The ECM protein nephronectin promotes kidney development via integrinα8β1-mediated stimulation of Gdnf expression

The metanephric kidney arises through instructive signaling between two cell populations, a mesenchymal population known as the nephric cord and an epithelial population termed the nephric duct (ND). Both of these populations arise from the intermediate mesoderm on approximately embryonic day (E) 9.0 in the mouse. Within the nephric cord, in a region adjacent to the posterior hindlimb bud, there develops a special group of cells known as the metanephric mesenchyme (MM). At ∼E11, a signal(s) from the MM elicits the formation of the ureteric bud (UB) from the ND. Once formed, the UB extends toward and then invades the MM, and by ∼E11.5 the UB begins branching, thus giving rise to a tubular network that will eventually mature into the collecting ducts of the adult kidney. As the UB invades and branches, it expresses genes that induce the MM to condense and differentiate into various cell types that comprise the nephron. In the absence of invasion by the UB into the MM, the metanephric kidney does not develop (reviewed in Saxen, 1987).

Glial cell line-derived neurotrophic factor (GDNF), a member of the transforming growth factor-β (TGFβ) superfamily, is a key signal in the initiation of UB formation and subsequent branching(Costantini and Shakya, 2006). GDNF is expressed in the MM and signals to the UB through a receptor complex consisting of the receptor tyrosine kinase RET and a coreceptor, GDNF family receptor alpha1 (GFRα1). GDNF has been shown to induce ectopic UB formation in culture, and mice lacking either the Gdnf, Ret or Gfra1 genes display renal agenesis at high penetrance(Cacalano et al., 1998; Moore et al., 1996; Pichel et al., 1996; Sainio et al., 1997; Sanchez et al., 1996; Schuchardt et al., 1994). Gdnf expression in the MM is regulated by a highly conserved network of transcription factors, and can be activated or inhibited by extracellular signaling molecules such as WNT11, GDF11 and SLIT2 at discrete stages of development (Brodbeck and Englert,2004; Esquela and Lee,2003; Grieshammer et al.,2004; Majumdar et al.,2003). It is also known that GDNF signaling is dependent on heparan sulphate glycosaminoglycans, which interact with or are constituents of the extracellular matrix (ECM) (Barnett et al., 2002).

Integrins are cell adhesion receptors that serve as a link between the ECM and the cytoskeleton. Integrins are heterodimers, consisting of two single-pass transmembrane proteins designated as the α and βsubunits. Integrins are thought to occupy either an inactive or active state depending on cues received from the extracelluar environment in the form of an appropriate ECM ligand or growth factor signal. They communicate information from the ECM to the cell interior through recruitment of cytoskeletal proteins and kinases. In addition, they are able to communicate information from the cell interior to the ECM, resulting in its proper deposition and remodeling(Delon and Brown, 2007; ffrench-Constant and Colognato,2004; Giancotti and Ruoslahti,1999; Hynes,2002).

Previously, we showed that the integrin subunit α8 (Itga8)is expressed throughout the nephric cord, including the MM(Muller et al., 1997). Furthermore, we demonstrated that loss of Itga8 function invariably results in a delay of invasion of the MM by the UB, which in turn results in a high frequency of kidney agenesis. The molecular mechanism by which Itga8 function in the MM influences the UB has yet to be determined. Several ECM constituents are known ligands for α8β1 integrin,including fibronectin, osteopontin, tenascin C and vitronectin(Denda et al., 1998a; Denda et al., 1998b; Varnum-Finney et al., 1995). However, data from expression and loss-of-function analyses indicated that these ligands are unlikely to be mediators of α8β1 integrin function in the developing kidney. Using an expression cloning strategy we identified a gene that encodes a novel ECM molecule, nephronectin(Npnt), that is expressed by the UB and the epithelia of several developing organs (Brandenberger et al.,2001). We demonstrated that α8β1 integrin recognizes nephronectin in binding assays and associates with nephronectin in vivo. In addition, we found that the localization of nephronectin in the kidney is consistent with it mediating α8β1 integrin function during development.

Here we report that mice lacking nephronectin frequently display kidney agenesis. We show that the phenotype arises during the early stages of metanephric development, when the UB is beginning to invade the MM, similar to the phenotype of mice lacking Itga8 (and therefore α8β1 integrin function). Thus, nephronectin is an ECM protein expressed by the UB that is required for α8β1 integrin function during early stages of UB invasion and branching. Significantly, we demonstrate that Gdnfexpression is reduced in both Npnt and Itga8 mutants at the time when the invasion of the MM by the UB is delayed. Finally, we present genetic data indicating that Gdnf dosage as well as signaling from the receptor tyrosine kinase, RET, impact the penetrance of the Itga8mutant phenotype. Taken together, our results suggest that the observed reduction of Gdnf expression in the MM is sufficient to explain the phenotypes observed in mice lacking either nephronectin or the α8 integrin subunit.

A probe derived from the 5′ region of the Npnt locus was used to screen a mouse 129Sv/J BAC library, RPCI-22 (Roswell Park Cancer Institute). One of 9 clones isolated was used to generate the floxneo targeting construct (Fig. 1B-D)by recombineering (Zhang et al.,1998). Clones harboring BACs that were correctly modified were identified by Southern blot hybridization, which revealed the presence of both the 5′ and 3′ arms. One clone, designated 273P10 NN-1, was electroporated into E14 embryonic stem (ES) cells.

Because of the large size of the targeting construct (300 kb), we used TaqMan (Applied Biosystems) real-time quantitative PCR to identify ES cells in which there was a reduction in copy number of a region that was targeted for insertion (Valenzuela et al.,2003). This involved the use of specific PCRs that amplify the wild-type allele, but fail to amplify the mutated allele because of the increase in sequence length. Candidate ES cell clones that were identified by sequential PCR screens were assessed for reduced wild-type copy number by quantitative Southern hybridization using densitometry. Two clones isolated from the screen were injected into C57BL/6 blastocysts (Transgenic/Targeted Mutagenesis Core, University of California, San Francisco, CA), and both were incorporated into the germ line.

We generated a null allele for nephronectin by crossing animals heterozygous for the Npnt^floxneo allele with animals expressing CRE recombinase under the β-actin promoter(Lewandoski et al., 1997). This cross produced mice carrying the Npnt^Δex1 allele in which the first exon of Npnt was excised without deletion of the neomycin-resistance gene expression cassette(Fig. 1D,E). All the analysis described here was performed on mice carrying this allele on a mixed genetic background (129Sv/J; C57BL/6; FVB/N).

To stage embryos, noon of the day on which a vaginal plug was detected was considered E0.5. Embryos were collected at various stages and the region containing the hindlimb buds was fixed in 4.0% paraformaldehyde(PFA)/phosphate-buffered saline (PBS) overnight at 4°C and cryosectioned at 14 μm. Analysis of gene expression using in situ hybridization with RNA probes was performed according to standard protocols. Data using the following probes are presented: Gdnf(Srinivas et al., 1999), Eya1 (Xu et al.,1999), Six2 (Xu et al., 2003), Pax2(Dressler et al., 1990).

Embryos at various stages and kidneys from newborn animals were fixed in 4.0% PFA/PBS overnight at 4°C. Tissues were cryosectioned and stained with hematoxylin and eosin according to standard protocols. Sections were stained with the following antibodies: anti-nephronectin (1:100)(Brandenberger et al., 2001),anti-EHS laminin (1:500) (Sigma L9393), anti-fibronectin (1:300) (Sigma F-6140), anti-calbindin D28K (1:600) (Swant CB-38a), anti-collagen type IV(1:500) (LB-1403; Cosmo Bio., Tokyo, Japan), and anti-pax2 (1:100) (PRB-276P;Covance, Princeton, NJ). Confocal imaging was performed on a Zeiss LSM 5 Pascal.

We generated a line of mice carrying, Npnt^Δex1, which lacks the first exon of nephronectin (see Materials and methods; Fig. 1). Animals homozygous for Npnt^Δex1 express neither nephronectin protein nor Npnt RNA (see Fig. 1G,H), demonstrating that Npnt^Δex1 is a null allele. Npnt^Δex1 homozygotes, hereafter referred to as Npnt mutants, were born at the expected Mendelian frequency (52 out of 219 mice; 24%). At birth they appeared to be of normal size without any obvious external defects. However, upon dissection, we found that Npnt mutant animals (n=52) frequently lacked one (23%)or both (46%) kidneys (Fig. 2A-C). The remaining Npnt mutants had two kidneys (31%),on average resulting in 58% agenesis (see legend to Fig. 2F). By contrast, 2 of 109 Npnt^Δex1 heterozygotes displayed bilateral kidney agenesis and none displayed unilateral agenesis (2%agenesis). No agenesis was observed in their wild-type littermates. In most cases the Npnt mutant kidneys were smaller than those of their wild-type littermates (Fig. 2A,B). Histological analysis revealed variability in development,with most Npnt mutant kidneys containing essentially normal nephrogenic regions (Fig. 2D,E), and a few displaying cystic and dysplastic elements (not shown). Significantly, Npnt mutants lacking kidneys did not show evidence of partial ureter development or survival of MM, suggesting that development is perturbed at an early stage, when the UB has yet to invade the MM.

To determine the developmental origin of the kidney agenesis observed at birth, we examined Npnt mutants and their wild-type littermates at early stages of metanephric development. At E11.0, the UB had formed and appeared similar in Npnt mutants (n=2) and their wild-type littermates (Fig. 3A,B). However, by E11.5, we found a significant difference: whereas in all wild-type embryos the UB had invaded the MM and branched, none of the mutant embryos examined, in wholemount or serial section (n=12), showed any evidence of UB invasion of the MM or branching, on either side of the embryo(Fig. 3C,D). By E12.5, the UB had invaded the MM and undergone some branching in all Npnt mutants examined (n=4). However, the extent of branching was variable,ranging from a single branching event to two rounds of branching, and therefore fewer than the three rounds in their wild-type littermates(Fig. 3E-G).

To determine whether the UB branching observed was sufficient to promote nephrogenesis, we analyzed kidney development at E13.5. In two of four Npnt mutants examined, we found that kidneys were developing on both sides in one embryo and on one side in the other. In all cases, the mutant kidneys were markedly smaller than normal. In the other two of four Npnt mutants, no kidney development was detected on either side and the MM was beginning to atrophy (five out of eight kidneys expected did not form), whereas kidneys were developing on both sides in four of four wild-type littermates (no agenesis) (Fig. 3H-K′, and data not shown). It seems likely that the three kidneys that were observed at E13.5 developed in embryos like those in which the UB had undergone several rounds of branching by E12.5 (see Fig. 3F). By contrast, the failure of kidney development at E13.5 may have occurred in embryos like those in which only a single branching event had occurred by E12.5 (see Fig. 3G). From these data it appears that in Npnt mutants, UB formation is initially normal, but then the UB fails to invade the MM at E11.5. Subsequently, invasion/branching does occur, but in many cases it appears that the extent of branching is too little to sustain kidney development, resulting in the kidney agenesis observed at birth. Significantly, the lack of invasion of the MM by the UB at E11.5 in the Npnt mutants appeared similar to that observed in embryos lacking α8β1 integrin function(Muller et al., 1997).

The ability of a structure such as the UB, a group of polarized epithelial cells, to invade the adjacent mesenchyme depends on proper remodeling of the basement membrane and other matrix constituents that surround the epithelial structures. We therefore were interested in determining whether the basement membranes surrounding the ND and the UB were abnormal in Npntmutants. To assess this, we examined two of the core components of the basement membrane, laminin (LN) and collagen IV (COL IV). At E11.5, when the mutant phenotype becomes obvious, staining for either LN or COL IV revealed that the basement membranes surrounding the ND and the UB appeared similar in Npnt mutants and their wild-type littermates (n=3)(Fig. 4A-F). In a Npntmutant in which the UB had invaded and branched at E13.5, the basement membrane likewise appeared similar to that in the wild-type controls, as assessed with antibodies to LN and COL IV(Fig. 4G-J). To assess the ability of the basement membrane to associate with other ECM proteins, we examined the distribution of fibronectin, an ECM protein that associates with,but is not a core component of, the basement membrane(Hynes, 1986). At E13.5 we found that the localization of this protein was similar in both a Npnt mutant and a wild-type littermate(Fig. 4K,L). From these results, we conclude that the basement membrane is comparatively normal in Npnt mutants.

The above results suggested that, rather than ensuring a normal basement membrane, nephronectin recognition by α8β1 integrin may facilitate key signaling events within the MM that promote kidney development. To investigate this possibility, we analyzed the expression, by in situ hybridization and antibody staining, of various genes known to have roles in kidney development. Because a lack of invasion of the MM by the UB was invariably observed in both Npnt- and Itga8-null mutants at E11.5, we performed our analysis at this stage. We examined the expression of key genes expressed in the ND and UB, including Pax2, Ret, Gfra1 and Emx2 and found that they all were expressed at comparable levels in both Npnt mutants and wild-type controls at E11.5(Fig. 5A-H, and data not shown). The expression of genes that mark the MM was also assessed, including Gdnf, Eya1, Six2 and Pax2. Interestingly, we did not detect Gdnf RNA in the Npnt mutant MM. Importantly, however, the level of Gdnf expression in the adjacent limb bud was comparable in the mutant and control embryos (Fig. 5A,B). Similar results were obtained in seven separate experiments with seven separate mutant embryos. The expression domains and levels of Eya1 and Six2 RNA and PAX2 protein, which are expressed in the MM and are known to have roles in controlling Gdnf expression(Brodbeck et al., 2004; Brophy et al., 2001; Xu et al., 1999), were similar in Npnt mutants and their wild-type littermates(Fig. 5C-H). Taken together,these results demonstrate that, at E11.5, the MM is present in Npntmutants and displays normal expression of several genes known to be upstream of Gdnf, and provide evidence that nephronectin has a role in promoting Gdnf expression in the MM at this stage.

To determine the stage-specificity of this effect, we examined expression of Gdnf and Eya1 at E10.5, before the phenotype becomes obvious, and at E13.5, when kidney development is progressing in some Npnt mutant embryos. Remarkably, we found that Gdnf RNA levels appeared normal in the two mutant embryos examined at E10.5(Fig. 5I-L) and in a mutant embryo in which kidney development was observed at E13.5(Fig. 5M-P). We could not determine whether Gdnf was also expressed in those mutants in which UB invasion and branching was not detected at E13.5 because the MM had begun to degenerate. These data suggest that nephronectin is transiently required for normal Gdnf expression in the MM at E11.5, the time when the phenotype in Npnt mutants is first obvious and is invariably observed.

In view of the apparent similarity between the kidney phenotypes of Npnt- and Itga8-null mice, we were interested to examine Gdnf expression in Itga8-null homozygotes, hereafter referred to as Itga8 mutants. We compared the expression at E11.5 of Gdnf and several other genes expressed in the MM, including the signaling molecule genes Bmp4 and Gdf11 and the transcription factor genes Eya1, Pax2, Six2, Wt1 and Sall1,in Itga8 mutants and their control littermates. Consistent with what we observed in Npnt mutants at E11.5, we detected no change in expression of any of these genes in Itga8 mutants at E11.5 except Gdnf (Fig. 6A-D and data not shown). Again, Gdnf RNA was not detected in the MM, but Gdnf expression appeared comparable in the adjacent limb buds of Itga8 mutant and control embryos(Fig. 6A,B). Similar results were obtained in five separate experiments with five separate mutant embryos. However, using a radiolabeled Gdnf probe we were able to detect a signal at low level (data not shown). In addition, in situ hybridization assays revealed no obvious difference between the Itga8 mutants and controls at E10.5 and E13.5 (Fig. 6E-L), indicating that, as in Npnt mutants, α8 integrin is transiently required for normal Gdnf expression in the MM at E11.5.

In order to test the hypothesis that the kidney agenesis we observed in Npnt and Itga8 mutants is because of a reduction in Gdnf expression, we took a genetic approach. One prediction of this hypothesis is that reducing the dosage of the Gdnf gene should increase the penetrance of the mutant kidney phenotype. For these studies we performed crosses to produce Itga8-null heterozygotes and homozygotes carrying one copy of a Gdnf-null allele, Gdnf^lacZ(Moore et al., 1996). To assess kidney development we scored kidney agenesis as described in the legend to Fig. 1F. We found that animals heterozygous for either Itga8- or Gdnf-null alleles displayed 0% and 9% agenesis, respectively. These data are consistent with previously published frequencies of kidney agenesis in these mutant heterozygotes (Moore et al.,1996; Muller et al.,1997). However, in animals heterozygous for both the Itga8- and Gdnf-null alleles(Itga8^+/-;Gdnf^lacZ/+), we observed 53%agenesis (Table 1). Furthermore, the frequency of kidney agenesis in animals homozygous for the Itga8-null allele was increased from 59% to 96% by heterozygosity for the Gdnf-null allele (Table 1). The one kidney that was found in the 11 Itga8^-/-;Gdnf^lacZ/+ animals examined was reduced in size and histological analysis revealed it to be highly dysplastic and most likely non-functional (data not shown). These results support the hypothesis that the reduction in Gdnf expression that results from a loss of α8β1 integrin function is responsible for the kidney agenesis in Npnt- and Itga8-null mutants.

To further test that hypothesis, we sought to rescue kidney development in Itga8-null mice by enhancing the GDNF signaling pathway in vivo. To do this, we made use of a null allele of the Sprouty1 (Spry1) gene,which has been shown to be involved in antagonizing the function in the UB of RET, the receptor for GDNF (Basson et al.,2005; Basson et al.,2006). Spry1-null mutants display ectopic ureteric budding from the ND, resulting in multiple ureters and multiplex kidneys as a consequence of excess GDNF signaling. We reasoned that if the kidney agenesis phenotype in the Itga8 mutants was due primarily to reduced Gdnf expression, it should be possible to overcome this deficiency by reducing Spry1 gene dosage and thereby increasing sensitivity of the UB to the small amount of GDNF produced in the Itga8 MM at E11.5. Indeed, we observed that heterozygosity for a Spry1-null allele in Itga8-null animals resulted in a significant rescue of kidney development: 25% agenesis in Itga8^-/-;Spry1^+/- animals(n=20) versus 71% agenesis in their Itga8^-/-;Spry1^+/+ littermates(n=14). In general, Itga8^-/-;Spry1^+/- kidneys appeared to be of normal size. Furthermore, no kidney agenesis was observed in animals homozygous for both the Itga8- and Spry1-null alleles(n=8) (Fig. 7A). These data provide strong support for the hypothesis that the kidney agenesis that results from the absence of the α8 integrin subunit is because of a reduction in Gdnf expression.

Histological analysis revealed that the rescue of the Itga8-null phenotype by reduction of Spry1 gene dosage takes place early in kidney development. Like Npnt mutants, all Itga8-null embryos displayed a lack of invasion of the MM by the UB at E11.5; however, in all Itga8-null embryos that were heterozygous for a Spry1-null allele (n=5) the UB on one or both sides was found to have invaded the MM at this stage (9/10 UBs invaded)(Fig. 7B,C). This finding provided us with an opportunity to address an important question: is the reduction in Gdnf expression that we observed at E11.5 in Itga8 mutants secondary to the lack of invasion of the MM by the UB at E11.5? The latter is a viable possibility because the UB is known to promote Gdnf expression in the MM at later stages by producing signaling molecules such as WNT11(Majumdar et al., 2003).

If the observed reduction in Gdnf expression was because of the lack of signals from the UB, one would predict that in those Itga8mutants in which UB invasion was rescued by reducing Spry1 gene dosage, Gdnf expression would be restored to the level found in wild-type embryos. However, when we assessed Gdnf expression in the MM of Itga8^-/-;Spry1^+/- embryos at E11.5, we found that the level of Gdnf RNA was substantially reduced compared with that in their control(Itga8^+/-;Spry1^+/-) littermates(Fig. 7D,E). By contrast, Eya1 expression was similar in the rescued Itga8 mutants and their control littermates (Fig. 7F,G). These data demonstrate that Gdnf expression in the MM is still reduced by the absence of the α8 integrin subunit in the rescued embryos. However, the signal was stronger than that found in Itga8-null embryonic kidneys (Fig. 7B), most likely because the presence of the UB in the rescued mutants produces signals that enhance Gdnf expression(Majumdar et al., 2003).

The importance of the ECM in the developing metanephric kidney has been appreciated since the kidney first began to be used as a model for organogenesis (Grobstein,1955). However, loss-of-function studies have revealed that few ECM proteins or their receptors have specific roles in the early events of metanephric development (Lelongt and Ronco, 2003). One exception is α8β1 integrin, which we have previously shown to be crucial for initial UB invasion into the MM and its subsequent branching, as well as for epithelialization of the MM(Muller et al., 1997). Our finding that Npnt-null embryos phenocopy the early phenotype of Itga8-null embryos (Muller et al., 1997) strongly suggests that nephronectin is the ligand mediating α8β1 integrin function during the early events of kidney development. In addition, we have found that the expression of Gdnf,a key factor in kidney development, is severely reduced in Npnt as well as in Itga8 mutant MM at E11.5, the time at which we invariably observed a lack of invasion of the MM by the UB in both mutants. These data suggested that α8β1 integrin might be involved in a pathway upstream of Gdnf expression. Our genetic analysis provides strong support for this hypothesis. It therefore appears that we have uncovered a novel pathway in which an integrin, through recognition of its ECM ligand, is directly involved in the regulation of an essential signaling molecule in the developing kidney.

Assessment of kidney development at birth in Npnt mutants revealed a high frequency of agenesis, with 46% of the mutants displaying bilateral agenesis and 23% unilateral agenesis. Only 31% of mutant animals had two kidneys, most of which were smaller than normal but otherwise appeared unaffected. Surprisingly, although nephronectin is expressed in several other organ systems during development, preliminary analysis has shown that all organs except the kidney appear grossly normal in Npnt mutants at birth. Consistent with this finding, Npnt mutants that survive beyond birth are healthy and fertile, and have an apparently normal life span. This suggests that in tissues other than the developing kidney, the presence of other ECM proteins compensates for the absence of nephronectin. Among the other ECM proteins that could replace nephronectin is Mam domain and EGf domain-containing protein (MAEG) (Buchner et al., 2000), which shares 41% overall amino acid identity with nephronectin and has been shown to be a ligand for α8β1 integrin(Osada et al., 2005). However,at present, little is known about MAEG function and expression, and it remains to be determined whether this ECM protein has roles in organogenesis.

Our analysis has revealed that although a UB forms in Npntmutants, it consistently fails to invade the MM at E11.5. Significantly, this phenotype very closely resembles the early phenotype of Itga8 mutants(Muller et al., 1997)(Fig. 6D). Since its identification, nephronectin has been a candidate ligand for α8β1 integrin in the developing kidney and this similarity in phenotype strongly points to nephronectin as an essential ligand for α8β1 integrin during the crucial early process of UB invasion.

Although the Npnt and Itga8 mutant phenotypes appear very similar at E11.5, there are some important differences at later stages. One is that Npnt mutants display kidney agenesis at a lower frequency than Itga8 mutants, 58% versus 83% agenesis, respectively(Muller et al., 1997)(Fig. 2F). A possible explanation for this is that there may be functional redundancy with another ligand(s) expressed by the UB, which can be recognized by α8β1 integrin and can mediate responsiveness of the MM. If so, the expression of this ligand might be responsible for enabling the UB in Npnt mutants to undergo the delayed invasion and branching that we observed at E12.5, which in some cases must be sufficient for kidney formation. By contrast, Itga8 mutants should be unable to respond to any ligand, and therefore display complete agenesis. The finding that kidneys occasionally form in Itga8 mutants raises the possibility that another integrin may compensate for the absence of α8β1 integrin.

Differences between the Npnt- and Itga8-null mutants might also reflect differences in the genetic backgrounds of the mice. Although the background of the Itga8 mutants was largely C57BL/6,with some minor contribution remaining from 129Sv/J, the Npnt allele has been maintained on a mixed background with contributions from C57BL/6,129Sv/J and FvB/N. In support of this explanation, we have observed that Itga8 mutant survival increases dramatically on an outbred background(J.M.L. and L.F.R., unpublished). Once the Npnt-null allele has been bred onto a pure background the penetrance of the homozygous phenotype may more closely resemble that in Itga8 mutants.

Here we have presented data that support a role for α8β1 integrin and its ligand nephronectin in a pathway that regulates the expression of Gdnf, an essential growth factor in kidney development. Using in situ hybridization, we have shown that Gdnf expression is severely reduced in Npnt- and Itga8-null embryos at a time when we invariably found that the UB has not invaded the MM. We have demonstrated that Itga8;Gdnf-compound-null heterozygotes display kidney agenesis at a fivefold higher frequency than is observed in Gdnf-null heterozygotes, and that reducing the level of Gdnfgene dosage increases the penetrance of the Itga8-null phenotype. Furthermore, we found that by reducing the dosage of a gene that encodes an attenuator of GDNF signaling, Sprouty1, and thus enhancing the sensitivity of Itga8-null mutants to GDNF, we decreased the penetrance of the Itga8-null phenotype. Taken together, these data provide genetic evidence that α8β1 integrin and GDNF function in a common pathway and suggest that α8β1 integrin and its ECM ligand play a role in regulating the expression of Gdnf.

Of special interest, our results show that the severe reduction in Gdnf expression in Npnt and Itga8 mutants is transient: in Npnt and Itga8 mutants at E10.5, GdnfRNA levels appeared normal, at E11.5 Gdnf RNA was barely detectable,and at E13.5 Gdnf RNA was readily detected in those mutants in which sufficient UB branching had occurred such that kidney development proceeded. This transient effect may be indicative of multiple factors working at different times during kidney development to produce the normal pattern of Gdnf expression. For example, WNT11, which has been shown to maintain Gdnf expression in the MM, seems to be required only after UB invasion (Majumdar et al.,2003). According to this hypothesis, lack of either nephronectin or α8β1 integrin results in a severe decrease in Gdnfexpression, which causes a delay in UB invasion that is subsequently overcome by the presence of other factors, possibly WNT11, or perhaps members of the fibroblast growth factor (FGF) or TGF-β families, which may have facilitating roles in regulating Gdnf expression.

An alternative explanation for the reduction in Gdnf expression is that it is a secondary effect of the absence of the UB from the MM in Npnt and Itga8 mutants. We have addressed this possibility by assaying for Gdnf expression in Itga8^-/-;Spry1^+/- embryos, in whichα8β1 integrin function is lacking but the UB has invaded the MM at E11.5. We found that in these Itga8^-/-;Spry1^+/- embryos the level of Gdnf expression at E11.5 was substantially reduced compared with that in their Itga8^+/-;Spry1^+/-littermates. This result demonstrates that loss of α8β1 integrin causes a substantial decrease in Gdnf expression in the MM even in the presence of a UB and, therefore, strongly supports our hypothesis that the recognition of nephronectin by α8β1 integrin in the developing kidney is necessary for robust Gdnf expression.

How might an integrin and its ECM ligand regulate Gdnf expression?Integrins are classically known as adhesion receptors, which have been shown to play roles in organizing the cytoskeleton and activating intercellular signaling pathways (ffrench-Constant and Colognato, 2004; Humphries et al., 2004; Hynes,2002). There is an extensive literature demonstrating that in cell culture, integrin-mediated cell adhesion together with growth factor signaling can promote mitogenesis, cell viability and gene expression(ffrench-Constant and Colognato,2004; Giancotti and Ruoslahti,1999). In mammary gland cultures, β1 integrins have been shown to synergize with prolactin signaling to activate Stat5 and thus to play a role in maintaining the differentiated state of the glandular epithelium and its expression of β-casein (Akhtar and Streuli, 2006; Faraldo et al.,1998; Naylor et al.,2005).

With respect to α8β1 integrin, it has been shown that its recognition of fibronectin activates both the MAPK and PI3K pathways in cell culture systems (Farias et al.,2005). These data raise the possibility that in the kidney,α8β1 integrin activates the MAPK cascade in the MM. In support of this, we have observed reduced levels of phospho-ERK in the MM of Itga8 mutants (J.M.L. and L.F.R., unpublished). Therefore, it is conceivable that signaling by α8β1 integrin synergizes with a growth factor signal in the MM to activate the MAPK cascade that then impinges on the transcriptional network involved in regulating Gdnfexpression.

The signaling properties of integrins have been appreciated for some time,but there is as yet very little in vivo evidence demonstrating roles for integrin signaling in regulating gene expression. The data presented here suggest that this is a key function of integrin in the developing kidney. Further studies will be needed to identify the specific signaling pathway(s)and target genes involved.

We thank Zhen Huang and Nigel Killeen for advice and guidance in working with BACs and quantitative PCR; Ayako Kuroda for ES cell culture and blastocyst injection and Natasha Shinsky-Bjorde for technical assistance;Denise Marciano, Dean Sheppard, Uta Grieshammer and members of the Reichardt laboratory for discussion and comments on the manuscript; Frank Costantini,Greg Dressler, Richard Maas and Pin Xu for providing the plasmids used to prepare in situ hybridization probes. This work has been supported by NIH grants R37 HD25331 and 5 R01 DK64338. L.F.R. is an investigator of the Howard Hughes Medical Institute.