Gremlin-mediated BMP antagonism induces the epithelial-mesenchymal feedback signaling controlling metanephric kidney and limb organogenesis

Vertebrate organogenesis is orchestrated by signaling centers with organizer properties that coordinate cell proliferation and survival with cell specification and differentiation. Signaling by cells with special organizer properties instructs undetermined neighboring cells with respect to their fate and differentiation potential. Such reciprocal epithelial-mesenchymal signaling interactions control growth and patterning of morphologically very diverse embryonic structures, including limbs, kidneys and lungs in vertebrates. In particular, the molecular epithelial-mesenchymal signaling interactions regulating branching morphogenesis have recently provided novel insights into how tissues are organized in space as organogenesis proceeds(Affolter et al., 2003). Two models of paradigmatic value to study morphogenetic epithelial-mesenchymal signaling interactions in vertebrate embryos are the limb bud(Tickle, 2003) and branching morphogenesis of the ureter during kidney organogenesis(Vainio and Lin, 2002). In particular, two main signaling centers control limb bud development: the sonic hedgehog (Shh)-expressing polarizing region, which is located in the posterior limb bud mesenchyme; and the apical ectodermal ridge (AER), a differentiated columnar epithelium expressing different types of signaling peptides. SHH signaling by the polarizing region controls patterning of distal limb structures and its expression is regulated by fibroblast growth factor(FGF) signaling from the AER (SHH/FGF feedback loop)(Panman and Zeller, 2003). Genetic analysis in the mouse indicates that the AER expressed FGFs, such as FGF8 and FGF4, cooperate to activate and positively regulate Shhexpression in the posterior limb bud mesenchyme(Lewandoski et al., 2000; Moon and Capecchi, 2000; Sun et al., 2002). The bone morphogenetic protein (BMP) antagonist gremlin (Grem1)(Hsu et al., 1998) is a cystein knot protein belonging to the CAN family that antagonizes preferentially BMP2 and BMP4(Avsian-Kretchmer and Hsueh,2003). Grem1 is expressed by a subset of SHH responsive mesenchymal cells and has been implicated in transducing the SHH signal to the posterior AER. This results in activation of Fgf4 expression and establishment of the SHH/FGF4 feedback loop(Capdevila et al., 1999; Zuniga et al., 1999).

In addition to the limb bud, Grem1 is expressed by a variety of embryonic structures including lung (Lu et al., 2001; Shi et al.,2001) and kidney rudiments (this study). Development of the definitive metanephric kidney is initiated by formation and growth of the ureteric bud. As the ureteric bud invades the metanephric mesenchyme, it induces condensation and nephrogenesis through reciprocal interactions(Saxén, 1987). Genetic analysis in the mouse shows that the Wt1 and Pax2transcription factors (Kreidberg et al.,1993; Torres et al.,1995) control the induction of metanephric development. By contrast, the extracellular signals that trigger ureteric bud formation in the posterior part of the Wolffian duct and initiate ureter growth and branching have so far remained largely elusive. The tips of the invading ureter express the tyrosine kinase receptor RET (Pachnis et al., 1993; Towers et al.,1998), whereas RET ligand GNDF is expressed in the condensing metanephric mesenchyme (Hellmich et al.,1996; Towers et al.,1998). Genetic analysis has established that the epithelial-mesenchymal signaling interactions mediated by RET and GNDF are essential for metanephric development(Vainio and Lin, 2002). BMP signaling has also been implicated in metanephric development as potential regulator of ureter growth, branching and nephrogenesis(Dudley et al., 1995; Luo et al., 1995; Martinez and Bertram, 2003; Miyazaki et al., 2000; Raatikainen-Ahokas et al.,2000). In particular Bmp4, which is expressed by the mesenchyme surrounding the Wolffian duct and ureter stalk seems to fulfill a dual function during early metanephric development. Heterozygous Bmp4mutant mouse embryos display a variable kidney phenotype characterized by defects in ureteric epithelium growth and induction of ectopic ureter branching (Miyazaki et al.,2000; Raatikainen-Ahokas et al., 2000). These studies have also provided evidence that BMP signaling regulates ureteric bud initiation and branching. An involvement of BMP antagonism has been postulated, but the relevant antagonist(s) remained to be identified (Miyazaki et al.,2000). To study the essential functions of the BMP antagonist Grem1, we have deleted the Grem1 open reading frame (ORF) by homologous recombination in mouse embryonic stem (ES) cells. We report that Grem1-deficient mice die shortly after birth because of disruption of kidney and lung organogenesis. During limb bud development, Grem1 is required for survival of core mesenchymal cells and to establish a functional AER expressing different types of signals, which regulate Shhexpression and progression of limb bud morphogenesis. During kidney organogenesis, Grem1 is required to initiate ureter growth and branching that in turn induces metanephric nephrogenesis. Together, these results reveal a common and essential role of Grem1-mediated BMP antagonism in initiating dynamic epithelial-mesenchymal signaling.

We generated the targeting vector using a 4.8 kb NdeI-XbaI and a 5.6 kb NsiI-NsiI Gre genomic fragment isolated from a 129/SvJ Lambda FIXII library(Stratagene). We inserted an IRES-lacZ gene and a PGK-Neo^R cassette flanked by two loxP sites in the same transcriptional orientation as the Grem1 gene. R1 ES-cells were electroporated with the NotI linearized targeting vector and screened by genomic Southern with an NsiI-EcoRI probe mapping outside the 3′ homology arm(Fig. 1A). Thirty-five homologous recombined ES-cell clones were obtained at a frequency of 10.8%. Correct recombination resulting in the deletion of the entire 552 base Grem1 ORF encoded by exon 2 and of 132 bases of the 3′ UTR was confirmed by extensive Southern blot and PCR analysis. ES cells carrying the Gre^ΔORF null allele were injected into C57BL/6 blastocysts and, following germline transmission, the mice were maintained in mixed B6;129S and CD1 backgrounds. PCR genotyping was used for all subsequent studies to allow specific detection of both the wild-type and Gre^ΔORF alleles. The floxed PGK-neo^R gene was removed by crossing Gre^ΔORF heterozygous mice with the Credeleter strain. The sequence of the murine Grem1 locus was obtained from the UCSC Genome Bioinformatics Website(http://genome.ucsc.edu/)and analyzed using the DNA Strider 1.3™ program. The Gre^ΔORF mutation was crossed into 129S3/SvImJ,C57BL/6 and CD1 strains as the penetrance of the kidney phenotype depends on genetic background. In the 129S and CD1 backgrounds, the kidney phenotype is fully penetrant (Fig. 2B).

Embryos and newborn mice were PCR genotyped and accurately staged by determining their somite numbers. Whole-mount and section RNA in situ hybridization were performed as previously described(Dono et al., 1998; Zuniga et al., 1999) using digoxigenin-UTP-labeled anti-sense riboprobes. Apoptotic cells were detected in situ by incorporating fluorescein-dUTP into fragmented DNA using terminal transferase (Roche Diagnostics). For histological analysis, sections were cut from formalin-fixed and paraffin-embedded material and stained with Hematoxylin and Eosin using standard protocols. For scanning electron microscopy (SEM), embryos were fixed in 1% gluteraldehyde (Sigma) for 1 hour at 4°C and processed for SEM.

Mouse forelimb buds were cultured and grafted as described(Zuniga et al., 1999) with the following modifications. Trunks with attached forelimb buds were isolated from either wild type, heterozygous or Grem1-deficient embryos. Embryos were staged by counting somites and genotyped by PCR. Spherical cell aggregates were grafted into the forelimb buds and trunks and were cultured for 15 hours in serum-free medium in 6.5% CO₂ at 37°C. The culture medium was prepared by supplementing high glucose DMEM (GIBCO BRL)medium, with L-glutamine, penicillin/streptomycin, non-essential amino acids,sodium pyruvate, D-glucose, L-ascorbic acid, lactic acid, d-biotin, vitamin B12 and PABA. QT6 fibroblast cells expressing Shh and Grem1under control of the CMV promoter were prepared using standard calcium phosphate transfection (Zuniga et al.,1999). One day after transfection, spherical cell aggregates were prepared by plating cells at high density on bacterial plates. The following day, cells were treated with mitomycin C for 1 hour to block proliferation. After washing the cell aggregates extensively, they were grafted into recipient limb buds (a detailed protocol for media preparation, limb bud grafting and culturing is available upon request).

The second exon of the Grem1 gene, which encodes the complete ORF,was inactivated by homologous recombination in R1 ES-cells as shown in Fig. 1A. The complete Grem1 coding exon 2 was deleted and replaced by a lacZmarker and a Neomycin resistance (Neo^R) gene flanked by two loxP sites. Correctly targeted ES-cell clones were identified by Southern blot screening (Fig. 1B,C) and two independent clones were used to generate Gre^ΔORF mice(Fig. 1D). Heterozygous mice of both strains appear normal and the distribution of Grem1 and Gre-lacZ fusion transcripts(Fig. 1A) are identical(Fig. 1E). By contrast, Gre^ΔORF homozygous newborn mice display limb defects(Fig. 2G-J) and die shortly after birth. Autopsy at birth reveals that Gre^ΔORFhomozygous newborn mice lack metanephric kidneys and ureters (compare Fig. 2A with 2B; data not shown), whereas the remainder of the urogenital system appears normal. In addition, newborn Grem1 deficient mice display respiratory problems(dyspnoe and cyanosis) that probably contributes significantly to their early death. Indeed, septation of the lung airway epithelium is affected as numbers of differentiated alveoli are reduced in Grem1-deficient newborn mice(compare Fig. 2C with 2D). Furthermore, the airway epithelium remains multi-layered in comparison with wild-type embryos (compare Fig. 2E with 2F). Cre recombinase-mediated removal of the Neo^R gene does not alter the phenotypes, confirming that they are due to the Grem1 deficiency (data not shown). As initial analysis indicated that the lung septation defects arise only during advanced lung organogenesis, the present study focuses on analyzing the early pattern defects that disrupt limb bud and kidney organogenesis.

The limb phenotypes observed in Grem1-deficient mice correspond to a strong and fully penetrant ld limb phenotype(Fig. 2G-J). The zeugopods of Grem1-deficient newborn mice are differentially affected as ulna and radius fuse during onset of ossification, while only one skeletal element forms in hind limbs (arrows, Fig. 2G-J). The autopods are severely truncated because of metacarpal fusions (arrowheads, Fig. 2H,J), reductions in digit numbers and loss of posterior identities together with soft tissue webbing(Fig. 2H,J; data not shown).

During limb bud morphogenesis, the number of Shh-expressing cells and transcript levels increase progressively in wild-type embryos(Fig. 3A)(Riddle et al., 1993). By contrast, the Shh expression domain remains small and levels stay low in limb buds of Gre^ΔORF homozygous embryos(Fig. 3A; data not shown). This failure to propagate SHH signaling has been attributed to the disruption of the SHH/FGF4 feedback loop (Haramis et al., 1995; Khokha et al.,2003; Zuniga et al.,1999). However, analysis of Gre^ΔORFhomozygous embryos reveals a general disruption of AER morphology and function(Figs 3, 4). Activation of Fgf8in the limb bud ectoderm and thereby initiation of AER formation occur normally in Grem1-deficient limb buds(Fig. 3B, E9.5). However, the Fgf8-expressing AER cells remain more spread out along the dorsoventral ectoderm, revealing the early disruption of AER morphology in Grem1-deficient limb buds (Fig. 3B, E9.75). As development proceeds, Fgf8-expressing cells become restricted to the apex, but the domain remains patchy in mutant limb buds (Fig. 5E,G; data not shown), owing to the defects in AER morphology(Fig. 4E,F). In addition, FGF signaling by the posterior AER (Martin,1998) is completely disrupted as neither Fgf4 nor Fgf9 nor Fgf17 is activated in Grem1-deficient limb buds (Fig. 3C and data not shown).

This general alteration of AER morphology and function could be a direct consequence of enhanced BMP signaling as it has been shown that BMPs, although required to induce the AER, interfere with formation of the mature and fully functional AER (Pizette and Niswander,1999; Pizette et al.,2001). Msx1 and Msx2 are targets of BMP signaling in the limb bud and can be used as in situ indicators of enhanced and/or ectopic BMP signaling (Pizette et al., 2001). Msx1 and Msx2 are activated normally(Fig. 4A and data not shown),but subsequently ectopically expressed in the distal to anterior sub-AER limb bud mesenchyme of Grem1-deficient embryos(Fig. 4B; data not shown). This upregulation is indicative of enhanced BMP signaling in the sub-AER mesenchyme due to lack of Grem1-mediated BMP antagonism. Indeed, expression of both Bmp4 and Bmp7 is maintained in the limb bud mesenchyme of Grem1-deficient embryos, while Bmp2 is reduced from early stages onwards (Fig. 4C; data not shown). Therefore, overall Bmp transcript levels appear unaffected in the sub-AER mesenchyme, where BMP signaling is enhanced due to lack of Grem1 function (compare Fig. 4B with 4C). Furthermore, Bmp expression is activated in the mutant AER, but not maintained during progression of limb bud morphogenesis(Fig. 4D; data not shown). Morphological analysis by scanning electron microscopy reveals that the apical ectodermal cells of Gre^ΔORF homozygous limb buds fail to adopt the characteristic ridge-like morphology(Fig. 4E), although AER-type cells are present (Fig. 4F). Taken together, these results establish that induction of Fgf8-expressing AER cells occurs normally, while formation of a morphologically distinct and functional AER depends critically on Grem1-mediated antagonism of BMP signaling in the distal/sub-AER mesenchyme.

Particularly Bmp2 has been considered a direct transcriptional target of SHH signaling in the mesenchyme(Drossopoulou et al., 2000). Therefore, reduced Bmp2 expression could be a consequence of reduced SHH signaling and thus secondary to disrupting Grem1. However,posterior grafts of Shh-expressing fibroblasts, which are capable of rescuing gene expression (Zuniga et al.,1999), fail to upregulate Bmp2 expression in limb buds of Gre^ΔORF homozygous embryos(Fig. 5A,B). By contrast,grafts of Grem1-expressing fibroblasts enhance mesenchymal Bmp2 transcription and restore Bmp2 expression in the AER of mutant limb buds (Fig. 5C,D). Similarly, Grem1 (Fig. 5G,H) but not Shh grafts(Fig. 5E,F) restore Fgf8,Fgf4 (Zuniga et al.,1999) and Fgf9 (data not shown) expression in the AER of mutant limb buds. These results establish that mesenchymal Grem1modulates Bmp2 and Fgf8 expression positively and is required for activation of FGF genes in the posterior AER.

The bilateral renal agenesis in Gre^ΔORFhomozygous newborn mice in the context of an otherwise normal urogenital system (Fig. 2B) indicates an unexpected essential role of Grem1 during metanephric kidney organogenesis. Metanephric kidney development is initiated by invasion and induction of the metanephric mesenchyme by the ureter between embryonic days 10.5 to 11.0 in mouse embryos (Vainio and Lin, 2002). Grem1 is initially expressed by the intermediate mesenchyme (Pearce et al.,1999) and from about embryonic day 9.5 onwards by the Wolffian duct and mesonephric tubules (Fig. 6A and data not shown). During onset of metanephric development, Grem1 is rapidly downregulated and restricted posteriorly in the Wolffian duct (Fig. 6B; data not shown). At this stage, Grem1 is expressed locally in the condensing metanephric mesenchyme, which surrounds the ureteric bud(Fig. 6C). The Pax2transcription factor is essential for urogenital development and acts upstream of the signal(s) initiating metanephric development(Torres et al., 1995). During ureteric bud formation, Pax2 expression is not affected in Gre^ΔORF homozygous embryos (data not shown). By contrast, invasion of the metanephric mesenchyme by the ureter and upregulation of Pax2 expression in the induced mesenchyme fail to occur in Grem1 deficient embryos(Fig. 6D). The failure to induce metanephric development becomes more apparent as development progresses. Pax2 expression is lost from the mutant metanephric mesenchyme by embryonic day 12.5, while nephrogenesis progresses in wild-type embryos (Fig. 6E). In addition, Bmp2, Bmp7 (Fig. 6F,G)and Wnt4 transcripts (data not shown) are absent from the mutant metanephric mesenchyme. These results are indicative of a possible failure to induce condensation of the metanephric mesenchyme.

The tyrosine kinase receptor RET and its ligand GDNF are expressed by the ureter epithelium and mesenchyme, respectively(Fig. 7A)(Hellmich et al., 1996; Pachnis et al., 1993; Towers et al., 1998). In Gre^ΔORF homozygous embryos, both Ret and Gdnf expression are activated and the ureteric bud forms(Fig. 7B; data not shown),possibly as a consequence of activating RET/GNDF signaling. However initiation of ureter growth (arrow, Fig. 7A) and Gdnf upregulation are completely blocked in Grem1-deficient embryos (Fig. 7B, compare with Fig. 7A). The ureter branches (arrowheads, Fig. 7C) as it invades the metanephric mesenchyme. The branching tips of the wild-type ureter express high levels of Ret and Gdnf is upregulated in the surrounding, condensing mesenchyme (Fig. 7E). By contrast, the ureteric bud does not branch in Grem1-deficient embryos (Fig. 7D) and mesenchymal Gdnf expression is rapidly lost,despite continued Ret expression by the arrested epithelium(Fig. 7D,F).

To understand how the molecular alterations give rise to the distal limb defects and result in elimination of the metanephric kidney, potential effects on programmed cell death were assayed. Massive cell death is observed in the core mesenchyme of Grem1-deficient limb buds by embryonic day 11.0(Fig. 8A). However, the superficial dorsal and ventral limb bud mesenchymal cells normally expressing Grem1 (Merino et al.,1999) survive in limb buds of Gre^ΔORFhomozygous embryos as indicated by the continued presence of lacZ-expressing cells (Fig. 8B). Similarly, lacZ-expressing cells remain in the Wolffian duct and mesenchyme of Grem1-deficient embryos (data not shown) in spite of massive aberrant cell death at embryonic day 11.5(Fig. 8C). Analysis of parallel sections shows that the metanephric mesenchyme, which expresses Pax2(Fig. 8D), is eliminated by apoptosis.

We show that inactivation of Grem1 in the mouse causes a limb phenotype in combination with complete renal agenesis and lung airway defects. In particular, the gross morphological appearance of the limb and kidney phenotypes is strikingly similar to newborn mice homozygous for the strongest ld alleles (Maas et al.,1994). Indeed, Khokha et al.(Khokha et al., 2003) reported allelism between their Grem1 loss-of-function mutation and the ld^J allele as trans-heterozygous mice display limb defects at birth. We have now identified the molecular lesions in both the ld^J and ld^OR alleles and shown that they disrupt, respectively, delete the Grem1 ORF(Zuniga et al., 2004). These rather unexpected results prompted us to re-evaluate the other ldalleles and the proposed function of formin in regulating Grem1expression in the limb bud mesenchyme(Zuniga et al., 1999) using extensive reverse genetic analysis of the two genes. These studies reveal that disruption of a cis regulatory element within formin, which is required for limb bud mesenchymal expression of Grem1 causes the ld limb phenotypes (Zuniga et al.,2004). In conclusion, these studies show that the existing ld alleles are either complete or limb bud specific Grem1loss-of-function mutations and that the ld phenotype is not caused by disruption of the C-terminal formin domain as so far assumed (reviewed by Evangelista et al., 2003; Zeller et al., 1999).

The present study establishes that Grem1-mediated antagonism of BMP signaling is required for proper AER formation and function. However in Grem1-deficient limb buds, the expression of genes controlling dorsoventral axis formation is normal (A.Z., unpublished) and both the AER and Fgf8 expression are induced indistinguishable form wild-type limb buds (this study). The AER and Fgf8 expression are induced by Wnt3/β-catenin and BMP signaling activities in the ectoderm during initiation of limb bud development and establishment of dorsoventral polarity (Barrow et al.,2003; Kawakami et al.,2001; Pizette et al.,2001; Soshnikova et al.,2003). Grem1 functions subsequently by antagonizing BMP signaling in the distal limb bud mesenchyme, which is obviously essential for progression of AER formation and establishment of multi-factorial AER signaling. In Grem1-deficient limb buds, enhanced mesenchymal BMP signaling blocks AER maturation and signaling at an early stage and disrupts distal limb bud morphogenesis, last but not least through apoptosis of core mesenchymal cells (see also below). These results corroborate previous studies in chicken embryos, which showed that mesenchymal BMP antagonism maintains the AER and promotes distal limb bud morphogenesis(Capdevila et al., 1999; Pizette and Niswander,1999).

Furthermore, Grem1-mediated BMP antagonism has been implicated in regulating Shh expression through its role in establishment of the SHH/FGF4 feedback loop (Capdevila et al.,1999; Zuniga et al.,1999; Khokha et al.,2003). During initiation of limb bud development, Shhexpression is activated in the posterior mesenchyme under the influence of FGF8 signaling by the AER, probably in combination with FGF4(Lewandoski et al., 2000; Moon and Capecchi, 2000; Sun et al., 2002). In particular, Shh is not activated in hindlimb buds lacking both Fgf8 and Fgf4, despite continued expression of Grem1 in the mesenchyme and Fgf9, Fgf17 and BMP genes in the mutant AER (Sun et al., 2002). These results together with our studies(Zuniga et al., 1999) also show that Grem1 functions initially independent of SHH in AER formation and FGF gene activation in the posterior AER. During progression of limb bud morphogenesis, Grem1 induced FGF signaling by the posterior AER participates in dynamic SHH regulation as Grem1 rescues Shh expression with kinetics similar to FGF4 in ld mutant limb buds (L.P., unpublished). The general disruption of AER-FGF signaling underlies the failure to upregulate Shh signaling in Grem1-deficient limb buds. Through establishment of feedback signaling, Grem1 mediates the dynamic regulation of both limb bud signaling centers. For example, the distal-anterior progression of mesenchymal Grem1 expression during limb bud morphogenesis causes anterior expansion of FGF signaling in the AER, which in turn regulates SHH signaling by the polarizing region (Zuniga et al.,1999). These dynamic changes alter the ratios of different peptide signals received by both AER cells and the underlying limb bud mesenchyme. Sanz-Ezquerro and Tickle (Sanz-Ezquerro and Tickle, 2000) have shown that the size and signaling strength of the Shh expression domain in limb buds is tightly regulated by apoptosis. Taken together, the analysis of epithelial-mesenchymal signaling in limb buds indicates that Shh expression is not regulated by a mere SHH/FGF feedback loop, but through complex and dynamic feedback signaling involving different types of mesenchyme and AER signals, and their antagonists belonging to the FGF, BMP and WNT gene families.

In Grem1-deficient mouse limb buds, prominent apoptotic cell death is observed in the core mesenchyme from about embryonic day 11.0 onwards. This cell death pattern is rather distinct from the ones observed in Shhdeficient (te Welscher et al.,2002) and Fgf4/Fgf8 double mutant(Sun et al., 2002) mouse embryos and following AER removal (Dudley et al., 2002). In addition, experiments in chicken embryos have provided evidence for a role of Grem1-mediated BMP antagonism in cell survival during digit formation and chondrogenesis(Merino et al., 1999). During the onset of chondrogenesis, Grem1 acts in a paracrine fashion on the adjacent (core-) mesenchyme to protect it from undergoing programmed cell death (this study). Therefore, this anti-apoptotic function of Grem1could provide an explanation for the reductions and fusions of distal limb skeletal elements observed in Grem1-deficient mouse embryos. It is possible that the effect of Grem1-mediated BMP antagonism on cell survival is direct and does not involve feedback signaling between mesenchyme and AER.

The complete renal agenesis in Grem1-deficient mice reveals that the BMP antagonist Grem1 is required for metanephric development. This study identifies Grem1 as the essential extracellular signal,which initiates metanephric kidney development by enabling the ureter to invade the metanephric mesenchyme. However, establishment of the two signaling centers controlling metanephric development, the ureteric bud (expressing RET)and metanephric mesenchyme (expressing GDNF), occurs without Grem1;while initiation of the ureter growth and branching depend on Grem1function. In analogy to its function in limb buds, Grem1 regulates the transition to dynamic signaling interactions to enable induction of metanephric organogenesis. During set-up of RET/GDNF signaling and ureteric bud formation, Grem1 is expressed by the Wolffian duct and locally by the metanephric mesenchyme, but the primary tissue affected in Grem1-deficient embryos could be the ureteric epithelium as is the case in ld homozygous embryos(Maas et al., 1994). Such impairment of epithelium to mesenchyme signaling disrupts upregulation of Gdnf, Pax2 and Ret expression in the mesenchyme and tips of the invading ureter, respectively. This disruption in turn leads to complete elimination of the metanephric mesenchyme by apoptotic cell death. This phenotype is strikingly similar to the one caused by inactivation of Sall1, a transcription factor expressed by the metanephric mesenchyme(Nishinakamura et al., 2001). However, Sall1 remains expressed in Grem1 mutant embryos(O.M. and A.Z., unpublished), which indicates that it is not a direct target.

Several BMP genes are expressed during initiation of metanephric kidney development and have been implicated in the early inductive events(Martinez and Bertram, 2003; Vainio and Lin, 2002). In particular, analysis of Bmp4 heterozygous embryos has provided evidence for its essential roles during ureter morphogenesis(Miyazaki et al., 2000; Raatikainen-Ahokas et al.,2000). BMP4 (possibly similar to BMP2)(Gupta et al., 1999) inhibits ectopic branching of the ureteric bud and is required for growth of the ureter stalk. These studies (Miyazaki et al.,2000; Raatikainen-Ahokas et al., 2000), together with ours, reveal the likely mechanism by which metanephric development is initiated. Ureteric bud formation by the Wolffian duct is independent of Grem1-mediated BMP antagonism, while it is required to induce ureter growth, branching and propagation of RET/GDNF feedback signaling. During branching morphogenesis, Grem1 is expressed locally in the mesenchyme surrounding the invading ureter (this study) and Bmp4 in mesenchyme adjacent to the ureter stalk(Dudley and Robertson, 1997; Miyazaki et al., 2000). Dynamic local changes in BMP activity as mediated by antagonistic Grem1-BMP2/4 interactions may regulate the temporal and spatial kinetics of ureter branching, while BMP signaling alone promotes ureter stalk elongation (Miyazaki et al.,2000; Raatikainen-Ahokas et al., 2000). Grem1-mediated BMP4 antagonism has also been implicated in branching morphogenesis and proximodistal patterning of embryonic lungs (Lu et al.,2001; Shi et al.,2001). Consistent with these results, airway epithelia are defective in lungs of Grem1-deficient newborn mice (this study). In summary, the present study reveals that Grem1-mediated BMP antagonism regulates the dynamic interactions of diverse epithelial and mesenchymal signaling centers during progression of vertebrate organogenesis.

The authors are grateful to I. Ginez, H. Goedemans, N. Lagarde, C. Lehmann and H. Schaller for technical assistance and mouse husbandry. We thank S. Kuc for performing the lacZ in situ analysis; the Zentrum für Mikroskopie der Universität Basel; M. Dürrenberger and M. Düggelin for assistance with SEM analysis; and K. O'Learyfor help in preparation of the manuscript. We thank S. Vainio for expert advice concerning metanephric kidney development; and F. Maina, M. Torres and S. Vainio for providing probes for in situ hybridization. We are grateful to J. Deschamps,R. Dono, A. Galli, G. Holländer, F. Meijlink, I. Mattaj and C. Torres de los Reyes for helpful discussions and comments on the manuscript. This research was supported by the Swiss National Science Foundation (to R.Z.),both cantons of Basel (to R.Z. and K.B.), the Faculty of Biology at Utrecht University, and grants from KNAW (to A.Z.) and NWO (to R.Z.).