Elucidation of the EP defect in Diamond-Blackfan anemia by characterization and prospective isolation of human EPs

Diamond-Blackfan anemia (DBA) is a rare inherited bone marrow (BM) failure syndrome caused by monoallelic mutations in ribosomal protein genes.^1-3  Normochromic, macrocytic anemia, and reticulocytopenia, the hematologic hallmarks of DBA, stem from a selective paucity of erythroid precursors in the BM.⁴ 

Animal models of DBA and in vitro studies of human unfractionated BM or total CD34⁺ cells have suggested that red cell aplasia in DBA arises from a defect in erythroid progenitors (EPs).^5,6  This defect has been defined functionally by a reduced ability of DBA BM to generate erythroid burst-forming unit (BFU-E) and erythroid colony-forming unit (CFU-E) colonies.^7,8  Since DBA EP have also been reported to have qualitative abnormalities, such as relative erythropoietin insensitivity,^9,10  elucidation of the relative roles of these qualitative and quantitative defects in DBA erythropoiesis and determination of the mechanism of steroid responsiveness in 40% patients,⁴  require knowledge of the markers that define the BM cells giving rise to CFU-E and BFU-E.

Although in mice the immunophenotypes of such cells have been described in fetal liver,^11,12  facilitating important insights into erythropoiesis,^12,13  in humans, definitions of corresponding EPs are limited, thus precluding their direct study in normal and aberrant erythropoiesis.

In this study, we sought to dissect the erythropoietic defect in DBA by prospective isolation and characterization of human EP.

BM samples were obtained from patients with DBA and healthy BM donors (see supplemental Table 1, available on the Blood Web site) under approval by the National Research Ethics Service, following written informed consent. Ribosomal protein gene mutations were identified as previously described.¹ 

Flow cytometry, cell sorting, clonogenic assays, liquid cultures, and quantitative polymerase chain reaction (PCR) are described in supplemental Materials and methods. Antibodies and TaqMan quantitative PCR probes are listed in supplemental Tables 2 and 3.

Descriptive statistics show mean ± SEM. The unpaired Student t test or Mann-Whitney U test were used as appropriate (GraphPad Prism, version 6).

Established models of human hematopoiesis suggest that common myeloid progenitors (CMPs) generate granulocyte-monocyte progenitors (GMPs) and megakaryocyte-EPs (MEPs) (Lin⁻CD34⁺CD38⁺CD45RA⁻CD123⁻ cells).¹⁴  The latter overlap with a CD45RA⁻CD123⁻ population enriched in committed EPs.^8,15,16  To study the erythropoietic defect in DBA, we examined the frequency and function of CMP, MEP, and GMP in BM. In transfusion-dependent (TD) DBA patients, these progenitor frequencies were similar to age-matched controls (supplemental Figure 1A). However, DBA MEP gave rise to significantly fewer BFU-E and CFU-E (P < .05) and generated small, poorly hemoglobinized erythroid cell clusters, which failed to develop into fully formed BFU-E (Figure 1A).

Next, we aimed to purify normal BM EP populations using a novel gating strategy. Specifically, we aimed to define surface markers relevant to murine and/or human erythropoiesis^11,17-19  on the CD123⁻CD45RA⁻ fraction of Lin⁻CD34⁺CD38⁺ BM cells (Figure 1B). Sorting into methylcellulose medium revealed that both CD105⁻CD36⁻ and CD105⁺CD36⁺ subpopulations of CD71⁺CD41a⁻ MEP (both of which, as recently described for CD34⁺ cord blood-derived BFU-E/ CFU-E,²⁰  express CD45; supplemental Figure 1B), generated erythroid but not myeloid colonies (Figure 1C). Moreover, CD105⁻CD36⁻ cells generated almost exclusively large and small BFU-E (Figure 1C and supplemental Figure 1C) with a purity of 96% and clonogenicity of 66.8% in single-cell assays (Figure 1D). We termed this early EP (EEP). In contrast, CD105⁺CD36⁺ cells, termed late EP (LEP), generated CFU-E but not BFU-E colonies with a purity of 97% and clonogenicity of 72.3% (Figure 1C-D). Notably, the frequencies of BFU-E and CFU-E derived from EEP and LEP (corrected for their clonogenicity at a single-cell level) were similar to BFU-E and CFU-E frequencies derived from age-matched, total BM-derived CD34⁺ cells (supplemental Figure 1D-E), suggesting that our sorting strategy captures the majority of EP activity in CD34⁺ cells. Further, consistent with previous work,^21,22  CFU-E activity in flow-sorted populations showed a fourfold enrichment in CD34⁺CD71⁺CD41a⁻CD105⁺CD36⁺ vs CD34⁻CD71⁺CD41a⁻CD105⁺CD36⁺ cells (supplemental Figure 1F), suggesting that CFU-E activity is contained in both CD34⁺ and CD34⁻ fractions.

Consistent with EEP and LEP being distinct progenitor populations, we found that ex vivo sorted EEP are small and blast-like with a higher nuclear-cytoplasmic ratio than the larger LEP (Figure 1E), similar to murine fetal EP.¹¹  Furthermore, under liquid culture conditions, EEP had a higher proliferative capacity than LEP (P < .05; Figure 1F) and a higher fraction of erythroid precursors, ie, erythroblasts derived from EEP were in S phase (P < .05; Figure 1G) consistent with LEP arising downstream of EEP in the erythropoietic hierarchy. Whereas flow-purified EEP generated LEP in short-term liquid cultures, LEP were unable to generate EEP (Figure 1H). Additionally, although both populations generated purely erythroblasts (ie, CD34⁻CD71⁺GlyA⁺CD14⁻CD11b⁻CD41a⁻), LEP differentiated more rapidly as shown by a higher fraction of GlyA⁺ (glycophorin A) cells lacking CD105¹⁹  on day 11 (supplemental Figure 1G). Finally, these populations exhibited distinct gene expression patterns: (1) in line with their immunophenotypes (Figure 1B) expression of CD36 progressively increased from EEP to erythroblasts, and (2) consistent with findings in murine²³  and human²⁰  EP, expression of the megakaryocyte-erythroid transcription factor GATA-1²⁴  increased from EEP to erythroblasts, whereas GATA-2 was downregulated at the LEP stage (Figure 1I).

Next, we sought to elucidate the EP defect in DBA BM using our novel EP definitions. The frequency of EEP was reduced in TD DBA BM (P < .001; Figure 2A-B), which was mainly attributable to the decreased frequency of CD71⁺CD41a⁻ cells (Figure 2A). In clonogenic assays, TD DBA EEP generated fewer colonies than controls (P < .05; Figure 2C), consistent with quantitative and qualitative defects in DBA EEP, which were even more pronounced when EEP frequency and their clonogenic output were considered together (supplemental Figure 2A). Single-cell assays confirmed the reduced clonogenicity of TD DBA EEP (P < .01; Figure 2D) and revealed heterogeneity within EEP in individual patients; whereas some EEP formed BFU-E of normal size and morphology, others produced abnormal, poorly hemoglobinized erythroid clusters (supplemental Figure 2B). Similar abnormalities were found in the frequency and clonogenicity of DBA LEP (Figure 2E-G). Consistent with these findings, DBA EEP generated fewer erythroblasts than controls in an erythroid liquid culture system (P < .05; Figure 2H). Notably, in steroid-responsive (SR) patients (supplemental Table 1), both frequency and clonogenicity of EEP were significantly higher than in TD patients (Figure 2A-B,D). Similarly, LEP frequency was higher in SR than in TD patients, although the impact on LEP clonogenicity was variable (Figure 2E,G).

To investigate whether the reduced frequency of EEP and LEP in TD DBA is associated with a reduced frequency of erythroblasts in vivo, we examined this population in fresh BM samples from the same patients studied above. To define erythroblasts, we sorted BM cells into five Lin⁻ populations (supplemental Figure 2C). Consistent with previous immunophenotypic descriptions of erythroblasts,¹⁸  Lin⁻CD34⁻CD36^hiCD71⁺GlyA^−/+ cells (population 3) had negligible colony-forming capacity (supplemental Figure 2C-D) and comprised proerythroblasts to normoblasts with occasional reticulocytes (supplemental Figure 2E). Compared with controls, population 3 was virtually absent in TD DBA (Figure 2I) but rescued in SR patients. Taken together, these data demonstrate that in DBA, as previously shown in normal erythropoiesis,^11,13,25  corticosteroids act predominantly at the level of BFU-E progenitors, in turn increasing downstream EP and precursor frequencies.

In conclusion, we present a novel approach allowing the prospective identification, enumeration, and isolation of human BM EP that give rise directly and selectively to BFU-E (Lin⁻CD34⁺CD38⁺CD45RA⁻CD123⁻CD105⁻CD36⁻ EEP) and CFU-E (Lin⁻CD34^+/−CD38⁺CD45RA⁻CD123⁻CD105⁺CD36⁺ LEP). Moreover, using this approach, we demonstrate that in DBA, cumulative defects in the frequency and function of EEP and LEP culminate in a striking paucity of erythroblasts, a defect that is restored in SR patients.

The authors thank all the patients, donors, and their families for generously donating samples. The authors also thank the hematology laboratories, nursing, and medical staff at St. Mary’s and Hammersmith Hospitals for their valuable contributions.

This work was supported by the Imperial College Healthcare National Health Service Trust, the CHAMPS charity, and the National Institute for Health Research Biomedical Research Centre at Imperial College London. D.I. is a recipient of a Leukaemia and Lymphoma Research Clinical Research Fellowship.

Contribution: D.I., B.P., G.G., A.C., and H.E.F. performed experiments and interpreted data; J.d.l.F., Y.H., and L.C.K. provided BM samples and clinical details; A.K. and I.R. conceived and supervised the work; and A.K., I.R., and D.I. wrote the paper with input from all authors.

Conflict-of-interest disclosure: The authors declare no competing financial interests.

Correspondence: Anastasios Karadimitris, Centre for Haematology, Department of Medicine, Imperial College London, Hammersmith Hospital, Du Cane Road, London W12 0NN, United Kingdom; e-mail: a.karadimitris@imperial.ac.uk.