Multiple, Distant Gata2 Enhancers Specify Temporally and Tissue-Specific Patterning in the Developing Urogenital System

Abstract

Transcription factor GATA-2 is expressed in a complex temporally and tissue-specific pattern within the developing embryo. Loss-of-function studies in the mouse showed that GATA-2 activity is first required during very early hematopoiesis. We subsequently showed that a 271-kbp yeast artificial chromosome (YAC) transgene could fully complement the loss of Gata2 hematopoietic function but that these YAC-rescued Gata2 null mutant mice die perinatally due to defective urogenital development. The rescuing YAC did not display appropriate urogenital expression of Gata2, implying the existence of a urogenital-specific enhancer(s) lying outside the boundaries of this transgene. Here we outline a coupled general strategy for regulatory sequence discovery, linking bioinformatics to functional genomics based on the bacterial artificial chromosome (BAC) libraries used to generate the mouse genome sequence. Exploiting this strategy, we screened >1 Mbp of genomic DNA surrounding Gata2 for urogenital enhancer activity. We found that the spatially and tissue-specific functions for Gata2 in the developing urogenital system are conferred by at least three separate regionally and temporally specific urogenital enhancer elements, two of which reside far 3′ to the Gata2 structural gene. Including the additional enhancers that were discovered using this strategy (called BAC trap) extends the functional realm of the Gata2 locus to greater than 1 Mbp.

In order to generate the enormous diversity of distinct cell types that comprise multicellular organisms, gene expression must be activated or extinguished in different tissues at precise times during development. To achieve this finely regulated control over gene expression, metazoans typically employ regulatory enhancers as independent, discrete functional units to control transcription at specific times during developmental time and space. Given the exquisite specificity of expression patterns displayed by a large number of vital regulatory molecules during development (e.g., cell surface receptors, signal transduction proteins, and transcription factors), the regulatory network responsible for conferring this control must be complex. However, in order to demonstrate epistatic, hypostatic, or parallel relationships in the regulatory circuitry leading to tissue elaboration and function, all of the factors, both cis and trans, that contribute to control over a given transcription unit must be identified and tested in vivo.

The identification of enhancer elements can be problematic as they can be located anywhere in relation to the target gene whose transcription they influence, and enhancers do not typically display characteristics that can be easily identified a priori from primary DNA sequence. For example, unbiased mapping of binding sites for transcription factors Sp1, c-myc, and p53 along chromosomes 21 and 22 showed that 42% of those factor binding sites do not lie in close proximity to a known gene (3). This observation suggests that a large number of regulatory elements may lie considerable distances from the genes that they regulate. A few such distant regulatory elements have in fact been identified, often by chance, either by transgene insertion (1) or as a consequence of natural inversions or other mutations (11), but there are no successful general strategies in common use for routine identification of these critical long-range control elements.

Because of the difficulties inherent in identifying these putative distant enhancers, various approaches using a combination of informatics and functional assays have been developed, with various degrees of success (13, 22, 24). However, current approaches have explicit limitations, either because they involve the use of large transgenes that contain the gene of interest (thus limiting the region to be studied to the size of these transgenes) or they involve testing many smaller candidate regions based on interspecies DNA sequence conservation, a time-consuming and only sporadically successful approach. Because mammalian enhancers may lie more than 1 Mbp (1, 24) 5′ or 3′ to the gene that they regulate, these limitations render such approaches critically deficient.

Studies of the regulation of the transcription factor gene Gata2 exemplify the difficulties inherent in identifying distant regulatory sequences. The GATA family of transcription factors is an evolutionarily conserved family of C₄ zinc finger transcription factors that play demonstrably critical roles in development. There are six members of the GATA family in vertebrates, and they have historically been subdivided into two subfamilies. GATA-1, -2, and -3 are all involved in various aspects of hematopoiesis: erythropoiesis, hematopoietic progenitor longevity, and T lymphopoiesis, respectively, among many other activities (26, 27, 29). Similarly, GATA factors 4, 5, and 6 have been shown to be involved in cardiac, genitourinary, and endodermal development (19-21), among other developmental functions.

GATA-2 was originally cloned from a chicken reticulocyte cDNA library (34) and was later shown to be critical for the proliferation and/or differentiation of early hematopoietic progenitors, as Gata2 null mutant mice die at early to midgestation due to a block in primitive hematopoiesis (29). Further examination of Gata2 gain of function and in vitro differentiation of Gata2 null mutant embryonic stem cells underscored the initial conclusions from the loss-of-function experiments and showed that GATA-2 plays a pivotal role in the proliferation of very early hematopoietic progenitors (2, 8, 30).

Previously, we found that transgenic mice bearing a yeast artificial chromosome (YAC) containing 271 kbp (−198 kbp/+73 kbp relative to the translational start site) surrounding the Gata2 locus are rescued from Gata2 mutant-induced embryonic lethality, demonstrating that this YAC contained all the regulatory information required to direct sufficient Gata2 expression in the hematopoietic compartment. However, compound mutant mice bearing this YAC but lacking endogenous Gata2 died perinatally due to hydroureteronephrosis, which developed coincident with the onset of urine excretion (36). This defect resulted from the failure of the Tg^Gata2YAC:Gata2^−/− compound mutant mice to develop a patent uretero-vesicular connection, causing the mice to develop cystic, nonfunctional kidneys, resulting in the observed perinatal lethality. This ureteric phenotype is accompanied by a broad spectrum of defects in the development of the hindgut and either the uterus or vas deferens, suggesting that Gata2 probably participates in even broader aspects of cloacal development than previously suspected.

Although Gata2 is normally expressed broadly in the urogenital system, analysis of the lacZ-tagged version of the 271-kbp Gata2 YAC transgene showed that it is unable to drive expression in any urogenital structures (36). These data suggested that the enhancer(s) required for the patterning of Gata2 in the developing urogenital system must lie outside of the boundaries spanned by the YAC, and thus the absence of this missing enhancer from the transgene was the developmental origin of the lethal urogenital phenotype observed in the partially rescued compound mutant mice.

In order to identify the regulatory sequence(s) that controls Gata2 expression in the developing urogenital system, we developed assays that combine functional and bioinformatics approaches. By using a bacterial artificial chromosome (BAC) transgenic assay designed to solve some of the previously noted limitations for regulatory element identification, we identified one BAC within an initial survey of 1 Mbp flanking the Gata2 locus that contained urogenital enhancer activity. Next, using comparative sequence conservation data, we then identified two separate enhancers that control different tissue and spatial domains of Gata2 activation within that BAC. Further refinement of the BAC transgenic data imply the existence of yet another urogenital regulatory element, which deductively must lie outside of the domain tested, and which is responsible for a third, separate tissue- and spatially restricted domain of Gata2 urogenital expression. The summary of these data indicates that GATA-2 plays a complex, multifaceted role in the development of the urogenital system, each aspect of which is regulated by a separate DNA element controlling the time and the specific cell type that will lead to proper extension of the nephric duct toward the cloaca, the outgrowth of the ureter, and its subsequent connection to the bladder.

MATERIALS AND METHODS

BAC identification.

We first identified BACs spanning the mouse Gata2 locus by filter hybridization using an RPCI-23 mouse genomic BAC library (25). When physical map data for the mouse genome became available (7), BACs were identified using the ENSEMBL website (http://www.ensembl.org/Mus_musculus/).BAC DNA was purified by standard alkaline miniprep from 5 ml of bacterial culture and was verified by restriction digestion with HindIII, followed by gel electrophoresis at 4°C (15).

BAC ends.

The BACs used in this study began and ended (5′/3′) at the following nucleotides (±0.5 kbp) relative to the Gata2 translation initiation site: 333I12, +60/+260; 272C07, +143/+390; 49L08, +230/+430; 97F04, +427/+560; and 356J11, +550/+787. The d16 YAC used in the original rescue experiments (36) spanned the Gata2 locus from kbp −198 to +73. The relative positions of these DNAs within the Gata2 locus are depicted by the diagram in Fig. 1A.

FIG. 1.

Strategy to identify Gata2 urogenital enhancer. (A) Diagram of Gata2 in its genomic position on mouse chromosome 6. Constructs referred to in the study are diagrammed. A list of end points of the BACs used in the study is available in Table 1. Other genes in the locus are denoted by black boxes. (B) Schematic diagram of the targeting cassette used to generate modified BACs. (C) Confirmation of modification was achieved using restriction digest fingerprinting with EcoRI (left panel). Diagnostic changes are indicated: R, recombinant vector band; P, parental vector band; T, targeting cassette band; Neo, neomycin resistance gene. Lanes 1 and 2, parental BAC; lane 3, recombinant BAC with neomycin resistance; lanes 3 to 7, recombinant BAC with Neo removed. These changes were verified by Southern analysis (right panel) using a PstI digest and Neo probe (top) and HindIII digest and LacZ probe (bottom). Expected sizes are indicated. Clones which have had the neomycin resistance gene removed by Flp induction are marked. Lane 8, parental BAC; lanes 9 and 11, recombinant BAC with neomycin resistance; lanes 10 and 12, recombinant BAC with Neo removed; lane 13, targeting vector positive control.

BAC modification.

BAC DNA was first introduced into Escherichia coli strains bearing a lambda prophage under the control of a temperature-sensitive repressor (EL250) (10, 35). Cells were grown overnight, diluted into 10 ml of fresh medium, and then grown to an optical density of 0.4 to 0.6. Cells were harvested by centrifugation at 1,000 × g for 10 min and resuspended in 10 ml of sterile water. The bacteria were incubated on ice for 15 min, pelleted, and resuspended in 5 ml of water. After 15 min on ice, cells were again pelleted and resuspended in 2 ml of 10% glycerol, followed by another 15-min incubation on ice. The cells were then pelleted and resuspended in a final volume of 200 μl of 10% glycerol. One hundred microliters of this E. coli solution and 0.2 to 1 μg of BAC DNA were used for electroporation. Electroporation was performed at 1.8 kV, 25 μF, and 200 Ω by using a Bio-Rad gene pulser. Τhe presence of the correct BAC was verified by HindIII digestion of miniprep DNA. Following transformation of the BAC into the EL250 strain, clones were prepared for homologous recombination using a similar protocol with the addition of a 15-min induction period in a 42°C shaking water bath prior to harvest. For recombination, we used 100 to 200 ng of linear targeting fragment for electroporation. All recombination experiments were verified by diagnostic restriction digests with EcoRI and NotI and confirmed by Southern analysis.

Arabinose induction of Flp recombinase.

To induce Flp recombinase for removal of neomycin resistance, an overnight culture of EL250 with the BAC was diluted 1:50 into 10 ml of L-broth plus chloramphenicol (20 μg/ml). The culture was grown for 2 h, after which l-arabinose was added to 0.01 to 1 mg/ml. The induction was shown to be dependent on both arabinose concentration and time of incubation (M. Khandekar, unpublished data). At time points of 30 min, 1 h, and 2 h, aliquots were taken and streaked onto appropriate selection plates. Recombination was assayed by replica spotting on both chloramphenicol and chloramphenicol-kanamycin (30 μg/ml) plates.

LacZ reporter cassette.

An Frt-flanked Neo gene was amplified by PCR from the pIGCN1 plasmid (the gift of Neal Copeland and Nancy Jenkins) and cloned into a SmaI site of pBluescriptII (Stratagene) to generate pFrtNeo. Two regions of homology to either end of the SacBII gene present in the pBACe3.6 vector backbone used to generate the RPCI-23 library were ligated to either end of pFrtNeo. The first region (homology 1, 2,977 to 3,123 bp of pBACe3.6 [5]) was generated using the following primers: forward, 5′-TTACTAGTTAATTAAGTTACGACTGCACTTCTGG-3′; reverse, 5′-TTACTAGTCGAATTGAGGCACTTGGT-3′.

The BAC homology 1 (H1) PCR fragment was purified, digested with SpeI, and cloned into the SpeI site of pFrtNeo to create pH1pFrtNeo. This plasmid was digested with PacI and EcoRV and cloned into the PacI and (Klenow-filled) XbaI site of pNEB193 to generate pNEB193H1FrtNeo. The second homology region (H2, 4,049 to 4,223 bp) was made with the following primers: forward, 5′-AACTCGAGCGATATTTACATGCTTGGTT-3′; reverse, 5′-AAGTCGACCATGTAGCTTGTGATAACCA-3′.

The BAC homology 2 (H2) PCR fragment was purified and digested with XhoI and SalI and cloned into the XhoI site of −4ISIGβ, a plasmid containing a mouse genomic fragment containing both Gata2 promoters as well as LacZ fused to the translational start site in exon 2 (18) to create plasmid −4ISIGβH2. This plasmid was digested with SalI and XhoI and cloned into the SalI site of NEB193H1FrtNeo. The linear fragment used for recombination with each of the BACs was prepared by digestion with PacI and SalI.

Construction of UG test plasmids.

Conserved noncoding sequences in the 70-kbp interval defining the urogenital enhancer activity were amplified by PCR from the 333I12 BAC using primers that added XhoI and SalI sites to facilitate subsequent subcloning. These fragments were cloned into the SalI site of the same −4ISIGβ plasmid described above. Constructs were prepared for microinjection by digestion with XhoI and SalI, purified using GenecleanII (Q Biogene) and then microinjected at 2.5 to 10 ng/μl. The primers used for PCR amplification of four candidate urogenital enhancers are as follows: UG1 forward, 5′-CCAACTCGAGTTGTCTGAAAATGGTTGCTCTCAT-3′, and reverse, 5′-ACGCGTCGACTCGTCCTCTACATAGTGAGTACAAAGC-3′; UG2 forward, 5′-CCAACTCGAGCCGGTGCTACACGCATTATTT-3′, and reverse, 5′-ACGCGTCGACGACCTCGGAGCACAGAGTCC-3′; UG3 forward, 5′-CCAACTCGAGGAGTGATGTTCAGAAGAAACCCATT-3′, and reverse, 5′-ACGCGTCGACTGATCCTTAAAAGTATTGAGGAGGTAGTCT-3′; and UG4 forward, 5′-CCAACTCGAGTTAGGGATGCTTAGCTAGT-3′, and reverse, 5′-ACGCGTCGACCACTATCCTCGCTTAGCC-3′.

Deletion of UG2 and UG4.

Deletion constructs were generated using oligonucleotides containing 40 to 50 bp of homology to flanking sequences of the region to be deleted and 20 bp that were complementary to the pFrtNeo plasmid (described above). The targeting fragment was amplified by PCR, purified, and digested with DpnI to remove methylated template DNA. The primers were as follows: 333dug2 forward, 5′TGGGAGGGTAACACAGACTTCATGCATGCCCAGCACGAGGCTCACAGCTACGCTCTAGAACTAGTGGATC-3′;333dug2 reverse, 5′CACAGTGTCAGTGTGTGGATAAGCTCCATGTCTTGAAATCTGCACATTCATCGAGGTCGACGGTATCG-3′; 333dug4 forward, 5′CCTATGTGTTGCTCAAAAAGTCTCAGATTTTAGGGCATTTTTGATTTTGCAGCTCTAGAACTAGAACTAGTGGATC-3′; and 333dug4 reverse, 5′TCAGTGTGTCTCTGCTGTGAGCCAGCTCGGGCAGGAGGAGCTCTCTGTTCTCGAGGTCGACGGTATCG-3′.

BAC purification for microinjection.

BACs were purified for microinjection from 400 ml of culture using two Clontech AX-500 columns according to the low-copy plasmid protocol (Clontech). Cell lysates were not centrifuged but rather cleared by gravity filtration. The final pellet was resuspended in 3 ml of Tris-EDTA (TE), pH 7.4, and prepared for CsCl gradient centrifugation using an NVT-90 rotor at 55,000 rpm for 18 h or a TLN-120 rotor at 2 h at 100,000 rpm, 1 h at 95,000 rpm, and 2 h at 90,000 rpm for a total of 5 h. Supercoiled bands were withdrawn using a 21-guage needle under long-wave UV, extracted with TE-saturated butanol, and dialyzed overnight against TE. Supercoiled DNA was verified by electrophoresis at 4°C for 24 h in the absence of ethidium bromide (32). DNA for injection was dialyzed overnight on a 0.05-μm-pore-size Millipore filter floating on sterile injection buffer (100 mM NaCl, 10 mM Tris-Cl [pH 7.5], 0.1 mM EDTA, 3 μM spermine, 7 μM spermidine). DNA was diluted in injection buffer and microinjected using standard techniques at concentrations of 0.5 to 2 ng/μl.

Embryo analysis.

Embryos were dissected into phosphate-buffered saline (PBS), rinsed several times, and fixed in 3% Formalin and 0.2% glutaraldehyde. After washing, embryos were stained in X-Gal (5-bromo-4-chloro-3-indolyl-β-d-galactopyranoside) overnight. For whole mount photography, embryos were cleared in 80% glycerol or benzyl alcohol-benzyl benzoate (2:1). Embryos for sectioning were fixed as described above, washed, incubated in 20% sucrose overnight, and then embedded in OCT and sectioned by cryostat. Sections were rinsed in PBS, incubated in X-Gal solution overnight at 37°C, and counterstained with nuclear fast red.

Antibody staining.

Embryos for antibody staining were fixed for 30 min in 4% paraformaldehyde at 4°C, washed in PBS, incubated in 20% sucrose overnight, embedded in OCT, and sectioned by cryostat. An anti-green fluorescent protein (anti-GFP) antibody (Molecular Probes; A11122) was used at 1:1,000 to 1:2,000 dilutions and detected using a horseradish peroxidase-conjugated anti-rabbit secondary antibody (Zymed) followed by diaminobenzidine staining. Colocalization of β-galactosidase (β-Gal) and GFP was performed by β-Gal immunofluorescence and direct GFP fluorescence. β-Gal immunofluorescence using a polyclonal β-Gal antibody (a gift of Robert Holmgren, Northwestern University) at 1:4,000 dilution was detected by a Cy3-conjugated anti-rabbit secondary antibody (Zymed). Fluorescence was visualized using a Leica DM inverted microscope using filters for Cy3 and fluorescein isothiocyanate. Separate images were taken and merged using OpenLab software.

RESULTS

Screening the genome in the vicinity of Gata2 for enhancer activity.

In order to locate Gata2 urogenital regulatory sequence(s), we developed a strategy to rapidly screen large genomic regions for enhancer activity. The strategy was conceptually based on a cell-based transfection assay developed 2 decades ago by Webber and colleagues, called enhancer trapping (33). The approach was designed to take advantage of the BAC libraries generated for the mouse genome sequencing effort by introducing a reporter construct into the vector backbone of BACs using homologous recombination in E. coli. When assayed in transgenic mice, classical enhancers present in the genomic insert of any BAC should be able to stimulate the Gata2 promoter and drive transcription of the reporter gene, yielding an in vivo readout of any enhancer activity harbored within a particular BAC. We anticipated that this strategy should circumvent the limitations of current strategies of regulatory sequence discovery.

We used the RPCI-23 mouse genomic library (25) as a source of C57BL/6 BAC clones defining the Gata2 locus because it has been well characterized and was used for the genome sequencing effort (7). Figure 1A depicts the designation and approximate position within the Gata2 locus of each of the BAC clones examined in this study (detailed in the legend to Fig. 1). Because Gata2 uses two alternative promoters (17), position 0 kbp was arbitrarily assigned as the Gata2 translational initiation site.

Having established a starting contig of BACs surrounding Gata2 by bioinformatics, we developed a reporter cassette to introduce the Gata2 promoter-directed lacZ gene into the BAC vector backbone (Fig. 1B). This cassette contained three critical elements. The first was a segment bearing both Gata2 promoters driving the lacZ reporter gene. For this purpose, we used a large genomic fragment of the Gata2 locus starting 4 kbp 5′ to the first alternative exon (1S) and ending with the lacZ reporter gene fused to the Gata2 translational start site. This fragment contains both the hematopoietic and general promoters for Gata2 (17, 18). We anticipated that by using this extended element, we would be able to identify enhancers that might specifically interact with either of the endogenous Gata2 promoters.

The second aspect of this targeting strategy was the incorporation of DNA sequences at both ends of the final targeting vector that would be used for bacterial homologous recombination. We chose two sequences within the SacBII gene in the vector backbone (used for negative selection during construction of the BAC library [5]). With these two elements flanking the reporter construct, homologous recombination was predicted to replace the SacBII gene with the Gata2/LacZ reporter. As the SacBII element is nonfunctional in the BAC, this modification should not result in any effects on BAC maintenance or stability.

The final element of the reporter cassette was incorporation of a neomycin resistance gene as a positive selection for homologous recombination in E. coli. Frt sites flanking the Neo gene allowed for its removal by induction of Flp recombinase prior to the generation of transgenic mice, eliminating the possibility of promoter interference effects (16).

By generating homologous recombinants in the EL250 E. coli strain (10), we were able to modify all of the BACs described here with this reporter fragment. We subsequently removed the Neo gene by arabinose induction of the Flp recombinase carried in EL250. The homologous recombination events were verified both by restriction enzyme digestion and Southern blotting (Fig. 1C). The EcoRI fingerprint digest displayed an appropriate increase in the size of the vector band, as well as the appearance of a new band derived from the targeting cassette (Fig. 1C). After removing the Neo gene by inducing Flp recombinase, we observed the anticipated change in size of the corresponding band (Fig. 1C, left panel). Proper homologous recombination and selection cassette excision were always verified in the modified BACs by Southern blotting, using probes for LacZ as well as for Neo (Fig. 1C, right panels).

Assessment of 1 Mbp surrounding the Gata2 gene for urogenital enhancer activity.

After modification of the BACs (described above), we assayed multiple transgenic founders for tissue-specific β-galactosidase activity. As shown in Fig. 2, the 333I12 BAC, which contains a genomic insert from kbp +60 to +260, shows strong X-Gal staining in the developing urogenital system (arrow, Fig. 2B). Similar staining was detected in six of seven 333I12/G2lacZ founder transgene-positive embryos (Table 1). The staining in the mid- and hindbrain and interdigital regions is detected in all transgenic mice as well as those bearing only the reporter cassette, and thus this pattern represents the background activity of the Gata2 promoter fragment incorporated into all of these transgenic mice (data not shown). The 272C07 BAC, bearing a 5′ end beginning at kbp +143, does not display similar activity (zero of seven transgenic embryos; Table 1), implying that the urogenital enhancer must lie within the interval between kbp +60 and +143 3′ to the Gata2 gene. Since the d16 YAC, which failed to rescue urogenital activity (36), ends at kbp +73 3′ to the Gata2 translational start site, we conclude that the urogenital enhancer activity within BAC 333I12 must be localized between kbp +73 and +143 3′ to the Gata2 structural gene.

FIG. 2.

Identification of BAC containing urogenital enhancer activity. (A) BACs were modified as described above and injected into fertilized oocytes. Founder embryos were collected on e12.5 and analyzed by X-Gal staining. Embryos bearing the 333I12-modified BAC show strong expression in the urogenital system (arrow). None of the other modified BACs show this activity. Other sites of expression are present in other BACs, indicating either the presence of other enhancers or ectopic sites. Pictures shown are representative founder embryos. The expression data are presented in tabular form in Table 2.

TABLE 1.

Summary of BAC expression data^a

BAC name	Start (kbp)	End (kbp)	Digits	BC	Ear	Eye	HB	MB	Nose	Other	UG	Vibrissae	Gut
179D23	−379	−134	2/7	0	2/7	1/7	7/7	6/7	2/7	3/7	0	0	6/7
333I12	+60	+260	5/7	5/7	6/7	0	5/7	5/7	7/7	3/7	6/7	0	0
272C07	+143	+390	5/7	5/7	6/7	0	5/7	5/7	6/7	1/7	0	0	0
49L08	+230	+430	0	0	3/8	0	4/8	4/8	3/8	3/8	0	0	0
97F04	+427	+560	1/8	0	1/8	3/8	8/8	7/8	2/8	2/8	0	0	0
356J11	+550	+787	5/6	0	0	0	6/6	6/6	0	5/6	0	6/6	0

^aBC, boundary cap cell; HB, hindbrain; MB, midbrain; UG, urogenital.

This “BAC trap” assay was designed to function as an enhancer trap for any positive regulatory activity that might lie within the genomic interval circumscribed by any BAC, and the initial survey data show that other enhancers must also be present in some of the sequences tested. Table 1 describes a summary of expression data for the BACs that span the kbp +60 to +787 interval of the Gata2 locus. For example, both 333I12- and 272C07-modified BACs show lacZ expression in the boundary cap cells of the spinal roots (31), implying the existence of a Gata2 boundary cap cell enhancer lying between kbp +143 and +260. Analysis of Gata2/GFP knock-in mice (described below) shows coincident GFP expression in the same cells. Similarly, mice bearing BAC 356J11/G2lacZ show strong transgenic staining in the vibrissae, while BAC 97FO4/G2lacZ do not, indicating that a tissue-specific enhancer conferring whisker-specific patterning lies between kbp +560 and +787 of the locus. Additionally, BAC 179D23/G2lacZ directs specific staining in a section of the gut, suggesting the presence of a Gata2 intestinal enhancer located between kbp −379 and −198. The 49L08- and 97F04-modified BAC transgenic mice show only variable staining, therefore probably reflecting weak ectopic staining due to variation in the integration sites of these transgenes. In summary, each of the transgenic BACs reveals consistent patterning by elements contained within them, demonstrating that this broad initial survey strategy is both robust and highly reproducible.

Comparison of BAC reporter to endogenous Gata2 expression.

Because the region of chromosome 6 surrounding the Gata2 locus is gene rich (Fig. 1A), we could not rule out the possibility that the BAC trap assay might detect enhancers that normally regulate one of these other interspersed genes, and thus might promiscuously activate the BAC reporter constructs. To verify that the urogenital enhancer activity detected in 333I12 was a bona fide Gata2 urogenital enhancer, we compared the distribution of X-Gal staining (from the reporter BAC) to the distribution of GFP in a Gata2/GFP knock-in mouse (N. Suzuki and M. Yamamoto, unpublished data). Transgenic lines bearing the 333I12-modified BAC were generated and crossed to the Gata2 GFP germ line mutant to generate compound mutant mice bearing both a lacZ-marked BAC as well as the GFP knock-in Gata2 allele.

Gata2 is expressed throughout the developing urogenital system, initiating expression in the developing urogenital ridge at embryonic day 10.5 (e10.5). Gata2 is expressed in the mesenchyme of the urogenital ridge but not in the epithelium of the e10.5 nephric duct (Fig. 3A). Although Gata2 is abundantly expressed in the mesonephric mesenchyme, no expression is detected in the mesonephric tubules (Fig. 3A) at this stage. Gata2 is also expressed in the cloaca, the site of formation of the urogenital sinus (Fig. 3B). By e12.5, Gata2 continues to be expressed in the mesenchyme of the mesonephros and surrounding the nephric and paramesonephric ducts but is additionally induced in the epithelium of the branching ureteric bud (Fig. 3C), as well as in the mesenchyme immediately surrounding the developing ureter. There is widespread expression in both the epithelium and mesenchyme surrounding the urogenital sinus (Fig. 3D). This pattern continues throughout development, and Gata2 continues to be expressed in the male derivatives of the nephric duct and in the ureteric bud, specifically the postnatal vas deferens (Fig. 3E), as well as in the ureter and bladder and the collecting ducts of the postnatal kidney (Fig. 3F).

FIG. 3.

Normal expression of Gata2 in the development of the mouse urogenital system. Immunohistochemistry using an anti-GFP antibody on frozen sections of mice bearing a Gata2/GFP knock-in allele. (A) Saggital section of e10.5 embryo showing GFP expression in the mesenchyme surrounding the nephric duct (ND) and mesonephric mesenchyme. There is no staining in the epithelium of the nephric duct per se, nor is there staining in the mesonephric tubular (mt) epithelium. (B) Saggital section of e10.5 embryo showing staining in the cloaca (cl), which is partitioned into the urogenital sinus and hindgut. (C) Transverse section of e12.5 embryo showing staining in the mesenchyme surrounding the nephric and paramesonephric ducts, as well as in the ureteric bud (ub). The epithelium of the nephric duct does not appear to stain, but both the epithelium and the mesenchyme surrounding the ureteric bud show strong GFP expression. There is also no staining in the mesonephric mesenchyme (mm). (D) Expression of GFP is widespread in the region surrounding the urogenital sinus (us). There is staining in both the mesenchyme and epithelium of the walls of the urogenital sinus, and surrounding the areas where the ureter meets the prospective bladder. (E) Whole mount GFP fluorescence of P14 male mouse, showing intense GFP signal from the derivatives of the nephric duct, the vas deferens (vd), and epididymis. k, kidney; t, testis; bl, bladder; u, ureter. There is no signal from the testes (t). The ureter and bladder both show strong GFP expression. (F) Section of P14 mouse kidney, showing scattered positive tubules in the kidney. These appear to be the collecting ducts, which are derived from the ureteric bud.

Analysis of adjacent serial sections from e14.5 compound mutant embryos indicates that the 333I12 BAC recapitulates several, but not all, of the sites where Gata2 is normally expressed. First, lacZ and GFP stain identically in several tissues outside the urogenital system, such as the nasal mesenchyme (Fig. 4A and D). However, in the urogenital system there is only partial overlap between transgene-derived and endogenous Gata2 expression. In the metanephros (Fig. 4B and E), the BAC-derived pattern largely overlaps that of endogenous GATA-2, as reflected by coincident lacZ and GFP staining, respectively, in most structures. LacZ is expressed in the mesenchyme of the degenerating mesonephros (Fig. 4C) and surrounding the nephric ducts, as well as the mesenchyme surrounding the ureter (Fig. 4B). In contrast, there is no X-Gal staining in the ureteric epithelium (Fig. 4E), whereas endogenous Gata2 is abundantly expressed there (Fig. 4B). The BAC-derived transgene is also not expressed in the branches of the ureteric bud within the metanephric mesenchyme (Fig. 4E), while endogenous Gata2 is prominently expressed there (Fig. 4B). Near the junction between the ureters and the bladder, the expression pattern of the transgene versus endogenous GATA-2 in the urogenital sinus is largely identical (Fig. 4C and F), again with the major difference being the expression of GFP (the endogenous GATA-2 pattern) in the epithelium of the ureter, where there is no corresponding lacZ staining from the transgene.

FIG. 4.

333I12 partially recapitulates Gata2 urogenital expression. Transgenic lines bearing the 333I12Z BAC transgene were mated to Gata2/GFP knock-in mice. Immunohistochemistry using anti-GFP antibody (A, B, and C) or X-Gal-stained adjacent sections (D, E, and F) on frozen sections from e14.5 mouse embryos shows that the BAC transgene expresses LacZ in some, but not all, of the places where GFP is expressed. (A and D) Transverse sections of the head show expression of lacZ in olfactory mesenchyme completely overlaps with expression of GFP. (B and E) Transverse sections at the level of the metanephros and mesonephros show that expression in the mesonephric mesenchyme (mes) is identical between lacZ and GFP. However, the epithelium of the ureteric bud (ub) shows only GFP expression, not lacZ. The mesenchyme surrounding the ureteric bud shows both GFP and lacZ staining. (C and F) Transverse sections at the level of the urogenital sinus (us) show a good overlap of lacZ and GFP staining. The notable exception is the expression in the epithelium of the ureter (ue [t]), which shows only GFP staining (B). (G) Direct GFP fluorescence from a transverse section at the level of the metanephros shows signal from both the epithelium (ue) and the mesenchyme (um) of the ureteric bud. (H) Anti-LacZ immunofluorescence in the same section shows signal only from the mesenchyme surrounding the ureteric bud, not from the epithelium. (I) This lack of overlap can be seen clearly in the merged image, where there is colocalization of GFP and lacZ in the mesenchyme but only GFP signal in the epithelium.

To confirm the observed differences in transgene versus endogenous GATA-2 expression, indirect immunofluorescence (using a primary lacZ antibody) and direct GFP fluorescence were compared at the level of the metanephros in the same section. LacZ expression from the BAC transgene (Fig. 4G) overlaps that of GFP from the endogenous locus (Fig. 4H) in the peri-ureteral mesenchyme but not in the ureteric epithelium (merged image in Fig. 4I). The lack of expression in the epithelium indicates that the mesenchymal and epithelial expression domains of Gata2 are regulated separately and that this epithelial urogenital enhancer does not lie within the 333I12 BAC.

Localization of Gata2 urogenital regulatory sequences within BAC 333I12.

From the data presented thus far, we concluded that a presumptive mesenchymal urogenital enhancer within 333I12 must lie between kbp +73 and +143 from the Gata2 translational initiation site. In order to precisely map the location of this element, we used comparative genomics to identify highly conserved noncoding sequences within this interval under the assumption that such conserved sequences might represent regulatory elements (13, 24). A comparison between the human and mouse Gata2 loci in the critical 70-kbp interval in question using PIPMAKER (28) revealed several sequences which are highly conserved between humans and mice (Fig. 5A). By using criteria that considered only sequences longer than 200 bp of greater than 80% identity, we identified four candidate conserved noncoding sequences (CNS) for the urogenital enhancer within that 70 kbp (called UG1 through UG4; Table 2). We isolated each of these candidate core enhancer elements (including approximately 2 kbp of flanking sequence) by PCR and then cloned the individual elements adjacent to the Gata2 promoter directing lacZ expression and assayed for β-galactosidase activity in founder transgenic embryos at e12.5.

FIG. 5.

Isolation of urogenital enhancers by comparative genomics. (A) PIPMAKER analysis of the critical 70-kbp region that contains urogenital enhancer activity. The circled regions represent conserved noncoding sequences that were identified as longer than 200 bp with >80% conservation between the human and mouse. In the case of UG3, two of these elements were located in close proximity and were grouped together into one region. Other highly conserved regions that are not circled represent the exons of the Eefsec gene, which is transcribed in the opposite direction of Gata2. Length and percent identity of each CNS are listed in Table 2. Approximately 2 kbp surrounding these conserved regions were amplified by PCR and cloned into a vector containing the Gata2 promoter driving lacZ. (B) Transgenic founders bearing each of these conserved regions were analyzed by X-Gal staining. Founders bearing both UG2 and UG4 show expression of lacZ in the urogenital system (arrows). However, each transgene has a unique pattern of activity, where UG2 activates expression in a rostral urogenital domain and UG4 activates expression in a caudal domain. Photos are of representative embryos; total transgenic data are shown in Table 2.

TABLE 2.

Conserved noncoding sequences in the critical urogenital interval of BAC 333I12

CNS	UG1	UG2	UG3	UG4
Location (kbp)	+103	+113	+142	+75
Length (bp)	235	359	208; 219	246
Percent identity	86	94	90; 89	96
No. stained	9	6	12	7
No. urogenital	0	6	0	6
No. transgenic (PCR)	12	15	14	9

Mice bearing the UG1 and UG3 reporter constructs displayed activity only in the CNS and digital regions, a pattern identical to that seen with the promoter alone. In contrast, mice bearing the UG2 (centered at kbp +113) and UG4 (kbp +75) elements exhibited lacZ expression in the developing urogenital system (Fig. 5B). Intriguingly, these two enhancer fragments direct different mesenchymal-specific spatial distributions, with UG2 driving lacZ expression in the more rostral mesenchyme surrounding the nephric duct and mesonephros and UG4 governing activation surrounding the caudal urogenital sinus and the mesenchyme surrounding the ureteric bud. These data suggest that 333I12Z bears two separate urogenital mesenchymal enhancers, each defining a unique expression domain within the developing urogenital mesenchyme.

To verify that BAC333I12Z bears no other than these two distinct urogenital enhancer activities, we deleted each enhancer separately from the 333I12Z BAC. The deletions were generated by homologous recombination using PCR-generated fragments containing an Frt-flanked neomycin resistance gene with 50 bp of DNA homologous to the Gata2 genomic sequence surrounding either UG2 or UG4. Homologous recombination in E. coli resulted in replacement of the presumptive enhancer with the neomycin gene, which was subsequently removed by Flp recombinase induction (see Materials and Methods). These modified BAC constructs were verified by restriction enzyme fingerprint digests and Southern blotting as before (data not shown) and then used to generate transgenic mice.

Embryos bearing the 333I12ΔUG2 BAC display lacZ staining only in the caudal aspect of the urogenital system and do not stain the rostral part (Fig. 6A and B). Analysis of frozen sections demonstrates that this staining is limited to the urogenital sinus and caudal nephric duct as well as the mesenchyme surrounding the ureters (Fig. 6E and G). This transgene does not express lacZ in the mesonephric mesenchyme (Fig. 6F). This staining pattern perfectly recapitulates the pattern detected in transgenic mice bearing the UG4-directed LacZ transgene (Fig. 5B). Similarly, e12.5 embryos bearing the 333I12ΔUG4 BAC display lacZ staining in the mesenchyme surrounding the rostral nephric duct and degenerating mesonephros (Fig. 6C, D, and I), a pattern that is quite similar to the one observed with the UG2 transgene (Fig. 5B) and complementary to the pattern observed in 333I12ZΔUG2 mice. No staining is detected in the caudal regions or peri-ureteral mesenchyme (Fig. 6H and J). Adding the Tg^333I12ZΔUG2 and Tg^333I12ZΔUG4 patterns together, the two expression domains comprise the entire staining pattern of the parental 333I12Z BAC. From these data, it appears that these two enhancers independently control regional expression in mesenchymal tissues within two distinct domains in the urogenital system and represent the complete urogenital expression pattern contained within the 333I12Z BAC. In conjunction with the experiments described above, we conclude that the urogenital patterning of GATA-2 is conferred by a minimum of three separate enhancer activities: both rostral (UG4) and separate caudal (UG2) mesenchymal elements within BAC 333I12, and at least one other epithelial enhancer element that must lie beyond the 5′ (kbp −379) or 3′ (kbp +787) borders of the Gata2 locus that we have examined here.

FIG. 6.

Deletion of UG enhancers from 333I12Z demonstrates that two independent enhancers control Gata2 in urogenital mesenchyme. By using homologous recombination in E. coli, UG2 and UG4 were individually deleted from the parental 333I12Z BAC (A and B). Two representative 333I12ΔUG2 founder embryos at e12.5 show a similar staining pattern as 333I12 mice, except the rostral domain of urogenital staining is absent. This absent region corresponds to the region activated by the UG2 enhancer. (C and D) Two representative 333I12ΔUG4 founder embryos at e12.5 show a reciprocal pattern to the 333I12ΔUG2; they retain lacZ expression in the rostral domain but lose expression in the caudal urogenital system. This region of activation corresponds to the UG4 transgene. (E) Transverse section of 333I12ΔUG2 at the level of the metanephros shows staining in the mesenchyme surrounding the ureter (u), indicating that UG2 is not responsible for this domain. (F) Sections at the level of the mesonephros (m) show no staining, this is the rostral domain which is lost when UG2 is deleted (g, gonad). (G) Staining in the urogenital sinus (us) is preserved, indicating that this domain is not controlled by UG2. (H) Sections of 333I12ΔUG4 show no staining in the mesenchyme surrounding the ureter, indicating that UG4 controls activation in this region. (I) Sections show lacZ expression in the mesonephros, which confirms that this expression is not regulated by UG4. (J) No staining is present in the urogenital sinus, suggesting this region is controlled by UG4.

DISCUSSION

Linking functional genomics to bioinformatics in order to identify very distant transcriptional regulatory elements.

Here we report the development of a strategy using BAC transgenic mice linked to bioinformatics to identify and quickly localize regulatory elements that lie a great distance from the gene being investigated. This screen utilized the insertion of a reporter cassette into the vector backbone of a well-characterized BAC library, followed by screening of each modified BAC for enhancer activity in a transgenic mouse founder assay. We believe this approach has several advantages over previously described BAC screening assays. Unlike reporter-BAC coinjection (4), the reporter in this strategy was inserted into a defined location within the BAC, giving us a predictable transgene structure with a 1:1 ratio of reporter gene and BAC transgene. One can imagine that the existence of multiple promoters at a transgene integration site might result in a disruption of the normal interactions between the enhancer and promoter, making interpretation of the expression data more difficult.

Secondly, a single modification of the BAC followed by injection of circular DNA should yield a variety of possible structures and orientations of the promoter relative to any putative enhancer within the genomic interval borne on the BAC. This variation of transgene structure should circumvent problems that might arise due to inappropriate positioning of chromatin structural elements such as insulators or matrix attachment regions relative to the embedded enhancer and the promoter.

Finally, by inserting the reporter gene into the vector backbone, one is not limited to BACs containing the gene itself, greatly increasing the possible number of BACs for use in both the initial survey as well as for rapid localization of the position of critical sequences. Due to the random nature of the creation of BAC libraries, finding a particular BAC that has both the gene of interest and any given enhancer may be impossible, limiting the area to be examined by conventional reporter gene knock-in strategies. In practice, using the RPCI-23 library, this type of approach would be limited to searching less than 200 kbp (the average insert size) upstream and downstream of the gene of interest. The strategy described here circumvents this limitation, allowing us to extend the search as far away from the structural gene as required to identify the enhancer. Furthermore, by applying the same reporter and targeting fragment to overlapping BACs, we can more rapidly and precisely define the critical interval, thereby reducing the candidate conserved noncoding sequences within that span that must be tested. These advantages, translated into the data presented here, show that this method can be readily employed as a general approach to identify regulatory elements that lie significant distances from any gene of interest. We have recently employed an identical strategy to extend the known boundaries of the Gata3 locus (T. Moriguchi and J. D. Engel, unpublished data) and have identified enhancers lying several hundred kilobase pairs from the structural gene.

Somewhat surprisingly, a very reproducible staining pattern in founder embryos was observed when enhancers were present in the genomic intervals within the BACs. For example, the BAC bearing the urogenital enhancers failed to be expressed in only one of seven transgenic founders, indicating that the enhancer-promoter interaction was not affected by different orientations of these two elements after integration of the circular BAC DNA as a transgene. There was more ectopic expression in BACs where no enhancer activity was detected (for example, between positions +230 and +560 kbp), and we concluded that the staining detected in these transgenic embryos was likely due to variation among randomly encountered local elements according to each individual transgene integration site.

Despite the advantages of reproducibility, rapidity, and functional simplicity, a few conceptual considerations must be considered before employing this approach. First, the assay does not preserve the normal physical relationship between an enhancer and its target promoter, thereby theoretically barring the detection of enhancers that function in a nonclassical, orientation- or distance-dependent manner. Secondly, this assay will not detect enhancers that require the cooperation of other elements in the genome. Such a cooperative delocalized enhancer, as appears to be required for Gata2 hematopoietic activity, has been identified in the locus (unpublished observations), and it would not have been detected using this assay. Similarly, the primitive erythroid enhancer activity in the first intron of Gata1 is only exhibited when the 5′ definitive hematopoietic enhancer is linked to it in cis (23), demonstrating that some enhancer elements have stringent structural requirements for their activity. These important caveats must be considered when interpreting the results of this, or any similar, approach.

Localization of distant enhancers.

Combining the expression data from overlapping BACs, we were able to refine the position of the critical region for Gata2 urogenital expression from an initial survey of 1 Mbp of the Gata2 locus to a 70-kbp interval present in BAC 333I12. In comparing the human and mouse genome sequences within this interval, we identified four highly conserved noncoding, nonrepetitive sequence elements, two of which were shown to regulate expression of a Gata2-directed reporter gene in the urogenital system (Fig. 6A). Two interesting points should be noted. The genomic region surrounding Gata2 is gene rich and harbors a relatively large number of conserved noncoding sequences which are longer than 200 bp and have >80% sequence identity. Thus, we were unable to precisely define the location of these elements simply using the interspecies sequence conservation data. However, when combining this data with the functional data produced in the BAC trap assay, we were able to relatively rapidly reduce the number of candidate cis regulators to a manageable few that could be directly tested for activity. The combination of interspecies sequence conservation data and functional data thus led to rapid localization of two discrete enhancers present in the Gata2 locus.

Interestingly, these enhancers are interspersed among the genes surrounding Gata2. In fact, UG2 is present in an intron of the Eefsec gene, a selenocysteine-specific elongation factor that is not known to be expressed in a tissue-restricted manner. Furthermore, previous work has shown that a hematopoietic element must lie more than 100 kbp upstream, which lies 5′ to Ribophorin1, another presumptively ubiquitously expressed gene (14). We deduce (and speculate) that the promoters of these housekeeping genes may not be sensitive to the effects of these enhancers and are instead controlled only by local elements. Alternatively, a complex chromatin structure may exist in this region, ensuring conformational availability of these distant Gata2 enhancers to the promoter.

The identification of other enhancer activities present within these BACs suggests that this approach may be useful for the general identification of regulatory elements throughout the genome. In the 333I12- and 272C07-modified BACs, we observed consistent X-Gal staining in cells resembling the boundary cap cells of the spinal cord. Although a more detailed analysis is required for confirmation, we presume that a tissue-specific regulatory element driving expression in these cells is present in the overlapping 120 kbp of these two BACs. Similarly, we see very reproducible staining in the vibrissae of 356J11/G2LacZ BACs, suggesting the presence of a tissue-specific whisker regulatory element in this region. By taking this approach, we anticipate being able to identify a number of other tissue-specific elements whose existence was previously unknown. Of course, as with any enhancer trapping experiment, it is possible that these additional sites of expression are due to regulatory elements of other genes. Only a careful analysis of the expression of lacZ from these transgenes relative to the endogenous expression of Gata2, as performed here in the urogenital system, can confirm the identification of bona fide Gata2 enhancers.

Regulation of Gata2 in the developing urogenital system.

The data presented here suggest that the expression of Gata2 in the developing urogenital system is controlled in a complex manner. From these experiments, we deduce the existence of at least three separable regulatory elements controlling Gata2 expression (Fig. 6B). The first, UG2, is responsible for the activation of Gata2 in the mesonephric mesenchyme as well as the mesenchyme surrounding the mesonephric duct. This enhancer does not require either Gata2 promoter specifically and can activate a heterologous promoter and functions in an orientation-independent manner (data not shown). Deletion of the UG2 enhancer from the 333I12 BAC shows that it functions independently to activate reporter expression in the mesonephric region. The second enhancer, UG4, has an expression pattern complementary to that of UG2. It appears to govern Gata2 transcriptional activation in the urogenital sinus and in the mesenchyme surrounding the ureteric bud. Deletion of this element from the 333I12Z BAC shows that it also functions independently to activate reporter activity in this domain. The third postulated enhancer has not been definitively localized in this study but can be inferred from the data presented here (Fig. 7). Because the 333I12 BAC does not drive lacZ expression in the ureteric epithelium, we postulate that an epithelial-specific enhancer must lie somewhere outside of the region defined by this BAC, but not within the 1 Mbp surveyed in this study.

FIG. 7.

Summary of different activation domains controlled by Gata2 urogenital enhancers. (A) This study demonstrates the existence of two separate enhancers, lying more than 70 kbp downstream from the Gata2 promoter, which control activation in the urogenital system. These enhancers, in addition to enhancers present near the gene, as well as a hematopoietic regulatory element more than 100 kbp upstream of the gene, collaborate to give rise to the appropriate expression of Gata2. This study postulates the existence of a third urogenital enhancer, which cannot be localized in this study. Thus, the minimal Gata2 locus is at least 313 kbp, although the existence of a third, epithelial urogenital enhancer suggests that this locus must be larger than 1 Mbp, as we have now determined. (B) The spatial distributions of the UG2 and UG4 enhancers are clearly separated. Although both enhancers control expression in mesenchymal tissues in the urogenital system, our data suggests that UG2 is responsible for expression in the mesonephros and mesenchyme surrounding the nephric duct, while UG4 is responsible for expression in the urogenital sinus and mesenchyme surrounding the ureter. Mes, mesonephros; Met, metanephros; UG sinus, urogenital sinus; u, ureter; ND, nephric duct.

We note specifically that there are other potential explanations for the absence of ureteric epithelial staining that are related to the limitations of the assay noted previously. It is possible that there is a cooperating element, missing from the 333I12 BAC, which is necessary to generate ureteric epithelial expression. Alternatively, it is possible that this expression depends on the normal relationship and proximity of this putative enhancer to the Gata2 promoter, which is disrupted in this assay. If this were the case, analysis of the expression of a larger BAC containing both the promoter and the urogenital regulatory elements in their appropriate orientation would be predicted to shed light on this issue.

It was of some interest that such a complicated regulatory scheme evolved for the patterning of a single transcription factor in one tissue. Although a superficial analysis revealed a simple expression pattern for Gata2 in the urogenital system, it appears that the regulation of this expression is instead quite complicated. First, the regulation of Gata2 is separated into different enhancers for the epithelial and mesenchymal urogenital components. Within the mesenchymal compartment, two individual elements, separated by 30 kbp and lying approximately 100 kbp 3′ to the structural gene, control different spatial aspects of Gata2 mesenchymal expression. This separation of regulation implies that there may be a difference in Gata2 function in these two different domains, requiring the expression to be finely tuned in each tissue. Alternatively, differential Gata2 regulation in the urogenital mesenchyme might be controlled by two different morphogens, one arising from the rostral aspect and one from the caudal aspect of the embryo, and thus the two enhancers may be responsive to fundamentally different signaling cascades. According to this hypothesis, one would predict that these two spatially defined regions are sufficiently different to require distinct sets of transcription factors to activate the two enhancers. Given the well-defined boundaries of the mesenchymal domains controlled by these separate enhancers, one can speculate that this separation of activity reflects a boundary in the transcription factors and/or signaling molecules or receptors responsible for activating these enhancers. Further characterization of the transcription factor binding sites present within these enhancers should identify candidates for the upstream regulators acting in these two distinct compartments of the urogenital mesenchyme.

Because Gata2 appears to be required at the ureter-bladder junction, it is interesting to speculate whether all of these domains of expression are required for the single lethal phenotype exhibited as hydroureteronephrosis or if expression of Gata2 in one of these domains is more important than the other. Because Gata2 and Gata3 are coexpressed in the epithelial compartment of the ureteric bud (6, 9), one might anticipate that the functions of Gata2 and Gata3 could be redundant in the ureteric epithelium. However, the converse cannot be true; animals lacking Gata3 have been shown to have severe deficits in kidney formation (12), demonstrating that Gata2 is unable to fully compensate for the lack of Gata3 function there. Furthermore, it is tempting to speculate that the caudal expression around the urogenital sinus and ureteric bud may be the most critical, as they represent the site of the defect that we detect physiologically as hydroureteronephrosis (36). Ongoing transgenic rescue studies using each of these enhancer elements individually should shed some light on this issue.

This analysis, combined with the analysis of YAC transgenic animals, suggests the presence of more than a dozen individual enhancers required for Gata2 expression, including three hematopoietic (36), three CNS (37), one endothelial (Khandekar, unpublished), and the three urogenital enhancers described here, scattered over more than 400 kbp surrounding the Gata2 locus. We strongly suspect that Gata2 is not unique in this regard, as other developmentally prominent regulatory genes also appear to harbor multiple, often quite distant, tissue-specific enhancers. As the number and types of mutations present in human diseases become better defined, it seems likely that mutations in these distal enhancers may be responsible for variations in the expression of these critical regulatory genes. Since the mutations in enhancers (or in other chromatin modulatory elements) that control the expression of these critical regulatory proteins (e.g., receptor tyrosine kinases, intracellular kinases and phosphatases, or transcription factors) could lie very far from their structural genes, conventional efforts to map and identify the mutant “gene” would likely be confounded. Only by explicit identification of these enhancers, and analyzing their detailed activities, can we begin to understand the complex network of regulatory control that underlies both embryonic development and human disease.