Targeting elevated heme levels to treat a mouse model for Diamond-Blackfan Anemia

Abstract

Diamond-Blackfan anemia (DBA) is a rare genetic disorder in which patients present a scarcity of erythroid precursors in an otherwise normocellular bone marrow. Most, but not all, patients carry mutations in ribosomal proteins such as RPS19, suggesting that compromised mRNA translation and ribosomal stress are pathogenic mechanisms causing depletion of erythroid precursors. To gain further insight to disease mechanisms in DBA, we performed a custom short hairpin RNA (shRNA) based screen against 750 genes hypothesized to affect DBA pathophysiology. Among the hits were two shRNAs against the erythroid specific heme-regulated eIF2a kinase (HRI), which is a negative regulator of mRNA translation. This study shows that shRNA-mediated HRI silencing or loss of one HRI allele improves expansion of Rps19-deficient erythroid precursors, as well as improves the anemic phenotype in Rps19-deficient animals. We found that Rps19-deficient erythroblasts have elevated levels of unbound intracellular heme, which is normalized by HRI heterozygosity. Additionally, targeting elevated heme levels by treating cells with the heme scavenger alpha-1-microglobulin (A1M), increased proliferation of Rps19-deficient erythroid precursors and decreased heme levels in a disease-specific manner. HRI heterozygosity, but not A1M treatment, also decreased the elevated p53 activity observed in Rps19-deficient cells, indicating that p53 activation is caused by ribosomal stress and aberrant mRNA translation and not heme overload in Rps19-deficiency. Together, these findings suggest that targeting elevated heme levels is a promising new treatment strategy for DBA.

Graphical Abstract

INTRODUCTION

Diamond-Blackfan anemia (DBA) is a rare congenital disease characterized by macrocytic anemia and lack of erythroid precursors in an otherwise normocellular bone marrow [1]. The majority of all DBA patients have mutations causing haploinsufficiency of a specific ribosomal protein, of which the most common is ribosomal protein S19 (RPS19) [2]. Ribosomal protein deficiency leads to impaired ribosome synthesis, including disturbed ribosomal RNA processing [3,4]. The only pharmacological therapy available for patients suffering from DBA is life-long glucocorticoid treatment, which promotes erythropoiesis in DBA by mechanisms involving reduced premature erythroid maturation and depressed p53 activation⁵. Patients unresponsive to this treatment are referred to chronic blood transfusion therapy or allogenic stem cell transplantation [1,6]. Unfortunately, all current DBA treatments are associated with severe adverse effects, and therefore a novel treatment targeting the underlying mechanisms causing the anemia is urgently needed. In order to develop safer and more effective therapies it is essential to first identify novel drug targets. To develop such treatments it is vital to gain further understanding of mechanisms behind DBA pathology.

Mechanistic studies of DBA pathology have revealed increased activity of the tumor suppressor p53 in DBA, which results in marked reduction of expansion capacity and increase in apoptosis in erythroid cells [5],[7–10]. The elevated p53 levels in DBA could be a result of failing ribosome assembly, which leads to release of free ribosomal proteins that inhibit human double minute 2 homolog (HDM2), the principal negative regulator of p53 [11]. In addition, the ribosome deficiency in DBA causes reduced capacity for mRNA translation in general, and translation of certain transcripts such as the erythroid transcription factor Gata1 is particularly affected [12–14]. The mRNAs predominantly translated in cells affected by DBA are globin mRNAs. It has therefore been proposed that the protein component of hemoglobin becomes limited in DBA, leading to accumulation of the other component, heme. Indeed, elevated heme levels have been found in erythroid cells from one DBA patient, indicating that the balance between globin and heme is distorted in erythroid precursors and potentially contributes to the failing erythropoiesis [15].

In order to develop and evaluate therapies for DBA we previously generated a doxycycline-inducible mouse model for Rps19-deficient DBA [16]. Upon induction there is a down regulation of Rps19 expression that results in a proliferative block in erythroid development. Since this inducible DBA mouse model recapitulates key features of the human disease it is an excellent tool for identifying and characterizing chemical compounds, genes and molecular pathways to be utilized for DBA treatment [17]. Here we utilized erythroid progenitors from this inducible DBA model to perform a targeted in vitro shRNA suppressor screen to identify novel drug targets for DBA therapy.

METHODS

Custom shRNA screen and deep sequencing

A pooled lentiviral shRNA library was generated [18] at the Broad Institute containing 3771 shRNA targeting 754 genes selected on the basis of involvement in either the p53, ribosomal stress, Akt/mTOR pathways or suggested by others to modify Rps19-deficiency [19] (and unpublished data shared by Dr. Stefan Karlsson’s lab). Five shRNA per gene were used. E13.5–14.5 fetal livers from a Rps19-deficient mouse model [16] were extracted as described previously [5] and transduced with the library before Rps19-deficiency was induced with 0.5–1μg/ml Doxycycline (Sigma-Aldrich, Saint Louis, US) to half of all samples. A total of ten replicates per condition were performed in separate experiments and cell samples were obtained at day 3 and 7 of culture. The samples were sequenced at Broad Institute using next-generation sequencing described earlier [20]. The mapped shRNA counts were normalized into counts per million (CPM) and a pseudocount of 1 was added to all CPM values. The CPM values were log2 transformed and the frequency distribution of all day 3 shRNA log2(CPM) values was evaluated by histogram analysis. A day 3 median log2(CPM)≥4 cut-off was used for individual shRNAs for further statistical analysis. Statistical analyses of proliferation rescue and changes in hairpin abundance at gene level during culture were performed by comparing consolidated paired log2(CPM day7/day3) values for all shRNAs targeting the same gene against consolidated Luciferase shRNA values using a non-parametric Kruskal-Wallis test. Statistical analyses of changes in individual hairpin abundance during culture were performed by comparing log2(CPM) values at day 3 and 7 using non-parametric Wilcoxon matched-pairs signed rank tests. A gene was considered a hit if it had a significantly higher (p≤0,05) consolidated proliferation/shRNA abundance than shRNAs targeting Luciferase at day 7 compared to day 3, as well as if at least two individual shRNAs were significantly enriched (p≤0,05) at day 7 compared to day 3. P-values were not adjusted for multiple comparisons.

Mice

All animal work was approved by the local Swedish ethical committee for experimental animal research. Rps19-deficient mice [16] and HRI −/− mice [21] were crossed for at least 5 generations, here referred to as Rps19-HRI mice. Rps19-deficiency is induced by Doxycycline administration that together with a constitutively active trans-activator rtTA drives the expression of a shRNA against Rps19. For all experiments mice homozygote for the rtTA trans-activator was used. All mice had either wild type expression of HRI (HRI+/+) or were heterozygote (HRI+/−). 7–9 weeks old wild type C57/BL6 females were irradiated (900cGy) and transplanted with 2 million total bone marrow cells from mice either WT or heterozygote for Rps19 shRNA (via tail vein injection). Mice received Ciprofloxacin 125mg/ml (Arrow Generics, Kilmersdon, UK) 14 days post transplantation. After 8–11 weeks of engraftment Rps19-deficiency was induced for 10–11 days by 0.5mg/ml Doxycycline (Sigma-Aldrich, Saint Louis, US) and 10mg/ml sucrose in drinking water. Blood samples were collected from tail vein 7–14 days before and 9–10 days after Doxycycline administration and analyzed on KX-21N analyzer (Sysmex, Kobe, Japan).

Erythroid cell culture

Fetal livers from E13.5–14.5 or bone marrow from adult mice were made into single cell suspension and enriched for CD117⁺, and fetal liver cells were cultured as described previously [5] with or without 0.5–1μg/ml Doxycycline. Bone marrow cells were cultured in SFEM (Stem cell technologies, Vancouver, Canada), 1% Penicillin/Streptavidin (GE Healthcare, Chicago, US), 10% fetal bovine serum (Thermo-Fisher Scientific, Waltham, US), 100ng/ml mSCF (Peprotech, Rocky Hill, US), 300μg/ml h-holo-transferrin (Sigma-Aldrich), and 2U/ml hEPO (Johnson-Johnson, New Brunswick, US) with 0.2μg/ml Doxycycline to induce Rps19-deficiency. For the drug screen, the compounds administered 24 hours after Doxycycline induction were recombinant human A1M [51,50], N-methyl mesoporphyrin IX (AH Diagnostics, Aarhus, Denmark), Succinylacetone, Hemin, Deferoxamine, Ferrostatin and N-acetyl-L-cysteine and Hemopexin (all from Sigma-Aldrich). Cell expansion 5 days after drug administration was measured using CellTiter Glo (Promega, Madison, US), which measures the number of viable and metabolically active cells in culture based on quantitation of the ATP present. Plates were read on Victor 3 Multilabel counter (PerkinElmer, Waltham, US).

A1M treatment of human samples

Peripheral blood samples were collected at Lund University hospital using informed consent according to ethical permission granted by the Swedish ethical review board. The three DBA patients had mutations in RPS19, RPS26 and RPL35a respectively, and all received blood transfusions with chelation therapy. Mononuclear cells from DBA patients and healthy subjects were obtained using lymphoprep (Fresenius Kabi, Bad Homburg, Germany) and enriched for CD34 expression using magnetic beads (Miltenyi, Bergisch Gladbach, Germany) according to manufacturer’s instructions. Cells were cultured in SFEM (Stem cell technologies), 1% penicillin/streptomycin (GE Healthcare), 100ng/ml hSCF (Peprotech), 2U/ml Epo (Johnson-Johnson). 5μM A1M or vehicle (Tris-HCl/NaCl) were added every 2–3 days for a total of 7–8 days. Cell count was performed manually using a hemocytometer.

Flow cytometric analysis and cell sorting

Murine whole bone marrow cells were stained with antibodies against CD71 (BD Biosciences, Franklin Lakes, US), Ter119 (Thermi Fisher Scientific), B220, CD3, Gr1 (BioLegend, San Diego, US) on ice, washed and analyzed on Canto II, LSR II or LSR Fortessa (BD Biosciences). 7AAD was used as viability stain. Analysis of erythroblast fractions was done on CD71⁺ cells before entering the reticulocyte stage, after excluding cells positive for B220, CD3, Gr1 and 7AAD. Gating of erythroid stages of maturation was done according to reference [22], using Ter119 and CD71. Cell sorting was performed on AriaIII and AriaIIu (BD Biosciences). Analysis was done using FlowJo (FlowJo, Ashland, US). Analysis was performed using the FlowJo software (FlowJo).

Unbound intracellular heme measurement of bone marrow erythroblasts

Intracellular free heme was analyzed using the hemin colorimetric assay kit as described by the manufacturer (BioVision Inc., Milpitas, USA). CD71⁺ bone marrow cells were sorted in equal numbers on AriaIII or AriaIIu into lysis buffer (ThermoFisher Scientific, Waltham, US) with proteinase inhibitor cocktail (Roche, Basel, Switzerland), diluted in assay buffer and analyzed against a hemin standard curve. Samples and hemin standard utilized the peroxidase activity of non-protein bound hemin in order to measure color conversion measured at 570 nm.

Gene expression analyses

CD71⁺ cells from murine bone marrow were sorted (100 cells/well) in 96-well plates containing 4ml/well of lysis buffer, and were snap frozen to −80°C. Control wells with 0, 20, 50 or 100 cells were included. Thawed plates were processed according to published protocol [23] where cDNA was pre-amplified in multiplex using gene specific Taqmanprimers and Taq/SSIII reaction mix (Invitrogen, Waltham, US), with an initial 50°C incubation step for the reverse transcription (RT). Gene expression analysis simultaneously on diluted samples was enabled using the 48.48 Dynamic array (Fluidigm, South San Francisco, US) and the Biomark HD platform (Fluidigm). The RT quantitative PCR (RT-qPCR) was run according to published protocol. RNA extraction was done using the RNeasy micro kit (Qiagen, Hilden, Germany) according to the manufacturers’ instructions. Analysis was performed on a 7900 HT fast RT-PCR (ThermoFisher Scientific) and the primers used were Actin (#Mm02619580_g1), and HRI (#Mm01202300_m1), also from Thermi Fischer Scientific.

Statistical analysis

Statistical analyses were performed in GraphPad Prism 7 (GraphPad Software Inc., San Diego, US) using non-parametric analysis by Mann Whitney, or Kruskal-Wallis with Dunn’s multiple comparison test, unless otherwise stated. For multiple analyses, the groups compared have been 1) WT HRI +/+ to WT HRI +/−, 2) WT HRI +/+ to Rps19-HRI +/+, 3) WT-HRI +/− to Rps19-HRI +/−, and 4) Rps19-HRI +/+ to Rps19-HRI +/−. Graphs display mean with standard deviation (SD), unless otherwise stated. P-values: *≤0,05, **≤0,01, ***≤0,001, ****≤0,0001.

RESULTS

Custom shRNA screen was used to identify disease-modifying genes in DBA

To identify disease modifying genes capable of increasing expansion capacity of these precursors, a custom short hairpin RNA (shRNA)-based screen against 750 genes was performed on CD117⁺ (Kit) E13.5–14.5 fetal liver cells from our inducible Rps19-deficient mouse model [16] cultured under erythroid promoting conditions (Figure 1A). The genes targeted by the shRNA library were selected based on known roles in the p53 pathway, ribosomal stress pathways, Akt/mTOR pathways or suggested by others to modify ribosomal protein deficiency [19] (and unpublished data shared by Dr. Stefan Karlsson’s lab, Lund University, Sweden). DBA patients have insufficient numbers of erythroid precursors in the bone marrow, and these cells show a dramatically reduced ability to survive and expand in culture, which is also observed in cells from a murine DBA model (Figure 1B). This makes rescue of proliferative capacity a promising readout in search for new DBA therapies [5,24,25]. Harvested fetal livers from 10 independent donors were transduced with the shRNA library and cultured for 7 days. At day 3 and 7, cells were collected and genomic DNA were isolated and deep-sequenced to evaluate changes in hairpin abundance. For the majority of the hairpins included in the screen, infection was stable and reproducible. More than 96% of all hairpins were detected above the cut-off level of at least four log2 counts per million (CPM), with a median of 7.84 (Figure S1A). To identify gene hits in the screen, log2(CPM day7/day3) values from shRNAs targeting the same gene were consolidated and compared against Luciferase-targeting shRNAs, serving as negative controls. Due to the poor proliferative capacity of non-treated cells, 49 genes targeted in the screen had a significantly higher (p≤0,05, unadj) shRNA abundance at day 7 compared to the Luciferase control (Figure S1B). Not surprisingly, the included positive shRNA controls targeting p53 showed amongst the highest increases in hairpin abundance over time (Figure 1C). To reduce the risk of false positives and shRNA off-target effects as confounders, genes were also evaluated based on the effect of individual shRNAs. A gene would only be considered a hit in the screen if it had two individually significantly enriched (p≤0,05, unadj) shRNAs (Figure 1D). Applying these stringent criteria resulted in 14 of 750 genes being eligible for validation (Figure S1C, see Table S1 for unfiltered screen results). Amongst the hits were HRI, with two promising individual hairpins to validate.

Figure 1.

HRI silencing increase expansion of Rps19-deficient cells A) E13.5–14.5 CD117 (Kit)⁺ fetal liver cells were transduced with shRNA library at a multiplicity of infection (MOI) of 5–10 and cultured over 7 days in erythroid promoting medium. Cells were collected at day 3 and 7, DNA extracted and deep sequencing performed to analyze changes in shRNA abundance over time. Screen hits were validated in independent experiments. B) Proliferation of Rps19-deficient Kit⁺ murine bone marrow cells in erythroid promoting medium. The statistical analysis used was RM two-way ANOVA with Sidak’s multiple comparisons test, and shown are mean and SD C) Genes ranked from highest enriched to most depleted. Median log2(CPM day7/day3) values represents the consolidated weighted effect of all shRNAs (n=10). D) Changes in individual hairpin abundance was evaluated by comparing log2(CPM) values at day 7 and 3. Data presented using sum of signed ranks and −log10(p-values). Multiple Wilcoxon matched-pairs signed rank tests were performed to determine unadjusted p-values (n=10). E) Expansion of Rps19-deficient E13.5–14.5 Kit⁺ fetal liver cells in erythroid promoting culture transduced with two individual shRNA against HRI or luciferase. Mean values and SD are displayed. P-values: *≤0,05, **≤0,01, ***≤0,001, ****≤0,0001. F) Gene expression of HRI in E13.5–14.5 Kit⁺ fetal liver cells transduced with shRNA against luciferase or HRI, and cultured in erythroid promoting media for 72 hours. Expression was normalized to Actin and the statistical analysis is Kruskal-Wallis test with Dunn’s multiple comparison was done, comparing both HRI bars to luciferase. The graph displays mean values and SD.

Heme-regulated inhibitor (HRI) silencing increases expansion of Rps19-deficient erythroblasts

The screen picked up shRNAs targeting the erythroid specific heme regulated-inhibitor (HRI) as particularly interesting, due to the known role of HRI as the main negative regulator of mRNA translation in erythroid precursors [21]. Additionally, HRI has favorable drug target properties since small molecules can be targeted both to the heme- and ATP-binding pockets [26]. When erythroid precursors face insufficient heme levels, HRI phosphorylates serine 51 of eIF2α and thereby inhibits eIF2α-mediated translation [27,28]. When reducing HRI mRNA expression by 60–70% in Rps19-deficient erythroid precursors, the proliferation was significantly increased, further confirming the positive effect of reducing HRI expression in these cells (Figure 1E–F).

HRI heterozygosity improves erythroid bone marrow phenotype in Rps19-deficient mice

Mutations in ribosomal proteins result in deficient ribosomal assembly and has shown to alter both polysome profiling and translation of certain transcripts [12,13], [29–31]. Since HRI −/− mice show increased mRNA translation and a 7-fold increase in globin production [21], we hypothesize that increasing mRNA translation could be beneficial for Rps19-deficient erythroid precursors and potentially rescue the anemia of an Rps19-deficient mouse model. To examine this in vivo, we crossed the inducible Rps19-deficient mouse [16] with a conventional HRI knock out mouse [21], generating an Rps19-HRI mouse (Figure 2A). As shown previously, Rps19-deficient mice show significant decreases in red blood cell count, hematocrit and hemoglobin levels. While not statistically significant, deletion of one HRI allele instead caused a slightly milder anemia in Rps19-deficient mice (Figure S2). Importantly however, flow cytometric analysis of the bone marrow revealed a significant increase in all erythroblast stages in Rps19-deficient HRI+/−mice, compared to Rps19-deficient mice with intact HRI expression. Interestingly, this effect was not observed in WT HRI+/− mice (Figure 2B–C). Together, this data indicate that HRI heterozygosity increases the numbers of erythroid precursors in Rps19-deficient bone marrow.

Figure 2.

HRI heterozygosity increases the numbers of erythroid precursors in the bone marrow of Rps19-deficient mice A) An inducible Rps19-deficient mouse model for DBA was crossed to a conventional HRI knock out mouse to generate an Rps19-HRI mouse. B) Absolute numbers of erythroid precursors in the bone marrow of Rps19-HRI mice induced for 9–10 days to generate Rps19-deficiency is shown. Flow cytometric analysis of the pro- (I), basophilic (II), polychromatic (III), and orthochromatic (IV) erythroblast stages in the bone marrow and bone marrow cellularity from three individual experiments. C) Gates for CD71 and Ter119 expression was set on CD71⁺ bone marrow cells negative for B220, CD3, Gr1 and 7AAD. The gating strategy used to identify erythroblast populations have been adopted from [22].

Rps19-deficient erythroblasts have increased intracellular heme concentration

As mentioned previously, it is vital for developing erythroid cells to keep the balance between heme and globin tightly checked to avoid cellular toxicity. Increased heme levels have previously been reported in cells from one DBA patient, possibly contributing to the pathogenesis [15]. We therefore hypothesized that the balance between heme and globin is distorted in our DBA model. We further hypothesized that part of the increased proliferative capacity of Rps19-deficient cells upon HRI +/− in vitro and in vivo is explained by increased mRNA translation, which then normalize the levels of unbound heme. We examined the concentration of unbound intracellular heme in these cells and observed a dramatic increase of free heme in erythroid progenitors that were Rps19-deficient (Figure 3A). Interestingly, purified Rps19-deficient erythroblasts had a significantly lower intracellular heme concentration in a HRI+/− background. In contrast, HRI heterozygosity did not lead to reduction in unbound intracellular heme in cells with normal Rps19 expression. Heme synthesis is a well characterized stepwise process catalyzed by eight enzymes [32]. Four of eight heme synthesis enzymes were significantly upregulated in purified CD71⁺ Rps19-deficient erythroblasts, and among them the rate-limiting enzyme Alas2 (Figure 3B). Interestingly, HRI+/− decreased the expression of heme synthesis enzymes in Rps19-deficient erythroblasts, an effect not observed in cells with normal Rps19 expression. Together, this data show that free intracellular heme is increased in Rps19-deficient erythroid cells and that HRI+/− ameliorates this toxic condition. It also shows that multiple heme synthesis enzymes are upregulated in Rps19-deficient animals and that this is normalized in an HRI+/− background.

Figure 3.

HRI+/− ameliorates increased heme concentration and synthesis, as well as elevated p53 activity in Rps19-deficient animals A) Intracellular free heme concentration in sorted CD71⁺ erythroblasts from bone marrow of Rps19-HRI mice. n=3. B) Fluidigm-based gene expression of all eight enzymes governing heme synthesis. For gene expression analysis, 100 CD71⁺ erythroblasts from the bone marrow of the same mice as in Figure 2 were sorted using FACS. Mean values and SD are shown. C) Fluidigm-based gene expression of multiple p53 responsive genes on the same cells as in A and B. Gene expression was normalized to both Hprt and Actin, and displayed is mean with SD.

Elevated p53 activity of Rps19-deficient cells is dampened by HRI heterozygosity

Many studies have confirmed increased p53 activity in different types of ribosomal protein deficiency, and erythroid cells from Rps19-deficient mice display a dramatic increase in multiple p53-responsive genes [5], [7–10], [33]. Sorted CD71⁺ bone marrow erythroblasts from Rps19-deficient mice showed a significant upregulation of a panel of p53-responsive genes. However, all p53-responsive genes were upregulated to a much lesser extent in an Rps19-deficient HRI +/− background (Figure 3C). This was not observed in mice with normal Rps19 expression. Among the genes down regulated in the HRI +/− Rps19-deficient cells were the cell cycle regulators p21 and Ccng1. Reducing the up regulation of the p53-responsive genes improves survival and proliferative capacity of Rps19-deficient erythroid cells, as seen previously⁵. These results indicate that HRI heterozygosity dampens elevated p53 activity in Rps19-deficient erythroid cells. Since mRNA translation is elevated in HRI+/− cells, this suggests that inefficient mRNA translation is part of the mechanism leading to p53 activation in DBA.

Recombinant human alpha-1-microglobulin (A1M) increases proliferation and decreases elevated heme levels in erythroid Rps19-deficient cells without affecting p53 activity

To identify molecules able to decrease intracellular heme levels observed in Rps19-deficient erythroblasts pharmacologically, we performed a drug screen for compounds with potential to restore iron or heme availability in erythroid bone marrow cells from the Rps19-deficient mice (Figure 4A). Strikingly, A1M was the only compound tested that showed increased proliferation of Rps19-deficient erythroid precursors (Figure S3A). A1M administration had no effect on proliferation in WT cells, indicating a specific effect of A1M on Rps19-deficient erythroid precursors (Figure 4B). Importantly, A1M significantly reduced the level of unbound intracellular heme (Figure 4C). The expression of Alas2 remained unchanged by A1M treatment, indicating that A1M decreased the concentration of intracellular unbound heme without affecting heme synthesis (Figure S3B). Interestingly, A1M treatment also did not alter the expression of any of the p53 responsive genes tested, including p21 and Bax in either Rps19-deficient erythroid cells or WT controls, indicating a p53-independent mechanism behind A1M treatment (Figure S3C). Together, this data indicates that A1M treatment significantly lowers elevated heme levels in Rps19-deficient erythroid precursors, but not in healthy cells.

Figure 4.

A1M treatment improves expansion of erythroid cells from a murine DBA model and from DBA patients CD117 (Kit)+ bone marrow cells were treated with Doxycycline to induce Rps19-deficiency. Twenty four hours after Doxycycline administration, cells were treated with drugs interfering with iron or heme availability. The ATP measuring platform CellTiter Glo was used to monitor viable cells after 5 days of expansion. A) A schematic picture of the drug screen is shown. B) Treatment with A1M significantly improved proliferation of Rps19-deficient erythroid cells cultured in erythroid promoting media for 5 days. Shown here are mean and SD. Kruskal-Wallis non parametric test with Dunn’s multiple comparisons test was used for statistical analysis, and genotypes were compared to respective control, separately. P-values: *≤0,05, **≤0,01, ***≤0,001, ****≤0,0001. n=3. C) Concentration of free unbound intracellular heme in Kit⁺ bone marrow cells from the Rps19-deficient mouse expanded in erythroid culture with A1M or vehicle for 72 hours, displayed as mean with SD. Mann Whitney non parametric analysis was employed for statistical analysis within each genotype. D) Expansion of CD34⁺ erythroid precursors from peripheral blood of DBA patients with mutations in RPS19, RPL35a or RPS26, or healthy controls treated with A1M or vehicle in erythroid promoting media for 7 days. Cells from each donor is presented pairwise.

Early erythroid cells from DBA patients increase proliferation following A1M treatment

To confirm a positive effect by A1M treatment on proliferation in human erythroid cells, we cultured purified erythroid precursors from three DBA patients with 5μM A1M or vehicle. Erythroid precursors from all patients showed a trend towards improved expansion, while no such trend was observed in cells from healthy subjects (Figure 4D).

DISCUSSION

As a first step towards developing new disease specific therapies for DBA this study set out to identify new disease modifying genes through a targeted shRNA screen. Throughout several separate in vitro experiments, we found that the shRNA library quality and experimental design was robust and reproducible, supporting the conclusions of this study. In total, 750 genes were screened and 14 considered hits for further validation. One of those genes, HRI, stood out due to its erythroid specificity as well as its drug target potential [26]. Whereas Rps19-deficient mice have been reported to display a reduction in polysomal mRNA [16] and mRNA translation, HRI −/− and HRI +/− mice display the opposite phenotype with marked increase in globin synthesis [21]. Knock down of HRI by 60–70% in vitro using shRNAs had a dramatic effect on expansion of Rps19-deficient cells, rendering more than 7-fold more cells over 5 days of culture. When observing the relatively mild improvement of peripheral blood parameters in Rps19-deficient HRI+/− mice, we speculate a greater phenotype rescue is possible if HRI expression is lowered to levels below that of the heterozygote (50%) but above complete deletion. In our hands, Rps19-HRI−/− mice displayed a steady state anemia and were for this reason not further included in the study (Figure S4). It is known that HRI−/− mice are indeed very sensitive to low levels of iron, and insufficient iron intake could account for the steady state anemia we observe in Rps19-HRI−/− mice. To circumvent issues of HRI expression levels, treatment with an HRI inhibitor previously described [26], or a new small molecule inhibitor with high affinity binding to the heme-binding pocket of HRI could serve as a potential treatment option for DBA.

Our study identified elevated levels of unbound intracellular heme in Rps19-deficient erythroid cells. This is similar to what has been reported by Yang and colleagues in a DBA patient [15]. We indeed observe increased concentration of unbound intracellular heme in a distinct erythroid population in the bone marrow (sorted CD71⁺ erythroblasts of Rps19-deficient animals). Since cell division of maturing erythroblasts is highly regulated [36], it is likely that this sorted CD71⁺ erythroid population has the same distribution of pro-, basophilic-, polychromatic and orthochromatic erythroblasts in both WT and Rps19-deficient animals. Apart from being a heme sensor, HRI knock down has shown to increase γ-globin translation and thereby fetal hemoglobin (HbF) via the known globin switcher Bcl11a [37]. Interestingly, DBA patients are known to have higher levels of HbF in circulation¹, but whether this is due to HRI inhibition by excess heme levels remains to be investigated.

Heme is able to induce expression of many genes, and globin translation is induced by the presence of heme in erythroid cells. In this study we observed a significant increase in multiple heme synthesis enzymes in Rps19-deficient erythroblasts, among them the rate-limiting enzyme Alas2, indicating increased heme synthesis. Any distortion in the balance between heme and globin causes accumulation of either component, resulting in cell toxicity [21,38]. The relationship between heme synthesis and globin translation is illustrated by the increase in Alas2 and β-globin expression during erythroid maturation, accompanied by a decrease in the heme transporter feline leukemia virus subgroup C receptor (FLVCR), indicating that the need for heme export declines at increased globin translation [39]. Deletion of this exporter further underscores the harmful effects of excess cellular heme as FLVCR knock out mice display insufficient erythropoiesis, resembling the DBA phenotype [40]. Interestingly, we observed that much of the increase in heme concentration and heme enzyme expression was reversed by HRI+/− in Rps19-deficient erythroid cells. In Rps19-deficient erythroid cells, A1M treatment also significantly reduced the level of unbound intracellular heme. A1M protects against heme induced cell and tissue damage by scavenging and degrading heme [34,35,41] and a truncated form of A1M has been shown to have heme degrading properties[35,41] and is able to export heme from the cell membrane into the extracellular compartment [34]. Since treatment with the extracellular heme scavenger hemopexin had no effect on DBA cells, A1M possibly functions intracellularly on erythroid DBA cells. Although further studies are needed to unravel the mechanism by which A1M lowers free heme levels in a DBA context, this study for the first time shows that A1M is able to reduce intracellular heme in DBA. While not evaluated in this study, our findings suggest A1M has potential to reduce heme toxicity in other anemic conditions caused by ribosomal protein deficiency, such as del 5q-myelodysplastic syndrome [15].

Ribosomal protein mutations cause increased p53 activation in many, but not all, model systems for DBA [5], [7–10], [42–45]. Interestingly, p53 activity was not altered by A1M treatment, indicating that heme overload is not responsible for the increased p53 activity in Rps19-deficient cells. Instead, increasing mRNA translation by HRI+/− significantly reduced p53 activity in Rps19-deficient cells. Together with results from other studies linking ribosomal stress to an upregulation of p53 by Mdm2 inhibition [46–48], our studies support the idea that ribosomal stress and possibly aberrant mRNA translation are main causes of increased p53 activity.

Our study identified elevated levels of unbound intracellular heme and increased heme synthesis in purified CD71⁺ erythroblasts from Rps19-deficient mice. This is likely contributing to the reduced proliferative capacity and survival of these cells, causing subsequent anemia. We also demonstrated that haploinsufficiency of erythroid specific HRI both increased erythroid cell expansion and ameliorated the phenotype of Rps19-deficient mice. This was likely achieved through increasing mRNA translation including Rps19 and globin mRNAs [49] and thereby dampening both p53 activity and lowering heme toxicity, since more heme could be incorporated into hemoglobin. Here we propose a model of distorted balance between heme and globin that can be targeted by increasing mRNA translation by HRI heterozygosity or reducing heme levels by A1M treatment. In all, this study proposes two new treatment strategies for DBA by targeting mRNA translation or heme scavenging and thereby restoring the balance between globin and heme synthesis in Rps19-deficient cells.

HIGHLIGHTS.

• Increased levels of intracellular unbound heme is found in an Rps19-deficient mouse model for DBA

• Reduced levels of Eukaryotic translation initiation factor 2-alpha kinase 1 (HRI) dampen elevated levels of unbound intracellular heme and ameliorates p53 activation in Rps19-deficient erythroblasts

• Treatment with alpha-1-microglobulin (A1M) decreases unbound intracellular heme without affecting p53 activity in Rps19-deficient cells