Erythropoietin-dependent acquisition of CD71^hiCD105^hi phenotype within CD235a⁻ early erythroid progenitors

Abstract

The development of committed erythroid progenitors and their continued maturation into erythrocytes requires the cytokine erythropoietin (Epo). Here, we describe the immunophenotypic identification of a CD34⁻ colony-forming unit-erythroid (CFU-E) progenitor subtype, termed late CFU-E (lateC), that arises in an Epo-dependent manner during human early erythropoiesis (EE). LateC cells lack CD235a (glycophorin A) but have high levels of CD71 and CD105, characterized as Lin⁻CD123⁻CD235a⁻CD49d⁺CD117⁺CD34⁻CD71^hiCD105^hi. Analysis of ex vivo cultures of bone marrow (BM) CD34⁺ cells showed that acquisition of the CD71^hiCD105^hi phenotype in lateC occurs through the formation of 4 other EE subtypes. Of these, 2 are CD34⁺ burst-forming unit-erythroid (BFU-E) cells, distinguishable as CD71^loCD105^lo early BFU-E (earlyB) and CD71^hiCD105^lo late BFU-E (lateB), and 2 are CD34⁻ CFU-E, also distinguishable as CD71^loCD105^lo early CFU-E (earlyC) and CD71^hiCD105^lo mid CFU-E (midC). The EE transitions are accompanied by a rise in CD36 expression, such that all lateC cells are immunophenotypically CD36⁺. Patterns of CD34, CD36, and CD71 indicate 2 differentiation routes—in one, earlyB loses CD34 to form earlyC, and in another, earlyB gains CD36 and CD71^hi expression prior to losing CD34 to form midC, bypassing the earlyC stage. Regardless of the route, the transition from midC to lateC requires Epo. All 5 EE subtypes could be prospectively detected in human BM cells and, upon isolation and reculture, exhibited the potential to continue differentiating along the erythroid trajectory. Finally, we find that all 5 EE populations can also be detected in cultures of cord blood-derived CD34⁺ cells at levels similar to those observed in BM CD34⁺ cell cultures.

Graphical abstract

Graphical Abstract.

Significance statement.

Erythropoietin (Epo) is essential for erythroid progenitor survival and differentiation. In this study, we resolve the heterogeneity of early erythroid progenitors into 5 distinct subpopulations and identify the transition to a CD71^hiCD105^hi immunophenotype as the critical Epo-­dependent event. Our immunophenotyping strategy enables prospective identification of these subtypes in human bone marrow cells. These findings refine our understanding of Epo action and provide a framework for analyzing erythropoiesis and evaluating erythropoiesis-stimulating agents (ESAs) in both basic and translational hematopoietic research.

Introduction

Differentiation of hematopoietic stem cells (HSCs) to erythrocytes or red blood cells (RBCs) is dependent on erythropoietin (Epo), a cytokine produced and released by the kidneys.^1–3 Epo is necessary for the survival and continued differentiation of erythroid lineage committed progenitor colony-forming unit-erythroid (CFU-E) cells in both humans and mice.^4–6 Mice lacking Epo or its receptor (EpoR) die by embryonic day 13.5 due to a lack of mature RBCs.^7–10 Binding of Epo induces homodimerization of EpoR, which stimulates the receptor-­associated Janus Kinase 2 to phosphorylate EpoR and itself.^11,12 This process is associated with the downstream activation of a gene expression program that inhibits apoptosis and promotes survival, proliferation, and continued differentiation of CFU-E.^2,13–16

Classically, the development of erythrocytes from HSCs has been described as a multistage process that progresses through intermediate progenitors, including multipotent progenitor and common myeloid progenitor (CMP).^17–22 The latter give rise to the bipotent megakaryocyte-erythroid progenitor (MEP), from which the erythroid progenitor burst-forming unit-erythroid (BFU-E) cell arises. The formation of BFU-E from MEP and their conversion to CFU-E, a phase referred to as early erythropoiesis (EE), is followed by terminal erythroid differentiation (TED).^20,23,24 The transition of CFU-E to proerythroblast (ProE) marks the onset of TED and is associated with the canonical erythroid lineage marker CD235a (glycophorin A). ProE develops through basophilic erythroblast (BasoE) and polychromatic erythroblast (PolyE), which finally form orthochromatic erythroblast (OrthoE). The enucleation of OrthoE forms a reticulocyte, which ultimately matures into an RBC.^25–27

Immunophenotyping strategies have been described for the analysis of all 4 phases of human erythroid differentiation.^28–41 Strategies for EE cells include early markers that are present in hematopoietic stem and progenitor cells (HSPCs) and persist into TED, such as CD117 (c-kit) and CD49d (α-4 integrin), and transiently expressed markers, such as CD36 (fatty acid translocase), CD105 (endoglin), and CD71 (transferrin receptor).^{35–37,40,42–46} Mature markers that arise late in differentiation and persist on mature RBCs, such as CD235a and CD233 (Band 3) have aided the analysis of erythroblast populations and reticulocytes.^34–36,40 Conventionally, the BFU-E to CFU-E transition has been associated with loss of CD34, acquisition of CD36, and increased expression of CD71 and CD105, where BFU-E are characterized as CD34⁺CD36⁻ and CFU-E as CD34⁻CD36⁺.^40,43,44,47 Notably, an intermediate CD34⁺CD36⁺ population has also been reported.^42,47–50 While some studies have found expression of CD71 and CD105 in bipotent MEPs, others have shown that CD71 and CD105 arise later, at, or just before commitment to the erythroid lineage.^37,43–45

Early erythropoiesis cells have been shown to be heterogenous and immunophenotypically resolved in both humans and mice. In a prospective analysis of human BM cells, EE progenitors were resolved into 4 distinct subsets based on the ­expression of CD34 and CD105 on CD235a⁻CD117⁺CD45RA⁻CD41⁻CD123⁻CD71⁺ cells, including one BFU-E (EP1) and 3 CFU-E populations (EP2-4).⁴⁰ This strategy has been widely applied and has provided valuable information on proteomic changes and effects of cytokines.^51–53 Assessment of IL-3, SCF, and Epo treatment on the 4 sorted EP1-EP4 populations showed a lack of an effect of IL-3 on proliferation or differentiation, whereas SCF increased proliferation and decreased differentiation of all 4 EE populations into erythroblasts, as measured by the percentage of CD235a⁺ cells.⁵³ Epo, however, induced a variable response in the 4 subpopulations. Isolated EP1 and EP2 exhibited significantly higher proliferation than EP3 and EP4, whereas EP3 and EP4 differentiated more efficiently than EP1 and EP2 in response to Epo. Importantly, the lack of Epo increased cell death in EP2-EP4, suggesting that Epo not only plays a role in the proliferation of EP1-EP4 but is also necessary for the survival of EP2-EP4. In another immunophenotyping analysis, 2 CFU-E subtypes were resolved within the CD34⁺CD36⁺ cohort: immature CFU-E, which were CD34⁺CD36⁺CD71^hiCD105^med, and mature CFU-E, which were CD34⁺CD36⁺CD71^hiCD105^hi.⁴⁸ Of these 2, the immature CFU-E were found to be much more responsive to SCF and dexamethasone treatment. Additionally, in a single cell mass cytometry analysis of human cord blood (CB) CD34⁺ HSPC cultures, 4 CFU-E populations were detected, 2 of which were CD34⁻CD45RA⁻CD123⁻CD36⁺CD71⁺CD235a⁻ but differed based on CD38 positivity: CFU-E1 were CD38⁺ and CFU-E2 were CD38⁻.⁵⁴ The other 2, CFU-E3 and CFU-E-like, expressed the same combination of markers as CFU-E2, but at a lower level, and the CFU-E-like lacked CD36 and CD71. Studies in mice have also used CD71 and CD105, and in addition CD150, to stratify Lin⁻Kit⁺CD55⁺CD49f⁻CD105⁺ EE populations, identifying one BFU-E subtype as CD150⁺CD71⁻ and 2 CFU-E subtypes based on scaled expression of CD71: early CFU-E (CD71^loCD150⁺) and late CFU-E (CD71^hiCD150⁺).^55–57 All 3 EE populations were observed in murine BM and spleen; however, only early CFU-E exhibited significantly increased proliferation in response to Epo treatment.⁵⁶

In this study, we utilized immunophenotyping of ex vivo cultures of human BM CD34⁺ HSPCs to systematically characterize the Epo-dependence of heterogeneous EE population dynamics. Based on profiles of established markers, we resolved the EE compartment into 5 subpopulations, identifying 2 CD34⁺ BFU-E, termed early BFU-E (earlyB) and late BFU-E (lateB), and 3 CD34⁻ CFU-E subtypes: termed early CFU-E (earlyC), mid CFU-E (midC), and late CFU-E (lateC). These 5 EE populations could also be resolved prospectively in BM cells and detected during ex vivo differentiation of CB CD34⁺ HSPCs. Additionally, comparison of cells undergoing differentiation in the presence and absence of Epo identified the conversion of midC to lateC, which is associated with the acquisition of the CD71^hiCD105^hi immunophenotype, as the crucial Epo-dependent step. In medium lacking Epo, there was a significant decrease in the formation of lateC with a concomitant accumulation of earlyC and/or midC. Transition of earlyB to lateC was accompanied by a steady increase in CD36 expression, which, in combination with loss of CD34 and gain of CD71, identified 2 potential differentiation routes: one progresses through earlyC and the other through lateB formation. Overall, this study resolves heterogeneity in EE cells to pinpoint the key transitional stage which requires Epo for further development to the erythroid lineage during the differentiation of human BM and CB HSPCs.

Materials and methods

Culture and differentiation of CD34⁺ HSPCs

Human BM-derived CD34⁺ HSPCs and MNCs, all from female donors between the ages of 13 to 34 years, were purchased from Ossium Health and Stem Cell Technologies, respectively. For all donor samples, quality control by flow cytometry, indicating >90% viability, via staining with 7AAD, and >90% CD45 and CD34 positivity, measured by antibody staining, was provided by the supplier and confirmed by in-house immunophenotyping analysis. Furthermore, tests indicating negative results for HIV1/2 antibody, HBV surface antigen, HBV core antibody, HCV antibody, HIV/HBV/HCV nucleic acid amplification, cytomegalovirus, and syphilis were provided for all donor cells. CB units were obtained from Valleywise Health Medical Center, Arizona through a material transfer agreement. CB CD34⁺ HSPCs were isolated as previously described.^58–60 MS-5 cells (used as feeder cells for co-­culture with CD34⁺ HSPCs) were maintained in Dulbecco’s modified Eagle’s medium (DMEM) with high glucose and L-glutamine (Thermo Fisher Scientific), 10% fetal bovine serum (FBS; negative for mycoplasma; Omega Scientific), and 1% penicillin-streptomycin (Pen-Strep, Sigma-Aldrich). MS-5 cells are checked annually for the absence of mycoplasma contamination.

Culture of BM and CB CD34⁺ HSPCs was performed as previously described.^58–60 Briefly, HSPCs were incubated overnight at a density of 1 x 10⁵ cells/well of a 48-well plate in 200 µL of stimulation medium [serum-free X-VIVO 15 (Lonza, Basel, Switzerland) containing 2 mM L-glutamine, 50 ng/mL stem cell factor (SCF), 50 ng/mL thrombopoietin (TPO), 20 ng/mL interleukin-3 (IL-3), and 50 ng/mL Flt3-Ligand (Flt-3)]. The stimulated CD34⁺ cells were collected and plated onto MS-5 cells at a density of 3 × 10⁴ cells/well of a 48-well plate for analysis on day 4 and 1 × 10⁴ cells/well for analysis on day 7 or 10 in complete medium [DMEM with high glucose and L-glutamine, 10% FBS, and 1% Pen-Strep, supplemented with 50 ng/mL Tpo (Miltenyi Biotec), 5 ng/mL SCF, 5 ng/mL IL-3, and 5 ng/mL Flt-3, with or without 4U/mL Epo (BioBasic or StemCell Technologies)]. Half-medium change was performed on days 4 and 7.

For reculture experiments, the individual EE cells were sorted into complete medium (containing Epo) supplemented with 20% FBS as described below. The sorted cells were counted and plated on MS-5 cells at a density of either 3 × 10⁴ cells/200 µL or 1 × 10⁴ cells/200 µL/well in a 48-well plate in complete medium and incubated as described above.

Immunophenotyping and cell sorting

For flow cytometry analyses, cells were harvested and stained with fluorescently labeled antibodies as previously described.^58–60 Briefly, cells were washed and resuspended in phosphate-­buffered saline (PBS), counted using the Countess 3-FL Automated Cell Counter (Thermo Fisher Scientific), and then stained with an antibody cocktail, as indicated in the figures. Stained cells were washed with PBS, fixed with 1% paraformaldehyde (PFA) (Sigma Aldrich, St Louis, Missouri), and resuspended in flow cytometry buffer (PBS containing 2% FBS and 5 mM EDTA). Single-stained controls were prepared using compensation beads (UltraComp eBeads from Thermo Fisher Scientific) as per the manufacturer’s recommendations. All flow cytometry analysis was performed on a Canto II (BD Biosciences), and data were analyzed using FlowJo (version 10.9.0).

For fluorescence-activated cell sorting (FACS), cells were collected on day 4 of culture, washed, counted, and stained as described above. Stained cells were washed and resuspended in flow cytometry buffer at a density of 3–5 × 10⁶ cells/mL and then sorted into 4 populations (earlyB, lateB, earlyC/midC, and lateC) based on immunophenotypes (Table S1) using a FACS Aria II (BD Biosciences, San Jose, CA) into complete medium supplemented with 20% FBS. Sorted cells were counted and plated either in complete myeloid medium for reculture or in semisolid medium for CFC assays.

Colony forming cell assay

For CFC assays, MethoCult classic (H4434), classic without Epo (H4534), Epo-only (H4330), or without cytokines (H4230) were used (STEMCELL Technologies; Cambridge, MA). Cells were diluted to a density of 2.5–5 × 10³ cells/mL in MethoCult medium and plated in triplicate in 35 mm tissue culture dishes. After 14 days of incubation, colonies were counted and characterized based on the criteria described by Palis and Koniski.⁶¹ Images were obtained using an Olympus IX2-SP inverted microscope under 40× and 100× magnification.

Statistical analysis

At least 3 biological replicates were performed for each experiment; each biological replicate represented a unique HSPC donor. The exact number of replicates for individual experiments is indicated in the figure legends. Graphical representation plots and statistical analysis by Student’s T-test were performed in GraphPad Prism (version 10.0.3; La Jolla, CA). Data is represented as mean ± standard error of the mean (SEM), and P values <.05 were considered significant.

Results

Epo-associated dynamics of CD71 and CD105 expression

To examine Epo-induced development of erythroid cells, we applied a previously established protocol for ex vivo differentiation of human BM-derived CD34⁺ HSPCs.^58–60 Longitudinal immunophenotypic assessment of CD235a in cultures grown in medium containing Epo (+Epo) or lacking Epo (−Epo) revealed a robust CD235a⁺ population in +Epo cultures but not in −Epo cultures (Figure S1A and B). Additionally, assessment of erythroblast populations using the flow cytometry strategy developed by Yan et al. showed the presence of all 5 erythroblast populations (ProE, early BasoE, late BasoE, PolyE, and OrthoE) in +Epo but none in −Epo cultures (Figure S1C).⁴⁰ Although CD235a expression indicated an increase in total erythroblasts, there was some variability in the trend exhibited by the individual erythroblast populations (Figure S1D). The percentage of ProE decreased from day 4 to 10, early BasoE, late BasoE, and PolyE reached maximum levels on day 7, and OrthoE, the final erythroblast stage, increased steadily, indicating a continuous transition through TED.

Examination of markers that have previously been used to distinguish and characterize EE and TED populations, including CD49d, CD117, CD36, CD71, and CD105 showed that the percentages of cells expressing CD71 (day 4), CD105 (days 4 and 7), and CD36 (days 4, 7, and 10) were significantly reduced in −Epo cultures, while those expressing CD49d and CD117 were unaffected (Figure 1A and D, and Figure S2).^{17,35,37,40,43,44,62} CD71 and CD105 exhibited low and high expression levels during the course of differentiation (Figure 1B and E). Notably, the CD71^hi and CD105^hi cells were significantly reduced in −Epo cultures (Figure 1C and F). This decrease was accompanied by an increase in CD71^lo but not CD105^lo cells. This scaled expression and Epo-dependence was observed with CD71 and CD105 antibodies labeled with 2 different fluorophores and has been reported in previous studies (Figure 1 and Figure S4A and B).^{36,39,40,44,45} Overall, the results confirmed the Epo-dependent development of erythroblasts in our culture system and indicated Epo’s influence on the surface expression of CD71 and CD105.

Figure 1.

Epo induces high expression of CD105 and CD71 in human BM ex vivo cultures. BM CD34⁺ HSPCs were cultured in medium containing or lacking Epo. Surface expression of CD105 and CD71 was examined by flow cytometry, and percentages of positive cells within LiveLin⁻CD123⁻ cells are shown. (A) Overall expression and (B) representative plots for CD71 expression. Gates for low and high expression as determined by anti-CD71/FITC antibody staining are shown. (C) Percentages of CD71^hi (top) and CD71^lo (bottom) cells in each condition. (D) Overall expression and (E) representative scatter plots for CD105 expression. Gates for low and high expression as determined by anti-CD105/PE antibody staining are shown. (F) Percentages of CD105^hi (top) and CD105^lo (bottom) cells in each condition. P-values ≤ .05 were considered significant; *P < .05, **P < .01, ***P < .001 (n = 5). Abbreviations: Epo; erythropoietin, Lin; lineage markers, including CD3, CD10, CD11b, CD19, CD41.

Immunophenotypic resolution of early erythroid progenitors

Utilizing previously described markers, we developed an immunophenotyping strategy for monitoring the Epo-­dependent transition within EE progenitors, BFU-E and CFU-E (Figure 2). First, CD3, CD10, CD19, CD11b, CD41, and CD123 were used to gate out non-erythroid lineages and progenitors; CMP and GMP are both expected to be CD123 positive and, therefore, should be eliminated with this panel. From the LiveLin⁻CD123⁻ cells, CD71⁺CD235a⁻ cells were then selected and from this population, only CD49d⁺CD117⁺ cells were considered for further analysis, since these markers have been shown to be expressed in EE progenitors (Figure 2A i and ii).^40,44 As BFU-E, but not CFU-E, are expected to express CD34, all CD105⁺ cells were segregated based on CD34 positivity (Figure 2A iii). Therefore, all EE cells were LiveLin⁻CD123⁻CD235a⁻CD71⁺CD49d⁺CD117⁺, and within this population, total BFU-E were additionally CD34⁺CD105⁺, whereas total CFU-E were CD34⁻CD105⁺. Further stratification of total BFU-E and CFU-E populations based on CD105 and CD71 expression resolved heterogeneity within them (Figure 2A iv and v). BFU-E consisted of 2 populations: a large CD34⁺CD71^loCD105^lo population and another smaller CD34⁺CD71^hiCD105^lo population that were termed early BFU-E (earlyB) and late BFU-E (lateB), respectively. Within CFU-E, 3 populations emerged: CD34⁻CD71^loCD105^lo early CFU-E (earlyC), CD34⁻CD71^hiCD105^lo mid CFU-E (midC), and CD34⁻CD71^hiCD105^hi late CFU-E (lateC) (Table S1).

Figure 2.

Immunophenotyping strategy for Epo-dependent resolution of EE populations. Cells from ex vivo cultures of BM CD34⁺ HSPCs incubated in medium containing or lacking Epo were analyzed on days 4, 7, and 10 for the indicated surface markers by flow cytometry (Panel 1 in Table S2). (A) Representative scatter plots from the analysis of EE populations in +Epo cultures on day 4. CD235a⁻CD71⁺ cells are gated from LiveLin⁻CD123⁻ cells (i), then CD117⁺CD49d⁺ cells are selected (ii), and CD105⁺ cells are stratified based on CD34 expression (iii)—all CD34⁺CD105⁺ cells are considered BFU-E and all CD34⁻CD105⁺ cells are considered CFU-E. BFU-E and CFU-E were further stratified based on CD71 and CD105. BFU-E showed 2 populations: CD71^loCD105^lo (earlyB) and CD71^hiCD105^lo (lateB) (iv). Within CFU-E, 3 populations were observed: CD71^loCD105^lo (earlyC), CD71^hiCD105^lo (midC), and CD71^hiCD105^hi (lateC) (v). (B) Representative scatter plots for analysis of EE populations in cultures lacking Epo on day 4 by the same strategy as in (A). Comparison of percentage of erythroid progenitor populations in BM HSPC cultures: (C) total BFU-E (CD34⁺CD105⁺) as a percentage of CD49d⁺CD117⁺ cells, (D) earlyB (CD34⁺CD71^loCD105^lo) and (E) lateB (CD34⁺CD71^hiCD105^lo) as a percentage of total BFU-E; (F) total CFU-E (CD34⁻CD105⁺) as a percentage of CD49d⁺CD117⁺ cells; (G) lateC (CD34⁻CD71^hiCD105^hi), (H) earlyC (CD34⁻CD71^loCD105^lo), and (I) midC (CD34⁻CD71^hiCD105^lo) as a percentage of total CFU-E. P-values ≤ .05 were considered significant (*P < .05, **P < .01, ***P < .001; n = 8 for EE populations). Abbreviations: BasoE, basophilic erythroblast; BFU-E, burst forming unit-erythroid; CFU-E, colony forming unit-erythroid; earlyB, early BFU-E; earlyC, early CFU-E; Epo, erythropoietin; lateB, late BFU-E; lateC, late CFU-E; Lin, lineage markers; midC, mid CFU-E; OrthoE, orthochromatic erythroblast; PolyE, polychromatic erythroblast; ProE, proerythroblast.

Assessment of CD36, which has previously been employed to distinguish BFU-E and CFU-E cells, showed that, amongst BFU-E, the fraction of CD36⁺ cells was markedly higher in lateB (∼59%) than in earlyB (∼1.5%) (Figure 3A i and ii and C).^38,43,63 Within CFU-E, the CD36⁺ cells increased from ∼25% in earlyC to ∼81% in midC and ∼99% in lateC (Figure 3A iii–v and C). These results indicate that within the EE compartment, there is a continuum of cells that are simultaneously losing CD34 and gaining CD36 expression. The observed patterns of CD34, CD71, CD105, and CD36 suggest the existence of 2 alternative differentiation routes (Figure 3D). In one route, loss of CD34 and gradual gain of CD36 transition earlyB to earlyC, both of which are CD71^loCD105^lo. EarlyC then sequentially differentiates through midC to lateC. In the other route, the onset of CD36 expression precedes the downregulation of CD34, leading to the transition of earlyB to lateB, which are CD71^hiCD105^lo. In this alternative route, the earlyC stage is likely bypassed, with lateB transitioning directly to midC, which forms lateC.

Figure 3.

CD36 expression within EE populations reveals 2 potential differentiation routes. Cells from ex vivo cultures of BM CD34⁺ HSPCs incubated in medium containing or lacking Epo were analyzed for expression of CD36 within all 5 EE subpopulations by flow cytometry. Representative scatter plots from day 4 analysis of cells cultured (A) with and (B) without Epo, depicting CD36 expression within each of the 5 EE populations—(i) earlyB, (ii) lateB, (iii) earlyC, (iv) midC, and (v) lateC. (C) Quantification of CD36 expression in all 5 EE populations on day 4 of culture in medium containing Epo (n = 5). Bar graphs for days 7 and 10 are shown in Figure S3. (D) Proposed differentiation routes for the transition of progenitors in EE. Abbreviations: BFU-E, burst forming unit-erythroid; CFU-E, colony forming unit-erythroid; earlyB, early BFU-E; earlyC; early CFU-E; Epo, erythropoietin; lateB, late BFU-E; lateC; late CFU-E; midC; mid CFU-E.

To confirm that the observed stratification of EE cells was not an artifact of the combination of fluorophores that were used in the antibody panel, we designed a second panel in which CD49d, CD71, CD105, and CD36 antibodies were conjugated to different fluorophores (Figure S4). The list of antibodies for the 2 panels is provided in Table S2. Application of the second panel also demonstrated the presence of 5 EE populations at percentages similar to those observed with the initial fluorophore combination (Figures S4C and S5). Additionally, the pattern of CD36 expression was also similar to that seen with the first panel (Figure S4E). Put together, these results confirm that heterogeneity within EE cells can be resolved based on CD71 and CD105 expression and that the rates of CD34 downregulation and CD36 upregulation may vary within each cell.

Erythropoietin is required for the development of late CFU-E

To examine the role of Epo in population dynamics within EE, cells from BM-derived CD34⁺ HSPCs cultured in +Epo medium were compared to those cultured in −Epo medium by immunophenotyping. Assessment of total BFU-E in +Epo culture showed an increase from day 4 to 7, which was followed by maintenance of their levels until day 10; this trend was largely sustained in the absence of Epo (Figure 2A iii, B iii, and C). In line with this observation, percentages of earlyB and lateB were also similar (Figure 2A iv, B iv, D, and E). In the presence of Epo, the percentage of total CFU-E was significantly higher but decreased steadily from day 4 to 10 (Figure 2F). The lack of difference in CFU-E percentages in +Epo and −Epo conditions on day 10 is likely due to their progression into TED in the presence of Epo. Interestingly, examination of the individual CFU-E populations revealed differential effects within the subtypes. The percentage of lateC was significantly lower in −Epo compared to +Epo cultures on all 3 days of assessment (Figure 2A v, B v, and G). Conversely, the percentages of both earlyC and midC were higher in the absence of Epo, but significant only on day 4 and not on days 7 and 10, most likely due to the continued differentiation (Figure 2H–I). These results suggest that the absence of Epo significantly diminishes the production of lateC with concomitant accumulation of earlyC and midC. The absence of Epo did not have a significant effect on earlyB and lateB or their transition to earlyC or midC, respectively. Similarly, analysis using the second antibody panel indicated comparable percentages of EE cells, absence of an effect of Epo on total BFU-E and BFU-E subpopulations, and an accumulation of earlyC accompanied by considerably reduced production of lateC in cultures lacking Epo (Figures S4C and D and S5).

In previous studies, BFU-E and CFU-E have been characterized as CD34⁺CD36⁻ and CD34⁻CD36⁺, respectively.^39,43 Interestingly, examination by both antibody panels employed in this study showed that CD36 expression within BFU-E and CFU-E populations was unaffected by the absence of Epo (Figure 3A and B, and Figures S3 and S4E and F). Further, application of the strategy proposed by Li et al. for the analysis of EE populations in +Epo and −Epo cultures failed to resolve Epo-dependence (Figure S6A). Furthermore, Yan et al. demonstrated heterogeneity within EE progenitors and proposed strategies for the resolution of progenitors and erythroblasts.^40,42 Based on the assessment of CD105 and CD235a, we were able to recapitulate the erythroblast populations within the CD235a⁺ cells (Figure S1C). However, we were unable to separate the BFU-E and CFU-E populations based on the patterns of CD34 and CD105 (Figure S6B). This could be because a CD34⁺ population that was CD105^hi was not observed in our culture system.

All BFU-E and CFU-E subtypes are present in human bone marrow

To ascertain that all observed EE subtypes were present in human BM, we performed prospective analyses. For BFU-E, frozen BM CD34⁺ cells were incubated overnight in stimulation medium, which does not contain Epo, and then analyzed by immunophenotyping (Table S1). Examination of CD34 expression revealed that >90% of the cells were CD34⁺, suggesting that the short overnight incubation did not result in significant progression beyond the HSPC stages. Immunophenotyping of these cells exhibited the presence of earlyB and lateB, with the percentage of lateB being slightly higher than that seen in BM CD34⁺ cell cultures (Figure 4A). To analyze CFU-E subtypes, total BM mononuclear cells (MNCs) were also immunophenotyped after an overnight incubation in stimulation medium. This analysis revealed the presence of both BFU-E and all 3 CFU-E populations in BM MNCs (Figure 4B). Thus, all EE populations were prospectively discernible in BM cells.

Figure 4.

Prospective analysis resolves all 5 EE populations in human BM. BM cells were thawed, stimulated overnight in medium lacking Epo (see Materials and Methods), then analyzed using the same immunophenotyping strategy as in Figure 2A. (A) EarlyB and lateB populations in BM-derived CD34⁺ HSPCs. (B) EE populations in BM-derived MNCs. CD235a⁻CD71⁺ cells are gated from LiveLin⁻CD123⁻ cells, then CD117⁺CD49d⁺ cells are selected, and CD105⁺ cells are stratified based on CD34 expression—all CD34⁺CD105⁺ cells are considered BFU-E, and all CD34^-CD105⁺ cells are considered CFU-E. BFU-E and CFU-E were further stratified based on CD71 and CD105. BFU-E showed 2 populations: CD71^loCD105^lo (earlyB) and CD71^hiCD105^lo (lateB). Within CFU-E, 3 populations were observed: CD71^loCD105^lo (earlyC), CD71^hiCD105^lo (midC), and CD71^hiCD105^hi (lateC). ­Abbreviations: Lin, lineage markers; MNC, mononuclear cells.

Isolated early erythroid populations exhibit expected differentiation trajectories

To further investigate the differentiation potential of each of the 5 EE populations, they were FACS-sorted on day 4 and recultured in suspension for immunophenotypic analysis and in semi-solid media for colony-forming cell (CFC) assays. In initial attempts, the numbers of earlyC and midC were found to be insufficient for downstream analysis, so they were sorted together. For suspension cultures, sorted cells were replated in complete medium containing Epo. During reculture, proliferation was observed to be highest for earlyB and lowest for lateC, while lateB and earlyC/midC exhibited similar intermediate expansion capacity (Figure S7A). On day 4, immunophenotyping of cells from cultures of sorted earlyB detected mostly earlyB, but by day 7, a large number of earlyC were seen, along with a few midC, lateC, and erythroblasts as well (Figure 5A). Sorted lateB, on the other hand, produced primarily lateB, midC, and lateC on day 4, which progressed to all erythroblast populations by day 7 (Figure 5B). Notably, earlyC were not observed in cultures of sorted lateB, thereby indicating that lateB may be transitioning directly to midC. Reculture of the earlyC/midC pool yielded all CFU-E populations and all erythroblast populations, with mainly ProE and BasoE arising by day 4 and PolyE and OrthoE forming by day 7 (Figure 5C). Sorted lateC also efficiently produced erythroblasts (Figure 5D). MidC and lateC were still present in lateB cultures on day 7, while this was not the case for cultures of earlyC/midC or lateC, because all cells had entered TED, as seen in the erythroblast scatter plots (Figure 5B–D). These results demonstrate the proliferation and erythroid differentiation potential of the sorted individual EE populations. Further, the inability of sorted lateB to produce earlyC supports the presence of 2 differentiation routes (Figure 3D).

Figure 5.

Sorted EE populations differentiate via the expected trajectories and enter TED. Cells from ex vivo cultures of BM-derived CD34⁺ HSPCs were incubated in complete medium containing Epo. EE populations were isolated by FACS on day 4, recultured in suspension, and analyzed using the strategies described in Figure 2A (EE) and by Yan et al.⁴⁰ (TED). Representative scatter plots for BFU-E, CFU-E, and erythroblast subpopulations arising from cultures of sorted (A) earlyB, (B) lateB, (C) earlyC/midC mix, and (D) lateC on days 4 and 7 of reculture. Abbreviations: BasoE, basophilic erythroblast; BFU-E; burst-forming unit erythroid; CFU-E, colony-forming unit erythroid; OrthoE, orthochromatic erythroblast; PolyE; polychromatic erythroblast; ProE, proerythroblast.

For additional functional analysis by CFC assays, the isolated EE cells were plated in semisolid medium containing either all cytokines (including Epo) or Epo alone, as conventionally, CFU-E have been characterized by their ability to form colonies in Epo-only medium.^40,43,61 Analysis of colonies formed by unsorted cells harvested on day 4 of culture confirmed the ability of semisolid medium to support the formation of colony-forming unit granulocyte-monocyte (CFU-GM), BFU-E, and CFU-E colonies (Figure 6A and B and Figure S7B). As previously reported, 2 types of BFU-E colonies were observed: large colonies that consisted of multiple clusters of hemoglobinized cells (large BFU-E) and smaller single-cluster colonies that were visible at 40× (small BFU-E).^61,64 The CFU-E colonies, on the other hand, were difficult to visualize at 40× but easily seen at 100× (Figure 6A and B).

Figure 6.

Colony-forming potential of sorted EE populations. Cells from ex vivo cultures of BM-derived CD34⁺ HSPCs were incubated in complete medium containing Epo. EE populations were isolated by FACS on day 4 according to the strategy in Figure 2A, then replated in methylcellulose containing SCF, IL-3, GM-CSF, and Epo or Epo-only. Representative images of CFU-GM, large BFU-E, small BFU-E, and CFU-E colonies at (A) 40× and (B) 100× magnification. Quantification of colonies formed by sorted (C) earlyB, (D) lateB, (E) earlyC/midC mix, and (F) lateC (n = 3). EE cells were also sorted based on CD36 expression within total BFU-E (CD34⁺CD105⁺) and total CFU-E (CD34⁻CD105⁺) (Figure 2A iii) to generate 4 populations: CD36⁺-BFU-E, CD36⁻-BFU-E, CD36⁺-CFU-E, and CD36⁻-CFU-E. Quantification of colony formation for each population is shown for (G) complete medium and (H) Epo-only medium (n = 2). All colonies were counted on day 14 and normalized to 10,000 cells. Abbreviations: BFU-E, burst-forming unit erythroid; CFU-E, colony-forming unit erythroid; CFU-GEMM, colony-forming unit-granulocyte, erythrocyte, monocyte, megakaryocyte; CFU-GM, colony-forming unit-­granulocyte monocyte; Epo; erythropoietin.

Examination of colonies formed from the sorted EE populations showed that earlyB formed small and large BFU-E colonies in complete medium and did not produce any colonies on Epo-only plates (Figure 6C). Interestingly, earlyB also formed CFU-GM colonies, indicating some plasticity. LateB formed both small and large BFU-E and CFU-E colonies in complete medium; surprisingly, they also produced CFU-E colonies in Epo-only medium (Figure 6D). While the earlyC/midC mix produced small BFU-E and CFU-E colonies in complete medium, they also formed CFU-E colonies in Epo-only medium (Figure 6E). The last EE population, lateC, yielded primarily CFU-E colonies in both types of media (Figure 6F). These results suggest that earlyB, which are CD34⁺CD71^loCD105^lo but largely CD36⁻ exhibit some plasticity, retaining the ability to form CFU-GM colonies. Other studies have also reported the formation of GM colonies from sorted BFU-E.^{40,  43,  65} GM and GEMM colonies from CD34⁺CD36⁻ cells have also been observed.⁵⁰ The other 4 EE populations are strictly erythroid-lineage committed. They also suggest that lateB and earlyC/midC populations, representing transitional stages, form similar types of colonies. The lateC represent a subset of CD34⁻CD36⁺ cells that only form CFU-E colonies and have completed the EE transition but have not yet entered TED, as they are CD235a⁻.

In the CFC-assays, we noted the formation of CFU-E colonies by sorted lateB and the earlyC/midC mix in complete medium and, surprisingly, also in Epo-only medium. Like lateC, these 2 subpopulations contain CD36⁺ cells. To examine if CD36 expression was a determinant in colony phenotype, total BFU-E (Lin⁻CD123⁻CD235a⁻CD71⁺CD49d⁺CD117⁺CD34⁺CD105⁺) and CFU-E (Lin⁻CD123⁻CD235a⁻CD71⁺CD49d⁺CD117⁺CD34⁻CD105⁺) were sorted based on CD36 into 4 populations—CD36⁺-BFU-E, CD36⁻-BFU-E, CD36⁺-CFU-E, and CD36⁻-CFU-E—and plated in either complete or Epo alone medium. In this analysis, CD36⁻-BFU-E exhibited the ability to form CFU-GM colonies and formed large and small BFU-E colonies in complete medium but did not form any colonies in Epo-only medium (Figure 6G and H). CD36⁺-BFU-E did not form CFU-GM but formed large and small BFU-E and CFU-E colonies. Interestingly, CD36⁺-BFU-E formed CFU-E colonies also in Epo-only medium (Figure 6H). CD36⁻-CFU-E primarily formed large and small BFU-E colonies in complete medium and none in Epo-only medium. CD36⁺-CFU-E formed small BFU-E and CFU-E colonies in complete medium, and just CFU-E colonies in Epo-only medium. The ability of sorted CD36⁺-BFU-E and CD36⁺-CFU-E, but not CD36⁻-BFU-E and CD36⁻-CFU-E, to form CFU-E colonies in Epo-only medium suggests that CD36 expression may be associated with the CFU-E colony phenotype. Additionally, the ability of CD36⁻-BFU-E but not of CD36⁺-BFU-E to form CFU-GM colonies suggests that CD36 expression may be linked to loss of plasticity. Overall, analysis by reculturing and CFC assays confirmed the potential of the isolated EE progenitor subtypes to progress along the erythroid trajectory and supported the existence of 2 parallel routes, one of which transitions through earlyC and the other through lateB, and also indicated a role for CD36 in determining the CFU-E colony phenotype.

Occurrence of EE subtypes during neonatal erythroid differentiation

To examine if the EE subpopulations were detectable during neonatal erythroid differentiation, we cultured CB-derived CD34⁺ HSPCs in +Epo and −Epo conditions. Immunophenotyping of the CB cultures revealed similar EE population dynamics as observed for BM cultures (Figure 2). As such, the expression of CD235a and all erythroblast populations was seen only in +Epo cultures (Figure S8). There was also a significant decrease in the percentage of CD71^hi, CD105^hi, and CD36⁺ cells and an accumulation of CD71^lo cells in the absence of Epo (Figures S9 and S10E and F). Furthermore, all 5 EE populations identified in BM cells were also resolved in CB cultures (Figure 7A). As with BM, the absence of Epo did not affect the formation of BFU-E, total, or the subtypes (Figure 7A and B and Figure S11). There was also a drastic and significant decrease in lateC formation in the absence of Epo that was accompanied by the accumulation of earlyC but not midC, thereby suggesting the Epo-dependence of lateC formation also in the neonatal system (Figure 7A–E). Finally, as with BM-­derived cells, CD36⁺ cells were <2% within earlyB but rose to ∼50% in lateB and also rose within CFU-E subtypes from ∼20% in earlyC, ∼90% in midC, and essentially 100% in lateC (Figure S12). CD36 expression within each population was still not affected by the absence of Epo.

Figure 7.

Occurrence of EE subtypes and Epo-dependence of lateC formation in CB-derived CD34⁺ HSPC cultures. Cells from ex vivo cultures of CB-derived CD34⁺ HSPCs incubated in medium containing or lacking Epo were analyzed on days 4, 7, and 10 using the immunophenotyping strategy described in Figure 2A. CD235a⁻CD71⁺ cells are gated from LiveLin⁻CD123⁻ cells (i), then CD117⁺CD49d⁺ cells are selected (ii), and CD105⁺ cells are stratified based on CD34 expression (iii)—all CD34⁺CD105⁺ cells are considered BFU-E, and all CD34⁻CD105⁺ cells are considered CFU-E. BFU-E and CFU-E were further stratified based on CD71 and CD105. BFU-E showed 2 populations: CD71^loCD105^lo (earlyB) and CD71^hiCD105^lo (lateB) (iv). Within CFU-E, 3 populations were observed: CD71^loCD105^lo (earlyC), CD71^hiCD105^lo (midC), and CD71^hiCD105^hi (lateC) (v). Representative scatter plots from day 7 of analysis for cells cultured in medium (A) with and (B) without Epo. Comparison of percentage of CFU-E subpopulations in CB HSPC cultures: (C) earlyC (CD34⁻CD71^loCD105^lo), (D) midC (CD34⁻CD71^hiCD105^lo), and (E) lateC (CD34⁻CD71^hiCD105^hi) are all depicted as a percentage of total CFU-E. P-values ≤ .05 were considered significant (*P < .05, **P < .01; n = 4). (F) Schematic representation of the 2 routes to the formation of lateC. One passes through the earlyC stage, whereas the other bypasses earlyC due to direct conversion of lateB to midC. EarlyB to lateB and earlyC to midC transitions are accompanied by an increase in the expression of CD71 and gain of CD36 positivity, whereas earlyB to earlyC and lateB to midC transitions are accompanied by the loss of CD34 positivity. Both paths converge at midC, which requires Epo to transition to lateC, characterized as CD34⁻CD71^hiCD105^hi within the CD235a⁻ compartment. Abbreviations: BFU-E, burst-forming unit erythroid; CFU-E, colony-forming unit erythroid; earlyB, early BFU-E; earlyC, early CFU-E; Epo, erythropoietin; lateB, late BFU-E; lateC, late CFU-E; Lin, lineage markers; midC, mid CFU-E

To discern any differences in erythroid population dynamics between adult and neonatal populations, we simultaneously cultured BM- and CB-derived cells in complete medium containing Epo. Assessment of marker expression and cell stage progression on days 4, 7, and 10 revealed similar percentages of individual markers, the BFU-E and CFU-E subpopulations, and CD36 expression within them (Figure S13). Comparison of erythroblast populations revealed a significantly higher percentage of ProE and early BasoE in the adult BM cultures than in neonatal CB cultures, although the differences in percentages were relatively low (Figure S14A and B). These results likely indicate a slightly delayed entry into TED in CB cultures, which is in alignment with a previously reported study comparing adult peripheral blood (PB) and CB cultures.⁴² Overall, the analysis demonstrates that, as with differentiation of BM CD34⁺ cells, the 5 EE subpopulations arise during the differentiation of CB CD34⁺ cells and likely progress through 2 differentiation paths.

Discussion

By immunophenotyping transiently expressed markers on cells from ex vivo cultures of BM-derived HSPCs, we resolved heterogeneous EE progenitor populations and identified the key Epo-dependent transition (Figure 7F). Based on the expression patterns of CD34, CD71, and CD105 within Lin⁻CD123⁻CD235a⁻CD71⁺CD49d⁺CD117⁺CD105⁺ cells, we detect 5 EE subtypes: 2 BFU-E and 3 CFU-E. The 2 BFU-E, earlyB and lateB are CD34⁺ but can be distinguished as CD71^loCD105^lo and CD71^hiCD105^lo, respectively. The 3 CFU-E, earlyC, midC, and lateC, are CD34⁻ and distinguishable as CD71^loCD105^lo, CD71^hiCD105^lo, and CD71^hiCD105^hi, respectively. Additionally, prospective analysis detected the 2 BFU-E subtypes and 3 CFU-E subtypes in BM cells, and, after isolation, each individual population exhibited the capacity to continue differentiation along the erythroid route. Furthermore, all 5 EE populations were also resolved in ex vivo cultures of CB-derived HSPCs at levels similar to those observed in BM cultures. Importantly, Epo was found to be essential for the acquisition of the CD71^hiCD105^hi immunophenotype during the midC to lateC transition in both BM and CB. Thus, while several previous studies have demonstrated heterogeneity within EE cells, this study identifies the explicit Epo-dependent transition. Additionally, the expression pattern of CD71, CD105, and CD36 allowed us to decipher 2 routes to lateC formation (Figure 7F). In the first, loss of CD34 expression occurs prior to the gain of CD36 and progresses from earlyB→earlyC→midC→lateC. The second pathway, on the other hand, bypasses earlyC because of the onset of CD36 expression in CD34⁺ cells, and the progression follows from earlyB→lateB→midC→lateC.

Although not entirely Epo-dependent, both CD71 and CD105 are dynamically expressed and sensitive to Epo in EE cells. While the function of CD71 (transferrin receptor) is well characterized, the role played by CD105 (endoglin) in erythroid differentiation is less clear.^66,67 Also known as transforming growth factor β (TGF-β) receptor type III, CD105 associates with type I and II receptors upon binding of TGF-β superfamily ligands to regulate gene expression via Smad signaling.^68–70 In mice, CD105 loss was reported to be associated with decreased erythroid output and have a potential role in regulating the proliferation of BFU-E and BasoE.^71–73 Expression of CD105 on immature human erythroid cells has been known for a long time, and its increase in MEP has been positively correlated with erythroid commitment.^{45,  74,75} We observed a basal level of CD105 on MEP, CMP, and GMP by flow cytometry (Schippel and Sharma, unpublished data). Notably, a CD71^hiCD105^hi mature CFU-E population within the CD34⁺CD36⁺ compartment was reported in human PB-, CB-, and BM-derived cells.⁴⁸ LateC identified in the present study is CD34⁻, thus different from the mature CD34⁺CD36⁺ CFU-E. Further characterization of target genes and cellular processes regulated by CD105 is needed to understand its precise role in EE cells.

In several studies, CD36 has been proposed to mark commitment to the erythroid lineage.^{35–37,  44,45,  50,  76} The onset of CD36 expression has been shown to be accompanied by the loss of CD34 during the BFU-E to CFU-E transition.^{43,  47} During differentiation of human PB CD34⁺ cells, a CD34⁺CD36⁺ population that gives rise to both BFU-E and CFU-E was identified.^{42,  50} Accumulation of a CD34⁺CD36⁺ population has also been detected in the PB of patients with hypercortisolemia and is associated with unresponsiveness to dexamethasone, a synthetic glucocorticoid commonly used for anemia treatment.^47–49 In our study, the CD34⁺CD36⁺ lateB were identified as an intermediate population that forms in an Epo-independent manner and transitions directly to the CD34⁻CD36⁺ midC. Interestingly, humans lacking the CD36 protein due to genetic mutations generally develop normally, but have thrombocytopenia, increased bleeding, and RBC abnormalities such as higher RBC distribution width (RDW).^77–79 CD36 knockout mice develop normally, although they also have thrombocytopenia, along with slight but significantly reduced RBC counts.⁸⁰ In contrast, in ex vivo experiments, CD34⁺ HSPCs from an individual lacking CD36 were found to undergo erythroid commitment and differentiation without much effect on either CD71 and CD235a expression or erythroblast development, as assessed by CD49d (α-4 Integrin) and CD233 (Band 3) expression according to the strategy proposed by Hu et al.^{35,  81} However, whether CD36 loss affects other markers, such as CD105, or the enucleation and maturation of erythroblasts remains to be determined. Our CFC analysis showed a potential link between CD36 expression and the CFU-E phenotype. The ability of lateB and earlyC/midC to form CFU-E colonies in Epo-only medium suggests that expression of CD36 may represent a step beyond which erythroid cells lose plasticity and require only Epo as an extracellular cytokine signal to progress through TED. Thus, CD36 may be sufficient to confer independence from other cytokines but not Epo. Additional analysis is needed to functionally characterize the role of CD36 during EE.

Anemia is a common health condition that can arise due to a multitude of health conditions, including iron or vitamin deficiency, inherited RBC disorders, infections (such as malaria), obstetric-gynecological conditions, and chronic diseases that lead to dysregulated erythropoiesis.^82,83 Administration of recombinant human Epo and dexamethasone are prevalent treatments for anemia; however, they are not always effective, and the observed improvements are often temporary.^84,85 Thus, there is a significant need to advance the knowledge of Epo’s influence on EE populations. Our study provides a framework for investigations into the regulation of key developmental stages by other factors that may promote erythroid differentiation to aid the identification of new therapeutic targets.

Author contributions

Natascha Schippel (Conceptualization, Data curation, Formal analysis, Investigation, Methodology, Visualization, Validation, Writing—original draft, Writing—review & editing), Mrinali Kala (Investigation and Methodology), Shalini Sharma (Conceptualization, Data curation, Formal analysis, Funding acquisition, Methodology, Project administration, Resources, Supervision, Investigation, Validation, Visualization, Writing—original draft, and Writing—review & editing)

Supplementary material

Supplementary material is available at Stem Cells online.

Funding

The authors would like to acknowledge funding support to SS from the National Institutes of General Medical Sciences (R01GM127464) and National Cancer Institute (P30CA023074) of the National Institutes of Health, the Valley Research Partnership Program (VRP P1-4009 and VRP77), and the Arizona Biomedical Research Center (ABRC: RFGA2022-010-30). The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institute of Health.

Conflicts of interest

The authors declare no conflicts of interest.

Data availability

The data underlying this article are available in the article and in its online supplementary material.

Ethical statement

Human bone marrow CD34⁺ HSPCs were purchased commercially from Ossium Health, and cord blood units were obtained from Valleywise Health Medical Center, Arizona through a material transfer agreement. No human subjects were recruited; therefore, no additional approvals were required. All experiments were conducted in vitro.