Iron deficiency affects early stages of embryonic hematopoiesis but not the endothelial to hematopoietic transition

Iron is an essential micronutrient for hematopoiesis and previous research suggested that iron deficiency in the pregnant female could cause anemia in the offspring. Since the development of all embryonic and adult blood cells begins in the embryo, we aimed to resolve the role of iron in embryonic hematopoiesis. For this purpose, we used an experimental system of mouse embryonic stem cells differentiation into embryonic hematopoietic progenitors. We modulated the iron status in cultures by adding either an iron chelator DFO for iron deficiency, or ferric ammonium citrate for iron excess, and followed the emergence of developing hematopoietic progenitors by flow cytometry. We found interestingly that iron deficiency by DFO did not block the endothelial to hematopoietic transition, the first step of hematopoiesis. However, it had a differential effect on the proliferation, survival and clonogenic capacity of hematopoietic progenitors. Surprisingly, iron deficiency affected erythro-myeloid Kitpos CD41+ progenitors significantly more than the primitive erythroid Kitneg CD41+. The Kitpos progenitors paradoxically died more, proliferated less and had more reduction in colony formation than Kitneg after 24 hours of DFO treatment. Kitpos progenitors expressed less transferrin-receptor on the cell surface and had less labile iron compared to Kitneg, which could reduce their capacity to compete for scarce iron and survive iron deficiency. We suggest that iron deficiency could disturb hematopoiesis already at an early embryonic stage by compromising survival, proliferation and differentiation of definitive hematopoietic progenitors.


Introduction
Embryonic hematopoiesis is an essential and complex process, which supplies blood to the developing embryo and the adult. It generates a wide range of hematopoietic progenitors and stem cells (HPSCs), from the primitive non-self-renewing erythroid progenitors of the yolk sac to the long-term self-renewing hematopoietic stem cells, which will reside in the adult bone marrow and generate all blood lineages when needed 1,2 . Despite its overall complexity, embryonic hematopoiesis can be simplified into a sequence of steps common for all HPSC types. It starts with the endothelial-to-hematopoietic transition (EHT), a step in which endothelial cells of a particular type (hemogenic endothelium) undergo through morphological and transcriptomic changes to become HPSCs [3][4][5] . Later on, these HPSCs proliferate, differentiate, and migrate to colonize their niches like the fetal liver, or the bone marrow 2, 6-8 . Recent work on human 9 and mouse embryonic stem cells 10,11 as well as on reprogrammed mouse embryonic fibroblasts 12 contributed a lot to the knowledge of the transcription factors and the growth factors controlling embryonic hematopoiesis.
Yet the role of iron in the steps of embryonic hematopoiesis is not completely understood.
Iron is an essential micronutrient required for catalysis, DNA synthesis, redox reactions and oxygen transport 13 . Iron deficiency through a knockout of iron import proteins like transferrin receptor (Tfrc) or Dmt1 (Slc11a2) causes anemia and embryonic or early postnatal lethality in mice 14,15 .
Nutritional iron deficiency in the pregnant female increases the risk of iron deficiency and iron deficiency anemia in the offspring, according to animal models and human epidemiological studies 16,17 . Hypotransferrinaemic hpx/hpx mice 18 or mice chimeric for Tfrc knockout 19 have a defect in T lymphoid differentiation, suggesting that the effects of iron deficiency might not be restricted only to the erythroid lineage. We therefore hypothesized that iron was important for an early step in embryonic hematopoiesis, which is common for all developing blood cells. Such a step could be either the EHT itself, or the steps right after it.
To dissect the step of embryonic hematopoiesis when iron is most required, we needed an experimental model where we could reproduce embryonic hematopoiesis, change the cellular iron status quickly and reversibly and test the effect of iron status on the steps of hematopoiesis in real time. To fulfill these requirements, we chose the experimental model of mouse embryonic stem cells progressively differentiating into blood through a hemangioblast stage 3 , similarly to what happens in yolk sac hematopoiesis 3 . The hemangioblast stage cultures start as Flk1 + mesoderm and differentiate with time into a mixed culture of endothelial, hematopoietic progenitor, and vascular smooth muscle cells 3 . In these cultures, we modified the cellular iron status by adding either an iron chelator (DFO) to cause iron deficiency or adding ferric ammonium citrate to cause iron excess 20 .
In this work, we demonstrated that iron deficiency by DFO did not block the EHT itself, but it differentially affected the proliferation, survival and differentiation of early hematopoietic progenitors. In contrast, iron excess had no adverse effects on hematopoietic progenitors. Thus, our findings offer broader understanding of how iron deficiency could affect embryonic hematopoiesis 16 .

Iron deficiency by DFO does not inhibit EHT.
We first tested whether iron deficiency would block the first step of embryonic hematopoiesis, which is the endothelial-to-hematopoietic transition (EHT) [3][4][5] . This hypothesis was based on previously published evidence that iron chelators were inhibiting the epithelial-to-mesenchymal transition (EMT), a mechanism similar to EHT 23 .
For this purpose, we differentiated mouse embryonic stem cells 21 into hemangioblast cultures 3,10,11 and followed the formation of hematopoietic progenitors from hemangioblast as a function of iron status and time. The time course of our experiments is schematically represented in Figure 1A Figure 1), suggesting no accumulation of these cells took place. We further observed an increase in the frequency of vascular smooth muscle cells after DFO treatment ( Figure   1E), but their absolute cell number was not increased (Supplemental Figure 1).
Excess iron added as 200 µM ferric ammonium citrate together with DFO abrogated all effects of DFO in culture, demonstrating that the observed effects were truly due to iron deficiency. Iron addition together with DFO or alone did not reduce the frequencies and absolute cell numbers of any cell type compared to untreated control ( Figure 1, Supplemental Figure 1), suggesting that in our conditions iron excess was at least not toxic to cells.

Iron deficiency differentially affects hematopoietic progenitors.
All hematopoietic progenitors in our cultures can be divided into two major subtypes: definitive Kit pos HPCs (Kit + CD41 + ) and primitive Kit neg HPCs (Kit -CD41 + ). The Kit pos HPCs mostly yield multilineage hematopoietic colonies when plated on methylcellulose with appropriate growth factors and are capable to reconstitute irradiated mice albeit transiently, while the Kit neg HPCs yield mostly primitive erythroid colonies on methylcellulose and have no reconstitution capacity 3,24 .
When we examined the effect of DFO on both kinds of HPCs, we saw that the frequency of Kit pos HPCs was significantly (p<0.0001) reduced compared to untreated control (Figure 2A,B) and the frequency of Kit neg HPCs was not significantly changed. The cell numbers of both progenitor types were significantly reduced, but the decrease in Kit pos HPCs was stronger (Supplemental Figure 2). This observation was surprising, since we did not expect to find a differential effect of iron deficiency on hematopoietic progenitors at such an early stage. Endothelial and vascular smooth muscle cells, viewed in these experiments as Kit + CD41and Kit -CD41respectively, behaved similarly to the experiments described in Figure 1 ( Figure 2 and Supplemental Figure 2).

Iron deficiency does not affect hematopoietic progenitor identity.
To find an explanation for the observed differential effect of iron deficiency on HPCs, we profiled gene expression in both kinds of progenitors by single-cell quantitative RT-PCR (sc-q-RT-PCR).

Iron deficiency differentially affects proliferation and survival of hematopoietic progenitors.
Since iron deficiency by DFO selectively reduced the frequency of Kit Pos HPCs, we investigated the causes of this reduction. Since the EHT is not inhibited by the DFO treatment (Figure 1), we hypothesized that the Kit Pos HPCs frequency could be reduced because of a decrease in proliferation or an increase in cell death. We measured cell proliferation in control, iron deficient and iron-excess conditions with the aid of a ClickIt-EdU kit, which only labels cells in S phase through incorporation of EdU 11 . Overall, hematopoietic progenitors were the most proliferating cells in culture having between 54-64% of S-phase cells, while endothelial and vascular smooth muscle cells proliferated less, with 23-30% ( Figure 4 and Table 2). In control conditions, both Kit Pos and Kit Neg HPCs had similar proliferation rate, but the decrease in proliferation after DFO was significantly stronger in Kit Pos HPCs. Adding excess iron together with DFO kept cell proliferation to control levels.
Our apoptosis measurements using AnnexinV and 7AAD 22 demonstrated that in control conditions the death rate of both progenitor types was not significantly different ( Figure 5A,B and Table 1).

DFO treatment increased the frequency of late apoptotic cells in both HPC types compared to
control or iron-treated groups ( Figure 5A). This increase in AnnexinV + 7AAD + cells was seen at 24 and 48 hours of DFO 50 µM treatment. Excess iron added together with DFO displayed apoptotic cell frequencies close to control levels. Since there was some extent of cell death in control conditions, we calculated the net cell death as delta apoptosis by subtracting the frequency of apoptotic cells in control conditions from the frequency of apoptotic cells with DFO (Δ DFO -control ).
The net cell death was significantly higher in Kit pos HPCs than in Kit neg HPCs in all time-points of DFO treatment ( Figure 5B). Together, our results suggest that DFO reduces Kit Pos HPCs frequency both by reducing proliferation and by increasing cell death.

Iron deficiency reduces colony-forming capacity of both Kitand Kit + hematopoietic progenitors.
We aimed to resolve whether iron deficiency would affect the differentiation of early hematopoietic progenitors into mature blood lineages. For this purpose, we sorted both kinds of HPs from control and DFO-treated conditions and put them on DFO-free methylcellulose plates for a week. As published previously 24 and in correspondence with our above-mentioned sc-q-RT-PCR results, Kit neg HPCs gave mostly primitive erythroid colonies after one week on methylcellulose; the Kit pos HPCs gave rise to erythroid, erythromyeloid and macrophage colonies. The clonogenic capacity of Kit pos HPCs was higher than of Kit neg , in agreement with previously published work 3,24 . DFO treatment significantly reduced the total amount of colonies formed from all HPCs together ( Figure   6) and the amount of erythroid and macrophage colonies from Kit pos HPCs. We did not observe a statistically significant effect of DFO on colony output from Kit neg HPCs. Overall, DFO treatment of early HPCs for 24 hours was sufficient to compromise HP clonogenic capacity, disturbing differentiation in the longer term.

Hematopoietic progenitors differ in their endogenous transferrin receptor protein expression and labile iron levels.
We hypothesized that the hematopoietic progenitors the most sensitive to iron deficiency should have higher metabolic demand for iron, reflected by higher expression of transferrin receptor (Tfrc) protein on the cell surface. However, our flow cytometry studies paradoxically showed the opposite. The more sensitive Kit Pos HPCs had significantly lower Tfrc expression than Kit neg HPCs, as seen by both the frequency (%) of Tfrc + cells and the mean Tfrc fluorescence ( Figure 7 and Table 3). Overall in culture, hematopoietic progenitors had higher Tfrc expression than nonhematopoietic cells, and most of the Tfrc fluorescence came from Kit Neg HPCs. Iron deficiency or iron excess gave a trend of increase or decrease of Tfrc expression respectively, but the Tfrc level was always higher in Kit Neg HPCs compared to other cell types ( Table 3). The Kit Neg HPCs also had the highest labile iron levels in culture as detected by cytosolic calcein (Table 3). Together, these observations could explain our paradox of HPCs sensitivity to iron deficiency: the more sensitive Kit Pos HPCs express less Tfrc and have less labile iron inside, therefore are less able to compete for scarce iron and survive through iron deficiency.

Discussion
In the present work, we aimed to resolve the role of iron in early hematopoiesis. While previous studies 14 , 15 demonstrated that iron deficiency causes anemia, some other works in Tfrc -/chimeric mice or the atransferrinemic mice showed that iron deficiency could also affect non-erythroid lineages [18][19] . Even in the total Tfrc knockout mouse, which had anemia and embryonic lethality by E12.5 at the latest 14 , there still was some yolk-sac erythropoiesis, suggesting that some hematopoietic progenitors and stem cells could be more sensitive to iron deficiency than others.
We therefore hypothesized that iron deficiency could affect not only erythroid development, but rather affect a hematopoietic step common for all blood lineages.
We used an in vitro model system of mouse embryonic hematopoiesis, which allowed us to visualize the early steps of the process in real-time 3,10,11 . It has the advantages of accessibly and reversibly modifying the cellular iron by chemical/pharmacological means, and lacking complications like embryonic lethality.
The addition of DFO to our hemangioblast cultures did not cause a massive accumulation of endothelial or Pre-HPCs characteristic of TGFβ treatment or overexpression of key transcription factors 10,11 . Neither the frequency nor the cell number of endothelial cells were increased (Figures 1,2 and Supplemental Figures 1, 2). Therefore, we conclude that DFO does not block the EHT per se.
In our cultures, we observed two kinds of hematopoietic progenitors, which differ by their Kit While the effects on proliferation and cell death were apparent right after completion of 24 or 48-hours treatment, the consequences on differentiation were long lasting. We could therefore suggest that iron deficiency could have both short and long-term effects on hematopoiesis.
We were interested in the reason for the differential sensitivity of our HPCs to DFO. In general, the sensitivity of a cell to iron deficiency could be determined by a combination of factors: the ironimport capacity of a cell; the iron content inside the cell; and the metabolic requirement for iron.
We initially assumed that the more sensitive progenitors would be more erythroid, requiring more iron for hemoglobin synthesis; and would express more transferrin receptor, but we actually observed the opposite. We unexpectedly found that the less erythroid Kit Pos HPCs were the more sensitive ones. When we measured the Tfrc and iron levels in both HPC types, we found that more sensitive Kit Pos HPCs had endogenously lower Tfrc expression and less intracellular iron than the Kit Neg HPCs (Figure 7 and Table 3). Having less Tfrc on the cell surface, the Kit Pos HPCs could be less capable to compete for scarce iron with Kit Neg . Having less iron inside, the Kit Pos HPCs could be less capable to survive iron deficiency.
Overall, we demonstrate for the first time that iron deficiency could affect hematopoiesis at an unexpectedly early embryonic stage. It is known from previous works that iron deficiency in pregnant rats reduced the total iron content in whole embryos 16 and fetal liver 25 , as well as hematocrit, hemoglobin and RBC counts in the offspring 16 . Our study suggests that iron deficiency could have consequences beyond erythropoiesis, also affecting developing myeloid cells in the embryo. Our work indicates that other aspects of hematopoiesis besides erythropoiesis need also to be assessed in cases of iron deficiency.

Maintenance and differentiaton of mouse embryonic stem cells (mESCs)
We used the A2lox Cre mESC cell line, which was a kind gift from Michael Kyba 21 . The detailed procedure of cell maintenance and differentiation is described in detail elsewhere 11

Hemangioblast cultures and experiments
Flk1 + cells were plated on gelatin-coated 6-well plates at 1x10 6  were added to the cells at a volume, which would not exceed 2% of the total well volume. In some cases, we used a well supplemented with 1-2% water as a control, which was similar to a well with

Cell proliferation assay
Cell proliferation was measured with the Click-It Plus EdU Alexa Fluor 488 flow cytometry assay kit from Invitrogen (C10633), according to manufacturer's instructions. Briefly, our blast cultures received 10µM EdU for 1 hour before cell harvesting and staining with above-mentioned antibodies. Stained cells were fixed and permeabilized and the ClickIt reaction was performed by manufacturer's instructions. Fluorescence was measured on FACS-Canto (BD).

Apoptosis assay
Apoptosis was measured by AnnexinV-FITC apoptosis detection kit (eBioscience #88 8005 72) according to manufacturer's instructions. Briefly, cells were harvested, stained with the anti-mouse cKit-APC and CD41-PE antibody mix, washed, incubated with Annexin-V-AF488 in AnnexinVbinding buffer, washed and put on ice. 7-AminoactinomycinD was used as a nuclear stain to label dead cells 22 . Fluorescence was measured on FACS-Canto (BD).

Hematopoietic Colony Forming Unit (CFU) assay
Hemangioblast cultures were grown as described above, but in 20-cm gelatinized dishes (Corning #10314601) instead of 6-well plates. 24 hours after plating, cells were treated or not with DFO 50 µM for 24 hours, then harvested and stained with a mix of anti-mouse CD144-efluor660, CD117-APC and CD41-PE antibodies as described above. Hematopoietic progenitors were gated as VE-Cad -CD41 + and sorted into separate sterile falcons according to their Kit/CD41 fluorescence profile (Kit + CD41 + or Kit -CD41 + ) with the FACSAria cell sorter. Dead cells were excluded by 7-AAD staining. The cells were plated into CFU mix with 55% methylcellulose, as previously described 3 .
Colony number was counted 1 week later. The experiment was independently repeated 5 times.

Single-cell quantitative RT-PCR
Hemangioblast cultures were grown as above and treated with either nothing or DFO for 24 hours.
Cells were harvested and stained with the same combination of anti-mouse antibodies used for colony assays. Single cells from each cell type (Kit -CD41 -, Kit + CD41 -, Kit + CD41 + and Kit -CD41 + ) were sorted with the FACSAria fluorescence cell sorter (BD Biosciences) into 96-well plates filled with 2 x lysis buffer from the CellsDirect One Step qRT-PCR Kit (Invitrogen, # 11753100) and snap-frozen on dry ice. RT/Specific target amplification reaction (RT-STA) was performed according to manufacturer's instructions. After RT-STA, individual single-cell qPCR was run on the Fluidigm Biomark HD system. Analysis was performed using R as previously described 11 .

Labile iron measurements
The method for labile iron measurements with calcein green was adapted from Shvartsman et al 20 .
Briefly, cells were stained with calcein-green AM 0.03 µM for 10min at 37 °C, washed, stained with antibodies as described above and sampled by flow cytometry in 10min intervals before (baseline) and after addition of a cell permeable chelator (L1 50 µM). The relative labile iron content is measured as the normalized delta fluorescence (ΔF) between baseline and chelator addition 20 .