Identification of PKDL, a Novel Polycystic Kidney Disease 2-Like Gene Whose Murine Homologue Is Deleted in Mice with Kidney and Retinal Defects*

Polycystin-1 and polycystin-2 are the products ofPKD1 and PKD2, genes that are mutated in most cases of autosomal dominant polycystic kidney disease. Polycystin-2 shares ∼46% homology with pore-forming domains of a number of cation channels. It has been suggested that polycystin-2 may function as a subunit of an ion channel whose activity is regulated by polycystin-1. Here we report the identification of a human gene, PKDL, which encodes a new member of the polycystin protein family designated polycystin-L. Polycystin-L has 50% amino acid sequence identity and 71% homology to polycystin-2 and has striking sequence and structural resemblance to the pore-forming α1 subunits of Ca2+channels, suggesting that polycystin-L may function as a subunit of an ion channel. The full-length transcript of PKDL is expressed at high levels in fetal tissues, including kidney and liver, and down-regulated in adult tissues. PKDL was assigned to 10q24 by fluorescence in situ hybridization and is linked to D10S603 by radiation hybrid mapping. There is no evidence of linkage to PKDL in six ADPKD families that are unlinked toPKD1 or PKD2. The mouse homologue ofPKDL is deleted in Krd mice, a deletion mutant with defects in the kidney and eye. We propose that PKDL is an excellent candidate for as yet unmapped cystic diseases in man and animals.

Polycystin-1 and polycystin-2 are the respective gene products of PKD1 and PKD2, mutations in which account for ϳ95% of cases of ADPKD. 1 ADPKD affects up to 1/1,000 individuals and is associated with a 50% incidence of end-stage renal failure by the sixth decade of life (1). At least one additional gene is known to be mutated in the ADPKD population (2,3) but has yet to be identified.
Polycystin-1 encodes a 4,303-amino acid plasma membrane protein with a large extracellular N-terminal domain that contains leucine-rich repeats, a C-type lectin domain, and an LDL-A-like domain, all three of which are involved in cell-cell or cell-matrix interactions in other proteins (4 -6). These domains are followed by 16 repeats of the so-called PKD domain and by an REJ (receptor for egg jelly in sea urchin sperm)-like domain.
The predicted amino acid sequence of the PKD2 gene is homologous to the C terminus of polycystin-1 (9,10). Polycystin-2 is a 968-amino acid protein with ϳ46% sequence similarity to each domain of the pore-forming ␣1 subunits of Ca 2ϩ and other cation channels, and like these channel subunits, it is predicted to have six transmembrane domains. Polycystin-2 has a putative Ca 2ϩ binding structure (EF-hand) in its Cterminal cytoplasmic domain. It interacts biochemically with polycystin-1 and with itself (7,8).
Here we report the identification, chromosomal localization, and expression of a third gene encoding a protein of the polycystin family. The product of this gene is an excellent candidate for a component of the pore-forming subunit of a polycystinrelated channel and is also a candidate for various human and murine cystic diseases.

EXPERIMENTAL PROCEDURES
Isolation of PKDL cDNAs-Overlapping EST sequences W27963 and W28231, derived from a retina cDNA library, were identified by their gene product homology to polycystin-2 (gb 189 U50928). A 340-base pair fragment in the overlap region of both ESTs was amplified from human adult kidney and brain poly(A)-selected RNA by reverse transcription-PCR using primers 5Ј-TCTTCGTGCTCCTGAACATG-3Ј and 5Ј-CCT-GTCGCATTTTTCCTGTT-3Ј. 5Ј-and 3Ј-rapid amplification of cDNA ends were performed with human skeletal muscle and kidney rapid amplification of cDNA ends kits (CLONTECH, Palo Alto, CA), respectively. Primers were designed based on PKDL reverse transcription-PCR products. Nested amplification was performed following manufacture's instructions. The 5Ј-rapid amplification of cDNA ends product was random-labeled with [ 32 P]dCTP and used to screen a human retina cDNA library (CLONTECH). Hybridization was performed in a buffer containing 5 ϫ SSC (1ϫ SSC, 0.15 M NaCl and 0.015 M sodium citrate), 50% formamide, 1% SDS, and 5ϫ Denhardt's solution at 42°C, overnight. Filters were washed three times in buffer (0.1ϫ SSC and 0.1% SDS) at 65°C. Positive signals were purified, and inserts were subcloned into pBluescript II (Stratagene, La Jolla, CA) and sequenced.
Sequence Analysis-Clones were sequenced from both strands, and the sequences were aligned to give an overall consensus sequence. The computer program MOTIFS of GCG (11) was used to identify putative glycosylation and phosphorylation sites (20). Prediction of coiled-coil structure by Lupas' algorithm (12) was performed with the COILS computer program with and without 2.5-fold weighting of positions a and d, whereas prediction by Berger's algorithm (13) was performed with Paircoil program. Kyte and Doolittle's hydropathy analysis (14) was performed using PEPPLOT (GCG). Secondary structure analysis was performed with PEPTIDESTRUCTURE (GCG) computer program and with the protein sequence analysis software using type-2 discrete state-space models (15,16). Analysis of transmembrane segments was performed with the TMpred computer program (17) using windows of 18 -33 residues and with SOSUI 2 and TMAP (19). Alignments with homologous sequences from polycystin-1, polycystin-2, and Ca 2ϩ channel ␣1 subunits were performed with LINEUP and BESTFIT (GCG) and optimized by visual comparison.
RNA Dot Blot and Northern Hybridization-Human adult and fetal RNA blots (CLONTECH) were hybridized with a randomly labeled 5Ј-most 1.5 kb of the coding sequence of PKDL in 5 ϫ SSC, 50% formamide, 1% SDS, 5ϫ Denhardt's solution at 42°C overnight. Filters were washed twice in 2 ϫ SSC, 0.1% SDS at room temperature and at 50°C. Signals were visualized by autoradiography.
Fluorescent in Situ Hybridization-A 1.7-kb human PKDL genomic fragment between cDNA positions 2,006 and 2,206 was PCR-amplified with exonic primers and subcloned into pCRII (Invitrogen, Carlsbad, CA). One g of this vector was labeled with digoxigenin-11-dUTP as described previously (20), coprecipitated with 10 g of Cot-1 DNA, and resuspended in 1ϫ Tris-EDTA at 200 g/ml. Hybridization of metaphase chromosome preparations from peripheral blood lymphocytes of normal human males was performed with PKDL at 10 g/ml in Hybrisol VI as described previously (21). Digoxigenin-labeled probe was detected using reagents supplied in the Oncor Kit (Oncor, Gaithersburg, MD) according to the manufacture's recommendations. Metaphase chromosomes were counterstained with 4,6-diamidino-2-phenylindole dihydrochloride. Map position of PKDL was determined by visual inspection of the fluorescent signal on the 4,6-diamidino-2-phenylindole dihydrochloride-stained metaphase chromosomes using a Zeiss Axiophot microscope. Images were captured and printed using the CytoVision Imaging System (Applied Imaging, Pittsburgh, PA). Twenty-one metaphases were assessed for probe localization.
Radiation Hybrid Mapping-An intron between cDNA positions 2,042 and 2,043 was amplified with exonic primers and sequenced. A set of primers was designed to amplify part of this intron. The Stanford G3 panel (22) was screened by PCR with this primer set. Data was processed at the Stanford Human Genome Center RH server.
Linkage Analysis-Two families, one Italian and one Spanish (F431 and F432) have been previously described (3,23,24), Four other families were studied: TOR1, a three-generation pedigree with 42 members of whom 26 were affected; TOR2, a two-generation pedigree with 8 members of whom 4 are affected 3 ; and Singa 1, a two-generation pedigree with 5 members of whom 4 are affected, and Bulga 1, a 3-generation pedigree with 9 members of whom 5 are affected (24).
Généthon polymorphic marker D10S603, which has the same distribution as PKDL by radiation hybrid mapping and two flanking markers (D10S198 -1.2 cM-D10S603-0.2 cM-D10S192) were selected to test for linkage to ADPKD in six families previously shown to be unlinked to the PKD1 and PKD2 loci. Genomic DNA from members of these families were used as templates for PCR. [ 32 P]dCTP-labeled PCR products were separated by polyacrylamide gel electrophoresis. Pairwise affected-only linkage analysis was performed using the FASTLINK suite of programs. A fully penetrant dominant model with a disease gene frequency of 0.0001 and equal allele frequencies was assumed. The data was calculated using two-point lod scores.

RESULTS
Through data base searches we identified two EST sequences of ϳ500 nucleotides, W27963 and W28231, with similarity to polycystin-2. The deduced amino acid sequences of W27963 and W28231 showed 78% homology and 56% identity (over residues 649 to 749) and 65% homology and 39% identity (over residues 678 to 786 with a single three-residue gap) to polycystin-2, respectively. The two EST sequences shared 94% identity over 421 base pairs. We tentatively concluded that these ESTs arose from the same gene.
Using primer sets based on these overlapping EST sequences, we amplified the same reverse transcription-PCR product from adult kidney and brain RNA whose translated amino acid sequence shows 67% homology and 46% identity to residues 670 to 779 of human polycystin-2. We further performed 5Ј-and 3Ј-rapid amplification of cDNA ends with skeletal muscle and kidney poly(A) RNA, respectively, and obtained 0.8 kb (5MR1) and 0.9 kb cDNAs (3MR20), respectively. Using 5MR1 as a probe, we screened a human retina library. Three clones, PKDL-6, PKDL-7 and PKDL-8, were obtained and sequenced. The consensus 3,044-base pair sequence revealed an open reading frame of 2,415 base pairs, which encodes a protein of 805 amino acids (Fig. 1). The putative translation start site at cDNA position 384 (5Ј-TTCCCCATGA-3Ј) is not accompanied by a typical Kozak sequence. A single inframe stop codon is found in the putative 5Ј-untranslated region. The open reading frame is followed by several in-frame stop codons, and the 3Ј-untranslated region contains a consensus polyadenylation signal (5Ј-AATAAA-3Ј) 10 nucleotides upstream from the poly(A) tail.
The deduced amino acid sequence of PKDL is shown in Fig.  1. Hydropathy analysis of the polycystin-L sequence showed five highly hydrophobic regions predicted to be transmembrane segments ( Fig. 2A). Three additional relatively hydrophobic peaks were identified. Polycystin-L showed significant homology to polycystin-2 as expected (71% homologous, 50% identical). This homology is generally higher in predicted transmembrane segments and in the loops between transmembrane segments (Fig. 2B). Polycystin-L also showed a moderate similarity (similarity 45%, identity 22%) to polycystin-1 over residues 1 to 797. This similarity is slightly higher in transmembrane segments, but there is one conserved positively charged short amino acid stretch in the first loop between transmembrane segments (Fig. 2B).
Polycystin-L, like polycystin-2, shows homology (similarity ϳ47%, identity ϳ21% overall) to each of the four domains of various Ca 2ϩ channel ␣1 subunits and other cation channels. Regions of homology are clustered in the last four transmembrane segments and the pore region of each domain of the Ca 2ϩ channel ␣1 subunits (Fig. 2B). In polycystin-L and polycystin-2, the regions corresponding to this pore region include the last of the three relatively hydrophobic peaks. The first two-thirds of this region is predicted to form a helical structure, which is characteristic for various cation channels.
Two algorithms (12,13) predict that polycystin-L has a coiled-coil domain in its C-terminal cytoplasmic tail (Fig. 1). Polycystin-L also has a putative Ca 2ϩ binding structure or EF-hand (25) that generally consists of two helices and a loop between them (Fig. 1). The C-terminal helix in the EF-hand of polycystin-L overlaps with the predicted coiled-coil region. Polycystin-L has a putative cAMP phosphorylation site in its C terminus. Putative protein kinase C phosphorylation sites are all in regions predicted to be cytoplasmic. Four of five putative casein kinase II phosphorylation sites with strong motif sequences (positions 249, 563, 674, 703, 719) are also found in the C-terminal cytoplasmic domain (Fig. 1). 2  Multiple tissue RNA dot blot analysis using the 5Ј 1.5-kb of the coding sequence as a probe revealed highest expression in adult heart and kidney (Fig. 3A). Northern blot analysis showed the presence of 5-and 1.5-kb bands in fetal tissues including kidney, liver, and brain (Fig. 3, B and C). This result suggests the presence of alternatively spliced forms. The abundance of the two splice variants is ϳ1:1 in fetal tissues. In adult tissues, however, the long transcript is only detected after prolonged autoradiography.
Chromosomal assignment of PKDL to 10q24 was achieved by fluorescent in situ hybridization on 4,6-diamidino-2-phenylindole-dihydrochloride-stained metaphase human chromosomes using a 1.7-kb genomic probe (Fig. 4). In 17 of 21 metaphase preparations analyzed, a hybridization signal was found to be present on the long arm of chromosome 10 in band q24. In six spreads, both copies of chromosome 10 were labeled, and in 11 metaphase spreads, a signal was detected on a single chromosome 10. No signals were observed on other chromosomes. With the Stanford G3 radiation hybrid panel, PKDL was found to have an identical distribution pattern as polymorphic marker D10S603 (lod score greater than 1,000). Linkage analysis of the PKDL locus using flanking markers D10S603, D10S198, and D10S192 gave negative lod scores in six ADPKD families previously documented to be unlinked to PKD1 and PKD2 loci ( Table I).
The human PKDL gene is located within a linkage group that is conserved on the distal portion of mouse chromosome 19 (26) (Fig. 5A). A 7-centimorgan deletion of this region has been described in Krd mice (27). To determine whether the mouse homologue, Pkdl, is located within the Krd deletion, we analyzed genomic DNA from F1 animals obtained from a cross of strain C57BL/6J-Krd with strain SPRET/Ei. To detect the C57BL/6J-Krd-derived allele in the F1 DNA, we utilized restriction fragment-length polymorphisms detected by hybridization with a 1.5-kb human PKDL cDNA probe (Fig. 5B). Strain C57BL/6J contains three hybridizing TaqI fragments of 5.5, 5.0, and 1.8 kb. Strain C3H, on which the Krd mutation originally arose, contains three hybridizing TaqI fragments of 5.5, 5.0, and 1.6 kb. Strain SPRET/Ei contains two hybridizing fragments of 8 and 4.5 kb. The (C57BL/6J-Krd X SPRET/Ei) F1 mouse inherited the hybridizing fragments contributed by the SPRET/Ei parent but did not inherit the fragments from C57BL/6J or C3H (Fig. 5B). This result indicates that the mouse Pkdl locus is located within the region that is deleted by the Krd mutation. DISCUSSION A Novel PKD2-like Gene-The manifestations of PKD1-and PKD2-linked ADPKD are generally similar, raising the likelihood that the gene products function in the same or parallel biological pathways. Homology between polycystin-2 and the pore-forming ␣1 subunits of voltage-activated Ca 2ϩ and Na ϩ channel proteins, combined with evidence of interaction between polycystin-1 and polycystin-2 (7,8), has led to the proposal that polycystin-2 forms homo-or heteromultimeric complexes with itself, with polycystin-1, or with another protein to function as an ion channel (9). Inasmuch as a small fraction of ADPKD families are not accounted for by PKD1 and PKD2 mutations and the function of polycystin family members may be cooperative, we postulated the existence of additional polycystin family members.
Here we report the identification and cloning of a third gene encoding a member of the polycystin superfamily, polycystin-L. Its gene, PKDL, is therefore an excellent candidate gene for human and murine cystic diseases. The temporal expression pattern of PKDL is similar to that of PKD1.
Sequence Analysis: Implications for Polycystin Function-The hydropathy patterns of polycystin-L and polycystin-2 are similar except in the region corresponding to the S4 segment of polycystin-2 where polycystin-L has a much lower hydrophobicity score, suggesting that this is a secondary membranespanning region. Polycystin-L and polycystin-2 both have putative EF-hand structures in their C-terminal cytoplasmic domains, suggesting that their functions are influenced by cytoplasmic Ca 2ϩ concentration. In several Ca 2ϩ channels, binding of Ca 2ϩ to EF-hand structures inactivates the channels (28).
Polycystin-L and polycystin-2 show moderate but significant sequence similarity to Ca 2ϩ and other cation channels, especially within their S3-S6 segments and the loop between the S5 and S6 segments. In addition, the last two membrane-spanning segments of polycystin-L, polycystin-2, and Ca 2ϩ channel ␣1 subunits share structural characteristics with the Streptomyces lividans K ϩ channel (KcsA) whose structure has been determined by crystallography (29). The common structural features include: lining residues of the last membrane-spanning segments that are mostly hydrophobic except for the negatively charged acidic amino acid near the end of these segments; loops between the last two membrane-spanning regions (pore region) that are mildly hydrophobic (Fig. 2A); first 2 ⁄3 of pore regions that are predicted to form short helical structures (pore-helix); and finally, last 1 ⁄3 of pore regions that begins with negatively charged residues which have been considered to determine the selectivity to Ca 2ϩ in known Ca 2ϩ channels (30).
Polycystin-L differs from polycystin-2 most significantly in the N-terminal cytoplasmic domain where it lacks a 100-amino acid segment. In the C-terminal cytoplasmic domain, polycystin-L is strongly predicted to have a coiled-coil structure, which has the potential to tightly interact with molecules with a similar structure like polycystin-1. Lupas' algorithm (12) also predicts a coiled-coil structure in polycystin-2, but it is not supported by Berger's algorithm (13).
Polycystin-L and polycystin-2 have three positively charged residues in S4 as opposed to five to eight in voltage-gated channels. Whereas the S4 region in voltage-gated Ca 2ϩ channels is considered to be a voltage sensor (31), it is not clear whether a membrane-spanning region with only three basic residues could act as a voltage sensor. Polycystin-L also has several putative phosphorylation sites: one cyclic nucleotide, two protein kinase C, and four casein kinase II phosphorylation FIG. 2. A, hydropathy analysis of polycystin-L (Pc-L) and polycystin-2 (Pc-2). Hydrophobic peaks that are considered to be primary membranespanning regions are described as S1, S2, S3, S5, and S6. Mild hydrophobic peaks indicating secondary transmembrane domains are labeled S1/2, S4, and p. a.a., amino acids. B, alignment of polycystin-L with polycystin-2 (gb 189 U50928), polycystin-1 (Pc-1) (gb 189 U24497), voltage-activated Ca 2ϩ channel ␣1G, ␣1C, and ␣1E (31), and transient receptor potential related channel 3, trpc3 (EMBL 189 Y13758). Roman numbers indicate domains of voltage-activated Ca 2ϩ channel ␣1 subunits. Positively charged residues in polycystin-shared motif and S4 segment are marked with a plus sign; negatively charged residues in pore-loop and S6 segment are marked with a minus sign.
sites with strong motif sequences in the C-terminal cytoplasmic domain. Two other putative protein kinase C phosphorylation sites are also found in the N-terminal cytoplasmic domain. Phosphorylation of these motif sequences may be involved in the gating process of the channel. Another scenario is that the channel is gated by a direct or indirect signal from associating proteins, e.g. polycystin-1. Given that polycystin-1 has domains that may be involved in cell-cell or cell-matrix interaction and is known to interact with polycystin-2 (7, 8), we hypothesize that the binding of ligand(s) to polycystin-1 may be associated with the gating of a polycystin-related channel.
Sequence analysis and comparison to other channels support the six or seven membrane-spanning plus one pore-region topology of polycystin-2 and polycystin-L. In addition to the five putative transmembrane segments, the middle of the three relatively hydrophobic peaks, which corresponds to S4 in ␣1 subunits of cation channels, is likely to be another transmembrane segment. Whether the N-terminal peak (S1/2) forms a membrane-spanning region is not clear.
One common feature of the polycystin-L/polycystin-2 structure that is rarely observed in known ion channels is that they both have relatively long extracellular loops between the first and the second putative transmembrane segments. Although this loop region does not show high homology to any known ion channels, polycystin-2 and polycystin-L maintain a high level of homology with each other in this region. Moreover, this region contains a 13-amino acid stretch with 3 to 4 basic residues that is conserved not only between polycystin-2 and polycystin-L but also with polycystin-1. The function of this polycystin-shared motif is not clear.
Chromosomal Assignment and Linkage Studies-Studies using D10S603, which maps to the same interval as PKDL by radiation hybrid mapping, and two adjacent markers, D10S192 and D10S198, did not reveal linkage in six non-PKD1, non-PKD2 families, making it unlikely that mutations in PKDL cause the disease in these families. Among other as yet unexplained human cystic kidney diseases, it is unlikely that PKDL plays a role in autosomal recessive polycystic kidney disease, as mutations in most autosomal recessive polycystic kidney disease families have been mapped to chromosome 6 (32). The PKDL locus can, however, be considered as a candidate for unmapped human genetic cystic disorders such as dominantly transmitted glomerulocystic kidney disease of postinfantile onset (33), isolated polycystic liver disease (34), and Hajdu-Cheney syndrome/serpentile fibula syndrome (35,36).
The region syntenic to the human PKDL locus is located on chromosome 19 in mice (26). This region is partially deleted in mice with the mutation Krd (Kidney and retinal defects) (27). The 7 centimorgans Krd deletion is located between Tdt and Cyp17 and includes the paired box gene Pax2. Mice heterozygous for a null mutation of Pax2 frequently demonstrate reduction in kidney weight, which ranges from 10 to 100% normal (37). The reduced size is due mainly to calyceal and proximal ureteral diminution as well as cortical thinning, with a reduced number of developing nephrons (37). In contrast, the phenotype of Krd/ϩ heterozygotes includes aplastic, hypoplastic, and cystic kidneys, as well as reduced viability on strain C57BL/6J (27). Our Southern analysis demonstrates that the mouse ortholog of PKDL is deleted in Krd mice. Further study is needed to clarify the contribution of Pkdl to the Krd phenotype.
Several other congenital murine and rat models with polycystic kidney disease are also known to exist, although the genetic defects in these models are as yet to be identified (38,39). Among mouse PKD models, loci for cpk, bpk, pcy, jck, jcpk, kd have been mapped to mouse chromosomes 12, 10, 9, 11, 10, and 10, respectively, and are unlikely to involve the mouse  homologue of PKDL. In Han:SPRD cy/ϩ rat, the disease gene was mapped to rat chromosome 5, whose human syntenic region resides on human chromosome 8 (18).