High Resolution Melt analysis for mutation screening in PKD1 and PKD2

Background Autosomal dominant polycystic kidney disease (ADPKD) is the most common hereditary kidney disorder. It is characterized by focal development and progressive enlargement of renal cysts leading to end-stage renal disease. PKD1 and PKD2 have been implicated in ADPKD pathogenesis but genetic features and the size of PKD1 make genetic diagnosis tedious. Methods We aim to prove that high resolution melt analysis (HRM), a recent technique in molecular biology, can facilitate molecular diagnosis of ADPKD. We screened for mutations in PKD1 and PKD2 with HRM in 37 unrelated patients with ADPKD. Results We identified 440 sequence variants in the 37 patients. One hundred and thirty eight were different. We found 28 pathogenic mutations (25 in PKD1 and 3 in PKD2 ) within 28 different patients, which is a diagnosis rate of 75% consistent with literature mean direct sequencing diagnosis rate. We describe 52 new sequence variants in PKD1 and two in PKD2. Conclusion HRM analysis is a sensitive and specific method for molecular diagnosis of ADPKD. HRM analysis is also costless and time sparing. Thus, this method is efficient and might be used for mutation pre-screening in ADPKD genes.


Background
Autosomal dominant polycystic kidney disease (ADPKD) is a hereditary kidney disorder affecting approximately one in 500 to one in 1000 human live births [1]. It is characterized by focal development and progressive enlargement of renal cysts, leading to end-stage renal disease (ESRD) [2].
ADPKD is genetically heterogeneous and involves two genes, PKD1 (MIM 601313, chromosome region 16p13.3) [3] and PKD2 (MIM 173910, 4q21-22) [4]. Mutations in PKD1 account for approximately 85% of ADPKD cases and are associated with a more severe disease than PKD2 . The median age at onset of ESRD is 54.3 [52.7-55.9] years for individuals with mutation in PKD1 compared to 74.0 [67. 2-80.8] years in PKD2 [5]. It has been hypothesized that a third gene could be implicated in ADPKD [6] but there is no evidence for this.
Genetic analysis of ADPKD is difficult owing to the existence of at least two distinct genes that can cause disease and the lack of an exhaustive list of PKD1 and PKD2 mutations that are associated with it. Genomic features of PKD1 also cause difficulty in identifying sequence variants [7]. The open reading frame of PKD1 is approximately 13 kb split in 46 exons. Exons 1-33 are duplicated six times at the HG locus on a proximal position on chromosome 16p. The homology between the 5' region of PKD1 and these six pseudo genes ranges from 94.5 to 96.7% in the duplicated area and almost 100% for the coding sequence [8]. Locus-specific amplifications of PKD1 are required for analysis which complicates genetic diagnosis [9]. Furthermore, PKD1 is a highly polymorphic gene with numerous sequence variants [10,11] (Human Gene Mutation Database [HGMD]; http://www.hgmd.cf. ac.uk/ac/index.php [12]). Single nucleotide variants in the coding regions of the PKD genes could result in the development of ADPKD. Therefore, a strategy has been proposed to investigate the pathogenic significance of sequence variants in PKD1 [11].
Definitive diagnosis of ADPKD is based on an age-specific cystic renal phenotype and a positive family history [13]. Diagnosis of ADPKD in younger patients can be difficult as renal ultrasonography can be inconclusive and if the family history is unknown. Molecular diagnosis could be of use in providing a definitive diagnosis.
High Resolution Melt (HRM) analysis, a recent advance in molecular biology used to detect variants in DNA sequences [14], has replaced dHPLC, the reference screening method, for analyzing genetic variants of BRCA1 and BRCA2 [15]. There are major drawbacks with dHPLC including chemical waste, high maintenance costs and the need for post-PCR manipulations. Moreover, dHPLC does not allow high-throughput sequence variation screening.
HRM is based on the analysis of the melting curve of a real-time PCR amplicon. Monitoring the melting curve of the amplicon involves detecting the change in fluorescence intensity induced by the release of an intercalating DNA dye from a DNA duplex as it denatures at high temperatures. Unlike dHPLC, this technique allows high throughput screening. Even if the sequence is the endpoint of molecular analysis, a pre-screening step always reduces the cost compared to the whole gene sequencing [16]. Therefore, with regard to PKD1 and PKD2 , HRM could be a rapid and cost effective technology for routine genetic analysis.
We show the results of an HRM screening strategy to identify sequence variants within PKD1 and PKD2 in 37 ADPKD patients and demonstrate that HRM is accurate to study ADPKD genes.

DNA samples
Thirty-seven unrelated patients from the Nephrology center of Marseille were screened for sequence variants in PKD1 and PKD2 . They all fulfilled the ADPKD unified diagnosis criteria [17]. The cohort is composed of 14 females and 23 males. The mean age is 51 years+/-11. They had normal renal function (2 patients) or chronic renal failure (6 patients) or end-stage renal disease (29 patients). Fifty healthy individuals are the control group. They are from the same area and ethnic group as patients. Genomic DNA was extracted from lymphocytes harvested by venous blood puncture after written informed consent. Quantification and quality of all experimental DNA samples were assessed using Nanodrop ® Technology (Coleman Technologies, Orlando, FL). DNA working solutions were diluted in ultrapure water to obtain the correct dilution. The study was conducted in compliance with the Helsinki Declaration and was approved by the Comité de Protection des Personnes (CPP) Sud-Méditerranée 2.
Long Range PCR PKD1 screening previously required amplification of five specific long range (LR) PCR with primers designed in the PKD1 sequence differing from the homologues; this step ensures to screen for sequence variants in PKD1 , not in homologues. Long range PCR assays were performed as described previously [18], LR1 amplified exon 1, LR2 amplified exon 2 to 12, LR3 amplified exon 13 to 15D, LR4 amplified exon 15E to 22, LR5 amplified exon 22 to 32. A sixth 7,838 bp LR PCR was designed to make HRM analysis for PKD1 exons 36 to 46 more specific with the forward and reverse primers: 5'-ACCTTCCCTC TAGGGAGGGAGCA-3' and 3'-GGCCAAGCTCG-CATCCAAGCA-5'. Amplification was performed using the GeneAmp High Fidelity PCR System (Applied Biosystems, Foster City, CA) in a final volume of 50 µl containing 100 ng gDNA, 2.5U enzyme, 15 µM of each primer, 5 mM dNTP, 5 µl manufacturer's supplied buffer, 2 µl supplied MgCl 2 and 10% DMSO. Cycling conditions were 95°C for 3 min, followed by 34 cycles at 96°C for 30 s, 62°C for 30 s, 68°C for 8 min, and a last step of 7 min at 72°C. All LR PCR products were verified on ethidium bromide stained gels.

Primers
We used oligonucleotide primers described in the literature to perform PCR on each PKD1 and PKD2 exons and on exon-intron junctions, which is important in order to identify variants affecting mRNA splicing. Primers from Tan et al. [18] and from Rossetti et al. [10] were used for PKD1 analysis; primers from Tan et al. [18] and from Hayashi et al. [19] were used for PKD2 . Each run included an activation step at 95°C for 10 min followed by 45 amplification cycles of 15 s denaturation, 15 s annealing and 15 s, elongation at 72°C. Denaturation and annealing temperatures differed according to each fragment these are detailed in additional file 1.
First, the products were heated to 96°C for 1 min and temperature sample was reduced at 40°C for 1 min. HRM was carried out with the temperature rising at a rate of 1°C per second with 25 acquisitions per degree. Initial and final temperatures also differed among amplicons (additional file 1). All reactions were performed in 96-well microtiter plates. Each fragment was tested in 10 patients per run.

HRM analysis
For HRM analysis, the melting curves were normalized and the temperature shifted (temp-shifted) so that samples were comparable. Modified curves were obtained with LC480 software in the gene-scanning module (version 1.3; Roche). The normalized and temp-shifted melting curves corresponded to the final curve after normalization. When an amplicon harbored a sequence variation, the normalized and temp-shifted melting curve had a different shape from wild-type amplicons.
In this study, normalization was manually adjusted; in a routine framework, normalization ranges are automatically filled out. Default software sensitivity setting was 30% in order to avoid false negative amplicons. Therefore, sensitivity was set by default to 40% for all amplicons.
Comparing several wild-type amplicons demonstrated that the derived normalized and temp-shifted difference plot varies within a range owing to small random differences. The variance is partly caused by the specificity of the primers. Every amplicon that had a different derived normalized temp-shifted curve from wild-type was sequenced.

Reverse transcription analysis
To confirm the signification of the suspected splice mutation c.7210-5C>G, we performed RT-PCR from RNA extracted of leucocytes as previously described [20], briefly lymphocytes were isolated using a Ficoll gradient and total RNA was extracted using the Rneasy Qiagen kit, according to the manufacturer's instructions (Qiagen SA). Total RNA was resuspended in Rnase free water. A total of 5 μg of total RNA was reverse transcribed using a RT-PCR kit (RT Life Technology GIBCO BRL), according to the manufacturer's instructions, and 3 μl of the RT products were amplified using a standard hot start PCR technique using the primers, for PKD1 specific PCR (6F 5'-AGCGCAAC-TACTTGGAGGCCC-3' and 6R 5'-ACCACAACGGAG TTGGCGG-3') and we diluted this RT-PCR product to 1/1000 and performed a nested PCR with the following primers (SAB26F2 5'-CAGGCCAATGTGACGGTGG-3' and SAB26B3 5'-CAGTCGCATGCCTGCACTGC-3').

Sequencing
Sequencing was used to confirm and investigate variants. All sequencing analysis was subcontracted to Cogenics ® (Cogenics, Meylan, France) and performed by Sanger sequencing method. Capillary sequencing is performed using Big Dye* Terminator chemistry and sequences are delineated with Applied Biosystems 3730 × l platforms (http://www.beckmangenomics.com/documents/services/ GS_Sanger_Sequencing.pdf). Samples were prepared by diluting 20 µl PCR product in 30 µl of ultrapure water. Primers were diluted to obtain a concentration of 2 mM as recommended by the subcontractor.

Sequence variation analysis and classification
All sequences were compared to the following references sequences NCBI RefSeq PKD1 : NM_000296.2; and PKD2 : NM_000297.2. The standard nomenclature recommended by HGVS (http://www.hgvs.org/mutnomen) was used to number nucleotides and name mutations.
For previously described sequence variants, we reported the clinical significance assessed in The Polycystic Kidney Disease Mutation Database accessible at http://pkdb.mayo.edu [12]. Pathogenic potential of novel mutations was assessed as follows: -nonsense or frameshift variants leading to a STOP codon as well as intronic mutations altering mRNA sequence proved with RT-PCR were considered as definitely pathogenic mutations.intronic and synonymous variants that did not alter predicted splicing (http://www.fruitfly.org/seq_tools/splice. html) were considered as polymorphisms, the others as indeterminate mutations. -substitutions were evaluated using the scoring method described by Rossetti et al [11] and with PolyPhen, a software for interspecies sequence variations examination (http://genetics.bwh.harvard.edu/ pph) [21]. When the two scores assessed a pathogenic variant the mutation was considered as probably pathogenic, when they assessed a non pathogenic variant it was considered as polymorphism. If the two tests were discordant, the variation was reported as indeterminate.

Results
Forty-three exons of PKD1 were analyzed using HRM in 56 fragments after PCR amplification. Of the 2,072 amplicons tested in the 37 patients, 302 (14.6%) were suspected to have a sequence variants compared to the wild-type amplicons analyzed (Figure 1). Each of the 302 amplicons was sequenced and 244 had at least one sequence variants. A total of 410 sequence variants were detected in PKD1 , corresponding to 11.1 ± 7.7 (mean ± SD) per patient; 3 exons were studied by direct sequencing. 129 different sequence variants were found.
Fourteen exons of PKD2 were analyzed in 14 fragments after PCR amplification. Of the 518 amplicons tested, HRM analysis identified 8 with sequence variants and 4 were shown to have at least one sequence variant after sequencing. A total of 30 sequence variants were detected in PKD2, one exon was studied by direct sequencing. 9 different sequence variants were found. Table 1 and Figure 2 show definitely pathogenic and probably pathogenic mutations. 25 mutations in PKD1 and 3 in PKD2 were identified in 28 patients. Of these 28 mutations, 9 were already described in literature. Of the 19 new mutations, 13 were identified as being definitely pathogenic because they lead to a STOP codon; 3 substitutions were scored as probably pathogenic (Additional file 2) as well as 1 in frame deletions, one intronic deletion and one intronic substitution.
The first in frame deletion, p.Thr2337_Phe2338del, was classified as probably pathogenic because the two residues are highly conserved during evolution, they are located in the functional REJ domain and this deletion was the only probably pathogenic mutation identified in this patient (Figure 3). The second deletion was p. Leu2433del. Since leucine 2433 is a conserved residue (in chicken, frog and takifugu but not in mouse or rat), this variant was also considered probably pathogenic. It was the only possible mutation causing disease identified in this patient. In addition, this mutation was previously described associated with lack of expression of PC-1 in primary cilia [22]. The two in frame deletions were not found in 100 control chromosomes tested.
The 19 base pair intronic deletion in intron 31 (c.10167 +24del19) was considered as probably pathogenic. Unfortunately, this patient was lost of follow up and no RT-PCR could be performed. A very similar deletion has previously been reported [23] as probably pathogenic in The Polycystic Kidney Disease Mutation Database. The deletion is present in three of the HG genes-PKD1 partial homologues on chromosome 16 (Figure 4) leading to a splice mutation associated with intron 31 retention. It was the only possible pathogenic mutation found in this patient and was not present in the 100 control chromosomes tested. The abnormal predicted splicing of intron 31 leads to a frameshift within exon 32. The presence of this deletion in three of the six PKD1 homologues is of interest as it is associated with retention of the intron. A possible hypothesis for the pathogenesis of this mutation is the occurrence of recombination between homologue and PKD1. If a crossover event has occurred, it must be a small crossover, because the deletion region is flanked by two bases differing from the homologues and identical to PKD1 (Figure 4). The mutation c.7210-5C>G was predicted to alter splicing in silico and no other indeterminate or pathogenic variation was found in this patient. We classified this mutation as pathogenic after we performed a specific RT-PCR as previously described [20]. The mutation is responsive of the retention of intron 18 leading to a frameshift with a stop codon in position 2514.
Polymorphisms are presented in Table 2. We identified 92 polymorphisms in PKD1 and 6 in PKD2 . Two intronic variants in PKD1 were predicted to alter splicing but were considered as polymorphisms, c.2985 +4G>A was present in 4 patients, a definite pathogenic mutation was identified in three and c.2985+5G>A was present in three individuals who already had a pathogenic mutation. The p.Glu107Asp novel substitution in PKD2 was classified polymorphism because it was found in 15 patients of the cohort, which is not consistent with a pathogenic mutation.
We found 12 indeterminate sequence variants in PKD1 and none in PKD2 . They are presented in Table 3. The majority of these variants were substitutions; their scoring is presented in Additional file 2. The small in frame insertion p.Asp3781_Val3782insGlu was considered as indeterminate because the patient had another indeterminate variant (p.Arg4276Trp) and we could not be sure of its pathogenicity. The three other indeterminate sequence variants (p.Ala2704Val, p.Ser2935Phe p. His3559Pro) are the sole mutation with potential pathogenicity observed in the three individuals screened.

Discussion
We describe the first sequence variants pre-screening method for PKD1 and PKD2 using HRM analysis. This is the first use of this technique in genetic nephrology. This method has a good diagnosis rate as evidenced by the identification of a mutation in 75% of our cohort (including definitely pathogenic and probably pathogenic mutations), which is the mean diagnosis rate reported with direct sequencing of the two genes [24]. In comparison, Rossetti et al found a diagnosis rate of 62.9% with direct sequencing [11] and 67% with dHPLC [10]. In addition a recent report of Hoefele et al found a mutation detection efficiency of 64,5% by direct sequencing [25]. More than 81% of amplicons suspected to have a sequence variant after HRM analysis carried one after sequencing, confirming the specificity of this technology. Overall, less than 15% of the fragments of PKD1 and 1.5% of the PKD2 fragments analyzed by HRM were sequenced. Moreover, we screened in a blind procedure four patients with previously known mutations [20] and identified all of them. Thus, HRM is efficient, sensitive and specific for mutation screening in ADPKD genes. HRM analysis reduced drastically the need for systematic sequencing and the time of sequence analysis.
We describe 52 new sequence variants in PKD1 : 18 were classified as mutation, no mutation was detected in more than one individual, 23 as polymorphism and 11 as indeterminate variant. We report one new mutation and one polymorphism in PKD2 .
Large deletions or duplications represent 4% of ADPKD pathogenic variants [26]. HRM technology allows genomic qPCR screening to be performed for large deletions/duplications in the same run. Successful simultaneous analysis by HRM and genomic qPCR has been described for MLH1 gene [27]. Therefore, it is likely that HRM and genomic qPCR can be performed simultaneously for PKD1 and PKD2 , which will require the introduction of control DNA and a reference gene in the microtiter plate. This will improve molecular diagnosis by directly identifying large deletions/duplications that have not been identified by HRM analysis or with sequencing. Therefore, HRM analysis could prove to be a highly integrated molecular diagnosis tool.
HRM analysis is a high throughput technique for sequence variants screening PKD1 and PKD2 . The format of PCR analysis in 96 or 394 well plates allows the process to be automated after DNA extraction and the first round of LR PCR for PKD1 . A major advantage of HRM is a drastic diminution in the needs for sequencing to identify a sequence variation. HRM could be more cost effective than direct sequencing for identification of molecular anomalies in ADPKD. Comparing to dHPLC, HRM is a closed tube and a non destructive technique. The PCR and analytical steps are in the same run with HRM. During dHPLC several separate steps require PCR products manipulation and can induce mistakes. After analysis, in HRM the PCR product can be directly sequenced; in dHPLC the PCR product is destroyed during the run and a new PCR must be done for sequencing. In our facility, the price for one amplicon analysis with HRM is 1.52 $ (1.10 €) compared to 4.64 $ (3.36 €) with dHPLC [28]. In addition, HRM allows large fragment rearrangements detection in the same step. This technique is very competitive compared to dHPLC. HRM is time sparing, provides greater security (closed tube during the whole process) and is cheaper. Assuming bidirectional sequencing costs 13.8 $ (10 €) per fragment, the cost for mutation screening in PKD1 and PKD2 with direct sequencing costs 966 $ (700 €) (70 fragments). With HRM, 70 fragments × 1.52 $ plus sequencing of positive fragments (i.e. 9 fragments × 13.8 $), the total cost is of 230.46 $ (167 €). Furthermore, we do not consider here the technician time sparing provided by HRM in this cost analysis. Thus, acquiring the needed hardware for HRM analysis is easily paid off. HRM pre-screening is cost and time efficient facing direct sequencing in ADPKD genes.
The technique efficiency will soon be further improved with the search of conditions allowing the screening of exons 1 of PKD1 and PKD2 and exons 42 and 43 of PKD1 . We were not able to conduct HRM analysis of these exons because the denaturation curves could not be analyzed. We suspected this failure is related to the high GC contents of the DNA sequence of exon 1 of PKD1 and PKD2 . For exon 42 and 43 of PKD1 , we suspected the formation of secondary structures not solved by denaturation temperature. HRM is a rather young technology and we can expect important developments increasing the efficiency of this sensitive and specific method of screening [28].
Molecular diagnosis of ADPKD is important for improving the care of patients [29]. Renal transplantation from a related living donor at risk of ADPKD is a concern since clinical and ultrasonographic diagnosis criteria fail to identify donors with ADPKD when they are aged less than 40 years [30]. Molecular diagnosis could be of use in this area as well as identifying mutations in ambiguous phenotypes [31], for example when the family history is unknown. Genetic diagnosis also results in a more accurate prognosis [32] as the genes mutated in ADPKD are the prognostic factors linked to kidney survival [33]. In addition, therapies development will probably require an accurate diagnosis in young

Conclusions
HRM analysis is an efficient, sensitive, specific and cost effective strategy to identify mutations in PKD1 and PKD2 .