Article Text

Original article
Large-scale targeted sequencing comparison highlights extreme genetic heterogeneity in nephronophthisis-related ciliopathies
  1. Markus Schueler1,
  2. Jan Halbritter1,2,
  3. Ian G Phelps3,
  4. Daniela A Braun1,
  5. Edgar A Otto4,
  6. Jonathan D Porath1,
  7. Heon Yung Gee1,
  8. Jay Shendure5,
  9. Brian J O'Roak6,
  10. Jennifer A Lawson1,
  11. Marwa M Nabhan7,8,
  12. Neveen A Soliman7,8,
  13. Dan Doherty3,
  14. Friedhelm Hildebrandt1,9
    1. 1Divison of Nephology, Department of Medicine, Boston Children's Hospital, Harvard Medical School, Boston, Massachusetts, USA
    2. 2Divison of Nephrology, Department of Internal Medicine, University Clinic Leipzig, Leipzig, Germany
    3. 3Department of Pediatrics, University of Washington School of Medicine, Seattle, Washington, USA
    4. 4Department of Pediatrics, University of Michigan, Ann Arbor, Michigan, USA
    5. 5University of Washington, Genome Sciences, Seattle, Washington, USA
    6. 6Oregon Health and Science University, Molecular and Medical Genetics, Portland, Oregon, USA
    7. 7Department of Pediatrics, Kasr Al Ainy School of Medicine, Cairo University, Cairo, Egypt
    8. 8Egyptian Group for Orphan Renal Diseases (EGORD), Cairo, Egypt
    9. 9Howard Hughes Medical Institute, Chevy Chase, Maryland, USA
    1. Correspondence to Professor Friedhelm Hildebrandt, Division of Nephrology, Boston Children's Hospital, 300 Longwood Avenue, Boston, MA 02115, USA; friedhelm.hildebrandt{at}childrens.harvard.edu Dan Doherty MD PhD Associate Professor of Pediatrics, Divisions of Developmental Medicine and Genetic Medicine, University of Washington, Seattle Children's Hospital, Seattle, WA 98195, USA; ddoher@uw.ed

    Abstract

    Background: The term nephronophthisis-related ciliopathies (NPHP-RC) describes a group of rare autosomal-recessive cystic kidney diseases, characterised by broad genetic and clinical heterogeneity. NPHP-RC is frequently associated with extrarenal manifestations and accounts for the majority of genetically caused chronic kidney disease (CKD) during childhood and adolescence. Generation of a molecular diagnosis has been impaired by this broad genetic heterogeneity. However, recently developed high-throughput exon sequencing techniques represent powerful and efficient tools to screen large cohorts for dozens of causative genes.

    Methods: Therefore, we performed massively multiplexed targeted sequencing using the modified molecular inversion probe strategy (MIPs) in an international cohort of 384 patients diagnosed with NPHP-RC.

    Results: As a result, we established the molecular diagnoses in 81/384 unrelated individuals (21.1%). We detected 127 likely disease-causing mutations in 18 of 34 evaluated NPHP-RC genes, 22 of which were novel. We further compared a subgroup of current findings to the results of a previous study in which we used an array-based microfluidic PCR technology in the same cohort. While 78 likely disease-causing mutations were previously detected by the array-based microfluidic PCR, the MIPs approach identified 94 likely pathogenic mutations. Compared with the previous approach, MIPs redetected 66 out of 78 variants and 28 previously unidentified variants, for a total of 94 variants.

    Conclusions: In summary, we demonstrate that the modified MIPs technology is a useful approach to screen large cohorts for a multitude of established NPHP genes in order to identify the underlying molecular cause. Combined application of two independent library preparation and sequencing techniques, however, may still be indicated for Mendelian diseases with extensive genetic heterogeneity in order to further increase diagnostic sensitivity.

    • Genetics
    • Diagnostics
    • Molecular genetics
    • Renal Medicine

    Statistics from Altmetric.com

    Introduction

    The term nephronophthisis-related ciliopathies (NPHP-RCs) summarise a group of rare autosomal-recessive cystic kidney diseases including nephronophthisis (NPHP), Senior-Løken syndrome (SLS), Joubert syndrome (JBTS) and Meckel-Gruber syndrome (MKS).1 ,2 NPHP-RCs are genetically heterogeneous disorders, characterised by mutations in genes which encode proteins that localise to primary cilia, basal bodies or centrosomes. The disruption of the ciliary function results in a broad phenotypical spectrum which is collectively termed ‘ciliopathies’.3 NPHP-RCs account for the majority of genetic end-stage renal diseases during the first three decades of life. The most prominent renal features are increased echogenicity and corticomedullary cysts on ultrasound. Renal histology reveals tubular atrophy, basement membrane disintegration, interstitial fibrosis and cyst formation. About 15% of the affected individuals with NPHP-RC show extrarenal organ involvement, most notably, progressive retinal dystrophy defined as SLS. Cerebellar malformations such as mid-hindbrain malformation and cerebellar vermis hypoplasia/aplasia are commonly observed in individuals with JBTS and are characterised by the ‘molar tooth sign’ in brain imaging studies.4 These malformations are associated with numerous neurological features including developmental delay, intellectual disability, muscle hypotonia, ataxia, oculomotor apraxia, nystagmus and abnormal respiratory control.4 MKS represents the most severe manifestation of the NPHP-RC spectrum as a perinatally lethal ciliopathy.5 To date, more than 90 genes have been implicated in the pathogenesis of ciliopathies, accounting for approximately 50% of affected individuals.6 JBTS and MKS are caused by mutations in a subset of these genes including: MKS1 (MIM# 609883), B9D1 (MIM# 614144), B9D2 (MIM# 611951), AHI1 (MIM# 608894), INPP5E (MIM# 613037), ARL13B (MIM# 608922), TMEM216 (MIM# 613277), CC2D2A (MIM# 612013), KIF7 (MIM# 611254), TCTN1 (MIM# 609863), TCTN2 (MIM# 613846), TCTN3 (MIM#614815), CEP41 (MIM# 610523), IFT172 (MIM#615630), CSPP1 (MIM#615636), RPGRIP1L (MIM#611560), OFD1 (MIM#300804), TMEM138 (MIM#614465), TMEM231 (MIM#614970), PDE6D (MIM#615665), and TMEM237 (MIM# 614423).7–26 Due to the large number of established NPHP–RC genes, molecular genetic diagnostics remains a major challenge and requires further technical advances. Recently, high-throughput and next-generation sequencing (NGS) platforms have become widely available and allow massively parallel low-cost sequencing.27 ,28 Previous studies from our group have established a microfluidic multiplex PCR technology with consecutive NGS. Screening for multiple established NPHP genes yielded the molecular cause in about 13% of cases in large cohorts of patients with NPHP-RC.29 ,30 The recently developed modified molecular inversion probe strategy (MIPs) represents another technique that allows rapid and cost-effective sequencing of candidate genes in large cohorts.31 In the present study, we applied the MIPs technology to 384 samples from affected unrelated individuals with NPHP-RC. We captured and amplified 760 coding exons of 34 established NPHP-RC genes, including NPHP1-NPHP12, PKHD1, the gene causing autosomal-recessive cystic kidney disease (ARPKD), and 21 established JBTS/MKS genes. We established the molecular diagnoses in 81/384 unrelated individuals (21.1%). Altogether, we detected 127 mutations. In addition, we compared the sensitivity, specificity and costs of the two above-mentioned targeted high-throughput technologies in our cohort.

    Materials and methods

    Human subjects

    We obtained blood samples, pedigrees and clinical information after receiving informed consent (http://www.renalgenes.org). A total cohort of 384 affected individuals diagnosed with NPHP-RC based on published clinical criteria were sequenced using the MIPs technology.32 This cohort contains 84 unrelated individuals from families with multiple affected individuals, and 300 single affected individuals. Consanguinity was known to be present in 67 (17.3%) families. Prior to study enrolment, individuals with homozygous deletions of NPHP1 were excluded by using a multiplex PCR-based deletion analysis as previously described.33

    Primer design and evaluation for the MIP system

    MIPs was designed to target 760 coding exons of 34 established NPHP-RC genes (see online supplementary table S1) using hg19 human genome reference and the dbSNP132 database to identity polymorphisms that might interfere with hybridisation. Each MIPs targeted a specific 112 bp genomic region using flanking extension (16–20 bp) and ligation (20–24 bp) arms. Following synthesis (Integrated DNA technologies, Coralville, Iowa, USA), MIPs was standardised to 100 µM, and equimolar amounts of each MIPs were pooled.31

    Target DNA enrichment by the MIP system

    Target-specific MIPs hybridised to 100 ng of genomic DNA input followed by polymerase-mediated gap filling and ligation. The captured DNA was PCR amplified using universal primers containing Illumina flowcell adaptor sequences and a unique 8-base bar code (Illumina, San Diego, California, USA). Next, four sets of 96 bar-coded sample libraries were pooled, and each was purified using 1.8X AMPure XP beads (Beckman Coulter, Brea, California, USA). The appropriate product size was confirmed by agarose gel electrophoresis.31

    MIP target rebalancing and NGS

    A pilot MIP capture was performed on a subset of samples and sequenced using the Illumina MiSeq. MIP targets were rebalanced by spiking the poorest performing (bottom 15% by read depth) MIP at 10×. The rebalanced MIP pool was used to create the 384 individual sample libraries as described above. Pooled libraries were sequenced on three lanes of an Illumina HiSeq 2000 instrument using 101 bp paired-end reads.31

    Bioinformatics pipeline

    Raw sequence reads were processed and aligned to the hg19 human reference sequence (UCSC Genome Browser) using the Burrows-Wheeler Aligner (BWA, V.0.5.9). BAM files were generated with Picard tools (see web resources), followed by single-nucleotide variant and indel calling using the Genome Analysis Toolkit's Unified-Genotyper (GATK V.2.2) with the depth of coverage parameter set to 5000. Called variants failing to meet the following quality metrics were removed from further analysis: quality >30, read depth >8, quality by depth >5. A window of 10 was set for clustered variants. Variants were annotated with SeattleSeq (see web resources) and further filtered against controls from the National Heart, Lung, and Blood Institute Exome Sequencing Project Exome Variant Server (EVS) ESP6500 data set with the use of an observed minor allele frequency (MAF) threshold <1%.31

    Variants were ranked by the criteria of whether mutations were likely to truncate the conceptual reading frame (nonsense, frameshift and obligatory splice mutations), whereas missense variants were ranked by the evolutionary conservation and by predicting the impact on the encoded protein using web-based programs (PolyPhen2, Phylo-P score, CADD score, Mutation Taster, SIFT).

    NPHP1 copy number variation detection by the MIP system

    The read depth of each MIP was normalised to the overall read depth for each sample, and then a standardised Z-score was calculated. Underperforming MIP with a median coverage <20× were removed. Singular value decomposition (SVD) was used to reduce systematic noise. Using the SVD-adjusted Z-scores, we identified samples where the majority of NPHP1-targeted MIPs had Z-scores ≥2.0 or ≤−2.0.31 Individual A2527, who was examined in an independent array previously, served as a positive control for successful copy number variation (CNV) calling.

    Sanger sequencing confirmation and segregation analysis

    Variants/mutations detected by NGS and predicted to be detrimental were subsequently confirmed by Sanger sequencing using original DNA samples from the respective affected individuals as PCR template. Whenever DNA was available from the index person, the affected siblings and parents, we performed segregation analysis. PCR was performed using a touchdown protocol described previously.34 Sequencing was performed using Mastercycler pro (Eppendorf, Hamburg, Germany). Sequence traces were analysed using CLC Genomics Workbench software (CLC-bio, Aarhus, Denmark).

    Results

    NGS and mapping statistics

    Sequencing was performed on an Illumina HiSeq 2000 platform after capture and amplification of 34 NPHP-RC genes in 384 affected individuals (384 families) with NPHP-RC (figure 1, see online supplementary table S2). The total output was 361 million reads of 101 bases yielding an average of 940 thousand reads per DNA sample. Alignment resulted in mean exon coverage of 505, with >97% of targeted exons having ≥25-fold average coverage (see online supplementary figure S1). Due to low DNA quality, sequencing failed for two samples, while seven samples had 10% and one sample 13% coverage at 25×.

    Figure 1

    Distribution of molecular nephronophthisis (NPHP) diagnoses in 18 established NPHP-related ciliopathies (RC) genes, detected in an cohort of 384 individuals with NPHP-RC. (A) Number of families detected with molecular diagnoses in 18 established NPHP-RC genes, which are sorted from left to right by frequency. All individuals were previously screened for the presence of a homozygous NPHP1 deletion prior to being entered into the present study. (B) Percentage of individuals with a molecular diagnosis versus individuals without a molecular diagnosis in our cohort of 384 unrelated individuals diagnosed with NPHP-RC (left); distribution of molecular diagnoses across 18 NPHP-RC genes (right).

    Variant filtering, validation and parameter setting

    Altogether, we obtained 27 133 single nucleotide variants and 646 insertion/deletion (indel) variants derived from the 384 individuals. In total, 20 371 (exonic and intronic) variants showed a MAF of >1% across the data set, which were considered to be frequent SNPs according to the EVS database and, therefore excluded.

    As previously reported,31 we removed variants from further analysis whenever the following parameters were present: coverage <8×, quality score <30 and a quality by depth <5. Variants below this threshold were considered as likely ‘false positives’. Furthermore, we included only missense variants that showed an evolutionary conservation down to Ciona intestinalis, a PolyPhen2-score >0.9 and a MAF <1% across the data set. Alternatively, variants were included when they had previously been reported as disease causing according to the HGMD-Professional mutation database ‘Biobase’ and were conserved down to Danio rerio.

    After filtering and ranking, 189 variants, including 110 potentially protein-truncating mutations (such as nonsense, frameshift and obligatory splice site mutations), and 79 missense variants/mutations, were selected for standard Sanger sequencing. We were able to confirm 93 of the 110 potential truncating mutations (85%) and 70 of the 79 selected missense variants (89%). A total of 163 out of 189 variants (specificity: 86%) were confirmed by Sanger sequencing. However, 35 of the 163 confirmed potential mutations were excluded as common SNPs according to the 1000 genomes, dbSNP138, EVS and ExAC Browser database or have been reported as homozygous SNPs. The remaining mutations were compared with the HGMD-Professional mutation database ‘Biobase’ (HGMD 2015.2 (copyright Cardiff University 2015)) and defined as novel if not present.

    Sensitivity for detecting known variants and comparison of two approaches

    In order to calculate the sensitivity of the MIPs technology in this study, we compared the detected disease-causing mutations within the coding regions of 12 established NPHP genes (NPHP1-NPHP12) with our previous mutation analysis.29 ,30 By using an array-based microfluidic PCR technology with NGS, we previously identified and Sanger confirmed 78 variants, which explained the underlying molecular cause in 49/384 families of the same cohort.29 ,30 Overall, 66 out of these 78 mutations were redetected in the present study (‘mutation’ sensitivity: 85%). In particular, eight mutations were not redetected due to low coverage, giving a sensitivity of 94% (66/70) at covered base pairs. A four base pair duplication in IQCB1 was missed in three individuals because one of the flanker arms of this MIP overlapped with the duplication. For three individuals the reason why the MIPs technology missed the previous detected mutation remains unclear (figure 2, see online supplementary table S3).

    Figure 2

    Comparison of two targeted high-throughput library preparation approaches; modified molecular inversion probe strategy (MIPs) and array-based microfluidic multiplex PCR. (A) Schematic design of general work flow of the two high-throughput mutation analysis approaches: MIPs technique (on the left) and array-based microfluidic multiplex PCR technology (Fluidigm) (on the right). For both techniques library preparation was followed by consecutive next-generation sequencing (NGS). (B) Overlapping findings in NPHP1-NPHP12 by application of either mutation analysis approach: MIPs technique (blue circle) and Fluidigm technology (yellow circle). Sixty-six (61%) variants were detected by both methods (overlapping circles). By applying MIPs 28 (26%) additional variants were detected. By using the array-based microfluidic multiplex PCR (Fluidigm) approach, an additional 12 (11%) variants were identified (see also online supplementary tables S2 and S3). dNTP, deoxy-nucleotide triphosphate.

    The estimated costs for library preparation and sequencing of 384 individuals using the modified MIPs technology are US$21 00031 whereas the calculated costs for sequencing the same cohort using an array-based microfluidic PCR technology amount is US$18 50029 in total. These costs comprise reagents and primers but do not include costs for acquisition of additional machinery, which is the case in the microfluidic PCR approach (see online supplementary table S4 and figure S2). Concerning the hands-on time for library preparation, sequencing and bioinformatics pipeline analysis, there are no major differences between both methods: approx. 142 h for the MIPs technology and 148 h if using an array-based microfluidic PCR technology.

    Identification of genetic causes in a cohort of 384 individuals

    The combination of capturing all coding exons of 34 established NPHP-RC genes by using the MIPs technology in an international cohort of 384 unrelated affected individuals (384 families) diagnosed with NPHP-RC revealed the molecular diagnosis in 81 individuals (81 families). Furthermore, 26 single heterozygous mutations were detected in 26 individuals, 15 of which were missense variants and 11 of which were single heterozygous nonsense variants. Additional Sanger sequencing of all remaining coding exons of the genes, wherein a single heterozygous variant was detected, revealed two additional heterozygous missense mutations, which segregated within the affected status. CNV analysis for NPHP1 revealed nine heterozygous deletions and six heterozygous duplications including the positive control (individual A2527); none was previously detected by using a multiplex PCR-based deletion analysis.

    In summary, by massively parallel-targeted sequencing using MIPs strategy we identified the molecular diagnosis in 81 of 384 (21.1%) unrelated patients/families. Recessive mutations were identified in the following genes (number of unrelated patients/families in parentheses): CEP290 (NPHP6) (15), PKHD1 (10), IQCB1 (NPHP5) (10), NPHP3 (9), NPHP1 (8), TMEM67 (7), NPHP4 (6), INVS (NPHP2) (3), CC2D2A (2), CEP164 (NPHP15) (2), TTC21B (NPHP12) (2), SDCCAG8 (NPHP10) (1), IFT172 (NPHP17) (1), OFD1 (1), TMEM216 (1), TMEM237 (1), TCTN1 (1) and AHI (1) (see online supplementary table S2). No causative mutations were identified in the genes GLIS2 (NPHP7), RPGRIP1L (NPHP8), NEK8 (NPHP9), ARL13B, B9D1, B9D2, C5ORF42, CEP41, INPP5E, KIF7, MKS1, TCTN2, TCTN3, TMEM231, TMEM138 and CSPP1. Overall, we identified 30 independent families with homozygous mutations, 5 families with hemizygous mutations and 46 families with compound heterozygous mutations. In 51 individuals truncating mutations (nonsense, frameshift or obligatory splice-site mutations) were found on both alleles, whereas 14 affected individuals showed one truncating mutation in combination with a non-synonymous missense mutation. The remaining 16 individuals exhibited missense mutations on both alleles. After additional evaluation and Sanger sequencing of all coding regions and intron/exon boundaries in the respective genes, 11 families remained with only one heterozygous truncating mutation. We discovered a total of 22 novel pathogenic mutations in the following genes (number of families in parentheses): PKHD1 (6), NPHP1 (4), CC2D2A (2), CEP290 (NPHP6) (2), NPHP4 (2), NPHP3 (2), INVS (NPHP2) (1), TTC21B (NPHP12) (1), OFD1 (1) and TMEM237 (1) (table 1).

    Table 1

    Number of novel mutations detected in this study compared with previously reported mutations (HGMD-Professional ‘Biobase’, 2015) in the genes NPHP1, INVS, NPHP3, NPHP4, CEP290, TTC21B, PKHD1, CC2D2A, OFD1 and TMEM237

    Discussion

    Massively multiplex targeted sequencing of 34 established NPHP-RC genes using the MIPs strategy in an international cohort of 384 affected individuals with NPHP-RC identified the molecular cause in 81 families. Altogether our study identified 127 pathogenic mutations. CEP290 (NPHP6) was found most commonly mutated (15 families). Interestingly, mutations in PKHD1 and IQCB1 (NPHP5) were the second most frequent molecular cause in our cohort (10 affected families each). This shows that the clinical presentation of ARPKD may mimic the NPHP-RC phenotype and is not necessarily distinguishable without the use of molecular genetics in clinical practice.

    Combined with our previous mutation analysis using an array-based microfluidic multiplex PCR technology, we obtain a representative distribution of genes in a cohort of individuals with NPHP-RC prescreened for the most common mutation, the NPHP1 deletion.30 After exclusion of the presence of a homozygous NPHP1 deletion, a disease- causing mutation was identified in 21% of the cases, leaving almost 80% of cases genetically unsolved when analysing 34 established NPHP, ARPKD and JBTS genes (see online supplementary table S2) (figure 1).

    Twenty-six individuals carry single heterozygous variants, which do not explain the phenotype based on an autosomal recessive hypothesis; however, it is possible that second variants were not detected in the few areas of low coverage or in non-coding regions. These unsolved cases may be explained by missed mutations in the targeted genes, mutations in other established NPHP-RC genes that were not part of this study, or by mutations in genes which phenocopy NPHP-RC, such as for example, CLCNKB, COL4A5 and AGXT.35 Second, these cases may be explained due to mutations in genes, which have not yet been implicated as a molecular cause of NPHP-RC. This fact further implies the presence of genetic heterogeneity in individuals with NPHP-RC. Lastly, a significant fraction may be explained by unidentified mutations (false negatives), as indicated by a mutation sensitivity of 85% when compared with our previous analysis of the same cohort (figure 2, see online supplementary table S3). As a conclusion, combination of two library preparation methods may be indicated in order to warrant sufficient sensitivity, at least in a diagnostic setting.

    In summary, we demonstrate that the MIPs is a useful library preparation method for targeted sequencing approaches and mutation analysis in large cohorts of individuals with NPHP-RC. Whereas Sanger sequencing has been the gold standard over the past 30 years, it has been proven to be tedious and expensive, especially for large cohorts and genetically heterogeneous conditions like NPHP-RC. Recent developments in high-throughput library preparation and massively parallel sequencing offer various novel and efficient methods that decrease sequencing time and costs. Due to the increasing number of established genes, the relevance and utility of high-throughput sequencing techniques is indispensable for molecular genetic diagnostics.

    As about 50% of affected individuals with NPHP-RC remain genetically unsolved, non-targeted sequencing approaches like whole exome/genome sequencing (WES/WGS) are needed to identify new candidate disease genes. For mutation analysis of panels of known disease genes in large cohorts, however, WES is still not cost-efficient. In our case, the price tag for sequencing a cohort of 384 affected individuals using WES would amount to about US$268 800 assuming US$700 per individual, whereas the costs for MIPs or array-based microfluidic PCR technology added up to about US$20 000 (approx. US$50–100 per individual for >200 samples) (see online supplementary table S4 and figure S2). For the proposed targeted sequencing approaches, though, a minimal cohort size of at least 100 probands is recommended in order to run them cost-efficiently (see online supplementary figure S2). While WES/WGS are essential tools to identify novel candidate disease genes in single cases, targeted high-throughput screening technologies, such as MIPs, can also be helpful to analyse large cohorts with the same phenotype to corroborate these candidates through detection of additional unrelated individuals with mutations in the same gene.

    Acknowledgments

    The authors thank the affected individuals and their families for participation in this study and the study coordinators Leslie Spaneas and Brittany Fisher.

    References

    Supplementary materials

    • Supplementary Data

      This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.

    Footnotes

    • MS, JH and IGP contributed equally.

    • Collaborators Contributing members of the GPN study group are: F Yalcinkaya (Ankara, Turkey); S Bakkaloglu (Ankara, Turkey); F Ozaltin (Ankara, Turkey); E Comak (Antalya, Turkey); F Krull (Aurich, Germany); Schmitz-Hübner (Bad Oeynhausen, Germany); H Rupprecht (Bayreuth, Germany); D Muller (Berlin, Germany); P Dahlem (Coburg, Germany); B Hoppe (Cologne, Germany); M Wolfe (Cologne, Germany); M Weber (Cologne, Germany); U Vester (Essen, Germany); K Bonzel (Essen, Germany); J Nikolay (Furth, Germany); I Hansmann (Halle, Germany); M Wiefel (Hamburg, Germany); U Orth (Hamburg, Germany); H Pfleiderer (Hamm, Germany); L Pape (Hannover , Germany); Morlot (Hannover, Germany); J Ehrich (Hannover, Germany); B Tonshoff (Heidelberg, Germany); F Schindera (Karlsruhe, Germany); J Hoefele (Martinsried, Germany); M Griebel (Munich, Germany); E Broeking (Münster, Germany); M Konrad (Münster, Germany); M Radke (Potsdam, Germany); M Brandis (Ravensburg, Germany); A Kirchhoff (Wurzburg, Germany); V Feygina (Brooklyn, New York, USA); J Springate (Buffalo, New York, USA); S Ahmadzdeh (Burlington, Vermont, USA); D Gipson (Chapel Hill, North Carolina, USA); A Becker (Dallas, Texas, USA); V Dharnidharka (Gainesville, Florida, USA); P Mark (Grand Rapids, Michigan, USA); P Srivaths (Houston, Texas, USA); A Wilson (Indianapolis, Indiana, USA); E Kamil (Los Angeles, California, USA); S Why (Milwaukee, Wisconsin, USA); C Pan (Milwaukee, Wisconsin, USA); C Kashtan (Minneapolis, MN, USA); C D’Alessandri (New Haven, Connecticut, USA); H Trachtman (Ney York city, New York, USA); B Kaplan (Philadelphia, Pennsylvania, USA); M Joseph (Phoenix, Arizona, USA); R Weiss (Valhalla, New York, USA); S Thomas (Ann Arbor, Michigan, USA); L Newberry (Aurora, Colorado, USA); M Koyun (Cairo, Egypt); H Fathy (Alexandria, Egypt); A Rybi—Szuminska (Bialystok, Poland); M Szczepanska (Zabrze, Poland); Z Dolezel (Brno, Czech Republic); M Malina (Prague, Czech Republic); T Seeman (Prague, Czech Republic); T Honzik (Prague, Czech Republic); P Ferreira (Calgary, Canada); M Ferguson (Halifax, Canada); E Harvey (Toronto, Canada); K Chong (Toronto, Canada); R Sandford (Cambridge, UK); D Josifova (London, UK); D Bockenhauer (London, UK); J Sayer (Newcastle upon Tyne, UK); C Johnson (Yorkshire, UK); P Senguttuvan (Chennai, India); I Pela (Firenze, Italy); N Knops (Leuven, Belgium); T Levart (Ljubljana, Slovenia); T Neuhaus (Luzern, Switzerland); C Ayuso (Madrid, Spain); A Kindi (Muscat, Sultanate of Oman); N Knoers (Nijmegen, The Netherlands); C Antignac (Paris, France); W Radauer (Salzburg, Austria); C Genzani (Sao Paulo, Brazil); U Berg (Stockhom, Sweden); C Klingenberg (Tromsø, Norway); C Jones (Victoria, Australia); R Savarirayan (Victoria, Australia); J Kausman (Victoria, Australia).

    • Contributors MS, JH, IGP, DAB, EAO, JDP, HYG and JAL: conceived and performed NGS, analysed and interpreted sequencing data and phenotype-genotype correlations; JS, BJO: provided MIPs technology and gave technical advice; NAS, MMN: had major clinical input, patient recruitment; MS, JH and DD: wrote the manuscript; FH and DD: coordinated and supervised the project.

    • Funding This work was supported by a grant from the National Institutes of Health to FH (RC4-DK076683) and the NephCure Foundation. This work was also supported by the National Institutes of Health through the University of Washington Intellectual and Developmental Disabilities Research Center Genetics Core (P30-HD002274) and R01-NS064077 to DD. We also acknowledge private donations from the families of children with Joubert syndrome.

    • Competing interests None declared.

    • Ethics approval University of Michigan and the Boston Children's Hospital Institutional Review Boards.

    • Provenance and peer review Not commissioned; externally peer reviewed.

    Request Permissions

    If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.