Evidence for a genetic link between bone and vascular measures in African ancestry families

Authors


Address correspondence to: Allison L Kuipers, PhD, University of Pittsburgh, Epidemiology, GSPH130 DeSoto Street, A521 Crabtree Hall, Pittsburgh PA 15261. E-mail: kuipers@pitt.edu

Abstract

Bone mineral density (BMD) has been inversely associated with subclinical and clinical cardiovascular disease (CVD) in population studies, but the potential mechanisms underlying this relationship are unclear. To test if there is a genetic basis underlying this association, we determined the phenotypic and genetic correlations between BMD and carotid artery ultrasound measures in families. Dual-energy X-ray absorptiometry and peripheral quantitative computed tomography were used to measure BMD in 461 individuals with African ancestry belonging to seven large, multigenerational families (mean family size 66; 3414 total relative pairs). Carotid artery ultrasound was used to measure adventitial diameter (AD) and intima-media thickness (IMT). Phenotypic and genetic correlations between BMD and carotid measures were determined using pedigree-based maximum likelihood methods. We adjusted for potential confounding factors, including age, sex, body weight, height, menopausal status, smoking, alcohol intake, walking for exercise, diabetes, hypertension, serum lipid and lipoprotein levels, inflammation markers, and kidney function. We found statistically significant phenotypic (ρ = 0.19) and genetic (ρG = 0.70) correlations (p < 0.05 for both) between lumbar spine BMD and AD in fully adjusted models. There was also a significant genetic correlation between trabecular BMD at the radius and IMT in fully adjusted models (ρG = 0.398; p < 0.05). Our findings indicate that the previously observed association between osteoporosis and CVD in population-based studies may be partly mediated by genetic factors and that the pleiotropic effects of these genes may operate independently of traditional risk pathways.

Introduction

Bone mineral density (BMD) has been inversely associated with subclinical and clinical cardiovascular disease (CVD) in population and clinical studies, even after adjusting for potential confounding factors.[1-4] Epidemiologic studies have predominately focused on investigating the relationship between BMD or osteoporotic fractures and measures of CVD in Caucasian, Mexican American, and Asian populations.[1, 5-10] Less is known about this relationship in other population groups, especially among individuals of African ancestry. Moreover, the potential mechanisms underlying the correlation between BMD and CVD are not completely understood. Part of the association may be due to shared genetic factors; in mice, genes influencing BMD have pleiotropic effects on development of atherosclerosis.[11] However, much less is known about the possible shared genetic effects on BMD and atherosclerotic CVD in humans. Using data from large, extended multigenerational families of African ancestry, we tested whether [1] there are inverse phenotypic correlations between BMD and carotid atherosclerosis measures, and [2] these correlations are due in part to pleiotropic effects of genes. We further tested if the relationship between BMD and carotid measures is independent of potential confounding factors such as age, gender, menopause, lifestyle factors, diabetes, hypertension, serum lipid and lipoprotein levels, biomarkers of inflammation, and kidney function.

Subjects and Methods

Study sample

Participants for this analysis were from the Tobago Family Health Study.[12] Briefly, eight probands were originally recruited without regard to their medical history from a population-based cohort study of bone mineral density and body composition on the Caribbean island of Tobago.[12] Probands were eligible if they had a spouse willing to participate and had at least six living offspring and/or siblings aged ≥18 years and who were residing in Tobago. All first-degree, second-degree, and third-degree relatives of the probands and their spouses were invited to participate. In 2003 to 2004, we recruited 471 individuals belonging to seven large families. The families consist of 21, 26, 28, 49, 96, 98, and 153 individuals, respectively (mean family size 66; 3414 relative pairs). An ancillary study in 2007 invited all participants in the family study to complete carotid ultrasound imaging and 395 individuals completed the scans (84% of the original study) and form the basis of the current analyses. Written informed consent was obtained from each participant. The Tobago Division of Health and Social Services and the University of Pittsburgh Institutional Review Boards approved this study.

BMD measures

Integral areal BMD at the spine and proximal femur were measured by dual-energy X-ray absorptiometry (DXA) using a Hologic QDR-4500W densitometer (Hologic Inc., Bedford, MA, USA). The short-term in vivo precision of the DXA measurements for 12 subjects were all ≤1.16%. Trabecular and cortical volumetric BMD at the nondominant forearm and left tibia was measured by peripheral quantitative computed tomography (pQCT) using an XCT-2000 scanner (Stratec Medizintechnik, Pforzheim, Germany). Technicians followed standardized protocols for patient positioning and scanning. A scout view was obtained prior to the pQCT scan to define an anatomic reference line for the relative location of the subsequent scans (4% and 33% of the total length) at the radius and tibia. Tibia length was measured from the medial malleolus to the medial condyle of the tibia, and forearm length was measured from the olecranon to the ulna styloid process. A single axial slice of 2.5 mm thickness with a voxel size of 0.5 mm and a speed of 20 mm/s was taken at all locations. Image processing was performed using the Stratec software package (version 5.5E). The short-term in vivo precision of the pQCT measurements for 15 subjects ranged from 0.65% (tibia cortical BMD) to 2.1% (tibial trabecular BMD).

Carotid artery ultrasound traits

The common carotid artery was imaged with B-mode ultrasonography using an Acuson Cypress portable ultrasound machine (Siemens Medical Solutions, Malvern, PA, USA). Both the near and far walls of the distal common carotid artery were captured for 1 cm proximal to the carotid bulb. Only the common carotid artery could be reliably imaged with the portable technology. Adventitial diameter (AD) and intima-media thickness (IMT) were obtained using a semiautomated reading software system (Artery Measurement Software [AMS]; Department of Clinical Physiology, Sahlgrenska Hospital, Gothenburg University, Gothenburg, Sweden[13]). IMT measures correspond to the mean IMT across all pixels of both the near and far wall of the common carotid artery. AD measures were obtained from the same 1-cm region and correspond to the mean distance between near and far wall medial-adventitial borders. Two images were taken on each side and both sides of each participant were averaged to obtain the overall mean measures. The mean difference between right and left side measures was 0.410 mm for AD and 0.063 mm for IMT. All images were read centrally at the Department of Epidemiology's Ultrasound Research Laboratory (University of Pittsburgh, Pittsburgh, PA, USA). Reproducibility analyses were conducted on 35 participants. The intersonographer intraclass correlation (ICC) was 0.97 for mean IMT and 0.95 for mean AD and the interreader ICC was 0.99 for both mean IMT and mean AD.

Covariates

Demographic, lifestyle, and medical history variables were collected by trained clinic staff through administration of a questionnaire and interview. Body weight was measured to the nearest 0.1 kg on a balance beam scale. Standing height was measured to the nearest 0.1 cm, without shoes, using a wall-mounted stadiometer. Race was based on self-report of grandparental ethnicity. The Tobago population is predominantly of West African origin with low admixture based on ancestry informative markers.[14] Smoking status was classified as either current or not (yes/no), and participants reporting ever smoking <100 cigarettes in their lifetime were considered nonsmokers. Alcohol consumption was assessed by questionnaire and was coded based on having >1 drink per week (yes/no) because there was a very low prevalence of substantial alcohol intake. Physical activity was assessed by the number of minutes walked per week for exercise and participants were dichotomized into “not active” or “active” determined by a median split based on walking greater than 0 minutes for exercise in the past week. Participants were asked to bring current medications to their interview, and staff recorded each medication. Diabetes was defined as a fasting glucose level ≥126 mg/dL or current use of diabetes medication. Hypertension was defined as a seated diastolic blood pressure ≥90 mmHg, systolic pressure ≥140 mmHg and/or current use of antihypertensive medication.

Biochemical assays

Venous blood samples were collected in the morning after an overnight fast. Sera were separated, aliquotted into cryovials and stored at 80°C until the time of assay. Glucose was determined using a coupled enzymatic reaction similar to the procedure described by Bodnar and Mead.[15] high-density lipoprotein cholesterol (HDL-c) was determined using the selective heparin/manganese chloride precipitation method. Low-density lipoprotein cholesterol (LDL-c) was calculated by the Friedewald equation. Triglycerides were determined enzymatically using the procedure of Bucolo and David.[16] Serum adiponectin concentrations were measured with a commercial radioimmunoassay developed by Linco Research (Linco Research, Inc., St. Charles, MO, USA). Serum C-reactive protein (CRP) was measured turbidimetrically after reaction with goat anti-CRP-antibodies coated on latex particles (Carolina Liquid Chemicals, Brea, CA, USA). Serum creatinine was quantitatively determined by the VITROS CREA slide method (Ortho Clinical Diagnostics, Raritan, NJ, USA). Standards, serum controls, and duplicate samples were run with each assay, which were traceable to a gas chromatography isotope dilution mass spectrometry (GC/IDMS) method and National Institute of Standards and Technology (NIST) SRM 914 creatinine standard reference material. The Modification of Diet in Renal Disease Study formula was used to estimate glomerular filtration rate as: GFR [mL/min/1.73 m2] = 175 × (serum creatinine [mg/dL]−1.154 × age [years]−0.203 [× 0.742, if female] [× 1.212, because they are all of African ancestry]).[17]

Statistical analysis

All non-normally distributed traits, including IMT, triglycerides, CRP, and adiponectin, were successfully transformed using a log transformation. Outliers, defined as ≥4 SD from the mean, were removed for each trait in order to satisfy the statistical assumption of normality. No more than two observations were removed from any trait and conclusions remained the same in analyses that included these observations. We first estimated the residual heritability math formula for each trait, while simultaneously adjusting for the effects of age, sex, body weight, height, and menopausal status. Residual heritability and the variance attributable to the fixed covariate effects for each trait were estimated by maximum-likelihood based methods using the program SOLAR.[18]

We next determined the extent of genetic and environmental correlation between the variance components of BMD and carotid artery measures.[18, 19] Phenotypic correlations (ρ) between BMD and carotid artery measures were estimated from residual heritability, genetic correlation (ρG) estimates as: math formula. The statistical significance of ρG being different from zero was then tested with a likelihood ratio test of models in which the parameters were constrained or unconstrained. Finally, to determine whether the correlation between BMD and carotid artery traits was influenced by known CVD risk factors, we included parameters for lifestyle factors, serum lipid and lipoproteins, inflammation, and estimated GFR (eGFR) into the base model, which was adjusted for age, sex, body weight, height, and menopausal status.

Results

Characteristics of family members

The family members ranged in age from 18 to 86 years (mean = 42.7 years) (Table 1). The sample was 60.1% female and 18.8% were postmenopausal. Men and women had similar body weight (mean = 82.3 kg), but men were taller; therefore, women had a greater body mass index (BMI). Smoking and drinking at least one alcoholic drink per week were significantly more common in men than women. The frequency of walking for exercise was similar in men and women. Diabetes was present in 9% and hypertension was present in 28.0% of the sample. There was no difference in diabetes or hypertension prevalence by sex. Women had higher levels of LDL-c, CRP, and adiponectin than men (p < 0.05 for all), but there was no difference in eGFR between men and women.

Table 1. Characteristics of the African Ancestry Family Members
TraitAll (n = 461)Men (n = 184)Women (n = 277)
  1. Values are mean ± SD, median [IQR], or %.
  2. LDL-c = low-density lipoprotein cholesterol; HDL-c = high-density lipoprotein cholesterol; CRP = C-reactive protein; eGFR = estimated glomerular filtration rate.
  3. aFrequency in women only.
  4. bDifference by sex significant (p < 0.05).
Age (years)42.7 ± 16.642.9 ± 17.042.6 ± 16.4
Female sex (%)60.1
Postmenopausal (%)a18.818.8
Body weight (kg)82.3 ± 18.484.0 ± 17.381.2 ± 19.0
Height (cm)170.7 ± 8.6b177.4 ± 7.1166.3 ± 6.4
Body mass index (kg/m2)28.3 ± 6.3b26.7 ± 4.929.4 ± 6.9
Current smoker (%)4.8b11.60.4
>1 drink per week (%)42.1b61.829.0
Walk for exercise (%)70.873.269.2
Diabetes (%)9.16.910.5
Hypertension (%)28.030.126.6
LDL-c (mg/dL)131.9 ± 40.2b127.2 ± 37.4134.9 ± 41.66
HDL-c (mg/dL)40.0 ± 12.340.8 ± 12.039.5 ± 12.48
Triglycerides (mg/dL)77.0 [59.0–104.0]81.0 [61.5–108.0]74.5 [58.0–97.0]
CRP (mg/L)1.0 [0.4–2.3]b0.8 [0.3–1.7]1.4 [0.5–2.5]
Adiponectin (ng/mL)8.7 [6.4–11.9]b6.9 [5.5–8.8]9.8 [7.6–13.6]
eGFR (mL/min/1.73 m2)103.8 ± 27.4106.3 ± 23.1102.3 ± 29.7

Heritability of BMD and carotid artery measures

All BMD and carotid artery measures were heritable after adjusting for age, sex, body weight, height, and menopausal status (Table 2). Residual heritabilities of BMD at the femoral neck and lumbar spine were 0.55 and 0.59 (p < 0.001 for both). Cortical BMD at the radius and tibia had the lowest residual heritabilities of any BMD trait (0.32 and 0.37, respectively; p < 0.001 for both). In contrast, trabecular BMD had the greatest residual heritability (0.70 and 0.67 at the radius and tibia, respectively; p < 0.001 for both). The residual heritability of AD and IMT were 0.58 and 0.46, respectively (p < 0.001 for both). Covariates explained between 19% and 56% of the variance in BMD and carotid artery traits.

Table 2. Estimates of Heritability and Effects of Covariates for BMD and Carotid Artery Measures in African Ancestry Families
TraitMean ± SDaGenetic effect (math formula ± SE)bCovariate effects (r2)
  1. Genetic effect estimates are adjusted for age, sex, body weight, height, and menopausal status.
  2. BMD = bone mineral density; AD = adventitial diameter; IMT = intima-media thickness.
  3. aUnadjusted mean and SD.
  4. bAll residual heritability estimates had a corresponding p < 0.001.
Femoral neck BMD (g/cm2)1.0 ± 0.20.55 ± 0.100.33
Lumbar spine BMD (g/cm2)1.1 ± 0.20.59 ± 0.110.21
Cortical BMD (mg/cm3)
Radius1217.1 ± 27.50.32 ± 0.090.19
Tibia1183.3 ± 31.00.37 ± 0.090.20
Trabecular BMD (mg/cm3)
Radius221.6 ± 42.60.70 ± 0.100.27
Tibia248.8 ± 35.80.67 ± 0.100.28
Carotid traits
Mean AD (mm)7.2 ± 0.70.58 ± 0.120.24
Mean IMT (mm)0.7 ± 0.10.46 ± 0.100.56

Phenotypic and genetic correlation between BMD and carotid artery measures

The genetic correlation between AD and IMT was 0.55 (p = 0.002, data not shown). All BMD measures were phenotypically, inversely correlated with AD adjusted for age, sex, body weight, height, and menopausal status (r: 0.17 to 0.23; p < 0.05; Table 3). However, BMD was not phenotypically correlated with IMT.

Table 3. Genetic and Phenotypic Correlation Between BMD and Carotid Artery Measures in African Ancestry Families Adjusted for Base Model and Full Model Covariates
Traitmath formulaAdventitial diameter (math formula = 0.58)Intima-media thickness (math formula = 0.46)
BaseaFullbBaseaFullb
ρGρρGρρGρρGρ
  1. Bold indicates p < 0.05 different than zero; italics indicates borderline significance (p < 0.10); ρ significance was assessed by variance component modeling (outlined in Subjects and Methods).
  2. ρG = genetic correlation; ρ = phenotypic correlation (calculated).
  3. aBase model correlation adjusted for age, sex, body weight, height, and menopausal status.
  4. bFull model correlations adjusted base model plus smoking, alcohol intake, walking for exercise, LDL-c, HDL-c, triglycerides, CRP, adiponectin and eGFR.
Femoral neck BMD0.55−0.511−0.206−0.490−0.241−0.259−0.080−0.296−0.079
Lumbar spine BMD0.59−0.517−0.193−0.702−0.2680.300−0.038−0.328−0.072
Radius cortical BMD0.32−0.235−0.240−0.410−0.2690.0980.0010.1090.003
Tibia cortical BMD0.37−0.375−0.262−0.632−0.2710.0880.0430.031−0.011
Radius trabecular BMD0.70−0.503−0.135−0.676−0.3170.298−0.019−0.398−0.125
Tibia trabecular BMD0.67−0.503−0.192−0.535−0.202−0.172−0.041−0.147−0.068

Part of the phenotypic correlation between BMD and AD appears to be genetic; ρG ranged from 0.38 to 0.52 (p < 0.05 for all). There was also moderate inverse genetic correlation of lumbar spine BMD (ρG = 0.30; p = 0.09) and trabecular BMD at the radius (ρG = 0.30; p = 0.08) with IMT adjusted for age, sex, body weight, height, and menopausal status.

Correlation of BMD with carotid artery traits is independent of osteoporosis and CVD risk factors

To assess the impact of potential confounders and intermediate pathways on correlations between BMD and arterial traits, we conducted more extensive modeling adjusting for hypertension, diabetes, serum lipid and lipoproteins, biomarkers of inflammation, and kidney function. Adjusting for these additional factors did not attenuate the phenotypic or genetic correlations between BMD and AD; all phenotypic and genetic correlations remained inverse and statistically significant (p < 0.05; Tables 3). In Table 4, as an example, we present more detailed results of the effects of including additional variables and confounders on the genetic and phenotypic correlation between lumbar spine BMD and arterial traits. If a particular pathway, such as lipid metabolism, mediated the genetic correlation between two traits, inclusion of lipid traits as covariates should reduce the genetic correlation. As can be seen, the addition of lifestyle factors, lipid measures, inflammation markers, etc., did not reduce the genetic correlation between lumbar spine BMD and AD.

Table 4. Multivariable Adjusted Phenotypic and Genetic Correlations Between Lumbar Spine BMD and Carotid Artery Measures in African Ancestry Families
ModelAdventitial diameterIntima-media thickness
ρGρρGρ
  1. Bold indicates p < 0.05 different than zero; italics indicates borderline significance (p < 0.10); ρ significance was assessed by variance component modeling (outlined in Subjects and Methods).
  2. ρG = genetic correlation; ρ = phenotypic correlation (calculated).
  3. aBaseline model adjusted for age, sex, body weight, height, and menopausal status.
  4. bLifestyle factors include smoking, alcohol intake, and walking for exercise.
  5. cLipid measures include LDL-c, HDL-c, and triglycerides.
  6. dInflammation markers include CRP and adiponectin.
Baseline modela–0.517–0.193–0.300–0.038
Add lifestyle factorsb–0.596–0.210–0.351–0.065
Add hypertension–0.618–0.209–0.364–0.059
Add diabetes–0.602–0.229–0.337–0.066
Add lipoproteinsc–0.670–0.268–0.388–0.097
Add inflammationd–0.710–0.271–0.318–0.071
Add eGFR–0.702–0.329–0.328–0.087

Discussion

We examined the phenotypic and genetic correlation between BMD and indices of vascular health in large, multigenerational families of African ancestry. We identified significant, inverse phenotypic and genetic correlations between areal and volumetric measures of BMD and carotid adventitial diameter. For both AD and IMT, a smaller value is indicative of a healthier vasculature. The inverse phenotypic correlation between BMD and vascular measures is consistent with previous epidemiologic studies.[1-10] The genetic correlation between noncortical BMD measures and IMT was also negative. Our results support previous observations among predominantly white individuals, but also raise the possibility that a genetic mechanism underlies the relationship between indices of bone and cardiovascular health. Moreover, the relationship appears to be independent of traditional osteoporosis and CVD risk factors. To our knowledge, this is the first study to quantify the potential genetic relationship between BMD and measures of carotid artery structure in families. Collectively, these observations suggest that part of the phenotypic correlation between BMD and carotid atherosclerosis may be due to the pleiotropic effects of a common set of genes.

Previous studies of the relationship between BMD and subclinical CVD indices have focused on women of non-African descent; the correlation between BMD and carotid artery measures in African ancestry individuals has not been rigorously evaluated. In our family-based analysis, we observed that greater BMD was correlated with smaller carotid artery vessel diameter. This finding is consistent with previous reports of an inverse association of BMD and CVD indices among unrelated individuals.[1-4] However, this is the first report of a relationship between BMD and adventitial diameter; the majority of previous reports have focused on arterial calcification as a marker of subclinical CVD.[1, 5-9] Unfortunately, we did not have measures of arterial calcification in our families to enable a comparison of findings with previous studies. We also observed an inverse genetic, but not phenotypic, correlation between carotid IMT and areal BMD at the lumbar spine and femoral neck, and trabecular BMD at the radius and tibia. However, the correlation between IMT and BMD was only significant for trabecular BMD at the radius. These findings are consistent with previous studies that reported a correlation between lower BMD and higher carotid IMT among unrelated individuals.[10, 20-22] In our study, the correlation coefficients with carotid IMT were negative for trabecular BMD at the radius. However, the magnitude of association was much smaller than for AD and achieved statistical significance only after adjusting for age, sex, body weight, height, menopausal status, lifestyle factors, hypertension, and eGFR. The association would not withstand a Bonferroni adjustment for multiple comparisons.

We also found significant, inverse genetic correlations between BMD and adventitial diameter and IMT. These observations suggest that variation at a similar set of genes may influence these traits. Changes in AD arise from changes in the hemodynamics within the vasculature and occur before changes in IMT, which arise from plaque burden.[23] This suggests that the strong and consistent genetic correlation between AD and BMD is not reflective of plaque formation, but rather of earlier vascular remodeling. The strongest genetic correlations were observed for trabecular BMD in the appendicular skeleton and integral BMD at the lumbar spine, also a mainly trabecular-rich skeletal site. This observation may reflect the higher heritability and genetic determination of trabecular BMD in these families or to the higher turnover rate of trabecular compared with cortical bone. Genes with pleiotropic effects on bone and atherosclerosis have previously been found in inbred strains of mice.[11] Our study is consistent with this report, and is the first to investigate the potential shared genetics between measures of bone and vascular health in human families.

Many hypotheses have been proposed to explain the relationship between BMD and CVD, including a shared pathophysiology such as diabetes, hypertension, dyslipidemia, inflammation, or sex steroid effects.[24-30] Using our extensive phenotypic data, we were able to statistically adjust for all of these intermediate factors except sex steroids. In addition, we were able to test the impact of kidney health, which may be an important mediating factor because patients with chronic kidney disease are more likely to experience both osteoporosis and CVD.[5, 9, 24, 31]

Other hypotheses for the underlying cause of the correlation between bone and CVD include a loss of inhibition of bone formation signals with aging[32] and calcium misappropriation.[33] Calcifying vascular cells exhibit characteristics similar to bone,[29] and osteoblasts and osteoclasts have been found in calcified regions in the blood vessels,[30, 32] leading some to hypothesize that calcium deposition within the arteries occurs when the natural mechanism to inhibit this process is lost.[29]

In conclusion, we have identified an inverse phenotypic correlation between BMD and carotid adventitial diameter and IMT in large, multigenerational African ancestry families that appears to have a strong genetic basis. The genetic correlation with carotid artery measures was greater for trabecular bone than cortical bone, which may help advance our understanding of the shared mechanisms. Adjustment for previously hypothesized intermediate pathways had no effect on the genetic correlations between BMD and carotid measures. Further research is needed to better understand the potential physiologic and genetic mechanisms for the association of BMD with measures of atherosclerotic CVD.

Disclosures

The authors state that they have no conflicts of interest.

Acknowledgments

This work was supported by NIH grants R03-AR050107 and R01-AR049747 from the National Institute of Arthritis and Musculoskeletal and Skin Diseases. ALK is supported by a National Heart Lung and Blood Disorders Institute postdoctoral grant T32-HL083825. IM was supported by a Mentored Research Scientist Development Award from the National Institute of Diabetes and Digestive and Kidney Diseases grant K01-DK083029. We thank Dr. Genevieve Woodard for her assistance in data collection for this manuscript.

Authors' roles: ALK: manuscript writing and revising, data analysis and interpretation, and data collection; IM: manuscript revision and data collection; CMK: manuscript revision and interpretation of data; RWE: data collection; CHB: manuscript revision, data collection and obtained funding for parent studies; ALP: data collection and obtained funding for parent studies; VWW: data collection and obtained funding for parent studies; KST: manuscript revision and data collection; JMZ: study conception and design, manuscript revision, data interpretation, and obtained funding for parent studies. All authors have approved the final version of this manuscript.

Ancillary