Comprehensive Molecular Diagnostics in Autosomal Dominant Polycystic Kidney Disease

Mutation Screening

DNA for mutation screening was available from 202 families who had at least one individual enrolled in CRISP. Thirty-two families were multiplexed within CRISP, giving a total of 239 patients. The initial PKD1 and PKD2 screening of the 202 probands by denaturing HPLC (DHPLC) identified 81 definite mutations (nonsense, frame-shifting deletions or insertions, typical splicing, or in-frame changes of five or more amino acids [aa]); defined as mutation group (MG) = A in Table IA). A second round of screening by direct sequencing (DS) of both genes in all patients in whom no definite mutation was detected, plus 29 control subjects (a total of 150), revealed an additional 37 definite mutations. In addition, analysis of a subset by field inversion gel electrophoresis (FIGE) and visualization of PKD1 long-range PCR (LR-PCR) products revealed four large deletion mutations, giving 122 probands with MG = A (Table IA).

In four atypical splicing cases, RNA was available to test the splicing predictions, and in a fifth, IVS43+14del20, skipping of exon 43 had been previously demonstrated.²⁸ In two patients, substitutions close to the end of IVS15 generated novel AG dinucleotides predicted to form novel splice acceptors and cause frame-shifting mutations. Reverse transcription–PCR (RT-PCR) and sequencing confirmed this abnormal splicing (Figure 1A). A similar mutation, IVS37-10C→A, was predicted to abolish the normal acceptor site by weakening the poly-pyrimidine tract (Figure 1B) and RT-PCR showed skipping of exon 38, plus a product that included the final 180 bp of IVS37. In the final case, PKD2 IVS4+3delAAGT, the splice donor site was predicted to be destroyed (Figure 1C). No abnormality was seen with RT-PCR across the region, but amplification with a primer in IVS4 showed that all or part of IVS4 is incorporated into the sequence. These mutations were therefore classed as MG = A, giving a total of 127 families with definite mutations.

• Download figure
• Open in new tab
• Download powerpoint

Figure 1.

Examples of atypical splicing events in PKD1 (A and B) and PKD2 (C). (A, top) Diagram of wild-type and mutant cDNA sequence illustrating the effects of two IVS15 substitutions. Novel sequence included in the transcript is boxed in red. (Bottom) Genomic sequence from these cases. Exonic sequence is in boldface, intron/exon boundaries are indicated with vertical lines, and polypyrimidine tracts are underlined. Substituted nucleotides are in red, and novel acceptor sites and polypyrimidine tracts are shown with red vertical and horizontal lines, respectively. Novel exonic sequence is in blue boldface. (B) Wild-type and mutated sequence of IVS37-10C→A. This substitution weakens the normal polypyrimidine tract (7/10; dashed black line) and eliminates the normal acceptor site. Reverse transcription–PCR (bottom right: M, mutant; N, normal) shows that exon 38 is skipped (the skipped product is also found at a low level normally) and a product including part of IVS37 is also seen. A diagram shows the wild-type and mutant transcripts (bottom left); the genomic sequence of the novel donor site in IVS37 is in red. (C) Diagram of the wild-type and mutant sequence of the PKD2 mutant IVS4+3del4 (top) and mutant cDNA sequence (bottom). This deletion destroys the normal donor site (underlined), and cDNA sequence (bottom) shows that a product including part of IVS4 is present in the mutant.

Scoring of UCV

All UCV that were not previously described as polymorphisms were analyzed and evaluated as possible disease-associated mutations. Those variants that ultimately were scored as probably pathogenic are summarized in Table II. The detailed scoring of UCV is shown in Table II (probably pathogenic), Table III (indeterminate), and Table IV (likely neutral polymorphisms). The algorithm that was used to evaluate the UCV is described in detail in the Materials and Methods section. To illustrate how this algorithm works, two examples are provided here. The PKD1 change G381S is moderately conservative but at a highly conserved site, giving a GD/GV score of +5. The change is not in a conserved domain, and no abnormal splicing was predicted (+0). This change was found in three different families, two of which had no other likely mutation and one had the indeterminate variant R3063C (+5). Segregation was demonstrated in one family with nine affected individuals (+4); and another mutation, G381C, has been described at this residue (+1) giving a variant score (VS) = +15. This is a high score (MG = B) reflecting the recurrent nature of the change, demonstrated segregation, and highly conserved residue that is mutated. In contrast, the same substitution at another site, G3651S, is less clearly pathogenic. The GD/GV score is the same (+5), but it is not in a conserved domain, it is not predicted to change splicing, and segregation has not been demonstrated (+0). No other variant was found in this individual (+2), giving a VS = +7 (MG = C). This lower level of certainty reflects that the substitution is relatively conservative and that the mutation is novel.

As a result of this analysis, 44 missense, two atypical splicing, and seven small in-frame deletions were defined as probably pathogenic, giving a total 180 (89.1%) of 202 with a probable mutation characterized. In 37 probands, a highly likely mutation (MG = B) was defined, with a likely mutation (MG = C) in another 16 (Table II). The pathologic significance of 22 additional variants was not clear, and they were defined as indeterminate (I; Table III). The remainder of novel variants were defined as likely neutral polymorphisms and are shown in Table IV.

The most important factor in determining whether a missense change was likely pathogenic was the degree to which it was conserved in orthologs and in other proteins with the same domain, with the GD almost always substantially greater than the GV (Table II). This contrasted with the 48 newly described neutral variants (Table IV), where the residue was rarely conserved and the GV was often higher than the GD. Recurrence of a variant in two or more patients with no other clear mutation also strongly supported a pathogenic likelihood. A factor that was important in excluding UCV as likely pathogenic was the finding of a more likely mutation in the same patient (Table IV). Four PKD1 UCV that previously were described as mutations were defined as indeterminate, R1340W and R4276W, or neutral polymorphisms, R324L and R2200C (Tables III and IV). The most interesting polymorphism was a 3-aa deletion (PKD1: 2894delANS) that did not segregate with the disease in a family with a segregating PKD2 missense change, R325Q. Analysis of orthologs in this area (Figure 2A) shows that this region is present only in the human sequence, not in the HG pseudogenes or other mammals. The presence of these three residues is, therefore, the result of a recent duplication in the human sequence. Segregation analysis also helped to identify likely mutations; for instance, in case 380166, two possible changes were found (F3168L and R2477C), but only the former segregated with the disease.

• Download figure
• Open in new tab
• Download powerpoint

Figure 2.

Multisequence alignment illustrating how comparative analysis can inform about the likely pathogenicity of variants. (A) Polycystin (PC1) sequence from CRISP case 187456 (pedigree 120010), mammalian orthologs, as indicated, and an example of the HG sequence, showing the position of the 2894delANS variant. This 3–amino acid region is present just in the human sequence. (B) Alignment of human polycystin-1 PC1 and PC2 homologs and sea urchin REJ sequences, as indicated, in the region of the polycystin-A motif (PC-A).²⁹ Four missense mutations, two PC1 (red; R3753W and R3753Q) and two PC2 (blue; R322W and R325Q), affect highly conserved basic residues in this motif. (C) C-type lectin domains of human PC1, low-affinity Ig epsilon Fc receptor (FCER2), macrophage mannose receptor 1 (MRC1), asialoglycoprotein receptor 2 (ASGR2), and C-lectin consensus (Pfam). Mutations are shown in red (A432V, C436Y, and C508R), and indeterminate changes are shown in green (V466L and V466M). (D) Alignment of the PKD repeats of human PC1. The position of pathogenic changes are shown in red (W967R, Y1412D, G1503R, N1870H, G1999S, and G1999V), indeterminate changes are shown in green (N970K, H1093D, R1340W, S1352N, R1411C, and R1698W), and neutral changes are shown in blue (S903G, N10345, H1093Y, L1106V, V1339M, R1351W, A1422T, A1516T, H1777P, A1790V, E1811D, and A1871T). The position of the residue changed is shown colored.

Characteristics of Mutations

Mutation analysis of the CRISP population defined 85.0% as PKD1 and 15.0% as PKD2. These values correspond closely to previous studies,⁵,⁶ although these are the first data that are based on mutation detection. The breakdown of mutation type in the two genes is summarized in Table 1. Overall, >70% of mutations were predicted to truncate the protein, >80% in PKD2. Although the majority of mutations were novel, 27 had previously been described and seven others were seen more than once in the study; 30% of families had a recurrent mutation (Table 1). The most common mutation in the study was 5014delAG (five [2.5%] families), followed by Q2556X (four [2.0%] families); two truncating and two missense changes were found three times, and six mutations were found twice (see Table I for details).

Important Functional Residues in PC1 and PC2

The large number of missense changes identified several residues or regions that were disrupted more than once, suggesting particular functional significances. In the PC-A domain (in the first extracellular loop of PC2 and third extracellular loop of PC1²⁹), two basic residues were substituted a total of four times (two PC1 and two PC2; Figure 2B) and four additional times in the literature (see Table I). Several mutations were also seen in the C-type lectin with three highly conserved residues substituted with a nonconservative or, in one case, a relatively conservative change (Figure 2C). Two other interesting substitutions, L4137P and L4139P, are predicted to disrupt the helical structure of the G-protein binding region.³⁰ A clear indication of how conservation within a domain helped determine whether a substitution is likely pathogenic is shown in the PKD repeats (Figure 2D). Likely pathogenic changes were usually nonconservative in nature and at well-conserved residues, whereas the likely neutral or indeterminate changes were most often found at nonconserved residues and were often conservative changes (in some cases matching the residue found in other PKD repeats).

Mutation Location

Mutations were spread throughout both the PKD1 and PKD2 genes with at least three mutations found in each 5% interval across PKD1 (except one with a single event) and at least one in each 12.5% interval across PKD2 (Figure 3). Of the 46 PKD1 exons (including immediate flanking regions), 40 harbored mutations, as did nine of the 15 in PKD2. Analysis of the distributions show that they were NS differently from uniform in PKD1 (P > 0.05) or PKD2 (P > 0.10). The median mutation position in PKD1 was 6873nt (median gene position 6454.5nt) with 55.6% of mutation in the 3′ half of the gene. Missense changes were relatively evenly distributed in PKD1 (P > 0.10), but truncating changes were nonuniform (P < 0.01; Figure 3A). In particular, there was a cluster of truncations at the end of exon 15 to exon 19, corresponding to the junction of the PKD repeats and the REJ domain. This clustering was not due to recurrent mutations (Table I). IVS21 has previously been suggested as a possible hotspot for mutations as a result of a long polypyrimidine tract,³¹ and although this could be a factor, the peak did not correspond precisely to that region. Too few missense changes were seen in PKD2 to determine the significance of the distribution (Figure 3B).

• Download figure
• Open in new tab
• Download powerpoint

Figure 3.

Positions of mutations in PKD1 (A) and PKD2 (B). Graph of number of mutations, truncating (light blue) or in-frame (dark blue), in intervals of the genes (PKD1, 10 divisions; PKD2, eight divisions). Corresponding exonic regions of the transcript are shown below each graph (PKD1, green; PKD2, red) with exon number indicated. Protein structures with domain regions indicated are shown above the graphs for PC1 (A) and PC2 (B). Only truncating mutations in PKD1 were distributed significantly differently from uniform (expected per interval 10.7 mutations).

Mutation-Negative Cases

Twenty-two families did not have a probable mutation defined (Table V). In 10 of these, no variants other than known polymorphisms were detected, whereas in five others, only a single unique synonymous change was detected, which was not predicted to change splicing significantly. Two cases had a substitution of the same residue, V466, suggesting significance; however, both changes (to leucine or methionine) were highly conservative, and this residue is not well conserved (leucine in frog) and variable in the C-type lectin (Figure 2C). The unique change S1352N was similarly not scored as a mutation, because the substitution matched that found in fish, whereas another, L2696P, was rejected because of the highly nonconservative nature of this site in orthologs (see Table IIIA). Two cases, 103227 (V609G and N970K) and 118641 (V690G and R1340W), had two indeterminate changes, and a third had a single such change (476972; H1093D). V690G is at a residue previously described as the site of a pathogenic change (V690D), whereas H1093D is a nonconservative change at the site of a likely polymorphism (H1093Y). Although these both are suggestive of pathogenic changes, they did not meet the threshold that we set in this article (Table IIIA). Some cases may have larger deletions because not all were screened by FIGE, and, indeed, suggestive changes (dosage or possible aberrant fragments) or regions of apparent homozygosity (hemizygosity?) were seen in four cases, but the DNA changes have not been characterized (Table V).

Polymorphic Variants

PKD1 is a highly polymorphic gene with 194 different neutral or indeterminate variants detected within the coding and immediate IVS flanking regions. The corresponding figure for PKD2 was 8 (see Table IV for novel nonsynonymous changes). The average number of variant alleles per patient in PKD1 was 10.1, with a range from 0 to 55, whereas the corresponding figures for PKD2 were 0.8 (range 0 to 3). Only one polymorphic variant in PKD2 had a minor allele frequency >5% (R28P, 30.9% of alleles). In PKD1, the majority of variants were rare; 94 (51.1%) were found just once within the study, with 24 (12.6%) having a minor allele frequency >5%. The greatest variability in PKD1 was in the 26 black patients, in whom an average of 20.6 variant alleles was found. The high level of polymorphisms at PKD1 compared with PKD2 is likely explained by the larger coding region (four to five times greater) and GC richness of the region, resulting in a much higher level of hypermutable CpG dinucleotides.¹⁸,³² Additional special mechanisms are probably not required to explain the higher polymorphism rate in PKD1.³¹,³³

Details of the CRISP Study Cohort

The CRISP study cohort consists of 239 patients who have ADPKD, are aged 15 to 46 yr, and have a GFR of >70 ml/mm at enrollment.¹⁵,¹⁶ For each proband, a blood sample was obtained for DNA extraction, and Epstein Barr virus (EBV)-transformed lymphoblast cell lines were established. Samples are stored at the National Institute of Diabetes and Digestive and Kidney Disease Center for Genetic Studies.

Mutation Analysis of PKD1 and PKD2 by DHPLC

DHPLC was performed as described previously.¹⁷ The duplicated region of PKD1 was amplified as five PKD1-specific fragments by LR-PCR, and exons were amplified from these fragments after dilution 1:1000 to avoid genomic DNA carryover. LR-PCR products were checked on 0.8% agarose gels before the nested PCR to detect large deletions/insertions. DHPLC was performed using a Wave System HT (Transgenomic, Omaha, NE), as described previously.¹⁷ Each amplicon was run at two temperatures, typically the melting temperature and +2°C, as determined by Wavemaker version 4.0.32 (Transgenomic). Chromatograms were subjectively grouped, depending on the differences in the profile from normals and known polymorphic variants.⁴⁵ Novel profiles were sequenced, plus a representative profile that resembled a known variant; the remainder were identified by signature-based genotyping.³⁴

DS of PKD1 and PKD2 and Confirmation of Mutations

DS was performed on a fee-for-service basis using a commercial diagnostic test (Athena Diagnostics, Worcester, MA), and a subset were analyzed at Emory for PKD2. This sequence data are now stored at the CRISP web site.

All likely pathogenic changes were double-checked in a different aliquot of the same sample, and, when samples were available, the segregation of variants was tested by DS. A subset of samples were also analyzed for large genomic rearrangements using FIGE, as described previously.¹ The precise mutation in these cases was determined by amplifying across the deletion and sequencing the aberrant product. The designation of all variants in both genes was numbered from their translational start. The PKD1 sequence also included the alternatively spliced codon at the start of exon 32 so that the total length of the protein was 4303 aa.

RT-PCR to Analyze Atypical Splicing

RNA was isolated from lymphoblast cell lines using TRizol (Invitrogen, Carlsbad, CA), and cDNA was generated with the Superscript III cDNA synthesis kit (Invitrogen). For the IVS15 variants, LR-RT-PCR was performed with the SPEC-4 primers,¹⁸ followed by nested PCR with primers in exons 15 and 18. The IVS37 mutation was analyzed with primers in exons 36 and 39, and the PKD2 IVS4 mutation was analyzed with primers in exons 3, 7, and IVS4 (primer sequences available on request). All PCR products were sequenced by standard methods.

Classification of UCV

The GMS²⁰ was used to determine the GD of each substitution. An MSA was generated using the software ClustalW (MacVector Inc, Cary, NC) of PC1 orthologs from human, rat, mouse, chicken, frog (Xenopus tropicalis), and fish (Fugu rubripes). Other fish species, Tetraodon nigroviridis and Danio rerio, were also used in some cases. For PC2, human, dog, mouse, chicken, Danio, and Fugu were used, plus in some cases Drosophila melanogaster, sea urchin, and Caenorhabditis elegans. Sequences were obtained from NCBI or Ensembl. Predicted sequences for chicken, frog, and fish species were edited as necessary. The GV was determined as the largest GMS between vertebrate orthologs. Conservation within defined motifs used published MSA: WSC⁴⁶; PKD repeat²,³,⁴⁷; REJ domain⁴⁸; GPS³⁸,⁴⁶; PLAT domain⁴⁹; and polycystin motifs, PC-A and B²⁹; plus MSA of PC1 and PC2 homologs (PC1L1, PC1L2, PC1L3, PKDREJ, PC2L1, and PC2L2) and sea urchin REJ1 to 4 homologs. Changes within transmembrane regions and the G-protein binding region were also assessed for disruption of secondary structure. Predicted splicing changes were scored using the Splicing Predictor by Neural Network (http://www.fruitfly.org/seq_tools/splice.html).⁵⁰ Details of previous descriptions of the UCV were obtained from the HGMD (http://www.hgmd-cf.ac.uk) and/or the ADPKD Mutation Database (http:/pkdb.mayo.edu).

For all of the UVC, a VS was calculated. A matrix comparing the GD and GV was used with the GMS divided into six groups (0, 1 to 35, 36 to 80, 81 to 130, 131 to 179, and >180) and scores from +8 (e.g., GD >131, GV <35) to −4 (e.g., GD <35, GV >81) given. An additional −4 was allotted when the substitution matched the residue in an ortholog. Conservation of the reside in a domain scored +2 to +4 (highly conserved to invariant) with −2 when substitution matched the residue in an ortholog. Splicing changes were scored from +20 when an abnormal splice product was demonstrated by RT-PCR to −5 when an intronic change was not predicted to alter splicing. Additional points were given when a variant segregated in siblings or a family (+2 to +4); failure to segregate = −20. Other descriptions of the change as a mutation added +2 per time described. If another likely mutation (groups A through C) or an indeterminate (I) change was found with the variant, −5 (A) to −1 (I) was subtracted; otherwise +2 was added. The sum of these scores gave the VS. VS ≥11 were classed very likely mutations (MG = B); 5 to 10 as likely mutations (MG = C); −4 to 4 as variants of indeterminate (I) pathogenicity; and ≤5 as neutral polymorphisms (P). The χ² test was used to examine whether the mutation location in each gene was uniform.