To identify novel morphoregulatory molecules possibly involved in kidney development, we generated a murine monoclonal antibody (mAb) against a cell surface antigen of the CD cell line RC.SVtsA58 (Prié et al., 1991). This source of antigen was chosen because the cell line has a principal cell phenotype, but with some characteristics of intercalated cells (Prié et al., 1994), suggesting that it could express developmentally regulated antigens. In this work, we show that the mAb-recognized antigen is a splicing isoform of the neural cell adhesion molecule L1.

L1 is a member of the immunoglobulin superfamily (IgSF). It is composed of six Ig-like domains followed by five fibronectin type III-like repeats, a transmembrane domain, and a cytoplasmic domain that confers cytoskeletal association via linker proteins (Davis and Bennett, 1994; Brümmendorf and Rathjen, 1996; Dahlin et al., 1997). This molecule is engaged in complex extracellular interactions, with multiple binding partners on cell surface and in the extracellular matrix, triggering distinct signal transduction pathways (Kamiguchi and Lemmon, 1997; Walsh and Doherty, 1997). L1 is primarily expressed in the embryonic and adult nervous system, where it is involved in diverse neuronal processes (Brümmendorf and Rathjen, 1996; Hortsch, 1996). Consistent with a role in development, mutations in the human L1 gene are associated with a variable phenotype including mental retardation and anomalous development of the nervous system, referred to as “CRASH” syndrome (Fransen et al., 1997; Brümmendorf et al., 1998). Similarly, mutations in the Drosophila L1 homologue, neuroglian, result in abnormal guidance of motorneurons (Bieber et al., 1989; Hall and Bieber, 1997). Phenotypes of L1-deficient mice (Cohen et al., 1997; Dahme et al., 1997) confirm that L1 is crucial for embryonic brain histogenesis, in particular for the development of axon tracts and pyramidal decussation. Outside the nervous system, the effects of L1 gene mutation or invalidation have not been analyzed; however, L1 appears to be important for interactions between leukocytes and between epithelial crypt cells of the small intestine (Brümmendorf and Rathjen, 1995; Kadmon and Altevogt, 1997).

To understand the role of L1 in kidney morphogenesis, we first studied its ontogeny by a combined immunohistochemical and immunoblotting approach. We showed that among organs developing by branching morphogenesis, the kidney was the only one that expressed L1. L1 was detected since the mesonephric stage and was developmentally regulated both in its segment distribution and molecular processing. In vitro using the organotypic culture model, we found that addition of anti-L1 antibody perturbed the development of the ureter bud epithelium, suggesting a functional role for L1 in CD development. Based on the properties of L1 and its distribution pattern, we further suggest a possible role of L1 in the guidance of developing distal tubule and in the generation and maintenance of the terminal phenotype in the CD.

In some experiments, SDS denatured membrane proteins were deglycosylated with PNGase F (Boehringer Mannheim GmbH, Mannheim, Germany) using buffers and conditions as recommended by the manufacturer.

For macrodissection of cortex, outer, and inner medulla, E29 kidneys were excised, and the three zones were isolated with dissection scissors. For microdissection of nephron segments, the adult rabbit kidney was perfused with cold saline until blood free, followed by cold DME (Sigma Chemical Co., St. Louis, MO) containing 1.5 mg/ml collagenase. The kidney was removed, cut into coronal slices, and incubated at 37°C for 30 min in the DME-collagenase solution. Slices were rinsed with ice-cold PBS, placed into DME medium containing 1.0% FBS, and maintained at 4°C during microdissection. Glomeruli, proximal convoluted tubule, medullary thick ascending limb, cortical thick ascending limb, cortical collecting duct (CCD), outer medullary collecting duct (OMCD), and inner medullary collecting duct (IMCD) segments were dissected under microscope.

Aliquots of solubilized membranes prepared from kidney, brain, or cells were loaded onto the affinity column at a flow rate of 3 ml/h. The column was extensively washed with three buffers successively: (a) 20 mM Tris, pH 7.4, 1% NP-40; (b) 20 mM Tris, pH 8.0, 0.5% NP-40, and 0.5 M NaCl; and (c) 50 mM Tris, pH 9, 0.5 M NaCl. Finally the antigen was eluted with 50 mM triethanolamine, pH 11.5, containing 150 mM NaCl and 1% octyl glucoside, and quickly neutralized with HCl. Fractions were analyzed using SDS-PAGE, and peak fractions were concentrated in a Centricon-30 (Amicon, Inc., Beverly, MA).

For protein sequencing, the immunopurified samples were deglycosylated with PNGase F (Boehringer Mannheim GmbH) and separated on 8% SDS gels. After transfer to PVDF membrane (Trans-Blot Transfer Medium; Bio-Rad Laboratories) protein spots were excised from Coomassie brilliant blue–stained membranes. NH₂-terminal sequencing was performed by automated Edman degradation using a protein sequenator (model 473A; Applied Biosystems, Inc., Foster City, CA).

Immunolocalization of L1 in the adult kidney was studied using techniques on frozen ultramicrotome sections. Tissues were prepared and fixed as previously described (Christensen et al., 1995). Ultrathin sections of 0.8 μm and 70–90 nm were obtained at −80°C for light microscopy and −100°C for electron microscopy, respectively, using a cryoultramicrotome (model Ultracut S; Reichert-Jung, Vienna, Austria), and collected as described previously (Christensen et al., 1992). For immunolabeling, sections were preincubated in PBS containing 0.05 M glycine and 0.1% nonfat dry milk and subsequently incubated with mAb 272 or a control isotype-matched murine IgG (5–10 μg/ml) overnight at 4°C. For light microscopy, sections were further incubated with peroxidase-conjugated secondary antibody (Daco A/S, Glostrups, Denmark). Peroxidase was visualized with diaminobenzidine. For electron microscopy, bound immunoglobulins were visualized with 10-nm gold-labeled goat anti–mouse Ig (Bio Cell, Cardiff, UK). The sections were embedded in methylcellulose and examined with an electron microscope (model CM 100; Phillips Electron Optics, Mahwah, NJ). Immunolocalization of L1 by electron microscopy in E29 kidney was done by using the same method as for the adult tissues.

Figure 1 Immunodetection and purification of the antigen identified by mAb 272. (A and B) Western blot analysis of proteins solubilized from membrane fractions prepared from CD cell lines or kidney and brain isolated from rabbit and human species. Equal (more ...)

To characterize the kidney and brain rabbit antigen, the NP-40–solubilized membranes were submitted to immunoaffinity chromatography using mAb 272. The unreduced eluate was analyzed by SDS-PAGE and silver staining (Fig. 1 C) and by Western blotting (Fig. 1 D) before and after PNGase F treatment. In addition to the expected bands previously detected by Western blotting in kidney and brain membrane extracts (Fig. 1, A and B), an additional protein component was copurified both in kidney and brain. After PNGase F treatment, the immunopurified antigen resolved in three bands at 150, 100, and 50 kD in the kidney as well in the brain (Fig. 1 C), but the 50-kD band was not reactive with the antibody (Fig. 1 D), which suggests that this protein component is strongly associated with the first two by noncovalent bonds.

To identify the nature of the antigen and of the tightly associated protein eluted from the immunoaffinity column, the eluate was resolved by SDS-PAGE and electrotransferred to PVDF membrane, and the NH₂-terminal sequence of each of the three bands was determined. We analyzed the antigen purified from rabbit kidney and brain because of its peculiar abundance in these two tissues. We also sequenced the antigen isolated from the rabbit CD cell line RC.SV3 because CD cells represent a minor cell population in the whole kidney, which is composed of many other different cell types. All NH₂-terminal sequences are shown in Table I. The search for homologous sequences in the Genbank protein sequence database revealed close homology (85 to 90% identity) of the NH₂-terminal sequences of the rabbit brain 150-, 100-, and 50-kD components with corresponding protein sequences of the mouse (Moos et al., 1988), rat (Miura et al., 1991), and human (Hlavin and Lemmon, 1991) neuronal cell adhesion molecule L1. The degree of homology between rabbit and the other species was quite similar, to the degree of interspecies homology previously reported. We therefore concluded that the kidney and brain antigen was the rabbit orthologue of the murine and human L1.

Table I NH2-terminal Sequences of Immunoaffinity Chromatography–purified Rabbit Proteins: Homologies with Known Proteins

The basic structural plan of L1 molecule is presented in Fig. 2. The NH₂-terminal sequence of the 150- and 100-kD components corresponds to that of the mature protein predicted by the cDNA sequence, whereas the NH₂-terminal end of the 50-kD component starts at amino acid 845 in the middle of the third fibronectin type III repeat. As the other members of its group, L1 is synthesized as a 150-kD precursor polypeptide, subsequently cleaved at a conserved sequence in the third fibronectin type III repeat. The resulting 100- and 50-kD polypeptides remain tightly associated in cell membranes as heterodimers. This explains why although the antibody was not reactive with the 50-kD component, this polypeptide was purified with its 100-kD partner, which expresses the reactive epitope.

Figure 2 Model of the domain structure of L1. Ig domains are shown as loops and fibronectin III repeats are shown as boxes. The transmembrane region is indicated by a vertical bar. The arrow indicates the phylogenetically conserved cleavage site at amino acid (more ...)

Sequencing the 100-kD component of the rabbit kidney L1 revealed an unexpected finding. Five amino acids were missing in both the kidney and the RC.SV3 cell line antigens compared with the sequence identified in the brain peptide (Table I). In addition, the first amino acid after this deletion was substituted, with a leucine in the kidney replacing a valine in the brain antigen (Table I). Comparison with the genomic structure of the mouse and human genes (Coutelle et al., 1998) shows that the missing sequence and the substitution of the next amino acid correspond exactly to the splicing event of the mini-exon 3, which was called exon 2 or B until an additional untranslated exon was identified at the 5′ end of the gene (Kallunki et al., 1997). Thus, the kidney epithelium expresses an alternative splicing isoform of the brain L1 characterized by the skipping of exon 3 sequence.

Figure 3 Immunohistochem-ical localization of L1 protein during development of rabbit kidney. Cryosections were incubated with mAb 272 and visualized by indirect peroxidase staining. (A and B) L1 is expressed at low level in the Wolffian duct (B) but not in (more ...)

The formation of definitive kidney (metanephros) is initiated by interaction of the ureteric bud with the metanephric blastema that occurs at E11.0 in the rabbit (Saxen, 1987). At E11.0, the rising ureteric bud grows caudally into an adjacent region of the metanephric mesenchyme. At E11.5, after ureteric bud–mesenchyme interaction, a weak staining became visible in the ureteric bud epithelium, mainly at the apical aspect of the cells (Fig. 3 C). The antigen was not detected in the induced condensing mesenchyme (Fig. 3 C). At E12.5, L1 was clearly expressed in the T-shaped ureteric bud that has just divided within the metanephric mesenchyme, as well as in the developing ureter (Fig. 3 E). Staining predominated on the apico–lateral cell membranes, whereas the basal side remained negative (Fig. 3 F). Later on, the ureter bud branches repeatedly and the condensing mesenchyme surrounding the tips of the bud branches starts its epithelial differentiation. At E15 (Fig. 3, G and H), L1 expression was still restricted to the epithelium of the ureter bud derivatives (Fig. 3 G). No staining was observed in mesenchyme-derived structures, including condensates, comma-shaped bodies, and S-shaped bodies (Fig. 3 H). At E27 (Fig. 3, I–K), the S-shaped bodies elongate and join with the terminal ampulla of the CD (Fig. 3 I). L1 immunoreactivity then appeared at the lateral membranes of the cells, which form the most distal part of the newly formed primitive nephron (Fig. 3 I). In more mature nephrons, in which the distal tubule has further elongated (Fig. 3 J), L1 was still detected in connecting tubule segments in the cortex (Fig. 3 J) as well as in the elongating tubules that will form the loops of Henle (Fig. 3 K). L1 expression in these tubule segments was, however, transient and declined after birth. L1 immunoreactivity was not observed at any time in proximal tubules and glomeruli. It was not detected on the ureter smooth muscle cells, vascular endothelium, or stroma cells.

In the adult kidney, we had previously reported that the antigen recognized by mAb 272 was a marker of P cells in the cortical segment of the CD (Prié et al., 1994). Our new ontogenic data, and particularly L1 expression in the embryonic distal tubule, led us to reappraise L1 distribution in the adult kidney. In the distal tubule, L1 did persist albeit in small amounts, on the basal plasma membrane of connecting cells, but it was not detected on IC cells (Fig. 4 A). In the CD, there was no reaction on IC cells but strong reactivity was seen on the basolateral cell membrane of the principal cells in cortex, as well as in outer and inner medulla (Fig. 4, A–C). Serial adjacent sections showed absence of L1 staining from CD cells that expressed H⁺-ATPase, a marker for IC cells (not shown). In the deep segment of the IMCD where IC cells and ciliated P cells are replaced by progressively taller cells with less basolateral plasma membrane infoldings, L1 protein then became restricted to the lateral membrane of the tall cells (Fig. 4 D). The same subcellular distribution was seen in papilla-forming cells.

Figure 4 Immunolocalization of L1 protein within different adult renal segments. Semithin cryosections were incubated with mAb 272 and visualized by indirect peroxidase staining. (A) Transition between the connecting tubule (CNT) and the initial collecting (more ...)

L1 was not detected in endoderm-derived structures, including urethra and urinary bladder (not shown). It was also not detected during ontogeny of other organs, including the lung, intestine, salivary gland, and pancreas, which develop as the kidney by branching morphogenesis of an epithelial structure (not shown).

Figure 5 Immunogold electron micrograph showing ultrastructural localization of L1 in the embryonic and fully differentiated adult CD epithelium. (A) Embryonic kidney at E27, outer medulla. The gold particles are seen along the lateral membranes of two epithelial (more ...)

Figure 6 Immunochemical analysis of L1 ontogeny in rabbit kidney. Kidneys were isolated at indicated times from embryonic, postnatal and adult rabbits. (A–D) Western blot analysis. Proteins were solubilized from membrane fractions prepared from whole (more ...)

To support the assumption of a maturation process of L1 during development, L1 expression was examined by Western blotting in the zonal sections of E29 kidney (Fig. 6, C and D) and by immunoaffinity chromatography of E14 and E29 samples (Fig. 6 E). At E29, the cortical area consists predominantly of the growing CD tips, undifferentiated mesenchyme, and its early differentiation stages. This zone essentially showed the 150-kD component (Fig. 6 D). In deeper parts of the kidney that contain late and terminal stages of epithelial differentiation, the 100-kD component dominated. By comparison in the adult kidney, where the nephrogenic zone has disappeared from the upper cortex, L1 was expressed as a heterodimer with a prevailing 100-kD component in the CD segments throughout kidney zones (Fig. 6 B). Since our antibody did not recognize the 50-kD polypeptide, further characterization of the embryonic kidney antigen was performed by immunoaffinity chromatography followed by SDS-PAGE after antigen deglycosylation (Fig. 6 E). At E14 and in the superficial cortex of E29 kidney, L1 was predominantly expressed as the 150-kD precursor, whereas in the rest of E29 kidney, L1 was mainly composed of the 100- and 50-kD polypeptides as in the adult kidney.

Figure 7 Effect of anti-L1 antibody on kidney and lung branching morphogenesis. Kidney (A–E) and lung rudiments (F and G) were isolated from E11.5 rabbit embryos. Kidney explants sampled at E11.5 (A) were grown on Nucleopore filters for 24 h (B and C (more ...)

Figure 8 Photomicrographs of paraffin sections from E11.5 kidneys grown for 96 h in control IgG (A) and in the presence of anti-L1 mAb (B and C). Control IgG and anti-L1 mAb were added to culture media at a concentration of 150 μg/ml. Sections were (more ...)

E11.5 kidneys were then cultured in the presence of monoclonal anti-L1 antibody (mAb 272) or mouse monoclonal IgG as a control. Antibody penetration in tissue was demonstrated by direct immunofluorescence of kidneys with FITC-linked anti–mouse IgG to localize anti-L1 mAb (not shown). Treatment with anti-L1 antibody did not alter T-shaped branching of the ureter bud (Fig. 7 C) in the first 24 h of culture. By contrast, after 96 h (Fig. 7 E), it induced dramatic alterations of branching morphogenesis and a marked reduction of kidney size (Fig. 7 E and Fig. 8, B and C). The ureteric bud and its primary and secondary divisions were notably shorter and dilated, while the tertiary divisions were rudimentary. The tips of the ureteric bud branches, the site of nephron formation, were blunted. Nevertheless, the ureteric bud was still able to induce nephrogenesis since condensates and early epithelial differentiation stages, including renal vesicles and comma-shaped bodies, had developed in the periphery of the organ (Fig. 8, B and C). Impairment of ureter bud morphogenesis by anti-L1 antibody was concentration dependent (Fig. 9). The effects of anti-L1 antibody became visible at 50 μg/ml and were maximum at 150 μg /ml. No alteration was observed in the kidney grown in basal medium or in the presence of mouse IgG in the same concentration range.

Figure 9 Concentration-dependent effect of anti-L1 mAb on branching morphogenesis of kidney. Kidney explants were cultured for 96 h in the presence of basal medium alone (Control) and basal medium with mouse IgG at a concentration of 50, 100, and 150 μg/ml, (more ...)

Lung rudiments sampled from the same E 11.5 embryos were grown in parallel with kidneys under the same experimental conditions. Contrary to kidney explants, anti-L1 antibody did not induce significant alterations of lung branching morphogenesis after 96 h of culture (Fig. 7, F and G). These results indicate that the antibody has no effect on nonrenal epithelial structures developing by branching morphogenesis but that it specifically impairs kidney development.

In this work, we have identified the antigen defined by mAb 272 in the CD epithelium as an isoform of the neuronal cell adhesion molecule L1, a member of IgSF. We analyzed its renal expression during ontogeny by a combined biochemical and immunohistochemical approach. In addition, we first demonstrate that L1 is implicated in branching morphogenesis of the ureter bud in organ culture.

Identification of the CD antigen as L1 is based on the following data: (a) The rabbit immunopurified polypeptides have 85–90% amino acid identity with the corresponding regions of human, mouse, and rat L1 protein. (b) The 100- and 50-kD polypeptides originate from one 150-kD precursor molecule and are noncovalently associated in the cell membrane, a common feature of all members of the L1 subfamily. (c) The missing sequence in the NH₂ terminus of the rabbit kidney antigen corresponds exactly to the mini-exon detected in the human and mouse L1 genes (Coutelle et al., 1998), which suggests that at least at the 5′ end of the coding sequence, the rabbit L1 gene has the same organization as those genes. The skipping of exon 3 that we detected in the kidney has already been described in B lymphocytes (Ebeling et al., 1996), but it has never been observed in cells of neuronal origin (Miura et al., 1991). The region encoded by exon 3 precedes the Ig domains endowed with strong homophilic binding activity (Zhao and Siu, 1995). A recently found mutation in a patient with X-linked hydrocephalus leads to exon 3 skipping (Jouet and Kenwrick, 1995). Thus, the region encoded by exon 3 appears to be crucial to the function of the neuronal L1 protein but not to that of the renal and the lymphocytic proteins. L1 protein is encoded by a single gene composed of 29 exons, with only two splicing products as yet identified. The other reported splicing concerns exon 28, a 12-bp exon located in the 3′ region of the gene, which is spliced out in L1 mRNA in the peripheral nervous system (Miura et al., 1991) and leukocytes (Ebeling et al., 1996). By using reverse transcription PCR surrounding exon 28, we found that this exon was also spliced out in the rabbit epithelial cell line RC.SV3 (data not shown). Whether splicing of small exons 3 and 28 can occur independently in renal cells and whether it is developmentally regulated is not yet known. A formal proof that these particular combinations of exons may exist as RNA transcripts would require additional work beyond the scope of this paper.

For some time, kidney expression of L1 has been controversial. Searches for L1 protein (Mujoo et al., 1986) or mRNA (Moscoso and Sanes, 1995) in renal tissue were unsuccessful. On the other hand, it was reported that anti-L1 antibody, used as a marker for neurons in the metanephros, also stained ureteric bud epithelial structures (Sainio et al., 1994, 1997). However neither the precise localization of L1 in the kidney nor its role in nephrogenesis had been defined. We show that the protein is expressed in a temporally dynamic and spatially restricted pattern within the developing and adult kidney epithelium, but with contrasting kinetics in ureter bud and mesenchyme derivatives. L1 expression can be considered as constitutive in epithelial structures that originate from the mesonephric duct (Wolffian duct)—that is, the collecting duct system and the urogenital tract epithelia (Kujat et al., 1995)— since it appears early and persists in all cells except for intercalated cells in which L1 is downregulated in the terminally differentiated CD. By contrast in the metanephric mesenchyme derivatives, L1 appears late in the distal portion of elongating newly formed tubules and disappears from the distal tubule after its patterning has been completed. Only connecting cells show a persistent but weak basal expression.

In parallel to its spatio–temporal regulation during ontogeny, L1 undergoes substantial posttranslational modifications, including proteolytic cleavage and glycosylation. These modifications may play an important role in renal organogenesis. The degree of proteolytic cleavage increases during renal development parallel with epithelial maturation. Cleavage of the 150-kD precursor within the third fibronectin-like domain yields two chains, each containing structural motifs that could contribute to the different functions of L1 (Burgoon et al., 1995). Immunopurification experiments have shown that the resulting 100- and 50-kD fragments remain tightly associated. In addition, the 100-kD polypeptide, which lacks a transmembrane domain, may also become anchored in the cell membrane via other cell surface proteins such as integrins (Montgomery et al., 1996), GPI-linked proteins (Olive et al., 1995), or yet unidentified molecules leading to the generation of secondary response mediated by these receptors. Alternatively, the 100-kD fragment may be shed and become deposited in the extracellular matrix. Thus, increased proteolytic cleavage may stimulate a variety of autocrine or paracrine intracellular pathways. The augmented carbohydrate content of L1 when renal maturation proceeds may also play an important role since carbohydrate structures are directly involved in cell–cell interactions (Dwek, 1995). For example, carbohydrates expressed on the neuronal form of L1 determine its interactions with N-CAM and integrins (Horstkorte et al., 1993; Hall et al., 1997).

To evaluate the role of L1 in the first events of kidney morphogenesis, functional studies were performed. Addition of mAb 272 to kidney organotypic cultures induced dysmorphogenesis of the ureteric bud epithelium in a concentration-dependent manner but did not prevent mesenchyme induction or vesicle and comma-shaped body formation. The anti-L1 antibody used in this study reacts with the extracellular domain, a major adhesion site for several hetero- and homophilic ligands (Appel et al., 1993). Moreover, bivalent mAbs that bind to extracellular L1 domains have also been shown to trigger L1 association with signaling receptors (Klinz et al., 1995). Therefore, it is likely that mAb 272 perturbed interactions of L1 with its natural ligands and signaling systems, and this in turn resulted in abnormal ureteric bud development. Explant experiments were performed at 11.5–15.5 d, before appearance of the 100-kD band as a single entity, indicating that only the 150-kD precursor could be the target of mAb 272 at that time, although the 100-kD cleaved subunit might also play a developmental role later on. Involvement of L1 in bud initiation or initial inductive interactions seems unlikely since its expression follows rather than precedes these events, and T-shaped branching is not inhibited by mAb 272. Furthermore L1 is not detected at the basolateral side in the tips of the ureteric bud, where the presumptive inducing signals are believed to act. These data suggest that L1 could control ureter bud development in later stages, i.e., after intervention of the GDNF/c-ret axis, which seems to be responsible for the early steps of ureteric bud formation and extension (Sainio et al., 1997).

Generation of the arborescent collecting duct system is a complex process requiring constant cytoskeleton rearrangements at the growing ureteric bud tips (Sakurai and Nigam, 1998). Tubulogenic growth factors and their receptors, such as HGF/c-met tyrosine kinase receptor, regulate branching morphogenesis by modulating the function of linker molecules between membrane and cytoskeleton (Cantley, 1996; Weidner et al., 1996; Crepaldi et al., 1997). A parallel can be drawn between renal tubulogenesis and the development of axon tracts, which is profoundly altered in L1 knock-out mice (Cohen et al., 1997; Dahme et al., 1997). L1-mediated guidance of axons is based on its interaction with the actin cytoskeleton via linker proteins, including ankyrin (Davis and Bennett, 1994) and a still unknown linker protein (Dahlin et al., 1997). The connection of L1 with linker proteins is reversible and controlled by a kinase-phosphatase system activated in response to specific extracellular interactions of L1 with its binding partners (Maness et al., 1996; Garver et al., 1997; Kamiguchi and Lemmon, 1997). Whether the cytoplasmic tail of L1 kidney isoform associates with the actin linker molecules in response to extracellular interaction remains to be determined. To clarify the role of L1 in these morphogenetic events, it will be essential to identify L1 ligands and signaling pathways in the ureter bud epithelium.

The other major site of L1 expression is the developing nephron. Development of renal tubules results from epithelial transdifferentiation of the mesenchyme and subsequent elongation. The late and transient L1 expression in the elongating distal tubules argues for its role in the guidance of tubule growth into the surrounding loose mesenchyme and in the arcade forming process, rather than for its implication in the maintenance of tubule segment specificity.

We have demonstrated by electron microscopy that L1 is expressed at the lateral membrane by all progenitor cells in early stages of CD development but is absent from IC cells in mature CD. Recent works have suggested that P and IC cells originate from a single precursor ureteric bud cell type (Aigner et al., 1995; Jamous et al., 1993, 1995). It has been established that cell surface receptors and their ligands participate in cell-to-cell and cell–matrix signaling events that influence cell fate decision during development (Greenwald and Rubin, 1992; Schmidt et al., 1993). Thus, L1 could also play a role in the terminal differentiation of CD epithelium. Identification of the interacting components involved in L1 mediated signaling is the next step toward understanding how this molecule could drive the generation of two distinct phenotypes in CD epithelium.

Finally, the potential of L1 to associate with ankyrin, which interacts also with the cytoplasmic domains of several ion channels and pumps (Davis et al., 1996; Dubreuil et al., 1996), could be helpful during maturation in the organization of epithelial cell polarity by the segregation of membrane components. L1 characteristic expression in the basolateral membrane of differentiated P cells is compatible with this function. In addition, in the tall cells of the IMCD and the papilla where L1 is localized laterally, the multicomplexes formed by L1, integral membrane proteins and cytoskeleton, could provide a structural support allowing distribution of shear stress throughout the tissue.

Epithelial–mesenchymal interactions drive branching morphogenesis during the development of a number of organs that fail to express L1 in the epithelial compartment. Thus, our results indicate that L1 has a kidney-specific function during morphogenesis. In addition, the involvement of L1 in multiple stages of renal epithelial morphogenesis may bear a clinical interest as in the nervous system. Mutations in L1 gene or disturbances in posttranslational modifications might lead to developmental abnormalities affecting the integrated physiology of the CD system.

We are grateful to Dr. Marie-Hélène Metz-Boutigue (INSERM U 338, Strasbourg, France) for amino acid sequencing, Dr. Brigitte Lelongt for help with embryo dissection and organ culture, Madelaine Delauche, Inger Blenker Kristoffersen, and Hanne Sidelmann for excellent technical assistance in histochemistry and electron microscopy, Béatrice Baudouin for growing the cells, and Christine Vial for secretarial help.

This work was supported by grants from INSERM, Paris VI University, the Danish Medical Research Council, and the Novo Nordic Foundation.