Angioblast-mesenchyme induction of early kidney development is mediated by Wt1 and Vegfa

Epithelial-mesenchymal inductive interactions have long been the hallmark of organ development, especially of those organs that develop specialized epithelial and/or undergo branching morphogenesis, such as the kidney, lung,liver, pancreas and other secretory glands. In the case of the kidney,interaction of the epithelial ureteric bud with the metanephric mesenchyme, a histologically distinct patch of cells in the urogenital ridge, leads to condensation of the mesenchyme around the branching ureteric bud, and the subsequent induction of a mesenchymal-to-epithelial transformation of the condensed mesenchyme into tubular elements called nephrons(Saxen, 1987). This process is reiterated by continual branching of the ureteric bud and expansion of the mesenchyme until, in the human kidney, approximately one million nephrons have been induced. Many genes have been identified with a function required for these initial interactions. For example, the transcription factors Wt1,Pax2, Six1 and Eya1 are expressed in the metanephric mesenchyme,and in the absence of any one of these factors, the ureteric bud either fails to emerge from the Wolffian duct or fails to invade the metanephric mesenchyme, thus blocking kidney development at its earliest stages(Kreidberg et al., 1993; Torres et al., 1995; Xu et al., 1999; Xu et al., 2003).

Many recent studies have significantly advanced our understanding of the molecular basis of ureteric bud-mesenchymal interactions that initiate kidney development. For example, Pax2 has been shown to regulate the expression of glial cell line derived neurotrophic factor (Gdnf)(Brophy et al., 2001), a member of the neurotrophin family expressed by the metanephric and condensed mesenchyme, which binds the c-Ret receptor expressed at the tip of the ureteric bud to stimulate growth and branching of the bud(Durbec et al., 1996; Sanicola et al., 1997; Trupp et al., 1996). In Gdnf-deficient embryos, there is no outgrowth of the ureteric bud(Moore et al., 1996; Pichel et al., 1996; Sanchez et al., 1996). Compared with that of Pax2, our understanding of how the Wt1gene regulates the interactions that initiate kidney development is less well understood. Wt1 encodes a zinc finger transcription factor; this gene was first identified as a tumor suppressor gene for Wilms' tumor, a neoplasm of the kidney that occurs in young children(Call et al., 1990; Gessler et al., 1990; Glaser et al., 1989). Wt1 is expressed in the metanephric mesenchyme and increases in expression as the mesenchyme condenses around the ureteric bud, finally being restricted to the podocyte population of the glomerulus, where it continues to be expressed in the mature kidney. Many genes have been identified as potential targets for Wt1(Scharnhorst et al., 2001),but it has been difficult to reconcile these findings with the renal agenesis phenotype of Wt1-deficient embryos(Kreidberg et al., 1993). For example, amphiregulin has been identified as a target for Wt1(Lee et al., 1999) and stimulates branching morphogenesis in kidney organ cultures, but amphiregulin mutant mice demonstrate normal kidney development(Luetteke et al., 1999). Whether this can be accounted for by redundancy among EGF family members is not known.

Recently, we reported the phenotype of transgenic mice in which a truncated form of Wt1, expected to act in a dominant-negative fashion, was specifically expressed in glomerular podocytes, which are Wt1-expressing cells that form part of the filtration unit of the kidney (Natoli et al., 2002a). Surprisingly, instead of yielding defects in podocyte differentiation,expression of the mutant form of Wt1 resulted in abnormal development of the glomerular capillaries that form in intimate contact with the podocytes. This observation has led us to hypothesize that one function of Wt1 is to regulate the expression of growth factors that are involved in vascular development. A role for growth factors that stimulate vascular development in kidney development was first suggested by Tufro, who demonstrated that vascular endothelial growth factor A (Vegfa), or hypoxic culture conditions, could stimulate proliferation and nephrogenesis in metanephric kidney organ cultures (Tufro,2000; Tufro et al.,1999; Tufro-McReddie et al.,1997). Here we present results demonstrating that Wt1does indeed regulate the expression of Vegfa. Furthermore, it is shown that Vegfa produced by the mesenchyme stimulates branching morphogenesis and nephrogenesis in the early kidney through reciprocal inductive events that involve a population of Flk1 (Vegfr2; Kdr - Mouse Genome Informatics)-expressing angioblasts. Thus, similarly to recent studies identifying a role for angiogenic cells in hepatogenesis and pancreatic development (Lammert et al.,2001; Matsumoto et al.,2001), this work identifies an interaction between angioblasts and the condensed mesenchyme as an additional crucial interaction involved in early kidney development, in addition to the classically recognized inductive interaction between the mesenchyme and the ureteric bud(Grobstein, 1953).

Recombinant mouse Vegfa (#493-MV), recombinant rat Gdnf (#512-GF) and goat polyclonal anti-Vegfa antibody (#AF-493-NA) were purchased from R&D Systems (Minneapolis, MN); rat polyclonal anti-Flk1 antibody (#550549) for immunostaining was purchased from BD Biosciences (San Diego, CA); rat polyclonal anti-Flk1 antibody (clone DC101) was obtained from ImClone Inc.(New York, NY); mouse polyclonal anti-pan-cytokeratin antibody was purchased from Sigma (St Louis, MO); rabbit polyclonal anti-Pax2 antibody (#71-600) was purchased from Zymed (South San Francisco, CA) or obtained from Dr Greg Dressler, (University of Michigan); rabbit polyclonal anti-Pecam antibody(H-300)), goat polyclonal anti-Pecam antibody (M-20) and rat-IgG were purchased from Santa Cruz Biotechnology (Santa Cruz, CA); rabbit polyclonal anti-Brush Border antibody (Miettinen and Linder, 1976) was obtained from Hannu Sariola (University of Helsinki). All secondary antibodies were purchased from Jackson ImmunoResearch(West Grove, PA). Tyrphostin SU1498 (#T-2710) was purchased from Calbiochem(La Jolla, CA); DMSO was purchased from Fisher Scientific (Pittsburgh, PA). A plasmid encoding a truncated version of Flk1(Tsou et al., 2002) was obtained from Dr F. Frank Isik, (University of Washington, Seattle). Lotus lectin (FL-1321) was purchased from Vector laboratories, Inc. (Burlingame,CA).

The electroporation system modified from that previously used to microinject chick neural tubes (Itasaki et al., 1999; Nakamura et al.,2000) is depicted in Fig. 1A. Metanephric kidney organ cultures were placed in a dish under a drop of buffer, and plasmid vectors were injected into the culture using 15-20 μm diameter glass needles prepared using a Sutter Instrument micropipette puller model P-87 (Sutter Instruments, Novato, CA). Settings on the pipette puller were: Heat 400; Pull 150; Velocity 100; Time 100. The glass tubes for making needles were obtained from Sutter, (#BF100-50-10), outer diameter 1.0 mm, inner diameter 0.5 mm, 10 cm length. Needles were connected by thin tubing to a FemtoJet microinjection device (Eppendorf, Hamburg,Germany) using the following settings: injection Pressure 6.00 psi;compensation pressure 1.00 psi; injection time 0.4 seconds, two injections per injection site. A square wave electroporator (BTX ECM830, Genetronics Inc.,San Diego, CA) was used to pulse the culture immediately after the injection,using rectangular electrodes (gold plated, Model 516, Genetronics Inc.) placed in parallel on either side of the organ culture before the injection. The parameters used here were voltage: 36 V; number of pulse: 5; pulse length: 50 ms; internal time: 100 ms. After growing for 6-48 hours, cultures were analyzed to examine gene expression, culture growth and differentiation.

Embryonic day (E) 11.5 metanephric rudiments were isolated from embryos of FVB mice, placed on nitrocellulose filters (0.1 μm pore size, Nuclepore Track-Etch Membrane, Whatman 0930059), suspended over DMEM/10% fetal calf serum (FCS) and cultured at the air/medium interface for 24, 36, 48 or 72 hours at 37°C with 5% CO₂. Blocking antibodies or pharmacological agents were added to the growth medium as mentioned for each experiment. Kidney cultures were fixed in 4% paraformaldehyde (PFA) for in situ hybridization with RNA probes, or fixed in methanol and stained with antibodies. Branch tips were quantified by manual counting after anti-cytokeratin staining, nephron proximal tubule units were identified by staining with Lotus lectin or Brush Border antibody. Results were statistically analyzed by Student's t-test. Antibody staining: kidney cultures were fixed in ice-cold methanol for 10 minutes, washed in PBS containing 1% bovine serum albumin (BSA) at room temperature and incubated overnight in primary antibodies at 4°C. The samples were then washed three times for 2 hours each in PBS at room temperature and incubated overnight in secondary antibodies at 4°C. Finally, after washing three times for 2 hours each in PBS at room temperature, organ cultures were analyzed using an inverted fluorescence microscope (Nikon Eclipse TE 300), and imaged using a Spot 1.4.0 digital camera (Diagnostic Instruments, Sterling Heights, MI). Images were processed on Macintosh computers using Photoshop 6.0 (Adobe Systems, Inc., San Jose, CA). Fluorescent alkaline phosphatase (AP) detection used for Flk1 staining was essentially as published(Natoli et al., 2004).

Riboprobes were obtained or generated from the coding region of mouse Gdnf (Srinivas et al.,1999) (obtained from F. Costantini, Columbia University), Wt1 (Pelletier et al.,1991), Vegf164 (bases 75-669 in the murine cDNA,amplified by PCR), Pax2 [(Dressler et al., 1990), obtained from P. Grus], Nanog [generated in our laboratory cDNA 558-1140 (Chambers et al., 2003; Mitsui et al.,2003)], Osr1 [(So and Danielian, 1999), obtained from P. Danielian, Chester Beatty Laboratories, London] and Wnt4[(Stark et al., 1994),obtained from S. Vainio, University of Oulu, Finland], and subcloned in pCDNA3 or pCRII-TOPO (Invitrogen). Sense and antisense probes were synthesized and labeled with digoxigenin-UTP (Roche). The protocol used for whole-mount in situ hybridization was as published(Wilkinson and Nieto, 1993). For section in situs, material was fixed in 4% paraformaldehyde and embedded in paraffin. Sections were cut at 8uM and deparaffinized through xylene and a standard ethanol series.

As specified in the Results section, RNA was prepared either from whole organ cultures or from sections co-injected with a GFP expression plasmid, in which case GFP-expressing sections were micro-dissected apart from the rest of the organ culture. Total RNA was isolated from kidney organ cultures, digested with DNase 1 (Qiagen) and reverse-transcribed with SuperScript First-Strand Synthesis System kit (Invitrogen) to get cDNA template. For Smart Cycler Real-Time PCR reaction, a mastermix of the following reaction components was prepared to the indicated end-concentration: Water (RNase free), forward primer (0.3 μmol/l), reverse primer (0.3 μmol/l), TaqMan probe (0.2μmol/l, Cepheid), Q solution (1×, Qiagen), dNTPs (0.2 mmol/l,Biolabs), PCR buffer (1×, containing 15 mmol/l MgCl2, Qiagen), Taq DNA polymerase (0.1 unit/μl, Biolabs), cDNA template (5 ng/μl). Smart Cycler reaction mastermix was filled in the Smart Cycler reaction tubes [Cepheid(Sunnyvale, CA)], which were closed, centrifuged and placed into the Smart Cycler processing block (Cepheid). The following protocols were used: for Gapdh and Vegfa, 94°C for 120 seconds; 94°C for 60 seconds, 58°C for 30 seconds, 72°C for 30 seconds, 40 cycles; cool to 4°C. For Pax2, 94°C for 120 seconds; 94°C for 30 seconds,58°C for 30 seconds, 72°C for 45 seconds, 40 cycles; cool to 4°C. For Gdnf, 94°C for 120 seconds; 94°C for 30 seconds, 60°C for 45 seconds, 72°C for 45 seconds, 40 cycles; cool to 4°C. Primers used were: gapdh (Forward: TCC ACC CAT GGC AAA TTC A; Reverse: TCG CTC CTG GAA GAT GGT), Vegfa (Forward: AGC AAC ATC ACC ATG CAG AT;Reverse: TCA CAG TGA TTT TCT GGC TTT G), Pax2 (Forward: TGG CTG TGT CAG CAA AAT CCT; Reverse: ATTCGGCAATCTTGTCCACCA), and Gdnf (Forward:TTC GCG CTG ACC AGT GA; Reverse: CTC TCT TCG AGG AAG CGC T). Standard curves of Gapdh, Vegfa, Pax2 and Gdnf were generated using cDNA dilutions from 10^-3 μg/μl to 10^-7 μg/μl(data not shown); each dilution was run in triplicate, and Ct values all varied by less than ±0.4 unit cycles. Vegfa, Pax2 and Gdnf Ct values were normalized using Gapdh Ct values using the following equation: Vegfa corrected Ct (sample 1) = Vegfa obtained Ct(sample 1) × Gapdh Ct (sample 2) / Gapdh Ct (sample 1). Ct values were converted to relative mRNA values using the Vegfa, Pax2 or Gdnf standard curves.

Using Wt1^-/- embryos to evaluate potential downstream targets of the Wt1 gene has been problematic, as the metanephric mesenchyme in these mutant embryos begins to undergo apoptosis almost immediately upon its formation as a histologically distinct structure(Kreidberg et al., 1993). Therefore, whether a gene is expressed or not may be due to the apoptotic program rather than whether it is a valid target of Wt1. To circumvent this issue, we adapted a microinjection/electroporation system previously used in chick neural tubes(Itasaki et al., 1999; Nakamura et al., 2000) to the study of Wt1 function in murine organ culture(Fig. 1A) because it allows higher-throughput experimentation than the derivation of a novel transgenic mouse line for each individual transgene. Additionally, this approach allows gene expression in locations for which no appropriate promoter is available,such as the metanephric mesenchyme. Thus, rather than studying loss-of-function phenotypes in Wt1 mutant embryos, this system allows the observation of gain-of-function phenotypes in organ cultures in which Wt1 or other genes have been overexpressed. This system also allows the evaluation of Wt1 target genes in non-immortalized cells,avoiding the possibility that Wt1 function is modified when co-expressed with immortalizing proteins. As described in the Materials and methods section, an organ culture was placed between two electrodes, DNA plasmid expression constructs were microinjected into the mesenchymal component of the organ rudiment, and the culture was electroporated immediately after the injection. After an additional 6-48 hours in culture,these cultures were analyzed to examine gene expression, culture growth and differentiation.

As a first test of this system, a green fluorescent protein (GFP)expression construct (pCS2-EGFP) was microinjected into the E11 kidney organ cultures (Fig. 1A, part c). GFP was highly expressed in multiple cell layers at the site of injection (see Fig. S1 in the supplementary material; additionally, Fig. S2 shows a slight decrease in branching number as a consequence of electroporation, but the pattern of Pax2 and Wt1 expression is preserved). To further examine the ability of this system to manipulate biologically relevant aspects of kidney development, a vector encoding Gdnf, the major growth factor responsible for stimulating growth of the ureteric bud(Sainio et al., 1997; Sanicola et al., 1997), was introduced into the organ culture by microinjection/electroporation. Localized expression of Gdnf adjacent to the Wolffian duct led to outgrowth of ectopic ureteric buds from the Wolffian duct, adjacent to the site of injection (Fig. 1B, compare with Fig. 1C). This was observed in 100% of injections (the success rate for all experiments is presented in Table 1). Injection of an empty vector that did not contain the Gdnf cDNA never resulted in ectopic bud outgrowth from the Wolffian duct(Fig. 1C). In situ hybridization with a Gdnf RNA probe showed strong expression of Gdnf at the site of injection(Fig. 1D, compare with sense probe, Fig. 1E).

Having established the efficacy of the microinjection/electroporation system for use in metanephric kidney organ culture, Wt1 was overexpressed in an attempt to identify possible Wt1 target genes. In situ hybridization with a Wt1 antisense probe showed increased signal at the site of the Wt1 plasmid injection(Fig. 2A, part a, compare with sense probe Fig. 2A, part c),whereas no increased signal was found after injection of a control GFP expression vector (Fig. 2A,part b). Overexpression of Wt1 in the condensed mesenchyme (after injection at multiple sites) led to increased branching of the ureteric bud(Fig. 2A, parts d,f, compare with Fig. 2A, part e). The condensed mesenchyme refers to the portion of the metanephric mesenchyme that condenses around the ureteric bud upon its invasion of the metanephric mesenchyme, and in which expression of genes such as Pax2 and Wt1 increases as a consequence of induction by the bud(Schedl and Hastie, 2000). These results indicate that this system might be suitable for the evaluation of potential Wt1 target genes involved in early kidney development.

Our previous work suggested that Wt1 may regulate the expression of factors that regulate vascular development(Natoli et al., 2002a). One such factor is Vegfa (Neufeld et al., 1999). Wt1 and Vegfa are both highly expressed by podocytes in the kidney glomerulus(Armstrong et al., 1993; Eremina and Quaggin, 2004),leading us to hypothesize that Vegfa may be a regulatory target of Wt1. Injection of a Wt1 expression plasmid containing a Wt1 cDNA, without alternatively spliced exon 5 or the KTS sequence present at the end of exon 9, led to high-level expression of Vegfaat the site of injection [Fig. 2B, part a, compare with sense probe (part c) and with control injection of a vector containing the Gfp cDNA (part b)]. Injection of a vector encoding Pax2 (Dressler et al., 1990) also failed to induce expression of Vegfa(data not shown). Real time-PCR for Vegfa also demonstrated increased expression of Vegfa RNA at the Wt1 injection site(Fig. 2B, part d). Previous studies have ascribed distinct functions to the different splice forms of Wt1. However, the ability to induce Vegfa expression was not restricted to a single splice form of Wt1. Rather, all four major splice forms were able to induce expression of Vegfa(Table 1, data not shown). This finding is, perhaps, not surprising in view of recent results in which we had shown that a targeted deletion of alternatively spliced exon 5 in mice had no apparent phenotype (Natoli et al.,2002b), and Hammes et al.(Hammes et al., 2001) had shown that embryos singly expressing either the -KTS or +KTS versions of Wt1 experience normal early kidney development (see Discussion for additional detail). Because we found no apparent difference in the ability of Wt1 alternatively spliced isoforms to induce expression of Vegfa, the remainder of the studies in this report used the version of the Wt1 cDNA that did not include either exon 5 or the KTS sequence.

The expression pattern of Vegfa in the early kidney is shown in Fig. 3A. Vegfa expression was present in the condensed mesenchyme and highest in pretubular aggregates(Fig. 3A). Lower levels of Vegfa were also present in the ureteric bud. In older kidneys, expression was highest in podocytes of maturing glomeruli. The mesenchymal expression of Vegfa overlapped with that of Wt1, which was also in condensed mesenchyme and in pretubular aggregates, and highest in podocytes of older kidneys (data not shown) (see Armstrong et al.,1993; Eremina et al.,2003).

Vegf164 is a predominant isoform of Vegfa(Neufeld et al., 1999). The findings of Tufro, that Vegfa could stimulate branching morphogenesis and induction of nephrons were confirmed both by adding Vegfa to the organ culture(Fig. 3B), and by microinjection/electroporation of a Vegfa expression vector (not shown). Addition of Vegfa neutralizing antibodies to the organ culture also resulted in decreased branching and nephron induction(Fig. 3C).

Understanding the role of Vegfa in early kidney development requires identification and localization of the Vegfa receptors expressed at the initiation of kidney development. The two major receptors for Vegfa are Flt1 (Vegfr1) and Flk1 (Vegfr2) (Neufeld et al., 1999); Flk1 appears to be essential for endothelial differentiation, whereas Flt1 appears to have a more crucial role in vascular assembly (Fong et al., 1995; Fong et al., 1999; Shalaby et al., 1997). Previous reports using Flk1-lacZ knock-in mice and immunostaining for Flk1 have identified a population of Flk1-expressing cells at the periphery of the condensed mesenchyme and adjacent to the stalk of the ureteric bud(Loughna et al., 1998; Robert et al., 1998; Tufro, 2000; Tufro et al., 1999; Tufro-McReddie et al., 1997). Immunostaining of E13 kidneys and organ cultures with an Flk1 antibody confirmed this expression pattern (Fig. 4A, part a), and co-staining with anti-Pax2(Fig. 4A, part b), a marker of the condensed mesenchyme and the ureteric bud, demonstrated their non-overlapping expression patterns (Fig. 4A, part c), thus demonstrating that the Flk1- and Pax2-expressing cells represent distinct compartments in the early kidney. Furthermore,staining for Pecam, a marker of mature endothelial cells(Newman, 1997), identified a small subset of the Flk1-positive cells(Fig. 4A, parts g,h,i),indicating that the majority of Flk1-expressing cells in the early kidney are angioblasts or immature endothelial cells.

Next it was determined whether the stimulation of kidney development by Vegfa required signaling through Flk1. Signaling by Flk1 was blocked by three independent methods, all of which yielded similar results. Addition of an Flk1-blocking antibody (DC101 from ImClone Systems)(Fig. 4B), addition of a chemical inhibitor of the kinase activity of Flk1 (tyrphostin SU1498) (data not shown) or microinjection and electroporation of a dominant-negative truncation mutant form of Flk1 (Tsou et al., 2002) (data not shown), all led to reduced branching of the ureteric bud and reduced numbers of induced nephrons(Fig. 4B).

To further examine the effect of angioblast-derived signals on the differentiation of the mesenchymal component of the early kidney, it was determined whether signaling through Flk1 is involved in maintaining expression of several genes characteristically expressed by the condensed mesenchyme. The expression of several known, and one novel, marker of the condensed mesenchyme was examined by whole-mount in situ hybridization in the presence or absence of the Flk1-blocking antibody. Pax2 expression in the mesenchyme was decreased in the presence of the Flk1-blocking antibody, as shown both by in situ hybridization and real-time PCR, the latter indicating an approximately fourfold decrease in expression level(Fig. 5A,B,K). Whether the expression of Pax2 in the ureteric bud was affected is difficult to determine using whole-mount in situ hybridization.

Pax2 has previously been shown to regulate the expression of Gdnf (Brophy et al.,2001); and, as would be predicted in the presence of decreased Pax2, Gdnf RNA levels measured by real-time PCR were also decreased in the presence of the Flk1-blocking antibody(Fig. 5L).

Osr1 (odd-skipped related 1) is a marker of the intermediate mesoderm that continues to be expressed in the condensed mesenchyme(So and Danielian, 1999). By contrast to those of Pax2, Osr1 expression levels appeared unchanged in the presence of the Flk1-blocking antibody(Fig. 5C,D), indicating that Flk1 blockade did not non-specifically affect expression of all genes in the organ culture.

Nanog, a homeobox-containing gene recently shown to be required to preserve the multipotent state of embryonic stem cells(Chambers et al., 2003; Mitsui et al., 2003) was also identified as a novel marker of the condensed mesenchyme, with an expression that decreased after treatment with Flk1-blocking antibody(Fig. 5E,F), suggesting that it may be not a marker of the most undifferentiated mesenchyme. Wnt4 is a signaling molecule required for the mesenchymal-to-epithelial transformation by which aggregates of condensed mesenchyme begin to form tubules(Stark et al., 1994). As would be predicted for a later marker involved in the differentiation of nephrons, Wnt4 expression also decreased in the presence of the Flk1-blocking antibody (Fig. 5G,H). By contrast to Pax2 expression, which overlaps Wt1 in the condensed mesenchyme, Wt1 expression itself did not decrease in the presence of Flk1 blockade. However, instead of the usual pattern of Wt1 expression, which is highest in the discrete mesenchymal condensates around each derivative of the ureteric bud(Fig. 5I) a more diffuse Wt1 expression pattern was apparent(Fig. 5J), indicative of the decreased branching and failure to form the usual discrete mesenchymal condensates around each tip of the ureteric bud derivatives. Similar results for each gene were obtained using SU1498, the pharmacological inhibitor of signaling through Flk1 (data not shown).

The reduced branching of the ureteric bud observed after blockade of signaling though Flk1 could be due to a signal from the angioblasts that acts directly on the ureteric bud, or that acts on the mesenchyme, to indirectly affect branching. To distinguish these possibilities, Vegfa was expressed by microinjection/electroporation at the periphery of the condensed mesenchyme. This resulted in a localized increase in Pax2 expression adjacent to the injection site (Fig. 6A,parts a,b), that was apparent within 6 hours of the injection, and that could be inhibited by either the Flk1-blocking antibody(Fig. 6A, parts d,e) or SU1498(data not shown). This observation is most consistent with the possibility that the Flk1-dependent signal is acting directly on the mesenchyme. By contrast, if the Flk1-dependent signal were acting directly on the ureteric bud, it would be expected that the ureteric bud would then induce higher levels of Pax2 in its usual pattern within the condensed mesenchyme around the ureteric bud.

As signaling through Flk1 is apparently required to maintain high levels of Pax2 RNA in the condensed mesenchyme, it can be hypothesized that reduced branching of the ureteric bud upon blockade of Flk1 signaling would be due to reduced expression of Pax2 and consequent decreased stimulation of Gdnf expression. To further examine the role of the Flk1-dependent signal, Gdnf expression in response to Pax2microinjection/electroporation in the condensed mesenchyme was examined, in the presence or absence of Flk1 blockade. Gdnf was abundantly expressed in response to Pax2, but Flk1 blockade eliminated Gdnf induction by Pax2(Fig. 6B).

To further examine the mechanism by which the Flk1-dependent signal may regulate Pax2, the levels of Pax2 RNA expression were measured in response to either a generalized blockade of transcription by actinomycin-D (act-D) or a specific blockade of Flk1 signaling by a pharmacological agent, SU1498. As measured by real-time PCR, levels of Pax2 decreased 1000-fold over an 8-hour treatment period with act-D(Fig. 6C). By contrast, on treatment with SU1498, Pax2 mRNA levels initially decreased but reached a steady-state level significantly higher than that resulting from treatment with act-D. No additive effects were observed when act-D and SU1498 were combined. At least three possible explanations can be offered to account for these observations: (1) the Flk1-dependent signal acts to stabilize Pax2 RNA in the mesenchyme, and the lower level observed after Flk1 blockade reflects a high-turnover state of non-stabilized Pax2 RNA;(2) the difference in Pax2 RNA levels between act-D treatment and Flk1 blockade is due to retention of ureteric bud expression of Pax2in the latter treatment, compared with complete loss of Pax2 RNA in the former treatment; (3) a third possibility, that the angioblast signal stimulates supra-basal transcription of Pax2, is also possible,although the failure of ectopically expressed Pax2 to induce expression of Gdnf under conditions of Flk1 blockade strongly suggests some degree of post-transcriptional regulation.

Organ culture experiments do not allow the use of nuclear run-on experiments to directly measure transcription rates. However, the hypothesis that Flk1-dependent signals regulate post-transcriptional regulation of Pax2 expression was tested further by examining Pax2expression itself in response to injection and electroporation of the Pax2 expression construct, in the presence or absence of inhibitors of Flk1. It was observed that when Flk1 inhibitors were present, much less Pax2 RNA or protein could be detected in response to Pax2injection (compare Fig. 6D,parts a,c and b,d; and Fig. 6D,parts e,g and f,h). By contrast, Flk1 blockade had no effect on expression of a Gfp construct expressed from the same CMV promoter-enhancer used to express Pax2 (Fig. 6D,parts i-l), demonstrating that Flk1 blockade is not simply inhibiting transcription from the CMV promoter used in the Pax2 (or Gfp) expression vector(s). Thus, the results of this last experiment are also consistent with post-transcriptional regulation of Pax2 by an Flk1-dependent signal.

To further test the hypothesis that the Flk1-dependent signal acts primarily on the mesenchyme, Gdnf was added to the organ culture medium in the presence of the Flk1 blockade. As Gdnf acts directly on the ureteric bud, this should be able to overcome any deficiency in the condensed mesenchyme due to the absence of the Flk1-dependent signal from the angioblasts. Addition of Gdnf in the presence of the Flk1 blockade stimulated branching of the ureteric bud (Fig. 7A). Surprisingly,addition of Gdnf also rescued induction of nephrons(Fig. 7A); this was not predicted because expression of Pax2 and Wnt4 were greatly decreased upon blockade of the Flk1-dependent signal. However, it was also observed that expression of Pax2 in the condensed mesenchyme was also rescued by addition of Gdnf (Fig. 7B). Together these results suggested that the Flk1-dependent signal was required to initiate kidney development, but that once Pax2 was able to stimulate expression of Gdnf, and Gdnf was able to elicit an inductive signal from the ureteric bud, the Flk1-dependent signal might no longer be required. As further support for this possibility, it was observed that addition of the Flk1-blocking antibody after organ cultures had already been established for 48 hours, had little effect on the extent of branching and nephron induction (Fig. 7C).

Previous studies on the induction of the metanephric kidney have considered the interaction of two cell lineages, the metanephric mesenchyme and the epithelial cells of the ureteric bud(Schedl and Hastie, 2000). Our results provide support for the requirement for a third cell type, the angioblast, which appears to be involved in maintaining the condensed mesenchyme in a differentiated state capable of responding to the inductive influence of the ureteric bud. We have shown that the Wt1 tumor suppressor gene, previously shown to be required for the initiation of kidney development, stimulates the expression of Vegfa within the mesenchyme. The target cells of Vegfa appear to be a population of Flk1-expressing angioblasts found at the periphery of the mesenchymal condensations. There appears to be a reciprocal set of interactions between the angioblasts and the mesenchyme that are required to maintain the mesenchyme in differentiated state characterized by high expression of Pax2 and Gdnf. Vegfa appears to be the major component of the mesenchyme-derived component of this reciprocal set of inductive signals,and the angioblast signal remains to be identified. However, once Gdnf expression has stimulated the initial branching of the ureteric bud and the induction of nephrons has begun, the Flk1-dependent signal appears not to be required, and kidney development is maintained by the interaction of the ureteric bud and the mesenchyme.

In embryos carrying a homozygous targeted mutation in the Wt1gene, the ureteric bud fails to grow out from the Wolffian duct, and there is consequent apoptosis of the metanephric mesenchyme(Kreidberg et al., 1993). It has been difficult to identify potential targets of Wt1 for which a deficiency could account for the renal agenesis phenotype, even though several studies have identified potential Wt1 target genes in the gonad,including Sry (Hammes et al.,2001; Hossain and Saunders,2001) and Sf1 (Zfp162 - Mouse Genome Informatics) (Wilhelm and Englert,2002). Low levels of Gdnf were detected in the metanephric mesenchyme of Wt1-deficient embryos(Donovan et al., 1999),suggesting that Gdnf may not be a target of Wt1, although it remains possible that Wt1 is involved in boosting levels of Gdnf to achieve bud outgrowth. More recently, siRNA to Wt1was used in an organ culture system to confirm the requirement for Wt1 in early kidney development, but this approach was less amenable for identifying novel target genes, because of the requirement for Wt1 to maintain cell viability(Davies et al., 2004). The present study suggests Vegfa as a novel candidate target gene for Wt1. The angioblast-mesenchyme interaction mediated by Vegfa appears to be crucial in maintaining the initial differentiated state of the mesenchyme, and in the absence of this interaction, it is likely that insufficient levels of Pax2 and Gdnf are present to elicit outgrowth of the ureteric bud.

The conclusion that Flk1-expressing angioblasts express a signal that is required to initiate kidney development depends on their identification as the major, if not sole, cell type expressing Flk1 in the early metanephric kidney. Immunostaining for Flk1 and the use of Flk1-LacZ knock-in mice has failed to detect expression in other cell types, and particularly not in the ureteric bud. Furthermore, the observations that (1) localized Vegfainjection/electroporation results in adjacent Pax2 expression, as opposed to enhanced Pax2 expression around the ureteric buds, and (2) addition of Gdnf can overcome the Flk1 blockade, are both more consistent with a model that invokes the requirement for angioblasts, rather than direct stimulation of the ureteric bud by Vegfa. However, the possibility that very low levels of Flk1,not detectable by immunostaining or the use of knock-in mice, are expressed by the ureteric bud and involved in Vegfa stimulation of branching, remains a possibility that might be examined by genetic ablation of the Flk1gene in the ureteric bud.

There are four major splice forms of the Wt1 gene, due to alternative splicing of exon 5, and alternative insertion of a three amino acid sequence, lysine-threonine-serine (KTS), at the end of exon 9(Haber et al., 1991). We recently showed that elimination of exon 5 had no effect on kidney development and function, and the role of this alternatively spliced exon remains unknown(Natoli et al., 2002b). Hammes et al. (Hammes et al., 2001)have published the results of gene targeting experiments that allowed expression of either the +KTS or -KTS forms of Wt1. In each case,homozygous mutant embryos showed defective kidney development, with poorly developed glomeruli, a phenotype more pronounced in the +KTS-only kidneys. The difference in these phenotypes raised the question of whether this indicated that the two splice forms of Wt1 have entirely distinct functions, or whether the two phenotypes are due to differences in severity, but due to the same function of the different splice forms. Supporting the former possibility are observations that the +KTS form of Wt1 associates with spliceosomes(Larsson et al., 1995), and has a speckled nuclear staining pattern, whereas the -KTS form has a more diffuse staining pattern in the nucleus. In our microinjection and electroporation system it has been observed that both the +KTS and -KTS forms of Wt1 are capable of inducing expression of Vegfa. However, this does not necessarily conflict with the likely possibility that the +KTS and-KTS forms of Wt1 have common functions early in kidney development but acquire distinct functions during podocyte differentiation. Indeed, both+KTS- and -KTS-only embryos apparently undergo normal early kidney development(Hammes et al., 2001),implying that they are interchangeable with regard to early functions of Wt1.

Previous studies examining liver and pancreas development have identified roles for the vascular system, and in particular Flk1-expressing cells in organogenesis (Lammert et al.,2001; Matsumoto et al.,2001). Thus it appears to be an emerging paradigm that organs require signals from angioblast-type cells to stimulate or maintain organ-specific patterns of differentiation. It is not known whether liver- and pancreas-associated angioblastic cells are responding to Vegfa produced by the mesenchymal components of these organs. It will also be of great interest to eventually determine whether these signal elaborated by pancreas-, liver- and kidney-associated angioblasts are identical or differ between these organs.

The authors thank: Drs Greg Dressler, Frank Isik, Frank Costantini, Seppo Vainio and Thomas Schultheiss for the sharing of plasmids and antibodies; and ImClone Systems for the DC101 antibody. This work was supported by a National Kidney Foundation Fellowship Grant to X.G. and by grants from the March of Dimes Birth Defects Foundation and the NIDDK to J.K. The authors also acknowledge support from the Jo Ann Webb Fund for Kidney Research and the Russo Family Charitable Foundation.