Mammalian Genome

, Volume 24, Issue 3, pp 119–126

Glomerulopathy and mutations in NPHS1 and KIRREL2 in soft-coated Wheaten Terrier dogs

Authors

    • Department of Clinical StudiesUniversity of Pennsylvania School of Veterinary Medicine-Philadelphia
  • Claire A. Wiley
    • Department of Clinical StudiesUniversity of Pennsylvania School of Veterinary Medicine-Philadelphia
  • Michael G. Raducha
    • Department of Clinical StudiesUniversity of Pennsylvania School of Veterinary Medicine-Philadelphia
  • Paula S. Henthorn
    • Department of Clinical StudiesUniversity of Pennsylvania School of Veterinary Medicine-Philadelphia
Article

DOI: 10.1007/s00335-012-9445-8

Cite this article as:
Littman, M.P., Wiley, C.A., Raducha, M.G. et al. Mamm Genome (2013) 24: 119. doi:10.1007/s00335-012-9445-8

Abstract

Dogs of the soft-coated wheaten terrier breed (SCWT) are predisposed to adult-onset, genetically complex, protein-losing nephropathy (average onset age = 6.3 ± 2.0 years). A genome-wide association study using 62 dogs revealed a chromosomal region containing three statistically significant SNPs (praw ≤ 4.13 × 10−8; pgenome ≤ 0.005) when comparing DNA samples from affected and geriatric (≥14 years) unaffected SCWTs. Sequencing of candidate genes in the region revealed single nucleotide changes in each of two closely linked genes, NPHS1 and KIRREL2, which encode the slit diaphragm proteins nephrin and Neph3/filtrin, respectively. In humans, mutations in nephrin and decreased expression of Neph3 are associated with podocytopathy and protein-losing nephropathy. The base substitutions change a glycine to arginine in the fibronectin type 3 domain of nephrin and a proline to arginine in a conserved proline-rich region in Neph3. These novel mutations are not described in other species, nor were they found in 550 dogs of 105 other breeds, except in 3 dogs, including an affected Airedale terrier, homozygous for both substitutions. Risk for nephropathy is highest in dogs homozygous for the mutations (OR = 9.06; 95 % CI = 4.24–19.35). This is the first molecular characterization of an inherited podocytopathy in dogs and may serve as a model for continued studies of complex genetic and environmental interactions in glomerular disease.

Introduction

Protein-losing nephropathy (PLN) in dogs is associated with proteinuria, hypoalbuminemia, hypercholesterolemia, systemic hypertension, thromboembolic events, edema/effusions, and progressive renal failure. More than 25 dog breeds are genetically predisposed to PLN but the mode of inheritance and mechanism is usually unknown (Littman 2011). Only a few breeds have specific genetic mutations identified, and they are models of Alport syndrome (Zheng et al. 1994; Davidson et al. 2007; Nowend et al. 2012).

Approximately 10–15 % of soft-coated wheaten terrier dogs (SCWT) are affected with PLN (Littman et al. 2000). The SCWT Open Registry (www.scwtca.org/openregistry/index.htm) lists 722 SCWTs diagnosed with PLN since 1997 and shows no limitation for age of onset. Since PLN has a late onset (mean age = 6.3 ± 2.0 years), dogs are typically bred before diagnosis, passing risk of disease to progeny. No proven predictive biological markers are known, so identifying the dogs that are clear of risk for developing the disease is challenging. The mode of inheritance appears complex: affected dogs may have long-lived parents, consistent with recessive inheritance, yet conversely, mixed breed dogs with one affected SCWT parent may be affected, implicating dominant inheritance (Littman et al. 2000; Vaden et al. 2002). Glomerulonephritis and/or glomerulosclerosis are described in affected SCWTs (Littman et al. 2000). Functional immunodysregulatory as well as structural developmental theories of pathogenesis for PLN are proposed because the breed is also predisposed to protein-losing enteropathy (PLE), inflammatory bowel disease (IBD), intestinal lymphangiectasia, food allergies, atopy, perinuclear anti-neutrophil cytoplasmic antibodies (p-ANCA), Addison’s disease, and renal dysplasia (Eriksen and Grondalen 1984; Nash et al. 1984; Littman 1999; Littman et al. 2000; Vaden et al. 2000a, b; Allenspach et al. 2008; Wieland et al. 2012). The myriad of genes in these pathways makes a candidate gene approach to defining the underlying molecular mechanism cumbersome. However, recently available high-throughput technologies to genotype single-nucleotide polymorphisms (SNP-chips) have made canine genome-wide association studies (GWAS) possible. In addition, combined with the extensive linkage disequilibrium and long haplotypes resulting from the creation of dog breeds (Ostrander 2012), GWAS studies in dogs are proving to be an effective approach for the study of complex diseases that occur in both humans and dogs. Recent GWAS studies using as few as 22–138 dogs per study have made progress in complex diseases such as systemic lupus erythematosus-related disease complex, atopic dermatitis, and epilepsy (Wood et al. 2009; Wilbe et al. 2010; Seppälä et al. 2011). Considering the potency of the technology, a case-control GWAS seemed appropriate to unravel the genetic basis for PLN in the SCWT breed.

Methods

Samples

Owners requested that their local veterinarians ship DNA samples from PLN-affected and geriatric (14–18 years old) unaffected control SCWT dogs (all were of North American origin) to the PennVet SCWT DNA Bank. Samples of DNA from other breeds were isolated from residual blood samples from the Clinical Laboratory of the Matthew J. Ryan Small Animal Hospital of the University of Pennsylvania and from one affected Airedale terrier from Cornell University Hospital for Animals.

Diagnostic criteria

Disease status was determined by signalment, history, blood, urine, and histopathology findings, as previously described (Littman et al. 2000). Geriatric control dogs were considered free of PLN disease by survival to age 14 years and absence of PLN criteria.

Genome-wide association study

Genomic DNA was extracted from 42 PLN-affected SCWT and 20 geriatric SCWT control dogs using a standard phenol-chloroform extraction protocol. Genotyping was performed using the Illumina Infinium CanineHD Beadchip (Illumina, Inc., San Diego, CA). Association and cluster analyses were performed with the software package PLINK (PLINK 1.07, http://pngu.mgh.harvard.edu/purcell/plink/). Cluster analysis using PLINK confirmed that the affected dogs and geriatric controls were from the same population. After removing SNPs with a minor allele frequency (MAF) ≤ 0.01, a total of 123,738 SNPs were included in the analysis. Bonferroni correction was applied to calculate genome-wide significance.

Candidate gene sequencing

PCR primer pairs for amplification of protein-coding exons and flanking splicing signals were designed using the software program DNASTAR (DNASTAR, Madison, WI) for four candidate genes (NPHS1, KIRREL2, PSENEN, and WTIP). All exons were sequenced for NPHS1, KIRREL2, and PSENEN. However, we were unable to sequence across gaps in the WTIP gene canine reference sequence, despite multiple attempts. Therefore, only 5.5 of the 8 exons of WTIP were sequenced. DNA fragments were amplified by standard PCR conditions, using annealing temperatures determined by gradient PCR (primer sequences and PCR conditions available upon request), and submitted to the University of Pennsylvania’s Sequencing Facility. The use of KOD Xtreme™ Hot Start DNA Polymerase (Novagen, Madison, WI) was necessary for the amplification of several GC-rich fragments. DNA from at least three dogs was sequenced, including one PLN-affected dog homozygous for the PLN-associated haplotype, one geriatric control dog homozygous for the normal haplotype, and one heterozygous dog. Primer sequences and PCR conditions are available upon request.

Additional genotyping

The missense mutation identified in the NPHS1 gene creates a restriction fragment length polymorphism that was used for genotyping additional SCWTs and dogs of other breeds. Briefly, the PCR of NPHS1 missense SNP was carried out in a standard PCR reaction using 20 μM each of oligonucleotide primers (GCCCGGATCCCCCAACA and CACCCCATAAGTCCCCTGAG). The PCR product was then digested with the restriction enzyme MspA1I (New England Biolabs, Beverly, MA) and visualized on a 1 % agarose gel. In addition, custom TaqMan genotyping assays (Illumina) were developed for the NPHS1 and KIRREL2 polymorphisms by the company and performed under standard conditions.

Genotyping of the KIRREL2 SNP was performed through direct sequencing of a PCR fragment containing exon 15 of the gene. PCR products were purified using the Qiaspin Purification Column (Qiagen, Valencia, CA) and submitted to the University of Pennsylvania’s Sequencing Facility.

Protein amino acid alignments were performed using the MegAlign portion of the LaserGene sequence analysis package (DNASTAR, Madison, WI).

PolyPhen-2 analysis of NPHS1 and KIRREL2 mutations utilized the website http://genetics.bwh.harvard.edu/pph2/ to predict the possible impact of the altered proteins’ function (Adzhubei et al. 2010).

Results

Genome-wide association study

We performed a GWAS using the Illumina CanineHD Beadchip on DNA from 42 SCWTs affected with PLN (with or without accompanying protein-losing enteropathy) and 20 geriatric control SCWTs. Association analysis identified three SNPs (rs22026481 at CFA1 position 116,617,009, rs21953589 at CFA1 position 116,931,543, and rs21972858 at CFA1 position 116,937,625) with a praw value of 4.13 × 10−8 and encompassing a 320,805-bp segment of canine chromosome 1 (CFA1; pgenome after Bonferroni correction ≤0.005; Fig. 1a, b). Sixty-four of the 100 SNPs with lowest P values were in or surrounding this region, spanning 9.7 Mb near the end of CFA1. The highest scoring SNPs outside of this CFA1 interval, an isolated SNP on CFA30 (praw ≤ 1.1 × 10−5; pgenome ≤ 1) and a cluster of SNPs on CFA20 (best scoring SNP of praw ≤ 6.5 × 10−6; pgenome ≤ 0.74), did not warrant further consideration due to lack of genome-wide support.
https://static-content.springer.com/image/art%3A10.1007%2Fs00335-012-9445-8/MediaObjects/335_2012_9445_Fig1_HTML.gif
Fig. 1

a Manhattan plot showing the chromosomal distribution (x axis) of the 123,738 informative SNPs with their p values (y axis) when comparing PLN SCWT dogs vs. geriatric normal SCWT dogs. Chromosome 1 has 64 of the lowest 100 p values, including three SNPs at praw ≤ 4.13 × 10−8. b Expanded Manhattan plot of the end of chromosome 1. The gray-shaded region shows the location for the three best scoring SNPs, which encompasses the NPHS1 and KIRREL2 genes (also shaded in Supplementary Table 1). The hatched region shows the region containing the 11 SNPs with p < 6.7 × 10−7 (−log10 > 6.0), evaluated for candidate genes

Candidate gene analysis

The three most significant SNPs are in complete linkage disequilibrium in the 62 dogs analyzed in the GWAS. A region of approximately 2.3 Mb (CFA1 position 115,854,614–118,115,697) surrounding these three SNPs includes the 11 best scoring SNPs (pgenome ≤ 0.077). Within this region is an area of homozygosity among affected dogs homozygous for the disease-associated haplotype (bounded by positions 115,977,846–117,917,907). Genes from the 2.3-Mb region were examined for known or putative function and literature reports of association with mammalian kidney disease (Supplementary Table 1). Four genes in this region warranted further investigation. NPHS1 encodes the protein nephrin, which is a component of the podocyte slit diaphragm and is found mutated in patients with nephrotic syndrome and with focal segmental glomerulosclerosis (OMIM #256300) (Pollak 2009; Welsh and Saleem 2010). KIRREL2 encodes the protein variously called Neph3 or filtrin, a member of the NEPH family of proteins which interact with podocin, another glomerular slit diaphragm protein (Sellin et al. 2003). PSENEN encodes a protein in the Notch pathway; expression of a cleaved Notch fragment correlates with albuminuria and glomerulosclerosis (Murea et al. 2010). WTIP expression maintains glomerular integrity, especially after podocyte injury (Kim et al. 2012).

Two SNPs were detected in NPHS1 protein-coding exons. At CFA1 position 116,800,965 in exon 14, a missense mutation, from A in the CanFam3.1 reference sequence to G, seen in all four SCWT dogs sequenced (including control and PLN-affected dogs), causes a glutamine-to-arginine substitution of amino acid 624. Since arginine is found at this position in humans and many other mammals and is not associated with the disease phenotype, this substitution was not considered to be a candidate PLN mutation. Another nonsynonymous SNP was found in exon 22 of NPHS1, at amino acid position 1,023. This G- (in the canine reference sequence, CFA1 position 116,806,124) to-A substitution causes an amino acid change from glycine to arginine and was homozygous for A in the PLN-predisposing haplotype and homozygous for G in the normal geriatric control. Amino acid G1023 is at the carboxyl end of the fibronectin type 3 (FN3) domain, located next to the transmembrane domain on the extracellular side. Glycine in this position is conserved in mammalian species in which the FN3 domain is also conserved (Fig. 2).
https://static-content.springer.com/image/art%3A10.1007%2Fs00335-012-9445-8/MediaObjects/335_2012_9445_Fig2_HTML.gif
Fig. 2

Multispecies alignment of a portion of the NPHS1 gene product (nephrin) surrounding the novel SNP associated with PLN in SCWT. Numbers are amino acid positions. The position of the G1023R amino acid substitution is shaded. Dots represent identical residues to the human sequence. Amino acids are indicated using the standard nomenclature. Asterisks indicate residues with variant alleles associated with PLN in human patients (Abid et al. 2012)

Sequencing of the exons of KIRREL2 revealed that exon 7 contains multiple sequence differences when compared to the reference genome sequence. All differences were common to all of the SCWTs sequenced, including control and PLN-affected dogs. However, the quality of the dog genome sequence is poor in this exon (as indicated by the quality scores across this region, as visualized in the UCSC genome browser, http://genome.ucsc.edu/). The predicted protein sequence from our sequence data differs from the dog reference genome sequence by one single base insertion and five base substitutions (causing a frameshift mutation and several nonsynonymous substitutions), and it shows better protein homology with other mammalian sequences. Given the lack of sequence differences among the SCWTs sequenced, these sequence differences in exon 7 were considered not relevant to this study. However, a single SNP in exon 15 (CFA1 position 116,785,027), changing a cytosine to a guanine, causes a nonsynonymous substitution of proline to arginine at amino acid P626 (P630 in human isoform 3 of the KIRREL2 gene) in a proline-rich region that is conserved among mammals (Fig. 3).
https://static-content.springer.com/image/art%3A10.1007%2Fs00335-012-9445-8/MediaObjects/335_2012_9445_Fig3_HTML.gif
Fig. 3

Multispecies alignment of a portion of the KIRREL2 gene product (Neph3) surrounding the novel SNP associated with PLN in SCWT. Numbers are amino acid positions. The position of the P626R amino acid substitution is shaded. Dots represent identical residues to the human sequence. Dashes are gaps inserted to optimize the alignment. Amino acid substitutions are indicated using the standard amino acid nomenclature

There were no PLN-associated mutations found in the PSENEN and WTIP genes. PolyPhen-2 analysis scored the NPHS1 G1023R as “probably damaging” (0.996) and the KIRREL2 P626R as “possibly damaging” (0.945).

Genotyping additional dogs

The NPHS1 G1023R and KIRREL2 P626R mutations were genotyped in a total of 145 PLN-affected (83 dogs) and geriatric (62 dogs) SCWTs. Among PLN-affected dogs, 63 were homozygous for PLN-predisposing SNPs, 18 were heterozygous, and 2 were homozygous for the normal/reference genome alleles. Among geriatric controls, 16 were homozygous for PLN-predisposing alleles, 31 heterozygous, and 15 homozygous for the control alleles. The significance of these haplotype distributions (based on Fisher’s exact test of the 2 × 3 contingency table, 2-tailed) is 7.78 × 10−10. Odds ratios (ORs) were calculated to assess the significance of haplotypes among affected and control dogs (Table 1). Dogs homozygous for the PLN-predisposing haplotype, and thus for the amino acid substitutions in NPHS1 and KIRREL2 found in this haplotype, are at very high increased risk of developing PLN compared to dogs that are not homozygous, or are homozygous for the “normal” haplotype or alleles.
Table 1

Odds ratio (OR) comparisons of haplotypes among affected and control dogs

Haplotypes compared

OR (95 % CI)

Fisher’s exact p value

1-tailed

2-tailed

Predisposing vs. NL

29.5313 (6.1183, 142.5383)

2.07 e-7

2.07 e-7

Predisposing vs. HET

6.7813 (3.05, 15.0771)

1.12 e-6

1.16 e-6

Predisposing vs. (NL + HET)

9.0563 (4.2379, 19.3529)

1.51 e-9

1.87 e-9

HET vs. NL

4.3548 (0.8919, 21.2621)

0.0472

0.0693

NL normal/reference genome, HET heterozygous

We searched for the presence of NPHS1 allele G1023R in the DNA of 747 dogs of 114 other breeds, including 91 dogs of 16 other terrier breeds, dogs of other breeds that are predisposed to PLN (Littman 2011), and 9 dogs of other breeds that had been diagnosed with PLN. The only dogs not homozygous for the normal/reference genome allele were two Airedale terriers. One was affected with PLN and homozygous while the other was heterozygous for the PLN-predisposing NPHS1 allele. The KIRREL2 allele P626R was genotyped in a subset of 550 dogs (105 breeds) of the 747 dogs, including all of the terrier dogs. In this group of dogs, only three dogs were not homozygous for the normal KIRREL2 allele. As expected, they included the two Airedale terriers, one affected and homozygous and the other heterozygous for the PLN-predisposing KIRREL2 allele. One bloodhound was heterozygous at only the KIRREL2 SNP.

Discussion

With the goal of understanding the genetic and pathological basis of PLN in SCWT dogs, we used a GWAS to compare DNA samples from PLN-affected and geriatric unaffected SCWTs to find loci of interest that would allow us to narrow the pool of gene candidates for further examination. The GWAS showed strongest support for association of PLN to a locus on chromosome 1. Examination of this region identified two promising candidate genes, NPHS1 and KIRREL2, which encode the podocyte slit diaphragm proteins nephrin and Neph3. We sequenced the exons and intron/exon boundaries of these two genes from SCWT dogs diagnosed with PLN or PLN combined with PLE, as well as from control healthy geriatric SCWTs, and discovered a previously unidentified nonsynonymous SNP in each gene. For both proteins, the amino acid changes are in an amino acid residue conserved among mammals, and replace smaller amino acids (glycine and proline) with the larger arginine residue, which may alter protein structure and lead to deleterious effects in the glomerular filtration barrier. PolyPhen2 assessment of the impact of these changes indicated that the protein alterations identified in NPHS1 and KIRREL2 are likely to interfere with their normal functions, compromising the permeability of the glomerular filtration barrier. While our studies do not address the issue of which amino acid substitution, or both together, may be causative, published support for the role of NPHS1 in kidney function and disease is stronger.

Examination of a larger population of SCWT dogs determined that homozygosity for the variant alleles of these SNPs is a predisposing factor for the development of PLN in this breed and in one additional terrier breed (Airedale).

Nephrin and Neph3 are members of the immunoglobulin superfamily, a subset of which function together as transmembrane proteins at the slit diaphragm, along with a myriad of other proteins, in a 3-dimensional dynamic construct connecting to the podocyte actin cytoskeleton. More than 176 mutations in NPHS1 have been associated with human PLN, including infantile, childhood, and adult onset of the disease (Narita et al. 2003; Löwik et al. 2009; Machuca et al. 2009; Santin et al. 2009; Zenker et al. 2009; Caridi et al. 2010; Schoeb et al. 2010; Welsh and Saleem 2010; Abid et al. 2012; Ovunc et al. 2012). We identified a novel nonsynonymous SNP in exon 22 of the NPHS1 gene in affected SCWTs, but not in controls. The disease-associated mutation G1023R is at the carboxy-terminal boundary of the single fibronectin type-3 domain in the nephrin protein. Missense mutations G1020 V and S1016 N in the FN3 domain have been observed in nephrotic syndrome patients (Fig. 2) (Abid et al. 2012).

A nonsynonymous mutation in exon 15 of the gene that encodes canine Neph3 (KIRREL2) is also associated with SCWT PLN. This mutation is in a proline-rich region that is conserved in mammals. While no disease-associated mutations in Neph3 have been identified, its interaction, homology, and structural similarity to nephrin, Neph1, and Neph2 and its downregulation in human proteinuric diseases suggest that it may play a role in the structure and function of the slit diaphragm (Ihalmo et al. 2003, 2007; Hartleben et al. 2008; Heikkilä et al. 2011; Ristola et al. 2012).

Additional support for the concept that one or both of the nonsynonymous SCWT SNPs are PLN-associated comes from the fact that they are extremely rare in dogs of other breeds. Only three dogs among the 550 dogs tested for both SNPs had any copies. Two were dogs of a terrier breed (Airedale), with the heterozygous dog ascertained from a population screen and the homozygous dog was itself affected with PLN. Our data do not differentiate which mutation, in NPHS1 or KIRREL2 or both, is associated with PLN in SCWT dogs. These genes reside close together in the genome, and the resulting linkage disequilibrium may account for the association of both with disease. The identification of dogs in which the NPHS1 and KIRREL2 PLN-predisposing alleles are no longer linked may by possible through the study of additional dogs, particularly bloodhounds, since we found one bloodhound with a disease-associated KIRREL2 allele linked to the normal NPHS1 allele on one chromosome.

In addition to determining that NPHS1 and KIRREL2 missense mutations were associated with PLN in SCWTs, we assessed phenotypic risk among the genotypes in a larger SCWT population. Dogs homozygous for both mutations are at significantly increased risk to develop PLN. Heterozygotes appear at intermediate risk, although based on the 95 % confidence interval, it is possible that heterozygous dogs do not differ from controls, or they may be at increased risk, possibly if combined with immunodysregulatory disorders, environmental triggers, or other predisposing loci. Dogs lacking the predisposing alleles may have PLN due to infectious, inflammatory, immune-mediated, neoplastic, or other causes. Although the phenotype–genotype relationship is not perfect, this fact is not surprising; data from the SCWT Open Registry showed that PLN is not inherited in a simple Mendelian manner and therefore may involve multiple genes and environmental triggers leading to the observed variable expression and incomplete penetrance.

The pathogenesis of SCWT PLN has been thought to involve immune-complex glomerulonephritis, since food allergies, IBD, and/or PLE are often seen prior to or combined with PLN. The SCWT breed is also at risk for atopy, p-ANCA antibodies, and Addison’s disease. Renal histopathology has most often been available from end-stage necropsy sampling and studied only by light microscopy, which could confuse glomerulonephritis and glomerulosclerosis. In field studies, the renal lesion has been described to look most like focal segmental glomerulosclerosis (FSGS) (Littman et al. 2000). A handful of affected SCWT and affected mixed-breed dogs with one SCWT parent (“Wheagles”), with early-stage renal histopathology, including electron microscopy and immunofluorescent studies, showed mild focal proliferative glomerulonephritis with IgA, IgM, and C3 deposition in 4 % of glomeruli at 2 years of age and glomerulosclerosis in 11 % of glomeruli at 4 years of age (Afrouzian et al. 2001). The combination of immunodysregulatory predisposition and structural defects may work synergistically. For example, structural defects that result from these PLN-associated nonsynonymous SNPs may predispose the glomerulus to immune-complex deposition, resulting in an increased risk for glomerulonephritis, not just FSGS. Similarly, people with IgA nephropathy had more severe disease if they also carried one or more copies of the E117K NPHS1 population variant (Narita et al. 2003).

In summary, we propose that PLN in the SCWT breed represents a canine genetic podocytopathy, analogous to that seen in humans. It is associated, and probably caused by, an amino acid substitution in either or both of the proteins nephrin and Neph3, encoded by NPHS1 and KIRREL2, respectively. By offering breeders a DNA-based test for the SNPs causing these amino acid substitutions, it should be possible to decrease the incidence of PLN in this breed through careful selective breeding that maintains genetic diversity in the breed. In addition, early detection and monitoring of dogs carrying these disease-associated SNPs will allow for earlier intervention to decrease progression of disease. The identification of a disease-predisposing locus for a complex genetic disease with 62 dogs in the initial GWAS demonstrates the power of mammalian genetic disease studies using the isolated populations provided by dog breeds. Additional genetic and clinical studies with the other dog breeds predisposed to PLN may reveal additional models for human glomerular disease and increase our understanding of the molecular basis of these entities.

Acknowledgments

The authors thank Rachel Cianciolo, Donna Dambach, Mattie Hendrick, Rebecca Kessler, George Lees, Junlong Liu, Alisa Newton, Shelly Vaden, Brian Wilcock, and the many veterinarians, students, breeders, and SCWT owners who helped with this work. We received financial support from the Soft Coated Wheaten Terrier Club of America Endowment, Inc., and American Kennel Club – Canine Health Foundation.

Supplementary material

335_2012_9445_MOESM1_ESM.rtf (136 kb)
Genes located in the PLN-associated interval. The region between the three most significant SNPs is shown in light gray, with the flanking regions encompassing all SNPs with p < 6.7 × 10−7 (RTF 137 kb)

Copyright information

© Springer Science+Business Media New York 2013