|Home | About | Journals | Submit | Contact Us | Français|
The transcription factor GATA-2 plays vital roles in quite diverse developmental programs, including hematopoietic stem cell (HSC) survival and proliferation. We previously identified a vascular endothelial (VE) enhancer that regulates GATA-2 activity in pan-endothelial cells. To more thoroughly define the in vivo regulatory properties of this enhancer, we generated a tamoxifen-inducible Cre transgenic mouse line using the Gata2 VE enhancer (Gata2 VECre) and utilized it to temporally direct tissue-specific conditional loss of Gata2. Here, we report that Gata2 VECre–mediated loss of GATA-2 led to anemia, hemorrhage, and eventual death in edematous embryos. We further determined that the etiology of anemia in conditional Gata2 mutant embryos involved HSC loss in the fetal liver, as demonstrated by in vitro colony-forming and immunophenotypic as well as in vivo long-term competitive repopulation experiments. We further documented that the edema and hemorrhage in conditional Gata2 mutant embryos were due to defective lymphatic development. Thus, we unexpectedly discovered that in addition to its contribution to endothelial cell development, the VE enhancer also regulates GATA-2 expression in definitive fetal liver and adult BM HSCs, and that GATA-2 function is required for proper lymphatic vascular development during embryogenesis.
GATA factors belong to an evolutionarily conserved family of C4 zinc finger transcription factors that play demonstrably crucial roles in quite diverse developmental programs, including hematopoietic, urogenital, otic, and neuronal developmental elaboration (1–11). GATA-2 was first demonstrated to be essential for hematopoiesis, as Gata2 homozygous null mutant (Gata2–/–) embryos die at E10.5 during embryogenesis due to a failure in initial blood cell generation (2). Analyses of null mutant embryonic stem cells and their contribution to in vitro differentiation or to definitive hematopoiesis in reconstituted chimeras supported the argument that GATA-2 is required for the proliferation and/or survival of hematopoietic stem cells (HSCs) (12). Recently, it was shown that Gata2 haploinsufficiency resulted in altered integrity of the definitive HSC compartment, leading to a reduction in the number of HSCs by essentially one-half (13, 14).
We previously identified and characterized a Gata2 intron 4 enhancer that conferred reporter gene activity in transgenic mice in both blood and lymphatic endothelial cells (LECs) as well as in poorly characterized subsets of hematopoietic cells (5). Here, we report the generation of conditionally inducible Gata2 vascular endothelial (VE) enhancer–regulated Cre transgenic lines and the consequences of their induced activity in a floxed Gata2 (Gata2fl) genetic background. To circumvent the normal E10.5 demise encountered in Gata2–/– embryos, we utilized a version of Cre recombinase fused to a tamoxifen-sensitive (Tx-sensitive) ligand-binding domain of the estrogen receptor (CreERT2, ref. 15). This strategy allowed us to administer Tx, thereby activating Cre and inactivating the Gata2fl allele, after the time when embryos would normally encounter the first lethal block in primitive erythropoiesis. Analyses of Tx-treated doubly transgenic Gata2 compound mutant (TgVE:Gata2–/fl) embryos revealed that the VE enhancer harbors what we believe to be a previously undiscovered Gata2 transcriptional regulatory activity in definitive (fetal and adult) HSCs, and unexpectedly that GATA-2 deficiency in the endothelial lineage results in edema and hemorrhage, leading to late gestational lethality. Histological examination of Tx-treated TgVE:Gata2–/fl embryos revealed blood pooling in the lymphatic vasculature due to failed lymphatic-venous abscission. Thus, these data show that the Gata2-null mutant animals that survive the initial primitive erythroid embryonic lethal deficiency normally encountered at E10.5 subsequently die of LEC deficiency at E16.5, thus providing the first evidence to our knowledge that GATA-2 is required for an essential endothelial (either autonomous or non-autonomous) function.
Recently, several groups reported that human patients with GATA2 mutations presented symptoms of primary lymphedema (16–18), underscoring the vital role of transcription factor GATA-2 in lymphatic development. Through analysis of the conditionally mutant Gata2 mice described here, we hope to further elucidate the role of GATA-2 in lymphangiogenesis.
We generated transgenic mice bearing integrated copies of the VECreERT2 transgene in which the Gata2 VE enhancer independently directed the transcription of an inducible Cre recombinase or the fluorescent mCherry (mCh) reporter gene (Figure (Figure1A1A and ref. 19). Of the transgenic founder animals that were determined by PCR to harbor both transgenes (11 of 44; hereafter referred to as TgVE), some (7 of 11) stably transmitted both. Their progeny (F2–F5 generations) were used for Cre transgene copy number (and other, see below) analyses, and they ranged from 5 to 47 (using actin as a normalization control) (ref. 20 and Figure Figure11B).
When we examined the expression of the co-integrated transgenes, we found that robust mCh epifluorescence in an endothelia-restricted pattern was detected in E10.5 embryos from 3 lines (56, 62, and 73), while other lines (60, 457, and 473) exhibited no or only very faint vascular mCh fluorescence (Figure (Figure1C1C and data not shown). Furthermore, in the 3 lines that displayed robust mCh staining, Cre mRNA (normalized to Flk1 mRNA) in the vascularized heart and kidneys of P0–P2 transgenic pups was easily detectable by RT-qPCR in TgVE56 and TgVE62 but not in TgVE73 mice (Figure (Figure1B).1B). Therefore, only TgVE56 and TgVE62 lines were used for subsequent studies.
To determine whether the vascular expression pattern of the mCh reporter gene reflected genuine GATA-2 expression, we bred TgVE62 to a Gata2-eGFP knock-in mouse (21). As depicted in Figure Figure1D,1D, coincident mCh transgenic expression precisely reflected eGFP fluorescence produced from the endogenous Gata2 locus, while eGFP alone was additionally (and exclusively) expressed in a cluster of deep ventral midbrain neurons that could be visualized in the same fields (22, 23). Finally, we asked whether mCh fluorescence matched the tissue specificity of a known vascular endothelia-specific gene, the Tie2 tyrosine kinase (24). As shown in Figure Figure1E,1E, the transgenic mCh expression pattern in endothelial cells was indistinguishable from that of GFP directed by the Tie2 gene promoter and enhancer (24) in the intersomitic vasculature.
We further investigated the inducibility and tissue specificity of CreERT2 activity in these two lines by examining transgene activity in the ROSA26 reporter R26R strain, in which the lacZ gene, flanked by transcriptional termination sites and loxP sequences, is integrated at the ROSA26 locus (25). In this configuration, Cre recombinase activity is reported by β-gal expression only in those tissues in which Cre becomes activated. Without Tx, we could detect no X-gal staining in the doubly transgenic TgVE:R26R embryo or a littermate R26R embryo (Supplemental Figure 1A; supplemental material available online with this article; doi: 10.1172/JCI61619DS1). However, RT-qPCR analyses of total RNA isolated from their respective yolk sacs confirmed that Cre (as well as mCh, normalized to Flk1 mRNA) was expressed only in the doubly transgenic TgVE:R26R embryo but not in its littermate R26R embryo (Supplemental Figure 1B). In the presence of Tx (administered by daily gavage of pregnant dams from E9 to E11 during embryogenesis), the TgVE:R26R embryo, unlike its control R26R sibling (Figure (Figure2A),2A), displayed strong X-gal staining in the blood vasculature, mirroring mCh expression (Figure (Figure1C).1C). Thus, we concluded that the Cre activity in TgVE is stringently ligand dependent and is functionally expressed in a tissue-specific manner, as anticipated.
When we examined transverse sections of the E11.5 TgVE:R26R embryo (Figure (Figure2A),2A), X-gal accumulation was observed in the endocardium and the endocardial cushions of the developing heart (Figure (Figure2B),2B), as well as in the cardinal vein (Figure (Figure2C),2C), from which a subset of endothelial cells first gain competence to become LECs as they begin to express Prox1 at E9.5 of embryogenesis (26). LECs then delaminate and migrate dorsolaterally to eventually give rise to the first bona fide lymphatic structure, the jugular lymph sac. Further remodeling and elaboration of the embryonic lymph sacs eventually give rise to the mature lymphatic vascular system that is responsible for modulating fluid homeostasis (27).
More interestingly, X-gal–stained individual rounded cells that appeared to be budding from distended endothelial cells into the lumen of the dorsal aorta were routinely observed (Figure (Figure2,2, D and E, arrowheads) and also appeared in clusters (Figure (Figure2,2, E and F, arrows). When we examined serial transverse sections along the anteroposterior length of Tx-treated TgVE:R26R embryos (n = 3; >80 serial sections per embryo), we found that 70% ± 6% of hematopoietic clusters stained with X-gal. Such rounded cells budding into the dorsal aorta at this stage have been reported to include HSCs (28, 29). Within the fetal liver (FL), which is the major hematopoietic organ during mid-embryogenesis, we also observed discrete cells displaying β-gal activity (Figure (Figure2G).2G). We were able to validate the previous characterization of the Gata2 VE element as a vascular endothelial enhancer (5) using this independent strategy, while at the same time discovering that the enhancer might also bear very early hematopoietic lineage specificity.
To functionally assess the activity of the VE enhancer, we bred the CreERT2 lines in order to conditionally ablate the floxed Gata2 (Gata2fl) allele by removing its fifth exon, which encodes the primary DNA-binding zinc finger (10). In the absence of Tx administration, progeny from these intercrosses were recovered in a normal Mendelian distribution. In particular, Gata2–/fl and TgVE:Gata2–/fl sibling animals from litters that were not exposed to Tx were healthy and lived a normal life span. In order to avoid purported effects of Tx administration or Cre expression on embryonic survivability, we tested the progeny of 15 litters from TgVE:Gata2+/– × Gata2+/fl intercrosses, with and without different concentrations of Tx, to determine: (a) whether the Cre (or theoretically, the mCh) transgene “poisoned” normal Mendelian survival among the progeny; and (b) the precise dosage of Tx and method of administration that led to robust Cre activation accompanied by no (unusual) embryonic death. Since constitutive loss of GATA-2 results in early embryonic lethality at about E10.5 from a failure to properly execute primitive hematopoiesis (2, 3), Tx was administered after primitive hematopoiesis (beginning between E7 and E7.5) was already in process. Tx was administered to pregnant dams from Gata2+/fl × TgVE:Gata2+/– intercrosses beginning on E9 for 3 consecutive days by oral gavage. (Previous studies showed that when Tx was administered to pregnant dams, there was a lag time of 12 hours before levels peaked in the serum of treated animals [ref. 30].) We finally employed an experimentally determined optimal Tx dose of 0.07 mg/g maternal body weight.
Of 89 Tx-treated embryos recovered from 10 litters that were analyzed between gestational days 13.5 and 15.5, TgVE:Gata2–/fl embryos were recovered at the anticipated Mendelian ratios (Table (Table1).1). However, we noticed that compared with their littermates, Tx-treated TgVE:Gata2–/fl embryos and yolk sacs appeared to be anemic, with death occurring by E16.5 (see below). We were curious to determine whether the observed pale appearance of the Tx-treated TgVE:Gata2–/fl embryos might be attributable to disruption of a function for GATA-2 in erythropoiesis, since Grass et al (31) independently identified the same Gata2 enhancer and reported that the VE element was able to potentiate reporter gene expression in transfected murine erythroleukemia cells. Alternatively, the VE enhancer could regulate Gata2 transcription at some very early stage of hematopoiesis. Our earlier observations, of rounded X-gal–stained cells apparently budding from the dorsal aorta (Figure (Figure2,2, D–F), lend credence to the latter hypothesis.
To delve more deeply into the etiology of the observed anemia in these embryos, we examined Tx-treated FL cells from E14.5 embryos in triplicate in colony-forming assays in order to determine their ability to generate CFU of either erythroid (burst-forming unit–erythroid [BFU-E]), myeloid (granulocytic, macrophage, or both [CFU-G, CFU-M, or CFU-GM]), or mixed (CFU-GEMM) lineages. Compared with Tx-treated E14.5 wild-type control embryos (Gata2+/+ with or without TgVE, n = 11), the total FL cell number in Tx-treated heterozygotes (Gata2+/–, TgVE:Gata2+/fl, n = 13) and TgVE:Gata2–/fl embryos (n = 6) was correspondingly reduced with increasing allelic loss of GATA-2 (Figure (Figure3A).3A). In the latter group of embryos, total FL cell number was reduced to 19% of that in wild-type embryos.
FL cells from these Tx-treated embryos were then seeded into methylcellulose medium supplemented with growth factors (M3434, Stem Cell Technologies). Gata2 heterozygotes (Gata2+/–, TgVE:Gata2+/fl, n = 13) and TgVE:Gata2–/fl embryos (n = 6) also generated lower numbers of erythroid, myeloid, and mixed-lineage colonies (Figure (Figure3B).3B). Notably, Tx-treated TgVE:Gata2–/fl FL cells failed to generate BFU-E, CFU-G/M/GM, or CFU-GEMM colonies (Figure (Figure3B),3B), initially suggesting that a very early definitive hematopoietic progenitor compartment was severely compromised in Tx-treated TgVE:Gata2–/fl embryos.
Successful Cre-mediated excision of the Gata2 floxed allele was determined at the RNA and genomic DNA levels. Remaining GATA-2 transcript abundance (if any) in Tx-treated FL samples that were used for CFU assays was examined by RT-qPCR. In TgVE:Gata2–/fl FL cells, Gata2 mRNA (relative to endogenous Hprt mRNA) was reduced to 15% of that in wild-type embryos (Figure (Figure3C).3C). We also quantified Cre-mediated excision of the floxed Gata2 allele using qPCR assay to detect the presence of exon 5 (the one flanked by loxP sites in the floxed Gata2 allele, normalized to Actin) in genomic DNAs extracted from individual colonies isolated from Tx-treated TgVE:Gata2+/fl methylcellulose cultures. Of the 35 TgVE:Gata2+/fl individual colonies analyzed, Gata2 exon 5 was detected at 58% ± 6% relative to the wild-type control. Hence, the calculated excision efficiency of the single floxed Gata2 allele achieved was 84% (Figure (Figure3D3D and Supplemental Figure 2). Together, our results confirmed that the in vivo induction of CreERT2 activity led to effective Gata2 inactivation in TgVE:Gata2–/fl embryos.
Since the in vitro colony-forming assays could not be used to pinpoint where the defect was in the affected progenitor population (i.e., in multipotent hematopoietic progenitors or even more primitive HSCs), more precise and quantitative hematopoietic lineage analysis was undertaken using flow cytometry. To investigate in greater depth the hypothesis that the VE enhancer regulates Cre-mediated inactivation of Gata2, thereby ablating a very early hematopoietic progenitor compartment, we isolated E14.5 FLs from embryos that had been administered Tx daily from E9 through E11 and then performed flow cytometry to determine the immunophenotype corresponding to different early progenitor populations. The combinatorial analysis of a subset of Slam family receptors (CD150 and CD48) in conjunction with conventional HSC surface markers (mature hematopoietic lineage–negative, Sca-1–positive, and c-Kit–positive [LSK]) facilitates the identification of a highly enriched HSC cellular fraction in which 1 in every 2.1 recovered cells is capable of conferring long-term multilineage reconstitution in lethally irradiated host animals (32, 33).
We found that Tx-treated embryos bearing a single Gata2 null allele (TgVE:Gata2+/–, TgVE:Gata2+/fl, Gata2–/fl; n = 9) had reduced LSK and LSKS (lin–Sca-1+c-Kit+CD150+CD48–) cells when compared with wild-type or pseudo-wild-type (Gata2+/+, TgVE:Gata2+/+, Gata2+/fl; n = 6) embryos (Figure (Figure4).4). The total number of HSCs recovered from Tx-treated embryos bearing two wild-type alleles was between 2.2 × 104 to 5.7 × 104 (LSK) and 1.2 × 103 to 4.6 × 103 (LSKS), whereas only 1.3 × 104 to 4.0 × 104 (LSK) or 0.5 × 103 to 1.6 × 103 (LSKS) cells were recovered from Tx-treated embryos bearing a single Gata2 null allele. These data mirrored previous reports showing that Gata2 haploinsufficiency results in compromised HSC homeostasis (13, 14). Intriguingly, we found that when TgVE:Gata2–/fl embryos (n = 4) were administered Tx between E9 and E11, both the LSK and the more highly HSC-enriched LSKS populations displayed an exponential decline (Figure (Figure4,4, C–E). A small number of LSK cells (i.e., 98–477) were recovered from Tx-treated TgVE:Gata2–/fl (Figure (Figure4C;4C; n = 4) embryos, and likewise a minuscule number of LSKS cells were recovered from these FLs (Figure (Figure4D).4D). We obtained the same results when Tx was administered later during gestation (between E11 and E13), at a time during embryogenesis when definitive hematopoiesis was already fully engaged (Figure (Figure4B).4B). This is consistent with the observation that in embryos hemizygous for an active TgVE transgene, a majority of both LSK and LSKS cells were mCh+ (64% ± 4% and 87% ± 6%, respectively, n = 7; Figure Figure4E4E and data not shown). We conclude that the loss of GATA-2 activity in those tissues regulated by the Gata2 VE enhancer results in a virtually complete loss of HSCs in the most highly refined fraction of fetal HSCs, revealing for the first time to our knowledge the localization of a transcriptional regulatory element that is capable of controlling HSC activity of the Gata2 gene.
To determine whether the Gata2 VE enhancer displays stage-specific differential activity among definitive HSCs, we asked whether the TgVE transgene would ablate Gata2 expression in adult BM HSCs as it did in FL HSCs, distinct but closely related definitive HSC populations (34). Gata2–/fl and Gata2+/fl adult mice (without or with the TgVE, from 8 to 12 weeks of age) were gavaged with Tx (5 mg/d) for 5 consecutive days. Two days after the final Tx administration, BM cells were harvested and analyzed by flow cytometry. Once again, we employed the immunophenotype defined by Kiel et al. (32) to define the most enriched definitive HSC population (LSKS cells).
In comparing the absolute LSK and LSKS cell numbers between TgVE:Gata2+/fl and TgVE:Gata2–/fl adult mice that had not been administered Tx (Figure (Figure5B5B and data not shown) and between Gata2+/fl and Gata2–/fl mice exposed to Tx (Figure (Figure5B5B and data not shown), we again noted that reduced Gata2 gene dosage correlated perfectly with diminished LSKS cell number. This is reminiscent of the same phenomenon observed in E14.5 FL HSCs (Figure (Figure4)4) that was also previously reported by others (14). When Gata2+/fl and Gata2–/fl mice bearing the TgVE were exposed to Tx, we saw a further precipitous decline in the LSKS population as a consequence of reduced GATA-2 activity (Figure (Figure5,5, A–C). Explicitly, in Tx-treated adult TgVE:Gata2–/fl mice, the LSKS population was depleted by approximately 2 orders of magnitude within a week after the first Tx exposure (Figure (Figure5,5, B and C). Thus, the Gata2 VE enhancer is functional in both FL and adult BM definitive HSCs, and loss of GATA-2 adversely impacts HSC survival and/or proliferation.
To functionally detect HSCs in Tx-treated mice, we performed competitive repopulation assays. Total CD45.2 FL cells (5 × 105) from control (Gata2+/fl [n = 1] or TgVE:Gata2+/+ [n = 3]) or from TgVE:Gata2–/fl (n = 2) E14.5 embryos, all of which had been exposed to Tx from E9 to E11, were coinjected at a 1:1 ratio with wild-type CD45.1 adult BM cells (5 × 105) into lethally irradiated CD45.1 animals (5 mice per group; Figure Figure6A).6A). Peripheral blood reconstitution from CD45.2 donor FL-derived cells was monitored from week 4 to 16 after transplantation. In all animals receiving Gata2+/fl (n = 5) or TgVE:Gata2+/+ (n = 15) FL cell transplants, the FL-derived donor cells contributed to the myeloid (>80%), B (>80%), and T (>60%) lymphoid lineages in the peripheral blood (Figure (Figure6B6B and Supplemental Figure 3A). In contrast, in transplant recipients injected with Tx-treated TgVE:Gata2–/fl FL cells, only a weak multilineage reconstitution was observed in most animals (8 of 10), while some level (<30%) of multilineage reconstitution was seen in the remaining two animals (Figure (Figure6B,6B, Supplemental Figure 3A, and data not shown).
When we analyzed the highly HSC-enriched LSKS fraction in the BM of transplant recipients, we found that in control animals, CD45.2 FL–derived LSKS constituted the major population (60%–80%; Figure Figure6,6, C and D). In contrast, 8 of 10 animals that received TgVE:Gata2–/fl FL cells had more than 90% of the LSKS fraction derived from wild-type CD45.1 BM cells. The two remaining transplant animals, in which some CD45.2 progenitor cells were detected in the peripheral blood (Figure (Figure6B6B and Supplemental Figure 3A), had some (20% and 40%) CD45.2 FL-derived LSKS in their BM (Figure (Figure6,6, C and D, Supplemental Figure 3B). When we analyzed FACS-sorted CD45.2-positive LSK cells from the latter two animals by qPCR, we found that these cells indeed contained an intact Gata2fl allele (data not shown), indicating that approximately 20% of LSKS cells per TgVE:Gata2–/fl embryo had escaped VECre-mediated Gata2 inactivation, in good agreement with the efficiency of conditional GATA-2 loss (Figure (Figure33).
Within the adult hematopoietic cell hierarchy, we noted that mCh epifluorescence was absent in mature lymphoid, myeloid, and erythroid lineages isolated from the BM, spleen, or thymi of wild-type adult mice harboring TgVE (Figure (Figure7),7), in stark contrast to endogenous GATA-2 expression in the erythroid and myeloid compartments (35, 36). Most conspicuously, mCh expression was almost entirely restricted to the BM LSK fraction that contained all stem cells. These additional data further underscored the highly restricted and early-acting (within hematopoietic lineages) activity of the Gata2 VE enhancer within the hematopoietic cell hierarchy. Thus, these data demonstrate that induction of VE-regulated Cre leads to functional depletion of HSCs and that the Gata2 VE element possesses HSC- and early progenitor–restricted enhancer activity.
We noted that live TgVE:Gata2–/fl embryos began to appear edematous by E13.5, and they invariably developed severe subcutaneous edema, hemorrhage, and anemia and died by E16.5 (Figure (Figure8A,8A, arrows). The identical phenotype was encountered when we employed a second, independent conditional Cre transgenic line (TgVE56; Figure Figure1B1B and legend to Figure Figure8).8). Importantly, we did not observe these phenotypes in Tx-treated embryos that did not bear a Cre transgene, and therefore we were confident that the phenotypes were not due to Tx toxicity; the acquisition of these phenotypes correlated only with Cre induction in the embryos that co-inherited Gata2 null and floxed alleles. Further experiments performed with only a single Tx delivery on day 9 after fertilization or after 3 consecutive daily doses from E11 to E13, instead of the generally employed Tx regimen of 3 consecutive daily doses from E9 to E11, generated the identical phenotypes of anemia, edema, and hemorrhaging in TgVE:Gata2–/fl embryos. PCR genotyping of yolk sac genomic DNAs confirmed the conversion of the Gata2 floxed allele to the Gata2 exon 5–deleted allele (Gata2Δ, Figure Figure8B).8B). The Gata2Δ amplicon was detected exclusively in embryos bearing both the Gata2 floxed allele and TgVE from dams that had been treated with Tx. Since CreERT2 expression is restricted to endothelial cells in the yolk sac, the (unaltered) Gata2 floxed allele was still detected in total yolk sac genomic DNA of Tx-treated TgVE:Gata2–/fl embryos (Figure (Figure8B).8B). Thus, the final phenotype revealed by employing this binary conditional inactivation strategy was that loss of GATA-2 in the endothelial lineage leads to massive mid-gestational hemorrhage, edema, and anemia, followed shortly thereafter by embryonic demise by around E16.5.
We noted massive blood pooling in the vicinity of the jugular lymph sacs in E14.5 Tx-treated mutant, but not control, embryos (Figure (Figure9,9, A and B). Detailed examination of serial histological sections of E13.5 Tx-treated (from E9 to E11) embryos revealed that in the control embryo, the jugular lymph sac was distinctly separated from the jugular vein, while in the mutant only a single vessel lumen could be found (Figure (Figure9,9, C–E and I). When we examined serial sections of TgVE:Gata2–/fl embryos more anteriorly and posteriorly, we saw that the lymph sacs appeared hypoplastic compared with that of the control embryo (Figure (Figure9,9, E and F). Additionally, erythrocytes were aberrantly present in the lymph sacs (Figure (Figure9,9, G and H) of the E13.5 TgVE:Gata2–/fl mutant embryo, which typically had not presented visibly outward signs of gross hemorrhage by this gestational day (Figure (Figure8).8). Thus, employing this binary conditional inactivation strategy revealed yet another striking phenotype: GATA-2 function is required for proper lymphatic morphogenesis during embryonic development.
The study of Gata2 transcriptional regulation in hematopoiesis has a long and storied history. More than a decade ago, we discovered that a 271-kbp Gata2 yeast artificial chromosome (YAC) transgene, which circumscribed genomic sequences from –198 kbp to +73 kbp (with respect to the translational start site) of the Gata2 locus, could fully rescue the embryonic lethality in Gata2–/– embryos (3). Hence, this YAC functionally defined the genomic limits of the Gata2 locus that contained all of the regulatory element(s) required for fully elaborating primitive and definitive hematopoiesis, but the YAC only incompletely rescued all of the GATA-2 functions, since the P0 pups died from incomplete patterning of the urogenital system (3, 9, 20).
In contemporaneous studies, Kobayashi-Osaki et al. (37) reported that the Gata2 GATA-2–EHRD element, located 3.1 kbp 5′ to the Gata2 1S first exon, was capable of directing GFP reporter gene expression within the para-aortic splanchnopleura, the immediate developmental precursor of the aorta-gonad-mesonephros (AGM), from which definitive HSCs first emerge in mice (38). This promoter-proximal element was later shown to modulate GATA-2 abundance in E12.5 FL HSCs and its immature progenitor populations, but not in adult HSCs (39).
Suspicions regarding the potential complexity of Gata2 hematopoietic transcriptional regulation were underscored when Bresnick and colleagues identified 5 GATA motif–bearing regions (–77, –3.9, –2.8, –1.8, and +9.5 relative to the Gata2 1S transcription start site) that are bound by GATA-2 but are replaced by GATA-1 during erythroid differentiation (40), suggesting that each of these “GATA switch” sites might also play roles in regulating Gata2 expression during hematopoiesis. When either the –1.8 or –2.8 site was systemically deleted from the germ line, GATA-2 transcripts were altered in HSCs, but this did not translate into a detectable change in HSC number (39, 41); this implies that other regulatory element(s) within the Gata2 locus are responsible for its HSC activity.
In our continuing efforts to elucidate the transcriptional regulation of GATA-2, we have in the past examined reporter gene expression using large (yeast and bacterial artificial chromosomes) and small (plasmid) expression vectors in transgenic studies. As a result of these analyses, we previously identified and characterized an intronic Gata2 VE enhancer predominantly as an endothelial regulatory element using lacZ-tagged plasmid expression vectors in founder transgenic studies (5). The data presented here demonstrate that the Gata2 VE enhancer is transcriptionally active in both FL and adult BM HSCs and that VE-regulated conditional Gata2 loss leads to massive depletion of the most highly purified immunophenotypic fetal and adult definitive HSCs. On closer scrutiny, we noted that VE-regulated mCh expression co-segregated with most (87% ± 6%) of the most highly purified fraction of TgVE-bearing wild-type E14.5 FL LSKS HSCs (Figure (Figure4E),4E), while only about 60% of the TgVE-bearing wild-type LSKS population recovered from adult BM expressed mCh (Figure (Figure5C).5C). This difference in mCh expression in the most highly purified fraction of FL versus adult BM HSCs could reflect the differential activity of the Gata2 VE enhancer in the HSC cellular milieu at different developmental stages; alternatively, the data could as easily be interpreted to support the hypothesis that the adult BM HSC compartment is intrinsically more heterogeneous than the fetal compartment, a contention strongly supported by recent literature (42–44). The data could also be explained simply as a consequence of diminished levels of mCh expression in adult LSKS HSC or altered stability of the protein in (perhaps longer-lived) adult LSKS cells.
The constitutive loss of GATA-2 that leads to early embryonic lethality by E10.5 has been attributed to primitive hematopoietic failure (2, 3). Further insight into a role for GATA-2 in definitive hematopoiesis was gleaned from the examination of null mutant embryonic stem cells and their contribution to in vitro differentiation or to definitive hematopoiesis in reconstituted chimeras, where it was deductively concluded that GATA-2 is required for the proliferation and/or survival of HSCs (12). More recent analyses of Gata2 heterozygous mutant mice clearly demonstrated a quantitative reduction in HSC number and compromised HSC function in competitive BM reconstitution experiments (13, 14). Here, using a combinatorial conditional loss-of-function strategy, we were able to directly and formally demonstrate that GATA-2 is vital for immunophenotypic as well as functional HSC recovery from both mid-gestational embryos and adults. In light of the negligible effects of the –1.8 and –2.8 kb promoter Gata2 mutations on HSCs (above), it will be interesting to determine the consequence of deleting the Gata2 VE enhancer singly or in combination with other putative Gata2 regulatory elements to ascertain their individual and combinatorial contribution to Gata2 transcription in HSCs and, in particular, whether the VE HSC activity described here is a unique or redundant function within the locus.
In our previous and in current studies, we noted the presence of X-gal–stained hematopoietic cells that appeared to be budding into the lumen of the dorsal aorta in E10.5 embryos, possibly hinting that this enhancer might be active in hemogenic endothelium or hemangioblasts, the presumptive progenitor cell common to the endothelial and hematopoietic lineages. Indeed, using blast colony–forming cells derived from embryonic stem cells, Lugus et al. showed that temporally sensitive, conditionally forced induction of GATA-2 led to a transient expansion of hemangioblasts in cell culture (45). Whether it is required for the development of an even earlier-stage, common progenitor cell (30, 45) will require additional study.
While GATA-2 appears not to be required for endothelial patterning into the initial primary vasculature (as evident in E10.5 Gata2 homozygous null embryos), it may be necessary for the elaboration of the mature vasculature later during embryogenesis. At the time of embryonic demise at E16.5, the binary conditionally inactivated Gata2 mutant mice displayed severe subcutaneous edema and extensive hemorrhage. We show that the VE is active in the cardinal vein, where endothelial precursor cells with potential to differentiate into LECs are generated (26), and in nascent LECs that have sprouted from the cardinal vein (5). Robust GATA-2 expression has also been reported in E16.5 and adult murine lymphatic vessel and lymphatic valves (17). The abnormal accumulation of blood in the jugular lymph sacs in Tx-treated TgVE:Gata2–/fl embryos prompted us to histologically examine their embryonic lymphatic vasculature. Curiously, we found that in Tx-treated E13.5 TgVE:Gata2–/fl embryos, the lymphatic and blood vascular systems remained aberrantly connected, unlike in control embryos. Interestingly, there have been several recent reports of primary lymphedema in patients with either Emberger or monocytopenia and mycobacterial infection (MonoMAC) syndrome due to GATA2 mutations (16–18).
Much progress has been made over the last decade with regard to understanding embryonic lymphatic vascular development, and numerous genes that impact this morphogenic process have now been identified (46). In particular, the blood-lymphatic mixing phenotype had been reported in mice with mutations that affect podoplanin (47), Ruxn1 (26), Meis1 (48), Syk (49), Slp76 (50), and Plcg2 (51). A number of these mutants have platelet deficiency or dysfunction, as there is accumulating evidence that platelets play a central role in lympho-venous partitioning (46). Whether this pathophysiological developmental aberration in Tx-treated TgVE:Gata2–/fl embryos is due to loss of GATA-2 expression in the hematopoietic and/or lymphatic endothelial lineage will require further investigation. It will be of interest to now investigate the role and mechanism(s) by which GATA-2 intersects with the developmental programs specifying lymphangiogenic differentiation.
The lacZ insert in pTKSXβ (5), containing a 1.2-kbp SfiI/XbaI Gata2 intron 4 enhancer fragment in pTKβ vector (Clontech), was removed by NotI digestion and replaced with a Klenow-treated EcoR1 CreERT2 fragment (52) or a Klenow-treated BamHI/NotI fragment encoding the fluorescent reporter gene mCh (Clontech). Both resultant plasmids were sequentially treated with PvuII/SphI and Klenow polymerase. The excised inserts were separately ligated into PacI-digested pIV vector, which contained duplicated 1.2-kbp chicken HS4 insulators (53) sequentially cloned into the polylinker (SmaI and PmeI) sites of pNEB193 (NEB). The VECre and VEmCherry expression plasmids were verified by sequencing during and at the completion of construction.
Trangenic mice were generated by the the Transgenic Animal Model Core at the University of Michigan Medical School. Plasmids VECre and VEmCherry were digested with FspI, purified, and microinjected (1:1 ratio) into (C57BL/6 × SJL) F1 fertilized eggs. Transgenic animals were identified by PCR detection of Cre and mCh (Supplemental Table 1) using genomic DNAs isolated from tail snips as template.
For mouse lines that stably transmitted both Cre and mCh transgenes (referred to as TgVE lines), F2–F5 progeny were examined for Cre transgene copy number, Cre mRNA levels, and mCh fluorescence. For Cre transgene copy number analysis, tail snip genomic DNAs were used as template in SYBR Green qPCR as described previously (20) using endogenous Actin as a normalization control. Genomic DNAs from mice that were wild-type, heterozygous, or homozygous for the AyuI-Cre knock-in allele (21) were included as copy number controls for the Cre qPCR assays (data not shown). For determining Cre transcript levels, total RNAs were isolated from vascularized tissues (heart and kidney) of neonatal pups that were Cre transgene positive. Reverse transcription and SYBR Green qPCR were performed as previously detailed (54), and Cre mRNA abundance was expressed relative to endothelia-specific Flk1 mRNA. All primer pairs used for qPCR are listed in Supplemental Table 1.
Selected TgVE lines (62 and 56) that coexpressed intense endothelia-specific mCh fluorescence and robust Cre mRNAs were independently mated with R26R (25), TgTie2.GFP transgenic (24), or Gata2+/gfp (21) mice (in which a gfp gene was inserted at the Gata2 translational start site) to generate E10.5–E12.5 compound transgenic embryos. Embryos from the first intercross were isolated from Tx-gavaged pregnant females and processed for X-gal staining and then subjected to whole-mount photography or cryosectioning and counterstaining with nuclear fast red (5). E10.5 TgVE:TgTie2.gfp and TgVE:Gata2+/gfp embryos were examined by fluorescence confocal microscopy. Genomic DNAs from yolk sacs or embryonic tails were used for PCR genotyping (21, 24, 25).
TgVE56 and TgVE62 lines (of mixed background) were also bred to CD1 outbred Gata2+/– mice (2) to generate compound TgVE:Gata2+/– animals. Timed matings between TgVE:GATA2+/– and Gata2+/fl (10) animals were carried out. The morning of vaginal plug detection was defined as E0.5. Pregnant dams were gavaged daily with Tx (Sigma-Aldrich; 0.07 mg/g body weight) from E9 to E11. This dosage was experimentally determined to result in minimal embryonic lethality due to Tx toxicity (55). Embryos were isolated from Tx-treated pregnant dams on the gestational days described and were fixed in 4% paraformaldehyde and then digitally photographed as whole mounts, or gradually cleared in benzyl benzoate:benzyl alcohol before whole-mount photography. Alternatively, embryos were processed for paraffin embedding followed by H&E staining. Genomic DNAs of embryos isolated from yolk sacs were genotyped by PCR (Supplemental Table 1).
Embryos (E14.5) were harvested from Tx-treated pregnant dams. Yolk sacs were used for PCR genotyping. Individual FLs were isolated and triturated to generate single-cell suspensions. The cells were then filtered through cell strainers before being stained with unconjugated monoclonal antibodies (B220, CD3e, CD4, CD5, CD8a, Gr-1, and TER-119) that recognize mature hematopoietic lineages as well as directly conjugated antibodies (c-Kit, Sca-1, CD150, CD48), thus allowing the immunophenotyping of HSC-enriched fractions (56). Antibodies were purchased from Biolegend and eBioscience. Note that anti-Mac1 antibody was excluded from the lin– cocktail, as FL HSCs are Mac1lo (57).
For ascertaining the effects of conditionally inactivating Gata2 in the adult BM, 8- to 12-week-old mice (of mixed background) were gavaged with Tx (5 mg/day) for 5 consecutive days, and their BM cells were isolated for flow cytometric analyses 48 hours after the final Tx administration. BM cells were subjected to rbc lysis and then filtered through cell strainers, followed by magnetic depletion of mature hematopoietic lineage cells. The resultant lineage-negative (lin–) cells were then stained with various antibodies prior to their staging by flow cytometry as described previously (58).
All flow studies were performed using a FACSAria (BD). The resulting cell distribution files were analyzed using FACSDiva (BD) or FlowJo (Tree Star Inc.) software. The statistical significance of differences between two groups of animals (as represented by the P value) was determined using Student’s t test.
For methylcellulose colony-forming assays, single-cell suspensions were prepared from individual Tx-treated control and TgVE:Gata2–/fl E14.5 FLs. Unfractionated FL cells from individual embryos were seeded in triplicate in methylcellulose supplemented with various cytokines optimized to support the growth and differentiation of hematopoietic progenitors (M3434, Stem Cell Technologies). For control embryos, 2 × 104 total FL cells were seeded per plate. For TgVE:Gata2–/fl embryos, cells were seeded at multiple concentrations (2 × 104, 5 × 104, and 1 × 105 per plate) in triplicate. Colonies (comprising >30 cells) were scored as mixed (CFU-GEMM), myeloid (CFU-G, CFU-M, and CFU-GM), or erythroid (BFU-E) after 7–12 days of culture according to the manufacturer’s instructions.
TgVE56 and TgVE62 lines (of mixed background) were backcrossed to wild-type C57BL/6J mice for 5–6 generations and then bred to CD57BL/6-Ly5.2 Gata2+/fl mice to generate compound mutant TgVE:Gata2+/fl animals. Timed matings of TgVE:Gata2+/fl × Gata2+/– mice were initiated, and pregnant dams were gavaged daily from E9 to E11 with 0.07 mg/g body weight of Tx (Sigma-Aldrich). Embryos were harvested on gestational day 14.5 and kept in medium on ice while PCR genotyping was performed using yolk sac DNAs. After genotyping (~3 hours later), single-cell suspensions of FLs were prepared from the embryos of desired genotypes. CD57BL/6-Ly5.1 (B6-SJL, CD45.1) mice from 6 to 8 weeks of age were purchased from The Jackson Laboratory and were irradiated as previously described (58) for use as transplant recipients. CD45.2 FL cells (5 × 105) from control and test embryos were isolated from individual E14.5 embryos that had been exposed to Tx during development and were mixed at a 1:1 ratio with an equal number (5 × 105) of competitor CD45.1 adult BM cells. Cell mixtures were injected intravenously into lethally irradiated hosts. Donor chimerism in peripheral blood was determined by flow cytometry at 4, 8, 12, and 16 weeks after transplantation. Animals were considered positive for engraftment if more than 1% of tested cells were in the Gr-1+Mac1+ myeloid cell population. After 16 weeks, transplanted animals were sacrificed, and all of the BM cells recovered from 2 tibias and 2 femurs were processed for immunophenotyping of CD45.1 versus CD45.2 LSKS cells, as described above.
Student’s 1-tailed t test P values less than 0.05 in the pairwise comparisons were considered to be statistically significant.
All animal experiments were carried out under a protocol approved by the University of Michigan IACUC (PRO00001725).
We gratefully acknowledge the helpful advice and assistance of, and spirited discussion with, members of our laboratory, as well as our colleagues Stephen J. Weiss, Sean Morrison, Jun-Lin Guan, and Ivan Maillard for critical evaluation of the manuscript. This work was supported initially by NIH grant R01 GM28896 and subsequently by the G. Carl Huber Professorship and NIH grant R01 AI94642 (to J.D. Engel).
Conflict of interest: The authors have declared that no conflict of interest exists.
Citation for this article: J Clin Invest. 2012;122(10):3705–3717. doi:10.1172/JCI61619.