The BH3-only proteins BIM and PUMA are not critical for the reticulocyte apoptosis caused by loss of the pro-survival protein BCL-XL

Anaemia is a major global health problem arising from diverse causes and for which improved therapeutic strategies are needed. Erythroid cells can undergo apoptotic cell death and loss of pro-survival BCL-XL is known to trigger apoptosis during late-stage erythroid development. However, the mechanism by which loss or pharmacological blockade of BCL-XL leads to erythroid cell apoptosis remains unclear. Here we sought to identify the precise stage of erythropoiesis that depends on BCL-XL. We also tested whether deficiency of BIM or PUMA, the two main pro-apoptotic antagonists of BCL-XL, could prevent reticulocyte death and anaemia caused by BCL-XL loss. Using an in vivo mouse model of tamoxifen-inducible Bclx gene deletion and in vitro assays with a BCL-XL-selective inhibitor, we interrogated each stage of erythrocyte differentiation for BCL-XL dependency. This revealed that reticulocytes, but not orthochromatic erythroblasts, require BCL-XL for their survival. Surprisingly, concurrent loss of BIM or PUMA had no significant impact on the development of anemia following acute BCL-XL deletion in vivo. However, analysis of mixed bone marrow chimaeric mice revealed that loss of PUMA, but not loss of BIM, partially alleviated impaired erythropoiesis caused by BCL-XL deficiency. Insight into how the network of pro-survival and pro-apoptotic proteins works will assist the development of strategies to mitigate the effects of abnormal cell death during erythropoiesis and prevent anaemia in patients treated with BCL-XL-specific BH3-mimetic drugs.

Anaemia affects over 1.5 billion people world-wide 1 and is a major cause of morbidity that requires improved therapeutic interventions. Underlying causes for anaemia are diverse and include haemoglobinopathies, red cell enzyme disorders, nutritional deficiencies, cytoskeletal abnormalities and autoimmune diseases. Furthermore, anaemia is a frequent toxicity associated with cancer therapy, including both conventional chemotherapy as well as emerging targeted therapies, such as some of the BH3-mimetics. 2,3 Understanding the requirements for normal erythroid development and survival will help develop improved treatment strategies to support erythropoietic function in patients.
In the adult, mature red blood cells derive from haematopoietic stem cells. Erythropoiesis proceeds through several differentiation steps whereby the potential for alternate blood cell lineages is progressively lost and immature progenitors expand in number to facilitate production of sufficient mature red blood cells, with half-lives spanning 100-120 days in humans and 8-20 days in mice. [4][5][6][7] Should bone marrow erythropoiesis become compromiseddue to bone marrow failure/dysplasia or infiltration, or excessive loss/destruction of peripheral red blood cellsextramedullary erythropoiesis can occur in the spleen to facilitate enhanced erythrocyte production in mice. Robust prosurvival signalling is essential to prevent excessive apoptotic cell death during erythropoiesis and the maintenance of adequate red blood cell numbers. Expression analysis has identified the pro-survival gene Bclx as particularly highly expressed in the erythroid lineage, 8 suggestive of its critical importance for countering apoptosis during erythropoiesis. Aberrant apoptosis of red blood cells and erythroid progenitors is a common feature of chronic anaemias and is likely to contribute to their severity. For example, thalassaemia patients have elevated numbers of early precursor cells in the bone marrow but abnormally high rates of apoptosis in more mature progenitors, 9 whereas in anaemia of chronic disease, elevated interferon-γ levels are associated with increased apoptosis of erythroid progenitors and inversely correlated with reticulocyte numbers and haemoglobin levels. 10 Aberrant apoptosis is also a key feature of Diamond-Blackfan Anaemia where ribosomal stress drives abnormal TP53 activation and cell death. 11 TP53 is known to regulate the expression of several key initiators of apoptosis, including Puma, Noxa and Bax. [12][13][14][15] The intrinsic (also called 'BCL-2-regulated', 'stress' or 'mitochondrial') apoptotic pathway is regulated by three major classes of proteins. 16,17 The pro-survival proteins (e.g., BCL-2, BCL-XL, MCL-1) antagonise the pro-apoptotic effectors BAX and BAK. Upon stress stimuli pro-apoptotic BH3-only proteins, such as BIM and PUMA, are induced transcriptionally or post-transcriptionally in a cell type and stimulus-specific manner, and initiate apoptosis by displacing BAX/BAK from their restraint by the pro-survival proteins and possibly also by direct activation of BAX/BAK. 18 Activated BAX and BAK mediate disruption of the outer mitochondrial membrane. This represents the 'point of no return' for the cell, triggering activation of the caspase cascade, resulting in proteolysis of many substrates and cellular demolition. 19 In accordance with their roles in apoptosis signalling, loss of pro-apoptotic BIM or combined loss of BAX and BAK result in abnormal accumulation of diverse haematopoietic cell types, 20,21 whereas conversely loss of pro-survival MCL-1 results in bone marrow failure. 22,23 In the erythroid lineage BCL-XL is of critical importance. Germline deletion of Bclx results in embryonic death of mice at~E13.5 in part owing to increased apoptosis of erythroid progenitors; interestingly, this death of erythroid progenitors could be rescued by concomitant loss of BIM. 24,25 Tissue-restricted deletion of Bclx results in severe anaemia, splenomegaly due to erythroblast accumulation, and thrombocytopenia. 26 Loss or inhibition of BCL-XL (e.g., using the BH3-mimetics ABT-737 or ABT-263 ref. 27,28), but not of BCL-2 (e.g., using ABT-199 ref. 29), results in thrombocytopenia in patients and mouse models and also anaemia in mice. 2,24,26,27,[30][31][32][33][34] It has yet to be established whether pharmacological inhibition of BCL-XL will also cause anaemia in patients. Published data demonstrate that both BAX and BAK must be removed to prevent anaemia caused by loss or drug mediated inhibition of BCL-XL, as they have largely overlapping roles in the execution of apoptosis. 26,35 It is, however, still unknown which BH3-only protein is responsible for the initiation of apoptosis in this context.
We defined the requirement for BCL-XL at various stages of adult erythropoiesis by using a discriminating flow cytometry method 36 and a tamoxifen-inducible, acute Bclx gene deletion mouse model. Given that BIM is essential for the aberrant apoptosis of erythroid progenitors in embryonic mice caused by the absence of BCL-XL, 25 we investigated the role of this pro-apoptotic BH3-only protein in adult erythropoiesis. We found that BCL-XL is critical for the survival of reticulocytes and that BIM is not essential for the anaemia that is caused by acute loss of BCL-XL, whereas pro-apoptotic PUMA has a minor role. These discoveries inform the development of strategies to alleviate anaemia caused, for example, by inherited mutations, infections or treatment with anti-cancer agents, including the new BH3-mimetic drugs.

Results
Acute loss of BCL-XL causes profound anaemia in adult mice owing to failure of erythropoiesis. To confirm and extend published data characterising the role of BCL-XL in erythroid cell survival, 24,26,30 we generated mice bearing floxed Bclx (Bcl2l1) alleles that could be deleted in an inducible manner by tamoxifen-dependent CreERT2-recombinase activity. Bclx fl mice 26 were crossed with Rosa-CreERT2 Ki mice 37 to generate Bclx fl/fl ;RosaCreERT2 Ki/+ compound mutant mice. In these mice, tamoxifen administration activates the latent CreERT2 recombinase to facilitate recombination of the floxed Bclx alleles, leading to loss of BCL-XL expression.
A cohort of Bclx fl/fl ;RosaCreERT2 Ki/+ mice and Bclx fl/fl control mice were treated with tamoxifen at 8 weeks of age. At 1 month post treatment these adult mice were analysed to determine the effects of BCL-XL loss. Peripheral blood analysis confirmed previous reports, 26,31 with profound anaemia observed in the Bclx fl/fl ;RosaCreERT2 Ki/+ mice. Haemoglobin (P = 0.0001) levels, haematocrit (Po0.0001) and reticulocyte as well as mature red blood cell numbers were all significantly reduced ( Figure 1a). Conversely, both spleen weight and cellularity were increased in the Bclx fl/fl ;Rosa-CreERT2 Ki/+ mice (Figure 1b). In contrast, the tamoxifentreated Bclx fl/fl control mice retained normal blood and spleen cell counts.
A previous study reported that haemolysis was associated with BCL-XL deletion. 26 We therefore examined the impact of inducible adult loss of BCL-XL on red blood cell turnover.  Figure 1A). Consistent with normal hepatic function, no increases in the levels of serum albumin, alkaline phosphatase, aspartate aminotransferase and gammaglutamyl transferase (o4 U/l; data not shown) were observed. There was a marginal increase in alanine aminotransferase; however, the levels observed were below those indicative of significant liver damage (Supplementary Figure 1B). Hence, we conclude that liver dysfunction did not account for the accumulation of serum bilirubin. The serum haptoglobin levels were very low (o0.08 g/l; data not shown), consistent with abnormal red blood cell breakdown and release of haemoglobin. However, the lactate dehydrogenase levels were not increased, which suggests that red blood cell destruction may occur predominantly within the extra-vascular space (bone marrow and spleen) (Supplementary Figure 1B). Blood films were prepared from tamoxifen-treated Bclx fl/fl ; RosaCreERT2 Ki/+ and RosaCreERT2 Ki/+ (control) mice to identify changes in red blood cells caused by loss of BCL-XL. Remarkably, almost complete disappearance of reticulocytes was seen by day 2 (Supplementary Figure 2A). This was preceded by morphologic changes in the reticulocytes of coarse basophilic stippling (Supplementary Figure 2B). There was no evidence of spherocytosis, red blood cell fragmentation (schistocytes) or specific changes typical of oxidative haemolysis ('bite' or 'blister' cells) following BCL-XL loss. Mild CRE recombinase toxicity was observed in control animals at day 5, with a minor reduction in reticulocytes. However, this effect was transient and haematopoietic reconstitution experiments revealed that transient CreERT2 activation exerted no long-term consequences (Supplementary Figure 2C).
Red blood cells constitute 495% of all blood cells and must be continuously replaced as they are lost or expended. Consequently, perturbation of bone marrow stem/progenitor cell activity may lead to the observed anaemia. Hence, we examined whether the reduction in red blood cells in the tamoxifen-treated Bclx fl/fl ;RosaCreERT2 Ki/+ mice was a consequence of defective bone marrow stem/progenitor cell activity. Flow cytometry was used to determine the proportions and numbers of stem and progenitor cell subsets in the bone marrow of these mice. This revealed that after tamoxifen treatment  Figure 3A). As a functional readout of progenitor cells, we performed colony-formation assays with bone marrow and spleen cells in semi-solid agar. No significant differences were observed in the relative proportions of the various myeloidcommitted colony-forming cells in the bone marrow and spleen between tamoxifen-treated Bclx fl/fl ;RosaCreERT2 Ki/+ and Bclx fl/fl (control) mice (Table 1). Therefore, we conclude that BCL-XL is dispensable for the survival of haematopoietic stem/progenitor cells in adult mice at steady state.
We next wanted to define the precise stage at which red blood cell production is perturbed to identify the cause of the anaemia caused by loss of BCL-XL. One month after deletion of Bclx, red blood cell maturation was analysed by flow cytometry based on expression of TER119, CD44 and cell size (FSC) (ref. 36; Figure 2a). Following Bclx gene deletion we observed significantly decreased percentages of mature red blood cells, but not reticulocytes in the bone marrow with a corresponding increase in the percentages of the more immature precursors (Figure 2b). This phenomenon was mirrored in the spleen of the tamoxifen-treated Bclx fl/fl ; RosaCreERT2 Ki/+ mice, where normally only few immature erythroid progenitor cells are found. Colony-formation assays revealed a pronounced increase in the numbers of colonyforming units-erythroid (CFU-e) in the spleen of tamoxifentreated Bclx fl/fl ;RosaCreERT2 Ki/+ mice compared with Bclx fl/fl mice, whereas no differences were observed in the bone marrow (Supplementary Figure 3B). These findings are consistent with erythropoietic insufficiency in the tamoxifentreated Bclx fl/fl ;RosaCreERT2 Ki/+ mice, whereupon the spleen has been recruited as a site for extramedullary erythropoiesis in a reactive attempt to overcome the low numbers of mature red blood cells in the circulation.
Next, we harvested bone marrow cells from Bclx fl/fl ; RosaCreERT2 Ki/+ and Bclx fl/fl mice one month following tamoxifen treatment and tested their ability to reconstitute the haematopoietic system of lethally irradiated recipient mice. The animals reconstituted with Bclx fl/fl ;RosaCreERT2 Ki/+ cells developed symptoms of erythropoietic failure within 3 weeks post-reconstitution. Blood transfusion was performed, however by 4 weeks post-reconstitution these mice reached ethical endpoint and were killed for analysis. Flow cytometric Number of colonies per 50 000 cells BCL-XL is essential for the survival of reticulocytes.
Analysis of mitochondrial content revealed that, as expected, 95% of reticulocytes contained mitochondria and only mature red blood cells were devoid of mitochondria (Figures 3a  and b). Having confirmed that reticulocytes and erythroblasts contain the organelle required for the mitochondrial apoptotic pathway, we examined whether the loss of reticulocytes resulting from the absence of BCL-XL was caused by apoptosis. Wild-type reticulocytes were FACS-sorted and treated in vitro with BH3-mimetic drugs that specifically inhibit BCL-XL (A-1331852) 38 or BCL-2 (ABT-199; used as a control) ( Figure 3c). Treatment with A-1331852 resulted in the rapid death of reticulocytes, and this could be substantially reduced during the first 24 h by blocking caspase activation with QVD-OPH. In contrast, treatment with ABT-199 had no effect on reticulocyte viability (Figure 3c). Interestingly, the marked dependency on BCL-XL for survival was specific to reticulocytes, as orthochromatic erythroblasts were far less sensitive to A-1331852. Furthermore, following Bclx gene deletion a significant increase in the exposure of phosphatidyl-serine (detected by Annexin V staining)an early apoptotic markerwas observed on reticulocytes, but not mature red blood cells, from the bone marrow and spleen (Figure 3d). BCL-XL was implicated in mitochondria specific-autophagy (mitophagy) owing to the established role of Nip3-like protein-X (NIX) in the clearance of mitochondria from maturing red blood cells 39 and the ability of BCL-XL to interact with the NIX homologue BNIP3. 40 To investigate the involvement of mitophagy in the loss of reticulocytes caused by BCL-XL deletion, cell viability was measured following treatment of erythroid progenitors with the autophagy inhibitor - bafilomycinthat prevents fusion between autophagosomes and lysosomes. 41 Bafilomycin had no impact on erythroid precursor survival (Figure 3e), consistent with the notion that autophagy plays no role in the loss of reticulocytes caused by BCL-XL deletion.
Hematopoietic-specific loss of BCL-XL provokes anaemia in adult mice that cannot be alleviated by concomitant loss of BIM or PUMA. We next investigated which proapoptotic BH3-only protein might be critical for reticulocyte apoptosis following loss of BCL-XL. BIM and PUMA are prime candidates because they bind to all pro-survival BCL-2 proteins and are critical for apoptosis induction in many haematopoietic cell types, particularly lymphocytes, 20,42-44 after exposure to diverse cytotoxic insults. 22,45,46 To investigate the roles of BIM and PUMA, we reconstituted lethally irradiated mice with bone marrow from Bclx fl/fl ;RosaCreERT2 Ki/+ ;Bim −/− , Bclx fl/fl ;Rosa-CreERT2 Ki/+ ;Puma −/− or relevant control mice (Figure 4a). Mice expressing GFP in all cells, importantly including erythroid cells, 47 were used as recipients to distinguish donor-derived haematopoietic cells (GFP − ) from residual recipient-derived cells (GFP + ) (Supplementary Figure 5). Efficient haematopoietic reconstitution was confirmed 8 weeks post-transplantation with donor-derived (GFP − ) red blood cells present at 495% (Figure 4b). Bclx gene deletion was induced with tamoxifen and after 1 month bone marrow, spleen and peripheral blood cells were analysed. As expected, 26 tamoxifen-treated Bclx fl/fl ;RosaCreERT2 Ki/+ reconstituted mice developed severe anaemia evident by low red blood cell numbers and splenomegaly, whereas RosaCreERT2 Ki/+ reconstituted mice presented with normal blood counts and spleen cellularity (Figure 4c). Surprisingly, all tamoxifen-treated Bclx fl/fl ;RosaCreERT2 Ki/+ ;Bim −/− reconstituted mice presented with splenomegaly and anaemia ( Figure 4c). Hence, in contrast to its importance in embryonic erythropoiesis, loss of BIM failed to mitigate the erythropoietic insufficiency caused by BCL-XL deletion in adult mice. Loss of PUMA also provided no protection. Accordingly, no differences in PUMA has a minor role in anaemia induced in adult mice by acute BCL-XL loss. To further investigate whether BIM or PUMA may have a minor role in the apoptosis of erythroid cells caused by BCL-XL loss, we adopted a mixed bone marrow reconstitution approach (Figure 4a). GFP + mice were reconstituted with a 1:1 mix of bone marrow cells comprising wild-type (GFP + ) 'competitor' cells and 'test' cells of the following genotypes: RosaCreERT2 Ki/+ , Bclx fl/fl ;Rosa-CreERT2 Ki/+ , Bclx fl/fl ;RosaCreERT2 Ki/+ ;Bim −/− and Bclx fl/fl ; RosaCreERT2 Ki/+ ;Puma −/− (all GFP − ). As expected, the presence of the wild-type cells prevented the anaemia caused by BCL-XL deletion (Figure 5a). When the ratio of test to competitor derived cells was determined within the red blood cell compartment, relative to the RosaCreERT2 Ki/+ control cells, the Bclx fl/fl ;RosaCreERT2 Ki/+ cells were profoundly depleted (Figure 5b). This deficit was mildly alleviated by concomitant loss of PUMA, whereas loss of BIM had no impact (Figure 5b). These results indicate that PUMA has a minor role in the loss of red blood cells resulting from acute BCL-XL loss.

Discussion
Throughout life large numbers of red blood cells must be continuously produced in the bone marrow to offset their ongoing loss in the periphery. The high rate of proliferation of progenitors and the marked changes in cellular architecture during differentiation must impose stresses that activate apoptosis signalling. Thus, inhibitors of apoptosis are expected to be critical to sustain survival of differentiating erythroid cells and previous studies identified BCL-XL as the critical factor. Germline deficiency for BCL-XL causes embryonic lethality around E13.5 owing to aberrant death of erythroid and neuronal cells. 24 Loss of BIM was shown to reduce the excess apoptosis of haematopoietic fetal liver cells in Bclx −/− mice (although this did not prevent embryonic lethality). 25 Studies in vitro concluded that BCL-XL is needed for the survival of late-stage erythroid progenitors. 30,48 However, the precise stage during erythropoiesis at which BCL-XL is critical for survival was not resolved. Cell type-specific conditional Bclx gene deletion in adult mice caused severe anaemia, accompanied by splenomegaly and elevated numbers of erythroblasts. This led to the conclusion that BCL-XL is only essential during terminal erythroid maturation. Because of the elevated reticulocyte numbers, the authors of that study attributed the anaemia to excessive haemolysis. 26 Conversely, a subsequent study using the same model, but a later time point for analysis, reported reduced reticulocyte numbers. The authors of the second study concluded that an erythrocyte production defect was responsible for the anaemia caused by loss of BCL-XL and proposed that mature erythroblasts rather than reticulocytes depend on BCL-XL for survival. 31 Using genetic and pharmacological approaches coupled with FACS analysis to delineate discrete stages of erythroid differentiation, 36 we found that BCL-XL is essential for reticulocyte survival but largely dispensable for the survival of orthochromatic erythroblasts and earlier progenitors. This is consistent with reports that the highest levels of BCL-XL are observed during terminal erythrocyte differentiation when haemoglobin production is maximal. 8 Apoptosis of reticulocytes following acute Bclx gene deletiondocumented by appearance of apoptosis-specific markersresults in a compensatory increase in erythroblasts in the bone marrow and the spleen. Our findings clarify previously reported observations. Our use of the RosaCreERT2 allele enabled efficient inducible Bclx gene deletion resulting in anaemia within 1 month, compared with the 3-4 months in the previously described models 26 . This allowed primary effects of BCL-XL loss to be studied in the absence of potentially confounding secondary effects or compensatory processes. Our acute Bclx deletion model closely mimics the scenario of therapeutic BCL-XL inhibition by BH3-mimetic drugs and therefore provides insight into their likely impact on patients.
Pro-apoptotic PUMA and BIM were investigated as potential initiators of reticulocyte death caused by loss of BCL-XL. In contrast to the published role for BIM as a critical initiator of erythrocyte precursor death during embryogenesis in Bclx deficient mice, 25 we found that concomitant loss of BIM did not alleviate the anaemia caused by acute BCL-XL loss in the adult. Loss of PUMA also provided no rescue. However, experiments using mixed bone marrow reconstituted mice provided evidence that PUMA plays a minor role in the reticulocyte death. Hence, we conclude that other BH3-only proteins may be critical (either by themselves or together with PUMA and BIM) for the induction of apoptosis following BCL-XL loss in reticulocytes.
Our findings demonstrate that reticulocytes are the key erythroid precursors that require BCL-XL for survival and that acute loss of BCL-XL results in anaemia owing to increased apoptosis of reticulocytes and thus diminished production of mature red blood cells. In mice, mature red blood cells spontaneously lyse after 8-20 days, unless taken up by macrophages for retrieval of iron. Hence, anaemia will ensue upon loss of BCL-XL-dependent reticulocytes, although in mice this can be delayed by compensatory erythropoiesis in the spleen. This accounts for the relatively slow development of anaemia, the low numbers of reticulocytes as well as the elevated numbers of the more primitive erythroid precursors.
In humans, red blood cells survive in the circulation for up to 120 days, significantly longer than mice. The extended survival of red blood cells in humans may provide a therapeutic window for intervention with BCL-XL inhibitors, unless erythropoiesis is already compromised prior to treatment.   20 Puma −/− , 42 UBC-GFP Tg 47 and RosaCreERT2 37 ). All mice were either generated on a C57BL/6 background or crossed onto this background for at least 20 generations before commencement of our studies. Tamoxifen administration was performed by oral gavage, 3.6 μg/day on 3 consecutive days. Blood transfusion was performed with blood isolated from UBC-GFP Tg mice (GFP + ), 250 μl blood/recipient. Genotyping was performed as previously reported. 22 Oligonucleotide sequences for genotyping of the mutant alleles will be provided on request.
Generation of bone marrow chimaeric mice. Haematopoietic reconstitutions were performed as previously described. 22 Histology and blood film analysis. Sternum and spleen specimens were fixed in 10% formalin. Paraffin tissue sections were prepared at 7.5 μm and stained with haematoxylin and eosin. Images (sternum × 20; spleen × 40) were taken with an Olympus BX43 (Olympus, Shinyuku, Japan) using the acquisition program CellSens Standard (Olympus). Slides for analysis of blood cell morphology were prepared by spreading freshly venesected blood by hand onto slides, which were allowed to air-dry. Blood films were fixed in methanol and stained using May-Grunwald and Giemsa stains. Blood film images were acquired using a Nikon Eclipse 90i microscope. Images presented were taken using a × 100 oil objective and inset views for reticulocyte features utilised digital zoom.
Serum analysis. Blood from mice was collected 1 month after tamoxifen treatment via cardiac puncture. Bleeds were left to coagulate at room temperature for 5 min and centrifuged for 1 min at 13 000 rpm to collect serum. Sera were stored at − 20°C and analysed using Architect c1600 (Abbott Diagnostics, Santa Clara, CA, USA).
Flow cytometric analysis. Spleen and bone marrow cells were harvested and single-cell suspensions prepared. Cells were counted using the CasyCell Counter (Schaefe System GmbH, Neunkirchen, Germany). Retro-orbital bleeds were collected into EDTA for differential cell counts using an ADVIA 2120 analyser (Siemens Healthcare PTY, Ltd, Bayswater, VIC, Australia). The analysis of red blood cells and leukocytes from either wild-type or test bone marrow in the haematopoietic reconstituted mice was determined by flow cytometric analysis of GFP and staining for haematopoietic subset-specific surface markers ( . Antibodies were conjugated to FITC, R-PE or APC. For erythroid precursor analysis, mature cells were gated out prior to CD44 versus FSC analysis using a 'DUMP' channel with the following surface markers: B220 or CD19, GR-1, MAC-1, CD45.2, CD3. GFP + (wild-type bone marrow derived) and GFP − (test bone marrow derived) cells were used to determine relative contributions in the haematopoietic reconstituted mice. Samples were analysed in a LSR-II flow cytometer (BD Biosciences, San Jose, CA, USA). Dead cells were excluded using forward and side light scatter plus propidium iodide exclusion (Sigma Aldrich).
Mitochondrial content analysis. Cells were incubated with 500 nM MitoTracker Deep Red for 20 min at 37°C (Thermo Fisher, #M22426). Cells were then stained for the appropriate lineage markers, TER119 and CD44 to delineate specific erythroid cell populations by flow cytometry.
Caspase-9 activity assay. 5 × 10 4 bone marrow cells were cultured for 4 or 24 h with or without the BCL-XL inhibitor A-1331852 in a 96-well plate in 100 μl of medium. Plates were allowed to equilibrate at room temperature prior to adding 100 μl Caspase-9 Glo reagent (Promega, Madison, WI, USA). Assays were analysed after 30 min incubation according to the manufacturer's instructions.
Clonogenic assays. CFU-e were enumerated by culturing single-cell suspensions of bone marrow (25 000) or spleen (50 000) in 1 ml of MethoCult 3234 medium (Stem Cell Technologies, Vancouver, BC, Canada) supplemented with 2 U/ml erythropoietin (EPO, Janssen-Cilag Ltd, Buckinghamshire, UK) with incubation in 5% CO 2 in air for 2-3 days. The numbers of myeloid colonyforming cells in single-cell suspensions of bone marrow (25 000) or spleen (50 000) were assessed in 1 ml cultures of 0.3% agar in Dulbecco/s modified Eagles medium containing 20% newborn calf serum, stem cell factor (SCF, 100 ng/ml, WEHI), EPO (2 U/ml, Janssen-Cilag Ltd) and interleukin-3 (IL-3, 10 ng/ml). Cultures were incubated at 37°C for 7 days in 10% CO 2 in air. Cultures were fixed, dried onto glass slides and stained for acetylcholinesterase followed by Luxol fast blue and haematoxylin to determine the numbers and type of colonies. Statistical analysis. Cellularity, weight, blood parameters, flow cytometric results and in vitro cell survival were plotted and analysed with GraphPad Prism (GraphPad Software Inc, La Jolla, CA, USA). Statistical comparisons were conducted in a pair-wise manner using unpaired two-tailed Student's t-test assuming equal variance. For Figure 2c we utilised the Sidak-Bonferroni method for multiple comparisons correction. For Figure 4 we performed two-way ANOVA analysis with Tukey's multiple comparisons test. Stars indicate significant values. Error bars are presented as standard error of mean (± S.E.M.).