Developmental dysplasia of the hip (DDH) is a debilitating condition characterized by incomplete formation of the acetabulum leading to dislocation of the femur, suboptimal joint function, and accelerated wear of the articular cartilage resulting in arthritis. DDH affects 1 in 1000 newborns in the United States; there are well-defined “pockets” of high prevalence in Japan, and in Italy and other Mediterranean countries. Although reasonably accurate for detecting gross forms of hip dysplasia, existing techniques fail to find milder forms of dysplasia. Undetected hip dysplasia is the leading cause of osteoarthritis of the hip in young individuals, causing over 40% of cases in this age group. A sensitive and specific test for DDH has remained a desirable yet elusive goal in orthopedics for a long time. A 72-member, four-generation affected family has been recruited, and DNA from its members retrieved. Genomewide linkage analysis revealed a 2.61-Mb candidate region (38.7–41.31 Mb from the p term of chromosome 3) co-inherited by all affected members with a maximum logarithm (base 10) of odds (LOD) score of 3.31. Whole exome sequencing and analysis of this candidate region in four severely affected family members revealed one shared variant, rs3732378, that causes a threonine (polar) to methionine (non-polar) alteration at position 280 in the transmembrane domain of CX3CR1. This mutation is predicted to have a deleterious effect on its encoded protein, which functions as a receptor for the ligand fractalkine. By Sanger sequencing this variant was found to be present in the DNA of all affected individuals and obligate heterozygotes. CX3CR1 mediates cellular adhesive and migratory functions and is known to be expressed in mesenchymal stem cells destined to become chondrocytes. A genetic risk factor that might be among the etiologic factors for the family in this study has been identified, along with other possible aggravating mutations shared by four severely affected family members. These findings might illuminate the molecular pathways affecting chondrocyte maturation and bone formation. © 2013 American Society for Bone and Mineral Research.
Developmental dysplasia of the hip (DDH) is a debilitating condition characterized by incomplete formation of the acetabulum leading to dislocation of the femur, suboptimal joint function, and accelerated wear of the articular cartilage resulting in arthritis. DDH affects 1 in 1000 newborns in the United States; there are well-defined “pockets” of high prevalence in Japan, and in Italy and other Mediterranean countries. Because of its high prevalence and undesirable consequences, screening programs involving manipulation of the femur or ultrasound imaging of the hip in infants are in place in most countries. Although reasonably accurate for detecting gross forms of hip dysplasia, existing techniques fail to find milder forms of dysplasia. Undetected hip dysplasia is the leading cause of osteoarthritis of the hip in young individuals, causing over 40% of cases in the 20-year-old to 40-year-old age group. A sensitive and specific test for DDH has remained a desirable yet elusive goal in orthopedic medicine for a long time.
DDH is a complex disorder having an etiology that is both environmental and genetic.[5-7] Environmental risk factors include breech presentation (with feet toward cervix), oligohydramnios (deficiency of amniotic fluid), and primiparity (first-born).[3, 8, 9] Evidence for a genetic cause is well established and includes a higher concordance between monozygotic (41%) than dizygotic (2.8%) twins, and a 12-fold increase of DDH among first-degree relatives of those affected by the disorder.[10, 11] DDH appears to be transmitted in an autosomal dominant fashion in several families and, perhaps because of its complex etiology, exhibits incomplete penetrance. Our hypothesis is that DDH-affected individuals have mutation(s) or genetic variants that make them susceptible to the disorder. The goal of this study is to identify genetic susceptibility factors for DDH, and in so doing, lay the foundation for a genetic test to accurately identify susceptible newborns so that intervention with a device such as a Pavlik harness can be used to allow complete development of the acetabular labrum. Should this goal be attained our understanding of molecular pathways responsible for acetabular development will be enhanced.
Subjects and Methods
Clinical diagnostic criteria
Before the initiation of this study, informed consent was obtained from each participant and approval was obtained from the Institutional Review Board of Thomas Jefferson University. We have recruited a large family from Utah that shows transmission of DDH through four generations and have isolated DNA from 72 family members (Fig. 1). Individual family members were diagnosed using clinical exams and supine anterior posterior radiographs of the pelvis. Imaging of the hips was evaluated by three orthopedic surgeons, with clinical opinions of two additional surgeons elicited in any cases of disagreement. Shenton's line (disrupted = affected), center edge angle (<20 degrees = affected), Tonnis angle (>10 degrees = affected), extrusion distance (>10 mm = affected), and femoral neck angle were measured in each radiograph and compared to control values derived from 11 independent studies. Detailed clinical evaluation and criteria for diagnosis of this family has been described. In general, those individual family members with one or two signs of DDH were deemed questionable, those with three or more signs were deemed affected.
Genomewide genotyping was performed on DNA isolated from venous blood or cheek epithelial cells obtained from members of this family. The samples were analyzed using the Affymetrix GeneChip Human Mapping 250K Nsp Array following Affymetrix's protocol (Affymetrix, Santa Clara, CA, USA). Briefly, total genomic DNA (250 ng per sample) at 50 ng/µL was digested with the restriction enzyme, Nsp I. Adaptors that recognized the cohesive 4-bp overhangs from the digestion reactions were ligated to the products. Following the ligation reaction, a PCR primer, complementary to the ligated adaptor sequence, was used to amplify the DNA fragments. The PCR products were verified and analyzed on a 2% agarose gel. The products from the separate PCR reactions were pooled and purified using AMPure XP beads (Beckman-Coulter Genomics, Danvers, MA, USA). Purified products were quantitated using the NanoDrop ND-1000, fragmented, and analyzed on a 4% agarose gel. The samples were then labeled, hybridized to the 250K Nsp Arrays, washed on a Fluidics Station 450, and scanned according to Affymetrix's protocols on a GeneChip Scanner 3000G with Autoloader. Quality control assessment of the sample files was performed using Genotyping Console v4.0.
Using 261,961 SNPs as a starting point, only autosomal markers were used because male to male transmission within the pedigree ruled out X-linked inheritance. Individual samples were removed if the proportion of missing SNPs was greater than 0.1. The following filters were then applied: (1) SNPs with a missing genotype rate greater than 0.01 were excluded; (2) SNPs with a minor allele frequency less than 0.1 were excluded; and (3) SNPs with a Mendelian error rate >0 (as defined by PLINK version 1.07) were excluded. Using the HapMap Phase 2 CEU population (variants found in 180 DNA samples of Utah residents of northern and western European descent), SNPs were further pruned to remove those in high linkage disequilibrium. Specifically, we used a method that considered SNPs pairwise in a sliding window (50 base pairs [bp] wide and 5-bp step size), and removed one from each pair with a correlation coefficient of 0.1 or more. Finally, we excluded SNPs showing additional Mendelian inconsistencies detected by MERLIN version 1.12. Thus the final number of SNPs analyzed was 18,778.
Model-based and model-free multipoint maximum logarithm (base 10) of odds (LOD) score analysis were performed using the MERLIN software. Because of the size of the pedigree and the limitations of MERLIN, genomewide analysis was performed by including only the affected individuals and their parents. Once a likely candidate region was identified, analysis of the informative markers in the candidate region was repeated using LINKMAP in the entire pedigree. In both cases, an affected-only model-based analysis was run assuming a dominant model with no phenocopies and a very low disease allele frequency of 0.0001. All individuals with three or more signs of disease were coded as affected. All individuals with two signs or less, and unaffected individuals who are in the blood line (not married into the family) were coded as unknown, whereas unaffected individuals who married into the family were coded as unaffected.
To test whether the linkage signal on chromosome 3p can be explained by the variant identified by exome sequencing, linkage analysis conditional on rs3732378 was performed by means of the approach described by Sun and colleagues including informative markers in the candidate region. This approach uses the T2 statistical test, which is often used for studies of complex traits. It uses multiple SNP markers simultaneously and considers the effects of multiple disease susceptibility loci. The rationale for this approach is that if a variant in a candidate linkage region influences susceptibility to the disease, there should not be additional evidence for linkage when conditioning on the genotypes at the candidate variant. The software Stepc (http://genemed.bsd.uchicago.edu/old/software.html) implements this approach as a test of the null hypothesis that the variant tested is a disease susceptibility variant. A Z-score is calculated before and after conditioning on the candidate SNP to evaluate the evidence for linkage in a given region. A nonsignificant result after conditioning, which indicates no residual evidence for linkage, is in agreement with the hypothesis that the SNP tested is a disease susceptibility variant.
Four severely affected family members with three or more signs of the disorder were selected for whole exome sequencing and for analysis of shared exonic variants. Control reference sequence was derived from the 1000 Genome project (http://www.1000genomes.org) and from the GRCh37 assembly of the National Center for Biotechnology Information (NCBI). Exome capture on each individual was performed using a SureSelect Human All Exome kit v.1 (Agilent Technologies, New Castle, DE, USA) and sequencing was performed using a Genome Analyzer IIx (Illumina, Inc, San Diego, CA, USA). Sequence reads were mapped to the reference genome using the Burrows Wheller Aligner Tool (BWA). An average of 97.25 million reads were available for each individual, 92.5% of which aligned to the targeted region. An average of 91.6% targeted regions were covered at a read depth of at least 20× (Supplementary Fig. S1). The transitions versus transversions in SNPs (Ts/Tv) ratios were 2.1, 2.099, 2.171, and 2.098 for individuals 1, 3, 9, and 10, respectively (see Supplementary Fig. S3 for additional sequencing metrics).
The list of variants from all four affected individuals was filtered and the following variants were retained:
- Variants shared by all four affected individuals with a quality score greater than 20. Quality scores measure the probability that a base is called incorrectly. The quality score of a given base, Q, is defined by the equation Q = –10log10(e), where e is the estimated probability of the base call being wrong. Thus, a higher quality score indicates a smaller probability of error. For example, a quality score of 20 represents an error rate of 1 in 100, with a corresponding call accuracy of 99%.
- Shared variants that mapped to the linkage candidate region.
- New mutations (those not found in dbSNP build 135), nonsynonymous SNPs, and splice-site–causing variants within the subset.
Nonsynonymous variants shared by all four affected individuals were evaluated for functional effects on the protein they encoded using Polymorphism Phenotyping v2 (PolyPhen-2) and Sorts Intolerant From Tolerant (SIFT).[18, 19] PolyPhen-2 uses structural features of the protein as well as information on evolutionary conservation to predict whether a mutation affects protein function. SIFT sorts nucleotide changes that are tolerated from those that are not based on their evolutionary conservation. Variants were screened for biological significance using by determining Genomic Evolutionary Rate Profiling (GERP) scores. GERP identifies evolutionarily constrained elements in multiple alignments by quantifying substitution deficits. These deficits represent substitutions that would have occurred if the element were neutral DNA, but did not occur because the element has been under functional evolutionary constraint.
Validation of the sequencing results in the four severely affected members and in other family members and in-laws (individuals married-into the family) was performed using Sanger DNA sequencing.
Filtering whole-exome variants for relevance to bone was done using Ingenuity software. Starting with 1,518,628 variants spanning 20,018 genes, the following variants were excluded: (1) those observed with an allele frequency greater than or equal to 0.1% of the genomes in the 1000 genomes project; (2) those greater than or equal to 0.1% of the public Complete Genomics genomes; or (3) those greater than or equal to 0.1% of the National Heart Lung and Blood Institute Exome Sequencing Project (NHLBI ESP) exomes. Variants retained were those (1) experimentally observed to be associated with a phenotype: pathogenic, possibly pathogenic, unknown significance; (2) with established gain of function in the literature or gene fusions or inferred activating mutations by Ingenuity; (3) with predicted gain of function by SIFT; (4) that resided in a microRNA binding site; (5) that were nonsynonymous and not predicted to be innocuous by SIFT; (6) that disrupted a splice site; or (7) that were deleterious to a microRNA. Retained data also included those data that are associated with gain of function or were compound heterozygous or homozygous or heterozygous or haploinsufficient or hemizygous and have call quality greater than 20 and occur in at least four of the case samples at the variant level in the case samples. Finally, data passing these screens were filtered for those variants that are known or predicted to directly affect bone formation.
The versions of software used in this analysis were as follows: Ingenuity Variant Analysis version 1.2.20121206 Content versions: Ingenuity Knowledge Base (Vega 121015.000), COSMIC (v61), dbSNP (Build 137), 1000 Genome Frequency (v3), TargetScan (v6.1), EVS (ESP6500), JASPAR (10/12/2009), PhyloP hg18 (11/2009), PhyloP hg19 (01/2009), Vista Enhancer hg18 (10/27/2007), Vista Enhancer hg19 (12/26/2010), CGI Genomes (11/2011), SIFT (06/2012), BSIFT (06/2012), and TCGA (5/14/2012).
Description of the affected family
Diagnostic criteria for the family recruited for this study have been described in detail. Briefly, this four-generation, 72-member family from Utah shows transmission of DDH in a manner consistent with an autosomal dominant mode of inheritance with incomplete penetrance (Fig. 1). Eleven patients had three or more signs of DDH and were considered to be unequivocally affected (see Subjects and Methods). Thirteen individuals had one or two signs of DDH and had questionable diagnoses. Originally, individual 27, an adolescent, was classified as affected. Subsequent review by a panel of orthopedic surgeons changed this diagnosis to questionable. She was therefore classified as unknown for the purpose of linkage analysis in the current study. Poor quality of DNA in affected individuals 2 and 6 resulted in missing genotype rate >0.1 and caused them to be excluded from the analysis.
Model-free and model-based analyses were performed using the linkage analysis program MERLIN. Both genomewide analyses revealed an identical candidate region on chromosome 3 with both analyses producing a maximum LOD score of 3.31 in an interval delimited by SNPs rs4481097 and rs4626072 at 38.91 to 40.66 Mb from the p-terminal end of chromosome 3 (Fig. 2). There was no other region in the genome that yielded LOD scores greater than or equal to 1.055. Crossovers in individuals 47 and 33 determined the proximal and distal boundaries of the candidate region to be 38.7 Mb and 41.31 Mb, respectively, from the p-terminus of chromosome 3 (Supplementary Fig. S2). Model-based affected-only analysis in the candidate region using the software LINKMAP that allowed inclusion of all individuals resulted in a maximum LOD score of 3.28.
Whole exome analysis of four severely affected family members (individuals 1, 3, 9, and 10 in Fig. 1) was performed. Supplementary Table S1 shows the 96 shared variants that mapped to the candidate region, including seven novel mutations. All seven shared new mutations were intronic or found in untranslated regions except for the CATG/C indel found in the coding region of the DLEC gene. This indel was found to reside in a region of the DLEC gene that is not conserved in evolution. No shared splice-site mutations or variants within the candidate region were found (Supplementary Table S1). Shared nonsynonymous variants that mapped to the chromosome 3 candidate region were further evaluated for functional significance by PolyPhen-2 and SIFT. Of the nine nonsynonymous shared SNPs in the candidate region, only one (rs3732378) was predicted to have a deleterious effect on its encoded protein (Table 1). PolyPhen-2 predicted that this variant was “possibly damaging.” SIFT predicted that this variant was damaging (score of 0.01, with a range of 0 = damaging to 0.05 = tolerated). This SNP is in the coding sequence of the chemokine (C-X3-C motif) receptor 1 (CX3CR1) that functions as a receptor for the ligand fractalkine, which mediates cellular adhesive and migratory functions. Fractalkine or CX3CL1 is a 373–amino acid protein that has a chemokine domain located on top of a mucin-like stalk. Many cell types secrete CX3CL1, which exists in both membrane-bound and soluble forms. The membrane-bound CX3CL1 can serve as an adhesion molecule for cells expressing the fractalkine receptor (CX3CR1). The C to T missense variant in the CX3CR1 receptor causes a threonine (polar) to methionine (non-polar) alteration at position 280 (hence referred to as M280) in the transmembrane domain of the protein.
|Chromosome||Position||Chromosome position||Gene||Region||Type||dbSNP||Sift assessment||PolyPhen 2 assessment|
A number of regulatory elements were found near rs3732378 from the ENCODE database. The histone methylation sites H3K4Me1 and H3K4Me2 were found to overlap the position of the variant rs3732378 on chromosome 3 DNA. These sites are often located near regulatory elements. This DNA site is also involved in binding the transcription factors GATA2 and JunD in the K562 cell line. The RNA binding protein polyA binding protein C1 (PABPC1) also binds to this site in the GM12878 cell line (data not shown).[25, 26] This cell line is lymphoblastoid with a relatively normal karyotype. Both this cell line and cartilage are of mesodermal origin. PABPC1 mRNA and protein are expressed in the cartilage of 6-year-old to 12-year-old children and adolescents, and this gene is expressed in hypertrophic growth plate cartilage.
Validation by Sanger sequencing
The presence of SNP rs3732378 was validated by Sanger sequencing in the four affected individuals and in the DNA of all 13 other affected individuals and obligate heterozygotes within the family. Sanger sequencing was also performed to discover the presence of this SNP in individuals with fewer signs of the DDH as well as in seemingly unaffected in-laws (see Fig. 3). Individuals 8, 18, 32, 38, 44, 45, and 59, who appeared to have fewer signs of DDH, were found to carry this variant. This variant was also present in the DNA of married-in individuals 21, 46, 14, and 64. (Fig. 3).
Conditional linkage analysis
To test whether SNP rs3732378 explained the linkage signal on chromosome 3p, we used the T2 test implemented in Stepc (see Sun and colleagues) to evaluate the overall evidence for linkage in the critical region before and after conditioning on the genotype at rs3732378. Although evidence for linkage before conditioning was significant with a Z-score of 2.3849 (p = 0.0085), analysis conditional on genotypes at rs3732378 resulted in a Z-score of 0.9399 (p = 0.1736), thus not rejecting the null hypothesis that rs3732378 is a disease susceptibility variant.
Screening shared variants outside candidate region for relevance to bone formation
Using Ingenuity software, variants were filtered for relevance to bone formation (see Subjects and Methods for detailed description). After validation by Sanger sequencing, six variants were found to pass all filters (Table 2). Four of these base changes were previously documented variants, two of them are novel mutations. Three of the six variants involve insertions or deletions of a triplet codon encoding an amino acid. One variant on chromosome 17p11.2 causes a frameshift mutation in RAI 1. Another novel mutation on chromosome 17q21.2, a deletion of 12 bases, resides in the promoter region of one gene (TMEM99) and in the coding region of another (KRT10). Within this deleted segment are three CpG “islands” from 38975330-38975338 bp from the p-terminal end of chromosome 17 (data not shown). One variant found in intron 1 of BMP8b on chromosome 1p34.2 is predicted to be activating. Activating mutations are those predicted by the SIFT algorithm to have the greatest potential for increasing reproductive fitness. Data for the predicted effects of the other known variants in this group was unavailable in either the SIFT or the PolyPhen-2 databases.
|Chromosome||Position||Reference||Sample||Genomic region||Gene symbol||Transcript variant||Protein variant||Individual 1||Individual 3||Individual 9||Individual 10||Predicted impact Sift||SNP or novel||Notes|
|1||40235489||T||C||Exon; intron||oxct2; bmp8||1439A > G||E480G||C/T||C/T||C/T||C/T||Activating||rs79011683|
|11||65325325||10 CAGs||12 CAGs||Exon||LTBP3||105_106 insertion||35_36 insertion of 2 Leu||10 CAG/12 CAG||10 CAG/12 CAG||10 CAG/12 CAG||10 CAG/12 CAG||Unknown||rs71036212|
|16||67229793||13 CAGs||14 CAGs||Exon||E2F4||917_918 insertion||307_308 insertion of 1 Asp||13/14 CAG||13/14 CAG||13/14 CAG||13/14 CAG||Unknown||Novel||Mutation is in highly conserved DNA sequence|
|17||17697094||13 GACs||12 GACs||Exon||RAI1||832_834 deletion||278delQ||12/13 GACs||12/12 GACs||12/12 GAC's||12/13 GACs||Unknown||rs11303801, rs71944489||Mutation results in removal of 1 conserved G in series of 14|
|17||17697102||G||Deleted G||Exon||RAI1||840delG||Q280 frameshift||del/G||del/del||del/del||del/G||Unknown||rs34083643|
|17||38975329||G||(See note)||Promoter||TMEM99||1457_1451||487_488in||del/del||wt/del||wt/del||wt/del||Unknown||Novel||Mutation is detection of 12 bp underlined GCCGCCGCCG > TGGCCGCCGCCGGAGCTTCCGC|
We have identified and retrieved DNA from one of the largest documented families showing intergenerational transmission of DDH. The goal of this study is to identify the molecular basis of the disease in this family using the approach of genomewide linkage analysis together with whole-exome sequencing, as illustrated schematically in Fig. 4. By performing genome-wide linkage analysis, which makes no assumptions about where a mutation might reside, a 2.61-Mb candidate region on chromosome 3p22.2 has been identified with a high degree of certainty. However, classical linkage analysis requires unambiguous knowledge of who is affected in a given pedigree. Because this disorder is complex, with both environmental and genetic causes, and shows incomplete penetrance, not all individuals who appear unaffected will, in fact, be free of the disease allele. One method that we have chosen to address the problem of unknown, incomplete penetrance in our analysis is to define as affected only those family members whose diagnosis is certain (ie, those having three or more signs of DDH). Penetrance for these affected individuals is irrelevant and unaffected individuals are scored as having an unknown phenotype. Using this “affected only” model-based analysis assuming an autosomal dominant mode of inheritance, linkage analysis was performed and the results compared to model-free linkage analysis in which no assumption about the mode of inheritance was made. In both analyses the same candidate region on chromosome 3 resulted, with no other genomic region producing LOD scores higher than 0 in the model-based analysis or 2.055 in the model-free analysis.
The maximum LOD score of 3.31 was found in an interval spanning 38.91 to 40.66 Mb from the p-terminal end of chromosome 3 (Fig. 2). This 2.61-Mb region has within it several candidate genes that might explain many of the signs seen in individuals affected with DDH. To further scrutinize these candidates, exomes of four severely affected family members were sequenced. Results of sequencing for the chromosome 3 candidate region shown in Table 1 reflect a spectrum of variants including synonymous and nonsynonymous SNPs, variants that mapped to the untranslated region of various genes, deletions, and SNPs in intronic and intergenic regions, as well as seven novel mutations. Novel mutations are those that appear in Supplementary Table S1 without a dbSNP designation and are not found in the single nucleotide database 135. One of these novel mutations, an indel in the DLEC gene, was located in a part of the coding sequence found to be highly variable in evolution. The six other novel mutations occurred in intronic regions not related to mRNA splicing or in untranslated regions of their respective genes. None of these were in conserved genomic regions and it is difficult to evaluate their significance at this time.
In order to predict whether shared nonsynonymous variants might be deleterious to the function of the respective genes, they were analyzed by both PolyPhen-2 and SIFT. Among the subset of nine shared nonsynonymous variants, only one was predicted to be detrimental to protein function. The variant rs3732378 causes a C to T transition in the coding region of the chemokine receptor (CX3CR1). This missense mutation converts a threonine (polar) to methionine (non-polar) at position 280 in the transmembrane domain. PolyPhen-2, which uses structural features of the protein as well as information on evolutionary conservation, predicted that this mutation was “possibly damaging.” SIFT predicts that this mutation is damaging (score of 0.01, with a range of 0 = damaging to 0.05 = tolerated) (Supplementary Table S1).
Sequence conservation in evolution is often considered a measure of relative biologic importance. GERP is a method for producing position-specific estimates of evolutionary constraint using maximum likelihood evolutionary rate estimation. None of the candidate region variants (with the exception of rs3732378) were found to localize in regions of the genome that were highly evolutionarily conserved. At position 280 in CX3CR1, the threonine to methionine mutation, shared by all affected individuals in this family, has a GERP score of 3.98; any positive GERP score >2 is considered evidence of evolutionary constraint. Threonine 280 is conserved in all mammals, the chicken, and the D. rerio species of fish, lending support to this analysis.
The presence of this SNP was validated by Sanger sequencing in the four affected individuals and in the DNA of all 13 other affected individuals and obligate heterozygotes within the family. Sanger sequencing was also performed to discover the presence of this SNP in individuals with fewer signs of DDH as well as in seemingly unaffected married-in individuals (see Fig. 3). Individuals 8, 18, 32, 38, 44, 45, and 59, who appeared to have fewer signs of DDH, were found to carry this variant. Interestingly, only one questionable child (individual 59) was found to have the disease variant, whereas in the DNA of the other adolescents who had been thought to have signs of DDH (individuals 27, 30, 31, 62, and 63), the mutation was missing. Five married in individuals (14, 21, 46, 58, 64) were also found to have this mutation in their DNA, supporting the fact that the prevalence of this allele is very high in the indigenous Utah population from which this family originates. Analysis of genomewide SNPs did not show any evidence of high levels of homozygosity in the genomes of the family members, which would have been suggestive of inbreeding, and none of the spouses appeared to be related by SNP analysis.
The incomplete penetrance of DDH seen in some members of this family is understandable in light of the fact that DDH is a complex disorder with multiple genetic, epigenetic, and environmental causes. Indeed, the allele frequency of this relatively common variant varies from 3% in the Han Chinese to 8% in whites (Caucasians) and is higher in other populations, including this one (Utah residents with Northern and Western European ancestry), whose allele frequency is 0.173, one of the highest recorded in the United States. As found in most other multifactorial disorders, a number of individuals including the obligate heterozygotes and some “married-in” individuals in this family carry this allele but manifest no signs of DDH. This supports the idea that in this family this variant may be necessary but not sufficient by itself to cause the disorder. We are vigorously pursuing other etiologic factors that might explain the severity of DDH in this family and plan to address this issue at a later point in this continuing study.
Epigenetic factors that may cause altered gene expression of CX3CR1 include the histone modifications H3k4Me1 and Me2. These methylation sites, which are often located near regulatory elements, span this variant sequence in seven cell lines. This DNA site also includes a binding site for the transcription factor Jun D in the K562 cell line. Jun D is required for bone morphogenic protein 4 (BMP4)-induced hematopoiesis in Xenopus. The significance of the presence of a PABPC1 spanning this DNA segment is not clear.
Although this study appears to have identified a significant genetic risk factor shared by all affected members of this family, it is not known how prevalent this variant will be in the overall DDH patient population. We are in the process of validating this mutation in our DNA databank of over 30 dysplastics. Additionally, exome sequencing does not detect mutations that could exist in intergenic regions within the chromosome 3 candidate region. Important regulatory sequences that affect gene expression could reside there. Also not known is whether the association of the CX3CR1 variant with DDH susceptibility will prove to be an etiologic factor. In vitro and in vivo experiments are currently underway to answer this question. Finally, pleiotropy with age-related macular degeneration (AMD) and cardiovascular disease cannot be ruled out or confirmed in this family.
This receptor, CX3CR1, which is involved in cell adhesion and migration, is known to play a role in chondrocyte maturation, being expressed at higher levels in mesenchymal stem cells and becoming downregulated during the process of chondrocyte development. Recently in human mesenchymal stem cells that were induced to differentiate to chondrocytes, Djouad and colleagues observed that CX3CR1 was present at higher levels in the stem cell and not expressed in the differentiated chondrocyte. Additionally, Cristino and colleagues, using a three-dimensional hyaluronic acid scaffold, have found that the chemokine receptors CXCR4 and CXCR5 were modulated during in vitro chondrogenic differentiation, suggesting a role for the CXC class of chemokines in the differentiation and maturation of a cartilage-like structure in vitro. The ability of mesenchymal cells to migrate and to receive regulatory signals from their environment via chemokine receptors may play a role in the developing structure of the cartilage anlage of the acetabulum.
The specific variant found in the DNA of all affected family members in this study is known to have significant biologic effects in a number of organ systems, all related to CX3Cr1's capacity to regulate cell migration or adhesion. In 2007, Combadière and colleagues found that homozygosity for the CX3CR1 M280 allele was consistently more frequent in AMD patients compared to controls, observing that impaired migration of microglial cells occurred in affected patients. They also found that chemotaxis of monocytes from individuals with homozygous M280 was impaired in the presence of bound Cx3CR1. Cx3CR1 is a human immunodeficiency virus (HIV) coreceptor as well as a leukocyte chemotactic/adhesion receptor for fractalkine and is overexpressed in the lymph nodes of HIV patients. Individuals with the M280 variant are more susceptible to HIV infection resulting from compromised migration of immune cells. Finally, in cardiovascular disease, the M280 variant provided a protective effect against carotid and coronary artery disease and stroke by decreasing cellular adhesion.
All severely and moderately affected individuals were heterozygous for the M280 variant with a 50% complement of unmutated CX3CR1. The effect of this possibly-decreased amount of fully functional receptor on the development of the acetabular labrum is unknown and is currently being investigated in animal models. The lack of effect of the variant on other organ systems in which it is present could be a result of a lack of sensitivity during a particular phase of development and the fact that they can function within their biological limits on a possibly lower concentration of this protein. Additionally, the unique morphology of the developing human hip might account for the tissue-specific phenotypic effect of this heterozygous mutant.
To explain the severity DDH in some individuals but not in others, we hypothesized that severely affected individuals might have a second aggravating mutation. To determine which variants these might be, we searched throughout the human exome for variants shared by four severely affected individuals that caused various kinds of mutations (see Subjects and Methods) and that resided in pathways related to bone formation. The variants described in Table 2 map outside the chromosome 3 candidate region and for this reason are unlikely to be shared by all affecteds and by obligate heterozygotes because linkage analysis has shown that there is likely to be only one co-inherited genetic locus. One rare variant (rs77857664 allele frequency 0.361%), a T to C transition, was found in intron 1 of BMP8B, a gene known to be involved in development of the skeletal system. OXCT2 is a relatively small gene located entirely within intron 1 of BMP8B and has a function unrelated to bone metabolism. This T to C transition, which also causes a glutamic acid to glycine amino acid change in the OXCT2 protein, is located in a transcriptionally active region, one with possible regulatory elements, as shown by a DNase I hypersensitivity cluster and an H3k27 acetylation mark found within this region by the ENCODE consortium. SIFT analysis predicts this variant to have the potential to increase reproductive fitness.
Three of the observed variants shared by four severely affected family members involve insertion or deletions of repetitive amino acid(s). One of these in-frame variants in the transcription factor E2F4, on chromosome 16q22.1, is a novel mutation involving the insertion of the nucleotide triplet CAG in a highly conserved DNA sequence. This transcription factor participates in the transforming growth factor-beta (TGF-beta) superfamily mediated signaling pathway, and deficiency of the E2F4 transcript leads to abnormal turbinate bone formation. Another variant that causes an in-frame insertion of two leucines is latent transforming growth factor beta binding protein (LTBP3) on chromosome 11q12. This protein, which forms a complex with TGF-beta, may be involved in their subcellular localization. Dabovic and colleagues created an Ltbp3-null mutation in the mouse by gene targeting. Mice homozygous for the mutation developed craniofacial malformations by day 10. At 2 months, there was a pronounced rounding of the cranial vault, extension of the mandible beyond the maxilla, and kyphosis. Between 6 and 9 months of age, mutant mice also developed osteosclerosis and osteoarthritis. The pathologic changes were consistent with perturbed TGF-beta signaling in the skull and long bones. The third variant causes an in-frame deletion of glutamine in retinoic acid inducible gene 1 (RAI 1) on chromosome 17p11.2. Inactivation of RAI1 in the mouse recapitulates phenotypes seen in Smith Magenis syndrome. Bone-related signs of this syndrome are deficient rib and nose cartilage formation.
The last two variants shared by four severely affected family members are potentially more disruptive. A deleted G in exon 2 of the RAI1 gene causes a frameshift mutation at position 280 in the protein. Two individuals (3 and 9) are homozygous for this deletion. Finally, the last variant found is a novel mutation, a deletion of a 12-bp segment in the promoter of the TMEM99 gene on chromosome 17q21.2. This segment of chromosome 17 located within exon 6 of the KRT10 gene co-encodes the repetitive amino acid motif glycine-histidine in the protein sequence of connective tissue protein, Keratin 10. Individual 1 is homozygous for this mutation the other three family members are heterozygous. Although the function of TMEM99 is unknown, it (and many other members of the TMEM gene family) are strongly expressed in cartilage. This deleted promoter sequence encodes at least three predicted CpG methylation sites. These CpG ”islands” are often methylated in promoters of genes that are inactivated. This finding might provide insight into the mechanism of incomplete penetrance often seen in familial DDH. This TMEM promoter is also the binding site of C/EBP beta, a transcription factor known to be involved in bone regulation. Last, and interestingly, this novel promoter mutation in TMEM maps to chromosome 17q21 very near the proximal border of the candidate region of another smaller DDH family analyzed by Feldman and colleagues. Further validation of these novel mutations/variants is ongoing.
In summary, a novel role is being proposed for an existing polymorphic nucleotide that is known to be biologically significant, causing both beneficial and detrimental effects in the cardiovascular and ocular systems, respectively. Gene dosage, timing of expression, and the unique morphology of the developing hip all create the microenvironment in which this variant acts, possibly explaining the phenotype of those with deficient labrum formation seen in DDH. Variants and new mutations that may contribute to the DDH phenotype severity have been found. Along with other genetic risk factors that we expect to be found in other families (because hip formation is a complex process involving the timed interaction of many proteins), these findings might illuminate the molecular pathways affecting chondrocyte maturation and bone formation.
All authors state that they have no conflicts of interest.
Authors' roles: Study design: GF, JP, CP Study conduct: GF. Data collection: JE. Data analysis: GF, KS, ML, PF. Data interpretation: GF, JP, MD, ML. Drafting manuscript: GF, MD. Revising manuscript content: GF, JP, MD, JE, CP. Approving final version of manuscript: GF, JP, MD, PF, CP. GF takes responsibility for the integrity of the data analysis.