A comprehensive candidate gene approach identifies genetic variation associated with osteosarcoma

Osteosarcoma (OS) is the most common primary malignant bone tumor and typically occurs in adolescents and young adults (Damron et al, 2007; Mascarenhas L, 2006; Stiller CA, 2006). OS incidence has a bimodal age distribution; the primary peak occurs during adolescence and a second, much smaller peak is present in the elderly [1, 2]. In young patients, OS incidence correlates with puberty and bone growth. The peak incidence of both OS and puberty tend to occur earlier in females. OS incidence is higher in males, who usually grow taller than females, and it typically occurs at sites of rapid bone growth (e.g., the metaphyses of long bones) [1]. The incidence peak in adolescence is followed by a rapid decline and a plateau when bone growth is complete (after age 24 years) [3]. Several studies have suggested that being taller than average at diagnosis is associated with increased OS risk [4‐11]. A recent meta-analysis of height at diagnosis and birth-weight as OS risk factors found that high birth-weight (OR 1.35, 95% CI 1.01-1.79, compared to average birth-weight subjects) and being taller than average were significant OS risk factors (for those ≥90^th percentile of height: OR 2.63, 95% CI 1.98-3.49, compared to those ≤50^th percentile of height) [12]. In aggregate, these data suggest that growth and development during puberty, and possibly in utero, contributes to OS etiology.

OS frequently occurs in several cancer predisposition syndromes, including Li-Fraumeni Syndrome [13], hereditary bilateral retinoblastoma [14], Bloom, Werner, Rothmund Thomson syndromes [15], and Diamond-Blackfan anemia [16]. It also occurs more frequently in individuals with Paget's disease [17]. However, in the majority of OS cases, there is no known predisposing factor. The data regarding the role of common germline genetic variation in OS risk are sparse. Positive or suggestive associations have been observed for SNPs in the vitamin D receptor (VDR; FokI polymorphism) [8], tumor necrosis factor-alpha (TNF-a; promoter region -238 SNP) [18], insulin-like growth factor 2 receptor (IGF2R) [19], Fas [20], transforming growth factor-beta receptor 1 (TGFBR1) [21], and MDM2 [22] genes, but no associations were observed for the estrogen receptor (ER) [8], collagen I[alpha]1 (COL1A1) [8], or TP53 [23] genes.

Peak levels of endogenous sex hormones, growth hormones, and IGF-I levels occur during puberty which also corresponds to peak bone growth rates. It is possible that variation in genes important in bone development, growth, and puberty are modifiers of OS risk. In addition, insulin-like growth factors are known to play critical roles in carcinogenesis [24, 25]. Chromosomal aneuploidy in OS cells [26, 27] and the increased OS risk observed with genetic syndromes caused by mutations in DNA repair pathways [e.g., TP53 [13], WRN, BLM, RECQL4 [15]] suggests that variants in DNA repair genes may be associated with OS risk. Genes in DNA repair and tumor suppressor pathways may also contribute to OS pathogenesis, because they help maintain the integrity of critical cellular processes and defects in these genes often lead to carcinogenesis. Diamond-Blackfan anemia is associated with an increased frequency of OS and mutations in ribosomal genes (i.e., RPS19, RPS24 and RPS17) [16, 28]. Thus, it is also feasible that variation in these genes may contribute to OS risk.

There are numerous genes that contribute to bone growth and puberty, and DNA repair which could contribute to OS which have not yet been evaluated. We evaluated these hypotheses in an OS association study of candidate genes from the following pathways: growth and hormone metabolism, bone formation, tumor suppressor and DNA repair, and ribosomal. We genotyped 4836 tag-SNPs across 255 candidate genes from these four pathways in 96 OS cases and 1426 cancer-free controls. This approach identified several SNPs in candidate genes from biologically plausible pathways that were associated with OS risk.

The characteristics of the study participants are shown in Table 1. The median age of the 96 OS cases was 20.3 years (age range 8 to 80.5). The median age of the 63 orthopedic controls from the BDISO was 18.5 years (age range 7.2-68.5), and the PLCO controls were older with a median age of 62 years (age range 55-75). All participants were self-identified Caucasians and from the continental United States. Additional file 1, Table S1 shows the genes analyzed in each pathway, the number of SNPs and most significant SNP in each gene, and the gene- and pathway-level P values. There were 161 genes (2835 SNPs) in the DNA repair pathway, 62 genes (1448 SNPs) in the growth/hormone metabolism pathway, 28 genes (534 SNPs) in the bone formation pathway, and 4 genes (19 SNPs) in the ribosomal pathway. We took three approaches to the analyses by evaluating associations with OS at the individual SNP level, gene level, and pathway level. We used the conservative Bonferroni correction to correct for multiple statistical tests.

The biology of OS pathogenesis is complex and there are limited data on risk factors in the more common sporadic form of OS. Epidemiologic studies of OS suggest that growth and development play a role in etiology [2, 32, 33]. It occurs primarily in adolescents during puberty [1] when bone growth is rapid and endogenous sex hormones and growth hormones are at their highest, so variation in a gene involved in regulating sex hormones is biologically plausible. Rapidly growing tissue, such as bone during puberty, is known to be highly susceptible to carcinogenesis, possibly due to rapidly proliferating osteogenic cells being more vulnerable to DNA repair errors [4, 34]. In addition, chromosomal aneuploidy is extensive in somatic OS cells, which suggests the presence of chromosomal instability[26, 27]. The increased frequency of OS in genetic predisposition syndromes [13, 15] characterized by mutations in DNA repair pathways suggests that variants in genes involved in DNA repair are also reasonable candidates.

We evaluated the association between OS and 255 candidate genes, including 4836 SNPs, from four functional pathways (growth and hormone metabolism, bone formation, DNA repair, and ribosomal). We used 3 approaches to comprehensively evaluate these biologically plausible pathways: analyses were performed at the individual SNP level, gene level and pathway level with conservative statistical corrections for multiple testing. While no genes or pathways were significantly associated with OS after correction for multiple tests, the SNP based approach identified some potentially important candidates. A total of twelve SNPs in genes from the growth and hormone metabolism, and DNA repair pathways were significantly associated with OS risk after correction for multiple tests. Two genes had multiple significant SNPs associated with risk, FANCM and GH1, after correction for multiple tests.

FANCM contained 4 SNPs significantly associated with a similar 2-fold increased risk of OS using a dominant inheritance model, the most in any gene studied. These SNPs were not correlated in our controls. One SNP is located in exon 14 and the minor or risk allele results in a nonsynonymous change from valine to leucine (Ex14+316, Val878Leu). The minor allele of this SNP is the ancestral allele and is highly conserved among other mammalian species. The three other SNPs in FANCM were intronic. FANCM has DNA-dependent ATPase activity, promotes the dissociation of DNA triplexes, and with other Fanconi anemia-associated proteins, may repair DNA at stalled replication forks [35, 36]. DNA repair must be accurate to preserve genome stability for long-term cellular viability; genetic instability is characteristic of cancer cells, and may be due, at least in part, to mutations or variation in genes that function to ensure DNA integrity [37].

Two significant SNPs were located downstream of GH1. They appear to be highly correlated in our controls, although one SNP was associated with an increased risk of OS and the other was protective. The variant in IGF1 (rs7956547, IVS2+10605) associated with a decreased risk of OS was another interesting candidate involved in growth. The insulin-like growth factor signaling system is important in the formation and homeostasis of bone, and differential expression of IGF1 has been observed in osteosarcomas [38‐40]. IGFI expression is stimulated by growth hormone, and OS incidence peaks during puberty with the release of growth hormone. OS cells have been shown to be IGF1-dependent for growth, and inhibiting growth hormone release in mice decreased IGF1 serum levels and inhibited tumor growth [41‐43]. In addition, animal model data from dogs, which develop OS similar to human patients (similar sites, histology and treatment response) and large breeds have a 185-fold increased risk of OS compared with small dog breeds [44, 45], suggest that a SNP in IGF1 is a main determinant of small size in dogs and is virtually absent in giant breeds [46]. The data suggest that GH1 and IGF1 may play a role in the etiology of OS.

We also specifically evaluated the genes found to be associated with OS in other studies. We previously identified a SNP in IGF2R (rs998075; Ex16+88G >A) to be significantly associated with OS risk in study of variation in genes critical in growth regulation [19]. This SNP was also included in our dataset and the significant association replicated in an analysis limited to our BDISO cases and controls (P = 0.01), as expected because it is the same study population. However, the association did not replicate with the addition of our 1363 PLCO controls (P = 0.12), which suggests that the original study may have been limited by the number of controls, or possibly these older PLCO controls have a different ethnic mix within whites related to this polymorphism. Others have found significant associations between OS and polymorphisms in VDR [8], TGFBR1 [21], and MDM2 [22], which were also included in our dataset. We found no significant associations with individual SNPs within VDR or TGFBR1 before or after correction for multiple tests, although our dataset did not include the VDR FokI polymorphism [8] or TGFBR1*6A variant [21], or at the gene level. One SNP downstream of the DNA repair gene MDM2, rs1690916, was significantly associated with a decreased risk of OS after correction (OR 0.62, 95% CI 0.45-0.85, P _adj = 0.026), and 3 intronic SNPs were significantly associated with an increased risk of OS before correction (ORs 1.6-1.8, P = 0.008-0.02). The previously identified [22]MDM2 T309G (rs2279744) polymorphism was marginally non-significantly associated with an increased risk of OS (OR 1.31, 95% CI 0.97-1.77, P = 0.07). At the gene level, MDM2 was found to be significantly associated with OS (Gene P = 0.016), but not after correction for multiple tests. As others have shown [8, 23], the current study did not confirm associations between polymorphisms in ESR1/2 (including previously analyzed PvuII and XbaI polymorphisms), COL1A1, or TP53 and OS after correction, or at the gene level.

We also investigated many of the genes with germline mutations in cancer predisposition syndromes associated with an increased frequency of OS [15], including the DNA repair genes TP53 (mutated in Li-Fraumeni Syndrome), WRN (mutated in Werner syndrome), BLM (mutated in Bloom syndrome), RECQL4 (mutated in Rothmund-Thomson Syndrome), and the ribosomal genes RPS19 and RPS24 (mutated in Diamond-Blackfan anemia) [28, 47, 48]. OS also occurs more commonly in older individuals with Paget's disease[17]. Two genes involved in bone formation, SQSTM1 and TNFRSF11A, have been shown to cause Paget's disease [49]. We found that no common SNPs within these genes that predispose to cancer predisposition syndromes or Paget's disease were significantly associated with OS after correction for multiple tests.

A limitation of the current study was the small number of cases; however, the inclusion of 14 controls per case improved the statistical power to detect SNPs with strong effects. For the additive model and our 96 cases and 1426 controls, we had greater than 80% statistical power to detect an OR of 1.82 for MAFs of 0.1 and an OR of 2.15 for MAFs of 0.05 (with a baseline population risk of 0.0000001 and type 1 error of 0.05). Another potential limitation of our study was the use of controls from PLCO that were older than the BDISO case and control population. However, the PLCO controls had no history of any cancer, including osteosarcoma, they were limited to Caucasians from the continental US, as were the BDISO cases and controls, and we found no evidence of population stratification in between groups.

Strengths of the current study include the detailed genotyping of biologically plausible pathways which give a higher a priori likelihood. We used three statistical approaches to comprehensively evaluate associations with OS (at the SNP level, gene level, and pathway level). In addition, we used a stringent Bonferroni correction and conservatively interpreted our results to reduce the probability of a Type 1 error.

We have conducted the largest study of genetic variation in candidate genes associated with OS to date. We identified the presence of genetic variants in candidate genes from biologically plausible pathways important in growth and hormone metabolism, and DNA repair associated with OS, even with conservative statistical corrections. The strongest candidates are FANCM, GH1 and IGF1. The potential functional implications of the variation in these genes are currently unknown. However, since OS occurs during a period of rapid bone growth, genes important in growth, puberty, and DNA repair are biologically plausible contributors to OS pathogenesis because of their function in critical cellular processes. Larger studies of common genetic variation in OS and functional studies of the SNPs identified here are required to confirm the significance of our findings.