Identification of Quantitative Trait Loci InfluencingSkeletal Architecture in Mice: Emergence of Cdh11as a Primary Candidate Gene Regulating FemoralMorphologyCharles R Farber,1,2Scott A Kelly,3Ethan Baruch,1Daniel Yu,1Kunjie Hua,3Derrick L Nehrenberg,3Fernando Pardo-Manuel de Villena ,3Ryan J Buus ,3Theodore Garland Jr.,4and Daniel Pomp3,51Center for Public Health Genomics, University of Virginia, Charlottesville, VA, USA2Departments of Medicine and Biochemistry and Molecular Genetics, University of Virginia, Charlottesville, VA, USA3Department of Genetics and Carolina Center for Genome Science, University of North Carolina, Chapel Hill, NC, USA4Department of Biology, University of California Riverside, Riverside, CA, USA5Department of Nutrition, Department of Cell and Molecular Physiology, Carolina Center for Genome Science,University of North Carolina, Chapel Hill, NC, USAABSTRACTBone strength is influenced by many properties intrinsic to bone, including its mass, geometry, and mineralization. To further advanceour understanding of the genetic basis of bone-strength-related traits, we used a large (n ¼ 815), moderately (G4) advanced intercrossline (AIL) of mice derived from a high-runner selection line (HR) and the C57BL/6J inbred strain. In total, 16 quantitative trait loci (QTLs)were identified that affected areal bone mineral density (aBMD) and femoral length and width. Four significant (p < .05) and onesuggestive (p < .10) QTLs were identified for three aBMD measurements: total body, vertebral, and femoral. A QTL on chromosome (Chr.)3 influenced all three aBMD measures, whereas the other four QTLs were unique to a single measure. A total of 10 significant and onesuggestive QTLs were identified for femoral length (FL) and two measures of femoral width, anteroposterior (AP) and mediolateral (ML).FL QTLs were distinct from loci affecting AP and ML width, and of the 7 AP QTLs, only three affected ML. A QTL on Chr. 8 that explained7.1% and 4.0% of the variance in AP and ML, respectively, was mapped to a 6-Mb region harboring 12 protein-coding genes. The patternof haplotype diversity across the QTL region and expression profiles of QTL genes suggested that of the 12, cadherin 11 (Cdh11) was mostlikely the causal gene. These findings, when combined with existing data from gene knockouts, identify Cdh11 as a strong candidategene within which genetic variation may affect bone morphology. ß 2011 American Society for Bone and Mineral Research.KEY WORDS:BONE MINERAL DENSITY; BONE MORPHOLOGY; BONE GEOMETRY; OSTEOPOROSIS; MOUSE GENETICS; ADVANCED INTERCROSS LINEIntroductionOsteoporosis is a common and complex disorder character-ized by bone fragility. Bone fragility is influenced by amyriad of factors, and of those intrinsic to bone, bone mineraldensity (BMD) and bone geometry are two of the mostimportant.(1–3)Although these traits are influenced by bothenvironmental and genetic factors, most (60% to 80%) of theirvariance is genetically based.(4)Thus, bone fragility is primarily agenetic disorder, and studies aimed at elucidating its geneticbasis are critical for the development of a comprehensiveunderstanding of osteoporosis.Over the last decade, the mouse has been used extensively toinvestigate the genetic basis of bone traits. Because of its clinicalrelevance, most studies have focused on BMD,(5,6)although otherskeletal traits, such as bone geometry, also have been subjectedto genetic analysis.(7)With regard to BMD, much of this work wassummarized recently in the reanalysis of genetic data from 11 F2crosses that placed over 150 BMD quantitative trait loci (QTLs) ona standardized mouse genetic map.(5)To gauge the usefulness ofthe mouse for the discovery of BMD genes, the authors evaluatedthe genomic overlap between human BMD genome-wideassociations (GWAs) and newly positioned mouse QTLs. Of the28 human GWAs identified at the time, 26 overlapped withORIGINAL ARTICLEJJBMRReceived in original form January 14, 2011; revised form April 11, 2011; accepted May 17, 2011. Published online June 2, 2011.Address correspondence to: Charles R Farber, PhD, Center for Public Health Genomics, PO Box 800717, University of Virginia, Charlottesville, VA 22908, USA.E-mail: [email protected] of Bone and Mineral Research, Vol. 26, No. 9, September 2011, pp 2174–2183DOI: 10.1002/jbmr.436ß 2011 American Society for Bone and Mineral Research2174mouse BMD QTLs. These data suggest that there may be asignificant overlap in genes harboring natural variation thatperturb skeletal development and maintenance in humans andmice. In addition, human GWA studies that include upwards of20,000 subjects have only been able to explain approximately 3%of the genetic variance for BMD,(8)and there are a number ofdifficulties inherent in assessing all traits that contribute to bonefragility in human populations. Therefore, it is likely that mousegenetics has much to contribute with regard to the discovery ofbone fragility genes.In this study, we used a G4advanced intercross line (AIL)to identify QTLs that modulate skeletal architecture. This G4population originated from a reciprocal cross between mice witha genetic propensity for increased voluntary exercise [high-runner (HR) line] and the C57BL/6J (B6) inbred strain.(10,14)Ouranalysis revealed a complex genetic architecture for both arealBMD (aBMD) and femoral morphology. Furthermore, as a steptoward gene discovery, we investigated a QTL that affects femurwidth on chromosome (Chr.) 8 in more detail. This locus was themost statistically significant and possessed the smallest 1-LOD(LOG of odds) drop confidence interval, a region harboring only12 genes. The combination of gene expression data and ananalysis of identity by descent (IBD) suggested that of the12 genes, cadherin 11 (Cdh11) was the most likely candidate.These results increase our understanding of the geneticinfluences on skeletal architecture and suggest that Cdh11 isinvolved in the regulation of femur morphology.Materials and MethodsMethods relevant to the creation, phenotyping, and genotypingof the mapping population used here have been describedpreviously.(10,14)Additionally, a complete list of the final set ofsingle-nucleotide polymorphisms (SNPs; n ¼ 530) used for theQTL analyses can be found elsewhere.(10,14)SNP locations arefrom Mouse Build 36 of the Mouse Diversity Genotyping Array(http://cgd.jax.org/tools/diversityarray.shtml). Only methods rel-evant to the current phenotypes and statistical