Associations between circulating sex steroid hormones and leukocyte telomere length in men in the National Health and Nutrition Examination Survey

Telomeres are DNA-protein complexes that cap and shield the ends of chromosomes. Several processes, particularly mitotic cell division, lead to telomere shortening over time ( Blackburn, 2005 ; Jain & Cooper, 2010 ). Shortening telomere length—which occurs with age and is enhanced in individuals with telomeropathies—has been associated with risk of several outcomes including atherosclerosis, cardiovascular disease, cancer and death ( Armanios & Blackburn, 2012 ; Fyhrquist et al., 2013 ; Needham et al., 2015 ; Samani et al., 2001 ; Wentzensen et al., 2011 ). Telomere shortening can be slowed by a reverse transcriptase enzyme, telomerase, which is expressed in a limited number of normal human adult cells (typically stem cells) and provides a DNA template to overcome replication complications at chromosome ends ( Blackburn, 1997 ). It has been widely hypothesized that enhancing telomerase expression and activity may provide a therapeutic intervention for individuals diagnosed with telomeropathies ( Holohan et al., 2014 ) and may even avert age-related conditions and increase longevity ( Bar et al., 2015 ; Bernardes de Jesus et al., 2012 ; de Jesus, 2011 ). Further, preliminary evidence has demonstrated that sex steroid hormones may increase telomerase activity in certain cell lines and tissue types ( Calado et al., 2009 ; Kyo et al., 1999 ; Nanni et al., 2002 ; Nourbakhsh et al., 2010 ; Sato et al., 2004 ). A recent phase 1–2 clinical trial provided evidence that oral Danazol—a synthetic steroid sex hormone with weak androgenic effects and functional anti-estrogenic effects ( Hoffman, 2016 ; Knuth, 2010 )—led to leukocyte telomere elongation in individuals with telomeropathies ( Townsley et al., 2016 ). Despite the potential clinical applications of exogenous hormones among individuals with telomeropathies, exploration of relationships between endogenous hormones and telomere length has been limited. Further, such investigations may help clarify the underlying mechanism of how Danazol may affect telomere length. Therefore, we assessed circulating sex steroid hormone and sex hormone-binding globulin (SHBG) concentrations in relation to leukocyte telomere length (LTL) utilizing a relatively large, and ethnically diverse cross-sectional study.

The subsample identified for analysis, with data on circulating sex steroid hormones, SHBG, and LTL, comprised 499 male subjects ( Figure 1 ). Descriptive statistics and estimates of association between hormone metrics and telomere lengths were weighted to the US population using NHANES weights as described. Sample-weighted population characteristics included a median age of 42 years and a median BMI of 27.1 ( Table 1 ). Current cigarette smokers comprised 26.1%, 51.7% were non-Hispanic white and 27.3% had some college education. Median biological parameters for all hormones and SHGB were within normal reference ranges for adult males ( Becker, 2001 ; Gardner et al., 2007 ). As expected, in univariate sample-weighted linear regression, continuous age was inversely associated with leukocyte telomere length ( β continuous per ½ IQR = −0.17, p<0.001) (results not tabulated). Age-adjusted, sample-weighted linear regressions demonstrated associations of telomere length with estradiol ( β continuous per ½ IQR = −53, p=0.05) and free estradiol ( β continuous per ½ IQR = −60, p=0.048) ( Table 2 ). Additional adjustment for race/ethnicity, education, BMI, and smoking status slightly strengthened these inverse associations (estradiol β continuous per ½ IQR = −61, p=0.04; free estradiol β continuous per ½ IQR = −67, p=0.03; Table 2 ). There was no evidence for any other relationship between sex steroid hormones or SHBG and LTL. Visual assessment of the regression shape, and results of the Wald tests confirmed that associations of all hormones and SHGB with LTL were linear (p>0.05). Reinstatement of subjects with values in the top and bottom 1 percentiles of hormone distributions did not affect these results, with the exception that the inclusion of two subjects with the highest supraphysiological levels of estradiol (360 pg/mL and 130.5 pg/mL respectively, compared with the normal physiological range of 10–50 pg/mL for adult men ( Gardner et al., 2007 )) did cause the estradiol–LTL relation to become non-linear (p<0.05), though categorical quartile analyses remained marginally statistically significant (results not tabulated). There was no statistical evidence for interaction by age, race/ethnicity, or BMI, except for androstanediol glucuronide by BMI (p interaction =0.03) which gave variable β continuous per ½ IQR values of −75, 39, and −40 for BMI categories of <25, 25–29.9, and ≥30 kg/m 2 , respectively (results not tabled).

In this study of NHANES data we found an inverse association between circulating and free estradiol concentrations with LTL. Specifically, for every quartile (~8 pg/mL) increase in circulating estradiol, LTL decreased by 61 base pairs. For all other hormones and SHBG, there was no association with LTL in minimally and maximally adjusted analyses. A recent study of endogenous estradiol and LTL provides discordant results. This larger study, based within the Busselton Health Study of Western Australia, included 980 men aged 17–90 years and found a very modest positive correlation between estradiol and telomere length using the T/S ratio (r=0.068, p=0.034) ( Yeap et al., 2016 ). Although this analysis only adjusted for age, additional adjustment in our models for other covariates only served to slightly strengthen the inverse relationships observed. The mean age in the Busselton study was slightly younger (42 years) than that of our study (weighted mean=44 years), and we did not find any effect modification by age. In contrast, a smaller study of 110 men (aged 71–86 years) based in Ghent, Belgium found no correlation between estradiol and telomere length ( Bekaert et al., 2005 ), although this study used laboratory assays that are less sensitive (RFLP and immunoassays) than our study and the Busselton study (qPCR and mass spectrometry). The inconsistent results from studies of endogenous sex steroid hormones (specifically estradiol) and telomere length highlight the complexity of these relationships and mirror the intricate pharmacology of Danazol, which has been described as a weak androgen, weak functional anti-estrogen, weak anti-gonadotropin, weak progestogen, competitive inhibitor of SHBG, and a down regulator of SHBG ( Hoffman, 2016 ; Knuth, 2010 ). Although the authors exploring Danazol and LTL hypothesized that an estrogen response element (ERE) in the TERT promoter may underlie the observed positive association between Danazol and telomere length ( Townsley et al., 2016 ), there is evidence that Danazol does not bind to the estrogen receptor ( Knuth, 2010 ) and the ERE has high specificity for the estrogen receptor. Utilizing NHANES provided a large and nationally representative study population, and enabled us to adjust for important covariates in our analysis of endogenous sex steroid hormones and LTL. However, the cross-sectional nature of this analysis limits our ability to establish a temporal association between hormones and telomere length. Additionally, serum hormone and SHBG levels were derived from one sample which may not represent typical concentrations, though the inclusion of morning samples only should mitigate variation in diurnal patterns. Overall, the findings from this analysis and similar previous studies demonstrate the complex and equivocal role of sex steroid hormones in the association between Danazol and LTL, as well as the need for further longitudinal research. The observed inverse association between endogenous estradiol and LTL, coupled with prior evidence that Danazol has hypoestrogenic effects in women ( Hoffman, 2016 ; Knuth, 2010 ), suggests a Danazol-induced hypoestrogenic state may underlie the observed association between Danazol treatment and increased LTL. Thus, further investigation of sex steroid hormones should be considered to further elucidate the mechanisms underlying the association between Danazol and LTL.

The National Health and Nutrition Examination Survey (NHANES) conducted by The National Center for Health Statistics (NCHS), Centers for Disease Control and Prevention (CDC) is a series of nationally representative, cross-sectional surveys of the non-institutionalized, U.S. population, consisting of approximately 10,000 persons every two-years. The survey includes collection of demographic, socioeconomic, dietary, and health-related information, and the physical examinations include medical, dental, and physiological measurements, as well as laboratory tests. NHANES 1999–2002 contains 21,004 individuals aged 2 months and older with a 4-stage sampling design: 1) primary sampling units (PSUs) consisting of single or contiguous counties, 2) area segments within sampled PSUs, 3) households within sampled segment areas, and 4) persons within sampled households. NHANES oversamples certain strata to enhance accuracy of estimates. During 1999–2002, individuals with low-incomes, individuals aged 12–19 years, individuals aged 60 years and older, African-Americans, and Mexican-Americans were oversampled. Sex steroid hormones and sex hormone-binding globulin (SHBG) were previously measured among men who were 20 years and older who participated in NHANES 1999–2004 surveys, as described previously ( Nyante et al., 2012 ). Eligible male subjects from 1999–2002 only were included in this analysis because DNA samples were not collected in 2003 or 2004, as described below. Serum androstanediol glucuronide, testosterone, estradiol, and SHBG were quantified at Children’s Hospital, Boston, MA. Androstanediol glucuronide was measured using ELISA (ALPCO Diagnostics, Salem, NH). Testosterone, estradiol, and SHBG were measured using the Elecsys 2010 system (Roche Diagnostics, Laval, QC, Canada). Coefficients of variation were 5.2% for androstanediol glucuronide, 3.4% for testosterone, 15.8% for estradiol, and 4.7% for SHBG. We calculated free estradiol ( Sodergard et al., 1982 ) and free testosterone ( Vermeulen et al., 1999 ) using formulas that include the individual hormone, SHBG and a constant for albumin (43 g/L). DNA samples collected in NHANES 1999–2002 were purified from whole blood and were thus amenable to telomere length estimation, as previously described ( Needham et al., 2013 ). In brief, 7,826 of the 10,291 eligible adults (age 20 years and older) provided both DNA and consent for future genetic research and underwent telomere length assessment using quantitative polymerase chain reaction (PCR) method to measure telomere length relative to standard reference DNA (T/S ratio). Each sample was assayed twice and the mean T/S ratio was calculated. The inter-assay coefficient of variation was 4.4%. For statistical analysis, we converted the mean T/S ratio to base pairs (bp) using the formula 3,274 + 2,413 * (T/S) which was based on a comparison of Southern blot telomeric restriction fragment (TRF) lengths and T/S ratios ( Needham et al., 2013 ). Among the 3,077 men aged 20 or older who also provided DNA samples and consent, selection was restricted to individuals who provided a morning blood sample (n=1,717) to reduce effects of diurnal variation in sex hormone production, and to individuals who were previously selected for persistent organic pollutant (POP) analysis (n=575). These restrictions are the same criteria applied to a previous analysis in NHANES of sex steroid hormones in order to use existing hormone and SHBG data as well as the corresponding sample weights ( Trabert et al., 2012 ). With these restrictions, there were 527 men with data on relevant sex steroid hormones and SHBG, and had telomere length quantified ( Figure 1 ). Age (continuous), race/ethnicity (non-Hispanic white/non-Hispanic black/Mexican-American/Other/Other Hispanic), cigarette smoking (never/former/current), and education (less than 9th grade/9th−11th grade/high school graduate/some college/college graduate) were collected from self-report based on the NHANES interview. Body mass index (BMI, kg/m 2 ) was calculated from height and weight, which were measured during the physical examination. Individuals within or equal to the top or bottom one percentile of any sex steroid hormone or SHBG (n=28) were omitted from the defined subpopulation for statistical analysis to reduce the potential effects of outliers, given that observations within these percentiles were largely outside of reported physiological ranges ( Becker, 2001 ; Gardner et al., 2007 ). After excluding these values, there were 499 men available for analysis ( Figure 1 ). All descriptive statistics and inferential analyses used survey procedures to account for the sample weights and the stratification and cluster sampling aspects for the complex survey design of NHANES to ensure that results were representative of the US civilian non-institutionalized population. Because the men used in the analysis were a subsample of NHANES that had readily available weights, we used existing NHANES sample weights to derive analytic sample weights for this study. The analytic sample weights were calculated as the product of the 4-year (1999–2002) persistent organic pollutant weight ( wtspo4yr ) and the 4-year (1999–2002) fasting weight ( wtsaf4yr ) divided by the 4-year (1999–2002) Mobile Exam Center (MEC) weight ( wtmec4yr ). Sample-weighted median and interquartile ranges of continuous variables were estimated as well as percentages of categorical variables. Sample-weighted linear regression was used to estimate age-adjusted and multivariable-adjusted estimates of associations between circulating sex steroid hormones, SHBG, and LTL. Independent variables (hormones and SHBG) were modelled as continuous variables as was the dependent variable of LTL (bp). Estimates of association were rescaled to represent telomere length change in base pairs per half the value of the interquartile range of the independent variable. This approximates a single quartile increase in exposure and facilitates comparison of effects across independent variables. Linearity of association was assessed through modeling the independent variables (hormones and SHBG) as quartiles, visual assessment of the regression shape using restricted cubic spline of the independent variable with 3-knots at the 10 th , 50 th , and 90 th percentiles, and a Wald test to examine whether the spline regression coefficient was equal to zero. Age adjustment in all models used a cubic spline with three knots at the 10 th , 50 th , and 90 th percentiles to account for a potential nonlinear relationship with LTL. Multivariable adjusted models included age, race/ethnicity, education, BMI, and smoking status. Effect modification by age (20–44, 45–69, 70–84 years), race (non-Hispanic white, non-Hispanic black, Mexican-American), and BMI (<25, 25–29.9, ≥30 kg/m 2 ) were assessed using the Wald test to determine whether interaction parameters significantly differed ( Korn & Graubard, 1999 ).