In vivo Mapping of Notch Pathway Activity in Normal and Stress Hematopoiesis

Notch receptor in vivo lineage tracing reveals a division of labor during early hematopoiesis

To lineage-trace Notch receptor expression in hematopoiesis we have used mice with the Cre-ER^T2 cassette knocked into the endogenous loci of each of the Notch receptors. We crossed them to the ROSA26-RFP reporter strain (Luche et al., 2007) (Figure 1A). After tamoxifen administration, Notch(1-4)^CreER mice were analyzed following various periods of chase and only Notch1 and Notch2 were detectable in bone marrow progenitors. After both 3 and 7 day chase, Notch1^CreER predominantly labeled bone marrow progenitors with lymphoid potential including lymphoid-primed multipotent progenitors (L-MPP) and common lymphoid progenitors (CLP). (Figure 1B-D, Figure S1A, B). In contrast, Notch2^CreER-labeled cells were found mostly in non-lymphoid progenitors indicating that there is lineage-specific expression of Notch receptors in stem/progenitor cells. The peaks of Notch2 labeling were within the HSC and pre-erythrocytic stages of differentiation. Interestingly, short chase labeling experiments indicated an almost complete absence of Notch1 labeling within all non- lymphoid progenitor subsets, likely reflecting the lack of receptor expression of Notch1 in all such subsets. Intriguingly, even after 20wks of chase there were sustained higher levels of Notch2 labeling within the myelo-erythroid fraction (Figure S1C, D).

Closer analysis of progenitor subsets in Notch2^CreER mice revealed that the percentage of RFP⁺ cells was significantly higher in erythroid and megakaryocytic progenitors (pre-MegE, CFU-E, and MkP) than in granulocyte/monocyte progenitors (GMP) at both 3 and 7 days after tamoxifen injection. After a 3-day chase, approximately 15% and 30% of pre-MegE and CFU-E respectively were RFP⁺ compared to 2% in the GMP subset (Figure 1B). This implied that expression of Notch2 could be important in the megakaryocyte-erythrocyte (Meg-E) versus granulocyte-monocyte (GM) cell fate decision at an early progenitor stage. Additionally, Notch2^CreER-labeled cells were found with increasing frequency within the erythroid progenitor fraction (up to 35%) in comparison to megakaryocyte progenitors (MkP<10%) (Figure 1B,D). From the bipotent pre-megakarocyte-erythrocyte progenitor (pre-MegE) to the CFU-E erythroid progenitor, there was a sustained increase in RFP⁺ cells whereas the frequency actually decreased between the pre-MegE and the MkP at both time points. Notch3 and Notch4 labeling was not detected in the bone marrow (Figure 1B-D). Notch3 labeling appeared in early thymic T cell progenitors and was detectable in mature T cells with longer periods of chase (Figure 1C and data not shown). Notch4 labeling was extremely infrequent throughout the hematopoietic system but was detectable in splenic CD8⁺CD11c⁺ dendritic cells (DC), in agreement with previous qPCR data (Sekine et al., 2009) (Figure S1F). To confirm lineage tracing results at the level of transcription we carried quantitative RT-PCR analysis for the four Notch receptors (Figure S1E). Similar to our lineage tracing results, mainly Notch2 mRNA was detected in hematopoietic stem cells and only low levels of Notch1 mRNA. In the myelo-erythroid progenitor compartment, Notch1 was detected at low level in GMP and pre-MegE whereas Notch2 was detected at significantly higher levels. Only Notch2 mRNA was detected in CFU-E and MkP progenitors. Consistently, no expression of Notch3 and Notch4 mRNA was detected in these populations.

Furthermore, we were able to confirm these fate-mapping studies using receptor-specific antibodies (Fiorini et al., 2009). Indeed, Notch1 expression increased during commitment to the lymphocytic lineage and did not appear to be detectable outside the lymphoid lineage. Conversely, Notch2 was not expressed in early lymphoid progenitors (Figure 1E). Expression of Notch2 was low in CMP and GMP, but significantly higher in erythroid progenitors (Pre-MegE, CFU-E, ProE). Notch2 surface expression reached a peak in CFU-E progenitors and pro-erythroblasts but began to decrease at the basophilic erythroblast stage (Figure 1E and not shown). In comparison to erythroid progenitors, MkP expressed lower levels of surface Notch2. Whereas only a fraction of cells from populations like HSC were stained in the lineage tracing experiments, most of these cells appeared to be stained using antibodies. This discrepancy is most likely reflecting a partial recombination efficiency of Cre-ER, particularly after limited number of tamoxifen injection. Consistently with the lineage tracing results, Notch4 surface expression was detectable in a subset of splenic DC (Figure S1G) and undetectable in hematopoietic stem and progenitor cells in the bone marrow (Figure S1H), once more suggesting that Notch4 signaling plays no role during early hematopoiesis. Our combined lineage tracing and antibody labeling studies suggested a very tightly regulated division of labor between the Notch1 and Notch2 receptors during early stages of hematopoiesis.

In vivo mapping of Notch pathway activation during early hematopoiesis

To combine lineage tracing and receptor expression to pathway activation we used a novel reporter knock-in allele (Fre et al., 2011), in which the Hes1 locus, a well-known direct transcriptional target of Notch, drives expression of GFP (Hirata et al., 2002; Ntziachristos et al., 2012) (Figure 2A). To validate the model, we initially characterized the thymus, where the role of Notch1 is well established. We found that Hes1 mRNA expression was more than 50-fold higher in cells expressing Hes1^GFP and that expression of other well-characterized Notch target genes was also increased in Hes1^GFP+ cells, confirming that Hes1-expressing cells reflected the activation of the Notch pathway (Figure S2A). In the thymus, Hes1^GFP expression was first detected in a fraction of early T cell progenitors (ETP) and increased during T cell commitment up to the DN3 (CD4⁻CD8⁻CD25⁺cKit⁻CD44⁻) stage where Notch activation reached its peak (Figure S2B,C). After this stage of differentiation, Notch signaling activity is down-regulated (Kleinmann et al., 2008). Accordingly, we found that in CD4⁺CD8⁺ double positive T cells (DP), the expression of Hes1^GFP was also down-regulated (Figure S2B,C). Consistent with this differentiation stage-specific Hes1 expression, GFP⁺ cells were located primarily in the thymic cortex (Figure S2D). These studies validated the use of the HES1^GFP mice as a faithful Notch reporter in the immune system. In agreement with this idea, high levels of Hes1 expression were also found in marginal zone B cells (MZB), which require Notch signaling (Moriyama et al., 2008; Tanigaki et al., 2002) (not shown).

Notch pathway activation marks initiation of commitment to the erythroid lineage

In the bone marrow, Hes1^GFP expression was detected within the Lin^negc-Kit⁺ hematopoietic fraction containing stem and progenitor cells. Interestingly, Hes1 expression was detected within the LT-HSC fraction and was virtually undetectable at the downstream multipotential progenitor (MPP) subsets (Figure 2B). Whole-transcriptome studies of Hes1^GFP-positive and negative Lin^neg/cKit⁺/Sca1⁺ (LSK) cells supported our phenotypic analysis as they correlated the Hes1^GFP fraction to the expression of genes, characteristic of the HSC stage (Figure S3A). Interestingly, Hes1^GFP LSK cells also expressed gene signatures correlating to erythrocytic differentiation, suggesting an early molecular bias towards the erythrocytic and against the GM/Mk lineages (Figure S3B-C). Further fractionation of LT-HSC based on their GFP expression showed that the observed erythroid bias seen at the level of LSK population was also evident in this compartment. Indeed, LT-HSC expressed higher levels of key erythroid genes such as Klf1 or Epor and lower levels of myelo/lymphoid genes such as Sfpi1 (Figure S3D). Furthermore, when LSK GFP⁺ and GFP⁻ were tested for their erythroid potential in methylcellulose CFU-E colony assay, LSK cells expressing Hes1 showed higher erythroid colony forming potential (Figure S3E). In agreement with this expression profiling and our aforementioned lineage tracing experiments, which demonstrated an erythroid bias in Notch2^CreER labeling, Hes1 expression was virtually undetectable at the CMP and GMP progenitor subsets (Figure S2E) and significantly de-repressed at the bipotent pre-MegE stage and subsequently increased as cells committed to the erythroid lineage, at the pre CFU-E and CFU-E stages (Figure 2C). The extent of Notch activation was more pronounced in the spleen, a major site of erythrocytic differentiation in the mouse, as high levels of Hes1^GFP were detected in erythrocytic progenitors (Figure 2D, E). This higher pathway activity in the spleen could not be explained by differential surface expression of Notch2 and most likely reflects differential ligand availability in this tissue (data not shown). Interestingly, the activation of the Notch pathway was transient in both tissues as it was down-regulated at later stages of erythrocytic differentiation (Figure S2F). Hes1 expression remained low in MkP suggesting that Notch is not active (or is attenuated) in early megakaryocyte development (Figure S2G).

To directly connect Hes1^GFP expression to Notch signaling we have crossed the Hes1^GFP mice to animals that conditionally lack gamma-secretase activity and thus Notch signaling (Vav1-cre Ncstn^f/f). We had previously shown that Ncstn deletion phenocopies Notch receptor loss and that Ncstn^−/− phenotypes could be rescued totally by the expression of an activated form of Notch (Klinakis et al., 2011). Using Vav1-cre Ncstn^f/f Hes1^GFP mice we were able to show that Notch inactivation led to an almost complete reduction of Hes1^GFP levels in erythrocyte progenitors, suggesting direct regulation of Hes1 expression by Notch signaling in erythrocytic progenitors (Figure 2F,G).

Differential patterns of Notch activity between fetal and adult hematopoiesis

As initiation of erythropoiesis is a crucial event during fetal hematopoiesis we decided to study the activation of Notch signaling and its relationship to red blood cell differentiation in the embryo. We first looked at primitive hematopoiesis in the yolk sack. We carried quantitative RT-PCR for each of the Notch receptors as well as for Hes1. Yolk sack HSPC expressed both Notch1 and Notch2 but showed no Hes1 expression suggesting that Notch signaling is not activated at this stage. Interestingly, megakaryocytes and red blood cells showed only expression of Notch2 and red blood cells also expressed moderate but significant level of Hes1, suggesting activation of the Notch pathway (Figure S3F-G). We then focused on fetal liver embryonic hematopoiesis. We probed Notch receptor expression by RT-PCR and found mainly Notch2 expression (Figure 3A). Notch1 had low expression level on HSC, CMP and GMP but was not expressed in MEP, consistent with our observation in adult stem and progenitor cells. We then used the Hes1-GFP reporter strain to probe activation of the Notch pathway. Interestingly, we found that, fetal liver hematopoiesis follows a distinct pattern of activation than what we reported in adults. At embryonic day 13.5, the majority of the phenotypic LT-HSC cells expressed significant amounts of Hes1 however this expression was absent from more committed progenitors (Figure 3B). Hes1^GFP+ HSC cells were bona fide LT-HSC capable of long-term full lineage reconstitution of lethally irradiated recipients (data not shown). Similar to sorted Hes1^GFP cells from adult mice, fetal liver HSC expressing Hes1 were characterized by a Notch transcriptional signature (Figure 3C, D), as judged by transcriptome profiling and qPCR studies. Interestingly, Hes1^GFP expressing fetal HSC also revealed a lymphocytic gene expression bias in contrast to the erythroid bias found in adult Hes1^GFP expressing HSC (Figure 3D, E). Furthermore, Hes1 expression and Notch pathway activation was undetectable in fetal erythroid progenitors, underlining the differences between fetal and adult hematopoiesis in regard to Notch activation.

Hes1⁺ progenitors show increased erythroid potential and an erythroid transcriptional profile

Next, we returned to adult hematopoiesis and further focused on the connection of Notch activity to erythropoiesis. Lin^neg cKit⁺ Sca1⁻ progenitors (cKit⁺ progenitors) were sorted based on Hes1^GFP expression and plated into methylcellulose or collagen-based semi-solid medium to assay for lineage potential. Hes1 expressing cells were enriched in erythroid lineage potential whereas Hes1^neg mainly differentiated into granulocytes/monocytes or megakaryocytes (Figure 4A). FACS analysis demonstrated Gr1⁺CD11b⁺ cells were virtually absent in colonies generated by Hes1⁺ progenitors and that most cells expressed the erythroid marker Ter119. This enhancement in erythroid potential was supported by CFU-E assays, which confirmed this striking difference in the presence of Hes1 expression (Figure 4B). In addition, there were less CD41⁺ cells in colonies from Hes1⁺ progenitors (Figure 4A). We further addressed the megakaryocyte potential of c-Kit⁺ progenitors in acetylcholinesterase (AchE) stained collagen based semi-solid cultures, which is the definitive way to measure CFU-Mk potential (Jackson, 1973). Hes1⁺ cells formed significantly fewer AchE⁺ megakaryocyte colonies than Hes1^neg cells (Figure 4C). Furthermore, the infrequent CFU-Mk colonies generated by Hes1⁺ progenitors were usually much smaller than CFU-Mk colonies from cells that were Hes1^neg (data not shown). The absence of a correlation between Notch activation and promotion of the megakaryocytic (Mk) lineage further questions previous experiments suggesting a role for Notch4 signaling in this lineage.

To analyze the transcriptional program of progenitors with an active Notch pathway, we sorted cKit⁺ progenitors based on expression level of Hes1^GFP and used them for whole transcriptome analysis. Cluster analysis of microarrays from Hes1^GFP mice revealed that Hes1⁺ progenitors have a transcriptional profile almost identical to PreCFU-E erythroid progenitors (Pronk et al., 2007), whereas Hes1⁻ progenitors clustered closely with PreGM and MkP populations (Figure 4D). Also, we compiled genes important in erythroid, granulocyte/monocyte (GM), and megakaryocyte lineage specific differentiation and found that essential erythroid and Notch target genes were both upregulated in Hes1⁺ progenitors (Figure 4E). The expression of GM and Mk-specific genes was down-regulated. Changes in expression of selected genes were verified by quantitative qPCR. As expected, in the Hes1^GFP+ cells, Hes1 and Nrarp expression was significantly upregulated as was the expression of erythroid genes (Epor, Gypa) (Figure 4E and Figure S4A). The transcription factors Runx1 and the thrombopoietin receptor (Mpl) are essential for generation of normal megakaryocytes and platelets. In GFP⁺ progenitors, expression of both Runx1 and Mpl were decreased compared to GFP⁻. Moreover, the expression of transcription factors known to be important in the GM lineage, such as Pu.1 (Sfpi1) and Cebpa were reduced in GFP⁺ progenitors (Fig 4E and Figure S4A).

To further compare the gene expression profiles of Hes1⁺ to Hes1⁻ Lin^negcKit⁺ progenitors, we used gene set enrichment analysis (GSEA) and analyzed expression of lineage-affiliated gene sets establish by comparing the transcriptional profile of HSC to more committed progenitors such as LMPP, GMP, and MEP (Ng et al., 2009). Genetic signatures expressed in early (s-ery) and more committed (d-ery) erythroid progenitors were significantly enriched in Hes1⁺ Lin^negcKit⁺ cells. In contrast, gene sets characteristic of granulocyte/monocyte and lymphoid progenitors (r-myly) showed a negative enrichment (Figure S4B). As previously mentioned, our microarray analysis indicated that key megakaryocyte associated genes were repressed in Hes1⁺ progenitors. Notably, using similar gene sets (Mercher et al., 2008) we revealed an inverse relationship between the expression of Hes1 and megakaryocyte-related genes. Transcriptional profiling therefore supports the notion that Hes1 expression is characteristic of progenitors committed (or destined to commit) to the erythroid lineage.

To probe the role of Notch signaling in human erythropoiesis we examined public gene expression databases for expression of Notch-related genes. In agreement with the results presented thus far, we found, using meta-analysis of whole transcriptome data (Novershtern et al., 2011), that Notch pathway genes were upregulated during human erythroid differentiation (Figure S4E). More importantly, expression of Notch direct transcriptional targets (including Hes1, Hey1, Gata3 and Tcf7) increased in human erythroid progenitors. We therefore propose that Notch activation is an evolutionarily conserved hallmark of erythroid differentiation not only in mice but also in humans suggesting that our findings may be of broader significance.

Notch2 signaling commits bipotent Pre-MegE progenitors to the erythroid lineage

The bipotent Pre-MegE progenitors generate erythroid and megakaryocyte progenitors both in vitro and in vivo (Pronk et al., 2007). To investigate whether Notch signaling can function at this branch-point we sorted Pre-MegE based on Hes1^GFP expression and plated cells in semisolid media. Hes1⁺ cells generated a significantly higher number of pure erythroid colonies and fewer mixed erythroid-megakaryocyte colonies or pure megakaryocyte colonies (Figure 4F). Conversely, Hes1^neg Pre-MegE developed more mixed megakaryocyte-erythrocyte colonies and pure megakaryocyte colonies. This suggested that in the absence of Notch signaling, Pre-MegE retain megakaryocytic potential. Also, in CFU-E assays, Pre-MegE that expressed Hes1 generated more CFU-E than the Hes1^neg (Figure 4G). We then focused on Notch2, as it is the only Notch receptor expressed on pre-MegE progenitors and the down-stream early erythroid precursors (Figure 1, Figure S1). In agreement with our previous findings, Pre-MegE sorted for surface Notch2 expression generated more CFU-E erythroid colonies than Notch2^low progenitors (Figure S4C). Finally, gene-set enrichment analysis of Pre-MegE, sorted based on Hes1^GFP expression demonstrated that erythroid genes are enriched in the Hes1 expressing fraction in comparison to the GFP negative fraction. Analysis of megakaryocyte transcription gene sets (Mercher et al., 2008) revealed an inverse relationship between the expression of Hes1 and megakaryocyte-related genes as we previously found with cKit⁺ progenitors (Figure S4D).

Notch2 gain of function promotes early erythroid progenitor differentiation in vivo

If Notch signaling activation characterizes commitment of adult hematopoietic progenitors to the erythrocytic lineage it is conceivable that genetic manipulation of the pathway could promote or impede adult erythropoiesis in vivo. We thus generated transgenic mice expressing the active domain of each Notch receptor from the ubiquitously expressed ROSA26 promoter (Figure 5A). Using these animals, we conditionally expressed intracellular Notch-IC(1-4) (ICN1-4) using inducible Cre-recombinase strains and monitored expression of the transgene using YFP (Kühn et al., 1995; Ruzankina et al., 2007; Seibler et al., 2003). This approach allowed us to compare the functional consequences of expressing each of the four ICNs from the same promoter in hematopoiesis. Interestingly, although we could demonstrate that all transgenes were expressed at comparable levels (Figure S5A and not shown), only expression of ICN1 was sufficient to induce T cell leukemia in vivo (Figure S5B), suggesting that each ICN defines distinct thresholds of signaling strength. Since Notch2 is the only receptor expressed during early erythropoiesis, we focused our subsequent gain of function studies on the effect of Notch2 activation (ICN2) in bone marrow hematopoietic progenitors. Activation of ICN2 expression in hematopoietic cells using the polyI:polyC inducible Mx1-Cre resulted in a significant increase of CD105⁺ CFU-E erythroid progenitors and CD71⁺ erythroblasts in the bone marrow and spleen (Figure 5B-D). Platelet counts were significantly reduced in mice when ICN2 was expressed (Figure 5E and Table S1). Histological analysis (Figure 5F) showed that a dramatic increase in nucleated cells was apparent in the splenic red pulp at low power, and the margins of B cell follicles appeared to be blurred. ICN2 nucleated cells clearly displayed the morphology of erythroid blast cells. Numerous early erythroid progenitors and erythroblasts were abundant. In the bone marrow, there was an increase in erythroid progenitor cells, also confirmed by staining and differential counts of bone marrow smears (data not shown). Strikingly, there was 5-fold reduction of megakaryocytes in ICN2 bone marrow (Figure 5F), in agreement to the suggested negative role of the Notch pathway in megakaryopoiesis.

We next sought to examine gene expression of sorted cKit⁺ progenitors that express activated ICN2 in vivo. As seen in Hes1⁺ progenitors, ICN2-expressing progenitors upregulated erythroid genes and suppressed GM and megakaryocyte genes (Figure 5G). Transcriptional targets of Notch were upregulated in ICN2 expressing progenitors, confirming the ICN2 functionality. As with Hes1-expressing cKit⁺ progenitors and Pre-MegE, the expression of erythroid gene sets (dery) was enriched, while the expression of genes expressed in lymphoid or GM progenitors (rmyly) was suppressed (Ng et al., 2009) (Figure 5H). To further address the involvement of Notch in the choice between the erythroid and megakaryocyte lineages, we imposed the expression of ICN2 in liquid cultures of c-Kit⁺ progenitors using a ROSA26-CreER driver and 4-hydroxytamoxifen (4-OHT). In the ROSA26^YFP control, after induction with 4-OHT, both CD71⁺ erythroblasts and megakaryocytes were generated in culture. However, induction of ICN2 expression directed the differentiation of the progenitors into mostly CD71⁺ erythroblasts and inhibited the generation of FSC^highCD41⁺ megakaryocytes (Figure S5C). These combined studies further support our main hypothesis as they demonstrated that Notch2 gain-of-function enforces erythrocytic commitment both in vitro and in vivo.

Notch loss-of-function inhibits early erythroid progenitor differentiation

To study effects of Notch loss-of-function, we used either deletion of Notch1 and Notch2 or Nicastrin (a non-redundant part of the γ-secretase complex) to avoid putative compensatory receptor functions (Klinakis et al., 2011). We conditionally deleted Notch signaling in adult mice using the Mx1-Cre deleter strain and analyzed bone marrow progenitor distribution focusing on erythroblasts. In agreement with previous reports no overt anemia under steady-state conditions was noticed (Maillard et al., 2008; Mancini et al., 2005). However, in the absence of Notch signaling, we found that both the frequency and absolute numbers of CD105⁺ CFU-E progenitors and CD45⁻CD71⁺ erythroblasts were significantly reduced in comparison to control mice (Figure 6A-C). Furthermore, CFU-E and CFU-Mk assays using flow purified Lin^neg/cKit⁺/Sca1⁻ progenitors showed decrease ability of Notch deficient progenitors to generate CFU-E and increased ability to generate CFU-Mk (Figure S6A). In contrast to the aforementioned gain-of-function analysis using ICN2 ectopic expression, Notch loss-of-function (in both Notch1^−/−2^−/− and Ncstn^−/− mice) lead to increased platelet counts and numbers of splenic megakaryocytes (Figure 6D, Figure S6B). Cell cycle and apoptosis analysis of pre-MegE, pre-CFU-E and CFU-E populations revealed no significant differences between Ncstn^−/− and littermate control (Figure S6C-D), suggesting that the reduction of CFU-E population observed in Notch loss of function models is mainly due to differentiation bias. These data are consistent with our previous in vitro and in vivo studies, enforcing the idea that in a subset of progenitors Notch signaling is important for early stages of lineage commitment and erythroid progenitor differentiation.

Notch signaling is essential for optimal progenitor responses during stress erythropoiesis

To further test the importance of Notch signaling for the generation of adult red blood cells (RBC) and their progenitors, we deleted Ncstn specifically in the hematopoietic system usingNcstn^f/f Vav1-cre⁺animals and studied response to stress caused initially by the administration of phenylhydrazine (PHZ), an oxidative agent able to cause severe hemolytic anemia. Whereas no significant difference in the peripheral blood red cell compartment was observed at steady state (Figure S6B), Ncstn-deficient animals showed a significant delay of recovery from stress as demonstrated by lower numbers of total RBC (Figure 6E) circulating in the peripheral blood. As PHZ mainly targets mature RBC and we have shown that the activation of the Notch pathway is only transient and not evident in the later stages of RBC differentiation we focused on an additional stress stimulus. We selected sublethal ionizing radiation, which rapidly eliminates marrow and splenic erythroid progenitors while sparing mature peripheral RBC (Peslak et al., 2012). Using this stress stimulus we initially demonstrated that irradiation significantly increased the abundance of Hes1^GFP erythrocytic progenitors (Figure 6F) in the bone marrow, a response initiated as early as the pre-MegE stage. Interestingly, Notch2 receptor cell surface expression was increased in pre-MegE cells in response to stress (Figure S6D). These studies suggested an enhanced activation of the Notch pathway in response to erythrocytic stress. Most importantly, irradiation of Ncstn^−/− animals revealed a profound inability to mount a response to radiation-induced erythroid stress in the absence of Notch signaling. Indeed, both early (CFU-E, pre-CFU-E, pre-MegE) and late (ProE, BasoE, OrthoE) stages of red blood cell differentiation were virtually absent in mice lacking Notch signaling (Figure 6G-H). On the other hand, recovery of the pre-GM population was not significantly altered in Ncstn^−/− suggesting that the defects are specific to the erythroid lineage.

As Notch2 is the only Notch receptor expressed on the surface of erythrocytic progenitors we have also tested whether Notch2 deficiency is sufficient to impede differentiation of erythrocyte progenitors in response to stress. To exclude contributions from the microenvironment we have transplanted Vav1-cre⁺Notch2^f/fCD45.2⁺ bone marrow to irradiated congenic CD45.1⁺ recipients. As control we have used Vav1-cre⁻Notch2^f/fCD45.2⁺ bone marrow from littermate animals. Five weeks post-transplant and after we verified that recipient animal bone marrow was comprised by more than 90% Notch2^neg donor cells (Figure S7A-B), we have sub-lethaly irradiated the animals and analyzed erythrocytic stress response as previously described. Four days post-irradiation Vav1-cre⁺Notch2^f/f reconstituted animals showed signs of defective stress response including smaller spleens and significantly decreased numbers of pre-CFU-E and CFU-E progenitors (Figure S7C-E) in the spleen. On the other hand there were no significant differences in the recovery of granulocyte/monocyte and megakaryocyte progenitors suggesting specificity for the erythrocytic lineage. All these studies highlight the importance of Notch signaling, and specifically Notch2, in stress erythropoiesis.

Animals

Notch(1-4)CreER and Hes1-GFP knock-in mice were recently described (Fre et al., 2011). ROSA26-ICN(1-4) mice were generated by insertion of a loxP flanked splice acceptor NEO-ATG cassette with two polyA sites followed by ICN(1-4)-IRES-YFP into the ROSA26 locus, allowing the ROSA26 promoter to drive expression of the NEO-ATG cassette. Cre recombinase mediated excision of NEO-ATG results in use of the splice acceptor in the ICN(1-4)-IRES-YFP cassette and irreversible expression of the transgene and the IRES-YFP bi-cistronic mRNA allows expression to be monitored by YFP expression. ROSA26-RFP mice (gift from H-J. Fehling, Ulm University) and ROSA26-YFP animals were (gift from D. Littman, NYU School of Medicine) have been described (Luche et al., 2007; Srinivas et al., 2001). Inducible Cre animals used include: the tamoxifen inducible human ubiquitin C promoter driven CreER (Ubc-CreER) (Ruzankina et al., 2007)(gift from D. Littman, NYU School of Medicine), tamoxifen inducible ROSA26-CreER (Seibler et al., 2003)(gift from D. Littman, NYU School of Medicine), and polyI:polyC inducible Mx1Cre (Jackson Labs). N1^f/f N2^f/f and Ncstn^f/f mice were previously described (Klinakis et al., 2011). Hematopoietic specifique Vav1-cre was previously described (Stadtfeld and Graf, 2005). All animal experiments were done in accordance to the guidelines of the NYU School of Medicine Institutional Animal Care and Use Committee.

Notch lineage tracing and LSK transplants

Tamoxifen (Sigma Aldrich) was solubilized in corn oil (Sigma Aldrich) at a concentration of 20mg/mL and injected intraperitoneally at 0.2mg/g body weight. Following (1 or 3) daily injections and a 3 day, 7 day, or 20 weeks chase, animals were euthanized for analysis of peripheral blood and tissues by FACS. For transplantation of RFP labeled LSK cells, ROSA26^CreER, Notch1^CreER, and Notch2^CreER lineage tracer mice were crossed to ROSA26-RFP reporter mice and injected with tamoxifen daily for 3 days (0.2mg/g mouse). Two days after the last injection, bone marrow was isolated from femurs and tibias. Lineage depleted cells were stained as described and RFP⁺ LSK cells were sorted and 500-750 cells were transplanted with 2.5×10⁵ helper bone marrow cells into lethally irradiated recipients. Peripheral blood was analyzed 4 weeks and 24 weeks post-transplant.

Stress erythropoiesis

Phenylhydrazine was dissolved in PBS and injected intraperitoneally at 50 mg/kg on two consecutive days to induce acute anemia. Peripheral blood (50μl) was collected and analyzed using the Hemavet 950 (Drew Scientific) hematology system. Sublethal total body irradiation of 4 Gy radiation was used to model endogenous stress erythropoiesis as previously described (Peslak et al., 2012). Progenitor responses were analyzed at day 4 following sublethal irradiation.

Antibodies and Flow Cytometry

Freshly dissected femurs and tibias were dissected and bone marrow was flushed with a 3cc syringe and 25g needle into PBS with 3% fetal bovine serum. The bone marrow suspension was centrifuged at 400 rcf for 10min at 4°C, washed and resuspended in PBS with 3% FBS. Antibody staining and FACS analysis was performed as previously described (Klinakis et al., 2011). All antibodies were purchased from BD-Pharmingen or e-Bioscience. We used the following fluorochrome or biotin conjugated antibodies: CD117 (2B8), Sca-1 (D7), CD11b (M1/70), Gr-1 (RB6-8C5), NK1.1 (PK136), TER-119, CD3 (145-2C11), CD19 (1D3), CD21 (7E9), CD23 (B3B4), CD127 (A7R34), CD34 (RAM34), FcγRII/III (2.4G2 or 93), CD135 (A2F10.1), CD4 (RM4-5), CD8 (53-6.7), CD150 (9D1), CD41 (MWReg30), B220 (RA3-6B2), CD48 (HM481), CD105 (MJ7/18), F4/80 (BM8), CD71 (Rl7217). Bone marrow lineage antibody cocktail included CD11b, Gr-1, NK1.1, TER-119, CD4, CD8, CD3, B220, Il7Rα. For analysis of erythroid progenitors, TER-119 was not included in lineage cocktail as previously described (Pronk et al., 2007).

Microarray and Gene set enrichment analysis

Human hematopoietic population micro array data have been previously described (Novershtern et al., 2011) and are available at GEO database (http://www.ncbi.nlm.nih.gov/geo/) under accession number GSE24759. Data were normalized using the Robust Multi-array Average algorithm using Genespring GX software (Agilent) and gene expression were leveled according to GAPDH expression. Mouse erythoid and myeloid progenitor population micro arrays have been previously described (Pronk et al., 2007) and are available at GEO database under accession number GSE8407.

Microarray analysis was performed as previously described (Klinakis et al., 2011). For each experiment, freshly isolated cells from three individual mice were sorted by surface marker expression and GFP expression, total RNA was extracted using the RNeasy Plus Micro kit (Qiagen). The Ovation RNA Amplification System V2 or Ovation Pico amplification (NuGEN) kits were used for amplification. Amplified RNA was labeled and hybridized to the Mouse 430.2 microarrays (Affymetrix). The Affymetrix gene expression profiling data were normalized using the previously published robust multi-array average algorithm using the GeneSpring GX software (Agilent). The gene-expression intensity presentations were generated with Multi Experiment Viewer software (http://www.tm4.org/mev/). Gene set enrichment analysis was performed using Gene Set Enrichment Analysis software (http://www.broadinstitute.org/gsea) using gene set as permutation type, 1,000 permutations and log₂ ratio of classes as metric for ranking genes. Genesets used in this study have been previously published (Mercher et al., 2008; Ng et al., 2009; Pronk et al., 2007). Newly generated microarray data are available at GEO database under accession number GSE46726.

In vitro differentiation assays

Sorted LSK, Lin^neg/cKit⁺/Sca1⁻ progenitors (500) or Pre-MegE (500) were plated in triplicate into cytokine-supplemented methylcellulose medium (MethoCult 3434, Stem Cell Technologies). Colonies were scored after 10 days of culture and cells were collected for analysis by FACS. For CFU-E assays cells were plated in methylcellulose medium supplemented with EPO (MethoCult 3334, Stem Cell Technologies) and colonies were counted 2 days after plating. Collagen gel assays for CFU-Mk were performed using Megacult-C (supplemented TPO and IL-3, Peprotech). Collagen gels were acetone fixed and acetylcholinesterase staining was performed according to manufacturers guidelines (MegaCult-C, Stem Cell Technologies).

Statistical analysis

The means of each data set were analyzed using Student's t-test, with a two-tailed distribution and assuming equal sample variance.