Genotype frequencies observed for the –1997 G→T polymorphism in our group of postmenopausal women were similar to those reported from Spain [22]. Those for the Sp1 polymorphism were closer to those in two Southern European populations [20,34] than in other Caucasians [11,16,22,35-37]. The Sp1 polymorphism is absent in Asian populations [38,39].

For the Sp1 polymorphism the highest LS BMD was found in SS homozygotes, while the influence on FN BMD was minimal but the differences did not reach significance for both sites. This trend agrees with that reported by Garnero and co-workers [40] who found SS homozygotes to have the highest and statistically significant LS BMD, but no association with FN BMD. An influence of Sp1 polymorphism on LS BMD was observed by others [11,35] and also an association with FN BMD [37]. Other investigators did not find the Sp1 polymorphism to affect BMD [21,22] or found it to be associated with an increased fracture risk at the LS, FN and other skeletal sites [17,20,36]. The increased risk of fracture at the FN has been explained by differences in FN geometry between different genotypes [41].

Our results for the –1997 G→T SNP differed from those observed in Spain [22], where the lowest BMD was found in TT, and not in GG homozygotes, in contrast to our results. As reported in the Spanish [22] and American [42] populations, strong linkage disequilibrium between the –1997 G→T and Sp1 polymorphism was also observed in our study, although it do not seem to affect BMD. It has been suggested [22] that binding of osteoblast nuclear factors may be affected by the –1997 G→T nucleotide change in the COL1A1 gene since the G allele has a higher binding capacity to single-stranded binding proteins. Polymorphisms within the promoter region act in synergy with the Sp1 SNP to affect binding capacities of transcriptional factors such as Nmp4/CIZ [43]. In a recent study the haplotype of these two polymorphisms was shown to have a greater effect on BMD than when analyzed separately, suggesting that interactions between the Sp1 and –1997 polymorphisms may affect the phenotype [42]. The binding affinity of Sp1 may be influenced by other nearby polymorphisms, including the –1997 G→T, and result in an increased expression of COL1A1 in relation to COL1A2, thus affecting bone quality [42].

At the phenotypic level, these polymorphisms can be influenced by interactions with other genes and with various environmental factors. We analyzed a common C677T polymorphism within the MTHFR gene for any correlation with BMD and found that TT homozygotes had the highest, but not statistically significant BMD at all anatomical sites, and that the C allele had a negative effect on trochanter BMD. This contrasts with what was reported on Japanese women where the T allele was associated with low BMD [27]. However, in a Danish study the T allele was associated with a decreased risk of fractures independently of BMD [12]. A degree of genotypic and allelic variations exist among ethnic groups such that there is a higher prevalence of TT homozygotes in southern European populations, Hispanics and Mexicans [44]. The frequency of TT homozygotes we observed is similar to that reported for Spanish, French [44] and Portuguese [45] populations, but lower than that for Italians [44] and Japanese [27], and higher than that for northern Europeans [12,44].

One reason for these conflicting results may be differences in folate intake in these populations. Recently, TT homozygotes were found to be associated with low BMD only in the presence of low plasma folate concentration, whereas the highest femoral and trochanter BMD were found when plasma folate concentrations were above 4 ng/mL [24]. It has been suggested that the north-south increase in the prevalence of TT homozygotes may be influenced by the higher folic acid content of the Mediterranean diet [46]. Low plasma folate may result in hyperhomocysteinemia, independently of the C677T polymorphism [47]. Levels of riboflavin and vitamins B_{12 and B6 also modify the effect of the MTHFR TT genotype on BMD in the Danish population, although it was estimated that only 2% of the population will benefit from vitamin B supplementation [48].}

The C677T SNP may also be influenced by nearby polymorphisms that are in linkage disequilibrium with the C677T SNP, an example being the 1298A→C that is known to affect plasma homocysteine levels [45]. Different phenotypic expression of this polymorphism might be affected by other variants within nearby genes on the same candidate region 1p36 [13,9], as well as by genes coding for enzymes involved in the same metabolic pathway as methionine synthase reductase and cystathionine synthase. The C677T polymorphism in the MTHFR gene is also known to affect DNA methylation, which is a major epigenetic controller of gene expression. It has been reported that TT homozygotes had a diminished level of DNA methylation when compared to CC homozygotes only when folate levels were low [49].

We found no association between two COL1A1 gene variants and an increased risk for low BMD in Maltese postmenopausal women. It is clear that allelic heterogeneity exists between different populations, and that different genes and gene variants may be responsible for osteoporosis in different ethnic groups. Future studies should focus on gene expression and how this may affect the phenotype in relation to the environment and other epigenetic mechanisms that play important roles in gene regulation and in proteinomics.