Abstract
Mutation of the Cys1 gene underlies the renal cystic disease in the Cys1cpk/cpk (cpk) mouse that phenocopies human autosomal recessive polycystic kidney disease (ARPKD). Cystin, the protein product of Cys1, is expressed in the primary apical cilia of renal ductal epithelial cells. In previous studies, we showed that cystin regulates Myc expression via interaction with the tumor suppressor, necdin. Here, we demonstrate rescue of the cpk renal phenotype by kidney-specific expression of a cystin-GFP fusion protein encoded by a transgene integrated into the Rosa26 locus. In addition, we show that expression of the cystin-GFP fusion protein in collecting duct cells down-regulates expression of Myc in cpk kidneys. Finally, we report the first human patient with an ARPKD phenotype due to homozygosity for a predicted deleterious splicing defect in CYS1. These findings suggest that mutations in the Cys1 mouse and CYS1 human orthologues cause an ARPKD phenotype that is driven by overexpression of the Myc proto-oncogene.
Translational Statement The cystin-deficient cpk mouse is a model for the study of autosomal recessive polycystic kidney disease (ARPKD). We show that the cpk mouse phenotype is associated with altered Myc expression. To date, the clinical relevance of cystin deficiency to human disease was unclear, due to the absence of ARPKD cases associated with CYS1 mutations. We report the first case of ARPKD linked to a CYS1 mutation disrupting normal splicing. These findings confirm the relevance of cystin deficiency to human ARPKD, implicate Myc in disease initiation or progression, and validate the cpk mouse as a translationally relevant disease model.
Introduction
Autosomal recessive polycystic kidney disease (ARPKD; MIM 263200) affects 1:26,500 live births.1 Cohort studies indicate that 75-80% of patients with typical ARPKD have mutations in the Polycystic Kidney and Hepatic Disease 1 (PKHD1) gene.2–5 Mutations in the DZIP1L gene account for less than 0.1% of affected patients,6 while mutations in other hepato-renal fibrocystic disease genes, eg. HNF1B, PKD1, NPHP2, NPHP3, and NPHP13, can phenocopy ARPKD.7 For reasons yet to be explained, mice with targeted disruption of Pkhd1 exhibit little or no kidney disease.8–15 In the absence of a Pkhd1 mutant mouse model that accurately recapitulates the human disease phenotype, the cpk mouse carrying a spontaneous truncating mutation in Cys1 has been the most widely studied mouse model of ARPKD.16,17 Cystin, the Cys1 gene product, is a 145-amino acid cilia-associated protein that is expressed in mouse embryonic kidney and liver ductal epithelium.18 Disruption of cystin function results in elevated Myc expression in collecting duct epithelial cells19–22 and increased cell proliferation.19,23 In previous work, we have demonstrated that in renal collecting duct epithelia, cystin physically interacts with necdin in a regulatory complex that modulates Myc expression.24
Cystin deficiency-associated disruption of ciliary signaling and/or overexpression of Myc is associated with aberrant SMAD3 phosphorylation,25 overexpression of Fos and Kras proto-oncogenes,19–21 elevated levels of growth factors,26 aberrant localization and abundance of the epidermal growth factor receptor (EGFR) on the apical surface of collecting duct cells27, abnormal levels of basement membrane components,28–30 and epithelial cell adhesion molecules.31,32 Until now, the relevance of these effects of cystin deficiency for human disease was unclear in the absence of ARPKD patients with mutations in human CYS. Here we present the first case of human ARPKD due to homozygosity for a CYS1 mutation, in this case predicted to cause defective splicing. We also show that complementation of defective Cys1 in the kidneys of Cys1cpk/cpk (cpk) mice rescues both Myc overexpression and the collecting duct cyst phenotype. These studies suggest that up-regulation of Myc expression in vivo may play a central role in the pathogenesis of mouse recessive polycystic kidney disease, with important implications for human ARPKD.
Results
Phenotypic rescue of cpk mice by kidney-specific expression of a cystin-GFP fusion protein
We generated a conditional expression Cys1 transgenic Cys1cpk/cpk (cpk) mouse line carrying a Cys1-GFP transgene knock-in at the Rosa26 locus. In these mice, Cys1-GFP transgene expression is precluded by the presence of a loxP-flanked termination sequence consisting of a PGK-Neo cassette (Figure1A, TOFF allele). The Cys1-GFP transgene is expressed by the ROSA26 promoter only after Cre-mediated deletion of the loxP-flanked PGK-Neo cassette (Figure1A, TON allele). We crossed Rosa26-Cys1-GFP mice with Cys1cpk/+ mice to generate Cys1cpk/+;Rosa26-Cys1-GFP mice, which were then crossed with Ksp-Cre transgenic mice33 to generate Cys1cpk/+;Rosa26-Cys1-GFP;Ksp-Cre progeny. In these mice, Cre expression, controlled by the Ksp-cadherin regulatory elements, occurs exclusively in the developing distal renal tubular epithelium and the genitourinary tract34 resulting in high level expression of the cystin-GFP fusion protein in the collecting ducts and loops of Henle and low or no expression in the proximal tubules. Finally, the rescue experiments were carried out by crossing Cys1cpk/+;Rosa26-Cys1-GFP;Ksp-Cre mice with Cys1cpk/+ mice. Genotype-confirmed Cys1cpk/cpk;Rosa26-Cys1-GFP;Ksp-Cre experimental “rescue” (R) mice were compared to their Cys1+/+;Rosa26-Cys1-GFP;Ksp-Cre control (C) littermates (Figure 1B). While cpk mice are characteristically smaller than wild-type littermates and die by 21 days of age,35 no differences in body size or survival were observed between the rescue (R) mice and their littermate controls (Figure 1C, left panel). Kidney sizes at postnatal days 14 and 21 were not significantly different in R and wild-type (WT) mice (Figure 1C, right panel), while age-matched Cys1cpk/cpk (cpk) mice exhibited the characteristic cystic kidney phenotype. These results indicate that the gross renal phenotype of the cpk mouse was rescued by kidney-specific expression of cystin-GFP.
Expression of cystin-GFP fusion protein in the kidneys of rescued cpk mice
We examined the expression of the cystin-GFP fusion protein in the kidneys of R mice. Endogenous cystin was detectable in the kidneys of both WT and C mice and absent from the kidneys of R mice (Figure 1D). The cystin-GFP fusion protein of ~50kDa was detected in R and C mice (Figure 1D, lanes 3 and 4). These results demonstrate that cystin-GFP expression was associated with Cre-mediated excision of the PGK-Neo cassette.
Dual immunofluorescence staining with antibodies against GFP and aquaporin-2 (AQP2) was used to examine cystin-GFP expression in nephron segments of kidneys from R and C mice (Figure 2). AQP2 is expressed primarily on apical cell membranes of collecting duct cells.36,37 The cystin-GFP fusion protein was detected in AQP2-positive collecting ducts of R mice (Figure 2C and I) and C mice (Figure 2B and H), while cystin-GFP was absent in the Rosa26-Cys1-GFP mice that do not carry Ksp-Cre transgene (Figure 2A and G). Co-localization of AQP2 and cystin-GFP demonstrated cystin-GFP fusion protein expression in the collecting duct cells.
Histological evaluation of cystogenesis in rescued cpk mice
The gross evaluation of kidneys from R mice suggested that the renal histology would be normal. Histological evaluation showed that, while the majority of the nephrons in R kidneys appeared to have normal dimensions, occasional cystic structures were present (Figure 3C and F). Using DBA and LTA lectins, markers of distal and proximal tubules, respectively,38,39 we observed that the cystic structures stained with LTA (Figure 3I). These findings suggest that while expression of the cystin-GFP in collecting ducts markedly attenuated the cpk renal phenotype, sporadic cyst formation did occur in proximal tubular segments of these kidneys (Figure 3I).
Expression of Myc in rescued cpk mice
Myc overexpression in cpk kidneys is well-documented.19–22 In previous work we demonstrated that cystin physically interacts with the DNA-binding protein necdin in a regulatory complex that binds to the Myc P1 promoter.24 Necdin enhances Myc promoter activity and cystin antagonizes this effect. In a previous report, we proposed that Myc up-regulation in cpk kidneys results directly from disruption of the cystin-necdin interaction. In the current study, we examined the relative abundance of C-MYC protein in the kidneys of 14-day old R mice as compared to cpk mice. Quantitative immunoblotting revealed comparable levels of C-MYC in WT and R kidneys that were markedly lower than in kidneys of cpk mice (Figure 4). These results demonstrate that transgene rescue of the cystic kidney phenotype in cpk mice is associated with down-regulation of C-MYC protein expression, suggesting a central role for Myc overexpression in the renal cystogenesis in this mouse model.
A case of human ARPKD associated with homozygous CYS mutation
While the cpk mouse phenotype recapitulates important clinical features of ARPKD, to date no human cases of the disease have been linked to mutation of the CYS1 gene. Recently, a 5-year-old male patient (subject B783), the offspring of consanguineous parents, was evaluated for monogenic cystic renal disease. The patient presented with polyuria and polydipsia and ultrasound examination revealed multiple medullary and cortical cysts consistent with polycystic kidney disease. Renal function was mildly reduced (creatinine 0.6 mg/dL) for a child of the subject’s age. DNA samples obtained from from the proband and both parents were analyzed by trio whole exome sequencing (TRIO-WES). Homozygosity mapping of exome variant data for B783 indicated 108 Mbp of homozygosity by descent (Figure 5A), suggesting that the parents are approximately fourth degree relatives. Based on this mapping, it was hypothesized that a biallelic gene mutation residing within a homozygous peak region caused the subject’s renal disease. To identify the most probable disease-causing mutation, we used the following criteria for exome variant filtering40–43: (1) exclusion of all variants that did not change the amino-acid sequence or affected canonical splice sites (defined as ± 6 nucleotides surrounding the exon-intron boundary), (2) exclusion of variants reported in the homozygous state or with a minor allele frequency greater than 0.1% in a control cohort (ExAC and gnomAD genome databases), (3) inclusion of homozygous bi-allelic variants with appropriate parental segregation consistent with our above hypothesis, and (4) assessment of variants for deleteriousness based on in silico prediction of their impact on protein structure and/or splice site function. This approach, however, failed to yield a strong candidate mutation. Therefore, trio whole genome sequencing (TRIO-WGS) was performed in order to extend coverage to non-coding genome regions.
Employing the same filtering criteria described for TRIO-WES led to the identification of a mutation lying in a region of homozygosity of descent (Figure 5A) representing a splice-site mutation (c.318+5G>A) in the exon 1 donor site of CYS1 (Figure 5B,C). The following criteria strongly suggest a deleterious mutation: (i) the splice variant is extremely rare, being absent from the gnomAD database (gnomAD, version 2.1.1), and (ii) four independent in silico prediction tools indicate a deleterious effect of the mutation on splicing (Table 1). Importantly, neither TRIO-WES nor TRIO-WGS analysis revealed causative mutations in >100 cystic kidney disease genes, including PKHD1 and DZIP1L. Allele-specific PCR analysis confirmed homozygosity and heterozygosity of the mutation in the proband and both parents, respectively (Supplementary Figure S1).
In light of the above genetic findings, a further evaluation of the clinical history of subject B783 revealed mild congenital liver fibrosis, which is also consistent with the established phenotype of the cpk mouse.44–46 Our findings strongly support the first identification of a causative mutation in CYS1 in a human patient with ARPKD.
Discussion
In the current study, we demonstrate that expression of the Cys1 transgene in renal collecting ducts of cpk mice rescues the cystic phenotype and down-regulates expression of Myc in vivo. We also report the first ARPKD patient with a homozygous CYS1 mutation, a c.818+5G>A variant predicted to disrupt splicing. This variant affects the +5-position of the canonical donor splice site, AG/GURAGU. Pathogenic variants affecting +5 position of the donor splice site have been reported in other genes47–49 and detailed analysis of G-to-A sequence changes at the +5 position have revealed disrupted base pairing between donor splice site of a pre-mRNA and the U1snRNP of the spliceosome leading to decreased efficiency of the splice site recognition and exon skipping.50
The identification of a pathogenic CYS1 variant in a patient with an ARPKD-like phenotype confirms the importance of the CYS1 gene product for normal function of human collecting duct cells. While it is surprising that CYS1 deficiency causing ARPKD has been observed in only one family, it is important to note that this gene is GC-rich, particularly its first exon (Figure S3). Such GC-rich regions can be difficult to amplify and sequence using Sanger methodology51 and can be missed in next-generation sequencing because low sequence complexity prevents efficient capture prior to library construction.52
Development of collecting duct cysts in humans and mice with cystin deficiency suggests a shared pathobiology and possibly similar molecular mechanisms underlying cyst formation. In the cpk mouse model, renal cysts initially develop at embryonic day 15.5 (E15.5) and are restricted to proximal tubules.53,54 As cpk mice develop and disease progresses, cysts predominantly affect the distal collecting duct region.53 Specific expression of cystin-GFP in developing ureteric bud-derived collecting ducts rescued the renal cystic phenotype. Transgenic cpk mice expressing the fusion protein only in AQP2-positive collecting duct cells exhibited survival rates and kidney sizes similar to wild type mice. Interestingly, while collecting duct cysts were absent in rescued cpk mice, these animals did express proximal tubular cysts, suggesting that the initial phase of proximal tubular cystogenesis was not rescued. Proximal tubule cysts have also been observed in human ARPKD fetal specimens between 14 and 26 weeks of gestation, but not in the kidneys of fetuses older than 34 weeks of gestation.39 These observations suggest a gradual shift of cyst formation sites from proximal tubules to collecting ducts during the early fetal development in both human and mouse ARPKD.
Cystin is a cilium-associated protein that localizes to the basal bodies and the ciliary axoneme.17,18,55 Treatment of cpk mice with paclitaxel, which promotes microtubule assembly, prevents renal cyst formation, suggesting that cystin may stabilize microtubule assembly within the ciliary axoneme.56 The primary cilia of the collecting duct epithelium function as transmitters of mechano- and chemosensory stimuli to signaling pathways that regulate multiple key cellular processes including differentiation, proliferation, apoptosis, tissue homeostasis and cell polarity.57 We have previously demonstrated that cystin, with two functional nuclear localization signals, can be released from the ciliary membrane through a myristoyl-electrostatic switch and translocate to the nucleus where it forms a regulatory complex with necdin to modulate Myc expression.24
The Myc proto-oncogene plays a critical role in normal kidney development58 and several lines of evidence suggest a central role for dysregulated Myc expression in the pathophysiology of polycystic kidney disease. First, overexpression of Myc in the kidneys of SBM transgenic mice causes polycystic kidney disease.59 Furthermore, renal cystic disease remitted in SBM mice that underwent spontaneous reversion to normal kidney Myc expression.60 Second, treatment of cpk mice with antisense Myc oligonucleotides mitigated the cystic phenotype.22 Third, Myc is overexpressed in mouse models of autosomal dominant polycystic kidney disease (ADPKD)61,62 and Myc expression appears to be tightly regulated by PC1, the product of the Pkd1 gene.63 Fourth, pharmacological inhibition of glucogen synthase kinase 3 beta (GSK3beta), which accelerates cyst formation in cpk mice, leads to decreased Myc expression and amelioration of the cystic phenotype.64 Similarly, Myc is down-regulated in Cys1cpk/cpk;Smad3+/− mice and such double mutants have a milder phenotype than cpk mice.25 Our findings that complementation of mutant cpk with cystin-GFP rescues the cystic phenotype and restores normal Myc expression confirms that cystin acts in vivo as a negative regulator of Myc.
In summary, we demonstrate that cystin deficiency causes an ARPKD-like phenotype in mice that can be rescued by targeted renal expression of a cystin-GFP fusion protein, most probably by downregulating Myc expression in collecting duct cells. Our identification of the first case of human ARPKD associated with a CYS1 mutation confirms the relevance of the cpk mouse as an ARPKD model yielding important insights into molecular mechanisms underlying disease pathobiology.
Methods
Animal study approvals
All mouse experiments were approved by the Institutional Animal Care and Use Committees of Children’s National Research Institute and the University of Alabama at Birmingham (UAB). Knock-in transgenic mice were generated at the University of Alabama at Birmingham (UAB) Transgenic & Genetically Engineered Models Core facility. Ksp-Cre mice were obtained from Jackson Laboratory (Bar Harbor, ME). Mouse colonies were maintained in the animal facility at Children’s National Research Institute.
Antibodies and lectins
Anti-GAPDH antibody was purchased from Cell Signaling Technologies (# 2118). Anti-AQP2 antibody was purchased from Santa Cruz Biotechnologies (# SC9882). Alexa Fluor 488 conjugated anti-GFP antibody was obtained from Life Technologies (# A21311). Polyclonal rabbit anti-cystin antibody (70053) was generated in our lab and described previously.18 Rabbit monoclonal anti-c-Myc antibody was purchased from Abcam (# ab32072). Goat anti-rabbit HRP conjugated secondary antibody was purchased from American Qualex Solution Products (# A102PS). Donkey anti-Goat IgG Alexa Fluor 555 was obtained from Life Technologies (# A21432). Lectins LTA-FITC (# W0909) and DBA-Rhodamine (# Y0828) were obtained from Vector Laboratories.
Vector cloning
Cys1-GFP cDNA was amplified from previously described pEGFP-N1.18 Gateway PCR primers were used to add flanking attB sites. A one-tube Gateway reaction was performed using pDonr221 and pRosa26 Dest.65 The reaction product was used to transform competent STBL3 cells (Life Technologies, # C7373-03) that were plated on Ampicillin and Kanamycin to select for destination and entry clones, respectively. Destination clones were screened by restriction enzyme digest prior to sequencing. The pRosa26 Dest Cys1-GFP targeting vector was linearized with Kpn1 and electroporated into ES cells. G418 resistant colonies were screened by long-range PCR as described66 and positive clones were used to produce chimeric founder mice.
PCR genotyping
PCR conditions are described in Supplemental Table 1.
Immunoblotting
Kidney tissue was collected, homogenized, and processed for immunoblotting as previously described.24 For cystin and control western blots, immuno-reactive protein bands were visualized using SuperSignal West Dura chemiluminescent substrate (Thermo Fisher Scientific, # 34076) and exposed to film. For C-MYC and control western blots, images were obtained with ChemiDoc Imaging System (Bio-Rad laboratory, Inc.) Densitometry was analyzed using Image Lab (Bio-Rad laboratory, Inc., Version 6.0).
Kidney histology
Tissue samples were collected and fixed in 10% formalin (Fisher Scientific, # 23-245-684) for 2 days, then stored in 70% ethanol. The samples were dehydrated, paraffin embedded, cut into 5 μm sections and stained with hematoxylin and eosin (H&E) and slide-mounted for examination by the UAB Comparative Pathology Laboratory.
Immunofluorescence analysis
Tissue samples were collected from 6 week-old mice and processed, using published methods.33 Immunofluorescence detection and image acquisition were performed using an Olympus FLUOVIEW FV1000 confocal laser scanning microscope configured with both an Argon Laser (488 nm) and a Laser diode (405 nm, 440 nm, and 559 nm). Images were analyzed using Olympus FV10-ASW 3.0 Viewer software.
Lectin staining
Five μm sections of fixed paraffin embedded kidney tissues (prepared as described for histopathology) were stained with Rhodamine labeled DBA (Vector Laboratory, # RL-1032) and Fluorescein labeled LTA (Vector Laboratory, # FL-1321).67 Immunofluorescence detection, image acquisition, and analysis were performed as described above.
Human study approval
Approval for human subjects research was obtained from Institutional Review Boards of the University of Michigan, Boston Children’s Hospital, and local IRB equivalents.
Human research subjects
Blood samples and pedigrees were obtained following informed consent from individuals with cystic kidney disease or their legal guardians. The diagnosis of cystic kidney disease was based on published clinical criteria. Clinical data were obtained using a standardized questionnaire (http://www.renalgenes.org).
Whole genome sequencing and mutation calling
TRIO-WES and data processing were performed by the Genomics Platform at the Broad Institute of Harvard and MIT (Broad Institute, Cambridge, MA). Exome sequencing (>250 ng of DNA, at >2 ng/μl) was performed using Illumina exome capture (38 Mb target). Single nucleotide polymorphisms (SNPs) and insertions/deletions (indels) were jointly called across all samples using the Genome Analysis Toolkit (GATK) HaplotypeCaller. Default filters were applied to SNP and indel calls using the GATK Variant Quality Score Recalibration approach. Lastly, variants were annotated using the Variant Effect Predictor. For additional information, refer to the Supporting Information Section S1 in the exome aggregation consortium (ExAC) study.68 The variant call set was uploaded on to Seqr (https://seqr.broadinstitute.org) and analysis of the entire WES output was performed. TRIO-WGS and data processing were performed by the Genomics Platform at the Broad Institute of MIT and Harvard. PCR-free preparation of sample DNA (350 ng input at >2 ng/ul) was accomplished using Illumina HiSeq X Ten v2 chemistry. Libraries were sequenced to a mean target coverage of >30x. Genome sequencing data was processed through a pipeline based on Picard, using base quality score recalibration and local realignment at known indels. The BWA aligner was used for mapping reads to the human genome build 38. Single Nucleotide Variants (SNVs) and insertions/deletions (indels) were jointly called across all samples using Genome Analysis Toolkit (GATK) HaplotypeCaller package version 3.4. Default filters were applied to SNV and indel calls using the GATK Variant Quality Score Recalibration (VQSR) approach. Annotation was performed using Variant Effect Predictor (VEP). Lastly, the variant call set was uploaded to seqr for collaborative analysis between the CMG and investigator.
Mutation calling was performed in line with proposed guidelines,41 and the following criteria were employed as previously described42,43. The variants included were rare in the population with mean allele frequency <0.1% and with 0 homozygotes in the adult reference genome databases ExAC and gnomAD. Additionally, variants were non-synonymous and/or located within splice-sites. Based on an autosomal homozygous recessive hypothesis, homozygous variants were evaluated. Subsequently, variant severity was classified based on prediction of protein impact (truncating frameshift or nonsense mutations, essential or extended splice-site mutations, and missense mutations). Splice-site mutations were assessed by in silico tools MaxEnt, NNSPLICE, HSF, and CADD splice-site mutation prediction scores.69–72 Missense mutations were assessed based on SIFT, MutationTaster and PolyPhen 2.0 conservation prediction scores73–75 and evolutionary conservation based on manually derived multiple sequence alignments.
Homozygosity mapping (HM)
Homozygosity mapping was calculated based on whole exome sequencing data. In brief, aligned BAM files were processed using Picard and SAMtools4 as described previously.76 Single nucleotide variant calling was performed using Genome Analysis Tool Kit (GATK).77 The resulting VCF files were used to generate homozygosity mapping data and visual outputs using the program Homozygosity Mapper.78
Web resources
UCSC Genome Browser, genome.ucsc.edu
Ensembl Genome Browser, www.ensembl.org
gnomAD browser 2.0.3., gnomad.broadinstitute.org
Polyphen2, genetics.bwh.harvard.edu/pph2
Sorting Intolerant From Tolerant (SIFT), sift.jcvi.org
MutationTaster, www.mutationtaster.org
Combined Annotation Dependent Depletion, cadd.gs.washington.edu
NNSPLICE splice-site mutation prediction, www.fruitfly.org/seq_tools/splice.html
MaxEnt splice prediction, hollywood.mit.edu/burgelab/maxent/Xmaxentscan_scoreseq_acc.html
Human Splice Finder, www.umd.be/HSF/
Disclosures
F.H. is a co-founder of Goldfinch Biopharma Inc.
Acknowledgements
The authors thank Gene Siegel, MD, PhD (UAB) for histological analysis and the Cellular Imaging and Analysis Core at Children’s National Research Institute for assistance with microscopy. This work was supported by National Institute of Diabetes and Digestive and Kidney Diseases (NIDDK) grant P30 DK074038. The authors would like to thank members of the University of Alabama at Birmingham Transgenic & Genetically Engineered Models (TGEMs) facility for creating the Cys1-GFP transgene knock-in animals. TGEMs is supported by NIH National Cancer Institute Grant P30CA13148, NIH NIAMS Grant P30AR048311, and NIH NIDDK Grants P30 DK074038, P30 DK05336, and P60 DK079626 (to RAK). F.H. is the William E. Harmon Professor of Pediatrics. This research is supported by a grant from the National Institutes of Health to F.H. (DK-076683-13). A.J.M. was supported by an NIH Training Grant (T32DK-007726), by the 2017 Post-doctoral Fellowship Grant from the Harvard Stem Cell Institute, and by the American Society of Nephrology Lipps Research Program 2018 Polycystic Kidney Disease Foundation Jared J. Grantham Research Fellowship. F.B. was supported by a fellowship grant (404527522) from the German Research Foundation (DFG). Sequencing and analysis were provided by the Broad Institute of MIT and Harvard Center for Mendelian Genomics (Broad CMG) and was funded by the National Human Genome Research Institute, the National Eye Institute, and the National Heart, Lung and Blood Institute grant UM1 HG008900 and in part by National Human Genome Research Institute grant R01 HG009141.