Precisely how the first HSCs are generated in the vertebrate embryo has been a matter of controversy for several decades. Recent studies have strongly supported the postulate of hemogenic endothelium, ECs that transiently possess the ability to generate HSCs during vertebrate development
5. By targeting expression of the Cre recombinase specifically to cells of the vasculature, Zovein et al. showed, using floxed reporter genes, that HSCs were generated from Cdh5
+ (also referred to as VE-Cadherin) precursors, suggesting that HSCs arise from endothelium or shared endothelial precursors
1. Furthermore, experiments utilizing an inducible Cdh5:Cre
ERT2 transgene suggested that ECs within the region flanked by the aorta, gonads and mesonephros (AGM) in the midgestation mouse embryo contained the majority of HSC potential
1. In addition, conditional deletion of the
Runx1 transcription factor gene in Cdh5
+ cells led to loss of HSCs, suggesting that Runx1 function is key in the transition from endothelium to HSC
2.
In vitro studies have also suggested that ESC derivatives can generate hematopoietic cells through hemogenic endothelial intermediates
3,4,6. It remains to be determined, however, which regions of the embryo, or extraembryonic tissues, possess endothelium with hemogenic potential.
The appearance of cells having HSC characteristics has been observed in close association with arterial endothelium
5,7. In particular, the ventral floor of the dorsal aorta (DA) has been suggested by a number of investigators to be the primary birthplace of HSCs
8. In the zebrafish embryo, we
9 and others
10 previously demonstrated that expression of a
cmyb:eGFP transgene marks nascent HSCs along the ventral aspect of the DA between 28–48 hpf. To determine whether these cells arise directly from vascular precursors, we generated
cmyb:eGFP; kdrl:memCherry11 double transgenic animals and performed confocal timelapse imaging. Between 28–32 hpf, expression of the
kdrl trangene (also known as
flk1 and
vegfr2) within the zebrafish equivalent of the AGM region is localized to the aorta, vein and developing intersomitic vessels; hematopoietic expression of the
cmyb transgene initiates in cells along the DA around this time (). Four-dimensional imaging demonstrated that
cmyb:eGFP+ cells arose directly from
kdrl:memCherry+ cells specifically along the ventral aspect of the DA (
Supplemental Movies 1, 2). As shown in ,
kdrl:memCherry+ ECs displaying typical flattened morphology were occasionally observed to transform into spherical shapes, forming buds that extended into the lumen of the DA. By virtue of the membrane-specific expression of mCherry, buds were observed to initiate as
kdrl+cmyb− cells transitioned to
kdrl+cmyb+ cells (). In contrast to the proposed budding of mammalian HSCs into aortic circulation
5, we almost always observed HSCs to migrate ventrally towards the caudal vein (CV;
Supplementary Movies 1, 2). This is consistent with the observations of Kissa et al.
12, which suggested that AGM HSCs enter circulation via the dorsal wall of the CV in the zebrafish.
To confirm the hematopoietic nature of these budding AGM cells, we performed flow cytometry on dissociated
kdrl:RFP13; cmyb:eGFP embryos at 36 hpf, the timepoint at which we observed the peak in number of
kdrl+cmyb+ cells. Embryos were dissected to separate anterior, head tissues from the posterior trunk/tail region that contains the AGM. In accordance with our microscopic observations, no
kdrl+cmyb+ cells were observed in anterior regions () above background. By contrast, 0.25% of posterior cells were
kdrl+cmyb+ cells (). We thus reasoned that these double positive cells represented the nascent HSCs observed in our imaging experiments ().
kdrl+cmyb+ cells could be subdivided based upon differing levels of the
cmyb:eGFP transgene (); each subset, along with single positive posterior fractions were highly purified by FACS and queried for expression of hematopoietic and vascular genes by qPCR. As expected, expression of endothelial genes, including
kdrl, cdh5, and
lmo2, were highly expressed in
kdrl+cmyb− cells (). In general, early
kdrl+cmyblo precursors maintained similar expression levels of these vascular markers. As
cmyb:eGFP levels increased in maturing
kdrl+cmyb+ cells, however, expression of most endothelial genes dropped dramatically. By contrast, expression of
cd41, one of the earliest markers of mesodermal commitment to definitive hematopoiesis
14,15, initiated in
kdrl+cmyb− cells and increased as
kdrl+ cells became
cmyb+ (). We observed little to no
cd45 expression in either
kdrl+cmyb− or
kdrl+cmyblo cells. As these precursors matured, however, we observed concomitant upregulation of
cd45 in
kdrl+cmyb+ and
kdrl−cmyb+ subsets (). These results are consistent with findings in the mouse. On embryonic day 10, murine AGM HSCs do not express CD45 (encoded by
ptprc)
14,15. By day 11, however, embryonic HSCs become CD45
+14,15. Collectively, these results support the hypothesis that the
kdrl+cmyb+ cells observed to arise from the ventral wall of the DA are definitive hematopoietic precursors.
In order to trace the progeny of ECs in the zebrafish embryo, we employed an indelible marking system utilizing a floxed reporter transgene and a Cre driver that is specific to endothelium via
kdrl upstream promoter/enhancer elements
16. In the zebrafish, there exist two orthologues of the mammalian
flk1 gene,
kdr and
kdrl. The former exhibits a pan mesodermal expression pattern
17, whereas
kdrl and
Tg(kdrl:Cre)s898 (
Supplementary Figure 1) is expressed only in endothelium in a manner nearly identical to that of the murine
cdh5 gene
18,19. Thus, the
Tg(kdrl:Cre)s898 line used in our studies mimics the
cdh5:Cre knock-in mouse line employed in previous studies
1,2. We crossed
kdrl:Cre animals to animals carrying a
Tg(βactin2:loxP-STOP-loxP-DsRed-express)sd5 “
switch” reporter transgene in which 10.5kb of upstream βactin promoter/enhancer sequence is followed by a 5.7kb floxed “superstop” cassette. Immediately downstream is a DsRed
express gene that serves as a reporter for Cre-based removal of the superstop cassette. As presented in
Supplementary Figure 2, reporter gene expression was never observed in the absence of Cre, and induction of Cre in 24 hpf embryos led to stable reporter expression within nearly all hematopoietic cells for one year.
To test whether the
kdrl+; cmyb+ hematopoietic precursors observed in the AGM are
bona fide HSCs, we generated
kdrl:Cre; switch animals for long-term studies. Compared to whole kidney marrow (WKM) isolated from single transgenic
switch animals that showed no expression of DsRed (), double transgenic
kdrl:Cre; switch animals showed the vast majority of leukocytes to be labeled at six months of age (). Analysis of a large cohort of double transgenic animals showed that over 90% of WKM cells were marked at 3 months of age (). At 6 months of age, over 96% of WKM myeloid cells expressed the DsRed
express lineage tracer (). Since this cellular subset is comprised of over 90% neutrophils, which are characterized by lifespans restricted to a few days, this result suggests that the vast majority of, if not all, HSCs were marked by the
kdrl:Cre transgene during their embryonic formation. Finally, marked WKM was sorted by DsRed expression level and subjected to qPCR for lineage-affiliated genes. As presented in ,
pax5 (B lymphocyte-affiliated) and
pu.1 (myeloid-affiliated) expressing cells were contained within DsRed
high cells, and
gata1 expressing cells within the DsRed
low fraction. These data are consistent with our previous demonstration that the βactin promoter is highly expressed in leukocytes, but is silenced upon erythropoietic differentiation
20. Thus, lineage tracing of
kdrl+ hemogenic endothelium demonstrates robust, multilineage, long-term population of the adult hematopoietic organ. Furthermore, since expression of
kdrl is rapidly extinguished as HSCs arise from hemogenic endothelium (), and since
kdrl+cmyb+ transitional intermediates were no longer detectable in larval or adult stages (
Supplementary Figure 3), our lineage tracing results support the hypothesis that HSCs no longer arise
de novo following their specification in the embryo.
In summary, our imaging and lineage tracing studies demonstrate that the first HSCs born in the zebrafish embryo arise directly from hemogenic endothelium lining the ventral wall of the DA. These results complement previous studies in the avian
21, amphibian
22, and mammalian embryo
1,2,7 and suggest that the cellular mechanisms of HSC generation have been highly conserved across vertebrate evolution. The finding that HSC development requires transition through a hemogenic endothelial intermediate should aid efforts to instruct HSC formation
in vitro from pluripotent precursors, a necessity for therapies designed to replace the adult blood cell lineages.