Hematopoietic stem cells develop in the absence of endothelial cadherin 5 expression

During fetal development, rare aortic endothelial cells of the aorta-gonad mesonephros (AGM) region, termed hemogenic endothelial cells, transition to hematopoietic stem cells (HSCs). In zebrafish, the first definitive HSCs emerge from AGM between 30 and 36 hours postfertilization (hpf) and peak at 36 to 72 hpf. These HSCs migrate to caudal hematopoietic tissue (CHT) and later to kidney marrow and thymus.¹  Similarly, adult-type definitive HSCs first emerge in the aortic endothelium of the mouse AGM at embryonic day 10.5 (E10.5) and enter into circulation to seed the fetal liver where the production of hematopoietic progenitor cells occurs. Then, hematopoiesis shifts to the bone marrow and spleen, and the liver becomes the hub of metabolism.¹  Lineage tracing experiments have revealed that HSCs emerge from cadherin 5⁺ (Cdh5⁺) endothelial precursors.^2,3  Isolated populations of Cdh5⁺ cells from the E11.5 AGM of both murine embryos and mouse embryonic stem (mES) cells differentiate into hematopoietic cells.^4-7  Thus, Cdh5 is a marker of AGM-derived hemogenic endothelial cells transitioning to HSCs.

Sca1, Runx1, and Cdh5 are all markers of HSCs that emerge from hemogenic endothelial cells. Sca1 is itself dispensable for HSC emergence,⁸  whereas Runx1 is essential for the formation of erythroid/myeloid progenitors from the yolk sac and HSCs from the hemogenic endothelium.⁹  Cdh5, a cell-adhesion molecule, regulates endothelial permeability¹⁰  as well as vascular integrity.¹¹  Although Cdh5 is present in virtually all active HSCs derived from E11.5 AGM, no difference was observed in multilineage differentiation capacity between vascular endothelial (VE)-cadherin⁺CD45⁺ and VE-cadherin⁻CD45⁺ HSCs isolated from the E13.5 liver.⁷  In fact, there are conflicting reports suggesting that Cdh5 expression declines either in E13.5 fetal liver^6,7  or in E16.5 fetal liver HSCs.¹²  However, there is a consensus that Cdh5 is absent from adult bone marrow HSCs.^6,7,12  Prior studies have established that loss or truncation of Cdh5 does not significantly impair mouse primitive hematopoiesis in the E8.5 yolk sacs.¹³  However, Cdh5 mutant embryos die at and beyond E9.^13,14  Because HSCs emerge from hemogenic endothelium in the AGM around E10.5,^1,15  the direct role of Cdh5 in endothelial emergence and the development of HSCs as well as in adult hematopoiesis has not yet been firmly established. In this manuscript, we have addressed this question using zebrafish and murine systems.

All procedures were approved by the animal care and use committee of Boston Children’s Hospital. The malbec (mlb^bw306) mutant was recovered from our unbiased chemical mutagenesis screen. C57BL6/J, CD1, and CD45.1 (SJL) mice were purchased from The Jackson Laboratory. Cdh5^flox/flox mice were provided by Dietmar Vestweber (Max Planck Institute, Münster, Germany) and Scl-Cre-ER^T mice by Stuart H. Orkin (Boston Children’s Hospital, Boston, MA).

Bioinformatics tools were used to identify and validate cdh5 as the candidate gene disrupted in the mlb locus on the zebrafish chromosome.^16,17  The complementary DNA prepared from wild-type (WT) and mlb embryos were sequenced and the mutation in the exon 3 sequence was verified using an allele-specific oligohybridization technique.¹⁸  We performed quantitative reverse transcriptase–polymerase chain reaction (qRT-PCR) using TaqMan gene expression assays (zf-cdh5: Dr03089732_g1 spanning exon 3-4, Dr03089733_m1 spanning exon 4-5, Dr03089737_m1 spanning exon 8-9, Dr03089729_m1 spanning exon 11-12, and zf-hprt: Dr03138604_m1; Applied Biosystems) to measure transcript levels of cdh5 along the length of the gene. Morpholinos (Gene Tools) against splice sites (TACAAGACCGTCTACCTTTCCAATC) and translational (CCTCCTGGCACACTGTTTCATCATC) sites of cdh5 were designed and injected into WT embryos to analyze a loss-of-function phenotype. Thirty embryos were pooled in each experiment (n = 5) to isolate RNA for qRT-PCR analysis

We crossed cd41:eGFP¹⁹  with flk1:mCherry^20,21  zebrafish and runx1+23:NLS-mCherry²²  with flk1:eGFP²³  zebrafish and injected cdh5-morpholino in their transgenic embryos. We mounted cdh5-silenced transgenic embryos in low-melting-point agarose and used a spinning-disk confocal microscope to perform time-lapse confocal imaging of cd41:eGFP⁺ HSCs and runx1+23:NLS-mCherry⁺ hematopoietic stem and progenitor cells (HSPCs) emerging from flk1⁺ endothelium from 30 to 42 hpf.^22,24  We analyzed 15 cdh5-silenced cd41:eGFP::flk1:mCherry embryos (vs 13 control) and 23 cdh5-silenced runx1:mCherry::flk1:eGFP embryos (vs 11 control).

We injected cdh5-morpholino premixed with Dextran blue dye in 1- to 2-cell-stage-old cd41:eGFP and runx1+23:NLS-mCherry transgenic embryos, and allowed them to grow until their high to 30% epiboly stage. We used custom-made pulled glass needles to fuse the blastula of cdh5-silenced transgenic embryos with WT casper embryos. We allowed these fused embryos to grow in high-calcium Ringer's solution containing antibiotics, and then tested for the efficacy of fusions 24 hours past surgery.²⁵  Fused embryos viable after 24 hours of surgery were allowed to grow and were mounted in low-melting-point agarose to test the efficacy of HSC emergence in the morphant embryos. Their migration and engraftment into the WT CHT region between 32 and 48 hpf, and the subsequent differentiation into platelets between days 4 and 5 were captured using confocal microscopes. All 10 of 10 WT parabiots survived whereas 30 of 35 cdh5-silenced parabiots survived.

We injected cdh5-morpholino, premixed with rhodamine-dextran to label all of the donor cells, in lcr:eGFP transgenic embryos at the 1-cell stage and allowed them to grow until they were between high or dome stage. We harvested the blastula from cdh5-silenced transgenic embryos (donor) and injected ∼20 cells into age-matched WT embryos (recipient).^26,27  We let these embryos grow, and then tested for lcr:GFP⁺ erythroid cells in both morphant and transplanted (recipient) embryos at 36, 48, and 72 hpf (n = 14 recipients).

We designed clustered regularly interspaced short palindromic repeat (CRISPR) single guide RNA (sgRNA) sequences flanking exon 1 and exon 12 of the Cdh5 gene using CRISPR Design software (http://crispr.mit.edu/; supplemental Table 1, see supplemental Data available at the Blood Web site). We cloned these sequences in a vector containing Cas9.^28,29  We electroporated the plasmids targeting both the forward and reverse sgRNAs along with the H2B-mKO vector in mouse green fluorescent protein (GFP) embryonic stem (ES) cells (GFP-mESC LB10; GlobalStem) using the GenePulser II (Bio-Rad).³⁰  We coelectroporated CRISPR constructs with the mH2B-KO vector to allow us to isolate mES cells carrying CAS9, which provides a much greater screening efficiency. We sorted the GFP and orange-positive cells, plated them on mouse embryonic fibroblasts, and genotyped 1869 clones in order to detect a clone containing a homozygous deletion of the Cdh5 gene using the primers listed in supplemental Table 2.

We performed karyotyping of candidate Cdh5-deleted clones to eliminate a possibility of aneuploidy in their chromosomes during genetic manipulations.³¹ 

After engineering mES cells with ubiquitous expression of GFP, and identifying the Cdh5^−/−GFP⁺ mES clone, we karyotyped them to exclude aneuploidy. We injected these Cdh5^−/−GFP⁺ mES cells (∼35-50) into the C57/BL6-derived WT blastocysts to generate chimeric embryos, which were transferred to the oviduct of CD1 foster mothers.³²  Of the 318 injected embryos of 46 foster mothers, 24 chimeric embryos survived up to E11.5 and 11 chimeric embryos survived up to E13.5. In addition, we used 6 E10.5 chimeric embryos to perform immunohistochemistry of Runx1 expression in the AGM region. Extensive detail on numbers of chimeric mice embryos profiled at the E11.5 and E13.5 developmental stage is provided in supplemental Table 3.

We harvested E10.5 chimeric mouse embryos, embedded them in a paraffin block, performed transverse sections, and immunostained with Runx1, GFP, and 4,6 diamidino-2-phenylindole (DAPI) antibodies in order to detect their expression in the E10.5 AGM region.^33-35 

We harvested E11.5 AGM and E13.5 fetal liver tissues from chimeric mouse embryos and used their single-cell suspension for fluorescence-activated cell sorter (FACS), colony-forming unit capacity (CFU), and transplant analyses.^4,33,35  Specifically, we detected expression of GFP, CD41 (phycoerythrin-Cy7), c-Kit (allophycocyanin), CD45.2 (Pacific Blue) in the E11.5 AGM as well as Sca1 (phycoerythrin), c-Kit (allophycocyanin) expression in E13.5 fetal liver. Background staining and negative control for conjugated antibodies are provided in supplemental Figure 8. We also plated 2 embryo equivalent whole E11.5 AGM as well as sorted Cdh5-deleted GFP⁺ or sorted control GFP⁻ E13.5 fetal liver cells on M3434 media (StemCell Technologies) and analyzed their capacity to make granulocyte, erythroid, macrophage, megakaryocyte (GEMM), granulocyte macrophage, macrophage, and erythroid colonies.^4,35 

We transplanted 2 embryo equivalent (e.e.) chimeric E11.5 whole AGM cells into irradiated (9 cGy, 2 split dose at an interval of 2-3 hours) CD45.1 (SJL) mice. In contrast, we sorted GFP⁺ (Cdh5^−/−) and GFP⁻ (WT) cells from E13.5 fetal liver cells and transplanted them (∼250 000 cells) separately into irradiated CD45.1 mice. We used CD45.1 splenic helper cells for AGM and fetal liver transplants. All GFP⁺ transplants were positive for GFP expression. We then tested the efficiency of these transplanted cells to engraft and reconstitute B cells (CD19), T cells (CD3), and macrophages (Gr1 or Mac1) beginning after 4 weeks of transplantation and up to 16 weeks using FACS analysis.^4,33,35  Extensive detail on numbers of the E11.5 AGM and E13.5 fetal liver transplanted in irradiated SJL mice is provided in supplemental Table 3.

We developed Cdh5^flox/flox:Scl-Cre-ER^T (as described in Figure 6A) and verified its genotype (as described in Figure 6B). We injected Tamoxifen (2 mg/kg, subcutaneously [s.c.]) in pregnant Cdh5^flox/flox:Scl-Cre-ER^T every day from E5.5 to E10.5 (see Figure 6C). At E11.5, we harvested the AGM, seeded its single-cell suspension in M3434 media for CFU analyses, followed by the genotyping of colonies to analyze the Cdh5 deletion. Sequences for the Cdh5 and Cre primers are listed in supplemental Table 4.

Statistical analyses were performed by Student t tests. Significance was set at P < .05.

To investigate the role of cadherin 5 in blood development, we studied mlb, an embryonic lethal zebrafish mutant that emerged from an unbiased ethylnitrosourea screen for anemic embryos.¹⁶  We performed chromosomal walking to demonstrate cadherin 5 (cdh5, VE-CAD) as the most probable candidate gene for the mlb locus. To identify the genetic mutation in the mlb locus, we sequenced the cdh5 gene and found an A267T stop codon in exon 3 (supplemental Figure 1A). We then used qRT-PCR (supplemental Figure 1B) and whole-mount in situ hybridization (WISH; supplemental Figure 1C) to validate loss of the cdh5 messenger RNA (mRNA) levels in mlb embryos, suggesting that the A267T stop codon resulted in nonsense-mediated decay of cdh5 mRNA. As an antibody cross-reacting with zebrafish cdh5 protein was not available, we were unable to directly assay the cdh5 protein levels in mlb embryos.

To further verify the loss-of-function phenotype for cdh5, we injected 2 independently derived cdh5 antisense splice-blocking and translational-blocking morpholinos (MOs)¹⁷  to knockdown cdh5 expression in zebrafish embryos. The cdh5-silenced embryos (morphants) phenocopied mlb’s anemia (supplemental Figure 1D-E). We further found a loss of cdh5 mRNA levels in the morphant embryos, suggesting that the splice-blocking MOs accurately targeted cdh5 (supplemental Figure 1F). Injection of cdh5 complementary RNA is toxic to malbec embryos, so it is not technically feasible to study the gain-of-function phenotype. Our phylogenetic analysis and peptide alignments further demonstrate human (CDH5, chromosome 16) and mouse (Cdh5, chromosome 8) orthologs for cdh5 (supplemental Figure 1G). These findings, along with positional cloning, mutational analysis, and qRT-PCR, as well as WISH data provide convincing evidences that cdh5 is mutated in the mlb locus.

The cdh5 knockdown zebrafish embryos have impaired cardiac function and circulatory arrest due to poor formation of endocardial junctions, which results in leakage across the endothelial layer and a reduction in the density of the cardiac jelly.³⁶  We followed the circulation of lcr:eGFP⁺ erythroid cells using time-lapse confocal imaging of the lcr:eGFP::flk1:mCherry zebrafish embryos between 19 and 32 hpf, and observed no active circulation in mlb embryos despite the initiation of a heartbeat. We ruled out necrosis of hematopoietic cells using acridine orange staining of mlb embryos (supplemental Figure 2A). We thus concluded that the observed anemic phenotype of mlb results from pooling of erythroid cells in blood vessels (supplemental Figure 2B) in the absence of adequate blood circulation.

To study the role of cdh5 in zebrafish hematopoiesis, we first analyzed the expression of markers for primitive and definitive hematopoiesis. Unlike the mouse Cdh5 mutant, mlb zebrafish embryos survive, despite circulatory defects, past the stage when definitive hematopoiesis emerges. We found normal expression of markers for primitive erythroid progenitors (gata1) and myeloid (mpo, mpx) cells in the mlb or cdh5-MO embryos (Figure 1A-C). Analyzing these markers of definitive hematopoiesis, we detected normal HSCs (runx1 and c-Myb mRNA expression, mlb::c-myb:eGFP⁺ cells and mlb::cd41:eGFP^low levels) as well as definitive erythroid (slc4a1), lymphoid (rag1), and thrombocytic (cd41) cells (Figure 1D-I; supplemental Figure 2B), indicating that mlb has intact primitive and definitive hematopoiesis. Normal expression of runx1 and c-myb (supplemental Figure 2C-E) in cdh5-silenced embryos further demonstrates that the cdh5 morphant phenocopies definitive HSC formation as seen in mlb.

To demonstrate whether HSCs would emerge from cdh5-silenced hemogenic endothelium, we first performed time-lapse confocal imaging of cd41:eGFP⁺ HSC formation from flk1:mCherry⁺ endothelium in zebrafish blood vessels. We found that cd41:eGFP⁺ HSCs^19,37  are continuously being formed in the vascular endothelium of the dorsal aorta in cdh5-silenced cd41:eGFP::flk1:mCherry embryos (Figure 2A-B; supplemental Videos 1-2). We also found that arterial endothelium is intact in cdh5-silenced embryos (supplemental Figure 2F). To analyze whether cd41:eGFP⁺ HSCs emerging from cdh5-silenced cd41:eGFP⁺ embryos are null for cdh5 expression, we sorted cd41:eGFP⁺ cells, measured cdh5 mRNA levels, and found that cd41:eGFP⁺ HSCs indeed lack cdh5 expression (supplemental Figure 2G). These data indicate that arterial endothelium continues to transition to HSCs in the absence of cdh5. We independently verified these observations by tracking continuous endothelial emergence of runx1+23:NLS-mCherry⁺ HSPCs²²  from cdh5-silenced flk1:eGFP⁺ vascular endothelium (supplemental Figure 2H-J).

We performed parabiotic surgery²⁵  to fuse cdh5-silenced cd41:eGFP morphant embryos with transparent casper embryos, which were otherwise WT. We found that cd41:eGFP⁺ HSCs emerged from cdh5-silenced aortic endothelium and then migrated to the CHT (a region analogous to mammalian fetal liver) of the casper recipients. The cdh5-silenced cd41:eGFP⁺ HSCs engrafted, divided, and subsequently differentiated into circulating cd41:eGFP⁺ platelets at day 4 postfertilization (Figure 2C-F; supplemental Figure 2G; supplemental Videos 3-4). Similarly, we fused cdh5-silenced runx1+23:NLS-mCherry⁺ morphant embryos with transparent casper embryos, and found that runx1+23:NLS-mCherry⁺ HSPCs formed in the cdh5 morphant embryos migrated to the casper WT CHT for engraftment and differentiation (Figure 2G). These data demonstrate that cdh5 is dispensable for HSC formation in the zebrafish, and that cdh5-silenced HSCs can migrate, engraft, and differentiate into definitive hematopoietic lineages in a WT hematopoietic niche.

As both cdh5 morphant and mlb mutant embryos have circulation defects, we further analyzed whether the migration of runx1:mCherry⁺ HSPCs from cdh5 morphants to the WT niche was due to a partial recovery of cardiac function and circulation, or due to the exchange of the blastula cells early in development. Because surgically paired embryos appeared to have 1 fused heart, we hypothesized that partial or complete recovery of cardiac function from the WT zebrafish was responsible for the migration of runx1:mCherry⁺ HSPCs from cdh5 morphants to the casper niche (Figure 2G). To test this, we fused 2 zebrafish embryos along the lateral body and head, such that the cdh5 morphant zebrafish could be directly observed to lack active circulation. Indeed, we found that the runx1: mCherry⁺ HSPCs could not migrate from cdh5 morphant to casper CHT in these pairs (Figure 2H). These data reinforce the conclusion that runx1+23:NLS- mCherry⁺ hematopoietic cells are formed in the AGM of cdh5 morphant zebrafish embryos and subsequently migrate via passive circulation to the casper CHT to engraft and differentiate (Figure 2G-H). We independently verified our time-lapse confocal imaging data by documenting levels of cd41:eGFP^low HSCs at 36 hpf and their differentiation into cd41:eGFP^high platelets at day 4 in cdh5-morphant embryos (Figure 2I). However, we could not confirm whether the increase in cd41 expression translates into an increase in functional HSCs. Thus, HSCs are being formed in the absence of cdh5, and manifest their stem cell properties in a casper CHT.

To further investigate the cell-intrinsic roles of cdh5 in HSC differentiation, we silenced cdh5 in erythroid transgenic lcr:eGFP embryos. We then allowed these morphants to develop up to high or dome stage, and then transplanted the blastula cells from cdh5-silenced transgenic embryo into WT recipients. We not only found that cdh5-morphant embryos produced lcr:eGFP⁺ erythroid cells, but also that cdh5-silenced blastula transplant (donor) cells contributed to lcr:eGFP⁺ erythroid cells in WT recipients at the definitive hematopoiesis stage of embryonic development (Figure 3). These data demonstrate that blastula cells lacking cdh5 expression differentiate into erythroid cells.

As Cdh5 mutant mouse embryos die before first HSCs emerge in AGM, we set out to analyze chimeric mice (in which Cdh5^−/−GFP⁺ ES cells injected into Cdh5^+/+GFP⁻ blastocysts) to test whether functional HSCs emerge from Cdh5-deleted hemogenic endothelium (Figure 4A).

We first used CAS9 nuclease genome editing to target sequences flanking exon 1 and exon 12 of the Cdh5 gene in mice (supplemental Figure 3A; supplemental Table 1). We engineered mouse ES cells with ubiquitous expression of GFP and identified a Cdh5^−/−GFP⁺ mES clone (supplemental Table 2; supplemental Figure 3B). After verifying loss of Cdh5 gene expression (supplemental Figure 3B) and confirming their normal karyotype (supplemental Figure 3C), we generated blastocyst chimeras by injecting the Cdh5^−/−GFP⁺ mES cells into WT embryos, and then transplanted these chimeric embryos into CD1 foster mothers (Figure 4A). By western blot, we confirmed the absence of Cdh5 expression in GFP⁺ cells isolated by flow cytometric sorting from the E13.5 fetal livers of chimeric embryos (Figure 4B). The percentage of GFP chimerism in 11 E11.5 AGM, 6 E13.5 fetal liver, and peripheral blood of 3 live pups is provided in Figure 4C.

We recovered E10.5 chimeric embryos, embedded in paraffin, and performed sections. By immunohistochemistry,^33,34  we found that Runx1⁺GFP⁺ cells emerge from the aortic endothelium (Figure 4D; supplemental Figure 3D-E) in a pattern similar to Runx1⁺ expression in E10.5 WT mice embryos, thus demonstrating the emergence of hematopoietic clusters from Cdh5-deleted mammalian hemogenic endothelial cells.

To prove the emergence of functional HSCs, we isolated the AGM from E11.5-old chimeric embryos and analyzed the hematopoietic marker-expression profile, colony formation, and reconstitution capacity of Cdh5^−/−GFP⁺ or Cdh5^+/+GFP⁻ AGM-derived cells. These cells expressed CD45.2, c-Kit, and CD41 (Figure 5A; supplemental Figures 4A-D and 5A-D), and differentiated into multipotential GEMM progenitors, granulocyte-macrophage progenitors (CFU-GM, CFU-M, CFU-G), and erythroid colonies (CFU-E) (Figure 5B-C). These data demonstrate that the Cdh5 is dispensable for CD41, c-Kit, and CD45.2 constitutive expression as well as the hematopoietic colony-forming capacity of Cdh5^−/−GFP⁺ HSCs. We, however, could not distinguish whether hematopoietic progenitors in E11.5 AGM are from yolk sac or AGM. In addition, transplantation of 2 e.e. of chimeric E11.5 AGM resulted in the long-term (16 week) multilineage engraftment of Cdh5^−/−GFP⁺ B cells, macrophages, and T cells in irradiated B6.SJL-Ptprc^aPep3^b/BoyJ (CD45.1) recipient mice^4,35  (Figure 5D-E; supplemental Figures 4E-G and 5E-H; Figure 5J).

Similarly, the fetal liver of E13.5 chimeric embryos had Lin⁻Sca⁺c-Kit⁺GFP⁺ hematopoietic cells (Figure 5F; supplemental Figures 6A-C and 7A-C), which differentiated into CFU-GEMM, CFU-GM, CFU-M, and CFU-E hematopoietic colonies (Figure 5G). The transplantation of E13.5 Cdh5^−/−GFP⁺ or Cdh5^+/+GFP⁻ fetal liver cells into irradiated CD45.1 mice also produced engraftment with B cells, macrophages, and T cells, indicating production of functional HSCs from Cdh5-deficient cells (Figure 5H-I; supplemental Figures 6D-F and 7D-E). To analyze the percentage of donor Cdh5^−/−GFP⁺ cells contributing to adult lineages, we analyzed the percentage of GFP⁺CD45.2⁺ peripheral blood chimerism in E11.5 AGM or E13.5 fetal liver transplants (Figure 5J). We also measured the percentage of GFP⁻CD45.2⁺ peripheral blood chimerism in E11.5 AGM or E13.5 fetal liver transplants to measure WT donor cells contributing to adult lineages (supplemental Figure 7F). To exclude reversion of Cdh5 expression, we also sorted GFP⁺CD45.2⁺ cells from E13.5 fetal liver transplanted recipients and analyzed Cdh5 gene expression (supplemental Figure 7G). We found deletion of Cdh5 in sorted cells, suggesting that GFP⁺ blood lineages in transplants are derived from Cdh5-null GFP⁺ mES cells.

After generating a conditional mouse Cdh5 knockout (Cdh5^flox/flox:Scl-Cre-ER^T; Figure 6A-B), we deleted Cdh5 floxed alleles beginning at E5.5 using tamoxifen-regulated Cre driven from Scl regulatory sequences. Tamoxifen was injected into pregnant dams from E5.5 to E10.5 to execute the Cdh5 deletion (Figure 6C). We harvested E11.5 embryos, dissected the AGM, and enumerated CFU colonies in methylcellulose colony-forming assays. We compared the number of GEMM colonies in the Cdh5^fl/fl:Scl-Cre-ER^T embryos with the Cdh5^fl/+:Scl-Cre-ER^T embryos for potential Cre toxicity (Figure 6D). We assessed the deletion of the Cdh5 alleles in individual CFU colonies from Cdh5^fl/fl: Scl-Cre-ER^T and Cdh5^fl/+:Scl-Cre-ER^T by PCR (Figure 6E). We found that the deletion of Cdh5 did not alter numbers of GEMM colonies (Figure 6D). Therefore, Cdh5 is dispensable for hematopoietic colony formation from Scl⁺ cells, further strengthening the conclusion that Cdh5 is dispensable within blood lineages.

Lineage tracing,^2,3  in vitro differentiation,^4,5  and various mouse studies have demonstrated that HSCs emerge from AGM-derived Cdh5⁺ endothelial precursors,³⁸  implicating Cdh5 as a marker for the hemogenic endothelial cells that transition to HSCs.^39-41  Because Cdh5 mouse mutant embryos die^13,14  before AGM-derived HSCs emerge at E10.5,^1,15  whether Cdh5 is essential for differentiation of HSCs into definitive blood lineages has remained unclear.¹²  In this study, we demonstrate that functional HSCs emerge from Cdh5-deficient AGM-hemogenic endothelium during zebrafish and murine fetal development, and show that Cdh5-deleted HSCs engraft and reconstitute multilineage adult blood. Thus, we have established that Cdh5 is dispensable for HSC formation and differentiation (supplemental Figure 9).

Creation of genetic mosaics (chimeras) between WT and genetically modified cells is a well-established and classical strategy for demonstrating cell-intrinsic effects of a gene on cell functions.^42,43  Creation of zebrafish chimeras has been used to study the cell-intrinsic functions of cloche and bloodless genes in HSPC proliferation, survival, and differentiation.^26,27,44  Here, we demonstrated that transplanted cdh5-silenced blastula cells differentiate into red cells in WT zebrafish embryos. To test the cell-intrinsic role of Cdh5 in mammalian blood development and corroborate our findings from zebrafish, we bypassed the lethality of Cdh5 mutant mouse embryos by creating chimeras of WT mouse embryos with GFP⁺Cdh5-mutant ES cells. A similar approach has been used to study cell-intrinsic roles of Ascl2 and Eed.^45,46  Although we did not perform comprehensive functional analyses of Cdh5-deficient blood cells by competitive transplantation, we observed that Cdh5^−/−GFP⁺ cells isolated from E11.5 AGM and E13.5 fetal liver express a normal range of hematopoietic markers, differentiate normally into hematopoietic colonies, and reconstitute normal levels of adult B cells, macrophages, and T cells following transplantation into irradiated SJL mice, indicating the production of functional HSCs from Cdh5-deficient cells. We also developed a conditional mouse knockout of Cdh5 (Cdh5^flox/flox:Scl-Cre-ER^T) and independently validated that multipotent hematopoietic colonies form from Scl⁺ hemogenic endothelial cells despite the absence of Cdh5. Therefore, our zebrafish, mouse chimera, and conditional mouse knockout data demonstrate that although Cdh5 is a marker of definitive hemogenic endothelium, it need not be expressed within the blood lineage to enable HSC emergence and function. However, our data do not exclude the possibility that an unknown molecule might compensate for the function of Cdh5 critical to hematopoietic specification. Our analyses provide a refined view of the molecular features of hemogenic endothelium and illustrate the power of combining model organisms in order to investigate the factors regulating HSC development and differentiation.

The authors are grateful to Kathryn Crosier, Philip Crosier, Jeff Cooney, Iman Schultz, Matthew King, Yi Zhou, Christian Lawrence, Eric L. Pierce, Benjanim S. Brigham, Jessica Collantonio, Spencer K. Heggers, and Michelle Cassim for their technical help and/or for providing reagents.

D.I.S. is supported by grants from the American Society of Hematology, the Cooley’s Anemia Foundation, and the National Institutes of Health (NIH), National Institute of Diabetes and Digestive and Kidney Diseases (NIDDK) grants (K01DK085217 and R03DK100672). H.A. is supported by a fellowship from the Sigrid Juselius Foundation. E.J.H. is supported by a fellowship from the Helen Hay Whitney Foundation. P.N.G.R. is supported by a fellowship from the American Society of Hematology. O.J.T. is supported by a fellowship from the American Society of Hematology. G.A.M. is supported by a fellowship from the Alex’s Lemonade Stand Foundation. D.E.B. is supported by a grant from the NIH, NIDDK (K08DK093705). B.H.P. is supported by a grant from the NIH, National Heart, Lung, and Blood Institute (P01HL032262). M.F. and D.V. are supported by grants from the Deutsche Forschungsgemeinschaft and the Max Planck Society. L.I.Z. is supported by grants from the NIH National Cancer Institute (R01CA103846) and National Heart, Lung, and Blood Institute (P01HL032262 and R01HL04880). S.H.O. is supported by grants from the NIH NIDDK (P30DK049216) and National Heart, Lung, and Blood Institute (R01HL032259 and P01HL032262). G.Q.D. is supported by grants from the NIH National Heart, Lung and Blood Institute (Progenitor Cell Biology Consortium UO1-HL100001) and NIDDK (R24DK092760), the Alex’s Lemonade Foundation, and the Boston Children’s Hospital Stem Cell Program.

L.I.Z., S.H.O., and G.Q.D. are investigators of the Howard Hughes Medical Institute.

Contribution: D.I.S. originally conceived the project, designed, supervised, and performed the experiments, analyzed data, and wrote the manuscript; H.A. identified the Cdh5 mutation, and performed in situ and blastula transplantation; T.C.P. performed confocal imaging and parabiotic experiments and helped with chimeric mouse experiments; H.A., T.C.P., E.J.H., K.A.S., O.J.T., K.H., N.C.H., J.P.K., and G.V. helped with zebrafish husbandry, imaging, and analyses; T.C.P., P.N.G.R., P.G.K., M.J.C., G.A.M., M.N., and Y.F. helped with mice colony management, injection of mES cells, transplantation of hematopoietic tissues, and FACS analyses; D.E.B. helped to design CRISPR constructs; M.F. and D.V. provided Cdh5^flox/flox mice; and D.V., B.H.P., L.I.Z., S.H.O., and G.Q.D. analyzed data and wrote the manuscript.

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

The current affiliations are: H.A., University of Helsinki, Helsinki, Finland; K.A.S., University of Buffalo School of Medicine and Biomedical Sciences, Buffalo, NY; M.F., Uppsala University, Uppsala, Sweden; K.H., Max Planck Institute for Molecular Biomedicine, Münster, Germany; J.P.K., The Medical Foundation, Boston, MA; and G.V., Institut de Cancerologie de Lorraine, Vandoeuvre-lès-Nancy, France.

Correspondence: Dhvanit I. Shah, Division of Hematology, Brigham and Women’s Hospital, Harvard Medical School, Harvard Stem Cell Institute, Boston, MA 02115; e-mail: dshah@research.bwh.harvard.edu.