CNV is a type of recently increasingly being recognized DNA variation associated with human complex diseases. CNV presumably alters gene dosage, disrupts coding sequences, or exerts long-range positional effects on the gene expression pattern outside the CNV region, and consequently leads to phenotype variation. We performed the first genome-wide CNV association study in a large Chinese sample for an important bone strength phenotype, hip BS, and a follow-up replicate study in 2,286 unrelated US Caucasians sample, and identified that CNP267 region may be associated with hip BS. We also tried to find if there are any BS candidate genes in the regions of significant CNPs in initial Chinese sample. Although there were no well known BS candidate genes, such as VDR, ER, COL1A2, CYP17, PTH et al., we suggested that these genes should deserve more attention in further studies and presented the results in Table S2
CNP267 was located in 2q12.2 with physical position from 106,247,145 bp to 106,251,789 bp, where no important candidate gene has known function directly associated with osteoporosis or bone. However in the downstream of CNP267 there is an important candidate gene, the four and a half LIM domains 2 (FHL2), which plays important roles in bone metabolism by binding to several bone formation regulator. FHL2 is strongly expressed in all types of human osteoblasts 
. FHL2 can bind to several regulation factors that are important during bone formation. For example, it binds to insulin-like growth factor-binding protein 5 (IGFBP-5) 
, the most abundant IGFBP stored in bone, which plays an important role in regulating bone formation 
. FHL2 also increases transcriptional activity of androgen receptor (AR) 
, which is important in regulation of bone formation and resorption 
. In addition, FHL2 can also influence bone formation as one of modulators of the Wnt signaling pathway 
and TNF receptor–associated factor 6 (TRAF6)-mediated receptor activator of NF-κB (RANK) signaling pathway 
. Gene knockdown studies showed that FHL2-deficient mice have decreased BMD and BS due to decreased osteoblast activity 
. Conversely, enforced expression of FHL2 boosts mineralization in cell culture and increased bone mass in transgenic mice 
Genes located outside of CNV intervals can show disrupted expression patterns by long-range position effects. For example, transcription levels of some genes located in the diploid flanking sequence outside deletion region were altered in patients carrying the Williams-Beuren syndrome deletion 
. CNV17p11.2 can change the expression of some flanking genes, and thus cause Smith-Magenis syndrome 
. Whole genome association studies between CNVs and gene expression levels showed that more than half of identified genes significantly associated with CNVs were not located in the CNV intervals 
. These observations may change the traditional view, i.e., CNVs regulate gene expression via directly altering copy number of dosage-sensitive genes located in CNV intervals. However CNVs may alter the expression of genes located outside of CNV intervals by disrupting local chromatin environment and/or gene regulatory elements such as enhancers, silencers, and insulators. Previous studies reported that the relative expression levels of genes were modified through changes in copy numbers as far as 2–7 Mb away from the breakpoints 
In association study, the inconsistence of association directions between initial and replicate association studies is usually observed. There may be several reasons to address the opposite effect of CN on BS variation. First, it is well known that the same risk locus with diverse genetic background may generate different effects 
. Studies also showed that osteoporosis related phenotypes are clearly under ethnic specific genetic determination 
. Chinese and Caucasians are two evidently different populations in genetic background. The comparison of frequencies of CNs in the two samples shows significantly different with P<0.01 in chi-square test. The ethnic difference may lead to the reverse effect of association in our study. Second, in most situations, significant association suggests strong linkage disequilibrium between marker and causal locus. If the causal loci are different in two samples, it is possible to cause discordant direction of association. For example, the risk allele of marker is negatively correlated with a risk allele at one causal locus in one population, but positively associated with a protective allele at another causal locus in another population. Czarnomska et al. also reported such observation, i.e., a set of genes control the impact of the ApcMin mutation in two organs but with opposite effects 
. Finally, environmental covariates may influence the effect direction of gene variants. Eder et al. discovered opposite effects of CD14/−260 on serum IgE levels in children raised in different environments 
. Reneland et al. found that rs1498608 in PDE4D gene showed an opposite relationship with BMD variation, which indicating that the variant's effect may be context-dependent 
. The two samples come from different geographic regions, suggesting obviously different environment such as living condition and dietary habit. These differences may also influence the effect direction of the allele.
In this study, FHL2 may be recognized as an important candidate gene involved in the association between CNP267 and hip BS by the potential mechanism that CNP267 may influence the expression of FHL2 by causing changes in local chromatin structure or alterations in enhancer activity. This potential mechanism is waiting for being confirmed by functional studies.
In conclusion, this is the first genome-wide CNV association study for hip BS in Chinese Han sample. CNP267 was significantly associated with hip BS in both Chinese and Caucasian. Our findings suggest that CNP267 region may be associated with hip BS.