The human α-PY sequence (QPGPAPSPALTAPPPAYATLGPR) was obtained by annealing of complementary primers containing partial EcoRI and XhoI restriction sites (underlined) $\underset{¯}{\mathrm{AATTC}}$AGCCAGGCCCT GCTCCCTCTCCAGCCTTGACAGCCCCTCCCCCT GCCTATGCCACCCTGGGCCCCCGC$\underset{¯}{\mathrm{C}}$ and $\underset{¯}{\mathrm{TCGA}}$ $\underset{¯}{\mathrm{G}}$GGCGGGGGCCCAGGGTGGCATAGGCAGGGG GAGGGGCTGTCAAGGCTGGAGAGGGAGCAGG GCCTGGCTG, respectively. The rat α-ENaC C-terminus (residues 633 to 716) was amplified using primers CCC$\underset{¯}{\mathrm{GAATTC}}$GGAGCCGGTACTGGTCTCC and CC CC$\underset{¯}{\mathrm{CTCGAG}}$TCAGAGCGCCGCCA GGCAC. These fragments were cloned in pLEX12 (a modified pBMT116 plasmid, carrying the tetracycline resistance gene). The β- and γ-ENaC C-termini in pBMT116 were generous gifts from O. Staub (‘Institut de pharmacologie et toxicologie’, Lausanne, Suisse). Sequences and expression of hybrid proteins in yeast were verified.

Two-hybrid screenings were performed as described [21] in L40 strain expressing either LEXA-α-rENaC or LEXA-α-hPY fusion proteins, and transformed with 100 μg of oligo(dT) kidney cDNA library, from rat (established in our laboratory) in pGAD3S2X [21] or from human in pACT2 (Clontech, Ozyme, Saint-Quentin-en-Yvelines, France). Transformation efficiencies were estimated on selective medium lacking tryptophan and leucine (DO-WL). Screening with the α-rENaC bait was performed in addition of 5 mM 3-amino-1,2,4-triazole (3-AT) (Sigma-Aldrich, Saint-Quentin-Fallavier, France). His+ clones were selected after 3–5 days of growth at 30 °C on selective medium lacking tryptophan, leucine and histidine (DO-WLH).

Yeast mating was performed between the L40 haploid strain transformed with pLEX plasmids and its complementary haploid strain AMR70 established with pGAD plasmids. β-galactosidase activity of His+ clones was estimated either on filter, using X-Gal (5-bromo-4-chloro-3-indolyl-β-d-galactopyranoside), or in liquid assays, using ONPG (O-nitrophenyl β-d-galactopyranoside) on pellet corresponding to 1 ml of culture at OD₆₀₀ = 0.5. The results are expressed as units as defined by Miller (1972): Unit = (OD₄₂₀) × (1000/time)×(1/OD₆₀₀).

Clones were identified by sequencing of PCR-amplified products obtained from yeast plasmids (forward primer CGATGATGAAGATACCCCACC and reverse primers TAATACGACTCACTATAGGGCGA and GAACTTGCGGGGTTTTTCAG for rat and human libraries, respectively). Sequences were submitted to BLAST SEARCH (NCBI).

The α-rENaC C-terminus (residues 633 to 716) was expressed as GST-αrENaC fusion protein after 0.5 mM IPTG induction, from pGEX-2T plasmid (Amersham Biosciences, Saclay, France). After cell lysis by sonication in PBS buffer containing 1% Triton X-100 and 1% anti-protease cocktail (Sigma), GST-fusion proteins were purified on glutathione–sepharose columns (Amersham Biosciences). In vitro interactions were performed for 2 h at 4 °C with 10 μg of recombinant GST-peptides immobilised on Sepharose 4B glutathione beads and either ³⁵S-labelled peptides obtained by transcription and translation in vitro (TNT quicked coupled transcription/translation systems, Promega, Charbonnières, France) in 20 mM HEPES, pH 7.9, 100 mM KCl, 0.2 mM EDTA, 0.01–0.2% Nonidet NP-40, 1% anti-protease cocktail or 10 μg of His₆-tagged-UBC9 peptide (expressed from pQE80 plasmid; Qiagen, Courtaboeuf; France) in PBS buffer, 0.1% triton X-100. After extensive washings, bound proteins were eluted by GSH and evaluated after SDS-PAGE by either autoradiography or Instant Imager (Packard, Rungis, France) or Western-blots.

Sequence database searches were performed using BLAST programs at the National Center for Biotechnology Information (http://www.ncbi.nlm.nih.gov) and the ‘Pôle bio-informatique lyonnais’ (http://www.pbil.univ-lyon1.fr) websites. The selected sequences were submitted to SIM4 (pbil) and ModelMaker and EvidenceViewer (ncbi) programs to define putative intron–exon organisation taking into account consensus splice signals.

Chromosomal location of hWWP1 was performed by PCR using the Genebridge 4 Radiation Hybrid DNA panel (UK HGMP Resource Centre). Specific primers were chosen in the 3′ non-coding region of hWWP1: CAGCCAAGAAAAATTGCACAG and CAAATCCAGATCTTTATGAATGAAATG.

Two yeast two-hybrid screenings were performed on cDNA library either from rat kidney using the C-terminus of α-rENaC (including the PXXP and PY motifs), or from human kidney with a short bait containing the α-PY motif. Among the selected His+ LacZ+ clones, four partners have been found to interact specifically with α-ENaC (no interaction observed with empty pLEX plasmid) as analysed by yeast mating. Two of them are E3 ubiquitin ligases containing WW domains: Nedd4-2 (n=4), for which an interaction with ENaC has been functionally established [19,20] and WWP1 (n=63), for which an interaction with ENaC has been described, but without functional studies [22]. The two other potential partners UBC9 (n=10) and TSG101 (n=2), members of the E2 ubiquitin/SUMO-conjugating enzyme family, are new ENaC-interacting proteins.

Nedd4-2, member of the Nedd4 family, consists of an alternative C2 domain (Ca²⁺-dependent lipid binding domain) implicated in targeting to the membrane, 3–4 WW domains and a HECT (homologous to the E6-associated protein carboxyl terminus) domain responsible for the ubiquitin-ligase activity [19,20]. Compared to Nedd4-2a, also known as KIAA0439 (accession number AB007899) and Nedd4La (AF210730) isoforms, the Nedd4-2 clones selected by our screening correspond to a new isoform of Nedd4-2 (accession number AY312514), which we named Nedd4-2c (Fig. 1). This isoform is characterised by the presence of a new sequence at the 5′ end. This sequence contains an ATG in frame with the previously described start codon in Nedd4La, producing a protein with 120 additional residues and a complete C2 domain (SMART and PFAM programs). Nedd4-2a has a different 5′ sequence, leading to a partial different C2 domain (Fig. 2A). Our Nedd4-2c clones also contain a 60-nucleotide insert located between the first and second WW domains. Nedd4-2 isoforms are encoded by a gene located on 18q21.31. BLAST analysis of Nedd4-2c cDNA sequence against genomic sequence (contig NT_011085.9), followed by the search for putative exons (see materials and methods) revealed that Nedd4-2 gene is made of at least 34 exons spanning around 354 Kbp. The novel 5′ end sequence found in Nedd4-2c and coding for the beginning of the C2 domain corresponds to exons 1 and 2 followed by exons 5, 6, 7 and 9. The 5′ end described in Nedd4-2a corresponds to exon 3, 5, 6, 7 and 9. Nedd4La mRNA contains exons 4, 5, 6, 7 and 9, but the first ATG is in exon 10, leading to an isoform lacking C2 domain. The putative exon 8, evidenced by the web programs, was not recovered in any of the Nedd4-2 isoforms. The 60-nucleotide insert present in Nedd4-2c is encoded by exon 16. Thus, the clones isolated in the two-hybrid screening correspond to the largest isoform encoded by the Nedd4-2 gene, with a complete C2 domain, four WW domains, a 20-residue insert located between WW1 and WW2 domains, and a HECT domain.

The WWP1 protein shares structural organisation with Nedd4 family members [22,23]. The WWP1 clones selected by our screenings begin in the middle of the C2 domain and contain four WW domains and a complete HECT domain. This HECT domain contains the cysteine required for the ubiquitin-ligase activity, suggesting a potential enzymatic activity (Fig. 2C, black box). The first three WW domains from rat and human WWP1 have the ‘WWP’ consensus pattern of WW domain (two tryptophan separated by twenty residues and followed by a proline) [24], whereas the fourth WW domain of both rat and human WWP1 lacks the second conserved W (Fig. 2B, black boxes). The human WWP1 gene has been previously located in 3q2.1. However this sequence (AC016962) differed slightly from our WWP1 clones (95% identity), as well as from the one reported by Pirozzi [22]. Moreover, the presence of three stop codons and the absence of intronic sequences suggested that this sequence in 3q2.1 could correspond to a pseudogene. The data we obtained by analysis of Hybrid Radiation DNA Panel allowed to precise a unique location of the human WWP1 gene in 8q21. This location has been recently confirmed by FISH [25].

The clones corresponding to UB2I/UBC9 (NM_ 003345) contain the complete catalytic domain UBCc, usually found in E2 ubiquitin-conjugating enzyme. However UBC9 conjugates an ubiquitin homologue known as SUMO instead of ubiquitin [26]. Like UBC9, TSG101 (U52945) contains an UBCc domain. However, the UBCc domain of TSG101 does not contain the cysteine residue required for the conjugating activity. The TSG101 clones contain almost full-length sequence, beginning six residues after the initiating methionine.

The involvement of the PY motif in ENaC binding activity has been evaluated by the measurement of β-galactosidase activities in the two-hybrid system. Concerning Nedd4-2c, WWP1 and UBC9, the β-galactosidase activities were two to ten times higher in the presence of the complete C-terminus of α-ENaC than in the presence of the PY motif (Fig. 3A). These data confirm the involvement of the PY sequence in the interactions, but show that additional sequences reinforced these interactions, particularly for UBC9. In contrast, TSG101, which was only selected by the C-terminus of α-ENaC, did not interact with PY sequence, suggesting that either the PY motif is not the binding site for TSG101 or flanking sequences of PY are required for the interaction with TSG101.

As the three ENaC subunits are homologous, we tested the interaction between each ENaC subunit and the four α-ENaC-interacting proteins in the two-hybrid system (Fig. 3B). The highest β-galactosidase activity occurred in the interaction between WWP1 and γ-ENaC, whereas half of this activity value was found in the interaction with α- and β-subunits. The Nedd4-2c isoform interacted better with β- and γ-subunits than with α-subunit, whereas UBC9 interacted mainly with α- and γ-ENaC. Interestingly, TSG101 interacted specifically with α-ENaC, as no binding activity could be detected with β- and γ-ENaC. These results show that the selected proteins exhibit different interaction affinities according to the three ENaC subunits.

All the Nedd4-2c and WWP1 selected clones contain at least the last two WW domains (WW3 and WW4) with a partial sequence of WW2, suggesting that these domains are required in the interactions. We studied if additional sequences, in particular the C2 domain, might participate in the interactions, since isoforms lacking the C2 domain have been described [20,25]. We tested the effect of the complete C2 domain on the interaction between ENaC subunits and the Nedd4-2c isoform or WWP1 (Fig. 3C). The C2 domain has been found to reinforce the interaction between Nedd4-2c and each individual ENaC subunits when compared to Nedd4-2c lacking the C2 domain. In contrast, the presence of the C2 domain decreased the interaction between the WWP1 protein and ENaC subunits. These results suggest that the presence of the C2 domain could modulate the affinity of the Nedd4 family members for their target proteins.

Interactions between α-rENaC and the potential partners were analysed in vitro by GST-pull-down assays (Fig. 4). Rat and human WWP1 and rat TSG101 were produced as ³⁵S-methionine-labelled peptides by transcription and translation in vitro and human UBC9 was produced as His₆-tagged recombinant peptide. The C-terminus of α-rENaC, expressed as GST fusion protein, retained specifically partial rat (60 kDa) and complete human (115 kDa) WWP1, as well as rat TSG101 (43 kDa) in a higher extend compared to GST (Fig. 4A). Coomassie blue staining and Western-blot revealed that UBC9 (18 kDa) specifically bound to GST-α-rENaC (Fig. 4B and C).