Red blood cell production is regulated by an internal gene regulatory circuit that includes the master megakaryocyte-erythroid TF GATA1, expressed early, as well as other erythroid-specific TFs, such as KLF1, NFE2, and TAL1/SCL (44) ( Figure 2 ). Proteomic studies of erythroid TF networks revealed progressive positive-feedback regulatory links between these TFs during erythroid differentiation, as well as cross-antagonism with key TFs involved in cell fate decisions such as GATA1:PU.1 (myeloid lineage), GATA1:GATA2, and KLF1:FLI1 (megakaryocyte lineage) antagonism (45, 46).

To maintain adequate levels of circulating red blood cells in dynamic conditions, each phase of erythropoiesis is also sensitive to regulation by extrinsic factors (47). IL-3 enhances expansion of haematopoietic progenitors including BFU-Es in the early stages of erythropoiesis (48, 49). Stem Cell Factor (SCF)/KIT ligand promotes cell survival and proliferation of BFU-Es, CFU-Es, and proerythroblasts by binding to the KIT receptor (CD117) (50, 51). In the subsequent stages of maturation beginning with CFU-E progenitors, erythropoietin (EPO) signalling via the EPO receptor (EPO-R) is critical for both the proliferation and survival of differentiating erythroid cells before gradually losing EPO-R expression after entering terminal maturation (51–55) ( Figure 2 ). These cytokines and growth factors stimulate several intracellular effector pathways including JAK/STAT, MAPK, and phosphatidylinositol 3-kinase (PI3K), that are known to promote cell growth and survival (56). Gain-of-function mutations in these pathways, such as constitutively activating JAK2 mutations, result in erythrocytosis and myeloproliferative disease (57–59). Similarly, a human EPO mutation resulting in altered EPO-R binding kinetics and reduced JAK2 signalling was described as a rare cause of DBAS (60, 61), whereas EPO gain-of-function mutations are associated with familial erythrocytosis (62).

Under conditions of erythroid stress, such as severe acute or chronic blood loss or hypoxia, red cell production is dramatically increased in response to stress, in a process termed stress erythropoiesis (63, 64). For example, a sustained increase in serum EPO levels, driven by tissue hypoxia, enhances the differentiation of erythroid progenitors and induces an erythroid lineage-bias in multipotent progenitor cells (65, 66). This physiological response has classically been considered part of stress erythropoiesis, but it has been proposed that this mechanism should instead be labelled enhanced steady-state erythropoiesis to distinguish it from a separate inflammation-driven pathway (53). Importantly, inflammation inhibits steady-state erythropoiesis (67–69), skews haematopoiesis toward myelopoiesis (70–72), and increases red cell turnover (73, 74), but simultaneously compensates for these processes by driving inflammation-driven stress erythropoiesis (63, 75). This involves two mechanisms: first, murine experiments have shown that BMP4 and Hedgehog signalling drive the recruitment of additional haematopoietic stem cell populations to the erythroid lineage (76, 77); second, amplification of lineage-committed erythroid progenitors is enhanced to increase red cell production (78–83).

Cell culture systems used to study erythropoiesis include primary CD34+ stem/progenitors, human umbilical cord blood derived erythroid progenitor (HUDEP) cells (84), and systems used to generate large numbers of red blood cells to develop alternatives to allogeneic red blood cell transfusion [termed Human Erythroid Massive Amplification (HEMA) cultures] (85–88). These systems exploit knowledge of the phases of erythroid differentiation in their design. EPO and SCF are used throughout in vitro differentiation, although EPO levels are increased in the latter phases. IL-3 and dexamethasone are used to promote erythroid expansion in early erythropoiesis, but are withdrawn in the last phase, when transferrin and insulin are added to promote terminal erythroid differentiation. Other cytokines and growth factors have been added with varying success in the context of human induced pluripotent stem cells (iPSCs) and murine erythroid differentiation, including pro-inflammatory cytokines such BMP4 and TGF-β signalling effectors that have been implicated in stress erythropoiesis in mice (63, 78, 89, 90).

Intrinsic gene regulatory circuits- composed of TFs that determine cell fate- and extrinsic signalling by growth factors and cytokines, including GCs, are intimately intertwined. GC signalling via the GR has an important role in regulating erythroid output both in steady-state and stress conditions. The cellular mechanisms by which GCs affect erythropoiesis are discussed below.

GC signalling via the GR receptor plays an important role in modulating the decision between self-renewal and differentiation in erythroid cells. Early observations in avian erythroid cells noted that in the absence of GR signalling, progenitors lose their ability to self-renew and enter terminal differentiation (82). Subsequent studies of mice defective for GR (79) have confirmed a cell-autonomous role for GCs in enhancing proliferation of erythroid progenitors (91), and shown that treatment of mice with synthetic GC dexamethasone simultaneously increased the proliferative capacity of BFU-E and CFU-E and inhibited the expression of key differentiation markers (3, 92). The slower rate of differentiation was proposed to allow more BFU-E cell divisions before self-renewal capacity was lost. This model fits with more recent single-cell studies that suggest the switch from late CFU-E to terminal differentiation governs the rate of erythropoiesis (8).

There is a debate in the field on the identity of the erythroid cells that are induced to self-replicate by GC. Data support target populations comprising BFU-E (3, 92, 93), CFU-E/pro-erythroblast (94), or a stress-specific erythroid progenitor cell. The latter was originally defined in mice as specific BMP4 and GC-dependent multi-potent progenitor and BFU-E that generated extramedullary erythropoiesis in the context of irradiation-induced anaemia (95). Subsequent work showed that murine and human bone marrow contain immunophenotypically distinct erythroid progenitors that express foetal haemoglobin during stress-induced differentiation (96). It is not yet clear whether ‘stress’ and steady state progenitor cells are mutually exclusive entities or a dynamic cell state. Further studies are also needed to resolve the impact of species, age and tissue (spleen and foetal liver versus bone marrow) on the cells targeted by GC in both steady state and stress erythropoiesis.

The transcriptional mechanisms by which GC delay erythroid differentiation, particularly in stress conditions, have been explored in numerous studies. Several direct transcriptional effects of GC via their GRE have been observed in erythroid progenitors. First, GATA1, the key TF driving erythroid differentiation in response to EPO-R signalling, is directly inhibited by GC activity via transcriptional repression in mouse erythroleukemia cells and human primary erythroblasts (31, 97), while expression of TFs driving erythroid proliferation such as LMO2 and C-MYB is induced (69, 75). Second, the ZFP36L2 gene, encoding an RNA-binding protein, is upregulated by GC during murine foetal erythropoiesis and has been shown to maintain a progenitor self-renewal state via downregulation of key erythroid differentiation genes by binding to the 3’ UTR ATTA motif of target transcripts (81). The observation that ZFP36L2 is downregulated in patient erythroid progenitors in DBAS, a condition in which expansion of erythroid progenitors is severely abrogated, provides further support for this (98), though the link between ribosomal haploinsufficiency in DBAS and downregulation of ZFP36L2 is unknown.

Direct non-transcriptional mechanisms of GC action have also been proposed. For example, a study of human proerythroblasts showed that GC inhibit erythroid maturation by interacting with EPO signalling and delaying STAT-5 phosphorylation through a rapid, membrane-associated pathway. When erythroid cells were stimulated with both dexamethasone and EPO, the GR associated with the EPO receptor and STAT-5. This association prevented the phosphorylation and subsequent nuclear translocation of STAT-5, thus inhibiting the transcriptional activation of the erythroid differentiation gene programme (31).

The delayed initiation of the terminal erythroid maturation gene programmes may at least in part be mediated by the ability of GCs to regulate cell cycle progression. Single-cell studies of erythropoiesis have revealed extensive changes to the cell cycle in BFU-E and CFU-E progenitors, including a gradual shortening of G1 and regulation of the S-phase at the transition from CFU-E to erythroid maturation (8, 99). Specifically, the cell fate switch from erythroid progenitor to terminal erythroid maturation was shown to be dependent on regulation of the S-phase by p57/KIP2, which mediates replication fork slowing (8, 100). This work highlighted that slowing of replication fork progression by p57/KIP2 in mouse erythroid progenitors prolongs the S-phase, which in turn promotes the CFU-E proliferation state over the switch to erythroid maturation. Deletion of p57/KIP2 resulted in a short S-phase and a failure to maintain a self-renewal state, resulting in deficient erythropoiesis in mouse embryos and reduced erythroid progenitor expansion in vitro ( 99, 101). GC were shown to modulate this process by inducing expression of p57/KIP2 to prevent S phase shortening and increase CFU-E proliferation, resulting in a larger CFU-E pool and erythropoietic rate. When the GC dexamethasone was withdrawn, p57/KIP2 expression was rapidly lost combined with induction of erythroid maturation transcriptional programmes (100). The expression of p57/KIP2 in CFU-E from DBAS patients was subsequently found to be essential for their GC responsiveness, highlighting the potential importance of this mechanism in humans (94).

Furthermore, the study of a genetic variant in an enhancer of the CCND3 gene identified in genome-wide association studies identified Cyclin D3 as an important regulator of the cell cycle during erythropoiesis. Genetic variation affecting CCND3 expression was linked to erythrocyte traits including both erythrocyte size and number (102). Although the effect of GC on Cyclin D3 has not been directly studied in erythropoiesis, Cyclin D3 expression is known to be sensitive to GC in T-cell lymphoma and lymphoblastic leukaemia where its downregulation and increased p27/KIP1 expression in response to dexamethasone results in increased apoptosis and growth suppression (103). Indeed, these mechanisms may play a role in erythropoiesis as cell cycle exit at the start of terminal erythroid maturation is also regulated by p27/KIP1 (104).

Cross-talk between GC signalling and the p53 response is well-established, but is complex and multivalent (105–107). P53 regulates the cell cycle via downstream activation of p21/CDKN1A (108). The effects of GC on p53 signalling in erythropoiesis have been best studied in the context of DBAS, in which ribosomal protein mutations lead to p53 activation in both patient cells and experimental model systems (98, 109–115). In an RPS19-deficient DBAS mouse model, GCs dampen the upregulation of cell cycle regulator p21/CDKN1A by antagonising the p53 response, stimulating maintenance of erythroid progenitors while delaying differentiation (93).

Cross-talk between GCs and other signalling pathways involved in erythroid differentiation and the stress response synergise to promote erythroid proliferation. GCs modulate gene expression in concert with cytokine receptors involved in erythroid differentiation such as EPO-R and SCF/KIT ligand to drive progenitor proliferation (116). Induction of CXCR4/CXCL12 by combined GC and EPO/SCF stimulation may be an important mediator of this effect by promoting BM retention of CD34+ progenitors and formation of erythroblastic islands (116–118).

Similar evidence exists for crosstalk between GC signalling and other signalling pathways involved in regulation of the stress response and metabolism. The simultaneous stimulation of the hypoxia-signalling transcription factor alpha (HIF1a) and dexamethasone treatment resulted in synergistic BFU-E expansion (3). Similarly, GC signalling and regulation of lipid metabolism may interact via bi-directional cross-talk in erythroblasts (4, 119). The lipid-sensing nuclear receptor PPAR-α co-occupies chromatin sites with GR when activated by agonists GW7647 and fenofibrate, resulting in a synergistic effect on BFU-E self-renewal (4).

Finally, there is increasing evidence that in addition to the direct effects on erythroid progenitors, GCs may stimulate erythroid differentiation indirectly by promoting the generation of macrophages with erythroid supporting potential in the context of the EBI. The surface protein CD169 is present on a subset of macrophages with an important nurturing role in murine erythropoiesis (120) and may be directly involved in EBI formation (121). In human HEMA cultures derived from CD34⁺ haematopoietic stem cells, dexamethasone promoted the generation of CD169⁺ macrophages that were observed to associate with proerythroblasts and stimulate cytokinesis to increase proerythroblast proliferation (122). Similar properties were described in culture of CD169⁺ macrophages that were derived from human monocytes by exposure to dexamethasone (118), highlighting the ability of GC to increase erythroid output by expanding the erythropoietic niche.

Although adrenal insufficiency causes a reduction in red cell mass and normocytic anaemia, the molecular mechanisms underlying this have not been studied directly (10, 11). The erythrocytosis observed in Cushing syndrome has only recently been exploited as a natural model for the study of erythropoiesis in chronic GC excess (13). CD34+ erythroid progenitors isolated from hypercortisolaemic Cushing patients were able to generate large numbers of immature erythroid cells in vitro, both in the presence and absence of further GC stimulation, suggesting that constitutive GR activation or epigenetic memory may be established upon chronic exposure. This phenotype disappeared in CD34+ cells isolated from patients in remission following pituitary adenoma removal, which instead showed resistance to GC stimulation and reduced erythroid cell numbers. The mechanisms underpinning this long-lasting desensitisation to GC stimuli has not yet been studied, but may be related to cytoplasmic retention of GRα (117).

GCs remain the only widely used drug class in the treatment of red cell failure in DBAS (14, 128), since it was first used over half a century ago (129). This congenital BM failure syndrome is rare (5-7 cases per million births) and is characterised by poor growth, congenital abnormalities, and severe anaemia, usually from infancy (14, 128, 130). The anaemia initially responds to high-dose GC treatment in 60-80% of patients, after which the dose is gradually tapered to minimise treatment-related morbidity. Unfortunately, the development of treatment-resistance is common and only 20-30% of patients respond to low-dose GCs long-term (131). GC treatment failure commits patients to a regular transfusion programme or allogeneic haematopoietic stem cell transplantation (HSCT), which remains the only curative treatment available for the BM-mediated effects of DBAS. Both treatments carry considerable treatment-related morbidity and mortality (130). While lentiviral vector-based gene therapies expressing RPS19 or GATA1 are currently being explored, these are not yet in use clinically (128).

DBAS is a ribosomopathy and more than 70% of DBAS cases are caused by heterozygous loss-of-function mutations in one of the ribosomal proteins (RP) of the large (RPL) or small (RPS) ribosomal subunits resulting in haploinsufficiency. Mutations most commonly affect the RPS19 gene (25% of patients) but have been described in over 20 RPS and RPL genes (132). Pathogenic variants in other genes (TSR2, GATA1, ADA2, EPO, and HEATR3) have been described to phenocopy DBAS in rare cases and in a quarter of patients, no pathogenic cause can be identified in genetic screens. While the pathophysiology of the erythroid lineage defect seen in DBAS has not been fully elucidated, multiple cellular aberrations arise due to reduced or selective translation of mRNAs (133) and nucleolar stress both caused by RP dysfunction (134). These include p53-mediated apoptosis, reduced levels of key erythroid TF GATA1, heme toxicity due imbalanced heme/globin production, and increased inflammation in the BM niche [for detailed review see (128, 132)].

Progress has been made in elucidating the mechanisms by which GCs influence these pathogenic processes. GC treatment did not ameliorate heme toxicity in RPL11-haploinsufficient mice (135). Instead of increasing GATA1 levels, GCs further inhibit GATA1 expression (31, 97), thus delaying erythroid differentiation and simultaneously stimulating expansion of progenitor cells via the stress erythropoiesis pathway ( Figure 2 , see above). This process may be further stimulated by TNFα and IFNγ-mediated inflammation in bone marrow niche that was observed in studies of DBAS patient bone marrow (98) and circulating red blood cells from DBAS patients (136). Moreover, GCs counter p53 activation in DBAS. The study of RPS19-mutant patient-derived CFU-Es and HUDEP2 erythroid cells showed reduced p53 signalling and apoptosis upon GC-treatment (101) and GC prevent p53 activation in a RPS19 mouse model (93).

Recent hypotheses propose that GCs may also support enucleation via c-Myc repression and modulate autophagy via mTor signalling in DBAS but direct evidence for these roles in erythropoiesis is awaited (137). Mytophagy, a form or autophagy specific to erythroid maturation, is an important regulatory component in erythropoiesis which facilitates the removal of mitochondria and other organelles during terminal maturation (138, 139). GATA1-mediated induction of mytophagy gene transcription requires silencing of mTor signalling (140), and increased mTor signalling results in macrocytic anaemia in mice (141). Interestingly, a screen for novel therapeutic approaches for DBAS identified the small-molecule SMER28 that was able to rescue erythroid differentiation of patient iPSCs by inducing mTor-independent mytophagy (142). The mTor signalling pathway is increased in DBAS, suggesting inhibition may be a therapeutic option to further enhance mytophagy (143). Paradoxically, a subset of DBAS patients respond to the mTor-inducing amino acid L-Leucine (144). This bifunctional response is explained by the dual, stage-dependent function of mTor in erythropoiesis, both stimulating proliferation of BFU-E and CFU-E proliferation in early erythropoiesis and inhibiting mytophagy in definitive erythroid maturation (140) ( Figure 2 ). Inhibition of mTor signalling by GCs has been described in several non-haematopoietic cell types (145, 146), suggesting that GCs may contribute to erythropoiesis in DBAS by stimulating mytophagy while simultaneously restoring erythroid progenitor proliferation via stress erythropoiesis (137).

Resistance to GC treatment in DBAS is only partially understood (94), and may be driven by a wide range of mechanisms related to GR genetic variation (such as the A3669G polymorphism in GR) and transcriptional and post-transcriptional regulation of the GR (32, 147). Interestingly, foetal haemoglobin (HbF) expressing cord blood-derived CD34⁺ progenitors do not respond to GC stimulation, whereas peripheral blood-derived progenitors do (94). This was recently linked to the finding that expression of developmental globin-switching regulator BCL11A correlated to the ability of erythroid progenitors to proliferate upon GC treatment (148).

MDS is a heterogeneous clonal stem cell disorder characterised by rapid proliferation in the BM paired with dysplastic maturation and high cell turnover due to excessive apoptosis, explaining the peripheral cytopenias associated with the disease. The most common cytopenia is anaemia which is present in 80-90% of all patients (149). Cytopenic and dysplastic features are used to classify and risk-stratify MDS, based on the risk of transformation to AML present in this disorder (150). The only form to be classified predominantly by the presence of a cytogenetic abnormality is MDS with isolated del(5q), characterised clinically by a low risk of disease progression and a macrocytic anaemia with erythroid hypoplasia often resulting in transfusion-dependence (151, 152).

Interestingly, the hemizygous del(5q) deletion on the long arm of chromosome 5 contains the RPS14 gene. Analogous to DBAS, haploinsufficiency in del(5q) MDS drives ribosomal stress and activation of the p53 response in erythroid cells, predisposing cells to apoptosis (153). Unlike DBAS, GCs are not routinely used in treatment of del(5q) MDS, whereas lenalidomide restores transfusion-independence and improves survival (154). The effects of lenalidomide and GCs on erythropoiesis were compared in the in vitro erythroid differentiation of CD34+ stem cells. Whereas GCs stimulated erythroid progenitor expansion at the BFU-e stage, Lenalidomide stimulated CFU-e proliferation suggesting separate mechanisms of action. Both also stimulated erythropoiesis in RPS19 and RPS14 mutant CD34+ cells and showed additive effects when used in combination (155, 156). These in vitro findings were confirmed in del(5q) MDS patients who developed lenalidomide resistance. Transfusion-independence was restored in five out of eight patients when dexamethasone was added to lenalidomide treatment. This effect was suggested to be mediated by observed reduction of the p53 response, and could be replicated in vitro via specific suppression of p53 in patient-derived CD34+ cells by antisense oligonucleotide, Cenersen, which increased erythroid burst recovery (157).