Monocyte Epigenetic Age Acceleration is Linked to Non-Somatic Depressive Symptoms in Women with and Without HIV

Abstract

Depression disproportionately affects women living with HIV, yet symptom heterogeneity and the lack of observable biomarkers can impede detection. Accelerated aging of monocytes—key innate immune cells—may contribute to depression, particularly in this population. A DNA methylation clock, MonoDNAmAge, estimates monocyte biological age and has shown evidence of epigenetic age acceleration (EAA) in women with HIV. Here, we examine MonoDNAmAge as a biomarker of depression in women with and without HIV, differentiating non-somatic from somatic symptom domains. DNA methylation data and Center for Epidemiologic Studies Depression Scale (CES-D) scores were available from 440 Women’s Interagency HIV Study participants. Two biological age estimates (HorvathDNAmAge and MonoDNAmAge) were calculated and orthogonalized with chronological age. In the total sample and subsamples stratified by HIV status, we used multiple linear regression to assess how EAA_Mono and EAA_Horvath were associated with depressive symptoms. Standardized β coefficients are reported. The sample included 261 women with HIV (mean chronological age=43.7 (8.9) years; 38% Black; 48% Hispanic) and 179 women without HIV (mean chronological age=39.5 (10.0) years; 31% Black; 49% Hispanic). In the overall sample, EAA_Mono was associated with the non-somatic depressive symptom domain (β=0.125, p=0.018), and anhedonia specifically (β=0.354, p=0.007), adjusting for HIV, race, and ethnicity. This pattern persisted in the subsample with HIV (β=0.112, p=0.085). EAA_Horvath was not associated with depression severity or symptom domains. Monocyte aging may represent a sensitive biomarker of non-somatic depression symptoms in women with HIV. The dynamics of monocyte aging and depression warrant further study to clarify mechanistic links.

Introduction

Depression—an often chronic and debilitating mental disorder—is known to significantly increase morbidity, mortality, and diminish quality of life in persons living with HIV(1–4). Moreover, depression is two to three times more common in women with HIV than in men or unaffected women, with rates ranging from 35-60% (3–5). In addition to the suffering and impairment caused by depression, its downstream effects include decreased antiretroviral therapy (ART) adherence, reduced CD4⁺ T cell counts, and increased viral transmission (2,3,6). Therefore, timely recognition and effective treatment of depression is a critical goal in HIV care, vital to reach public health targets.

Depression is a heterogeneous phenotype, with over 1,500 possible symptom combinations spanning multiple domains (i.e., affective, cognitive, somatic) (7). Symptoms can include depressed mood, anhedonia, excessive feelings of guilt, hopelessness, worthlessness, changes in sleep or appetite, fatigue, impaired concentration, and psychomotor slowing or agitation (8). Unlike other conditions of similar scope and burden (e.g., type 2 diabetes), the identification of depression is based strictly on reported symptoms and ruling out other possible causes (e.g., hypothyroidism, substance withdrawal, bereavement) rather than confirmatory lab or physiologic testing (8). The heterogeneity of depression, combined with the lack of indicative biomarkers, is a barrier to timely clinical identification and the early initiation of strategies to mitigate severity and prevent chronicity (9,10).

Specificity and Precision: Linking Biological Markers to Depression in HIV

Although the underlying biological pathways that link depression and HIV co-occurrence are not fully understood, inflammation and accelerated biological aging have been proposed as explanatory mechanisms. In populations without comorbidities, depression has been associated with altered monocyte composition, specifically an increase in non-classical monocytes, and this compositional shift is also typical in HIV (11,12). It has been suggested that these mechanisms may be differentially relevant to specific depressive phenotypes (13,14). For example, an inflammatory or immunometabolic phenotype of depression has been characterized by symptoms of anhedonia, fatigue, hypersomnia, and poorer metabolic health (e.g., higher BMI, dyslipidemia)—underscoring the importance of examining specific symptoms rather than diagnostic categories alone when evaluating their links to biological markers (9,15).

Chronic HIV and its associated neuropsychiatric complications, including depression, are associated with accelerated biological aging using several different DNA methylation (DNAm)-based epigenetic clocks (16–18). These clocks apply the weighted value of DNAm at specific points (i.e., CpG sites) throughout the human genome to estimate biological age (DNAmAge) in contrast to chronological age (years lived). Although more sensitive and precise than previously proposed biological aging measures (e.g., telomere length), DNAm-based clocks vary in both specific CpG sites used to estimate aging as well as tissue type used to derive the component CpG sites that estimate biological aging. The Horvath clock (HorvathDNAmAge), for example, is a multi-tissue clock based on 353 CpG sites, while the Hannum clock (HannumDNAmAge) is comprised of 71 CpGs derived from leukocytes, with only 6 CpGs overlapping (19). Later clocks, such as PhenoDNAmAge and GrimDNAmAge, respectively include 513 and 1030 CpGs (with 41 and 1 CpG overlapping with the Horvath clock, respectively) and incorporate laboratory markers. The past decade has seen the rapid development of numerous clocks, but there is growing recognition that a principled rather than opportunistic approach in the match between clocks and clinical endpoints is warranted. Selection of specific clocks should be based on biological plausibility and the nuances of targeted clinical endpoints.

By design, all epigenetic clocks are highly correlated with chronological age. Therefore, epigenetic age acceleration (EAA), defined as the residuals of regressing DNAmAge on chronological age, is what signifies deviations in the rate of aging. HIV has shown associations with an increased EAA measured by the Horvath clock, in line with the early signs of aging experienced by many people living with HIV (16), but findings regarding EAA and depression have been less consistent. Some studies suggest that depression and depressive symptoms are associated with EAA (17,20,21), but considerable variation exists in the sample population, tissue type, laboratory methods, and the choice of DNAmAge clock. Domains of depressive symptomatology also show differential associations with EAA (17,20), but this heterogeneity is rarely considered. A recent study showed that medication-naïve patients with depression had significantly higher EAA determined by Hannam, GrimAge, and PhenoAge clocks, as well as a marked decrease in DNAm-based natural killer cells—a particularly interesting finding in the context of HIV, in which the function of natural killer cells is impaired (21).

Consideration of Cell or Tissue Inflammation in Depression-Related Aging Clocks

A consistent limitation across prior studies using epigenetic clocks is the lack of specificity for the tissue and cell type used. Over the past decade, the role of inflammation in depression has become increasingly appreciated. Pro-inflammatory states, such as the milieu created by HIV, can activate the degradation of tryptophan, a precursor to serotonin, via the indoleamine 2,3-dioxygenase (IDO) pathway (22,23). In this context, tryptophan is shunted toward kynurenine production and its downstream neurotoxic metabolites (e.g., quinolinic acid), resulting in the reduced synthesis and availability of serotonin implicated in depression (24). Monocytes are key innate immune cells involved in inflammation and demonstrate age-related genetic dysfunction in inflammation, metabolism, and immunity (25,26). Monocytes also play a significant role in the pathophysiology of HIV progression, including viral spread early in infection, as a viral reservoir, and in neuropsychiatric complications (27–30). A cell-type-specific clock, such as one derived from monocytes, should minimize potential confounding from different cell types that have distinct DNAm profiles, which may capture different facets of biological aging—reducing one source of heterogeneity (19). Recently, a novel epigenetic clock derived from monocytes (MonoDNAmAge) demonstrated EAA in women with HIV who were heavy alcohol users (19), but the clock’s links to depression and other monocyte-implicated conditions remain unexplored.

Aims and Hypothesis of the Current Study

To address these gaps, we examined monocyte EAA (EAA_Mono) as a potential biomarker of depression among women with and without HIV. Additionally, we focused on the links between EAA and domains of depressive symptoms (i.e., non-somatic versus somatic). Across all participants, we hypothesized that EAA in monocytes would be associated with higher depressive symptoms overall, with a stronger association noted than when using previously proposed clocks (e.g., Horvath clock), which lack tissue specificity. Due to the prevalent co-occurrence of inflammation and depressive symptoms among women with HIV, we anticipated that stronger associations between EAA in monocytes and depressive symptoms would be observed for women with HIV compared to women without HIV.

Methods and Materials

This analysis leveraged data from the Women’s Interagency HIV Study (WIHS) to examine EAA_Mono as a potential biomarker of depressive symptomatology among women with and without HIV. The WIHS is a prospective cohort study that examines the course and impact of HIV in women treated with antiretroviral therapy. The study includes an approximate 2.5:1 ratio of women with and without HIV (31,32). Since its establishment in 1994, the WIHS cohort has grown to over 2000 active participants across centers in Atlanta, GA; Birmingham AL; Brooklyn, NY; Bronx, NY; Chapel Hill, NC; Chicago, IL; Jackson, MS, Los Angeles, CA; Miami, FL; San Francisco, CA; and Washington, DC. Data were collected biannually at study visits involving comprehensive physical examinations, blood collection, and surveys to assess demographics, social factors, health history, and current symptoms. In 2019, the WIHS merged with the Multicenter AIDS Cohort (MACS), a study of men with and at risk for HIV acquisition, to form the MACS/WIHS Combined Cohort Study (MWCCS) (33). Women with DNAm data and Center for Epidemiologic Studies Depression Scale (CES-D) scores from the same WIHS visit window between 1996 and 2015 were included in this analysis (34).

Measures

Depressive Symptoms

The Center for Epidemiologic Studies Depression Scale (CES-D) is a validated and widely used 20-item, self-administered questionnaire commonly used in HIV studies (34,35). This scale assesses the frequency of depressive symptoms over the past week, and scores (ranging from 0-60) ≥ 16 are often used as a cut point for possible depressive disorder (36). In this study, we evaluated three sets of outcomes derived from the CES-D: (1) the total score, (2) summary scores of a) somatic domain items and b) non-somatic domain items, and (3) individual items (encoded as binary variables such that item responses of “rarely” or “some” = 0 and “occasionally” or “most” = 1). Aligned with previous factor analytic work, somatic domain symptoms (items 1, 2, 5, 7, 11, 13, and 20) in the CES-D encompass unpleasant or worrisome bodily sensations—such as changes in sleep, appetite, and concentration—while non-somatic symptoms (the rest of the items) include affective disturbances (e.g., anhedonia, negative affect) and interpersonal difficulties (See Table 1). Items 4, 8, 12, and 16 were reverse scored, so that higher scores consistently reflect greater severity of depressive symptoms (34,37).

Table 1.

Symptoms & Domain Classification of Center for Epidemiological Studies-Depression (CES-D)

Symptom & Domain	Item	Item no.
Non-Somatic
Blues	I felt I could not shake off the blues even with help from my family or friends	3
Self-worth	I felt I was just as good as other people*	4
Depressed Mood	I felt depressed	6
Hopelessness	I felt hopeful about the future*	8
Failure	I thought my life had been a failure	9
Fearful	I felt fearful	10
Happiness	I was happy*	12
Loneliness	I felt lonely	14
Negativity	People were unfriendly	15
Anhedonia	I enjoyed life*	16
Crying	I had crying spells	17
Sadness	I felt sad	18
Feeling disliked	I felt that people dislike me	19
Somatic
Bothered	I was bothered by things that usually don’t bother me	1
Poor Appetite	I did not feel like eating; my appetite was poor	2
Poor Concentration	I had trouble keeping my mind on what I was doing	5
Perceived Effort	I felt everything I did was an effort	7
Sleep Disturbance	My sleep was restless	11
Poverty of Speech	I talked less than usual	13
Motivation	I could not get “going”	20

Note:

^*Reverse-coded; Participants are asked to rate the frequency of each item during the past week

Covariates

Sociodemographic and clinical data, including depressive symptoms, were leveraged from the WIHS visit window coinciding with DNA specimen collection. Sociodemographic data included age and self-reported race and ethnicity. Clinical data included HIV serostatus, current CD4⁺ T cell count (cells/mm^3), blood HIV RNA viral load (seropositive only), number of alcoholic drinks consumed per week, current smoking, and use of marijuana, and use of cocaine, freebase, or heroin since the last study visit.

Laboratory Analysis

Illumina Human Methylation EPIC BeadChip (EPIC) microarrays were used to measure methylation levels in bisulfite-converted DNA from peripheral blood mononuclear cells (PBMCs). Samples were processed at the Yale Center for Genomic Analysis. DNAm analysis procedures have been previously described (19). Briefly, fluorescent signals from methylated and unmethylated probes indicated the methylation level at each CpG site (ß = max [M,0/|U| + |M| + 100]) using a detection p-value cut-off of 0.05. CpGs of single-nucleotide polymorphisms (SNPs) or those that overlap with repetitive elements or regions were filtered. Raw methylation data were retrieved using the minifi R package (version 1.18.1) and then normalized and batch-corrected.

Two estimates of biological age were calculated (DNAmAgeHorvath and DNAmAgeMono). Both clocks used elastic net regression. The Horvath clock used over 8000 samples from 82 publicly available Illumina datasets. ß values from 353 CpG sites specified by the Horvath model are used to estimate biological age and age acceleration (38). MonoDNAmAge was developed and validated across four large independent cohorts, one of which was the WIHS (19). MonoDNAmAge was derived from the CD14⁺ monocyte methylome, reducing heterogeneity from multiple cell types. The 186 CpGs included in MonoDNAmAge were selected as they showed maximum correlation between DNAmAge and chronological age (19).

Statistical Analyses

Descriptive statistics, presented as mean (SD), were calculated for the overall sample and subsamples of participants with and without HIV. Using each outcome of interest as the dependent variable, we conducted a series of linear (CES-D total score, somatic and non-somatic scores) and logistic regression (i.e., individual CES-D items) analyses stratified by HIV serostatus and RNA viral load, defining four groups: (1) the overall sample, (2) women without HIV, (3) women living with HIV (4) and women living with HIV with undetectable viral loads (a subset of the third group). HIV serostatus was included as a predictor in the overall sample analyses, whereas HIV RNA viral load detection status was included only in the third group. Based on prior research suggesting their associations with depressive symptoms among people with HIV, we adjusted in regression models for the following covariates in each analysis: race (Black / not Black), ethnicity (Hispanic / not Hispanic), current smoking status, number of alcoholic drinks consumed per week, use of marijuana, cocaine, or heroin, and CD4⁺ T cell count (cells/mm^3).

We used correlation, unpaired two-sample t-tests, and Fisher’s exact tests where appropriate to select the most relevant covariates to include, due to the limited sample size. All continuous variables were standardized to ensure equal weighting in the models. Additionally, due to high multicollinearity among the three age measures (chronological age, HorvathDNAmAge, and MonoDNAmAge), we applied the Gram-Schmidt orthogonalization process to eliminate linear dependencies and isolate the unique contribution of each variable. Specifically, we first kept chronological age unchanged. Then we regressed HorvathDNAmAge on chronological age and used the residuals to calculate EAA for the Horvath clock (EAA_Horvath). Next, we regressed MonoDNAmAge on both chronological age and the residualized HorvathDNAmAge, with the residuals representing EAA for the Mono clock (EAA_Mono) to reflect any additional (rather than overlapping) contribution of this clock compared to HorvathDNAmAge. These orthogonalized EAA measures were included in multivariable regression models along with covariates, ensuring that the coefficients for EAA_Horvath and EAA_Mono reflected their unique contributions beyond the other age variables. Because EAAHorvath and EAAMono were constructed via sequential residualization and included jointly in each model, they represent distinct, non-overlapping components of biological aging rather than multiple correlated ‘clock’ tests. Results from the overall and HIV-stratified samples are reported as conditional subgroup-specific estimates to aid interpretation. Specifically, the four analytic subsets represent pre-specified, clinically meaningful strata. Their results are presented as descriptive, stratum-specific estimates rather than as independent discovery analyses; our primary inference was based on the overall model. Consequently, we did not apply a multiple-comparison correction. The analyses were performed in R (version 4.4.0) (39).

Results

Descriptives

The overall sample of women in this analysis (N=440) had a mean (M) chronological age of 42.0 (SD=5.9) years, and many identified as Black (35%) and as Hispanic (48%). The total sample of women with and without HIV was split: N=261 (chronological age M=43.7 (8.9) years; 38% Black; 48% Hispanic) and N=179 (age M=39.5 (10.0) years; 31% Black; 49% Hispanic), respectively. CES-D total scores were similar among women with and without HIV. Table 2 displays additional descriptives, including HorvathDNAmAge and MonoDNAmAge in years for the overall sample and subsamples. Standardized coefficients are reported for the following analyses.

Table 2.

Demographic, Clinical, and Aging-Related Characteristics of the Overall and Stratified Samples

Characteristic	Overall Sample (N=440) N (%)	WWoH (N=179) N (%)	WWH (N=261) N (%)
Age in years
25-35	116 (26%)	72 (40%)	44 (17%)
36-45	170 (39%)	62 (35%)	108 (41%)
46-55	120 (27%)	32 (18%)	88 (34%)
≥55	34 (8%)	13 (7%)	21 (8%)
Horvath DNA m-based age in years
25-35	102 (23%)	74 (41%)	28 (11%)
36-45	172 (39%)	64 (36%)	108 (41%)
46-55	131 (30%)	36 (20%)	95 (36%)
≥55	35 (8%)	5 (3%)	30 (11%)
Monocyte DNA m-based age in years
25-35	198 (45%)	134 (75%)	64 (25%)
36-45	102 (23%)	27 (15%)	75 (29%)
46-55	79 (18%)	15 (8%)	64 (25%)
≥55	61 (14%)	3 (2%)	58 (22%)
Race
Black	153 (35%)	55 (31%)	98 (38%)
Non-Black	287 (65%)	124 (69%)	163 (62%)
Ethnicity
Hispanic	212 (48%)	87 (49%)	125 (48%)
Non-Hispanic	228 (52%)	92 (51%)	136 (52%)
CD4N, median (IQR)	662.5 (386.75, 938.25)	1015 (782.25, 1247.75)	492 (337.5, 646.5)
Depressive symptoms (CES-D score), M (SD)	12.3 (11.8)	11.3 (11.8)	12.9 (11.8)
CES-D score ≥16, n (%)	136 (31%)	46 (26%)	90 (34%)
Number of drinks/week since last visit, M (SD)	1.8 (5.5)	3.1 (7.4)	1 (3.4)
Current smoking, n (%)	163 (37%)	87 (49%)	76 (29%)
Use marijuana/hash since last visit, n (%)	60 (14%)	35 (20%)	25 (10%)
Use crack/freebase/cocaine/heroin since last visit, n (%)	10 (2%)	8 (4%)	2 (1%)

Note. WWoH= women without HIV; WWH = women living with HIV; CD4N= Current CD4⁺ T cell count (cells/mm^3), IQR= interquartile range; CES-D= Center for Epidemiologic Studies Depression Scale.

Total Depressive Symptom Score Analysis

In the overall sample, higher EAA_Mono (β= 0.111, p=0.036) and higher chronological age (β=0.131, p=0.009) and were significantly associated with higher total depressive symptom scores in the adjusted model (see Figure 1, Table 3). Chronological age was significantly associated with total depressive symptom scores (β=0.143, p=0.026) among the women with HIV subsample only. Neither EAA_Horvath nor EAA_Mono were associated with total depressive symptoms scores in subsamples.

Figure 1. Standardized Estimates Showing Associations Between Epigenetic Age Acceleration and Total Depressive Symptom Scores.

Note. WWoH= women without HIV; WWH = women living with HIV; WWH-UD= women with HIV with undetectable viral loads; AGE= chronological age; Horvath = EAA as determined by the Horvath multi-tissue epigenetic clock; Mono= EAA as determined by the Monocytespecific epigenetic clock

Table 3.

Standardized Coefficients Showing Associations Between Epigenetic Age Acceleration and Total Depressive Symptom Scores, Somatic Depressive Domain Scores, and Non-Somatic Depressive Domain Scores in the Overall Sample and Among Stratified Sub-groups of Interest

	Overall Sample (N=440)		WWoH (N=179)		WWH (N=261)		WWH-UD (N=219)
	Estimate	p-value	Estimate	p-value	Estimate	p-value	Estimate	p-value
Total Depressive Symptom Scores
HIV serostatus	0.077	0.577	-	-	-	-	-	-
Viral load	-	-	-	-	−0.210	0.230	-	-
CD4N	0.050	0.438	0.059	0.447	0.049	0.462	0.081	0.251
Black	−0.130	0.259	−0.263	0.161	−0.045	0.764	−0.161	0.327
Hispanic	0.161	0.154	−0.051	0.767	0.317	0.042	0.246	0.144
Chronological age	0.131	0.009 b	0.091	0.243	0.143	0.026 a	0.086	0.225
EAAHorvath	0.044	0.396	−0.041	0.596	0.050	0.427	0.077	0.261
EAAMono	0.111	0.036 a	0.048	0.531	0.100	0.129	0.114	0.104
Somatic Depressive Domain Scores
HIV serostatus	0.074	0.593	-	-	-	-	-	-
Viral load	-	-	-	-	−0.275	0.121	-	-
CD4N	0.064	0.326	0.085	0.276	0.056	0.407	0.075	0.295
Black	−0.016	0.887	−0.166	0.378	0.048	0.754	−0.092	0.583
Hispanic	0.055	0.632	−0.004	0.982	0.093	0.553	−0.046	0.789
Chronological age	0.133	0.009 b	0.076	0.333	0.161	0.013 a	0.111	0.124
EAAHorvath	0.005	0.918	−0.021	0.793	0.005	0.937	0.029	0.678
EAAMono	0.066	0.220	0.006	0.934	0.058	0.382	0.058	0.417
Non-Somatic Depressive Domain Scores
HIV serostatus	0.071	0.607	-	-	-	-	-	-
Viral load	-	-	-	-	−0.148	0.391	-	-
CD4N	0.038	0.560	0.041	0.597	0.039	0.551	0.075	0.280
Black	−0.178	0.118	−0.291	0.120	−0.094	0.532	−0.180	0.264
Hispanic	0.203	0.071d	−0.071	0.679	0.408	0.008 b	0.377	0.023 a
Chronological age	0.117	0.019 a	0.091	0.239	0.116	0.067	0.062	0.372
EAAHorvath	0.061	0.239	−0.049	0.531	0.069	0.261	0.094	0.161
EAAMono	0.125	0.018 a	0.065	0.391	0.112	0.085d	0.131	0.056d

Note. . WWoH= women without HIV; WWH = women living with HIV; WWH-UD= women with HIV with undetectable viral loads; CD4N = Current CD4⁺ T cell count (cells/mm^3); EAA= epigenetic age acceleration; EAA_Horvath = EAA as determined by the Horvath multi-tissue epigenetic clock; EAA_Mono = EAA as determined by the Monocyte-specific epigenetic clock;

^a=p<0.05;

^b=p<0.01;

^c=p<0.001;

^d=p<0.1

Somatic and Non-Somatic Domain Score Analysis

In the overall sample, EAA_Mono was significantly associated with non-somatic depressive domain scores (β=0.125, p=0.018; Figure 2, Table 3). This pattern persisted among women with (β=0.122, p=0.085) but not without HIV (β=0.065, p=0.391). Neither clock was associated with somatic depressive domain scores in the overall or stratified subsamples.

Figure 2. Comparison of Standardized Estimates Showing Associations Between Epigenetic Age Acceleration with Somatic and Non-Somatic Depressive Domain Symptoms Scores.

a) Somatic Symptom Domain Scores

b) Non-somatic Symptom Domain Scores

Note: WWoH= women without HIV; WWH = women living with HIV; WWH-UD= women with HIV with undetectable viral loads; AGE= chronological age; Horvath = EAA as determined by the Horvath multi-tissue epigenetic clock; Mono= EAA as determined by the Monocytespecific epigenetic clock

Individual Symptom Analysis

EAA_Mono was significantly associated with specific CES-D items in the overall sample, particularly anhedonia (β=0.354, p=0.007), hopelessness (β=0.344, p=0.007), feelings of failure (β=0.335, p=0.047), and sleep disturbance (β=0.289, p=0.016) (eFigure 1, eTable 1). These patterns were fairly consistent in the subsample of women with HIV (all p<0.05, except hopelessness), but not among women without HIV. EAA_Mono was negatively associated with a lack of motivation or drive (β=−0.625, p=0.019), among women without HIV. By contrast, EAA_Horvath was associated with negative affect items (i.e., inability to shake ‘the blues’ [β=0.285, p=0.049], crying [β=0.312, p=0.037]) and sleep disturbance [β=0.230, p=0.048]) in the overall sample and showed non-significant findings for all items across stratified subsamples.

Discussion

Previous efforts to link accelerated biological aging with depression using DNAm-based clocks have been limited by a lack of cell or tissue specificity. In this study, we leveraged a monocyte-derived epigenetic clock in a sample at heightened risk for depression—women with and without HIV—and demonstrated that monocyte age acceleration was selectively associated with specific depressive symptoms, particularly anhedonia, hopelessness, and feelings of failure as well as non-somatic symptoms overall. This relationship persisted among the smaller subsample of women with HIV, most of whom were virally suppressed, but was shy of our p< 0.05 cut-off for statistical significance. In contrast, the Horvath clock was not associated with total symptom scores or domain-specific symptom scores. These findings support the utility of monocyte-specific clocks in samples with HIV and domain-specific approaches to depression phenotyping in uncovering meaningful associations.

Although monocyte aging was associated with total depression scores in the overall sample, effect sizes were small to medium and did not reach statistical significance in stratified models, possibly due to reduced power. Importantly, the Horvath clock also failed to show significant associations, suggesting that the monocyte clock may be more sensitive to depression among women with or at risk for HIV, consistent with our hypothesis, which was informed by the biological plausibility of the role of monocytes. Moreover, our findings represent an advance in what has been done previously, as we accounted for the independent contributions of these correlated variables in our estimates. Our analysis of depression and EAA in monocytes is novel, and several studies using the Horvath clock have demonstrated null associations between EAA_Horvath and total depression scores in demographically similar samples and across different depression instruments (17,20). Findings are less consistent in studies that compare EAA_Horvath across major depressive disorder (MDD) cases and controls (21,40). When EAA is assessed in alternative clocks (e.g., Hannum, PhenoAge, GrimAge), this inconsistency among case-control studies remains present, with some investigators reporting positive (21,41) or non-significant associations (40,42), underscoring the importance of study design. As a diagnosis, MDD may have clinical utility, but our results are aligned with the notion that it is an overly broad label for research aimed to elucidate underlying mechanisms, as this label lumps together diverse presentations.

Our analysis juxtaposes somatic with non-somatic symptom domains to move beyond imprecise and limiting diagnostic categories. In doing so, we found that monocyte age acceleration was differently associated with non-somatic (i.e., affective and interpersonal) domain symptoms over somatic domain symptoms overall. Similarly, among a socioeconomically diverse sample of White and African-American middle-aged men and women (N=329), Beydoun et al. (2019) reported that EAA was positively associated with reduced positive affect, as quantified by the CES-D subscale. Interestingly, several items comprising the CES-D subscale assessing lack of positive affect (i.e., anhedonia and hopelessness about the future) were positively associated with monocyte age acceleration in our item-level analysis. Additionally, our findings that monocyte age acceleration was related to anhedonia, hopelessness, and feelings of failure, while the Horvath clock was related to feeling down and crying, support the notion that inconsistent findings across the literature may stem from the possibility that different clocks may be sensitive to distinct “depressions,” a nuance which is lost when symptom totals across multiple domains are the sole outcome variable. The symptoms reflecting anhedonia were strongly associated with monocyte age acceleration, which is in step with the proposed inflammatory or immunometabolic depression subtypes (14,15). However, other features of this proposed subtype such as hypersomnia, hyperphagia, and affective reactivity are not captured by CES-D.

To the best of our knowledge, this is the first study that has examined EAA links to depression among women with HIV, and the role of HIV in this relationship is not well understood. The association observed between non-somatic domain symptoms and EAA_Mono may be driven in part by the high proportion of women with HIV in the overall sample; however, this relationship was not significant in the smaller, HIV subsample, which was likely underpowered. A majority of the women with HIV in our study had undetectable HIV RNA viral loads, suggesting that associations may not be strictly related to the progression of HIV. This finding is in line with prior research that suggests that even with adequate ART treatment and adherence, people living with HIV have an early onset of major, chronic, age-related comorbidities (43). Others have suggested the role of monocytes, particularly intermediate monocytes, in neuropsychiatric complications among virally suppressed people with HIV, specifically processing speed and executive function (30). In addition, prior research demonstrates higher levels of intermediate and non-classical monocytes in MDD, and suggests therapeutic effects of ketamine through monocyte pathways (11). While speculative, it is possible that EAA_Mono and its association with depression may reflect the shift toward intermediate and non-classical monocytes that occur with both aging and HIV infection (12). These CD16+ monocytes are vitally important to the development of latent HIV reservoirs in the brain (44,45). The noted association between the EAA_Mono and depression among the subsample with undetectable viral loads may be related to this reservoir and monocyte composition. Alternatively, it is also possible that while the latest generation of ART is less neurotoxic, remaining metabolic alterations may contribute to aging and depression-linked biological pathways through alternative immune and inflammatory pathways (44). Whether links between monocyte aging and depression relate to additional inflammatory or physiologic pathways or stress-related experiences of living with HIV warrants further study.

It is important to note that contributors to accelerated aging and depression are complex and multi-factorial. Especially pertinent to women living with HIV, psychosocial factors cannot be ignored as this population is more likely to experience food or financial insecurity, intimate partner violence, and discrimination based on several intersecting social identities, which can contribute to depression (46). Such psychosocial factors have also been proposed to accelerate aging, described as embodiment or biological weathering, which refers to how one literally and biologically incorporates the material and social world in which one lives, as such experiences contribute to early health deterioration (47,48). Future longitudinal studies incorporating social and environmental factors into further examinations of accelerated aging and depression are warranted to help elucidate mechanisms of health gaps and identify strategic intervention targets specific to those most at risk for depression and its far-reaching effects.

Strengths and Limitations

Although this analysis represents a novel contribution and addresses several critical literature gaps, limitations must be considered in the interpretation of our findings. Firstly, we were limited by the nature of cross-sectional data to distinguish cause from consequence, but our unique findings concerning domain-specific (i.e., non-somatic) symptoms and EAA set the stage for future longitudinal analyses. Secondly, while the CES-D is a validated and widely used instrument, it does not comprehensively capture all relevant depressive symptoms, especially those indicative of atypical presentations such as affect reactivity, hypersomnia, leaden paralysis, or increased appetite (15). Still, use of this instrument enables comparison across many large datasets. Further, our findings can be more seamlessly compared to the abundance of studies using CES-D, as stark differences in content across multiple measures of depressive symptoms have been noted (49,50). Relatedly, data limitations prevented us from analyzing EAAMono with respect to observable clinical endpoints related to depression, such as psychiatric hospitalization and suicide. Finally, correction for multiple comparisons was not performed for our item-level analysis, as this was exploratory rather than hypothesis testing. Regardless, many of the items associated with monocyte age accelerations (e.g., anhedonia, hopelessness) were significant at p<0.01 and would have approached significance even with highly conservative Bonferroni corrections. Although these findings need to be validated, our efforts to reduce cell type and phenotypic heterogeneity in our analysis, combined with our use of orthogonalization to understand the added contribution of monocyte aging, are notable strengths upon which future studies can build. Finally, our focus on women with or at risk for HIV centers a demographically understudied population bearing disproportionate risk for depression, while revealing additional insights across HIV-stratified samples.

Conclusions and Implications

Monocyte-derived epigenetic clocks may serve as sensitive biomarkers of non-somatic depressive symptoms in women with HIV. The use of cell-specific clocks combined with attention to distinctions across depressive symptoms may also help disentangle the underlying biological mechanisms at play in specific depressive phenotypes. Such knowledge is critical to moving science forward and has future implications for guiding diagnosis and treatment decisions— reducing guesswork and trial-and-error approaches to interventions that remain common in practice. Our study expands on previous knowledge of depression heterogeneity and the biological underpinnings of symptoms in a population that bears disproportionate risk of depression and provides new insights to further the goals of precision mental health care and improving the lives and longevity of women living with HIV.