Introduction

Osteoporosis represents a major public health concern in aging societies, fragility fractures having the highest impact on both quality of life and health care systems worldwide [1]. Bone mineral density measurement (BMD) in hip and spine is used in the diagnosis of osteoporosis [2]. The association between low BMD and risk of bone fractures has been reported extensively [3–7]. In a recent review, advanced age, smoking, low body weight, physical limitations, and prevalent fracture (after age 50) were consistent risk factors for osteoporosis [8]. BMD declines with age, chronological age being a major determinant of BMD.

Recent studies into the mechanism of aging revealed various processes leading to replicative senescence [9]. Critically short telomere length is a marker of biological aging [10]. In addition to natural replicative attrition, the inflammatory process also enhances the rate of telomere length (TL) attrition, such that association between TL and chronic inflammatory states has been documented. Telomeres consist of short unique repeated DNA sequence (TTAGGG), and specialized proteins are located at both ends of linear chromosomes in humans. They act as the cap of the chromosome and maintain genomic integrity by preventing chromosome fusions and atypical recombination [11]. Due to the end replication problem of DNA, telomeres shorten in each cell division. Therefore, telomere length and its attrition reflect on cellular replication history and cellular senescence and is considered a new marker of biological aging, representing a new but different dimension of the aging process from chronological age. With respect to bone health, replicative aging of osteoblast precursor cells has been demonstrated and shown to be associated with telomere attrition in this cell type [12]. Furthermore, exacerbated telomere erosion is also a common feature in other diseases associated with chronic inflammation such as autoimmune and ischemic heart diseases [13, 14].

There are genetic and environmental determinants for BMD. Genetic determinants of BMD have been described extensively [15–18]. Genetic studies include candidate gene association, familial linkage, and genome-wide association (GWA) studies [19, 20]. On the other hand, smoking, immobility, and chronic inflammatory conditions have been documented as common environmental risk factors associated with osteoporosis. It is interesting to find that pro-inflammatory genes were also associated with osteoporosis in the latest GWA studies. For example, one of the predisposition genes for osteoporosis is the receptor activator of nuclear factor κB ligand (RANKL) [21]. Chronic inflammation with activated T cells induced the expression of RANKL gene and enhanced bone resorption. These results indicate a strong role of inflammatory response in bone loss [22–24].

Since both bone loss and telomere attrition have been linked to inflammation, it has been hypothesized that short telomere length is a risk factor for osteoporosis. In a small cohort of 84 elderly men, telomere length was not associated with baseline forearm BMD, but it was associated with bone loss at a borderline significance (P = 0.03) [25]. In another study of 2,150 women aged between 18 and 80, telomere length was significantly correlated with forearm and spine BMD [26]. These findings suggest that telomere length may be a predictor or marker for osteoporosis and impending bone loss in aging populations. To date, there have not been any studies in Chinese populations. As part of a territory-wide study on bone health in older people aged 65 and over, we examined the association between telomere length, baseline BMD, and its change over 4 years.

Methods

Study subjects

A community cohort of 2,000 men and 2,000 women aged 65 or above were recruited from community centers for the elderly and housing estates in Hong Kong. In the recruitment process, about one third of the volunteers were stratified to each of the age groups of 65–69, 70–74, and 75 or above. DNA samples were available and analyzed in a subgroup of 963 men and 904 women. At recruitment, participants completed a detailed questionnaire containing information on demographics, medical and drug histories, smoking status, and dietary intake [27]. The presence or absence of diseases was reported as diagnosed by their doctors. The study had been approved by the Clinical Research Ethics Committee of the Chinese University of Hong Kong, and informed consent was obtained from all participants.

Bone mineral density at the hip and femoral neck was measured using dual X-ray absorbtiometry (DXA) by Hologic QDR-4,500 W densitometers (Hologic, Waltham, MA, USA). The coefficient of variation (in vivo) in our laboratory was 0.7% at the hip [28]. Clinical osteoporosis at the hip was defined using the WHO criteria of a T-score less than −2.5. Body weight was measured with subjects wearing a light gown using the Physician Balance Beam Scale (Healthometer, Illinois, USA). Height was measured by the Holtain Harpenden standiometer (Holtain, Crosswell, UK). Body mass index (BMI) was calculated by dividing the weight in kilograms by the square of the height in meters.

Information about usual level of physical activity was also recorded using the Physical Activity Scale of the Elderly (PASE). This is a 12-item scale measuring the average number of hours per day spent in leisure, household, and occupational physical activities over the previous 7-day period. Activity weights for each item were determined based on the amount of energy expended, and each item score was calculated by multiplying the activity weight by activity daily frequency. A summary score of all the items reflect the daily physical activity level [29].

Four-year follow-up

All participants were invited for a 4-year follow up. During the follow-up examination, information on disease status and fracture history was recorded. Anthropometry and BMD were also measured. Changes in BMD were determined during the 4-year period. Seven hundred twenty-seven men and 735 women came back for follow-up.

Telomere length measurement by quantitative real-time polymerase chain reaction

The principle of the laboratory method has been previously described [27]. In short, DNA was extracted in the peripheral blood by phenol-chloroform method and stored at −80°C with concentration >100 ng/μl (whereas working samples of 20 ng/μl were prepared before analysis). Telomere length measurement followed the method published by Cawthon [30] with modification [31]. Unlike the traditional method using terminal restriction fragment (TRF) analysis, this method used the technique of quantitative real-time polymerase chain reaction (qRT-PCR), which has the advantage of high throughput, time-saving, and reproducible estimation of telomere length [30, 32–34]. Roche LightCycler 480 (Roche, Mannheim, Germany) was used to perform the qRT-PCR with primer sequences obtained from Cawthon [31]. The T/S ratio (Ct(telomere assay)/Ct(single-copy gene assay)) was used to assess the relative length of telomere, while C t is the fractional cycle number for a threshold fluorescence level to be reached during qRT-PCR. The T/S ratio (ΔΔCt) was then plotted against a standard calibration curve using samples with predetermined telomere lengths to obtain the telomere length in unit of kilobase pairs (kbp). The coefficient of variation % (CV%) of the Ct measurements of the telomere and the single-copy gene probes were 3.6% and 1.8%, respectively. Two control samples were analyzed in duplicates in each batch of assays; one sample was collected from volunteers of age of 20s representing a long telomere control (long QC, by TRF = 11.6 kbp) and another one from an elderly subject representing a short telomere control (short QC, by TRF = 8.2 kbp). Within-batch and between-batch analytical imprecisions were determined from over 20 batches of assays using these two samples. The within-batch and between-batch CV% of ΔΔCt was 11.9% and 11.2% for the long QC sample. For the short QC sample, they were 8.1% and 14.2%, respectively.

A calibration curve was drawn by using ΔΔCt values obtained from five different samples of predetermined TL by TRF method. The coefficient of determination (R 2) of the linear correlation between TL by TRF method and ΔΔCt was 0.63, which was similar to that reported (R 2 = 0.68) in the original description of the ΔΔCt method by Cawthon [22]. Calibration to TL in kbp was carried out independently for each batch which in a way partially corrected for between-batch variation in determination of ΔΔCt and TL. After calibration to TL in kbp, the within-batch and between-batch CV% of TL was 8.5% and 7.5% for the long QC sample. For the short QC sample, they were 6.3% and 6.1%, respectively. A reduction of between-batch CV% was noted which reflected a partial correction of the between-batch variation in the calibration process.

Statistical analysis

Descriptive and analytical statistics were carried out using SPSS v15.0 (SPSS, Chicago, USA). Analytical errors of ΔΔCt and TL were partitioned into within-batch and between-batch errors using a BASIC program [35]. Changes in hip BMD and TL were normally distributed, while heteroscedasticity was not demonstrated in the residual plots of regressions for BMD changes. This allows associations between telomere length and BMD at various sites to be analyzed using multivariate linear models. Covariates of TL and BMD were adjusted in the multivariate regression for both sexes, which included age, BMI, PASE score, daily dietary vitamin D intake, history of smoking, heart diseases, and fracture. In addition, age of menarche, age of menopause, numbers of children, history of breast feeding, and estrogen treatment were also included as covariates for women [36–38]. PASE scores were grouped by deciles in the multivariate analysis. P value less than 0.05 were considered as statistically significant.

Results

At baseline, 963 men and 904 women TL were measured in using real-time quantitative PCR assay. Characteristics of the subjects are shown in Table 1. Women had higher body mass index compared with men. Hip and femoral BMD were higher in men by 16% to 21% compared with women (Table 2). After dividing subjects into three 5-year interval age groups, similar percentage differences between the two sexes were confirmed (Table 3). During follow-up, 75% of men and 81% of women had returned at 4 years. Mean hip BMD loss was 0.007 g/cm2 in men and greater in women (0.016 g/cm2; P < 0.001, Table 2). Similarly, loss in femoral neck BMD was greater in women (P < 0.001). BMD loss was highest in the oldest age group (≥75 year old).

Table 1 Characteristics of the study cohort at baseline
Full size table
Table 2 BMDs of the study cohort at baseline, at 4-year follow-up, and the changes in BMD after 4 years
Full size table
Table 3 BMDs of the study cohort, separated by age groups, at baseline and the changes in BMD after 4 years
Full size table

Mean telomere length was 5.40 and 5.95 kb for men and women, respectively. Women had longer telomere than men (P < 0.001; Table 1). There was a general decreasing trend in TL with age in men, and the correlation between TL and age was statistically significant (P = 0.037; Fig. 1). There was no correlation between TL and age in women (P = 0.784).

Fig. 1
figure 1

Distribution of telomere length in each age group

Full size image

Table 4 shows the regression coefficients of the association between telomere length, other covariants, and hip and femoral neck BMD at baseline. Age and BMI showed the strongest association in both sexes. In women, age of menopause, age of menarche, estrogen treatment/replacement therapy, and history of previous fracture were also significant covariates. In addition, PASE score and dietary calcium intake were also significantly correlated with BMD at baseline. However, in these multivariate analyses, TL was not associated with BMD in either sex.

Table 4 Multiple linear regression for association between telomere length and covariants with baseline BMDs
Full size table

To examine if TL could predict bone loss, the analysis was repeated using change in BMD after 4 years (Table 5). Age was the only significant factor associated with change in hip BMD. PASE score was significantly correlated with bone loss at both total hip and femoral neck in men, while history of fracture was also associated with change in hip BMD in women. TL was not associated with change in BMD in either site in men or women.

Table 5 Multiple linear regressions for association between telomere length and covariants with changes in BMDs after 4 years
Full size table

To further explore the association between baseline or loss of BMD and TL, a subgroup analysis was performed. In order to minimize the effect of chronological age, subjects were stratified by sex and age groups of 5 years (65–69, 70–74, and 75 and above). The effect of TL and covariants were analyzed by linear regression in each of these three different age groups. Only a weak association between TL and baseline hip and femoral neck BMD was found in men between 70 and 74 years (Table 6). However, TL only accounts for up to 1.6% of the variance of baseline BMD. The effect of TL on the risk of osteoporosis was explored by logistic regressions with correction for confounding covariates. In both whole sample and age group stratified analyses, TL was not a significant risk factor for osteoporosis of the hip in either sex (Table 7).

Table 6 Multiple linear regressions for association between BMDs and tertiled telomere length groups in each of the three age groups
Full size table
Table 7 Logistic regressions for association between clinical osteoporosis at the hip and tertiled telomere length groups in each of the three age groups
Full size table

Discussion

Two recent clinical studies suggest that leukocyte telomere length could be a marker of biological aging of bone and bone loss [25, 26], and the need for studies with large sample size and prospective design was pointed out. Our report attempted to continue to address this research question using large numbers of men and women examining changes in BMD prospectively. No association between TL and baseline BMD or changes in BMD was observed, although associations between bone loss and other well-known factors such as age, BMI, physical activity, calcium intake, age of menopause, and estrogen use were observed. This conclusion may appear different from two previous studies [25, 26]. However, many factors should be considered in the comparison between this and previous studies. Firstly, the mean age of this cohort was approximately 72 years, compared with the age range of 71–86 years for men in one study [25] and 48 years for women in the other, although one third of women were postmenopausal [26]. In examining association between TL and BMD, large numbers of subjects over a wide age range would be preferable. Our cohort did not include young or middle-aged subjects. This fact may also explain the lack of association between TL and age in our study. Secondly, bone loss may vary among different sites, as a result of factors other than biological aging, such as mechanical factors (body weight, physical activity). The contributory role of TL and these other factors may differ depending on the site. For example, mechanical factors and gravity may exert a greater influence in hip BMD compared with forearm. Only hip BMD was available in this study. In the other studies, the association between TL and BMD was only observed in forearm or spine but not in the hip. The findings of this study confirm the lack of association between TL and hip BMD.

Methodological differences in epidemiological studies of telomeres need to be addressed in comparing findings from different studies. We took advantage of the latest method of qRT-PCR to measure TL with the benefits of high throughput and sensitivity to TL changes comparable to original TRF analysis [33]. This method allows fast and reliable determination of TL length in such a large-scale epidemiological study. However, it has been pointed out that many methodological issues remain unresolved. The accuracy of the qRT-PCR method (used in this study) versus the southern blot analysis (used in the two previous studies [25, 26]) may be inferior as no study comparing the two methods have been carried out. [39]. Ideally in prospective studies, the rate of change in TL should also be measured in addition to the outcome being examined. The duration of follow-up over which TL changes can be detected is uncertain. The ability of current techniques to detect small changes in TL is uncertain.

In spite of the limitations in laboratory methodology, the strength of this study lies in the large numbers of men and women and the longitudinal design allowing changes in BMD to be examined. Furthermore, it represents the first report in a different racial population with a different lifestyle. Although TL may be a biological marker of bone aging, changes in BMD will be differentially affected by many other genetic and environmental factors between populations. Genetic polymorphisms determining the rate of bone loss may differ between different ethnic groups since there are ethnic differences in cross-sectional association between BMD and genetic polymorphisms [40, 41]. Environmental factors may operate differently according to the BMD site being examined. For example, hip BMD may be more affected by lifestyle differences, being a weight-bearing part of the skeleton.

In conclusion, no association was observed between TL and change in hip BMD in Chinese men and women aged 65 years and over. This finding does not necessarily negate the concept of TL as a biological marker of bone aging, since the age range studied was not wide and BMD at the hip site may be more affected by mechanical factors compared with other sites.

Conclusion

Leukocyte telomere length is not associated with baseline hip BMD or rate of bone loss in the elderly population of our study. Age itself appears to be the most important determinant of hip BMD and rate of bone loss in this site rather than telomere length.