1. Introduction
Stress fracture (SF) is a prevalent overuse injury that usually affects bones of the lower extremities in physically active individuals, especially among athletes and military recruits of combat units during basic and advanced training (Hod et al., Reference Hod, Ashkenazi, Levi, Fire, Drori, Cohen, Bernstine and Horne2006; Murray et al., Reference Murray, Reeder, Udermann and Pettitt2006) SF is a multifactorial disorder that in all likelihood results from the combined effects of environmental factors in genetically susceptible individuals. Numerous SF risk factors have previously been reported: mechanical [e.g., training intensity and errors (Jones et al., Reference Jones, Bovee, Harris and Cowan1993), hip joint anatomy (Giladi et al., Reference Giladi, Milgrom, Simkin and Danon1991), tibia width (Giladi et al., Reference Giladi, Milgrom, Simkin, Stein, Kashtan, Margulies, Rand, Chisin, Steinberg, Aharonson, Kedem and Frankel1987), tibial cortical thickness and femoral neck diameter (Cosman et al., Reference Cosman, Ruffing, Zion, Uhorchak, Ralston, Tendy, McGuigan, Lindsay and Nieves2013)]; environmental [e.g., shoes (Schwellnus et al., Reference Schwellnus, Jordaan and Noakes1990) and training surface (Albisetti et al., Reference Albisetti, Perugia, De Bartolomeo, Tagliabue, Camerucci and Calori2010)]; and behavioural [e.g., nutritional habits (Frusztajer et al., Reference Frusztajer, Dhuper, Warren, Brooks-Gunn and Fox1990), smoking (Altarac et al., Reference Altarac, Gardner, Popovich, Potter, Knapik and Jones2000), motivation (Hallel et al., Reference Hallel, Amit and Segal1976) and pre-military induction training intensity (Cosman et al., Reference Cosman, Ruffing, Zion, Uhorchak, Ralston, Tendy, McGuigan, Lindsay and Nieves2013)]. In the largest study published to date encompassing 583 681 US military recruits over a 10-year period, the most significant SF risk factors were older age, lower body weight, lower body mass index (BMI) and race other than African American (Knapik et al., Reference Knapik, Montain, McGraw, Grier, Ely and Jones2012). A large study that included data of 311 pelvic, femoral and knee SF in Finnish military recruits that included 102 515 person years, reported that advanced age and female gender were the most significant SF risk factors in Finnish military personnel (Mattila et al., Reference Mattila, Niva, Kiuru and Pihlajamaki2007).
Multiple lines of evidence imply that genetic predisposition may contribute to SF: SF in monozygotic twins (Singer et al., Reference Singer, Ben-Yehuda, Ben-Ezra and Zaltzman1990), in the paediatric age group (Bachmann et al., Reference Bachmann, Gaston and Hefti2011; Lee et al., Reference Lee, Baek, Han and Park2005), the high rate (10·6%) of SF recurrence within 1 year at different anatomical sites (Giladi et al., Reference Giladi, Milgrom, Kashtan, Stein, Chisin and Dizian1986) and familial clustering of SF or bone diseases in American military personnel (Friedl et al., Reference Friedl, Nuovo, Patience and Dettori1992) and Israeli soldiers with SF (Givon et al., Reference Givon, Friedman, Reiner, Vered, Finestone and Shemer2000). However, the genetic basis of SF and the precise genes involved in this disorder are unknown.
There are a few studies reporting limited genetic analyses of sequence variants in candidate genes in a small number of SF cases and controls. Valimaki et al. (Reference Valimaki, Alfthan, Lehmuskallio, Loyttyniemi, Sahi, Suominen and Valimaki2005) genotyped 15 Finnish soldiers with SF and compared the rates of two polymorphisms in the oestrogen receptor alpha gene – ESR1 (MIM 133430), XbaI (351A→G) and PvuII (397T→C), and to the trinucleotide CAG repeats in the androgen receptor gene – AR (MIM 313700) in these 15 cases with that of 164 non-SF controls. No differences in polymorphisms rates were noted between cases and controls. Chatzipapas et al. (Reference Chatzipapas, Boikos, Drosos, Kazakos, Tripsianis, Serbis, Stergiopoulos, Tilkeridis, Verettas and Stratakis2009) assessed the rates of four polymorphisms (FokI, BsmI, ApaI and TaqI) in the vitamin D receptor– VDR (MIM 601769) in 32 Greek individuals with SF and 32 controls, and reported a significant 2·7-fold increased SF risk associated with the FokI allele. Korvala et al. (Reference Korvala, Hartikka, Pihlajamaki, Solovieva, Ruohola, Sahi, Barral, Ott, Ala-Kokko and Mannikko2010) analysed single nucleotide polymorphism (SNPs) in eight candidate genes in 72 Finnish soldiers with femoral neck SF and reported that a specific VDR haplotype and a SNP in the calcium-sensing receptor (CaSR MIM 601199) gene were more prevalent among cases than controls (OR = 3·22, 95% CI 1·38–7·49), especially in individuals with low BMI. We have previously reported on the association between the length of the CAG repeat in the androgen receptor gene and SF risk in Israeli soldiers (Yanovich et al., Reference Yanovich, Milgrom, Friedman and Moran2011), and statistically insignificant associations between several SNPs located within candidate genes and SF in Israeli soldiers (Yanovich et al., Reference Yanovich, Friedman, Milgrom, Oberman, Freedman and Moran2012). Cosman et al. (Reference Cosman, Ruffing, Zion, Uhorchak, Ralston, Tendy, McGuigan, Lindsay and Nieves2013) reported no significant differences in the rates of SNPs and polymorphisms between US military cadets with SF (n = 43 males; n = 26 females) compared with 712 male controls and 110 female controls in the VDR, COL1A1 (MIM 120150) and estrogen receptor (ER) genes. Thus, the aim of the study was to assess the association of intragenic sequence variants in a gene-naive manner, using whole-exome capture and massive parallel sequencing in Israeli Defense Forces (IDF) combat soldiers using a two-tires discovery-validation, case–control study design.
2. Materials and methods
(i) Participant identification and recruitment
Active duty soldiers in the IDF, who were referred to the central orthopaedic clinic of the IDF for clinical examination and whole-body bone scan for symptoms compatible with SFs between 1 January 2005 and 31 December 2010, were eligible for participation. Controls were classified as having no evidence of SF after both clinical examination and bone scan, and cases were soldiers with grade 3–4 SF or multiple grade 2 SF. This study was approved by the IDF Ethics Committee and the Ministry of Health's Ethics Committee for genetic studies, according to the ethical principles for medical research involving human subjects. Each participant, after signing a written informed consent, filled a questionnaire that detailed demographic data, personal and family history of SF and bone diseases, engagement in sports, smoking and alcohol consumption habits, and additional data relevant to SF risk, as previously detailed (Yanovich et al., Reference Yanovich, Friedman, Milgrom, Oberman, Freedman and Moran2012).
(ii) Clinical and imaging evaluation
All participants were clinically evaluated by an orthopaedic surgeon whose physical examination focused on the lower limbs. Imaging included Technetium-99 m methylene diphosphonate (Tc99) whole-body bone scan as routinely practiced in the Institute of Nuclear Medicine at the Medical Services and Supply Center for IDF soldiers (Moran et al., Reference Moran, Evans and Hadad2008). The results of the bone scan were interpreted by two different expert radiologists, and based on the results of the bone scan, the soldiers were classified as either having no evidence of SF, or having grade 1–4 SF, according to the practiced criteria and protocols (Zwas et al., Reference Zwas, Elkanovitch and Frank1987). Individuals with a single grade 1 or 2 SF as well as individuals with metatarsal SF were excluded from the study.
(iii) DNA extraction
Blood (20 ml) was drawn into EDTA-containing tubes by certified phlebotomists and the DNA extraction was completed within 4 h from the blood sample collection. DNA was extracted from peripheral blood leucocytes using the PUREGENE kit by Gentra Systems Inc. (Minneapolis, MN), following the manufacturer's recommended protocol.
(iv) DNA pools
Two pooled DNA samples from Israeli soldiers were generated: cases were soldiers with high grade SF (n = 34) and controls were ethnically and age-matched soldiers with no evidence of SF (n = 60). Equimolar amounts of DNA from each of the 34 cases for the 1st DNA pool and each of the 60 for the controls DNA pool were prepared according to the method previously described for whole-exome sequencing (Flanagan et al., Reference Flanagan, Sheehan, Linder, Howard, Wang, Hoppe, Aygun, Adams, Neale and Ware2013). Each of the pools was run on 2% agarose to assess the quality of the genomic DNA; degraded samples were removed and replaced by a non-degraded sample.
(v) Genetic analyses outline
(a) Discovery phase genotyping
The two DNA pools, the cases (n = 34) and the controls (n = 60) were initially subjected to whole-exome capture and massive parallel sequencing (see below).
(b) Validation phase
For the initial validation of the SNPs emerging from the pooled analysis, 32 SF subjects and 55 controls, all from the initial pooled genotyped participants, were subsequently individually genotyped using the Sequenom™ assay (see below). As an additional validation, 104 SF cases and 72 controls, who were similarly identified and recruited but were not genotyped in the Whole exome sequencing (WES) phase, were individually genotyped for the same sequence variants that emerged as significant from the pooled WES phase by applying the Sequenom™ assay. The flow of genotyping and filtering is shown in Fig. 1.
Fig. 1. Schematic view of the pipeline analysis.
(vi) Capture array
NimbleGen 2.1M Human Exome v1.0 Sequence Capture array (Roche NimbleGen, Inc. Madison, WI) was used to target the entire human exome for capturing.
(vii) Next generation sequencing
The capture sequences were subsequently subjected to massive parallel sequencing using the HiSeq2000 apparatus (Illumina, San Diego, CA).
(viii) Sequenom™ genotyping technology
For validation of a selected number of SNPs for each of the study participants, the Sequenom™ technology was applied, using the primer design and methodology as previously described (Jurinke et al., Reference Jurinke, van den Boom, Cantor and Koster2002).
(ix) Haplotype analysis
The haplotype association test between cases and controls was performed using Haploview 4.2 (http://www.broadinstitute.org/scientificcommunity/science/programs/medical-and-population-genetics/haploview/haploview).
(x) Ingenuity pathway analysis
We implemented the ingenuity pathway analysis (IPA) (www.ingenuity.com) to identify and classify networks, pathways, biological processes and molecular functions of the genes that emerged from the current study as putatively involved in SF predisposition.
(xi) Statistical analysis
Analyses were carried out by ‘Wiki-based Automated Sequence Processor (WASP)’ (Dubin et al., Reference Dubin, Jing, O'Broin, Calder, McLellan, Moskowitz, Suzuki and Greally2010), of sequence data first generate descriptive statistics to assess quality. Sequence variants that passed the QC threshold, were then aligned to the human reference genome using the Burrows–Wheeler alignment (BWA; Li & Durbin, Reference Li and Durbin2009; Li & Durbin, Reference Li and Durbin2010), Polymerase Chain Reaction (PCR) duplicates eliminated with Picard MarkDuplicates (http://picard.sourceforge.net) and local re-alignment of indel regions achieved with Genome Analysis Toolkit (GATK; McKenna et al., Reference McKenna, Hanna, Banks, Sivachenko, Cibulskis, Kernytsky, Garimella, Altshuler, Gabriel, Daly and DePristo2010; DePristo et al., Reference DePristo, Banks, Poplin, Garimella, Maguire, Hartl, Philippakis, del Angel, Rivas, Hanna, McKenna, Fennell, Kernytsky, Sivachenko, Cibulskis, Gabriel, Altshuler and Daly2011). Determination of population frequency (HapMap3 and 1000 genomes) and prediction of functional impact by BLOSUM62, Polyphen2 and SIFT (Sunyaev et al., Reference Sunyaev, Ramensky, Koch, Lathe, Kondrashov and Bork2001; Ng & Henikoff, Reference Ng and Henikoff2002; Eddy, Reference Eddy2004; Adzhubei et al., Reference Adzhubei, Schmidt, Peshkin, Ramensky, Gerasimova, Bork, Kondrashov and Sunayev2010) were implemented as part of the analysis pipeline. The mpileup option within Samtools (align sequences from multiple files) (Li et al., Reference Li, Handsaker, Wysoker, Fennell, Ruan, Homer, Marth, Abecasis and Durbin2009) and custom scripts that were developed at the Albert Einstein College of Medicine (Bronx, NY) were used to determine the average read depth per exon.
The raw data from the sequencing pipeline were uploaded to the GeneSifter engine (www.geospiza.com) where a variants list was generated, and chromosome, position, dbSNP, variant, type change, description, read depth, score, gene and function were provided. A list of candidate sequence variants based on the above outlined scheme was generated (Fig. 1).
Case–control analysis was performed using JMP Genomics 6.0 and probability of genotype, allele, dominant and recessive was calculated. False discovery rate (FDR) (Q value of the R package) adjustment for multiple comparisons was implemented at the initial screen analysis followed by the chi-square test of the allele distribution in the DNA pool. Finally, we applied the algorithm suggested by Mirina et al. (Reference Mirina, Atzmon, Ye and Bergman2012) to correct for any potential bias that stem from the gene size, for all variants that have had at least 600× coverage.
3. Results
(i) Participants’ characteristics
Overall, 161 male soldiers age ranging between 18 and 23 years (mean 20·1 ± 0·9 years), with grade 3–4 SF or multiple grade 2 SF were assigned ‘case’ status, and 147 male soldiers age ranging between 18 and 25 years (mean 20·1 ± 1·1 years), were classified as having no evidence of SF and were assigned ‘control’ status. Most cases (83%) and controls (77%) were within 4 months of recruitment and initiation of their basic military training. There were no statistically significant differences in anthropometric measures, physical activity habits prior to enrolment, smoking habits and alcohol consumption between the two study groups, as previously detailed (Yanovich et al., Reference Yanovich, Friedman, Milgrom, Oberman, Freedman and Moran2012). The ethnic origin of the study population consisted of 45·1% (139/308) Ashkenazi, 41·2% (127/308) non-Ashkenazi, and in 13·6% (42/308) ethnicity could not be ascertained as a result of refusal of the participants to answer this specific question. In the SF group, 105 participants (65·2%) were diagnosed with multiple SFs: 52·3% (55/105) suffered from bilateral tibial fractures, and 15·2% (16/105) had bilateral femoral fractures. Among the participants who sustained single SF (n = 54), 25 (46·2%) had tibial SF, 22 (40·7%) had femoral SF, and in seven (12·9%) SF was located at the fibula.
(ii) Genetic analyses – genotyping data
(a) Discovery phase genotyping
In the initial screen, 202 406 and 144 217 sequence variants (case and control pools, respectively) were generated. Following initial QC filtering, a list of 67 408 variants with 3–20 000 reads/variant was generated. Of these, a list of 1174 variants with more than 600 reads/variant (allowing 10× coverage for each individual in the largest pool) in both DNA pools were used for comparison between cases and controls. Allele frequencies in each group (cases and controls) during this phase were corrected for the number of alleles in each pool divided by the pool size (i.e., number of individuals in each pool). In total, 146 sequence variants showed statistically significant (P < 0·05) differences in rates between cases and controls. Correction for FDR of these 1174 variants of the initial screened pools, revealed 14 sequence variants that had maintained a statistically significant differential rates between cases and controls (Figure 2, Supplemental Table 1). In order to be as inclusive as possible, the list of 146 candidate sequence variants (including the 14 that survived the FDR correction) was used for designing 11 multiplexes for Sequenom™ assays (Supplemental Table 1) that were further used for individual genotyping as a first phase validation of subjects from the initial screened DNA pools.
Fig. 2. Manhattan plot shows that functional sequence variants are distributed across all chromosomes in the pooled groups. Results are plotted as negative log10 transformed P-values from a genotypic association chi test using allele frequency differences. The red line highlighted the FDR significant threshold.
(b) Initial validation
This step of validation of the initial case–control analysis resulted in 15 sequence variants in 15 genes that displayed either genotypic or allelic nominal significant (P < 0·05) different rates between cases and controls. Five of these 15 sequence variants in five genes (rs7426114, rs4073918, rs3801369, rs716745 and rs3752135) maintained statistical significance after FDR correction (Supplemental Table 2).
(c) Independent validation phase
Twelve sequence variants of the 146 initial screened were validated in the second independent group of cases and controls, and two variants in two genes, maintained statistical significance after FDR correction (Supplemental Table 3).
(d) Combined analysis step
The two genotyping steps of the 146 candidate sequence variants were combined and re-analysed. Following this analysis, 20 sequence variants in 18 genes (Table 1) displayed a nominal significant (P < 0·05) threshold in one of the models used for analysis (additive, allele, dominant and recessive). Of these, six variants in six genes maintained significance after FDR correction (rs7426114, rs2515941, rs4073918, rs716745, rs3752135 and rs2071856; Table 1). Notably, one sequence variant (rs4073918) in the SLC6A18 gene consistently displayed significantly different rates between SF cases and their controls in all stages of analysis (initial pool, each of the validation sets and the combined analysis steps) and significance was maintained after FDR correction. In addition, two genes (CR1 and OR10H3) harboured two candidate sequence variants each (rs3811381, rs2296160 and rs11670007, rs11670326, respectively) that were validated in the combined analysis step (Table 1). Subsequent haplotype analysis showed that the GG haplotype in the CR1 gene was overrepresented among cases compared with controls (P = 0·008) (Supplement Table 4).
Table 1. Significant variants
M, missense; S, synonymous; MA, minor allele; MAF, minor allele frequency.
(e) Pathway analysis
Employing IPA (http://www.ingenuity.com) for the six candidate genes that maintained significance in the combined analysis after FDR (NEB, GRK4, SLC6A18, LRRC55, SIGLEC12 and ELFN2) several major pathways emerged: cell development, morphology, survival and death, cell-to-cell signalling and interaction, humoral immune response, inflammatory response, nervous system development and function, and tissue development (Fig. 3 a). Functional (canonical) analysis of these candidate genes was done; they are primarily involved in cell communication, physiology and proliferation (Fig. 3 b). Although SLC6A18 was the only gene harbouring a sequence variant that consistently showed different rates at all analyses stages, it did not classify to within any of the above-mentioned networks and pathways. IPA analysis highlighted the role played by the UBC gene in pathway associations with the above-listed genes (NEB, GRK4, ELFN2 and CR1). Noteworthy, another algorithm [The UniProt Gene Ontology (GO) annotation program (http://www.ebi.ac.uk/GOA/)] failed to reproduce these network and pathway associations. However, the GO analysis highlighted the role that the NEB gene may play in determining SF phenotype based on its function, process and pathway involvement (Supplemental Table 5). While IPA identifies the functional networks and physiologic processes resulting while highlighting the gene–gene interaction, the GO analysis focuses on the gene product properties based on cross-species functional studies.
Fig. 3. Computational analysis of the genes that differed significantly among cases (SF) and controls, using two complimentary approaches: (a) IPA Network analysis. Genes are represented as nodes; solid and hatched lines depict direct and indirect interactions, respectively. Human gene functions are colour-coded as follows: Red, unknown; Blue, kinase; Yellow, enzyme; Beige, transmembrane receptor; Light blue, peptidase; Green, transporter; Orange, mphosphatase, Light green, complex; Magenta, chemical endogenous mammalian; Purple, transcription regulator; Brown, G-protein-coupled receptor; and (b) functional (canonical) analysis of candidate genes, ordered according to the Fisher's test.
4. Discussion
The present study identified several novel candidate genes that are seemingly involved in SF predisposition and pathogenesis: GRK4 (OMIM# 137026), a guanine nucleotide-binding protein whose function has been associated with hypertension (Allayee et al., Reference Allayee, de Bruin, Dominguez, Cheng, Ipp, Cantor, Krass, Keulen, Aouizerat, Lusis and Rotter2001) and activation of kidney dopamine receptor signalling cascades in human proximal tubule cells (Villar et al., Reference Villar, Jones, Armando, Palmes-Saloma, Yu, Pascua, Keever, Arnaldo, Wang, Luo, Felder and Jose2009). Nebulin (NEB-OMIM#161650), a muscle protein of the cytoskeletal matrix; SLC6A18 (OMIM# 610300), specific transporter for neurotransmitters; LRRC55 (OMIM# 615213), a protein that regulates voltage- and calcium-activated potassium channels involved in contractile tone of smooth muscles (Yan & Aldrich, Reference Yan and Aldrich2012). SIGLEC12 (OMIM# 606094) a putative adhesion molecule that mediates osteoclast differentiation (Takahata et al., Reference Takahata, Iwasaki, Nakagawa, Abe, Watanabe, Ito, Majima and Minami2007) and ELFN2 (UniProtKB: Q5R3F8) whose function is to inhibit phosphatase activity of protein phosphatase 1 (PP1), a complex that enhances bone accumulation (Zhang et al., Reference Zhang, Qiu, Wu, Wan, Shi, Wang, Chen, Wan, Clemens and Cao2009). Pathway analysis of these genes revealed that they are primarily involved in cell development, morphology, survival and death, cell-to-cell signalling and interaction, humoral immune response, inflammatory response, nervous system development and function, and tissue development. Furthermore, pathway analyses highlighted the role that the NEB and the UBC genes may play in SF pathogenesis. While the NEB gene may plausibly and mechanistically be associated with SF pathogenesis, given its role in muscle cytoskeleton matrix, the UBC gene (MIM # 191340) that encodes for a ubiquitin is associated with cell–cell adhesion, DNA damage repair, intracellular degradation and other processes, may be less obvious in its association with SF pathogenesis. Only subsequent validation of the contribution of these genes and pathways in additional SF cases as well as functional assessment of their contribution to the phenotype in animal model may unravel the mechanisms by which these genes and pathways may lead to SF. Noteworthy, some of these genes and pathways have not been considered ‘classical SF candidate genes’ and have never been reported to be analysed as candidate SF genes. The few previous studies on the genetic predisposition to SFs were all focused on seemingly logical candidate genes: genes involved in bone formation and bone-associated disorders. Despite several reports on a significant statistical association between SF phenotype and a genetic variant [e.g., CAG repeat in the androgen receptor gene (Valimaki et al., Reference Valimaki, Alfthan, Lehmuskallio, Loyttyniemi, Sahi, Suominen and Valimaki2005), VDR polymorphisms (Chatzipapas et al., Reference Chatzipapas, Boikos, Drosos, Kazakos, Tripsianis, Serbis, Stergiopoulos, Tilkeridis, Verettas and Stratakis2009)] all studies were notably based on a limited number of cases and have analysed a rather limited number of sequence variants and genes. The involvement of genes along the novel pathways that are indicated by this preliminary study may shed new light on the pathogenesis of SF. The results of this study are consistent with the notion that the genes directly involved in bone formation, metabolic and inherited bone diseases and osteoporosis contribute little, if anything, to SF pathogenesis. In support of this notion, analysis of the chromosomal location of the sequence variants found in the present study as putatively associated with SF pathogenesis (n = 146) showed that none co-localized with any of the SNPs or genes previously associated with metacarpal bone geometry (Karasik et al., Reference Karasik, Shimabuku, Zhou, Zhang, Cupples, Kiel and Demissie2008), trabecular and cortical volumetric bone mineral densities (Paternoster et al., Reference Paternoster, Ohlsson, Sayers, Vandenput, Lorentzon, Evans and Tobias2010, Reference Paternoster, Lorentzon, Lehtimaki, Eriksson, Kahonen, Raitakari, Laaksonen, Sievanen, Viikari, Lyytikainen, Mellstrom, Karlsson, Ljunggren, Grundberg, Kemp, Sayers, Nethander, Evans, Vandenput, Tobias and Ohlsson2013), osteoporosis-related phenotype (Karasik et al., Reference Karasik, Cheung, Zhou, Cupples, Kiel and Demissie2012), bone mass, hip geometry and fractures (Boudin et al., Reference Boudin, Steenackers, de Freitas, Nielsen, Andersen, Brixen, Van Hul and Piters2013; Garcia-Ibarbia et al., Reference Garcia-Ibarbia, Perez-Nunez, Olmos, Valero, Perez-Aguilar, Hernandez, Zarrabeitia, González-Macías and Riancho2013), bone mineral density, cortical bone thickness and bone strength (Zheng et al., Reference Zheng, Tobias, Duncan, Evans, Eriksson, Paternoster, Yerges-Armstrong, Lehtimaki, Bergstrom, Kahonen, Leo, Raitakari, Laaksonen, Nicholson, Viikari, Ladouceur, Lyytikainen, Medina-Gomez, Rivadeneira, Prince, Sievanen, Leslie, Mellstrom, Eisman, Moverare-Skrtic, Goltzman, Hanley, Jones, St Pourcain, Xiao, Timpson, Smith, Reid, Ring, Sambrook, Karlsson, Dennison, Kemp, Danoy, Sayers, Wilson, Nethander, McCloskey, Vandenput, Eastell, Liu, Spector, Mitchell, Streeten, Brommage, Pettersson-Kymmer, Brown, Ohlsson, Richards and Lorentzon2012), bone mineral density loci and osteoporotic fracture (Duncan et al., Reference Duncan, Danoy, Kemp, Leo, McCloskey, Nicholson, Eastell, Prince, Eisman, Jones, Sambrook, Reid, Dennison, Wark, Richards, Uitterlinden, Spector, Esapa, Cox, Brown, Thakker, Addison, Bradbury, Center, Cooper, Cremin, Estrada, Felsenberg, Gluer, Hadler, Henry, Hofman, Kotowicz, Makovey, Nguyen, Nguyen, Pasco, Pryce, Reid, Rivadeneira, Roux, Stefansson, Styrkarsdottir, Thorleifsson, Tichawangana, Evans and Brown2011; Estrada et al., Reference Estrada, Styrkarsdottir, Evangelou, Hsu, Duncan, Ntzani, Oei, Albagha, Amin, Kemp, Koller, Li, Liu, Minster, Moayyeri, Vandenput, Willner, Xiao, Yerges-Armstrong, Zheng, Alonso, Eriksson, Kammerer, Kaptoge, Leo, Thorleifsson, Wilson, Wilson, Aalto, Alen, Aragaki, Aspelund, Center, Dailiana, Duggan, Garcia, Garcia-Giralt, Giroux, Hallmans, Hocking, Husted, Jameson, Khusainovam, Kim, Kooperberg, Koromila, Kruk, Laaksonen, Lacroix, Lee, Leung, Lewis, Masi, Mencej-Bedrac, Nguyen, Nogues, Patel, Prezelj, Rose, Scollen, Siggeirsdottir, Smith, Svensson, Trompet, Trummer, van Schoor, Woo, Zhu, Balcells, Brandi, Buckley, Cheng, Christiansen, Cooper, Dedoussis, Ford, Frost, Goltzman, Gonzalez-Macias, Kahonen, Karlsson, Khusnutdinova, Koh, Kollia, Langdahl, Leslie, Lips, Ljunggren, Lorenc, Marc, Mellstrom, Obermayer-Pietsch, Olmos, Pettersson-Kymmer, Reid, Riancho, Ridker, Rousseau, Slagboom, Tang, Urreizti, Van Hul, Viikari, Zarrabeitia, Aulchenko, Castano-Betancourt, Grundberg, Herrera, Ingvarsson, Johannsdottir, Kwan, Li, Luben, Medina-Gomez, Palsson, Reppe, Rotter, Sigurdsson, van Meurs, Verlaan, Williams, Wood, Zhou, Gautvik, Pastinen, Raychaudhuri, Cauley, Chasman, Clarkm, Cummings, Danoy, Dennison, Eastell, Eisman, Gudnason, Hofman, Jackson, Jones, Jukema, Khaw, Lehtimaki, Liu, Lorentzon, McCloskey, Mitchell, Nandakumar, Nicholson, Oostra, Peacock, Pols, Prince, Raitakari, Reid, Robbins, Sambrook, Sham, Shuldiner, Tylavsky, van Duijn, Wareham, Cupples, Econs, Evans, Harris, Kung, Psaty, Reeve, Spector, Streeten, Zillikens, Thorsteinsdottir, Ohlsson, Karasik, Richards, Brown, Stefansson, Uitterlinden, Ralston, Ioannidis, Kiel and Rivadeneira2012), and femoral neck bone geometry (Zhao et al., Reference Zhao, Liu, Liu, Liu, Papasian, Sha, Pan, Guo, Wang, Yan, Xiong, Tang, Yang, Chen, Guo, Li, Shen, Zhang, Lei, Recker and Deng2010) (Supplemental Table 6), including the ones that were reviewed by Ralston & Uitterlinden (Reference Ralston and Uitterlinden2010) (Supplemental Table 7).
If the results of the present study are validated in a larger study, of ethnically diverse individuals with SF, the data can be used for several clinical applications: identifying pre-symptomatic individuals at high risk for developing SF and offering them a modified training regimen, and possibly using the up-regulated gene products as surrogate markers for diagnosing SF without using radiation-emitting modalities.
This study has several limitations that should be pointed out. It focused on Jewish Israeli active duty soldiers, and the validation of the findings in other ethnically diverse population is further stressed. Although this is the largest genetic study pertaining to SFs, the number of analysed individuals is still limited. While DNA pooling is an efficient mean of lowering the costs of whole-exome sequencing, unequal representation of all participants in the DNA pool and skewed DNA amplification may have eliminated some additional candidate regions and genes. Lastly, the molecular mechanisms leading to SF in soldiers during basic training may differ than those underlying the same phenotype in elite training athletes.
In conclusion, using whole-exome capture and sequencing, novel candidate genes and pathways have emerged as putatively contributing to SF pathogenesis and predisposition.
Supplementary material
To view supplementary material for this article, please visit http://dx.doi.org/10.1017/S001667231400007X.
This study was funded by a grant from TATRC (W81XWH-09-2-0054) to Eitan Friedman and Daniel S. Moran.
Conflict of interest statement
All authors declare that they have no conflict of interest regarding the data published.
