Megakaryocyte TGFβ1 partitions erythropoiesis into immature progenitor/stem cells and maturing precursors

It is estimated that one-third of the world’s population suffers from anemia.^1,2  Daily production of >200 billion erythrocytes is required to keep up with routine losses, so even minor disequilibrium between blood loss and erythrocyte production can lead to anemia.³  Although many transient anemias are easily treated, therapy for chronic anemias is limited. Red blood cell (RBC) transfusions and erythropoiesis-stimulating agents are not always effective yet are linked to significant expense, inconvenience, potential toxicity, and generally transient utility.^4,5  Despite these limitations, development of new approaches to treat chronic anemia has been limited by our incomplete understanding of erythropoiesis.

A great deal of what we know about erythropoiesis relates to the terminal maturation of erythroid precursors (EPs) leading to hemoglobinization, enucleation, and release of immature RBCs into the circulation.^6,7  In the setting of anemia, or otherwise reduced oxygen delivery, juxtaglomerular renal erythropoietin (EPO)-producing cells can increase EPO secretion many thousands-fold, thereby serving as a feedback signal reporting erythrocyte need to the marrow.^8-13  Nonetheless, EPO acts during a very narrow window of erythropoiesis, well after progenitor commitment to an exclusively erythroid fate.^6,8,9  It is not known whether these final steps of RBC maturation are coupled to the earlier stages of hematopoietic stem and progenitor cell (HSPC) differentiation, a process that begins almost 3 weeks earlier when a hematopoietic stem cell (HSC) starts its march toward committed RBC precursors via a series of branching cell-fate decisions.^7,14,15  In this study, we investigated whether terminal erythropoiesis is linked to the availability of EPO-insensitive HSPCs.

Transgenic C57BL/6 mice with Cre recombinase driven from the regulatory elements of Cxcl4 (platelet factor 4 [Pf4]), C57BL/6-Tg(Pf4-cre)Q3Rsko/J (Pf4-cre),¹⁶  were crossed with mice harboring a floxed transforming growth factor β1 (TGFβ1) locus¹⁷  (Tgfb1^tm2.1Doe, TGFβ1^FL/FL mice) to generate Pf4-Cre/TGFβ1^FL/FL mice with selective deletion of TGFβ1 in megakaryocytes (Mks) and platelets (TGFβ1^ΔMk/ΔMk mice). GFAP-CreERT2 (B6;Cg-Tg(GFAP-cre/ERT2)505Fmv/J) mice with the human glial fibrillary acidic protein (GFAP) promoter sequence directing expression of a Cre-estrogen receptor (Cre-ERT2) fusion protein were bred into the TGFβ1^FL/FL strain to generate GFAP-CreERT2/TGFβ1^FL/FL mice (TGFβ1^{iSchwn/iSchwn}). Deletion of Tgfb1 in Schwann cells (TGFβ1^{ΔSchwn/ΔSchwn}) was induced at 5 weeks of age via intraperitoneal injection of tamoxifen (80 mg/kg; Millipore/Sigma-Aldrich) daily × 5 days followed by 4 weeks of tamoxifen chow (Envigo). TGFβ1^{ΔSchwn/ΔSchwn} mice were rested for 1 week prior to analysis. Selective Cre activity was confirmed by immunofluorescence and/or flow cytometry by crossing the Cre-deleter strains (PF4-Cre or GFAP-CreERT2) into the Cre-reporter strain Rosa26-FSF-ZsGreen (B6;Cg-Gt(ROSA)26Sor^{tm6(CAG-ZsGreen1)Hze}/J).

Animals were maintained in the Weill Cornell Medicine Animal Facility. All protocols were approved by the Weill Cornell Medicine Institutional Animal Care and Use Committee.

Mice were euthanized by CO₂ asphyxiation; bones (femur, tibia, plus or minus humeri) were dissected free of muscle and tendons and crushed in Dulbecco modified Eagle medium using a mortar and pestle. The resulting cell suspension was filtered through a 40-µm mesh and washed in phosphate-buffered saline (PBS) EDTA buffer (2 mM EDTA, 0.2% bovine serum albumin in PBS, pH 7.4). Spleens were isolated, and then minced before grinding through a 40-µm mesh to generate single-cell suspensions. Lineage-negative (Lin⁻) cells were isolated using immunomagnetic beads using a biotinylated lineage cell depletion cocktail (Miltenyi Biotec).

Lin⁻ cells were then stained with peridinin-chlorophyll protein/Cy5.5-CD117 (BD), phycoerythrin (PE)/Cy7-Sca-1 (BioLegend), Alexa Fluor 700–CD48 (BioLegend) and PE-CD150 (BioLegend), allophycocyanin (APC)/Cy7-CD16/32 (FcR) (BioLegend), APC-CD105 (BioLegend). HSPC populations were identified as follows: long-term HSCs, Lin⁻Kit⁺Sca1⁺ (LKS⁺) CD48⁻CD150⁺; multipotent progenitors (MPPs), LKS⁺CD48⁺CD150⁻; granulocyte macrophage progenitors (GMPs), Lin⁻Kit⁺Sca1⁻ (LKS⁻) CD150⁻FcR⁻; pre-GMP, LKS⁻CD150⁺CD105⁻; pre-Mk erythroid progenitors (MEPs), LKS⁻CD150⁺FcR⁻CD105⁻; pre-colony-forming unit (pre-CFU) erythroid, LKS⁻CD150⁺FcR⁻CD105⁺; erythroid progenitors, LKS⁻CD150⁻FcR⁻CD105⁺.¹⁸ 

To analyze cell cycle or pSmad2/3 staining in HSPCs, lineage-depleted cells were stained with streptavidin (BioLegend), APC-CD117 (BD), and PE/Cy7-Sca-1 (BioLegend), Alexa Fluor 700–CD48 (BioLegend), and PE or BV605-CD150 (BioLegend) and then fixed and permeabilized using CytoFix/Perm (BD). For cell cycle, surface stained, fixed, permeabilized cells were stained with peridinin-chlorophyll protein/Cy5.5-Ki67 (BD) and Hoechst 33342 (Invitrogen). To analyze the pSmad2/3 signal, fixed, permeabilized cells were stained first with pSmad2/3 primary antibody (Cell Signaling Technology) and then with secondary anti-rabbit, antibody Cy3 (The Jackson Laboratory).

Specificity of pSmad2/3 staining was tested by treating C57BL/6J mice with either the TGFβ-neutralizing antibody 1D11 (10 mg/kg) or a nontargeting control antibody 13C4 (10 mg/kg) on days 5, 10, and 15. Mice were euthanized on day 16 for analysis. Smad2/3 phosphorylation was assessed by intracellular flow cytometry as described previously.

Analysis gates were set based upon the fluorescence minus 1 (FMO) fluorophore. Multidimensional fluorescence-activated cell sorter analysis was performed using a BD LSRII equipped with 5 lasers. FlowJo (TreeStar) and FCS Express (De Novo) were used to analyze flow cytometry data.

Mice were euthanized by CO₂ asphyxiation and bones or spleen were processed as above except lineage depletion was not performed. Cell suspensions were stained with APC-conjugated Ter119, APC/Cy7-conjugated CD44, Hoechst and, in the indicated experiments, cells were stained with a modified biotinylated lineage cocktail (CD3, B220, CD11b and Gr1) and streptavidin. In the indicated experiments, cells were also stained with PE/Cy7-Kit (Abcam) and EPO receptor (Epor) primary antibody (R&D Systems) and then stained with secondary, anti-goat Cy3 antibody (The Jackson Laboratory).

Primitive myeloid progenitors were enumerated using the spleen CFUs (CFU-S) after 12 days (CFU-S₁₂) assay. Recipient mice were sublethally irradiated (7 Gy) using a ¹³⁷Cs–γ-ray source and injected with 10⁵ bone marrow (BM) cells. Spleens were isolated 12 days after transplantation and fixed in Bouin solution. The number of macroscopic spleen colonies was counted and expressed as the number of CFU-S colonies per 10⁵ donor cells. Clonogenic myeloid progenitors were assessed by standard methylcellulose colony-forming cell (CFC) assays (MethoCult GF M3434; Stem Cell Technologies) using 1.5 × 10⁴ BM mononuclear cells per well (6-well plate). Colonies were scored after 7 days of incubation and expressed as the number of macroscopic spleen colonies after 12 days (CFU-S₁₂) per 1.5 × 10⁴ BM mononuclear cells before normalizing to total leg (femur/tibia) cell counts.

Recipient mice were irradiated with 9 Gy using a ¹³⁷Cs–γ-ray source. Three to 4 hours after irradiation, whole BM cells (marrow) from donor animals were infused via the tail vein of recipient animals. Competitive repopulation was used to assess the relative number of HSCs in marrow from TGFβ1^ΔMk/ΔMk and TGFβ1^FL/FL mice. TGFβ1^ΔMk/ΔMk or TGFβ1^FL/FL marrow cells were mixed with an equal number of marrow cells from congenic CD45.1^+/+ control mice. The 1:1 (test/competitor) mixture (2 × 106 cells total) was injected into lethally irradiated mice via tail vein. Peripheral blood was analyzed for lymphoid (CD3/CD19) and myeloid (Gr1/CD11b) engraftment by flow cytometry at various times after transplantation. Animals were euthanized after 16 weeks and marrow harvested for immunophenotyping of HSPC populations and for serial transplantation. Serial transplantation of secondary recipients was performed as for primary recipients except for the marrow donor cells, which were not mixed with additional congenic competitor cells.

To assess blood count recovery, we collected peripheral blood (50 µL) into EDTA-coated capillary tubes (Thermo Fisher Scientific). Differential blood counts were measured using an automated ADVIA 120 Multispecies Hematology Analyzer (Bayer HealthCare) calibrated for murine blood.

Apoptosis in mononuclear cell suspensions was assessed using the Annexin V-FITC Apoptosis Kit (BioLegend) according to manufacturer’s instructions and analyzed via flow cytometry. In some experiments, apoptosis was assessed using CellEvent Caspase-3/7 Green Detection Reagent by incubating marrow cells for 10 minutes prior to analysis by flow cytometry. Cleaved Caspase 3 was also assessed by intracellular staining of surface-stained mononuclear cell suspensions after permeabilization using CytoFix/Perm (BD) and subsequent staining with cleaved Caspase 3 primary antibody (Cell Signaling Technology) and anti-rabbit secondary antibody Cy3 (The Jackson Laboratory).

Femurs were fixed in 4% paraformaldehyde overnight, and then decalcified using 10% EDTA before freezing at optimal cutting temperature (Sakura Finetek). Immunofluorescence staining was performed on frozen sections. After blocking with blocking buffer (3% bovine serum albumin, 3% serum, and 0.03% Tween-20), sections were incubated in primary antibodies overnight at 4°C using anti-cleaved Caspase 3 (Cell Signaling Technology) antibody. After washing, sections were stained with secondary antibody (anti-rabbit) CY3 (The Jackson Laboratory). The specificity of staining was confirmed in sequential sections using the secondary antibody alone. Images were acquired using a Zeiss spinning disk confocal microscope or a Zeiss 710 laser-scanning confocal microscope.

Human recombinant EPO (300 U/kg body weight, ∼9 U per mouse) was injected subcutaneously for 5 consecutive days. On day 6, blood was sampled and mice were euthanized for analysis.

To induce hemolysis, mice were injected intraperitoneally with phenylhydrazine hydrochloride at 60 mg/kg (Sigma-Aldrich) on days 1 and 2. Blood was sampled on days 3, 5, 8, and 14 from staggered cohorts of mice.

To assess the ability of exogenous TGFβ1 to rescue the megakaryocytic TGFβ1 knockout, mice were treated with TGFβ1 (5 μg/kg per day) subcutaneously for 5 consecutive days as indicated. On day 6, blood was sampled and mice were euthanized for analysis.

Mononuclear cell suspensions were washed in PBS, pelleted, and lysed in lysis buffer (50 mM Tris-HCl [pH 7.5], 120 mM NaCl, and 0.4% NP-40 supplemented with proteinase inhibitors; samples were transferred to polyvinylidene difluoride [PVDF] membranes [EMD Millipore]) and blocked with 5% nonfat dried milk in PBS with 0.1% Tween 20. Primary and secondary antibodies were diluted in blocking solution. Primary antibodies against cleaved Caspase 3 (Cell Signaling Technology) were used. Secondary peroxidase-conjugated anti-rabbit antibody (EMD Millipore) was used before chemiluminescent visualization using the SuperSignal West Femto Substrate (Thermo Fisher Scientific).

All data are expressed as mean plus or minus standard deviation (SD). The Student t test (2-tailed) was used to analyze the statistical differences between groups, with the P values indicated in the plots (*P < .05, **P < .01, ***P < .001, or if not shown, the comparison was not significant).

Many HSCs reside in close proximity to Mks where they can be maintained in a quiescent state by niche signals such as TGFβ1 and CXCL4.^19,20  Although genetic ablation of Mks using an inducible diphtheria toxin receptor under control of Mk-specific Cxcl4-cre driver releases HSCs from quiescence and licenses expansion of the HSC pool, it is not known how this affects blood cell production or other aspects of hematopoiesis. We selectively deleted Tgfb1 (TGFβ1) in Mks (TGFβ1^ΔMk/ΔMk) and found that peripheral blood counts were entirely normal in TGFβ1^ΔMk/ΔMk mice compared with TGFβ1^FL/FL littermate controls despite the pool of primitive hematopoietic cells being expanded (Figure 1A). Sequestration of maturing cells within hematopoietic tissues could not explain this discrepancy because total BM and spleen cellularity were normal in TGFβ1^ΔMk/ΔMk mice (Figure 1B). Excess HSCs in TGFβ1^ΔMk/ΔMk mice appeared capable of robust differentiation because the number of immature Lin⁻ hematopoietic progenitor cells (HPCs) was increased in the marrows of TGFβ1^ΔMk/ΔMk mice (Figure 1C). Thus, it remained unexplained why the expanded number of HSCs and HPCs did not increase blood counts and marrow cellularity.

Hematopoietic cell population size is determined by the balance of cell gain (proliferation/self-renewal and differentiation) and cell loss (apoptosis and differentiation). Many late-acting hematopoietic cytokines are not required for lineage commitment yet provide essential proliferation, differentiation, and survival signals during maturation of hematopoietic precursors.^6,21  Thus, we hypothesized that the excess progenitors observed in the TGFβ1^ΔMk/ΔMk mice failed to increase blood counts because their progeny were unneeded, and inadequately supported by homeostatic levels of late-acting cytokines (Figure 1D). Indeed, BM apoptosis was markedly increased in the TGFβ1^ΔMk/ΔMk mice compared with controls, as reported by cleaved Caspase 3 (Figure 1E-G; supplemental Figure 1a-b, available on the Blood Web site) or annexin V detection (Figure 1H-K). The excess apoptotic cells seemed largely restricted to hematopoietic cells expressing the lineage markers (Lin) Gr1, CD11b, CD3, B220, or Ter119 (Figure 1L-M). Annexin V staining of LKS⁻ HPCs and LKS⁺ HSPCs was rare in both TGFβ1^ΔMk/ΔMk mice and littermate controls (Figure 1N-O). These results support an interpretation that excess, unneeded hematopoietic precursors are pruned by apoptosis during hematopoietic differentiation.

To rule out the possibility that progeny of HSPCs from TGFβ1^ΔMk/ΔMk mice are intrinsically defective, we performed competitive repopulation assays by cotransplanting BM from TGFβ1^ΔMk/ΔMk mice or TGFβ1^FL/FL littermates competitively with CD45.1 congenic donor cells. As expected, TGFβ1^FL/FLdonor cells engrafted and contributed to multilineage hematopoiesis indistinguishably from the congenic CD45.1 controls (Figure 2A; supplemental Figure 1d-f). In contrast, TGFβ1^ΔMk/ΔMk donor cells outcompeted control cells, confirming that TGFβ1^ΔMk/ΔMk donors are enriched for functional HSCs and indicating that these TGFβ1^ΔMk/ΔMk HSCs yield progeny fully capable of reconstituting hematopoiesis after transplant. Serial transplantation of recipients demonstrated HSC self-renewal and contribution to hematopoiesis was intrinsically normal in the TGFβ1^ΔMk/ΔMk donor cells (Figure 2B; supplemental Figure 1g-h).

Ex vivo HPC function appeared normal in TGFβ1^ΔMk/ΔMk mice, suggesting that these cells were not intrinsically defective in TGFβ1^ΔMk/ΔMk mice (Figure 2C). The number of GMPs, burst-forming unit erythroid progenitors (BFU-E) and CFU granulocyte, erythroid, monocyte, Mk progenitors (CFU-GEMM) were all increased in the marrow of TGFβ1^ΔMk/ΔMk mice as reported by CFC assays. Similarly, the CFU-S₁₂ assay showed a greater number of functionally normal MPPs in TGFβ1^ΔMk/ΔMk donor marrow compared with TGFβ1^FL/FL littermate controls (Figure 2D). Quantified HSPC immunophenotypes correlated well with functional measures, again demonstrating that the number of HSPCs is increased in TGFβ1^ΔMk/ΔMk mice (Figure 2E-G). The larger number of HSPCs and their more proliferative character in TGFβ1^ΔMk/ΔMk mice (Figure 2H-I) indicate that megakaryocytic TGFβ1 normally curbs the HSPC pool size. Yet, progeny of these excess HSPCs appear to undergo apoptosis rather than contribute to BM cellularity and mature blood cell counts.

The islands of apoptotic cells observed in BM of TGFβ1^ΔMk/ΔMk mice (Figure 1E) are reminiscent of erythroid islands, suggesting that excess EPs may be contributing to the formation of these clusters. Indeed, the proportion of annexin V–staining apoptotic Ter119⁺ EP cells (EPs) was approximately sixfold higher in the marrow and spleen of TGFβ1^ΔMk/ΔMk mice compared with TGFβ1^FL/FL littermate controls (Figure 3A-B). We quantified apoptosis within well-defined populations of maturing erythroid precursors^22-24  (Figure 3C) using annexin V. Although the number of immature EPs, such as pro/basophilic (EI), polychromatophilic (EII), and orthochromatophilic erythroblasts (EIII), was increased in TGFβ1^ΔMk/ΔMk mice, the excess seemed largely composed of cells staining for the apoptotic marker annexin V (Figure 3D). In contrast, the number of mature reticulocytes and erythrocytes (EIV/EV) in TGFβ1^ΔMk/ΔMk mice was indistinguishable from TGFβ1^FL/FL. Thus, the excess erythroid committed precursors are culled during early maturation, thereby producing normal numbers of erythrocytes and normal RBC counts.

Apoptotic Ter119⁺ EPs within TGFβ1^ΔMk/ΔMk marrow predominantly expressed the Epor but not Kit (Figure 3E). EPO provides a survival signal to Epor⁺ EPs, allowing them to escape apoptosis and continue differentiation.^8-10  Excess apoptosis of EPs in TGFβ1^ΔMk/ΔMk mice was not due to subnormal plasma EPO levels (Figure 3F). We reasoned that the surplus, unneeded Epor⁺ cells may not be supported by physiologic EPO levels. To test this, we treated mice with exogenous EPO (300 U/kg) (Figure 3G). Strikingly, exogenous Epo rescued the excess apoptosis of EPs in TGFβ1^ΔMk/ΔMk marrow and spleen (Figure 3H-K). The erythropoietic response of TGFβ1^ΔMk/ΔMk mice to Epo was much more robust compared with TGFβ1^FL/FL littermate controls (Figure 3L-M), suggesting that the rescued apoptosis resulted in increased erythropoietic output. Using phenylhydrazine hydrochloride–induced hemolysis as an anemia model, we found that the erythropoietic response of TGFβ1^ΔMk/ΔMk mice was much brisker and more robust than that of TGFβ1^FL/FL littermate controls (supplemental Figure 2). These results demonstrate that, although genetic deletion of TGFβ1 in MKs expands the number of committed erythroid progenitors, the resultant glut of their EP progeny requires nonhomeostatic Epo levels to promote survival, expansion, and maturation to erythrocytes.

Megakaryocytic TGFβ1 is necessary for robust phosphorylation of receptor-activated Smad2 and Smad3 (Smad2/3) in HSPCs, as assessed by flow cytometry with intracellular staining (Figure 4A). In contrast, we detected little phospho-Smad2/3 (pSmad2/3) in EPs in either the TGFβ1^ΔMk/ΔMk mice or TGFβ1^FL/FL littermate controls (Figure 4A). HSPCs in TGFβ1^ΔMk/ΔMk mice were capable of normal TGFβ1 signaling because exogenous TGFβ1 induced strong phosphorylation of Smad2/3, restoring the mean fluorescence intensity (MFI) to the level found in TGFβ1^FL/FL controls (Figure 4B-C). Genetic deletion of Tgfb1 in Schwann cells had no identifiable effect on BM cellularity, HSPC numbers, or HSC cell cycle (supplemental Figure 3). Thus, Mks serve as the major source of TGFβ triggering Smad2/3 phosphorylation in HSPCs.

It is possible that TGFβ1 sensitizes EPs to EPO, and, if true, exogenous TGFβ1 should rescue EP dropout in TGFβ1^ΔMk/ΔMk mice. To test this, we treated the mice with TGFβ1 for 5 days and assessed erythroid response (Figure 4B). Although exogenous TGFβ1 reestablished pSmad2/3 in TGFβ1^ΔMk/ΔMk HSPCs (Figure 4C), it did not trigger an erythropoietic response. Rather, exogenous TGFβ1 induced mild anemia in the TGFβ1^ΔMk/ΔMk mice (Figure 4D) coupled with worsened apoptosis in the marrow and spleen (Figure 4E-G). In contrast, exogenous TGFβ1 did not induce detectable changes in RBCs or apoptosis in TGFβ1^FL/FL controls (Figure 4E-G). Although these studies cannot rule out functional interactions between TGFβ1 and EPO signaling during terminal erythropoiesis, they indicate that the excess erythroid apoptosis found in TGFβ1^ΔMk/ΔMk mice does not directly result from insufficient Mk-derived TGFβ signaling in EPs. Thus, the effect of Mk TGFβ1 is restricted and predominantly acts on immature HSPCs.

Because we found that the Mks are responsible for the majority of TGFβ signaling in HSPCs, it is possible that blockade of TGFβ signaling could phenocopy these effects by inducing overproduction of erythroid-committed precursors. To test this, we pretreated C57BL6/J (B6) mice with a TGFβ1-neutralizing antibody (1D11) or nontargeting, isotype control antibody (13C4) and then either PBS or low-dose EPO (Figure 4H). TGFβ neutralization by 1D11 reduced pSmad2/3 MFI in HSPCs in wild-type mice whereas the 13C4 control had no effect, demonstrating on-target activity (Figure 4I). As we found in TGFβ1^ΔMk/ΔMk mice, B6 mice treated with endogenous TGFβ neutralized by 1D11 had an expanded pool of Lin⁻ and HSPCs in marrow compared with those receiving the control 13C4 antibody (supplemental Figure 4). Low-dose EPO triggered a brisk erythropoietic response in mice treated with 1D11 but not those treated with the 13C4 control (Figure 4J). Consistent with this finding, exogenous EPO rescued the EP dropout observed in B6 mice treated with 1D11 but did not affect the low apoptosis observed in mice treated with the 13C4 control (Figure 4K-L). Therefore, the boundary of megakaryocytic TGFβ1 activity is compartmentalized within the marrow with predominant effects on immature HSPCs while excluding their progeny (Figure 5).

Erythropoiesis is subject to modular regulation. EPO acts during a very limited stage of differentiation, supporting early EPs as they gear up for iron accumulation, heme synthesis, and globin gene transcription.^7,14,15  Accordingly, many causes of anemia are unresponsive to exogenous EPO. Recently, TGFβ superfamily activin receptor (Acvr2a/b) ligand traps have shown activity treating chronic EPO-unresponsive anemia in myelodysplastic syndrome and β-thalassemia, and are thought to promote erythroid maturation after EPO signaling.^25-29  However, although activins can support erythropoiesis in vitro and in vivo,^30,31  the molecular regulators of the EPO-unresponsive erythroid compartment are not well defined. Here, we show that the number of committed erythroid progenitors is indirectly controlled by the availability of megakaryocytic TGFβ1, adding a new level of erythroid regulation prior to the EPO-restriction point.

Megakaryocytic TGFβ1 helps to match the number of immature hematopoietic progenitor and stem cells to the production of mature effector cells. Deletion of Mk TGFβ leads to inefficient erythropoiesis because unneeded, excess EPs undergo apoptosis rather than contribute to blood cell output. Because Mk TGFβ1 ablation, or TGFβ blockade, broadly expands the pool of HSPCs, disruption of this pathway may disorder production of nonerythroid blood cell lines too. Direct testing will be required because the cytokines and maturation checkpoints controlling terminal granulopoiesis, thrombopoiesis, and lymphopoiesis differ substantially from erythropoiesis. Exogenous EPO or increased physiologic demand (supplemental Figure 2) rescued the excess EPs, leading to their brisk contribution to RBC counts. It is possible that other lineages could be similarly recruited from the enlarged pool of HSPCs if supported by appropriate cues. We previously found that TGFβ blockade during hematopoietic regeneration hastened multilineage reconstitution after chemotherapy, transplantation, or hemolysis.³²  These results suggest that in the appropriate context, when cytokine levels are enhanced and the microenvironment is primed, the expanded pool of HSPCs can effectively contribute to blood cell production.

It is likely that normally the release and/or activation of latent Mk TGFβ is tightly controlled. Our data suggest that disrupting homeostatic levels of Mk TGFβ could be used to augment blood cell production, but HSC exhaustion is possible if this approach were used continuously, for long periods of time. Conversely, abnormally high Mk TGFβ signaling may also have potentially deleterious effects on hematopoiesis. Indeed, excessive, dysregulated TGFβ signaling is commonly implicated in the pathology of fibrogenic diseases such as pulmonary fibrosis and myelofibrosis,^33-35  and excessive SMAD2/3 phosphorylation is found in the marrow of patients with myeloid malignancies^36,37  (P.K. and J.M.S., unpublished observations in human myeloproliferative neoplasms). Intriguingly, genetic deletion of phosphatidylinositol transfer proteins in Mks leads to excess TGFβ secretion and myelosuppression.³⁸  TGFβ is known to suppress erythropoiesis and megakaryopoiesis in vitro and in vivo,^39,40  suggesting that aberrant Mk TGFβ could contribute to the anemia and thrombocytopenia that characterizes many myeloid malignancies. In this study, we found that exogenous TGFβ1 worsened the erythroid apoptosis and caused anemia in TGFβ1^ΔMk/ΔMk mice. The biological effects of TGFβ1 differ based on ligand concentration, crosstalk with other signaling systems, and the type of downstream effector proteins used. For this reason, it can be reasonably expected that hematopoietic cells harboring pathogenic mutations could interpret TGFβ signals differently than normal HSPCs perhaps escaping some of the suppressive activities of this mediator. Elucidating the contributions of aberrant TGFβ signaling to abnormal hematopoiesis is a rich area for future research.

Although Mks and erythroids share a common progenitor, Mks also have a direct ontological link to a subset of Mk-biased HSCs.⁴¹  Understanding the evolutionary reasoning for Mks being placed at the helm of hematopoiesis and exploring the potential of manipulating this pathway clinically will be important.

This work was supported by grants from the National Institutes of Health, National Heart, Lung, and Blood Institute (HL1166436-01A1); TriSci (Tri-Sci 2013-030); and the Cancer Research and Treatment Fund (CR&T).

Contribution: S.D.G. designed and performed experiments, and analyzed data, wrote the manuscript; P.K. performed experiments, analyzed data, and edited the manuscript; N. Mollé, M.M.Y., G.A.-Z., T.S., F.B., N. Messali, and R.S. performed experiments; S.R. reviewed data and edited the manuscript; and J.M.S. designed the research, analyzed the data, and wrote the manuscript.

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

Correspondence: Joseph M. Scandura, Weill Cornell Medicine, 1300 York Ave, Box 113, New York, NY 10065; e-mail: jms2003@med.cornell.edu.