Epigenetic Aging and Musculoskeletal Outcomes in a Cohort of Women Living With HIV

Abstract

Background

The relationship between accelerated epigenetic aging and musculoskeletal outcomes in women with HIV (WWH) has not been studied.

Methods

We measured DNA methylation age using the Infinium MethylationEPIC BeadChip in a cohort from the Women's Interagency HIV Study (n = 190) with measures of bone mineral density (BMD) and physical function. We estimated 6 biomarkers of epigenetic aging—epigenetic age acceleration (EAA), extrinsic EAA, intrinsic EAA, GrimAge, PhenoAge, and DNA methylation–estimated telomere length—and evaluated associations of epigenetic aging measures with BMD and physical function. We also performed epigenome-wide association studies to examine associations of DNA methylation signatures with BMD and physical function.

Results

This study included 118 WWH (mean age, 49.7 years; 69% Black) and 72 without HIV (mean age, 48.9 years; 69% Black). WWH had higher EAA (mean ± SD, 1.44 ± 5.36 vs −1.88 ± 5.07; P < .001) and lower DNA methylation–estimated telomere length (7.13 ± 0.31 vs 7.34 ± 0.23, P < .001) than women without HIV. There were no significant associations between accelerated epigenetic aging and BMD. Rather, measures of accelerated epigenetic aging were associated with lower physical function.

Conclusions

Accelerated epigenetic aging was observed in WWH as compared with women without HIV and was associated with lower physical function in both groups.

bone mineral density, epigenetic aging, HIV, osteoporosis, physical function

Adults with HIV appear to have a higher prevalence of frailty, poor physical function, and osteoporosis than age- and sex-matched adults without HIV [1, 2]. For women with HIV (WWH), conditions of musculoskeletal aging, including osteoporosis and fractures, occur more frequently after the menopausal transition, and the differences in prevalence between WWH and women without HIV are greater among postmenopausal women [3].

DNA methylation (DNAm)—specifically, methylation of 5′-cytosine in CpG-rich regions of DNA—is the most commonly studied epigenetic change due to its stability and accessibility [4, 5]. DNAm age correlates with chronological age, and individuals with increased DNAm age as compared with their chronological age have epigenetic age acceleration (EAA) [6]. There are numerous methylation-based biomarkers of aging (ie, epigenetic clocks) that measure age acceleration, including EAA, extrinsic EAA (EEAA), intrinsic EAA (IEAA), GrimAge, and PhenoAge, as well as a DNAm estimator of telomere length [6–10]. Accelerated epigenetic age by DNAm has been reported among adults with HIV [11–15]; however, most of the data are based on men.

Few studies have examined the relationship between EAA and bone mineral density (BMD) or physical function in WWH. A study in children did not find any evidence of accelerated epigenetic aging at birth or age 7 years with BMD [16]. A previous study showed that increased DNAm of Alu, a cluster of interspersed DNA elements, is associated with accelerated aging and lower BMD in postmenopausal women with osteoporosis [17]. In the general population, GrimAge was associated with a decline in physical function, including lower performance on 6-minute and 10-m walk tests and on knee extension and ankle plantar flexion strength tests, over 3 years of follow-up in older women [18].

In the primary analysis of the Women's Interagency HIV Study (WIHS) Musculoskeletal Substudy (MSK) of pre-, peri-, and postmenopausal women, we found that WWH had lower areal BMD by dual energy x-ray absorptiometry (DXA) and lower volumetric BMD by quantitative computed tomography than age- and race/ethnicity-matched women without HIV [3]. In this secondary analysis, we compare methylation-based biomarkers of aging between WWH and women without HIV and evaluate associations of EAA with BMD and physical function. We explore associations with DNAm signatures using large-scale epigenome-wide association studies (EWASs) to identify potentially modifiable mechanistic pathways.

METHODS

Study Participants

This analysis includes data from participants in the MSK, which enrolled 340 participants aged 40 to 60 years from 3 WIHS sites (San Francisco, Bronx, and Chicago): one group had HIV, was undergoing antiretroviral therapy for >1 year, and had CD4 >100 cells/μL; the other was a comparison group of women without HIV, with similar age, ethnicity, and risk behaviors [3]. Exclusion criteria for the MSK included weight >264 lb, height greater than 6′1″, pregnant or breastfeeding in the past 6 months, and estimated glomerular filtration rate <60 mL/min/1.73 m² per the Modification of Diet in Renal Disease equation, as well as current hormone replacement therapy, osteoporosis treatment, or glucocorticoid use. For this epigenetic analysis, a total of 195 samples were selected: 89 from the Bronx, 65 from San Francisco, and 41 from Chicago. This study was approved by the institutional review boards of all participating institutions, and informed consent was provided by all participants.

Measurements

Demographics and Clinical Characteristics

Demographic and clinical information on age, race (Black vs non-Black), weight, body mass index, menopausal stage defined per the SWAN trial (Study of Women's Health Across the Nation) [3, 19], substance use, and comorbidities was extracted from the WIHS database. HIV-related variables included information on CD4 count, HIV RNA level, and use and type of antiretroviral regimens.

Bone Mineral Density

DXA was utilized to measure areal BMD at the lumbar spine and total hip. DXA scans were performed with Lunar Prodigy densitometers (GE Medical Systems) at all MSK study sites and read centrally at the Image Analysis Lab as described previously [3].

Physical Function

A battery of muscle strength, walking speed, balance, and endurance measures was developed per the Baltimore Longitudinal Study on Aging and other aging cohorts [20, 21]. Muscle strength was assessed by grip strength [22] and repeated chair stands. For grip strength, the participant was asked to hold a handheld Jamar dynamometer with the dominant hand and squeeze with maximum force in kilograms, and the best of 3 attempts was utilized in analysis. For the repeated chair stand, the participant was asked to stand from a seated position without the aid of the arms. The time to completion of 10 repetitions was recorded to minimize the ceiling effect for higher-functioning women [23]. Walking speed was defined by the faster of 2 measurements at a “normal, comfortable pace” over a 4-m course. Endurance was assessed by a 400-m walk, which measures the time taken to complete it. Static balance was assessed with the 4-stage standing balance test (parallel, semitandem, tandem, and single leg) [24], with the duration that participants were asked to hold each position increased to 30 seconds to reduce the possibility of a ceiling effect [25]. Only data from the single-leg stance test were included here. The functional reach test was also performed, with results reported as the mean scores of 3 reaches, where each individual would slide a hand as far forward as possible without losing balance or taking a step [26].

DNA Extraction and Genome-wide Methylation Profiling

DNA was extracted by the Qiagen QIAmp DNA Blood Midi Kit (Qiagen) and quantified with the Qubit dsDNA BR Assay Kit and Qubit Fluorometer (Thermo Fisher Scientific) at the Columbia University Irving Medical Center. DNAm levels were measured by the Infinium MethylationEPIC BeadChip (version 1.0; Illumina) at the Roswell Park Cancer Institute. Genomic DNA (50 ng/µL) was isolated and quantified with PicoGreen (Thermo Fisher Scientific) from peripheral blood mononuclear cells of the 195 participants. Infinium MethylationEPIC BeadChip arrays, which interrogate ∼850 000 CpG sites, were run at the Northwestern University Genomics Core Facility according to the manufacturer's protocol. Briefly, genomic DNA samples were bisulfite converted by the EZ DNA Methylation Kit (Zymo Research). Samples were amplified, enzymatically fragmented, and hybridized to the BeadChips. Following hybridization, the chips were stained, washed, and scanned with the Illumina HiScan System. Raw intensity data files (.idat) were obtained.

Bioinformatics Preprocessing of DNA Methylome Data

Standard preprocessing pipeline procedures, including filtering, quality control, and dye bias correction, were performed with the R package ewastools (version 1.5) [27, 28]. Control metrics were checked for quality control. With the threshold of single-nucleotide polymorphism outliers set as −4, 5 samples were removed, for a total of 190 remaining for analysis. An overall 843 393 CpG sites were obtained after removal of the control probes, non-CpG probes, failed probes with detection at P ≥ .05, single-nucleotide polymorphism–enriched probes, probes demonstrated to cross-hybridize nonspecifically in the genome, and sex chromosome probes. Estimated proportions of 6 cell types were estimated via the Houseman method: B cells, CD4 T cells, CD8 T cells, natural killer cells, granulocytes, and monocytes [29].

DNAm Age

Estimated DNAm age in years was obtained from the online calculator developed by Horvath (https://dnamage.genetics.ucla.edu/) [6]. Six methylation-based biomarkers of aging were calculated: EAA, extrinsic EAA (EEAA), intrinsic EAA (IEAA), GrimAge, PhenoAge, and DNAm-estimated telomere length (DNAmTL). EAA is the (raw) residual resulting from regressing the Horvath-estimated DNAm age onto the chronological age [6]. EEAA is the residual resulting from regressing the Hannum-estimated DNAm age upweighted for the contributions of age-related blood cell counts onto the chronological age [30, 31]. IEAA is the residual resulting from regressing the Horvath-estimated DNAm age onto the chronological age + CD8.naive + CD8pCD28nCD45RAn + PlasmaBlast + CD4T + NK + Mono + Gran. GrimAge is based on 1030 CpG sites that predict time to death [8]. PhenoAge is based on 513 CpG sites that predict morbidity and mortality [9]. For these 5 measures, positive and negative values indicate that the participant's biological age is older and younger than expected based on chronological age, respectively. The final measure, DNAmTL, estimates telomere length, where a shorter telomere length is indicative of accelerated biological aging [32].

Statistical Analysis

Continuous variables were descriptively summarized by means and standard deviations and categorical variables by percentages. Comparisons between WWH and women without HIV were performed with t tests for continuous variables and Fisher exact tests for categorical variables. Biomarkers of epigenetic aging were compared between groups with t tests and linear regression, unadjusted and adjusted for race and smoking status. Associations between biomarkers of epigenetic aging and continuous measures of bone and physical function were assessed with linear regression models.

For EWAS analyses, we fit a model using the empirical Bayes moderated linear regression approach implemented by limma [33] with DNAm as the dependent variable and HIV status as the primary independent variable. We conducted an unadjusted analysis, as well as an analysis adjusted for age, race, and smoking status. CpG sites were considered differentially methylated if they had a P value meeting the Holm-Bonferroni threshold (P < 5.92 × 10⁻⁸) and |Δβ|>0.05, where Δβ is the mean difference between the average DNAm of the groups. Gene annotations used in the analysis were based on the IlluminaHumanMethylationEPICanno.ilm10b2.hg19 database [34]. All analyses were performed with R statistical software (version 4.1.2).

RESULTS

Characteristics of the Study Population

A total of 118 WWH (mean age, 49.7 years; 69% Black) and 72 women without HIV (mean age, 48.9 years; 69% Black) were included, and characteristics are shown in Table 1. WWH had a weight and body mass index similar to women without HIV. WWH were also less likely to be a current smoker. Among the WWH, the median CD4 count was 567 cells/mm³ (IQR, 467–756), and 72% had an HIV RNA <50 copies/mL. WWH had lower BMD T scores at the lumbar spine and total hip when compared with women without HIV. There was little difference in physical function measures between women with and without HIV (grip strength, repeated chair stand, walk speed, single-leg stand, and functional reach).

Table 1.

Characteristics of Participants in the WIHS Musculoskeletal Substudy With DNA Methylation Data (n = 190)

Characteristic	Women With HIV (n = 118)	Women Without HIV (n = 72)	P Value
WIHS site			.55
Bronx/Manhattan	56 (47)	31 (43)
San Francisco	35 (30)	27 (38)
Chicago	27 (23)	14 (19)
Age, y, mean (SD; range)	49.7 (4.9; 39.5–60.5)	48.9 (5.8; 39.9–59.9)	.36
Race			.82
White	14 (12)	10 (14)
Black	81 (69)	50 (69)
Other	23 (19)	12 (17)
Weight, kg, mean (SD)	77.5 (16.1)	81.0 (16.3)	.16
BMI, kg/mg2			.52
< 18.0	1 (1)	1 (1)
18.0–24.9	26 (22)	12 (17)
25.0–29.9	39 (33)	20 (28)
>30.0	52 (44)	39 (54)
Menopausal stage			.44
Premenopause/early perimenopause	40 (34)	29 (40)
Late perimenopause/postmenopause	78 (66)	43 (60)
Smoking status			.0005
Never	14 (12)	12 (17)
Past	45 (38)	9 (12)
Current	59 (50)	51 (71)
Alcohol >12 drinks/wk	5 (4)	10 (14)	.025
At index visit
Injection drug use	1 (1)	1 (1)	>.99
Opiate use	1 (1)	1 (1)	>.99
Cocaine use	2 (2)	7 (10)	.028
Diabetes	34 (29)	21 (29)	>.99
Hepatitis C infection, RNA+	27 (23)	16 (22)	>.99
eGFR <60 mL/min	5 (4)	2 (3)	.71
History of use
Statin	36 (31)	17 (24)	.32
Metformin	6 (5)	2 (3)	.71
HIV parameter		…	…
Current CD4 count at index, cells/mm3, median (IQR)	567 (467–756)
CD4/CD8 ratio, mean (SD)	0.85 (0.45)
HIV RNA viral load, copies/mL		…	…
≤50	85 (72)
51 to <200	9 (8)
≥200 to <400	8 (7)
≥400	16 (13)
Antiretroviral use		…	…
Cumulative years, mean (SD)
ART	9.26 (5.54)
Tenofovir	3.62 (3.51)
History of use
Older PI	81 (69)
Thymidine analog NRTI (D4T, AZT, DDI)	78 (66)
Use at index visit
Any ART	116 (98.3)
TDF	59 (50)
NRTI	112 (95)
NNRTI	59 (50)
PI	56 (47)
INSTI	13 (11)
Areal BMD, mean (SD)
g/cm2
Lumbar spine	1.23 (0.18)	1.27 (0.19)	.12
Total hip	1.03 (0.15)	1.08 (0.15)	.019
T score
Lumbar spine	0.27 (1.58)	0.73 (1.61)	.06
Total hip	0.10 (1.18)	0.57 (1.20)	.012
Lumbar spine T score
<−1	24 (20)	7 (10)	.07
<−2	7 (6)	2 (3)	.49
Total hip T score
<−1	19 (16)	6 (8)	.18
<−2	3 (3)	0 (0)	.29
Physical function
Muscle strength, mean (SD)
Grip strength, kg	25.56 (6.16)	25.97 (5.57)	.65
Time to complete 10 chair stands, s	25.55 (6.17)	25.58 (6.40)	.98
Walking speed: time to complete 4-m walk, s, mean (SD)	4.04 (0.86)	3.99 (0.83)	.67
Endurance: time to complete 400-m walk, min, mean (SD)	5.12 (0.85)	5.35 (0.81)	.09
Balance
Able to hold a single-leg stand for 30 s	51 (43)	27 (38)	.45
Duration of single-leg stand, s, mean (SD)	23.59 (9.17)	22.49 (10.29)	.54
Functional reach, cm, mean (SD)	25.40 (8.44)	23.97 (6.79)	.23

Data are presented as No. (%) unless noted otherwise.

Abbreviations: ART, antiretroviral therapy; BMD, bone mineral density; BMI, body mass index; CD4, cluster of differentiation 4; eGFR, estimated glomerular filtration rate; INSTI, integrase strand transfer inhibitor; NNRTI, nonnucleoside reverse transcriptase inhibitor; NRTI, nucleoside reverse transcriptase inhibitor; PI, protease inhibitor; TDF, tenofovir disoproxil fumarate; WIHS, Women’s Interagency HIV Study.

HIV

WWH had a significantly higher EAA (mean ± SD, 1.44 ± 5.36 vs −1.88 ± 5.07; P < .001) and a lower DNAmTL (7.13 ± 0.31 vs 7.34 ± 0.23, P < .001) than women without HIV (Figure 1). EEAA, IEAA, GrimAge, and PhenoAge were not significantly different between the groups. Findings were similar when adjusted for smoking status and race.

Figure 1.

A–F, EAA, EEAA, IEAA, GrimAge, PhenoAge, and DNAmTL in women with HIV (HIV+, n = 118) and women without HIV (HIV–, n = 72). Comparisons between groups were performed with t tests. ***P < .05. Line, median; box, IQR; error bars, 1.5 × IQR; circles, outliers. DNAmTL, DNA methylation–estimated telomere length; EAA, epigenetic age acceleration; EEAA, extrinsic epigenetic age acceleration; IEAA, intrinsic epigenetic age acceleration; NS, not significant.

In an unadjusted model, we identified 2286 differentially methylated CpG sites associated with HIV that met the Holm-Bonferroni threshold and had |Δβ| > 0.05. In a model adjusted for age, race, and smoking status, we identified 2094 differentially methylated CpG sites associated with HIV. The top 50 CpG sites are shown in Supplementary Table 1.

Bone

There were no differences in biomarkers of aging between women with and without a lumbar spine BMD T score <−1 or between women with and without a total hip BMD T score <−1 for the overall cohort (Table 2). Among WWH, there were also no differences between the groups (data not shown). Among women without HIV, EAA was significantly lower for those with a lumbar spine BMD T score <−1 vs those with a T score >−1 (mean ± SD, −5.43 ± 3.25 vs −1.35 ± 5.14). When continuous bone outcomes were examined for the overall cohort, greater lumbar spine BMD (β = 0.238; 95% CI, .010–.467; P = .04) and total hip BMD (β = 0.316; 95% CI, .029–.602; P = .03) were associated with increased DNAmTL.

Table 2.

Methylation-Based Biomarkers of Aging by Women With Lumbar Spine and Total Hip BMD T Scores <−1 vs >−1

	BMD T Scores, Mean (SD)
	Lumbar Spine			Total Hip
Epigenetic Aging Measure	<−1 (n = 31)	>−1 (n = 153)	P Value	<−1 (n = 25)	>−1 (n = 156)	P Value
EAA	−0.19 (6.52)	0.25 (5.17)	.73	1.88 (6.43)	0.00 (5.34)	.18
EEAA	−1.00 (8.12)	0.46 (7.51)	.36	1.17 (8.14)	0.12 (7.63)	.55
IEAA	0.58 (4.41)	0.12 (4.52)	.60	1.79 (4.36)	−0.01 (4.65)	.07
GrimAge	−0.65 (5.79)	−0.14 (4.69)	.65	−0.60 (5.96)	−0.08 (4.74)	.68
PhenoAge	−1.09 (8.72)	0.16 (7.06)	.46	−0.35 (8.97)	0.01 (7.25)	.85
DNAmTL	7.13 (0.33)	7.22 (0.29)	.16	7.11 (0.33)	7.23 (0.29)	.12

Abbreviations: BMD, bone mineral density; DNAmTL, DNA methylation–estimated telomere length; EAA, epigenetic age acceleration; EEAA, extrinsic epigenetic age acceleration; IEAA, intrinsic epigenetic age acceleration.

In an EWAS, no BMD measures were associated with DNAm (lumbar spine T score or BMD, total hip T score or BMD). No CpG sites met criteria for association (Holm-Bonferroni significance and |Δβ|>0.05).

Physical Function

We examined associations between biomarkers of aging and various measures of physical function. Women who could not hold a single-leg stand for 30 seconds (n = 112) had higher EAA, EEAA, IEAA, GrimAge, and PhenoAge and lower DNAmTL vs women who could hold the single-leg stand for 30 seconds (n = 78; Table 3). When stratified by HIV status, WWH who could not hold a single-leg stand for 30 seconds had higher EEAA (mean ± SD, 2.17 ± 8.30 vs −0.95 ± 7.34; P = .03), GrimAge (0.78 ± 5.03 vs −1.51 ± 4.47, P = .01) and PhenoAge (2.01 ± 7.23 vs −1.29 ± 7.09, P = .01) than WWH who could hold the single-leg stand for 30 seconds (Supplementary Table 2). For women without HIV, those who could not hold a single-leg stand for 30 seconds had higher EAA (−1.02 ± 5.38 vs −3.32 ± 4.22, P = .048) and lower DNAmTL (7.30 ± 0.24 vs 7.41 ± 0.20, P = .045) vs those who could hold the single-leg stand for 30 seconds (Supplementary Table 2).

Table 3.

Methylation-Based Biomarkers of Aging by Women Who Can and Cannot Hold a Single-Leg Stand for 30 Seconds

	Single-Leg Stand: 30 s, Mean (SD)
Epigenetic Aging Measure	Cannot Hold (n = 112)	Can Hold (n = 78)	P Value
EAA	0.85 (5.58)	−0.78 (5.22)	.041
EEAA	1.20 (7.84)	−1.25 (6.90)	.024
IEAA	0.73 (4.68)	−0.59 (4.37)	.048
GrimAge	0.44 (4.80)	−1.11 (4.85)	.031
PhenoAge	0.84 (7.60)	−1.60 (7.15)	.025
DNAmTL	7.17 (0.32)	7.26 (0.26)	.033

Abbreviations: DNAmTL, DNA methylation–estimated telomere length; EAA, epigenetic age acceleration; EEAA, extrinsic epigenetic age acceleration; IEAA, intrinsic epigenetic age acceleration.

Longer time holding a single-leg stand was associated with lower EEAA (β = −0.141; 95% CI, −.274 to −.009; P = .037) and PhenoAge (β = −0.145; 95% CI, −.278 to −.011; P = .033; Table 4). The time in seconds to complete 10 repeated chair stands was associated with greater GrimAge acceleration (β = 0.147; 95% CI, .031–.263; P = .013) and lower DNA telomere length (β = −0.008; 95% CI, −.015 to −.001; P = .032). Longer time to complete a 4-m walk in seconds was associated with higher PhenoAge (β = 2.227; 95% CI, .918–3.542; P = .001).

Table 4.

Association of Methylation-Based Biomarkers of Aging With Measures of Physical Function in Linear Regression Models

	EAA			EEAA			IEAA
Physical Function Measure	Estimate	95% CI	P Value	Estimate	95% CI	P Value	Estimate	95% CI	P Value
Grip strength, kg	0.034	−.110 to .176	.64	0.007	−.185 to .199	.94	0.013	−.106 to .132	.82
Time to do 10 repeated chair stands, s	−0.006	−.199 to .075	.37	−0.033	−.216 to .150	.72	−0.056	−.169 to .058	.33
Walk speed, s	0.397	−.608 to 1.401	.44	0.284	−1.062 to 1.631	.68	0.790	−.039 to 1.620	.06
Time to walk 400 m, m	0.505	−.503 to 1.513	.32	0.613	−.738 to 1.963	.37	0.514	−.307 to 1.370	.21
Length of single-leg stand, s	−0.075	−.174 to .024	.14	−0.141	−.274 to −.009	.037	−0.065	−.148 to .017	.12
Reach test, cm	−0.047	−.155 to .061	.39	−0.051	−.196 to .094	.49	−0.039	−.129 to .051	.39

	GrimAge			PhenoAge			DNAmTL
Grip strength, kg	−0.002	−.126 to .121	.97	−0.040	−.233 to .153	.68	0.005	−.003 to .012	.24
Time to do 10 repeated chair stands, s	0.147	.031 to .263	.013	0.108	−.076 to .292	.25	−0.008	−.015 to −.001	.032
Walk speed, s	0.654	−.209 to 1.518	.14	2.227	.918 to 3.542	.001	−0.051	−.104 to .001	.0549
Time to walk 400 m, m	0.292	−.580 to 1.164	.51	0.940	−.419 to 2.298	.17	−0.037	−.090 to .016	.17
Length of single-leg stand, s	−0.054	−.138 to .030	.21	−0.145	−.278 to −.011	.033	0.004	−.001 to .010	.10
Reach test, cm	0.068	−.025 to .162	.15	−0.051	−.198 to .095	.49	0.001	−.005 to .007	.67

Bold indicates P < .05.

Abbreviations: DNAmTL, DNA methylation–estimated telomere length; EAA, epigenetic age acceleration; EEAA, extrinsic epigenetic age acceleration; IEAA, intrinsic epigenetic age acceleration.

We then performed EWAS of the association between DNAm and physical function. No CpG sites met criteria for association (Holm-Bonferroni significance and |Δβ| > 0.05).

DISCUSSION

We report, for the first time in a sample of only women (age, 40–60 years), that WWH have a higher EAA and shorter DNAmTL when compared with women without HIV. Accelerated aging among adults with HIV has been demonstrated by using various epigenetic aging estimates in samples with mostly men [14, 15]. WWH are important to study given that globally >50% of people living with HIV are women and that women accounted for an estimated 49% of all new infections in 2021 [35]. Our finding of accelerated epigenetic age in a cohort of older WWH (mean age, 49.7 years) is an important addition to the literature on differences in aging among people with HIV by biological sex [36].

We found associations between EAA and measures of physical function, including balance and gait speed. Similar findings for some but not all the measures were observed when stratified by HIV status, possibly due to smaller sample sizes. In a study of women aged 53 to 64 years enrolled in the National Survey for Health and Development, baseline EAA was not associated with balance, in contrast to our findings [37]. A study of 63 to 76-year-old women from the Finnish Twin Study on Aging found an association between higher GrimAge and decreased performance in 6-minute and 10-m walk tests, but it did not conduct a 4-m walk test as used in our study [18].

Two prior studies revealed associations between increased epigenetic age and decreased grip strength [37, 38]. Despite these previous results, we noted no associations of any epigenetic aging measures in our study with grip strength. A recent longitudinal study did not find an association of any epigenetic aging measure with functional assessments covering different domains of aging (eg, frailty, mobility, ability to perform activities of daily living) [39]. Given inconsistent findings, additional research is needed to corroborate these findings, particularly in cohorts of WWH.

Contrary to our hypothesis, we did not find any associations between epigenetic aging measures and bone outcomes. This is consistent with a small non-HIV study of 32 individuals with osteoporosis and 16 controls, which found no association between bone parameters and HorvathAge acceleration [40]. The lack of findings in that study and ours could be due to a lack of power for these outcomes, and additional larger studies are needed.

Similar to other studies, we found a large number of CpG sites associated with HIV. The top differentially methylated CpG site (cg07839457) was located on the gene NLRC5, which encodes a transcription factor that regulates major histocompatibility complex class I molecule expression. This CpG site has been implicated in other studies of children and adults with HIV [11, 41]. In contrast to a study comparing the bone samples of 27 patients with osteoporosis and 23 with osteoarthritis that identified 241 differentially methylated CpG sites [42], our EWAS analyses did not identify any associations between BMD and DNAm from blood samples. A recent EWAS of BMD in European individuals profiled DNAm in blood and found only 1 CpG site to be significantly associated with femoral neck BMD in women: cg23196985, located on the 5′ untranslated region of CES1 [43]. These studies have yet to be replicated and used older Illumina arrays: the HumanMethylation27 BeadChip (∼27 000 CpG sites) and the HumanMethylation450 array (∼450 000 CpG sites), respectively. Genome-wide association studies of BMD by DXA in the general population have identified variants at 73 trait-associated genetic loci, including several associated with fracture risk [44, 45]. While genetic advances may pave the way for precision medicine in osteoporosis, genetic variants explain only 5% to 12% of the total phenotypic variance in BMD. Beyond genetics, many other environmental and lifestyle factors, medications, and health conditions affect BMD, such as hormone levels, tobacco and alcohol use, physical activity, and other comorbidities [46–48].

For physical function measures, there were no associations at a Holm-Bonferroni threshold and a 5% methylation difference. Similarly, other EWASs of grip strength in the general population have not reported significant findings [38, 43, 49]. Taken together, there is limited evidence to support associations between epigenetic changes and physical function. Larger studies and longitudinal studies are needed to more fully assess these potential associations.

Our study is limited by its cross-sectional design. There could be reverse causation such that a change in physical function could result in changes in DNAm, as opposed to the direction that we hypothesized. We also did not adjust for inflammatory biomarkers or other factors that could be associated with DNAm change, BMD, and physical function. Last, a major limitation for all DNAm studies is tissue specificity. Epigenetic changes may be tissue specific, and we do not have bone- or muscle-specific DNAm data. Of note, women without HIV had higher rates of tobacco use, cocaine, and alcohol use in our study. Given evidence that these substances can be associated with increased accelerated epigenetic aging, our finding of accelerated aging in WWH may be conservative [50].

In conclusion, we found evidence of associations between certain methylation-based biomarkers of aging and measures of physical function in a cohort of WWH and women without HIV, but we did not find any significant associations in EWAS analyses with either BMD or functional outcomes. Future studies will need to assess whether these findings persist longitudinally and to evaluate the directionality of these associations.

Supplementary Data

Supplementary materials are available at The Journal of Infectious Diseases online (http://jid.oxfordjournals.org/). Supplementary materials consist of data provided by the author that are published to benefit the reader. The posted materials are not copyedited. The contents of all supplementary data are the sole responsibility of the authors. Questions or messages regarding errors should be addressed to the author.

Notes

Acknowledgments. WIHS data were collected by 5 sites: Bronx (principal investigator, Kathryn Anastos), Brooklyn (principal investigators, Howard Minkoff and Deborah Gustafson), Chicago (principal investigators, Mardge Cohen and Audrey French), metropolitan Washington (principal investigator, Seble Kassaye), Connie Wofsy Women's HIV Study–Northern California (principal investigators, Ruth Greenblatt, Bradley Aouizerat, and Phyllis Tien), and overseen by the WIHS Data Management and Analysis Center (principal investigators, Stephen Gange and Elizabeth Golub).

Data availability. Data are available upon request.

Disclaimer. The contents of this publication are solely the responsibility of the authors and do not represent the official views of the National Institutes of Health.

Financial support. This work was supported by the National Institute of Allergy and Infectious Diseases, with cofunding from the Eunice Kennedy Shriver National Institute of Child Health and Human Development, the National Cancer Institute, the National Institute on Drug Abuse, and the National Institute on Mental Health (Bronx WIHS, U01-AI-035004; Brooklyn WIHS, U01-AI-031834; Chicago WIHS, U01-AI-034993; metropolitan Washington WIHS, U01-AI-034994; Connie Wofsy Women's HIV Study–Northern California, U01-AI-034989 to P. C. T.; WIHS Data Management and Analysis Center, U01-AI-042590); the National Institute on Drug Abuse (K01-DA-053157 to S. S.); the Eunice Kennedy Shriver National Institute of Child Health and Human Development (R21-HD-104558, R01-HD-111550 to S. S.); and the National Institute of Allergy and Infectious Diseases (R01-AI-095089 and K24-AI-155230 to M. T. Y.).

Author notes

Potential conflicts of interest. All authors: No reported conflicts.