at 100 μg/mL (Sigma-Aldrich, St. Louis, MO), with chloramphenicol at 100 μM (Sigma-Aldrich) or with INK128 at 0.2 μM (LC Laboratories, Woburn, MA) for 15 min at 37°C, followed by the addition of puromycin (Sigma-Aldrich, St. Louis, MO) at 5 μg/mL. Cells were collected after 15, 30, 45 or 60 min for western blot analysis using anti-puromycin antibody to measure protein synthesis.

Measurements of oxygen consumption rate, mitochondrial mass and mitochondrial DNA

Oxygen consumption rate (OCR) was measured on a Seahorse XFe96 Analyzer with a Seahorse XFe96 FluxPak (Agilent, Santa Clara, CA). Erythroid cells from BM of mice were isolated as previously described (Zhang et al., 2018). After recovery in IMDM with 10% FBS and 3 U/mL Epo at 37°C for 1 hr, 300,000 cells/well were seeded onto an XF96 cell culture microplate coated with Cell-tak (Corning, Corning, NY) and subjected to seahorse assay according to the protocol of the manufacturer. Three OCR measurements — basal, ATP-linked and maximal respiration — were performed after sequential injections of 1 µM oligomycin, 1 µM FCCP, and 0.5 µM Antimycin A/Rotenone. Five technical replicas from the same sample were performed for OCR measurement. Mitochondrial mass was determined by Mitotracker through flow cytometry analysis. The mitochondrial DNA (mtDNA) content was measured by qPCR as previously described (Liu et al., 2017). Briefly, after DNA isolation from the purified erythroid cells using a DNeasy Blood and Tissue Kit (Qiagen, Germantown), qPCR with specific primers for mtDNA and genomic DNA (gDNA) (Supplementary file 3a) was performed to measure the ratio of mtDNA versus gDNA using the ΔΔCt method.

Enrichment of FL erythroid progenitors for ex vivo culture and differentiation

Erythroid progenitors from E13.5 FLs of Wt + Fe embryos were enriched by magnetic sorting using EasySep Magnet (StemCell Technologies, Vancouver, Canada) as described (Thom et al., 2014) (Figure 7—figure supplement 2). In brief, total FL cells were mechanically dissociated by pipetting in PBS containing 2% FBS (Atlanta Biologicals, Norcross, GA), 2.5 mM EDTA and 10 mM glucose. Cells were labeled with biotin-conjugated anti-B220, anti-CD3, anti-CD11b, anti-CD11c, anti-GR1, anti-Ter119 (all BioLegend, San Diego, CA) and anti-CD41 (eBioscience, San Diego, CA) antibodies. Lineage negative (LinTer119) cells were subjected to a second purification step to obtain LinTer119CD71 erythroid progenitors using biotin-conjugated anti-CD71 antibodies (BioLegend, San Diego, CA and BD Biosciences, San Jose, CA).

Purified LinTer119CD71 erythroid progenitors were cultured in the expansion medi described by Thom et al. (2014) for three days at an initial cell density of 0.5 × 106 cells per mL. This expansion medium was StemPro-34 medium complemented with 10% supplement, 2 mM L-glutamine, 1% penicillin-streptomycin (P-S), 10−4 M β-mercaptoethanol, 10−6 M dexamethasone, 0.5 U/mL Epo, and 100 ng/mL mouse stem cell factor (mSCF). Then, cells were washed and cultured in differentiation medium at an initial cell density of 1 × 106 cells per mL. The differentiation medium was Iscove-modified Dulbecco medium (IMDM) containing 10% FCS (fetal calf serum, Gemini Bio-Products, West Sacramento, CA), 10% plasma-derived serum (Animal Technologies, Tyler, TX), 2 mM L-glutamine, 1% P-S, 10−4 M β-mercaptoethanol, and 5 U/mL Epo. Cells were collected for analysis at different time points as indicated in Figure 7A.

Knockdown of Grb10 expression in FL EBs by shRNAs

The knockdown of Grb10 expression was performed by using recombinant retroviruses containing an shRNA-expressing murine stem cell retroviral vector, MSCV-pgkGFP-U3-U6P-Bbs, a kind gift from the laboratory of Dr. Harvey F Lodish (MIT). DNA sequences of eight shRNA oligonucleotides (Supplementary file 3c) targeting different regions of Grb10 mRNA were either obtained from the Genetic Perturbation Platform of the Broad Institute or as previously reported (Doiron et al., 2012; Zacharek et al., 2011; Park et al., 2013) and were synthesized by Integrated DNA Technologies. Preparations of the plasmid constructs and recombinant retroviruses were performed as described previously (Hu et al., 2014). LinTer119CD71 erythroid progenitors enriched from Wt + Fe E13.5 FLs were infected with retroviruses as described previously (Thom et al., 2014). Cells were expanded for 72 hr after retroviral infections followed by differentiation for up to 42 hr as indicated.

Analysis of erythroid differentiation and cell proliferation

Erythroid differentiation was performed by flow cytometry using anti-Ter119 and anti-CD71 antibodies (BioLegend, San Diego, CA) as described previously (Zhang et al., 2018) on FACS LSR II (BD Biosciences, San Jose, CA). 4′,6-diamidino-2-phenylindole (DAPI, Roche Diagnostics, Basel, Switzerland) was used to exclude the dead cells. Data were analyzed with FlowJo (Tree Star, Ashland, OR). Erythroid differentiation was also analyzed by cell morphology on cytospin slides stained with May-Grunwald/Giemsa staining (Sigma-Aldrich, St. Louis, MO). Cell proliferation was determination by counting nucleated cells daily using crystal violet stain.

Statistical analysis

The independent t test (two-tailed) was used to analyze the experimental data. Pearson correlation analysis was performed to determine the correlation coefficient. Data are presented as means ± SEs. *p<0.05 was considered statistically significant. **p<0.01; ***p<0.001.

Acknowledgements

This paper is dedicated to the memory of Irving M London and his generous and inspiring mentorship. This work was supported by National Institute of Health Grants RO1 DK087984 (to J-JC), and R01 DK103794 and R33 HL120791 (to VGS). We are grateful for the expert help of Dr. Nora Kory from Dr. David Sabatini’s lab with Seahorse assays, which were performed at the Whitehead Institute for Biomedical Research, Cambridge, MA.

Funding Statement

The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.

Contributor Information

Jane-Jane Chen, Email: j-jchen@mit.edu.

Jian Xu, University of Texas Southwestern Medical Center, United States.

David Ron, University of Cambridge, United Kingdom.

Funding Information

This paper was supported by the following grants:

  • National Institute of Diabetes and Digestive and Kidney Diseases RO1 DK087984 to Jane-Jane Chen.

  • National Institute of Diabetes and Digestive and Kidney Diseases R01 DK103794 to Vijay G Sankaran.

  • National Heart, Lung, and Blood Institute R33 HL120791 to Vijay G Sankaran.

Additional information

Competing interests

No competing interests declared.

Author contributions

Data curation, Formal analysis, Validation, Investigation, Visualization, Methodology, Writing—original draft.

Formal analysis, Validation, Investigation, Methodology, Writing—review and editing.

Formal analysis, Methodology, Writing—review and editing.

Investigation.

Formal analysis.

Formal analysis.

Formal analysis, Funding acquisition, Writing—review and editing.

Conceptualization, Resources, Data curation, Formal analysis, Supervision, Funding acquisition, Investigation, Methodology, Writing—original draft, Project administration.

Ethics

Animal experimentation: Mice were maintained at the Massachusetts Institute of Technology (MIT) animal facility, and all experiments were carried out using protocols (#1015-099-18) approved by the Division of Comparative Medicine, MIT.

Additional files

Supplementary file 1. List of diifferentially translated mRNAs by HRI and mTORC1.

Supplementary File 1a. Quality control and mapping of Ribo-seq and mRNA-seq. Supplementary File 1b. Complete gene list of differentially translated mRNAs. Supplementary File 1c. Eif2ak1, Ppp1r15a and Atf4 mRNAs are highly expressed in basophilic erythroblasts. Supplementary File 1d. Higher ribosome occupancy at the AUG of uORF1 in Atf4 mRNA.Supplementary File 1e. Complete gene list of differentially expressed mRNAs.

elife-46976-supp1.xls (2.2MB, xls)
DOI: 10.7554/eLife.46976.021
Supplementary file 2. List of mRNAs that are differentially translated between Hri–/– –Fe EBs and Wt –Fe EBs and known mTORC1 targets in the categories of (a) RPs and non-RPs and (b) mitochondrial proteins.
elife-46976-supp2.xls (43.5KB, xls)
DOI: 10.7554/eLife.46976.022
Supplementary file 3. List of RNA primers, antibodies and ShRNA oligonucleotides.

Supplementary File 3a. DNA sequence of qPCR primers. Supplementary File 3b. List of antibodies used for western blot analyses. Supplementary File 3c. DNA sequences of shRNA oligonucleotides.

elife-46976-supp3.xls (32KB, xls)
DOI: 10.7554/eLife.46976.023
Transparent reporting form
DOI: 10.7554/eLife.46976.024

Data availability

All sequencing data have been deposited in GEO under accession code GSE119365.

The following dataset was generated:

Zhang S, Chen J. 2018. Iron and Heme Coordinate Erythropoiesis through HRI-Mediated Regulation of Protein Translation and Gene Expression. NCBI Gene Expression Omnibus. GSE119365

References

  1. An X, Schulz VP, Li J, Wu K, Liu J, Xue F, Hu J, Mohandas N, Gallagher PG. Global transcriptome analyses of human and murine terminal erythroid differentiation. Blood. 2014;123:3466–3477. doi: 10.1182/blood-2014-01-548305. [DOI] [PMC free article] [PubMed] [Google Scholar]
  2. Bagger FO, Sasivarevic D, Sohi SH, Laursen LG, Pundhir S, Sønderby CK, Winther O, Rapin N, Porse BT. BloodSpot: a database of gene expression profiles and transcriptional programs for healthy and malignant haematopoiesis. Nucleic Acids Research. 2016;44:D917–D924. doi: 10.1093/nar/gkv1101. [DOI] [PMC free article] [PubMed] [Google Scholar]
  3. Bao XR, Ong SE, Goldberger O, Peng J, Sharma R, Thompson DA, Vafai SB, Cox AG, Marutani E, Ichinose F, Goessling W, Regev A, Carr SA, Clish CB, Mootha VK. Mitochondrial dysfunction remodels one-carbon metabolism in human cells. eLife. 2016;5:e10575. doi: 10.7554/eLife.10575. [DOI] [PMC free article] [PubMed] [Google Scholar]
  4. Calvo SE, Mootha VK. The mitochondrial proteome and human disease. Annual Review of Genomics and Human Genetics. 2010;11:25–44. doi: 10.1146/annurev-genom-082509-141720. [DOI] [PMC free article] [PubMed] [Google Scholar]
  5. Camaschella C. Iron deficiency. Blood. 2019;133:30–39. doi: 10.1182/blood-2018-05-815944. [DOI] [PubMed] [Google Scholar]
  6. Chen JJ. Regulation of protein synthesis by the heme-regulated eIF2alpha kinase: relevance to anemias. Blood. 2007;109:2693–2699. doi: 10.1182/blood-2006-08-041830. [DOI] [PMC free article] [PubMed] [Google Scholar]
  7. Chen JJ. Translational control by heme-regulated eIF2α kinase during erythropoiesis. Current Opinion in Hematology. 2014;21:172–178. doi: 10.1097/MOH.0000000000000030. [DOI] [PMC free article] [PubMed] [Google Scholar]
  8. Chiabrando D, Mercurio S, Tolosano E. Heme and erythropoieis: more than a structural role. Haematologica. 2014;99:973–983. doi: 10.3324/haematol.2013.091991. [DOI] [PMC free article] [PubMed] [Google Scholar]
  9. Dogan SA, Pujol C, Maiti P, Kukat A, Wang S, Hermans S, Senft K, Wibom R, Rugarli EI, Trifunovic A. Tissue-specific loss of DARS2 activates stress responses independently of respiratory chain deficiency in the heart. Cell Metabolism. 2014;19:458–469. doi: 10.1016/j.cmet.2014.02.004. [DOI] [PubMed] [Google Scholar]
  10. Doiron B, Hu W, Norton L, DeFronzo RA. Lentivirus shRNA Grb10 targeting the pancreas induces apoptosis and improved glucose tolerance due to decreased plasma glucagon levels. Diabetologia. 2012;55:719–728. doi: 10.1007/s00125-011-2414-z. [DOI] [PubMed] [Google Scholar]
  11. Fiorese CJ, Schulz AM, Lin YF, Rosin N, Pellegrino MW, Haynes CM. The transcription factor ATF5 mediates a mammalian mitochondrial UPR. Current Biology. 2016;26:2037–2043. doi: 10.1016/j.cub.2016.06.002. [DOI] [PMC free article] [PubMed] [Google Scholar]
  12. Goodnough LT. Erythropoietin and iron-restricted erythropoiesis. Experimental Hematology. 2007;35:167–172. doi: 10.1016/j.exphem.2007.01.026. [DOI] [PubMed] [Google Scholar]
  13. Guo H, Ingolia NT, Weissman JS, Bartel DP. Mammalian microRNAs predominantly act to decrease target mRNA levels. Nature. 2010;466:835–840. doi: 10.1038/nature09267. [DOI] [PMC free article] [PubMed] [Google Scholar]
  14. Han AP, Yu C, Lu L, Fujiwara Y, Browne C, Chin G, Fleming M, Leboulch P, Orkin SH, Chen JJ. Heme-regulated eIF2alpha kinase (HRI) is required for translational regulation and survival of erythroid precursors in iron deficiency. The EMBO Journal. 2001;20:6909–6918. doi: 10.1093/emboj/20.23.6909. [DOI] [PMC free article] [PubMed] [Google Scholar]
  15. Harding HP, Zhang Y, Zeng H, Novoa I, Lu PD, Calfon M, Sadri N, Yun C, Popko B, Paules R, Stojdl DF, Bell JC, Hettmann T, Leiden JM, Ron D. An integrated stress response regulates amino acid metabolism and resistance to oxidative stress. Molecular Cell. 2003;11:619–633. doi: 10.1016/S1097-2765(03)00105-9. [DOI] [PubMed] [Google Scholar]
  16. Hinnebusch AG. In: Translational Control of GCN4: Gene-Specific Regulation by Phosphorylation of eIF2. Hershey J.W.B, Mathews M.B, Sonenberg N, editors. Cold Spring Harbor Laboratory Press; 1996. [Google Scholar]
  17. Hu W, Yuan B, Lodish HF. Cpeb4-mediated translational regulatory circuitry controls terminal erythroid differentiation. Developmental Cell. 2014;30:660–672. doi: 10.1016/j.devcel.2014.07.008. [DOI] [PMC free article] [PubMed] [Google Scholar]
  18. Igarashi K, Sun J. The heme-Bach1 pathway in the regulation of oxidative stress response and erythroid differentiation. Antioxidants & Redox Signaling. 2006;8:107–118. doi: 10.1089/ars.2006.8.107. [DOI] [PubMed] [Google Scholar]
  19. Ingolia NT, Ghaemmaghami S, Newman JR, Weissman JS. Genome-wide analysis in vivo of translation with nucleotide resolution using ribosome profiling. Science. 2009;324:218–223. doi: 10.1126/science.1168978. [DOI] [PMC free article] [PubMed] [Google Scholar]
  20. Ingolia NT, Lareau LF, Weissman JS. Ribosome profiling of mouse embryonic stem cells reveals the complexity and dynamics of mammalian proteomes. Cell. 2011;147:789–802. doi: 10.1016/j.cell.2011.10.002. [DOI] [PMC free article] [PubMed] [Google Scholar]
  21. Ingolia NT. Ribosome profiling: new views of translation, from single codons to genome scale. Nature Reviews Genetics. 2014;15:205–213. doi: 10.1038/nrg3645. [DOI] [PubMed] [Google Scholar]
  22. Ji Z, Song R, Regev A, Struhl K. Many lncRNAs, 5'UTRs, and pseudogenes are translated and some are likely to express functional proteins. eLife. 2015;4:e08890. doi: 10.7554/eLife.08890. [DOI] [PMC free article] [PubMed] [Google Scholar]
  23. Kaufman RJ. In: Double-Stranded RNA-Activated Protein Kinase PKR. Hershey J. W. B, Sonenberg N, Matthews M. B, editors. Cold Spring Harbor Laboratory Press; 2000. [Google Scholar]
  24. Kerenyi MA, Orkin SH. Networking erythropoiesis. The Journal of Experimental Medicine. 2010;207:2537–2541. doi: 10.1084/jem.20102260. [DOI] [PMC free article] [PubMed] [Google Scholar]
  25. Khajuria RK, Munschauer M, Ulirsch JC, Fiorini C, Ludwig LS, McFarland SK, Abdulhay NJ, Specht H, Keshishian H, Mani DR, Jovanovic M, Ellis SR, Fulco CP, Engreitz JM, Schütz S, Lian J, Gripp KW, Weinberg OK, Pinkus GS, Gehrke L, Regev A, Lander ES, Gazda HT, Lee WY, Panse VG, Carr SA, Sankaran VG. Ribosome levels selectively regulate translation and lineage commitment in human hematopoiesis. Cell. 2018;173:90–103. doi: 10.1016/j.cell.2018.02.036. [DOI] [PMC free article] [PubMed] [Google Scholar]
  26. Kingsley PD, Greenfest-Allen E, Frame JM, Bushnell TP, Malik J, McGrath KE, Stoeckert CJ, Palis J. Ontogeny of erythroid gene expression. Blood. 2013;121:e5–e13. doi: 10.1182/blood-2012-04-422394. [DOI] [PMC free article] [PubMed] [Google Scholar]
  27. Lara-Astiaso D, Weiner A, Lorenzo-Vivas E, Zaretsky I, Jaitin DA, David E, Keren-Shaul H, Mildner A, Winter D, Jung S, Friedman N, Amit I. Immunogenetics. chromatin state dynamics during blood formation. Science. 2014;345:943–949. doi: 10.1126/science.1256271. [DOI] [PMC free article] [PubMed] [Google Scholar]
  28. Li BB, Qian C, Gameiro PA, Liu CC, Jiang T, Roberts TM, Struhl K, Zhao JJ. Targeted profiling of RNA translation reveals mTOR-4EBP1/2-independent translation regulation of mRNAs encoding ribosomal proteins. PNAS. 2018;115:E9325–E9332. doi: 10.1073/pnas.1805782115. [DOI] [PMC free article] [PubMed] [Google Scholar]
  29. Liu S, Bhattacharya S, Han A, Suragani RN, Zhao W, Fry RC, Chen JJ. Haem-regulated eIF2alpha kinase is necessary for adaptive gene expression in erythroid precursors under the stress of iron deficiency. British Journal of Haematology. 2008a;143:129–137. doi: 10.1111/j.1365-2141.2008.07293.x. [DOI] [PMC free article] [PubMed] [Google Scholar]
  30. Liu S, Suragani RN, Han A, Zhao W, Andrews NC, Chen JJ. Deficiency of heme-regulated eIF2alpha kinase decreases hepcidin expression and splenic iron in HFE-/- mice. Haematologica. 2008b;93:753–756. doi: 10.3324/haematol.12175. [DOI] [PMC free article] [PubMed] [Google Scholar]
  31. Liu X, Zhang Y, Ni M, Cao H, Signer RAJ, Li D, Li M, Gu Z, Hu Z, Dickerson KE, Weinberg SE, Chandel NS, DeBerardinis RJ, Zhou F, Shao Z, Xu J. Regulation of mitochondrial biogenesis in erythropoiesis by mTORC1-mediated protein translation. Nature Cell Biology. 2017;19:626–638. doi: 10.1038/ncb3527. [DOI] [PMC free article] [PubMed] [Google Scholar]
  32. Masuoka HC, Townes TM. Targeted disruption of the activating transcription factor 4 gene results in severe fetal anemia in mice. Blood. 2002;99:736–745. doi: 10.1182/blood.V99.3.736. [DOI] [PubMed] [Google Scholar]
  33. Mehta C, Johnson KD, Gao X, Ong IM, Katsumura KR, McIver SC, Ranheim EA, Bresnick EH. Integrating enhancer mechanisms to establish a hierarchical blood development program. Cell Reports. 2017;20:2966–2979. doi: 10.1016/j.celrep.2017.08.090. [DOI] [PMC free article] [PubMed] [Google Scholar]
  34. Mills EW, Wangen J, Green R, Ingolia NT. Dynamic regulation of a ribosome rescue pathway in erythroid cells and platelets. Cell Reports. 2016;17:1–10. doi: 10.1016/j.celrep.2016.08.088. [DOI] [PMC free article] [PubMed] [Google Scholar]
  35. Morita M, Gravel SP, Chénard V, Sikström K, Zheng L, Alain T, Gandin V, Avizonis D, Arguello M, Zakaria C, McLaughlan S, Nouet Y, Pause A, Pollak M, Gottlieb E, Larsson O, St-Pierre J, Topisirovic I, Sonenberg N. mTORC1 controls mitochondrial activity and biogenesis through 4E-BP-dependent translational regulation. Cell Metabolism. 2013;18:698–711. doi: 10.1016/j.cmet.2013.10.001. [DOI] [PubMed] [Google Scholar]
  36. Muckenthaler MU, Rivella S, Hentze MW, Galy B. A red carpet for iron metabolism. Cell. 2017;168:344–361. doi: 10.1016/j.cell.2016.12.034. [DOI] [PMC free article] [PubMed] [Google Scholar]
  37. Nelson JW, Sklenar J, Barnes AP, Minnier J. The START app: a web-based RNAseq analysis and visualization resource. Bioinformatics. 2017;33:447–449. doi: 10.1093/bioinformatics/btw624. [DOI] [PMC free article] [PubMed] [Google Scholar]
  38. Pakos-Zebrucka K, Koryga I, Mnich K, Ljujic M, Samali A, Gorman AM. The integrated stress response. EMBO Reports. 2016;17:1374–1395. doi: 10.15252/embr.201642195. [DOI] [PMC free article] [PubMed] [Google Scholar]
  39. Palam LR, Gore J, Craven KE, Wilson JL, Korc M. Integrated stress response is critical for gemcitabine resistance in pancreatic ductal adenocarcinoma. Cell Death & Disease. 2015;6:e1913. doi: 10.1038/cddis.2015.264. [DOI] [PMC free article] [PubMed] [Google Scholar]
  40. Park Y, Jin HS, Lopez J, Elly C, Kim G, Murai M, Kronenberg M, Liu YC. TSC1 regulates the balance between effector and regulatory T cells. Journal of Clinical Investigation. 2013;123:5165–5178. doi: 10.1172/JCI69751. [DOI] [PMC free article] [PubMed] [Google Scholar]
  41. Pavitt GD, Ron D. New Insights Into Translational Regulation in the Endoplasmic Reticulum Unfolded Protein Response. Cold Spring Harb Perspect Biol; 2012. [DOI] [PMC free article] [PubMed] [Google Scholar]
  42. Plasschaert RN, Bartolomei MS. Tissue-specific regulation and function of Grb10 during growth and neuronal commitment. PNAS. 2015;112:6841–6847. doi: 10.1073/pnas.1411254111. [DOI] [PMC free article] [PubMed] [Google Scholar]
  43. Priesnitz C, Becker T. Pathways to balance mitochondrial translation and protein import. Genes & Development. 2018;32:1285–1296. doi: 10.1101/gad.316547.118. [DOI] [PMC free article] [PubMed] [Google Scholar]
  44. Quirós PM, Prado MA, Zamboni N, D'Amico D, Williams RW, Finley D, Gygi SP, Auwerx J. Multi-omics analysis identifies ATF4 as a key regulator of the mitochondrial stress response in mammals. The Journal of Cell Biology. 2017;216:2027–2045. doi: 10.1083/jcb.201702058. [DOI] [PMC free article] [PubMed] [Google Scholar]
  45. Reimand J, Arak T, Adler P, Kolberg L, Reisberg S, Peterson H, Vilo J. G:profiler-a web server for functional interpretation of gene lists (2016 update) Nucleic Acids Research. 2016;44:W83–W89. doi: 10.1093/nar/gkw199. [DOI] [PMC free article] [PubMed] [Google Scholar]
  46. Richter U, Lahtinen T, Marttinen P, Myöhänen M, Greco D, Cannino G, Jacobs HT, Lietzén N, Nyman TA, Battersby BJ. A mitochondrial ribosomal and RNA decay pathway blocks cell proliferation. Current Biology. 2013;23:535–541. doi: 10.1016/j.cub.2013.02.019. [DOI] [PubMed] [Google Scholar]
  47. Ron D, Harding HP. In: PERK and Translational Control by Stress in Endoplasmic Reticulum. Sonenberg N, Hershey J W B, Mathews M B, editors. Cold Spring Harbor Laboratory Press; 2000. [Google Scholar]
  48. Scheuner D, Song B, McEwen E, Liu C, Laybutt R, Gillespie P, Saunders T, Bonner-Weir S, Kaufman RJ. Translational control is required for the unfolded protein response and in vivo glucose homeostasis. Molecular Cell. 2001;7:1165–1176. doi: 10.1016/S1097-2765(01)00265-9. [DOI] [PubMed] [Google Scholar]
  49. Schmidt EK, Clavarino G, Ceppi M, Pierre P. SUnSET, a nonradioactive method to monitor protein synthesis. Nature Methods. 2009;6:275–277. doi: 10.1038/nmeth.1314. [DOI] [PubMed] [Google Scholar]
  50. Shpilka T, Haynes CM. The mitochondrial UPR: mechanisms, physiological functions and implications in ageing. Nature Reviews Molecular Cell Biology. 2018;19:109–120. doi: 10.1038/nrm.2017.110. [DOI] [PubMed] [Google Scholar]
  51. Singh S, Dev A, Verma R, Pradeep A, Sathyanarayana P, Green JM, Narayanan A, Wojchowski DM. Defining an EPOR- regulated transcriptome for primary progenitors, including Tnfr-sf13c as a novel mediator of EPO- dependent erythroblast formation. PLOS ONE. 2012;7:e38530. doi: 10.1371/journal.pone.0038530. [DOI] [PMC free article] [PubMed] [Google Scholar]
  52. Suragani RN, Zachariah RS, Velazquez JG, Liu S, Sun CW, Townes TM, Chen JJ. Heme-regulated eIF2α kinase activated Atf4 signaling pathway in oxidative stress and erythropoiesis. Blood. 2012;119:5276–5284. doi: 10.1182/blood-2011-10-388132. [DOI] [PMC free article] [PubMed] [Google Scholar]
  53. Thom CS, Traxler EA, Khandros E, Nickas JM, Zhou OY, Lazarus JE, Silva AP, Prabhu D, Yao Y, Aribeana C, Fuchs SY, Mackay JP, Holzbaur EL, Weiss MJ. Trim58 degrades dynein and regulates terminal erythropoiesis. Developmental Cell. 2014;30:688–700. doi: 10.1016/j.devcel.2014.07.021. [DOI] [PMC free article] [PubMed] [Google Scholar]
  54. Thoreen CC, Chantranupong L, Keys HR, Wang T, Gray NS, Sabatini DM. A unifying model for mTORC1-mediated regulation of mRNA translation. Nature. 2012;485:109–113. doi: 10.1038/nature11083. [DOI] [PMC free article] [PubMed] [Google Scholar]
  55. Xiao Z, Zou Q, Liu Y, Yang X. Genome-wide assessment of differential translations with ribosome profiling data. Nature Communications. 2016;7:11194. doi: 10.1038/ncomms11194. [DOI] [PMC free article] [PubMed] [Google Scholar]
  56. Yan X, Himburg HA, Pohl K, Quarmyne M, Tran E, Zhang Y, Fang T, Kan J, Chao NJ, Zhao L, Doan PL, Chute JP. Deletion of the imprinted gene Grb10 promotes hematopoietic stem cell Self-Renewal and regeneration. Cell Reports. 2016;17:1584–1594. doi: 10.1016/j.celrep.2016.10.025. [DOI] [PMC free article] [PubMed] [Google Scholar]
  57. Zacharek SJ, Fillmore CM, Lau AN, Gludish DW, Chou A, Ho JW, Zamponi R, Gazit R, Bock C, Jäger N, Smith ZD, Kim TM, Saunders AH, Wong J, Lee JH, Roach RR, Rossi DJ, Meissner A, Gimelbrant AA, Park PJ, Kim CF. Lung stem cell self-renewal relies on BMI1-dependent control of expression at imprinted loci. Cell Stem Cell. 2011;9:272–281. doi: 10.1016/j.stem.2011.07.007. [DOI] [PMC free article] [PubMed] [Google Scholar]
  58. Zhang S, Macias-Garcia A, Velazquez J, Paltrinieri E, Kaufman RJ, Chen JJ. HRI coordinates translation by eIF2αP and mTORC1 to mitigate ineffective erythropoiesis in mice during iron deficiency. Blood. 2018;131:450–461. doi: 10.1182/blood-2017-08-799908. [DOI] [PMC free article] [PubMed] [Google Scholar]

Decision letter

Editor: Jian Xu1
Reviewed by: Jian Xu2, Emery H Bresnick3

In the interests of transparency, eLife includes the editorial decision letter and accompanying author responses. A lightly edited version of the letter sent to the authors after peer review is shown, indicating the most substantive concerns; minor comments are not usually included.

[Editors’ note: a previous version of this study was rejected after peer review, but the authors submitted for reconsideration. The first decision letter after peer review is shown below.]

Thank you for submitting your work entitled "Iron and Heme Coordinate Erythropoiesis through HRI-Mediated Regulation of Protein Translation and Gene Expression" for consideration by eLife. Your article has been reviewed by three peer reviewers, one of whom is a member of our Board of Reviewing Editors, and the evaluation has been overseen by a Senior Editor. The reviewers have opted to remain anonymous.

Our decision has been reached after consultation between the reviewers. Based on these discussions and the individual reviews below, we regret to inform you that your work will not be considered further for publication in eLife.

While the reviewers found that many aspects of your work were well done, they raise concerns about the degree to which it advances prior studies of similar mechanisms. They were also concerned about some aspects of the mechanisms, and the strength of conclusions that could be drawn based on loss-of-function studies with a single shRNA and experiments performed in culture. We believe your paper would be more appropriate for a more specialized journal.

Reviewer #1:

In this manuscript, Chen and colleagues characterized how iron, heme and the heme-regulated kinase HRI coordinate to regulate erythropoiesis under diet-induced iron deficiency (ID). By unbiased and comparative analysis of ribosome profiling and mRNA-seq datasets in mouse fetal liver basophilic erythroblasts (EBs) with or without ID, they validated the known regulator ATF4 as the most differentially translated mRNA in HRI and iron deficiencies. They also identified additional new pathways such as ribosomal protein mRNA translation. The authors then focused on HRI-ATF4 and described ATF4 as the major transcription factor under HRI translational control likely through uORF regulation. They also showed that heme-HRI-ATF4 pathway, but not iron-regulated pathways such as IRE-containing mRNAs, mediate the major translational and transcriptional changes in response to iron deficiency. Finally, they identified Grb10 as one of the ATF4 target genes that may function to inhibit Epo-mTORC1 in ID.

Overall, this study presents a set of interesting findings related to the role of heme-HRI-ATF4 pathway in erythropoiesis and iron deficiency anemia. These results not only validated ATF4 as a major factor in coordinating protein translation and gene expression during ID, but also identified regulators such as Grb10 in regulating erythropoiesis for the adaptation to iron deficiency anemia. In this regard, this manuscript expanded our understanding of the underlying basis for iron deficiency or ineffective erythropoiesis. The ribosome and mRNA profiling studies were appropriately designed and carefully analyzed. There are several important questions remained to be addressed related to the mechanistic details how ATF4 regulates Grb10 and how Grb10 regulates Epo-mTORC1 in the context of erythropoiesis. Moreover, the functional roles of these pathways in more physiologically relevant conditions need to be established. If the remaining questions can be adequately addressed, this study may provide important insights into heme-HRI-mediated pathways in iron deficiency and ineffective erythropoiesis, and be potentially impactful.

1) It is difficult to follow and somewhat confusing on the various comparisons of the ribosome profiling and mRNA-seq in WT+Fe, HRI-/-+Fe, WT-Fe and HRI-/--Fe (Figures 2-4, Figure 2—figure supplement 1). It would be helpful if the authors could summarize the findings in a diagram to list the most differentially regulated pathways and/or genes in the pairwise comparison. This diagram can be combined in Figure 7 or as a separate figure.

2) When comparing differentially expressed mRNAs in HRI and iron deficiencies, the authors identified Grb10 as one of the most upregulated ATF4 target genes in WT relative to HRI-/- basophilic erythroblasts. To establish the functional role of Grb10 in erythropoiesis, a single shRNA (G3) was used to knockdown Grb10 expression in ex vivo cultured erythroid cells. The authors measured phosphorylated AKTS473, which is slightly increased, as the evidence for the role of Grb10 as the negative regulator of mTORC1 signaling. It is important to characterize in greater details how Grb10 regulate mTORC1 or other pathways in erythropoiesis. The authors cited previous studies (Plasschaet and Bartolomer, 2015, and Yan et al., 2016) as evidence that Grb10 is part of a feedback mechanism to inhibit mTORC1, however it is important to determine whether similar or different mechanism may be utilized in the context of erythropoiesis. How does ATF4 regulate Grb10 transcription (or translation) in erythroid cells? Moreover, additional shRNAs should be screened to validate the original findings. Since this part of the study could provide potentially new insights compared to the already established HRI-ATF4 pathway in erythropoiesis, more in-depth mechanistic studies are required.

3) Since all the functional studies were performed using ex vivo cultured erythroid cells with defined medium and cytokines, the functional importance of the identified pathway (HRI-ATF4-Grb10) should be assessed in more physiologically relevant settings in vivo, such as in the already established Grb10 knockout mouse models or in the transplant setting.

Reviewer #2:

The manuscript by Zhang et al. describes an analysis of global mechanisms regulating protein translation by iron, heme and HRI in primary erythroblasts. HRI has been studied for many years by Dr. Chen, and it represents a component of cellular machinery that senses alterations in heme. Other cellular mechanisms of high importance also exist, including the IRE-IRP system, as well as transcriptional mechanisms involving the heme-regulated repressor BACH1. A major conclusion of the RNA-seq and Ribo-seq analyses is that iron deficiency most highly activates the integrated stress response and ATF4-mediated gene expression. While the studies appear to have been rigorously conducted, and the overall problem is significant, this major conclusion is not surprising based on prior published work. The global analyses did not transform or extend existing knowledge in a major way, which represents a significant limitation. Recommendations are presented below to build upon the work to elevate novelty and broad interest.

1) Subsection “Upregulation of in vivo translation of ISR mRNAs in Wt EBs compared to Hri-/- EBs”, last paragraph – The authors conclude that diet-induced iron deficiency does not significantly affect gene expression through IREs/IPRs in primary erythroblasts. I do not see how the global measurements and correlations support this rather strong conclusion. At the very least, this needs to be fleshed out more effectively.

2) "ATF4 target gene GRB10". GRB10 has been published by Mehta et al., 2017, to be a direct GATA2 target gene in primary mouse erythroid precursor cells. This may influence thinking regarding the role of GRB10 in mechanisms of interest to the authors.

3) Figure 6 – as noted above, the finding that Integrated Stress Response and ATF4 are important players does not transform existing knowledge. A limited loss of function analysis was performed with retrovirally-expressed shRNAs to evaluate GRB10 function. Only one shRNA was effective in reducing GRB10 levels in the differentiation culture. It will be important not to rely on results obtained with a single shRNA, especially given the recent findings by Traxler et al. (2017) Blood 131, 2733-2736 that shRNAs can often "non-specifically" inhibit erythroid terminal differentiation independent of their specific target – and the GRB10 knockdown phenotype with the single shRNA is inhibition of terminal differentiation.

4) Given the datasets generated, and the fact that GRB10 has been published as a GATA2 target gene in erythroid precursor cells and as a regulator of cKit signaling in HSCs (Yan et al., 2016), can additional loss of function studies be conducted to yield discoveries that considerably extend the existing work – for example to implicate proteins and/or pathways that have not been described as being important in the erythroid (or hematopoietic) system?

5) The Discussion was quite extensive, but does not seem to effectively integrate existing knowledge of translational and transcriptional mechanisms mediating iron- and heme-dependent cell regulation. Integration is important, since we know about various components, but very little is known about integration.

Reviewer #3:

The authors present extensive data concerning the differential translational of mRNA during erythtroid differentiation of mouse fetal liver cells in vitro. Although the results are interesting, they add little to the current knowledge vis-à-vis the main proteins involved in erythroid differentiation. The text of the manuscript is flooded with an overwhelming amount of data that is difficult to read, analyze and summarize. In other words, in the middle of so much data, it is hard pinpoint the main take-home message of the current article.

The authors claim that they have noticed only small changes in the translation of IRE-containing mRNAs when comparing normal and iron deficient conditions. The authors specifically mentioned that there are no differences in the translation of ALAS2 and TfR1 mRNA when compared normal and iron deficient conditions. The authors should extend the discussion regarding these findings, since they are in conflict with what has been shown in the literature in the past 30 years.

Moreover, it is necessary to better explain how the iron deficient condition was attained. According to the Materials and methods section, mice were kept on an iron deficient diet for an unspecified period, following which it was observed that the embryos obtained from these mice were "pale". The visual appearance of the embryos, although not unimportant, cannot be used as measure of the iron status of the embryos. A simple measurement of hemoglobin levels in the fetal livers would have directly answered that the cells were lacking iron.

Furthermore, the authors mentioned several papers published in 2015 and 2016 claiming that the protein NCOA4 is important for erythroid differentiation, since it is required for iron mobilization from ferritin. It is important to stress that these are questionable findings since the NCO4A knockout mice show hypochromic microcytic anemia only when exposed to an iron deficient diet; this is not at all surprising. In this context, unfortunately, the authors disregard extensive literature supporting the view that ferritin does not supply iron for heme and hemoglobin synthesis. These references are neglected by the authors when discussing the rate of translation of ferritin in their in vitro system.

[Editors’ note: what now follows is the decision letter after the authors submitted for further consideration.]

Thank you for resubmitting your work entitled "HRI coordinates translation necessary for protein homeostasis and mitochondrial function in erythropoiesis" for further consideration at eLife. Your revised article has been favorably evaluated by David Ron (Senior Editor), a Reviewing Editor, and 2 reviewers.

The manuscript has been improved and represents a significant extension of prior work on HRI function in erythropoiesis. As you will see from their comments (attached below) the reviewers find the new results of interest, but there are some remaining issues that need to be addressed before acceptance, as outlined below:

- As reviewer #1 points out, questions about mTORC1 activity in Hri-/- -Fe vs Hri-/- +Fe erythroid cells will need to be assessed.

- As reviewer #1 points out, questions about the strength of several statements on the role of HRI and mTORC1 in regulating translation of mitochondrial genes (point #2) and difference in mtDNA and mitochondrial mass (point #3) will need to be addressed with further clarifications.

- As reviewer #2 points out, the statement on GRB10 as a new regulator of erythropoiesis will need to be revised and include the relevant citation(s).

- All other reviewer concerns pertaining to strengthening existing data, discussion, clarifications and textual changes should also be addressed.

We therefore invite you to take these points into account when revising the manuscript. We look forward to receiving your revised manuscript and ask that you return it within four weeks. Please feel free to contact me if you require additional time or would like to discuss any of the reviewer comments further.

Reviewer #1:

In the revised manuscript, the authors removed some of the original results (i.e. IRE/IRPs and NCOA4) and shifted focus to an interesting new link between HRI-ATF4 and the translation of cytosolic and mitochondrial ribosomal proteins in the context of iron-deficiency (ID)-induced ineffective erythropoiesis. They also re-organized the presentation of the Ribo-seq and RNA-seq comparison in the first part of the manuscript. These efforts significantly improved both the strength of new conclusions and the clarity of the main messages. In particular, the new results support an important role of the heme-induced kinase HRI in coordinating integrated stress response (ISR), translation of cytosolic and mitochondrial proteins, and Epo-mTORC1 signaling during normal and iron-deficiency-induced ineffective erythropoiesis. These new findings significantly extend previous results, and provide new insights into the coordinated regulation of protein translation, mitochondria function, and ISR during normal and ineffective erythropoiesis. They also provide additional evidence for the role of GRB10 as a negative feedback regulator of Epo-mTORC1 and an ATF4 target gene during ID, and discuss other mechanisms (e.g. transcriptional regulation by ATF4 and/or GATA2) that may contribute to GRB10 regulation during erythropoiesis.

Overall, the authors provide a nice resource of transcriptional (mRNA-seq) and translational (Ribo-seq) changes in response to HRI loss during normal and ID-induced ineffective erythropoiesis, which will be of great interest to the erythropoiesis and a broader community. The new findings related to HRI-coordinated control of protein translation, ISR and mitochondrial function advance our current understanding of mechanisms that regulate normal and ineffective erythropoiesis.

There are several remaining questions and suggestions for area of improvement, which if addressed, will further enhance the strength of the conclusions and the potential impact of this study.

1) Is mTORC1 activity increased in Hri-/- -Fe erythroid cells compared to Hri-/- +Fe? From GSEA analysis of differentially translated mRNAs between Hri-/- -Fe and Wt -Fe (Figure 3C), it seems that mTORC1 signaling is significantly elevated, consistent with the role of HRI in negative regulation of mTORC1. However, it is unclear from the current analysis whether any changes in mTORC1 activity between Hri-/- -Fe and Hri-/- +Fe erythroid cells, or between Wt -Fe and Wt +Fe cells. Furthermore, the authors should consider measuring mTORC1 activity directly by the level of phosphorylated 4EBP1 and/or other downstream targets as in Figure 2E.

2) In the Discussion section (second paragraph), the authors discussed that the mitochondrial genes regulated by HRI are different from the targets of mTORC1, and suggested that HRI may regulate translation of mitochondrial genes independent of mTORC1. This statement is a bit confusing. The evidence supporting this statement is based on the lack of overlapped targets for mitochondria related proteins in Hri-/- -Fe vs mTORC1 (Figure 3D). Since HRI may negatively regulate mTORC1 through GRB10, one would expect to see some overlapped targets if HRI-GRB10-mTORC1 mechanism plays a significant role in erythropoiesis. Furthermore, HRI loss may also impact transcriptional programs and/or indirectly affect mitochondrial proteins independent of protein translation. The authors should revise the statement and/or provide additional clarification.

3) To determine the effect of Hri loss on mitochondria, the authors used mitotracker to estimate mitochondrial mass. It is important to note that mitotracker signal may also be affected by mitochondrial activity. In Figure 4—figure supplement 1, it seems that the mtDNA content is modestly increased in Hri-/- -Fe cells. Is the difference statistically significant? If the difference is significant, the results would suggest that Hri loss affect both mitochondrial biogenesis (mtDNA) and mitochondrial mass (or activity).

4) Some parts of the manuscript are still difficult to follow and confusing with various comparisons. The authors are advised to further improve the clarity and presentation of the manuscript by removing and/or simplifying non-essential texts, and focus more on main messages. The Discussion is extensive and could be further compressed.

5) Several typos and grammar errors are noted: Figure 3D legend "Venn" diagram, subsection “Increased cytoplasmic and mitochondrial protein synthesis but reduced mitochondrial respiratory activity in Hri-/--Fe EBs”, first paragraph etc.

Reviewer #2:

The manuscript is improved and represents a significant extension of prior work on HRI function in the context of erythropoiesis. Dr. Chen has conducted foundational work on HRI for many years, and this new work provides a new dimension that largely could not have been predicted from known mechanistic or biological concepts. As HRI was implicated recently by Gerd Blobel as a regulator of fetal hemoglobin production, there is renewed interest in the function of this important protein.

One minor, but significant, comment remains, which was not addressed in the revision.

The authors indicate in the cover letter and have related text in the manuscript that their study demonstrates: "These additional results strengthened our finding GRB10 as a previously unappreciated regulator of fetal liver erythropoiesis".

However, GRB10 was already published in Mehta et al., 2017, to be a direct GATA2 target gene in primary fetal liver erythroid precursor cells. Based on this work and known GRB10 biochemical activities, GRB10 is not a "previously unappreciated regulator of fetal liver erythropoiesis". This work should be cited in the context of what was already known about GRB10 in erythropoiesis.

eLife. 2019 Apr 29;8:e46976. doi: 10.7554/eLife.46976.029

Author response


[Editors’ note: the author responses to the first round of peer review follow.]

While the reviewers found that many aspects of your work were well done, they raise concerns about the degree to which it advances prior studies of similar mechanisms. They were also concerned about some aspects of the mechanisms, and the strength of conclusions that could be drawn based on loss-of-function studies with a single shRNA and experiments performed in culture. We believe your paper would be more appropriate for a more specialized journal.

The new manuscript is revised extensively and emphasizes on the translational regulation of HRI in iron deficiency (ID) to maintain protein homeostasis and mitochondrial function during erythropoiesis. Our global study supports an integrated model of heme and HRI regulated translation in developing erythroblasts for the adaptation to systemic ID (Figure 8). Activation of HRI in ID coordinates three distinct pathways of translation. First and foremost, inhibition of protein synthesis to avoid accumulation of unfolded proteins and maintain protein homeostasis. Second, induction of ATF4 mRNA translation to increase gene expression for mitochondrial unfolded protein response, redox homeostasis and metabolic reprogramming in order to maintain mitochondrial respiration and erythroid differentiation. Third, repression of Epo- mTORC1 signaling to inhibit protein synthesis and cell proliferation.

We provide new data on the novel findings of heme and HRI mediated translation of mRNAs of both cytosolic and mitochondrial ribosomal proteins in ID concomitant with inhibition of protein synthesis in both cellular compartments (Figure 3). The absence of HRI in ID elicits a prominent cytoplasmic unfolded protein response (UPR) in attempt to reduce proteotoxicity of excessive unfolded proteins produced as evidenced by the comprehensive induction of several families of cytoplasmic chaperones (Figure 5A-B).

Importantly, neither endoplasmic reticulum (ER) nor mitochondrial chaperones were induced. The lack of ER UPR is consistent with our earlier observation that erythroblasts do not display significant ER stress and that HRI is activated by cytoplasmic, but not ER stress (Lu L, Han AP, Chen JJ. Mol Cell Biol. 2001;21(23):7971-7980).

Integrated stress response (ISR) of eIF2a kinase has been shown to be activated in mitochondrial stress and mediates mitochondrial UPR. To date, the eIF2a kinase responsible for mitochondrial UPR remains unknown. The lack of mitochondrial UPR in HRI deficiency shown in our present study supports the notion that HRI is necessary for the induction of mitochondrial UPR in ID.

We also discover that Hri -/- primary erythroid cells have impairment of mitochondrial respiration (Figure 4). Importantly, ATF4 target genes are most highly activated during ID including serine-glycine biosynthesis, one carbon metabolism and redox homeostasis, for the maintenance of mitochondrial function and enable erythroid differentiation. We now have two shRNAs for Grb10 knockdown study, and showed that the additional shRNA_G7 worked similarly to shRNA_G3 in increasing cell proliferation and inhibiting erythroid

differentiation (Figure 7). These additional results strengthened our finding of GRB10 as a previously unappreciated regulator of fetal liver erythropoiesis connecting between HRI and erythropoietin signaling during ID.

Reviewer #1:

[…] 1) It is difficult to follow and somewhat confusing on the various comparisons of the ribosome profiling and mRNA-seq in WT+Fe, HRI-/-+Fe, WT-Fe and HRI-/--Fe (Figures 2-4, Figure 2—figure supplement 1). It would be helpful if the authors could summarize the findings in a diagram to list the most differentially regulated pathways and/or genes in the pairwise comparison. This diagram can be combined in Figure 7 or as a separate figure.

We thank the reviewer for the suggestion, and have expand the diagram in the model (Figure 8 in the updated version) summarizes our findings of ribosome profiling and mRNA-seq. This figure is referred in the Results section for the clarity of data presented.

2) When comparing differentially expressed mRNAs in HRI and iron deficiencies, the authors identified Grb10 as one of the most upregulated ATF4 target genes in WT relative to HRI-/- basophilic erythroblasts. To establish the functional role of Grb10 in erythropoiesis, a single shRNA (G3) was used to knockdown Grb10 expression in ex vivo cultured erythroid cells. The authors measured phosphorylated AKTS473, which is slightly increased, as the evidence for the role of Grb10 as the negative regulator of mTORC1 signaling. It is important to characterize in greater details how Grb10 regulate mTORC1 or other pathways in erythropoiesis. The authors cited previous studies (Plasschaet and Bartolomer, 2015, and Yan et al., 2016) as evidence that Grb10 is part of a feedback mechanism to inhibit mTORC1, however it is important to determine whether similar or different mechanism may be utilized in the context of erythropoiesis. How does ATF4 regulate Grb10 transcription (or translation) in erythroid cells? Moreover, additional shRNAs should be screened to validate the original findings. Since this part of the study could provide potentially new insights compared to the already established HRI-ATF4 pathway in erythropoiesis, more in-depth mechanistic studies are required.

We now have an additional shRNA_G7, which maintained knockdown efficiency at differentiation stage. Similar results were obtained by G7 and G3 shRNA in increasing cell proliferation and inhibition of erythroid differentiation (Figure 7). We found that protein level of cyclin D3, which is a known target of mTORC1 signaling, was increased upon Grb10 knockdown, supporting the role of Grb10 in repressing mTORC1 pathway. These are incorporated into the updated manuscript.

Our ribosome profiling and mRNA-seq results showed that Grb10 was regulated at transcription and not translation by ATF4. This is consistent with a recent study demonstrating that Grb10 expression is activated upon acute ER stress through ATF4 (Luo et al. 2018, J. Mol. Endocrinol. 60, 285).

We agree with the reviewer that it will be important to determine the molecular mechanism by which ATF4 regulates Grb10 expression as well as by which GRB10 regulates mTORC1 signaling. We feel these are outside the scope of the present study. In the updated manuscript, we focused on the global assessment and impact by HRI-mediated translation on ribosome synthesis and mitochondrial function.

3) Since all the functional studies were performed using ex vivo cultured erythroid cells with defined medium and cytokines, the functional importance of the identified pathway (HRI-ATF4-Grb10) should be assessed in more physiologically relevant settings in vivo, such as in the already established Grb10 knockout mouse models or in the transplant setting.

We agree with the reviewer, but feel it is also outside the scope of this study in the updated manuscript as described above.

Reviewer #2:

[…] 1) Subsection “Upregulation of in vivo translation of ISR mRNAs in Wt EBs compared to Hri-/- EBs”, last paragraph – The authors conclude that diet-induced iron deficiency does not significantly affect gene expression through IREs/IPRs in primary erythroblasts. I do not see how the global measurements and correlations support this rather strong conclusion. At the very least, this needs to be fleshed out more effectively.

We have removed IRE/IRP results from the updated manuscript. Perhaps, much more severe ID is necessary to detect IRE/IRP mediated translation in erythroid cells. The major point to be conveyed in our manuscript is that HRI plays a major role in gene expression of EBs during ID. Globally, translation of ribosomal proteins, both cytosolic and mitochondrial, is the most highly differentially regulated between Wt and Hri-/-EBs in ID. At the mRNA level, ATF4-ISR target genes were most highly activated in Wt-Fe EBs compared to Hri-/--Fe EBs.

2) "ATF4 target gene GRB10". GRB10 has been published by Mehta et al., 2017, to be a direct GATA2 target gene in primary mouse erythroid precursor cells. This may influence thinking regarding the role of GRB10 in mechanisms of interest to the authors.

We thank reviewer for bring our attention to this line of thoughts. In the study of Mehta et al. Grb10 was suggested to participate in the transition of BFU-E to CFU-E, which are earlier in erythroid differentiation than the EBs that we studied here. We will keep this in mind in our future study of erythropoiesis at early stage of erythropoiesis.

Since last submission of our manuscript, there was a recent report showing the ATF4-dependent Grb10 induction upon ER stress as responded above to reviewer 1. Our results also show decreased Grb10 expression during ex vivodifferentiation in Hri-/-, eAA and Atf4-/- fetal liver progenitors compared to Wt. Thus, there is a strong experimental support for the role of Grb10 in terminal erythropoiesis, where GATA2 expression is low.

3) Figure 6 – As noted above, the finding that Integrated Stress Response and ATF4 are important players does not transform existing knowledge. A limited loss of function analysis was performed with retrovirally-expressed shRNAs to evaluate GRB10 function. Only one shRNA was effective in reducing GRB10 levels in the differentiation culture. It will be important not to rely on results obtained with a single shRNA, especially given the recent findings by Traxler et al. (2017) Blood 131, 2733-2736 that shRNAs can often "non-specifically" inhibit erythroid terminal differentiation independent of their specific target – and the GRB10 knockdown phenotype with the single shRNA is inhibition of terminal differentiation.

We agree with the reviewer and have an additional shRNA_G7, which worked similarly to G3.

4) Given the datasets generated, and the fact that GRB10 has been published as a GATA2 target gene in erythroid precursor cells and as a regulator of cKit signaling in HSCs (Yan et al., 2016), can additional loss of function studies be conducted to yield discoveries that considerably extend the existing work – for example to implicate proteins and/or pathways that have not been described as being important in the erythroid (or hematopoietic) system?

We thank the reviewer for the interesting suggestion. We believe this is outside the scope of this study presently as addressed above in response to the comments of reviewer 1, point 2. In updated and extensively revised manuscript, we focused on the global assessment and impact by HRI-mediated translation on ribosome synthesis and mitochondrial function.

5) The Discussion was quite extensive, but does not seem to effectively integrate existing knowledge of translational and transcriptional mechanisms mediating iron- and heme-dependent cell regulation. Integration is important, since we know about various components, but very little is known about integration.

We have rewritten the Discussion and presented adistinct model of heme and HRI-regulated translational control that involves several separate and interacting pathways for the adaptation to systemic ID (Figure 8). Activation of HRI in ID elicits three distinct pathways of translation as described in the Discussion of the revised manuscript.

Reviewer #3:

The authors present extensive data concerning the differential translational of mRNA during erythtroid differentiation of mouse fetal liver cells in vitro. Although the results are interesting, they add little to the current knowledge vis-à-vis the main proteins involved in erythroid differentiation. The text of the manuscript is flooded with an overwhelming amount of data that is difficult to read, analyze and summarize. In other words, in the middle of so much data, it is hard pinpoint the main take-home message of the current article.

We appreciate the helpful comments and have revised the manuscript extensively, not only in writing, but also in the presentation in figures.

The authors claim that they have noticed only small changes in the translation of IRE-containing mRNAs when comparing normal and iron deficient conditions. The authors specifically mentioned that there are no differences in the translation of ALAS2 and TfR1 mRNA when compared normal and iron deficient conditions. The authors should extend the discussion regarding these findings, since they are in conflict with what has been shown in the literature in the past 30 years.

This is addressed above in response to comments of Reviewer 2, point 1.

Moreover, it is necessary to better explain how the iron deficient condition was attained. According to the Materials and methods section, mice were kept on an iron deficient diet for an unspecified period, following which it was observed that the embryos obtained from these mice were "pale". The visual appearance of the embryos, although not unimportant, cannot be used as measure of the iron status of the embryos. A simple measurement of hemoglobin levels in the fetal livers would have directly answered that the cells were lacking iron.

We have added the details of inducing ID in developing embryos in the Materials and methods. We have shown in our earlier publication that under our iron deficient conditions, E14.5 Hri-/-embryos were anemic with prolong primitive erythrocytes in the embryonic blood.

Furthermore, inclusions were observed in both primitive and definitive erythrocytes (Liu et al., 2008). Hri-/-embryos in ID died at E17.5.

Furthermore, the authors mentioned several papers published in 2015 and 2016 claiming that the protein NCOA4 is important for erythroid differentiation, since it is required for iron mobilization from ferritin. It is important to stress that these are questionable findings since the NCO4A knockout mice show hypochromic microcytic anemia only when exposed to an iron deficient diet; this is not at all surprising. In this context, unfortunately, the authors disregard extensive literature supporting the view that ferritin does not supply iron for heme and hemoglobin synthesis. These references are neglected by the authors when discussing the rate of translation of ferritin in their in vitro system.

We have removed the discussion about IRE/IRP and NCOA4 in the updated manuscript and focus on HRI pathway.

[Editors' note: the author responses to the re-review follow.]

Reviewer #1:

[…] 1) Is mTORC1 activity increased in Hri-/- -Fe erythroid cells compared to Hri-/- +Fe? From GSEA analysis of differentially translated mRNAs between Hri-/- -Fe and Wt -Fe (Figure 3C), it seems that mTORC1 signaling is significantly elevated, consistent with the role of HRI in negative regulation of mTORC1. However, it is unclear from the current analysis whether any changes in mTORC1 activity between Hri-/- -Fe and Hri-/- +Fe erythroid cells, or between Wt -Fe and Wt +Fe cells. Furthermore, the authors should consider measuring mTORC1 activity directly by the level of phosphorylated 4EBP1 and/or other downstream targets as in Figure 2E.

mTORC1 activity in Hri+/+ and Hri-/- erythroid precursors in +Fe and -Fe conditions were published last year in Zhang et al., 2018 (https://doi.org/10.1182/blood-2017-08-799908, Supplementary Figure 3A). mTORC1 activities as measured by pS6 and p4EBP1 were increased in primary Hri-/--Fe erythroid cells from blood, bone marrow and spleen as compared to Hri-/-+Fe cells. This is consistent with the mild erythroid phenotypes of Hri-/- mice in +Fe as compared to -Fe conditions.

TE analysis of ribosome profiling data in the present manuscript also showed increased TE for mRNAs of mTORC1 targets such as cytosolic ribosomal proteins, Eef1a1, Eef2, and Pabpc1 in Hri-/--Fe cells as compared to Hri-/-+Fe (Supplementary file 1B). These results support the increased mTORC1 activity in Hri-/--Fe cells as compared to Hri-/-+Fe cells. In contrast, TE for mTORC1 target mRNAs was not increased in Wt-Fe cells as compared to Wt+Fe cells (Supplementary file 1B).

2) In the Discussion section (second paragraph), the authors discussed that the mitochondrial genes regulated by HRI are different from the targets of mTORC1, and suggested that HRI may regulate translation of mitochondrial genes independent of mTORC1. This statement is a bit confusing. The evidence supporting this statement is based on the lack of overlapped targets for mitochondria related proteins in Hri-/- -Fe vs mTORC1 (Figure 3D). Since HRI may negatively regulate mTORC1 through GRB10, one would expect to see some overlapped targets if HRI-GRB10-mTORC1 mechanism plays a significant role in erythropoiesis. Furthermore, HRI loss may also impact transcriptional programs and/or indirectly affect mitochondrial proteins independent of protein translation. The authors should revise the statement and/or provide additional clarification.

We thank reviewer for pointing this out. We have employed INK128, a mTORC1 inhibitor to discern the contribution of HRI directly in the regulation of mitochondrial protein synthesis. As compared to Wt-Fe cells, Hri-/--Fe cells has increased mitochondrial protein synthesis, which was inhibited by INK128 to about 50% of the control (Figure 3—figure supplement 3A). However, mTORC1 activities as measured by pS6 and p4EBP1, were completely inhibited by INK128 treatment (Figure 3—figure supplement 3B). These results demonstrate that HRI also inhibits directly mitochondrial protein synthesis in addition to repress mTORC1 activity during iron deficiency. Thus, both HRI-eIF2αP and mTORC1 contribute to regulation of mitochondrial protein synthesis in erythroid cells during iron deficiency (subsection “Increased cytoplasmic and mitochondrial protein synthesis but reduced mitochondrial respiratory activity in Hri-/- -Fe EBs”, first paragraph and Discussion).

We have now modified the statement in Discussion to “both HRI-eIF2αP and mTORC1 pathways contribute to increased mitochondrial protein synthesis in Hri-/--Fe EBs”.

3) To determine the effect of Hri loss on mitochondria, the authors used mitotracker to estimate mitochondrial mass. It is important to note that mitotracker signal may also be affected by mitochondrial activity. In Figure 4—figure supplement 1, it seems that the mtDNA content is modestly increased in Hri-/- -Fe cells. Is the difference statistically significant? If the difference is significant, the results would suggest that Hri loss affect both mitochondrial biogenesis (mtDNA) and mitochondrial mass (or activity).

The difference in mtDNA content was not statistically significant as stated in the manuscript.

4) Some parts of the manuscript are still difficult to follow and confusing with various comparisons. The authors are advised to further improve the clarity and presentation of the manuscript by removing and/or simplifying non-essential texts, and focus more on main messages.

We have removed the comparison of ISR and Epo target genes (Figure 5C and 5E) from the main text to supplementary materials (Figure 6—figure supplement 1A and D) to simplify the main text and to improve clarity.

The Discussion is extensive and could be further compressed.

We have compressed the Discussion in the revised manuscript.

5) Several typos and grammar errors are noted: Figure 3D legend "Venn" diagram, subsection “Increased cytoplasmic and mitochondrial protein synthesis but reduced mitochondrial respiratory activity in Hri-/--Fe EBs”, first paragraph etc.

We have corrected the typo and grammar errors.

Reviewer #2:

[…] One minor, but significant, comment remains, which was not addressed in the revision.

The authors indicate in the cover letter and have related text in the manuscript that their study demonstrates: "These additional results strengthened our finding GRB10 as a previously unappreciated regulator of fetal liver erythropoiesis".

However, GRB10 was already published in Mehta et al., 2017 to be a direct GATA2 target gene in primary fetal liver erythroid precursor cells. Based on this work and known GRB10 biochemical activities, GRB10 is not a "previously unappreciated regulator of fetal liver erythropoiesis". This work should be cited in the context of what was already known about GRB10 in erythropoiesis.

We thank the reviewer for the comments. We have modified the Abstract to state that GRB10 as a previously unappreciated regulator of the terminal erythroid differentiation in fetal liver erythropoiesis. We have also included the citation and the role of Grb10 in the Discussion. Grb10 has also been shown to be a GATA2 target gene and play a role in the transition from BFU-E to CFU-E (Mehta et al., 2017). Our results support an additional role of ATF4 induced Grb10 expression in terminal erythropoiesis in which GATA2 expression is low.

Associated Data

    This section collects any data citations, data availability statements, or supplementary materials included in this article.

    Data Citations

    1. Zhang S, Chen J. 2018. Iron and Heme Coordinate Erythropoiesis through HRI-Mediated Regulation of Protein Translation and Gene Expression. NCBI Gene Expression Omnibus. GSE119365

    Supplementary Materials

    Supplementary file 1. List of diifferentially translated mRNAs by HRI and mTORC1.

    Supplementary File 1a. Quality control and mapping of Ribo-seq and mRNA-seq. Supplementary File 1b. Complete gene list of differentially translated mRNAs. Supplementary File 1c. Eif2ak1, Ppp1r15a and Atf4 mRNAs are highly expressed in basophilic erythroblasts. Supplementary File 1d. Higher ribosome occupancy at the AUG of uORF1 in Atf4 mRNA.Supplementary File 1e. Complete gene list of differentially expressed mRNAs.

    elife-46976-supp1.xls (2.2MB, xls)
    DOI: 10.7554/eLife.46976.021
    Supplementary file 2. List of mRNAs that are differentially translated between Hri–/– –Fe EBs and Wt –Fe EBs and known mTORC1 targets in the categories of (a) RPs and non-RPs and (b) mitochondrial proteins.
    elife-46976-supp2.xls (43.5KB, xls)
    DOI: 10.7554/eLife.46976.022
    Supplementary file 3. List of RNA primers, antibodies and ShRNA oligonucleotides.

    Supplementary File 3a. DNA sequence of qPCR primers. Supplementary File 3b. List of antibodies used for western blot analyses. Supplementary File 3c. DNA sequences of shRNA oligonucleotides.

    elife-46976-supp3.xls (32KB, xls)
    DOI: 10.7554/eLife.46976.023
    Transparent reporting form
    DOI: 10.7554/eLife.46976.024

    Data Availability Statement

    All sequencing data have been deposited in GEO under accession code GSE119365.

    The following dataset was generated:

    Zhang S, Chen J. 2018. Iron and Heme Coordinate Erythropoiesis through HRI-Mediated Regulation of Protein Translation and Gene Expression. NCBI Gene Expression Omnibus. GSE119365


    Articles from eLife are provided here courtesy of eLife Sciences Publications, Ltd

    RESOURCES