Dominant effects of RET receptor misexpression and ligand-independent RET signaling on ureteric bud development

Development of the metanephric kidney has long been known to depend on reciprocal inductive interactions between the ureteric bud (UB) and the metanephric mesenchyme (Saxen, 1987). While the mesenchyme induces the growth and branching of the UB, which gives rise to the renal collecting system (Erickson, 1968; Grobstein, 1953, 1955), the UB branches induce the surrounding mesenchymal cells to condense into epithelial vesicles, which differentiate into the various segments of the nephron (Grobstein, 1953, 1955; Saxen, 1970). While nephrogenesis has been intensively studied for many years, until recently less attention has been paid to the developmental control of UB growth and branching. However, this process is equally critical for normal renal development: as the kidney develops, each new branch of the UB induces new nephrons, so that the histoarchitecture of the mature kidney is determined largely by the pattern of growth and branching of the UB (Ekblom, 1992).

Gene expression studies (reviewed by Bard et al., 1994) have suggested a number of growth factors and receptors that may play a role in UB development, and the importance of several such gene products has been demonstrated through gene knock-out studies or through inhibition of gene expression in organ culture systems (reviewed by Lechner and Dressler, 1997). The first gene with a demonstrated role in UB development was the proto-oncogene RET, which encodes a receptor tyrosine kinase (Takahashi and Cooper, 1987). RET is initially expressed throughout the Wolffian duct, and when the ureteric bud evaginates from the Wolffian duct and branches within the metanephric mesenchyme, RET expression continues throughout the entire UB epithelium. By the time the UB has branched several times, expression of RET is restricted to the distal tips of the UB in the peripheral nephrogenic zone (Pachnis et al., 1993; Tsuzuki et al., 1995). In RET^−/− mouse embryos, the UB either fails to form, resulting in renal agenesis, or its growth and branching are severely retarded leading to hypodysplasia (Schuchardt et al., 1994, 1996). This led to the proposal that RET serves as the receptor for a ligand, produced by the metanephric mesenchyme, which stimulates the outgrowth of the UB from the Wolffian duct as well as its subsequent growth and repeated branching (Schuchardt et al., 1996).

This model was substantiated by the discovery of a RET ligand, glial cell line-derived neurotrophic factor (GDNF) (Durbec et al., 1996; Lin et al., 1993; Robertson and Mason, 1997; Trupp et al., 1996; Vega et al., 1996; Worby et al., 1996), which is expressed in the metanephric mesenchyme (Hellmich et al., 1996; Miyamoto et al., 1997; Suvanto et al., 1996). GDNF binds to RET in conjunction with GFRα-1, a glycosylphosphatidylinositol(GPI) linked cell surface coreceptor (Jing et al., 1996; Treanor et al., 1996). In the developing kidney, GFRα-1 mRNA is expressed at the highest levels in the peripheral portion of the UB, although in a broader domain than RET (Baloh et al., 1997; Sainio et al., 1997), and at lower levels in condensed mesenchyme and early epithelial tubules (Sainio et al., 1997; Suvanto et al., 1997). Soluble GFRα-1, produced by cleaving the GPI linkage, can also mediate the binding of GDNF to RET (Jing et al., 1996; Treanor et al., 1996). Targeted mutagenesis of either GDNF (Moore et al., 1996; Pichel et al., 1996; Sanchez et al., 1996) or GFRα-1 (Enomoto et al., 1998) results in bilateral renal agenesis in homozygotes. These observations have established that GDNF is an inductive factor produced by the metanephric mesenchyme, which signals through GFRα-1 and RET, thus stimulating the evagination of the UB from the Wolffian duct, and its subsequent growth and branching within the metanephros.

While these studies have demonstrated that GDNF/RET signaling is required for normal UB development, the specific consequences of this signal for the UB remain to be elucidated. Does the GDNF/RET signal promote survival, proliferation, or differentiation of the cells at the tips of the developing UB? Does it stimulate branching, or influence the specific pattern of growth and branching? While several recent studies have attempted to address these questions through the addition of GDNF to kidney rudiments growing in organ culture (Pepicelli et al., 1997; Sainio et al., 1997; Vega et al., 1996), they remain to be fully resolved. In this paper, we address the related question of the significance of the restricted sites of expression of RET and GDNF. After the first few branches of the UB, expression of RET is restricted to the distal tips of the UB branches, the sites at which new UB branches are formed. GDNF is expressed in the undifferentiated metanephric mesenchyme, which is also localized in the peripheral nephrogenic zone. It has been observed that the early UB will grow toward localized sources of GDNF (Durbec et al., 1996; Pepicelli et al., 1997; Sainio et al., 1997), and that addition of soluble GDNF to organ cultures results in increased numbers of UB tips (Pepicelli et al., 1997; Vega et al., 1996). Together, these observations suggested that the restricted expression of RET and GDNF might be important to regulate the extent, and possibly the pattern, of UB growth and branching.

To examine this question, we have produced transgenic mice in which a wild-type RET cDNA is expressed under the control of the Hoxb7 promoter. Unlike RET, Hoxb7 is expressed in a constitutive pattern throughout the developing UB, and its expression persists in the collecting ducts, pelvis and ureter in the adult kidney. Hoxb7 promoter sequences have been shown to direct expression of a lacZ reporter gene in the same pattern as the endogenous Hoxb7 gene (Kress et al., 1990; Vogels et al., 1993). Our hypothesis was that constitutive expression of wild-type RET would cause excessive or ectopic branching of the ureteric bud, and thus disrupt or transform the normal histoarchitecture of the kidney. However, we were concerned that RET receptors ectopically expressed in the UB trunk regions might be ineffective if GDNF were restricted to the periphery of the kidney. Therefore, we also produced transgenic mice in which a ligand-independent, constitutively active form of RET was expressed under the Hoxb7 promoter.

All constructs were injected into (B6×CBA)F₂ zygotes, and transgenic mice were bred to the same F₁ hybrid strain, or to RET+/− mice (on a mixed 129/SvEv and MF1 background). GDNF+/− mice (Sanchez et al., 1996) were obtained from Dr M. Barbacid (Bristol-Myers Squibb).

The Hoxb7 promoter fragment (sequences from −1316 to +81 of the Hoxb7 gene), obtained from Dr Jacqueline Deschamps (Kress et al., 1990), was excised from the pGEM-Blue vector with KpnI and AvaI, the ends blunted, and recloned into the EcoRV site of pBluescript KS⁺ in both orientations. A cDNA encoding the RET9 isoform, cloned into the EcoRI and SacI sites of pBluescript KS⁺, was provided by Dr Vassilis Pachnis. The PTC2 RET cDNA (Bongarzone et al., 1993), cloned into the XbaI site of the pMAM-neo vector, was obtained from Dr Marco Pierotti. The human β-globin sequence was a 1.7 kb BamHI/PstI fragment that included the last 20 base pairs of exon2, all of intron 2 and exon 3, and the polyadenylation signal. This fragment was cloned into the BamHI and PstI sites of pBluescript KS⁺.

To make the Hoxb7/RET9 construct, the human β-globin sequence was excised with BamHI and ClaI, and the BamHI site blunted with Klenow DNA polymerase. This fragment was ligated downstream of the RET cDNA by digesting the RET clone with EcoRV and ClaI. The ligated RET and globin sequences were subsequently excised with SacII and ClaI, and the SacII end was blunted. This fragment was ligated downstream of the Hoxb7 promoter by cutting the vector with HindIII and ClaI (the HindIII site was blunted prior to ligation). The entire transgene was excised for microinjection using XbaI and ClaI. The RET cDNA portion of the final RET9 construct was sequenced using multiple overlapping primers generated on the basis of the published sequence.

The PTC2 cDNA was excised from pMAM-neo by partial digestion with EcoRI, digested with DraI to delete the 3′ poly(A) tail, and cloned into pBluescript KS⁺ cut with EcoRI and SmaI. To produce the transgene, a NotI/HindIII Hoxb7 fragment and a HindIII/SpeI PTC2 fragment were ligated into a human β-globin-containing pBluescript KS⁺ vector cut with SpeI and NotI. The entire transgene was then excised with KpnI and NotI.

Transgene DNA for microinjection was prepared as described by Hogan et al. (1994). Briefly, the transgene fragment was gel purified and then subject to CsCl gradient centrifugation. Peak fractions were dialyzed against a vast excess of injection buffer (10 mM Tris-HCl, 0.1 mM EDTA) at 4°C for >48 hours. The transgene solution was then centrifuged through ‘Ultrafree-MC’ centrifugal filtration units (Millipore) to remove particulate impurities. The DNA was diluted to 3 ng/μl and used for pronuclear injection into mouse eggs. Potential transgenic mice were screened by the PCR and confirmed by Southern blotting.

The primers used to screen the RET9 mice spanned the RET/β-globin junction. The primer in RET (Ab: 5′TTCGGACTCACTGCTGTA-TGAC3′) and the primer in globin (K: 5′ACGATCCTGAGACTT-CCACACT3′) amplified a band of 370 bp. The primers used to screen the PTC2 mice spanned the junction of the Hoxb7 promoter and PTC2 cDNA. The Hoxb7 primer (R2: 5′GGGGGTCCTTTTGGTGTA-AATC3′) and the PTC2 primer (Ad: 5′CCGCATCTTCCTCCGTG-TAGAC3′) amplified a 550 bp band. The primers used to screen for the RET knockout allele (Schuchardt et al., 1994) were P1: 5′TGGGAGAAGGCGAGTTTGGAAA3′ and P2: 5′TTCAGGAAC-ACTGGCTACCATG3′ which amplified a wild-type band of 220 bp, and P3: 5′AGAGGCTATTCGGCTATGACTG3′ and P4: 5′CCTG-ATCGACAAGACCGGCTTC3′ which amplified a mutant band of 420 bp.

The primers used to screen for the GDNF knockout allele were BH128: 5′GCTGGCGAAAGGGGGATGTG3′, which anneals to the 5′ end of the wild-type allele, BH122: 5′GGAGGGAGCGG-TTCTTACAG3′ which anneals to the 3′ end of the wild-type allele, and BH129: 5′CTTGCCTGGTGCGGTTCTCT3′ which anneals to the 3′ end of the targeted allele. They amplified a 260 bp wild-type band and a 170 bp targeted band.

Southern and northern blot hybridization was performed using standard methods. Total RNA was isolated from adult brains and kidneys by homogenizing the organs in GNTC buffer (4 M guanidine thiocyanate, 25 mM sodium citrate, 0.5% Sarcosyl, 0.1 M β-mercaptoethanol), lowering the pH with 2 M sodium acetate pH 4.0, extracting with phenol and precipitating with isopropanol. Poly(A)⁺ RNA was selected using the PolyAttract mRNA isolation system (Promega). Probes for Southern and northern blots were random primed using the ‘Rediprime DNA labeling system’ (Amersham Life Sciences). Probes were labeled using [α³²P]dCTP (NEN Dupont). Hybridized blots were exposed to a Molecular Dynamics ‘Storage Phosphor Screen’ and developed using a Molecular Dynamics PhosphorImager.

A 1.1 kb HindIII/XhoI fragment of the mouse RET cDNA was used as a template in probe preparation for both Southern and northern blot hybridization on nucleic acids from Hoxb7/RET9 transgenic mice. This fragment contains a portion of the extracellular domain, the transmembrane domain and a portion of the intracellular domain. Genomic DNA was digested with ScaI, to cut once within the transgene, and Southern blots detected a transgene band at 6.5 kb in addition to endogenous bands. The entire PTC2 cDNA was used as a template for the probe used on Southern blots of DNA from Hoxb7/PTC2 mice. Genomic DNA was digested with XbaI, and the probe detected a 5.0 kb target band. We determined that none of the adult Hoxb7/PTC2 transgenic mice expressed the transgene by performing northern blot hybridization on poly(A)⁺ RNA from the brain and kidneys. A 1.1 kb EcoRI/XbaI fragment from the 3′ end of the PTC2 cDNA (corresponding to the intracellular domain of RET) was used as template for the probe. As a positive control, we used poly(A)⁺ RNA from the SK-N-SH human neuroblastoma cell line (ATCC).

The cross-sectional area of newborn kidneys was estimated by placing them on a 0.25 mm grid and measuring their length (rostrocaudal aspect) and width (mediolateral aspect) under a dissection microscope. For organ culture, kidneys dissected from E11.5 embryos were cultured on ‘Transwell Clear’ filter units (Costar) in DMEM/F12 medium containing 5% fetal bovine serum (Hyclone), 0.1 unit/ml penicillin/streptomycin (Gibco BRL) and 2 mM glutamine (Gibco BRL). Kidneys were cultured at 37°C in the presence of 5% CO₂. Lyophilized recombinant rat GDNF (R&D Technologies) was resuspended according to the manufacturers recommendations, and added directly to the culture medium to a final concentration of 100 ng/ml. Culture medium was replaced every 2 days. To visualize the ureteric bud branches, the cultured kidneys were fixed in fresh 2% paraformaldehyde in PBS at 4°C, permeabilized in 0.1% saponin in PBS at room temperature and stained in 50 μg/ml FITC-conjugated Dolichos biflorus lectin (DB) at 37°C. They were then washed, postfixed in 2% paraformaldehyde, mounted on glass slides and photographed. To quantitate the extent of ureteric bud branching, the numbers of bud tips in each DB-stained kidney was counted, without prior knowledge of genotype or culture conditions. The numbers of bud tips in wild-type or transgenic kidneys cultured with and without GDNF were compared using Student’s t-test for paired samples.

For histological analysis, kidneys were fixed in 10% formalin or 4% paraformaldehyde, washed in PBS, dehydrated and embedded in paraffin. Six micron sections were cut and stained with hematoxylin and eosin. Dolichos biflorus and Tetragonolobus lotus lectin staining of kidney sections was performed as described by D’Agati and Trudel (1992).

In situ hybridization was performed as described by Wilkinson (1992) with modifications. Briefly, samples were fixed in fresh ice cold 4% paraformaldehyde in PBS, washed in PBS, then saline, and then dehydrated through an ethanol series. They were then embedded in paraffin and sectioned at 8 μm. Hybridization was performed using digoxigenin-labeled riboprobes, which were detected using an alkaline-phosphatase-conjugated monoclonal antibody against digoxigenin (Boehringer Mannheim). Alkaline phosphatase activity was detected with BCIP and NBT (Boehringer Mannheim).

The entire RET cDNA was used as a template for riboprobe synthesis. For sense probe, the vector was linearized with EcoRV, and T7 polymerase was used to transcribe the RNA. For antisense probe, the vector was linearized with SacII, and transcribed using T3 polymerase. A GDNF cDNA, obtained from Dr Andreas Zimmer, was linearized with XbaI and transcribed using T7 polymerase (sense probe), or linearized with HindIII and transcribed with Sp6 polymerase (antisense probe). The template for the GFRα-1 probe was a partial cDNA cloned by RT-PCR from newborn mouse brain RNA, using the primers Gtop: 5′AACTGCCTGAGTCTGCAAG-ACTCCTGCAAG3′ and Gbot: 5′GGAGCAGCCATTGATTTTG-TGGTTATGTGG3′. The amplified cDNA fragment, which corresponds to 645 base pairs of the 3′ end of the coding region, was cloned into the pCRII vector in both orientations. To make the sense and antisense probes, the two vectors were linearized using HindIII, and transcribed with T7 polymerase.

GDNF expression was examined in E12.5 and E13.5 (as well as later) embryos by another method (data not shown). Hoxb7/RET9 transgenic mice were crossed with a mouse strain carrying a lacZ knock-in at the GDNF locus, in which lacZ expression serves as a reporter for the transcriptional activity of GDNF (Sanchez et al., 1996). Embryos heterozygous for the GDNF knock-in allele, and either transgenic or non-transgenic for Hoxb7/RET9, were stained for β-galactosidase (Hogan et al., 1994).

To ask whether the expression of RET throughout the ureteric bud, instead of only at its tips, would alter the pattern of UB growth and branching, we expressed a wild-type RET cDNA under the control of the Hoxb7 promoter (Fig. 1A). Four transgenic mouse lines were generated, all of which expressed the transgenic RET mRNA in the adult kidney as well as in the brain (Fig. 1B). The three lines with the highest expression in the kidney (Tg5, Tg6 and Tg8) were used for further studies. The sustained expression of RET in the kidneys of adult transgenic mice contrasts with the transient fetal expression of endogenous RET in the kidneys of non-transgenic mice (Pachnis et al., 1993), and confirms that the Hoxb7 promoter is constitutively active in the mature kidney (Kress et al., 1990; Vogels et al., 1993).

While some adult Hoxb7/RET transgenic mice appeared normal and healthy, many others became sick or died, and their kidneys were found to be small and cystic (data not shown). Subsequently, to avoid selecting for less severely affected individuals, the kidneys from a random sample of newborn transgenic progeny and their non-transgenic littermates were analyzed histologically. The transgenic kidneys ranged from normal, to moderately small and cystic, to very small and highly dysplastic (Fig. 2). The most severely affected kidneys were comparable in size and dysmorphology to the kidneys observed in some newborn mice homozygous for the RET knockout (e.g. Fig. 7E). However, in contrast to RET knockout mice, most of which display unilateral or bilateral renal agenesis, all of the transgenic mice had two kidneys and ureters. Similar results were obtained with all three transgenic lines.

In general, the size of the newborn transgenic kidneys showed an inverse correlation with the degree of dysplasia determined histologically. Therefore, the cross-sectional area of the newborn kidneys was employed as a quantitative measure of the degree of dominant abnormality in Hoxb7/RET transgenic mice. As shown in Fig. 3, most kidneys of newborn wild-type mice had a cross-sectional area of 6-10 mm². In contrast, the transgenic kidneys displayed a broad, bimodal size distribution, with about half in the near-normal size range (5-9 mm²) and the other half reduced several-fold in size (2-3 mm²). The variable severity of renal hypodysplasia did not correlate with level of expression of the transgene (data not shown), and is most likely a consequence of the mixed genetic background of the mice.

To examine the developmental basis of the renal abnormalities in Hoxb7/RET transgenic mice, fetal kidneys were dissected at various stages and subjected to histological analysis. This revealed that the first signs of developmental retardation could be observed at about E12.5 (data not shown). In situ hybridization studies of the fetal transgenic kidneys showed the expected ectopic expression of RET, as well as abnormalities in the expression of several other genes. RET mRNA was found throughout the ureter and medullary trunks of the ureteric bud (Fig. 4B), in contrast to its normal pattern at the UB tips (Fig. 4A). GFRα-1 mRNA, which is normally expressed in the tip and distal trunk regions of the UB, as well as in nephrogenic vesicles (Fig. 4C), showed greatly reduced expression (Fig. 4D). Similarly, GDNF mRNA, which is normally expressed in the undifferentiated mesenchyme surrounding the UB tips (Fig. 4E), was much reduced (Fig. 4F).

Next, we dissected metanephroi from transgenic E11.5 embryos and their non-transgenic littermates, and monitored their development in organ culture. In normal mouse embryos, the ureteric bud evaginates from the Wolffian duct at E10.5, and by E11.5 it has grown into the metanephric mesenchyme and branched once. All of the transgenic metanephroi contained normal T-shaped ureteric buds at E11.5, and they remained indistinguishable from wild type through the first day of culture (Fig. 5A). Subsequently, the transgenic kidneys began to show delayed development, and by the fifth day of culture many of them were grossly retarded, with relatively few ureteric bud branches and nephron elements (Fig. 5B). To quantitate this effect, we counted the number of UB tips in preparations stained with the UB-specific lectin Dolichos biflorus (DB) (Fig. 5C and data not shown). While wild-type kidneys developed an average of 41 (25 UB tips; n=16) by the 4^th-5^th day, the transgenic kidneys had, on average, fewer than half as many UB tips, 15 (14 (n=18, P=0.0003)

In summary, rather than causing increased or ectopic growth and branching of the UB, the Hoxb7/RET transgene caused a retardation of kidney development beginning at about E12.5, and reduced expression of two other genes involved in signal transduction through RET.

One potential explanation for the inhibitory effects of the transgene on kidney development was that ectopic RET signaling in the UB trunk cells interfered with their normal growth or differentiation. Alternatively, since the Hoxb7 promoter is also active in the UB tip cells, which express endogenous RET, overexpression of RET in these cells could result in an abnormally high level of RET signaling, which might also be deleterious. To test whether either of these explanations might be valid, we examined the effects of varying the level of GDNF to which the kidneys were exposed, either in vitro or in vivo. The rationale was that if RET signaling (either excessive or ectopic) were responsible for the observed developmental defects, then an increase in the level of GDNF should worsen the defects, while a moderate reduction in the level of GDNF should improve kidney development.

One metanephros from each transgenic or wild-type E11.5 embryo was cultured in medium containing 100 ng/ml of recombinant GDNF, and compared to the contralateral metanephros cultured in normal medium. After 4 or 5 days, the kidneys were stained with the DB lectin to examine the branching patterns (Fig. 5C,D). Consistent with previous reports (Pepicelli et al., 1997; Vega et al., 1996), wild-type kidneys cultured with added GDNF (n=16) showed a modest increase (average 23%) in the number of UB tips compared to those cultured in normal medium. The transgenic kidneys (n=18) showed an even greater improvement in growth and branching of the UB when cultured with 100 ng/ml GDNF, with an average 88% increase in the number of UB tips compared to the contralateral controls (P=0.05). Thus, addition of GDNF to the transgenic kidneys partially rescued the defect in UB branching, rather than worsening it.

In a complementary experiment, the Hoxb7/RET transgenic mice were crossed with GDNF heterozygous (+/−) knockout mice to reduce by half the level of GDNF to which the transgenic UB is exposed in vivo. The kidneys of newborn mice of various genotypes (transgenic or non-transgenic, GDNF +/+ or +/−) were dissected and measured. As observed in previous samples of newborn mice, most non-transgenic, wild-type (GDNF^+/+) kidneys had a cross sectional area in the range 7-10 mm² (Fig. 6) with a mean of 8.2 mm². Transgenic mice on a wildtype background showed a heterogeneous reduction in kidney size, with a mean of 5.45 mm², and 28% of kidneys 3 mm² or smaller. Nontransgenic GDNF^+/− mice showed a mild reduction in the kidney size distribution (mean 6.3 mm²), consistent with previous reports (Moore et al., 1996; Pichel et al., 1996; Sanchez et al., 1996). However, transgenic GDNF^+/− kidneys were more severely affected than transgenic GDNF^+/+ kidneys, with a mean area of 2.8 mm², and 69% with an area of 3 mm² or less. Therefore, a 2-fold reduction in the gene dosage of GDNF throughout prenatal development exacerbated the developmental defects inHoxb7/RET transgenic kidneys.

In conclusion, the results of both experiments argue against the hypothesis that the inhibition of kidney development in Hoxb7/RET transgenic mice is due to excessive or ectopic RET signaling. On the contrary, the ability of added GDNF to rescue partially the defects in vitro suggests that expression of the transgene can somehow interfere with normal GDNF/RET signaling.

Since the Hoxb7 promoter is active throughout the UB (including the normal site of RET expression), and since only a fraction of Hoxb7/RET transgenic animals show severe dominant defects, we expected that the transgene might be able to substitute for endogenous RET during kidney development, in at least some homozygous RET knockout mice. When the Hoxb7/RET transgene was bred into the RET^−/− background, it was indeed able to support apparently normal kidney development in some RET^−/− individuals, although in other cases the rescue was only partial (Fig. 7). This variability in the degree of rescue may be due to a combination of the incompletely penetrant, deleterious effect of the transgene on kidney development, and its ability to substitute for endogenous RET. Nevertheless, given that kidneys >1 mm² were never observed in RET^−/− mice, it is highly significant that >15% of the kidneys in Hoxb7/RET transgenic, RET^−/− mice were within the normal size range and histologically normal, and 70% were ≥2 mm² (Fig. 3).

The endogenous RET gene encodes two major protein isoforms, RET9 and RET51, which are produced by differential splicing and polyadenylation, and differ only in their C-terminal tail (Tahira et al., 1990). While the two isoforms are co-expressed in most if not all tissues where RET is expressed (Pachnis et al., 1993), it is not known whether they serve different functions. As the RET cDNA used to construct the Hoxb7/RET transgene encoded only the RET9 isoform, our results indicate that the RET51 isoform is not required for normal kidney development.

In addition to the restricted expression of RET in the developing kidney, the specific expression patterns of the RET ligand GDNF and the co-receptor GFRα-1 may also be important for the control of ureteric bud development. Because receptor tyrosine kinases are active only when induced to dimerize by ligand binding, RET can only transduce a signal in cells exposed to a soluble ligand, such as GDNF, and a membrane bound (or potentially, soluble) co-receptor such as GFRα-1. Since GDNF is only synthesized in the peripheral nephrogenic zone, and GFRα-1 is expressed primarily in the peripheral regions of the ureteric bud, RET receptors ectopically expressed in the medullary (trunk) portions of the UB might not be induced to signal. Therefore, the ectopic expression of wild-type RET receptors by the Hoxb7/RET transgene may not fully test its ability to alter the pattern of UB growth and branching.

To address this question, we attempted to produce transgenic mice expressing a ligand-independent, constitutively active form of RET, also under the Hoxb7 promoter (Fig. 1A). RET-PTC2 is a rearranged form of human RET found in certain papillary thyroid carcinomas, in which the RET intracellular domain is fused to a heterologous protein as a result of a chromosomal rearrangement. The RET-PTC2 fusion protein undergoes spontaneous dimerization, resulting in constitutive and ligand-independent tyrosine kinase activity (Bongarzone et al., 1993; Durick et al., 1995). All of the five founder transgenic mice produced with the Hoxb7/RET-PTC2 construct had very low transgene copy numbers, and when mated to normal mice they transmitted the transgene to only a small fraction of progeny, if any, suggesting that they were highly mosaic for the transgene (data not shown). In addition, we were unable to detect any expression of the transgene by northern blot analysis of adult kidney and brain RNA (data not shown). We therefore suspected that mice with high transgene copy numbers in every cell, expressing this transgene at high levels, had not survived to birth or to weaning age.

We therefore microinjected additional eggs with the Hoxb7/RET-PTC2 construct, and analyzed the resulting mice before birth. Four of the seven transgenic mice analyzed at E16-E19 displayed renal abnormalities, which were quite distinct from the abnormalities in Hoxb7/RET transgenic mice. The Hoxb7/RET-PTC2 transgenic kidneys were close to normal in size, and none of them had the cystic and dysplastic appearance of Hoxb7/RET kidneys. Instead, they contained abnormal nodules, either outside the kidney capsule in one case (Fig. 8A,B) or near the periphery of an otherwise normal kidney in other cases (e.g., Fig. 8C). All of the nodules consisted of branched tubular epithelial structures whose morphology was reminiscent of ureteric bud branches, surrounded by undifferentiated mesenchyme or stroma, with no visible glomeruli or comma- or S-shaped bodies (Fig. 8D-F). Staining with DB lectin confirmed the UB origin of the branched tubular structures (Fig. 9A-C). The nodules failed to stain with the lectin Tetragonolobus lotus, a specific marker for proximal tubules, confirming the morphological evidence that the nodules do not contain differentiated nephron elements (Fig. 9D-F).

Thus, expression of a constitutively active form of the RET tyrosine kinase under the Hoxb7 promoter does not inhibit kidney development in the same way as a similar transgene expressing wild-type RET, but it results in the unrestricted growth of the ureteric bud at the periphery of the kidney, or outside the normal confines of the kidney.

It is known that the secreted factor GDNF and its receptor, composed of the RET receptor tyrosine kinase and a co-receptor, GFRα-1, are essential for normal growth and branching of the ureteric bud during kidney development. However, exactly how this signal transduction pathway regulates the development of the UB is not well understood. RET and GFRα-1 are normally co-expressed only in the UB tips at the periphery of the kidney, while GDNF is expressed in the surrounding metanephric mesenchyme. In this study, we sought to test the developmental significance of these restricted expression patterns, by misexpressing two different forms of the RET tyrosine kinase throughout the UB, under the control of the Hoxb7 promoter. One transgene encoded a wild-type RET protein, and the second a ligand-independent, constitutively active form (RET-PTC2). Because the UB tips, where RET is normally expressed, are the sites of active growth and branching, we expected that these transgenes might alter the pattern of branching of the UB, and indeed both of them did, but in different ways. The Hoxb7/RET transgene, rather than causing extra or ectopic branches, inhibited kidney development in a manner reminiscent of (although less severe than) a RET knockout. Studies in which we manipulated the level of GDNF suggested that this defect was caused by a decrease, rather than an increase, in RET signaling. We suggest that ectopically expressed wild-type RET receptors act as a sink, decreasing the available concentration of GDNF around the UB tips. The expression of a ligand-independent form of RET kinase under the same promoter did not inhibit kidney development in the same manner, but instead allowed the UB to grow in abnormal patterns, and in one case to outgrow the normal confines of the kidney. This implies that the growth of the UB is normally restricted by the limited and spatially restricted supply of RET ligand.

Many of the Hoxb7/RET transgenic mice on a wild-type background had small, cystic kidneys, with a dysplastic histoarchitecture similar to that seen in the rudimentary kidneys that develop in some homozygous RET knockout mice. Histological and organ culture studies with transgenic fetal kidneys showed that these defects resulted from a deficiency in UB growth and branching beginning at about E12.5, after the first few branches of the UB formed normally. Although there was no specific defect in nephrogenesis, there was a progressive deterioration in kidney development, as indicated by the reduced expression of GDNF mRNA in the metanephric mesenchyme and GFRα-1mRNA in the UB and nephrogenic vesicles. As the transgene is expressed specifically in the UB, the effect on GDNF expression is likely to be an indirect one, in which defective growth of the UB tips leads to reduced proliferation of metanephric stem cells (which express GDNF), and in turn to further deterioration of UB development (as evinced by the weak expression of GFRα-1). This conclusion was supported by the observation that GDNF expression was normal at E12.5-E13.5, before the stage when kidney development becomes severely compromised (data not shown). A greatly reduced level of GDNF mRNA in the metanephric mesenchyme has also been seen in mice defective for Emx2, a homeobox gene expressed primarily in the UB, whose disruption inhibits UB growth and branching (Miyamoto et al., 1997).

Why did the expression of RET under the Hoxb7 promoter interfere with normal UB growth and branching? The hypothetical mechanisms can be divided into two classes: (1) those that involve signaling by the exogenous RET molecules, either excessive signaling in the UB tip cells due to increased RET expression, or ectopic signaling in regions of the UB that do not normally express RET; and (2) those that involve inhibition of endogenous RET function. Mechanisms that involve increased or ectopic RET signaling appear to be ruled out by two types of evidence. First, two different experiments, in which we manipulated the level of GDNF available to the transgenic kidneys, indicated that the developmental defects were caused by insufficient rather than excessive RET signaling. Addition of extra GDNF to transgenic kidney primordia cultured in vitro improved the growth and branching of the UB, while halving the GDNF gene dosage in vivo, by breeding the transgene into a GDNF heterozygous knockout background, exacerbated the deleterious effect of the transgene on kidney development. Second, the Hoxb7/RET-PTC2 transgene, which encodes a ligand-independent form of the RET tyrosine kinase that signals constitutively, did not inhibit kidney development in a similar manner, although it caused a different type of abnormality.

There are several potential mechanisms by which the transgene might have interfered with endogenous RET function. First, it might have carried a mutation so that it encoded a hypomorphic receptor, which could inhibit normal RET signaling through a dominant negative mechanism. This appeared unlikely, since the transgene was able to support normal kidney development in mice homozygous for a knockout of the endogenous RET gene. Nevertheless, to rule out this possibility, we sequenced the entire coding sequence of the transgene, and found that it matched the published RET cDNA sequence (Iwamoto et al., 1993) at all but one base pair, which appears to be either a polymorphism or an error in the published sequence¹. Second, co-expression of the RET9 and RET51 isoforms might be important for kidney development. Overexpression of RET9 (encoded by the transgene) would alter the ratio of the two isoforms, which might somehow interfere with normal signal transduction. This possibility has been ruled out by producing Hoxb7/RET51 transgenic mice and breeding this transgene together with the Hoxb7/RET9 transgene. The doubly transgenic mice showed no improvement in kidney development compared to the Hoxb7/RET9 transgenic mice (unpublished data).

We propose an alternative model to explain the dominant defect, in which the ectopically expressed RET receptors on UB trunk cells can bind and sequester GDNF, thus reducing the amount available to the UB tip cells, for whose growth the level of GDNF is critical (Fig. 9). It is possible that the UB trunk cells, which are phenotypically distinct from tip cells (Bard et al., 1994; Sariola and Sainio, 1997; Vainio and Muller, 1997), may be unable to respond to RET signaling with the same biological response as UB tip cells and so cannot overcome the defect. Several observations are consistent with this model. First, GDNF is diffusible, as localized sources of the factor can act at a distance to stimulate UB outgrowth from the Wolffian duct (Durbec et al., 1996; Sainio et al., 1997). It is therefore plausible that GDNF produced by the peripheral mesenchyme cells may bind to receptors in more proximal regions of the UB. Second, GDNF is produced in limiting quantities, as GDNF knockout heterozygotes display reduced kidney size (Fig. 6) and occasional renal agenesis (Moore et al., 1996; Pichel et al., 1996; Sanchez et al., 1996). Therefore, a reduction in the level of available GDNF could account for the observed growth inhibition, as well as the ability of added GDNF to partially rescue this defect. Third, GFRα-1 is expressed not only in the UB tip cells but also in the immediately proximal UB trunk cells (Baloh et al., 1997; Pachnis et al., 1993; Sainio et al., 1997), which may become a sink for GDNF when they co-express RET. Alternatively, soluble GFRα-1 whose GPI anchor has been cleaved might also mediate the binding of GDNF to RET receptors in more medullary regions of the UB in the transgenic mice. Fourth, at early stages (E10.5-E12.0) when RET is normally expressed throughout the UB (Pachnis et al., 1993; Pepicelli et al., 1997), the transgene has no deleterious effect. Only at the stage when RET expression normally becomes restricted to the UB tips (E12.5-E13.5) does the transgene begin to inhibit kidney development. A similar mechanism has been invoked to explain the dominant effects of several naturally occurring mutations that cause the abnormal expression of c-kit, a tyrosine kinase receptor for the Steel factor, in somites or other ectopic sites. In these mutants, which include W^sash, W^banded and Patch, ectopically expressed c-kit is believed to sequester Steel, which normally promotes the survival or proliferation of migrating melanoblasts, resulting in pigmentation defects (Duttlinger et al., 1995, 1993; Kluppel et al., 1997; Wehrle-Haller et al., 1996).

One question we initially posed was why the normal expression of RET is so tightly restricted to a small population of cells at the tips of the UB. We initially hypothesized that this pattern of gene regulation might be part of a mechanism to limit the sites of UB branching. However, in the cultured transgenic kidneys, either in normal medium or with added GDNF, we observed no signs of abnormal branching or unusual growth of UB trunks. Furthermore, expression of a constitutively active form of RET throughout the UB did not visibly perturb the overall histoarchitecture of the kidney, which might have been expected if it induced ectopic branching. Therefore, while our data do not rule out the initial hypothesis, they suggest another explanation. Because the RET ligand GDNF is produced in limiting quantity and is diffusible, the expression of RET in a less tightly regulated pattern (e.g., throughout the UB) would result in binding and sequestering of the ligand by UB trunk cells at the expense of the UB tip cells.

When Hoxb7/RET transgenic mice were bred with RET knockout mice, the transgene was able to rescue kidney development in some of the newborn RET^−/− mice (Figs 3 and 7). Fifteen percent of the kidneys in these mice were normal in size and histology, and an additional 55% were partially rescued (i.e., smaller than normal but larger than RET^−/− kidneys, displaying mild to moderate dysplasia). The incomplete rescue in some animals may be due to two competing effects of the transgene: the ability of exogenous RET to substitute for endogenous RET in the UB tips, and the incompletely penetrant, inhibitory effect of the ectopically expressed RET. The Hoxb7/RET transgene did not rescue the other defects in the RET^−/− mice (e.g., intestinal aganglionosis), presumably because it is not expressed in many of the normal sites of RET expression in the developing nervous system (Kress et al., 1990; Vogels et al., 1993).

Two important conclusions can be drawn from this result. First, it formally demonstrates that the renal agenesis phenotype in RET knockout mice is due to the RET mutation, and not to an inadvertently introduced mutation in another gene linked to RET. Second, it demonstrates that the RET9 isoform is sufficient to support normal kidney development. RET9 and RET51 differ only at the C terminus, where the last 9 amino acids of RET9 are replaced by a different 51 amino acid tail in RET51. One difference between the isoforms is that only RET51 contains binding sites for Grb2, an adaptor protein that links RET to Ras and the MAP kinase pathway (Borrello et al., 1994; Liu et al., 1996; Lorenzo et al., 1997). Apparently, the ability to bind Grb2 is not essential for the role of RET signaling in kidney development. This may be because both RET isoforms have binding sites for Shc, another adaptor protein through which RET signaling can activate Ras (Arighi et al., 1997; Asai et al., 1996; Lorenzo et al., 1997; Ohiwa et al., 1997). Our observation is consistent with previous findings that RET9 and RET51 are both biologically active in a number of systems that depend on Ras activation, including fibroblast transformation and induction of neuronal differentiation in PC12 cells (Asai et al., 1995; Iwashita et al., 1996; Rossel et al., 1997; Santoro et al., 1995).

Expression of RET-PTC2 under the Hoxb7 promoter resulted in abnormal growth of the UB in localized nodules at the periphery of the kidney, and in one case, in multiple nodules outside the kidney. Surprisingly, with the exception of these nodules, the kidneys of newborn transgenic mice were histologically normal, and their size was unaffected. This suggests that throughout the early stages of kidney development, the expression of a constitutively active RET kinase had little or no effect, perhaps because there is already sufficient GDNF to maintain RET signaling in the UB tips. As noted above, this result also suggests that ectopic RET signaling in the UB trunk cells is inconsequential. The ability of the UB to eventually outgrow the rest of the kidney, forming nodules that protrude through the nephrogenic zone, suggests that at later stages of kidney development, the centripetal growth of the UB tips is normally restricted by the availability of GDNF. GDNF is expressed by a thin layer of mesenchymal cells near the periphery of the kidney, and any UB tips that grow through this layer would normally outgrow their source of ligand. In transgenic mice expressing a ligand-independent RET kinase, this level of regulation of UB growth is apparently circumvented. This indicates that RET signaling is not only necessary for UB development (as shown by the RET and GDNF knockouts), but that RET signaling actively stimulates the growth of the UB. In previous in vitro studies, addition of GDNF to the culture medium was observed to stimulate UB cell proliferation within intact metanephroi (Pepicelli et al., 1997), but not in isolated ureteric buds (Sainio et al., 1997). While our observations are consistent with a mitogenic role for GDNF/RET signaling, other factors produced by the kidney or surrounding tissues may also be required for the stimulation of UB growth.

It is interesting that the tubular UB elements in the nodules were surrounded only by undifferentiated mesenchymal or stromal cells, lacking mature glomeruli or immature nephrogenic vesicles. One likely explanation is that the UB has outgrown the nephrogenic stem cells at the periphery of the kidney, or (in the animal with extra-renal nodules) has grown into regions of mesenchyme lacking any nephrogenic potential. Alternatively, the abnormally growing UB may have lost the ability to induce nephrogenesis.

The authors would like to thank Doris Herzlinger and Cathy Mendelsohn for critical reviews of the manuscript, Qais Al-Awqati, Vassilis Pachnis and Jonathan Barasch for helpful discussions, and Tony Ward and Xiaolin Liang for excellent technical assistance. We also thank Vassilis Pachnis, Mariano Barbacid, Bristol-Myers Squibb, Jacqueline Deschamps, Marco Pierotti and Andreas Zimmer for reagents. This work was supported by a grant from the NIDDK to F. C.