CD14+ monocytes repress gamma globin expression at early stages of erythropoiesis

In β-hemoglobinopathies, reactivation of gamma- at the expense of beta-globin is a prominent therapeutic option. Expression of the globin genes is not strictly intrinsically regulated during erythropoiesis, supported by the observation that fetal erythroid cells switch to adult hemoglobin expression when injected in mice. We show cultured erythroblasts are a mix of HbA restrictive and HbA/HbF expressing cells and that the proportion of cells in the latter population depends on the starting material. Cultures started from CD34+ cells contain more HbA/HbF expressing cells compared to erythroblasts cultured from total peripheral blood mononuclear cells (PBMC). Depletion of CD14+ cells from PBMC resulted in higher HbF/HbA percentages. Conversely, CD34+ co-culture with CD14+ cells reduced the HbF/HbA population through cell–cell proximity, indicating that CD14+ actively repressed HbF expression in adult erythroid cultures. RNA-sequencing showed that HbA and HbA/HbF populations contain a limited number of differentially expressed genes, aside from HBG1/2. Co-culture of CD14+ cells with sorted uncommitted hematopoietic progenitors and CD34-CD36+ erythroblasts showed that hematopoietic progenitors prior to the hemoglobinized erythroid stages are more readily influenced by CD14+ cells to downregulate expression of HBG1/2, suggesting temporal regulation of these genes. This possibly provides a novel therapeutic avenue to develop β-hemoglobinopathies treatments.

www.nature.com/scientificreports/ in PBMCs and by differentiated macrophages originating from PBMC-monocytes. We used this characteristic of the culture system to study regulation of globin subunits by extrinsic factors and demonstrated that PBMC derived erythroid cells have lower HbF compared to CD34+ derived erythroid cells. This is reflected by a smaller fraction of cells expressing gamma globins. The effect depended on cell-proximity between HSPC and the CD14+ fraction from PBMCs, but was absent upon co-culture of already committed erythroblasts with CD14+ cells. In addition, RNA sequencing showed limited differences between cells with and without HbF expression. These data suggest that CD14+ monocytes/macrophages can signal to hematopoietic progenitor cells priming globin expression to HBB1 at the expense of HBG1/2.

Methods
Cell sorting. CD14 and CD34 MicroBeads (Miltenyi Biotec; Bergisch Gladbach, Germany) were used for magnetic-activated cell sorting (MACS) of PBMC fractions (manufacturer protocol). Flowcytometric cell sorting was used to sort populations of progenitors based on CD34 and CD36. Sorting after intracellular staining for globin was performed on FACS Aria 4L (BD Biosciences; Oxford, UK).
Cell cultures. Peripheral blood mononuclear cells (PBMCs) were obtained according to the Declaration of Helsinki (seventh revision, 2013). Written informed consent to extract and use human material (peripheral blood) as part of regular donations was obtained with approval of the local medical ethics committee (MEC, Sanquin). Experimental protocols were approved by Sanquin (MEC). Human PBMC from whole blood were purified by density separation using Ficoll-Paque (manufacturer protocol). CD34+ cells were isolated from PBMCs by MACS. Both were cultured in a three-phase liquid erythroid culture system as previously described 6 .
Low iron experiments contained 100 mg/ml holotransferrin one third of the concentration normally used throughout culture.
Co-culture. CD34 and CD14 fractions were put in co-culture at the ratios indicated in the legends, (ratio in blood is ~ 1:100 CD34:CD14). Trans-well co-cultures were performed by seeding CD14+ cells at the bottom and CD34+ in the trans-well (0.4 µm polyester membrane, Corning; NY, USA), cultured for 8 days and differentiated for two days to allow hemoglobinization as described before 6 . For co-cultures with sorted CD34CD36 fractions, freshly isolated CD34+ cells were cultured for three days, sorted and co-cultured for an additional 5 days.
Flowcytometry. Cells  of the total hemoglobin content in erythrocytes. However, erythrocytes of individuals with hereditary persistence of fetal hemoglobin (HPFH), and also in vitro cultured adult erythroid cells, may express HbF at levels higher than 5% as determined by HPLC 6 . HPLC measures bulk hemoglobin and cannot discriminate between increased numbers of cells expressing gamma globin or increased gamma expression in the fraction of cells that already express HbF. To this end, we analyzed hemoglobin expression in ex-vivo and cultured erythroid cells by flow cytometry. Upon intracellular staining for beta globin and gamma globin, components of HbA and HbF respectively, adult erythrocytes showed mostly HbA, with a fraction of cells positive for HbF (Fig. 1A). Erythroblasts from human fetal liver predominantly showed HbF expression, indicating that the assay is specific for the two globin types (Fig. 1B). In contrast to adult and fetal ex-vivo erythroid cells, cultured adult erythroid cells showed two distinct populations; one expressing HbA and one expressing both HbF and HbA (Fig. 1C).
No cells expressing only HbF are found. Of note, erythroid cells cultured from cord blood are significantly enriched in HbF/HbA expressing cells, reflecting the ongoing globin switching occurring at this developmental stage (supplemental Fig. 1A). The increase in HbF expression in adult cultures was dependent on the starting source for erythroid cultures. Erythroid cultures initiated from total adult peripheral blood mononuclear cells (PBMC) showed a significantly lower HbF expression and higher HbA expression compared to cultures that were started from purified CD34+ cells from PBMC (Fig. 1C, D). The increase in HbF and the linked decrease in HbA suggests an active regulation of globin chain expression. HPLC measurements confirmed that cultures from PBMC contained 1-3% HbF compared to > 5% HbF in cultures started from selected CD34+ cells (supplemental www.nature.com/scientificreports/ the two populations (HbA or HbA/HbF) can arise from one single CD34+ cell or whether the ability to express HbF is segregated at the progenitor level. Single cell sorted adult CD34+ cells were differentiated to erythroid cells. The single cell cultures still showed two populations, one expressing high HbF and one expressing low to no HbF ruling against the presence of two different parental hematopoietic progenitors. Cultures from different clones showed between 56 and 63% of cells in the HbF high population (Fig. 1E), which was significantly higher compared to non-single cell derived CD34+ cultures. The presence of two populations in all clones shows that erythroid cells derived from a single CD34+ cells can express different hemoglobin expression profiles. The data also indicated that in a selection of cells the HBG1/2 remains silent. Of note, to obtain enough cells to perform flow cytometry for HbA and HbF the CD34+ single cells were cultured for a minimum of 25 days.

CD14+ monocyte/macrophages from PBMC reduce HbF expression in erythroid cultures.
Erythroid differentiation in the bone marrow occurs on erythroblast islands, structures of erythroid progenitors surrounding a central macrophage 8,9 . Defects in the interaction between macrophages and erythroid cells have been reported to result in erythropoiesis defects 14,19 . PBMC cultures contain significant amounts of cells capable of providing support and/or interactions with CD34+ , erythroid progenitors and erythroblasts. In fact, 14% of PBMCs are CD14+ monocytes that can differentiate to erythropoiesis supporting macrophages in our culture medium 15,20 . CD14+-derived macrophages from PBMC interact with, and support HSPC 15,21,22 . To test if a similar interaction was responsible for the repression of HbF, PBMCs were depleted for CD14+ cells.
Differentiating erythroblast from CD14+ depleted PBMCs showed a significantly increased frequency of HbA/ HbF expressing cells and a higher HbF mean fluorescent intensity (MFI) compared to erythroid cultures derived from total PBMCs ( Fig. 2A-C). The increased fetal hemoglobin expression did not result from overall lower hemoglobinization (supplemental Fig. 2). Of note, cultures that received suboptimal levels of holo-transferrin did show reduced hemoglobinization (supplemental Fig. 2). Next, we asked if CD14+ alone would be sufficient www.nature.com/scientificreports/ to reduce HbF in CD34+ cultures. The CD34+ fraction and the CD14+ fraction were isolated from buffy coats and co-cultured. Co-culture of CD34+ cells with CD14+ cells lead to a decrease in HbF expression compared to CD34+ alone ( Fig. 2D and supplemental Fig. 3A), which occurred in a dose dependent manner (Fig. 2E, supplemental Fig. 3B). To test if reduced HbF expression required direct contact or resulted from a secreted factor, CD34+ were differentiated to erythroid cells in a transwell co-culture setup. Although co-culture in direct contact with CD14+ positive cells again reduced HbF expression in erythroid cultures, erythroblasts co-cultured with CD14+ cells in a transwell setting did not reduce HbF expression (Fig. 2F, G). This suggests that HbF repression in culture requires cell-cell proximity of CD14+ and erythroid progenitor cells.

The transcriptomes of HbA/HbF co-expressing compared to HbA only expressing cells is highly similar. The presence of HbF/HbA co-expressing cells and HbA single expressing cells in culture, combined
with the observation that the balance between these populations can be regulated through specific interactions with monocyte/macrophage populations suggests that potential globin regulators may be differentially regulated upon comparing these two erythroid populations. To find such potential targets the transcriptome of HbF/HbA co-expressing and HbA expressing populations cultured from adult PBMCs was assessed by RNA sequencing of the erythroblast stage (CD71+/CD235+; supplemental Fig. 4A). RT-PCR on the RNA isolated from these population indicated beta globin expression in HbA or HbF/HbA cells, however beta-globin levels were ~ fourfold lower in HbF/HbA cells (Fig. 3A). As expected, gamma-globin levels were nihil in HbA only and high in HbF/ HbF cells. Principle component analysis of the corresponding, normalized transcriptome data revealed that the first two components (45% of the variation) separated samples by donor, while the third component (15% of  www.nature.com/scientificreports/ the variation), separated the HbF/HbA sorted population from the HbA population ( Fig. 3B. Interestingly, out of over 16,000 distinct RNA species detected only a surprisingly small number of 25 genes (8 up and 17 down in HbF/HbA vs HbA expressing cells) were differentially expressed with at least a two-fold difference between the two sorted populations (false discovery rate (FDR) < 0.01; fold change: − 2 < FC > 2; Log count per million CPM > 2; Fig. 3B-D; Table 1). HBG1 and HBG2, were the most differentially expressed between the two populations conforming the sorting results by HPLC and flow cytometry in Fig. 1 (supplemental Fig. 4A). Among the 8 RNAs upregulated in HbF/HbA cells, 5 were located within the beta globin locus. Besides HBG1/2, these were embryonic epsilon globin chain (HBE1), and the non-coding RNAs BGLT3 and HBBP1, both non-coding RNAs were significantly higher expressed. In addition, two mitochondrial RNAs and one transmembrane protein TMEM233 were upregulated (Table 1). GO analysis indicates oxygen carrying/binding (upregulated RNA's) and transforming growth factor binding (down regulated RNA's) as enriched GO terms (data not shown).
Among the genes that are expressed at lower levels in HbF/HbA cells compared to HbA only cells are Endoglin (ENG), Thrombosondin (THBS1), Pleckstrin (PLEK), Plexin D1 (PLXND1), Rap1-interacting adapter molecule (RIAM/ APBB1IP) and AHNAK, encoding a large intracellular scaffold protein, which may all alter response to environmental factors for instance signal transduction by TGFβ (transforming growth factor beta) family members. Similarly, reduced expression of the sphingosine-1-phosphate receptor 4 (S1PR4) may alter response to the environment. Several transcriptional regulators are known that control or influence the expression of globin subunits. However, known globin regulators, BCL11A, KLF1, ZBTB7A and SOX6 were not found to be differentially expressed (Supplemental Fig. 4C). Adding to this, we do not observe a reduction in KLF1 target genes (e.g. CD44, BCAM, CARM1, BCL11A; Supplemental Fig. 4C). Moreover, among the deregulated RNA's no other transcriptional regulator was identified.

Repression of HbF by CD14+ cells occurs before erythroblasts stage.
As the number of differentially expressed genes is surprisingly low between HbA/HbF and HbA sorted populations and cannot be attributed to RNA expression levels of known or novel transcriptional regulators, we hypothesized that posttranscriptional control mechanisms regulate the frequency of HbF/HbA cells, and/or (ii) globin subunit regulation occurred before the hemoglobinized CD71+/CD235+ erythroblast stage analyzed and that the molecular mechanism driving this differential globin expression may already be downregulated at the hemoglobinized erythroblast stages. The alternative explanation would fit with high HbF in the cultures from single cell CD34+ in Fig. 1E, if this were to result from prolonged expansion of the earliest progenitor stage. Indeed, CD34+ hematopoietic stem and progenitor cells (HSPC) are a mix of hematopoietic cells at different stages of differentiation from hematopoietic stem cells to more committed lineage progenitors 15 . Lineage specification progresses from CD34+ CD36-hematopoietic stem and progenitor cells to CD34+ CD36+ megakaryoid/erythroid common progenitors to CD34-CD36+ erythroblasts 15,23 . Co-cultures with these fractions and the CD14+ cells were assessed to determine the stage at which the repression occurred. Co-culture of CD34-CD36+ erythroblasts with CD14+ monocytes did not decrease HBG1/2 expression. This supports the RNA-sequencing data results that the regulation of HBG1/2 in erythroid cultures by CD14+ cells does not occur at the committed lineage restricted (pro)erythroblast stage (Fig. 4A). In contrast, CD14+ cells were able to repress HBG1/2 expression in CD34+CD36−/+ progenitors with the biggest repression occurring in the early CD34+CD36− co-cultures with CD14+ cells (Fig. 4B). Of note, repression of HbF lead to a concomitant increase in HbA. The data indicate that regulation of globin genes by CD14+ cells can occur early during hematopoiesis before the commitment to erythroid restricted progenitors.

Discussion
Hematopoietic stem cell differentiation to erythroid cells occurs within the bone marrow. This process depends on interactions with specific cells present within the different bone marrow niches. Distinctive supporting macrophages can be identified both in the hematopoietic stem and progenitor cell niche as well as within the erythroid niche, where they can provide signals to the differentiating hematopoietic cells 8,9,[24][25][26] . Reminiscent to these niche functions, PBMC-isolated CD14+ monocyte-derived macrophages were found to both support survival of HSPC and recapitulate central macrophage functions 15,20,21 . Here we find that the CD14+ monocytederived macrophages alter the balance between gamma and beta globin gene expression in erythroid cultures. Specifically, we find that gamma-globin down regulation occurs prior to the erythropoietin, stem cell factor and glucocorticoid responsive non-hemoglobinized committed erythroblast stage, defined as CD71 high /CD235 low/+6 . In fact, the highest repression was observed at the earliest CD34+CD36− hematopoietic stem and progenitor population, suggesting that erythroid specific factors like KLF1 are not involved in this regulation. Indeed, KLF1 was not deregulated between hemoglobinized erythroblasts expressing HbA-only or HbA/HbF. Moreover, the regular suspects of globin regulation, e.g. SOX6, BCL11A and LRF were not differentially expressed at the erythroid committed erythroblast stage. However, as some of these regulators (e.g. BCL11A) are expressed and functional at the HSPC stage or even in HSCs [27][28][29] , the data does not rule out a role for these factors to regulate HbA/HbF expression in early hematopoietic progenitors. The data suggests a novel concept that globin expression induced later during erythroid differentiation can be regulated at these early HSPC progenitor stages. This would fit with the proposed flexibility in globin-chain expression observed by Kobari et al., who have reported that SCD-iPSC derived erythroid progenitor cells that in vitro express fetal hemoglobin (HbF) will convert to adult beta globin (HbA) expression containing the mutated adult beta sickle cell HBS peak in HPLC upon injecting in vivo, revert to producing beta-globin expressing erythrocytes upon injection into NOD-SCID mice 7 . In addition, Singh et al. have shown that the gamma-globin promoter was highly methylated in the earliest stage of hematopoietic stem progenitor cells (CD34( +)CD36(−)) and that methylation progressively decreased as HSPC differentiation progressed in sorted adult bone marrow progenitors to committed erythroblasts 30  www.nature.com/scientificreports/ did not look at the methylation status of the promoters in our erythroid cultures, combined with the flexibility in globin chain expression it substantiates a hypothesis where the niche can instruct erythropoiesis to express specific globin subunits in a process that involves epigenetic modifications at the earliest differentiation stages of erythropoiesis. In line with the role of a niche cell, the CD14+-derived macrophages from PBMCs regulate globin expression through cell-cell contact. We and others have previously shown that these PBMC-CD14derived CD163/CD169+ macrophages resemble and function like erythroid island central macrophages 20,21 . We propose that also macrophages within the bone marrow niche may similarly exert signaling to regulate globin expression. We acknowledge that the ratio between the macrophages and the erythroid precursors is off and in favor of macrophages, whereas in the niche one macrophage is surrounded by > 10 erythroid cells. However, as a clear dose response curve of macrophages on the repression of HbF is found, this clearly indicates that signaling influences specific globin subunit expression. Importantly, here we use a macrophage co-culture system with erythroid precursors recapitulating only one component of the hematopoietic/erythroid niche in vitro and hence a simplification of processes occurring within the bone marrow. Despite these considerations, the data indicate that hematopoietic progenitors can be influenced by signaling to control globin expression and provides a possible mechanism through which the bone marrow niche specifically produces beta globin expressing erythroid cells. The precise identity of these signals and the induced signal transduction in erythroid progenitors remains unknown but are important to elucidate. Mapping the signal transduction induced by macrophages on erythroid/hematopoietic progenitors may link to the well-known regulators of the globin genes or may uncover novel pathways. These pathways could potentially be blocked to increase gamma expression providing novel therapeutic options to increase HbF in sickle cell or beta-thalassemia either on its own or in combination with other (HbF-inducing) therapy.