The Role of CD4⁺ T cell Differentiation in Cytomegalovirus-associated Parahippocampal Volume Reduction and Depressive Symptoms: A Dual-Sample Study

Abstract

Cytomegalovirus (CMV) establishes lifelong infection and promotes an immunosenescent T cell profile characterized by depletion of naïve cells and expansion of terminally differentiated effector memory cells re-expressing CD45RA (TEMRA). We examined whether CMV-related T cell differentiation was associated with brain structure and depressive symptoms in two complementary cohorts. In a neuroimaging cohort of 129 adults, higher CD4⁺ TEMRA percentages was associated with lower parahippocampal volume (β = −0.27, 95% CI [−0.42, −0.11], p_FDR = 0.022). However, no T cell measure significantly mediated the association between CMV serostatus and parahippocampal volume after FDR correction. In a population-based sample of 5,759 older adults from the Health and Retirement Study, higher CD4⁺ TEMRA:naïve ratios, reflecting relative expansion of TEMRA cells and depletion of naïve CD4⁺ T cells, were associated with greater depressive symptom severity across three assessments from 2016 to 2020 (β = 0.097, 95% CI [0.074, 0.119], p_FDR < 0.001). Structural equation models indicated significant indirect associations between CMV seropositivity and depressive symptoms through the CD4⁺ TEMRA:naïve ratio and CD4⁺ naïve cells. These findings link CMV-related T cell differentiation to reduced parahippocampal volume and increased depressive symptoms. Larger longitudinal studies integrating immunophenotyping, neuroimaging, and clinical outcomes in the same participants are needed to test this proposed pathway.

1. Introduction

Major depressive disorder (MDD) is a leading global cause of disability, yet its biological underpinnings remain incompletely defined (Ford et al., 2020; Herrman et al., 2022; Vos et al., 2020). Converging evidence from clinical, genetic, and experimental studies indicates that chronic inflammation may be a key mechanistic pathway underlying biologically distinct subtypes of depression (Beurel et al., 2020; Dantzer et al., 2008; Lynall et al., 2022; Lynall et al., 2020; Miller and Raison, 2016). Human cytomegalovirus (CMV), a herpesvirus that infects up to 75% of adults and establishes lifelong chronic infection (Zuhair et al., 2019), may contribute to this inflammatory vulnerability (Zheng and Savitz, 2022, 2023). Epidemiological studies have linked CMV seropositivity to increased depression risk (Zheng and Savitz, 2022). For instance, a population study of over 80,000 individuals found CMV infection precedes depression onset (Simanek et al., 2014). Extending these findings, we previously found that CMV seropositivity in MDD is linked to reduced white matter integrity (Zheng et al., 2021a), lower orbitofrontal and temporal gray matter volume (Zheng et al., 2020; Zheng et al., 2021b), and weakened salience–sensorimotor functional connectivity (Zheng et al., 2021b). These frontotemporal circuits are crucial for emotion and reward processing (Dolcos et al., 2004; Olson et al., 2007). More recently, we observed increased signs of neuroinflammation (i.e., increased expression of inflammation-related genes, more “activated” microglia) in CMV-positive postmortem samples; an increased risk for mood disorders and suicidal behavior was also associated with CMV-positivity, as determined by retrospective clinical history (Zheng et al., 2023). However, the immunological pathway that might connect CMV to brain alterations and depression remains unclear.

One leading hypothesis highlights CMV’s significant impact on the adaptive immune system; CMV infection promotes the expansion of terminally differentiated effector memory T cells re-expressing CD45RA (TEMRA) and a corresponding decline in naïve T cell populations across both CD4⁺ and CD8⁺ lineages (Ford et al., 2020; Klenerman and Oxenius, 2016). This skewed, “immunosenescent” T cell profile, often captured by the reduced naïve to memory ratio, is associated with diminished immune responsiveness and chronic low-grade inflammation (Klenerman and Oxenius, 2016; Nicoli et al., 2022). In humans, psychosocial stress is associated with a shift toward this immunosenescent T cell profile, a pattern that appears partially driven by CMV reactivation (Klopack et al., 2022). In experimental studies using mice, it has been demonstrated that T cells can access the CNS to alter white matter integrity and drive behavioral changes (Fan et al., 2019; Groh et al., 2025; Groh et al., 2021). Maladaptive microglia activation, an indicator of neuroinflammation, in aged mice has been shown to recruit cytotoxic CD8⁺ effector T cells to infiltrate white matter, causing axon degeneration (Groh et al., 2025). Meanwhile, stress-induced metabolic reprogramming of naïve CD4⁺ T cells drives excess production of xanthine (Fan et al., 2019), which has been clinically shown to be elevated in patients with depression (Ali-Sisto et al., 2016). Together, these findings suggest that CMV-associated T cell differentiation may be related to brain structural vulnerability and depressive symptoms.

The current study investigates whether CMV-related shifts in peripheral T cell differentiation, specifically higher TEMRA to naïve cell ratios, are associated with structural brain alterations and depressive symptoms in two complementary samples. In Sample 1, an exploratory whole-cortex neuroimaging analysis tested the hypothesis that a more differentiated T cell profile (i.e., CD4⁺ TEMRA:naïve ratios) would be associated with lower regional gray matter volume. Separate CD4⁺ naïve and TEMRA percentages were examined as secondary component measures, and the corresponding CD8⁺ measures were examined secondarily to assess lineage specificity. In Sample 2, longitudinal data from the Health and Retirement Study were used to test the hypothesis that the same T cell profile would be associated with greater depressive symptom severity across three assessments spanning 2016 to 2020. Within each sample, mediation analyses further tested whether T cell composition statistically accounted for the association between CMV serostatus and the respective outcome. By integrating immunophenotyping, neuroimaging, and longitudinal epidemiological data, we examined whether CMV-associated T cell differentiation represents a potential link between chronic infection, brain structure, and depressive symptoms.

2. Methods

2.1. Study design and participants

This study utilized data from two independent cohorts. Sample 1 (N = 129) consists of individuals from the Laureate Institute for Brain Research (LIBR) neuroimaging cohort. Sample 2 (N = 5,759) consists of individuals from the Health and Retirement Study (HRS). The two samples were used to examine complementary components of a common conceptual framework rather than as direct replication cohorts. Sample 1 combined immunophenotyping and neuroimaging to examine associations between CMV-related T cell differentiation and regional gray matter volume. Sample 2 used a larger population-based cohort to examine associations between the same T cell differentiation profile and repeated depressive symptom measurements. Analyses were conducted separately within each sample; data and effect estimates were not pooled or directly compared across cohorts. Demographic characteristics of both samples are presented in Table 1. A participant flowchart is available in Figure 1.

Table 1.

Participant characteristics and T cell profiles by CMV serostatus and depressive status across Sample 1 and Sample 2.

Sample 1	HC		MDD
Characteristic	CMV−	CMV+	CMV−	CMV+	p
n	40	46	18	25
CD4+ Naïve [M(SD)]	56.07 (14.00)	44.56 (12.93)	54.83 (14.03)	44.21 (14.08)	<0.001
CD8+ Naïve [M(SD)]	47.26 (18.56)	31.80 (15.63)	48.79 (19.48)	34.47 (16.75)	<0.001
CD4+ TEMRA [M(SD)]	3.09 (4.11)	33.37 (23.20)	4.20 (6.30)	34.50 (21.44)	<0.001
CD8+ TEMRA [M(SD)]	17.73 (12.58)	46.40 (15.59)	20.06 (14.07)	47.53 (18.69)	<0.001
CRP [M(SD)]	1.84 (0.93)	2.05 (0.91)	2.02 (1.14)	2.15 (1.21)	0.514
Age, years [M(SD)]	34.04 (12.61)	35.16 (8.77)	36.82 (10.96)	33.76 (9.56)	0.638
Sex, female [n(%)]	28 (70%)	33 (72%)	9 (50%)	24 (96%)	0.008
BMI [M(SD)]	25.51 (5.46)	29.12 (5.92)	27.35 (7.13)	28.30 (6.92)	0.027
MADRS [M(SD)]	1.23 (1.87)	0.93 (1.72)	24.11 (9.54)	24.67 (8.66)	<0.001
Race, white [n(%)]	36 (90%)	40 (87%)	16 (89%)	19 (76%)	0.428
Tobacco use, yes [n(%)]	4 (10%)	3 (6.5%)	3 (17%)	5 (20%)	0.331
Sample 2	Low depressive symptoms		High depressive symptoms
Characteristic	CMV−	CMV+	CMV−	CMV+	p
n	1,487	3,035	304	933
CD4+ Naïve [M(SD)]	0.49 (0.17)	0.41 (0.17)	0.47 (0.18)	0.39 (0.17)	<0.001
CD8+ Naïve [M(SD)]	0.31 (0.16)	0.22 (0.15)	0.33 (0.16)	0.23 (0.15)	<0.001
CD4+ TEMRA [M(SD)]	0.01 (0.02)	0.04 (0.05)	0.01 (0.02)	0.04 (0.05)	<0.001
CD8+ TEMRA [M(SD)]	0.29 (0.17)	0.49 (0.19)	0.27 (0.16)	0.49 (0.19)	<0.001
CRP [M(SD)]	4.20 (9.68)	4.93 (11.12)	5.41 (10.13)	6.16 (15.16)	<0.001
Age, years [M(SD)]	63.73 (8.11)	64.67 (9.61)	63.33 (9.17)	64.13 (9.90)	0.006
Sex, female [n(%)]	671 (45%)	1,826 (60%)	163 (54%)	659 (71%)	<0.001
BMI [M(SD)]	29.04 (5.84)	29.42 (6.24)	29.70 (6.76)	30.32 (7.40)	0.002
CES-D score [M(SD)]	0.48 (0.69)	0.60 (0.74)	4.79 (1.60)	4.94 (1.61)	<0.001
Race [n(%)]					<0.001
White	1,282 (86%)	1,522 (50%)	238 (78%)	330 (35%)
Black	112 (7.5%)	712 (23%)	39 (13%)	287 (31%)
Hispanic	64 (4.3%)	670 (22%)	21 (6.9%)	253 (27%)
Other	27 (1.8%)	128 (4.2%)	6 (2.0%)	62 (6.7%)
Tobacco use, yes [n(%)]	787 (53%)	1,663 (55%)	202 (66%)	552 (59%)	<0.001
Alcohol use, yes [n(%)]	1,079 (73%)	1,722 (57%)	185 (61%)	441 (47%)	<0.001

Note. In Sample 2, high depressive symptoms were defined as an eight-item Center for Epidemiologic Studies Depression Scale (CES-D) score ≥4 at the 2016 assessment; this classification does not represent a clinical diagnosis of MDD. P-values calculated with Kruskal-Wallis rank sum test for continuous variables, and with Pearson’s chi-squared test for categorical variables. Abbreviations: HC = healthy control; MDD = major depressive disorder; CMV⁻ = cytomegalovirus serostatus negative; CMV⁺ = cytomegalovirus serostatus positive.

Figure 1. Participant flowchart for Samples 1 and 2.

Note. This flowchart summarizes participant inclusion for both study samples. Abbreviations: CES-D = center for epidemiologic studies depression scale; CMV = cytomegalovirus.

2.1.1. Sample 1: LIBR neuroimaging cohort

Data from 462 participants (aged 18-55 years) were collected between October 2010 and November 2016 at the Laureate Institute for Brain Research. The sample was made up of individuals diagnosed with MDD, as determined by DSM-IV-TR criteria in structured and unstructured psychiatric interviews, and healthy controls (HC). General exclusion criteria included medical conditions or medications likely to influence CNS or immunological function, a history of drug or alcohol abuse within 6 months or a history of drug or alcohol dependence within 1 year (DSM-IV-TR criteria), and serious MRI contraindications (including ferromagnetic implants or severe claustrophobia) and failure of T1-weighted image or FreeSurfer segmentation quality control. More detailed information on exclusion criteria is available elsewhere (Ford et al., 2019; Savitz et al., 2015).

The analytical sample in the current study consisted of 129 participants with CMV serostatus data, flow cytometry data, MRI-T1-weighted imaging data, and data for a prespecified set of potential confounders (age, sex, BMI, ethnicity, C-reactive protein levels, and tobacco use). Depressive symptoms in Sample 1 were measured with the Montgomery-Asberg Depression Rating Scale (MADRS). The clinical protocol was approved by the Western Institutional Review Board under protocol number 20101614 and written informed consent was obtained from all participants.

2.1.2. Sample 2: Health and Retirement Study (HRS) cohort

Data from the Health and Retirement Study (HRS), a nationally representative longitudinal cohort of U.S. adults over age 50 (Sonnega et al., 2014), were utilized for Sample 2. Core HRS interviews in 2016, 2018, and 2020 (Sonnega et al., 2014) collected data on depressive symptoms using the Center for Epidemiologic Studies Depression scale (CES-D) (Radloff, 1977; Steffick, 2000), from which an eight-item CES-D score was calculated by summing six negative-symptom items and two reverse-coded positive-affect items, resulting in scores ranging from 0 to 8, with higher scores indicating greater depressive symptom severity. For the descriptive comparison in Table 1, participants with a 2016 CES-D score ≥4 were classified as having elevated depressive symptoms. This threshold was used for descriptive purposes and was not interpreted as a clinical diagnosis of major depressive disorder. CES-D scores were treated as continuous outcomes in longitudinal analyses. Within approximately four weeks of the 2016 core HRS interviews, 9,934 participants completed the 2016 HRS Venous Blood Study (VBS) (Eileen et al., 2017). The 2016 HRS VBS was a special home-visit for venous blood collection in which CMV serology and immunophenotyping took place. Thus, immunophenotyping was approximately concurrent with the 2016 CES-D assessment, whereas the 2018 and 2020 assessments were prospective follow-ups. A timeline of HRS data collection used in the current analyses can be found in Supplementary Figure S1.

The analytical sample in the current study comprised 5,759 participants (aged ≥51 years) with CMV serostatus data, CES-D data, flow cytometry data, and covariate data (age, sex, BMI, ethnicity, C-reactive protein levels, tobacco use, alcohol use). The HRS protocol, including the 2016 VBS, was approved by the University of Michigan Health Sciences and Behavioral Sciences Institutional Review Board (protocol HUM00061128). All participants provided informed consent in accordance with HRS procedures.

2.2. CMV serostatus and immunophenotyping

For Sample 1, the present analysis used a neuroimaging subsample of a previously characterized cohort (Ford et al., 2020). Blood samples were collected in the morning via venipuncture into Becton-Dickinson (BD) Vacutainer tubes containing EDTA for plasma and BD CPT Vacutainer tubes for isolating peripheral blood mononuclear cells (PBMCs). Samples were processed according to standard laboratory protocols. Isolated PBMCs were suspended in fetal bovine serum with 10% dimethyl sulfoxide (DMSO) and stored at −190 °C in liquid nitrogen, while plasma samples were aliquoted and stored at −80 °C until analysis.

In the parent study, plasma samples were thawed and tested, blind to diagnosis, for cytomegalovirus (CMV) IgG antibodies using a commercially available solid-phase ELISA kit (IBL America). Seropositivity was defined a priori by optical density values greater than 0.5, corresponding approximately to 10 international antibody units. Optical density values were transformed into z-scores. The intra- and inter-assay coefficients of variation for the ELISA were 5% and 10%, respectively. The present secondary analysis used the CMV serostatus classification established in the parent study; the cutoff was not re-estimated or modified for the current neuroimaging analysis. Raw, unstandardized plate-level optical-density readings were not included in the present analytic dataset. Additional assay details are reported in the parent publication (Ford et al., 2020).

A ten-color flow cytometry panel was utilized specifically to identify effector memory T cell subsets with re-expression of CD45RA (TEMRA) within the CD4⁺ and CD8⁺ populations. PBMCs were thawed and stained with fluorochrome-conjugated antibodies (Biolegend) using standard staining procedures. Briefly, PBMCs underwent sequential incubation steps on ice: blocking with IgG Fc blocker, antibody labeling, and viability assessment with Zombie Violet dye (Biolegend). Flow cytometry was performed on a BD LSRII cytometer, and data analysis was conducted using FlowJo software version 10.5.0. A time gate was applied during data collection to ensure measurement stability. Doublets, non-viable cells, and cells expressing markers for monocytes (CD14, CD16), B-cells (CD19), or natural killer cells (CD56, CD16) were excluded from analysis. CD3⁺ T cells were selected, followed by gating to identify CD4⁺ and CD8⁺ populations. Within CD4⁺ and CD8⁺ T cells, naïve T cells were defined as CCR7⁺CD45RA⁺ (gating strategy: CD4⁺ or CD8⁺, CD3⁺ CD19⁻, CD45RA⁺ CCR7⁺). Effector memory re-expressing CD45RA (TEMRA; CD45RA⁺ CCR7⁻) subsets were specifically analyzed within CD4⁺ and CD8⁺ T cells. Late differentiation within these subsets was defined by the absence of CD27 and CD28 expression (gating strategy: CD4⁺ or CD8⁺, CD3⁺CD19⁻, CD45RA⁺CCR7⁻ CD27⁻CD28⁻).

For Sample 2, CMV levels were determined on the basis of CMV IgG antibodies, measured as part of the 2016 VBS (Eileen et al., 2017). IgG antibody levels to CMV were measured on a Roche Cobas e411 analyzer (Roche Diagnostics Corporation, Indianapolis, Indiana). The cutoff index (COI) was used to identify non-reactive (<0.5 COI; coded 0), borderline (0.5 to <1.0 COI; coded 0), and reactive (>1.0 COI; coded 1) individuals. The borderline samples (<1% of total VBS sample) were excluded from the current study. We analyzed CMV as a binary variable comparing seropositive individuals with seronegative individuals. The assay was qualitative, with a lower detection limit of 0.15 COI. HRS reported inter-assay coefficients of variation of 3.4% at a mean of 1.23 COI and 2.9% at a mean of 141.4 COI. An intra-assay coefficient of variation was not reported in the official HRS VBS documentation (Eileen et al., 2017).

For Sample 2, absolute counts and percentages of 24 different types of immune cells were assessed by multiparameter flow cytometry using the standardized protocol by the Human Immunology Project (Maecker et al., 2012) with minor modifications performed on an LSRII or a Fortessa X20 flow cytometer (BD Biosciences, San Diego, CA). More detailed flow cytometry methods for this sample are published in the official HRS VBS documentation (Eileen et al., 2017). In the present study, we focused on naïve T cells and effector memory T cells re-expressing CD45RA (TEMRA) within the CD4⁺ and CD8⁺ T-cell compartments. Consistent with the marker definitions shown in Supplementary Table S1, naïve CD4⁺ and CD8⁺ T cells were phenotypically defined by co-expression of CD45RA, CCR7, and CD28 on CD3⁺ T cells (gating strategy: CD4⁺ or CD8⁺, CD3⁺CD19⁻, CD45RA⁺CCR7⁺CD28⁺). In contrast, TEMRA subsets were defined by re-expression of CD45RA with loss of CCR7 and CD28 (gating strategy: CD4⁺ or CD8⁺, CD3⁺CD19⁻, CD45RA⁺CCR7⁻CD28⁻). Both samples used CD45RA and CCR7 as the core markers distinguishing naïve (CD45RA⁺CCR7⁺) from TEMRA (CD45RA⁺CCR7⁻) cells. The Sample 2 panel additionally included CD28 to further refine these populations. Thus, although the marker panels were not identical, they captured the same underlying biological distinction between naïve and terminally differentiated T cells. All analyses were conducted separately within each cohort using internally consistent definitions.

A more detailed description of flow-cytometry methodology for both cohorts can be found in the Supplementary Methods alongside a comparison table of T cell definitions across samples (Supplementary Table S1) and a flow-cytometry gating hierarchy flowchart (Supplementary Figure S2).

2.3. Image acquisition and processing

Neuroimaging data were only available for Sample 1; T1-weighted anatomical images were acquired on two identical 3T scanners (GE Discovery MR750) using an MPRAGE sequence with the following parameters: FOV=240mm, 190 slices, slice thickness=0.9mm, image matrix=256×256, TR/TE=5/2.012 ms, acceleration factor R=2 in the phase encoding direction, flip angle=8 degrees.

Cortical reconstruction and volumetric segmentation were performed using FreeSurfer version 6.0.0 (Dale et al., 1999). Whole-brain gray matter volumes (GMVs) were estimated from individual anatomical images including 68 cortical regions (34 regions per hemisphere) using the Desikan-Killiany atlas. Visual inspection of all cortical segmentation was performed before analysis for quality assurance. FreeSurfer has been validated against histological measurements (Cardinale et al., 2014) and demonstrates good test-retest reliability (Iscan et al., 2015).

2.4. Statistical analysis

Analyses were conducted separately by sample. To align interpretation with the biological hypothesis and reduce multiplicity, the CD4⁺ TEMRA:naïve ratio was treated as the focal immune exposure because it jointly captures naïve-cell contraction and TEMRA expansion. This hierarchy was specified during revision rather than preregistered; CD4⁺ naïve and CD4⁺ TEMRA percentages were component analyses, and all CD8⁺ measures were exploratory. Covariates were selected a priori because they are established determinants of immune differentiation, brain morphology, or depressive symptoms. Age, sex, and BMI were included in both samples; total intracranial volume (TIV) was included in all MRI models; diagnostic status was added to fully adjusted Sample 1 models; and CRP, tobacco use, alcohol use (only available for Sample 2), and race/ethnicity were included to reduce confounding by inflammatory, behavioral, and sociodemographic factors (Ford et al., 2020; Klopack et al., 2022). Primary model assumptions were evaluated at the residual level using quantile-quantile plots, residual-versus-fitted plots, Shapiro-Wilk and Breusch-Pagan tests, and Cook’s distance (Supplementary Table S2 and Supplementary Figures S3–S5). All analyses were conducted in R (version 4.5.1). Statistical significance was defined as two-sided p < 0.05, and Benjamini-Hochberg false discovery rate (FDR) correction was applied to adjust for multiple comparisons.

2.4.1. Sample 1: LIBR neuroimaging analyses

For Sample 1, T-cell percentages and ratios were transformed as log(x + 1) to reduce right-skew and the influence of extreme values; normality of the predictors was not treated as a requirement of linear regression. Associations between each immune measure and each of 34 bilateral cortical regional volumes were estimated using linear regression (base stats package in R). Model 1 adjusted for age, sex, body mass index (BMI), and total intracranial volume (TIV), whereas Model 2 additionally controlled for MDD diagnostic status, C-reactive protein (CRP), tobacco use, and race/ethnicity. The neuroimaging analysis was an exploratory whole-cortex ROI analysis. Associations were tested across the 34 bilateral cortical regions (68 hemisphere specific regions total) defined by the Desikan–Killiany atlas. For each immune predictor or CMV serostatus and each covariate model separately, FDR correction was applied across the 34 regional tests. Hemisphere-specific analyses were also conducted as secondary analysis to assess potential laterality effects and are reported in Supplementary Table S3.

2.4.2. Sample 2: HRS cohort analyses

For Sample 2, associations of CMV serostatus and T cell measures with repeated CES-D scores were estimated using linear mixed-effects in lme4 (v1.1–37). A participant-specific random intercept accounted for within-person correlation across the 2016, 2018, and 2020 assessments, and survey wave was entered as a categorical fixed effect. Model 1 adjusted for age, sex, BMI, and wave. Model 2 additionally adjusted for CRP, tobacco use, alcohol use, and race/ethnicity. The model was specified as:

$$CES-D\sim \beta 0+\beta 1\left(Exposure\right)+\beta 2\left(Wave\right)+\beta \left(Co\mathrm{var}iates\right)+\left(1/ID\right)$$

Exposure variables were CMV serostatus and the log-transformed CD4⁺ naïve, CD4⁺ TEMRA, CD4⁺ TEMRA:naïve ratio, CD8⁺ naïve, CD8⁺ TEMRA, and CD8⁺ TEMRA:naïve ratio measures. Within each Sample 2 covariate model, FDR correction was applied across these seven exposure tests. Because the 2016 VBS and the baseline CES-D assessment occurred in the same 2016 assessment window, mixed-effects estimates are described as associations across waves rather than prospective effects.

2.4.3. Mediation analyses

Mediation analysis was conducted using structural equation modeling (SEM) in lavaan (v0.6–19) to test whether T cell composition might plausibly mediate associations between cytomegalovirus (CMV) seropositivity and study outcomes. The Sample 1 mediation analysis was designed specifically to examine whether T-cell differentiation statistically accounted for the association between CMV serostatus and gray matter volume, adjusting for age, sex, body mass index (BMI), and total intracranial volume (TIV). MDD diagnosis was not evaluated as a mediation outcome because it was a pre-existing recruitment classification rather than an outcome measured after the immune and neuroimaging assessments. Moreover, associations between MDD diagnosis and these T-cell populations had previously been examined in the parent cohort, with no significant diagnosis effects identified (Ford et al., 2020). Associations between T-cell measures and psychiatric symptoms were instead evaluated longitudinally in the larger Sample 2. In Sample 2, longitudinal SEM models tested whether the effects of CMV serostatus on depressive symptom severity (CES-D scores) might be mediated by T cell composition, accounting for repeated CES-D assessments across waves. All models were estimated using robust maximum likelihood with Huber–White standard errors (MLR) and full information maximum likelihood (FIML) for missing data; in Sample 2, clustering by participant was applied to correct for non-independence (within subjects repeated measures). Indirect (a×b), direct (c′), and total (a×b + c′) effects were estimated, and statistical inference was based on robust (z-based) confidence intervals.

2.4.4. Power analyses

Because the sample sizes were fixed before the present analyses, sensitivity analyses were conducted to estimate the minimum detectable standardized effect at 80% power and two-sided α = 0.05. These direct-effect calculations were not simulation-based. For Sample 1 (N = 129), an analytical calculation based on the noncentral t distribution indicated a minimum detectable effect of approximately |β| = 0.25. For Sample 2, the nominal number of repeated observations was adjusted for within-participant correlation using the design-effect formula N_effective = N_observations /[1 + (m – 1) × ICC]. Assuming 5,759 participants, three waves, and ICC values of 0.30–0.70, the effective sample size ranged from approximately 7,199 to 10,798, corresponding to minimum detectable effects of approximately |β| = 0.027–0.033. This design-effect approach provides an approximate sensitivity benchmark rather than an exact power calculation for the fitted mixed-effects model because it does not reproduce the complete variance, covariate, and observation structure of that model. The calculations were based on nominal α = 0.05 and did not account for the subsequent FDR correction (Supplementary Table S4).

3. Results

3.1. Study population characteristics

Participant characteristics for both cohorts are summarized in Table 1. The present study represents a neuroimaging subsample of the previously published parent cohort (Ford et al., 2020). From the original sample, 129 adults with available structural MRI data were included across diagnostic (MDD vs. HC) and CMV serostatus groups. Consistent with findings reported in the parent study (Ford et al., 2020), CMV-positive participants exhibited a pronounced shift in T cell composition, characterized by lower CD4⁺ and CD8⁺ naïve cell percentages and markedly higher TEMRA proportions compared with CMV-negative individuals (all p < 0.001). This pattern was consistent across both diagnostic groups, indicating robust CMV-related immune differentiation. CMV-positive participants also showed slightly higher BMI (p = 0.027) and a greater proportion of females among MDD cases (p = 0.008), while age, CRP, ethnicity, and tobacco use did not differ significantly across groups.

In Sample 2, which included 5,759 participants (3,968 CMV-positive and 1,791 CMV-negative), similar immunological trends were observed at the population level. CMV seropositivity was associated with reduced proportions of naïve T cells and increased TEMRA subsets across both CD4⁺ and CD8⁺ lineages (all p < 0.001). CMV-positive participants were slightly older, more often female, and had higher BMI and CRP levels compared to CMV-negative individuals (all p < 0.01). Racial and ethnic distributions also differed significantly by CMV status (p < 0.001). Overall, both samples demonstrated consistent CMV-related immunological shifts toward a more differentiated T cell phenotype, providing a biologically coherent foundation for subsequent analyses linking T cell aging to brain and behavioral outcomes.

3.2. The association between T cell subtypes and gray matter volume

In the LIBR neuroimaging cohort (Sample 1), associations were tested across 34 bilateral cortical regions. The parahippocampal region was the only region in which an association with T cell composition survived FDR correction in both covariate models. As shown in Figure 2 and Table 2, a higher CD4⁺ TEMRA:naïve ratio was associated with lower parahippocampal GMV in Model 1 (β = −0.248, 95% CI [−0.406, −0.089], p_FDR = 0.030); the association was of comparable magnitude in Model 2 but did not survive FDR correction (β = −0.258, 95% CI [−0.421, −0.095], p_FDR = 0.054). Secondary component analyses showed that higher CD4⁺ TEMRA cell percentages were associated with lower parahippocampal GMV in both models (Model 1: β = −0.265, 95% CI [−0.423, −0.106], p_FDR = 0.022; Model 2: β = −0.276, 95% CI [−0.439, −0.113], p_FDR = 0.040). CD4⁺ naïve cell percentages were not significantly associated with parahippocampal GMV in either model (Model 1: β = 0.041, 95% CI [−0.127, 0.209], p_FDR = 0.924; Model 2: β = 0.038, 95% CI [−0.135, 0.212], p_FDR = 0.893). Effects were in the same direction in the left and right parahippocampal gyrus, although the left-hemisphere associations were somewhat stronger (see Supplementary Table S3).

Figure 2. Associations between CMV serostatus, CD4⁺ and CD8⁺ subsets and neuroanatomical and behavioral outcomes across two independent cohorts.

Note. Forest plot of associations between CMV serostatus, T cell subtypes, and two independent outcomes from different cohorts: parahippocampal gray matter volume (GMV) in Sample 1, and depressive symptom severity (CES-D score) in Sample 2. Models varied by outcome/sample: Model 1 adjusted for age, sex, and BMI in both samples; total intracranial volume (TIV) was included for Sample 1 and wave was included for Sample 2. Model 2 additionally adjusted for C-reactive protein (CRP), tobacco use, and race/ethnicity in both samples; MDD diagnostic status was included for Sample 1, and alcohol use was included for Sample 2. Blue bars represent statistically significant results after false discovery rate (FDR) correction (p_FDR < 0.05). Higher CD4⁺ TEMRA:naïve ratios and higher CD4⁺ TEMRA percentages showed moderate negative associations with parahippocampal GMV. Both were significant in Model 1, but only the CD4⁺ TEMRA association remained significant after FDR correction in Model 2. All CD4⁺ and CD8⁺ subsets showed significant FDR-corrected associations with CES-D scores in Model 1. CD4⁺ and CD8⁺ TEMRA:naïve ratios and TEMRA cell percentages demonstrated small positive associations with CES-D, whereas CD4⁺ and CD8⁺ naïve percentages exhibited small negative associations. In Model 2, all associations remained significant except for CD8⁺ TEMRA. Abbreviations: β = standardized regression coefficient; 95% CI = 95% confidence interval; p_FDR= false discovery rate–adjusted p-value; GMV = gray matter volume; CMV = cytomegalovirus; TEMRA = terminally differentiated effector memory T cells re-expressing CD45RA; CES-D = center for epidemiologic studies depression scale.

Table 2.

Associations of CMV serostatus and T cell subsets with parahippocampal gray matter volume (Sample 1) and depressive symptoms (Sample 2).

Outcomea	Predictor	Model	β	95% CI	p	p FDR
Parahippocampal GMV (Sample 1)	CMV serostatus	Model 1	−0.411	[−0.738,−0.084]	0.015	0.171
		Model 2	−0.450	[−0.786, −0.114]	0.010	0.337
	CD4+ TEMRA:naïve	Model 1	−0.248	[−0.406, −0.089]	0.003	0.030*
		Model 2	−0.258	[−0.421, −0.095]	0.002	0.054
	CD4+ TEMRA	Model 1	−0.265	[−0.423, −0.106]	0.001	0.022*
		Model 2	−0.276	[−0.439, −0.113]	0.001	0.040*
	CD4+ naïve	Model 1	0.041	[−0.127, 0.209]	0.631	0.924
		Model 2	0.038	[−0.135, 0.212]	0.665	0.893
	CD8+ TEMRA:naïve	Model 1	−0.161	[−0.328, 0.007]	0.063	0.863
		Model 2	−0.168	[−0.340, 0.004]	0.058	0.841
	CD8+ TEMRA	Model 1	−0.120	[−0.281, 0.042]	0.149	0.785
		Model 2	−0.142	[−0.309, 0.025]	0.099	0.841
	CD8+ naïve	Model 1	0.164	[−0.015, 0.343]	0.075	0.846
		Model 2	0.158	[−0.031, 0.346]	0.103	0.878
CES-D (Sample 2)	CMV serostatus	Model 1	0.187	[0.138, 0.236]	<0.001	<0.001***
		Model 2	0.079	[0.027, 0.131]	0.003	0.003**
	CD4+ TEMRA:naïve	Model 1	0.097	[0.074, 0.119]	<0.001	<0.001***
		Model 2	0.062	[0.039, 0.085]	<0.001	<0.001***
	CD4+ TEMRA	Model 1	0.046	[0.023, 0.068]	<0.001	<0.001***
		Model 2	0.024	[0.002, 0.047]	0.032	0.049*
	CD4+ naïve	Model 1	−0.095	[−0.118, −0.072]	<0.001	<0.001***
		Model 2	−0.062	[−0.085, −0.039]	<0.001	<0.001***
	CD8+ TEMRA:naïve	Model 1	0.039	[0.016, 0.063]	0.001	0.001**
		Model 2	0.026	[0.002, 0.049]	0.033	0.033*
	CD8+ TEMRA	Model 1	0.035	[0.012, 0.058]	0.003	0.003**
		Model 2	0.017	[−0.006, 0.040]	0.148	0.148
	CD8+ naïve	Model 1	−0.037	[−0.061, −0.013]	0.003	0.003**
		Model 2	−0.032	[−0.056, −0.008]	0.010	0.010**

^aParahippocampal GMV was modeled with linear regression, and CES-D was modeled with linear mixed-effects.

Note. Results for the associations between CMV serostatus, T cell subsets, and (1) parahippocampal gray matter volume (GMV) in Sample 1 and (2) depressive symptoms (CES-D scores) in Sample 2. Parahippocampal GMV in Sample 1 was modeled using linear regression; repeated CES-D scores in Sample 2 were modeled using linear mixed-effects regression with a participant-level random intercept. Models varied by outcome/sample: Model 1 adjusted for age, sex, and BMI in both samples; total intracranial volume (TIV) was included for Sample 1 and wave was included for Sample 2. Model 2 additionally adjusted for C-reactive protein (CRP), tobacco use, and race/ethnicity in both samples; MDD diagnostic status was included for Sample 1, and alcohol use was included for Sample 2.

Asterisks indicate statistical significance: * p_FDR < 0.05;

^**p_FDR < 0.01;

^***p_FDR < 0.001.

Abbreviations: β = standardized regression coefficient; 95% CI = 95% confidence interval; p_FDR= false discovery rate–adjusted p-value; GMV = gray matter volume; CMV = cytomegalovirus; TEMRA = terminally differentiated effector memory T cells re-expressing CD45RA; CES-D = center for epidemiologic studies depression scale.

Several additional regions survived FDR correction in one covariate model, including the lateral orbitofrontal cortex, supramarginal gyrus, pars triangularis, and frontal pole. However, the parahippocampal association with CD4⁺ TEMRA percentages was the most consistent because it remained significant across both models. The other regional findings should be considered preliminary because they were not consistent across two covariate models (Supplementary Table S5).

CMV serostatus was negatively associated with parahippocampal GMV in both models (Model 1: β = −0.411, 95% CI [−0.738, −0.084], p = 0.015, p_FDR = 0.171; Model 2: β = −0.450, 95% CI [−0.786, −0.114], p = 0.010, p_FDR = 0.337), though these associations did not remain significant after FDR adjustment.

3.3. The association between T cell subtypes and depressive symptoms

In the HRS cohort (Sample 2), longitudinal mixed-effects models demonstrated associations between the CD4⁺ T cell differentiation profile and depressive symptom severity measured by the CES-D scale (Figure 2 and Table 2). For the focal immune measure, a higher CD4⁺ TEMRA:naïve ratio was associated with higher CES-D scores in both Model 1 (β = 0.097, 95% CI [0.074, 0.119], p_FDR < 0.001) and Model 2 (β = 0.062, 95% CI [0.039, 0.085], p_FDR < 0.001). Secondary component analyses showed that higher CD4⁺ TEMRA percentages were associated with higher CES-D scores (Model 1: β = 0.046, 95% CI [0.023, 0.068], p_FDR < 0.001; Model 2: β = 0.024, 95% CI [0.002, 0.047], p_FDR = 0.049), whereas higher CD4⁺ naïve-cell percentages were associated with lower CES-D scores (Model 1: β = −0.095, 95% CI [−0.118, −0.072], p_FDR < 0.001; Model 2: β = −0.062, 95% CI [−0.085, −0.039], p_FDR < 0.001).

CD8⁺ measures were examined as exploratory comparisons. A higher CD8⁺ TEMRA:naïve ratio was associated with higher CES-D scores in Model 1 (β = 0.039, 95% CI [0.016, 0.063], p_FDR = 0.001) and Model 2 (β = 0.026, 95% CI [0.002, 0.049], p_FDR = 0.033). Higher CD8⁺ naïve-cell percentages were associated with lower CES-D scores in both models (Model 1: β = −0.037, 95% CI [−0.061, −0.013], p_FDR = 0.003; Model 2: β = −0.032, 95% CI [−0.056, −0.008], p_FDR = 0.010). Higher CD8⁺ TEMRA percentages were associated with higher CES-D scores in Model 1 (β = 0.035, 95% CI [0.012, 0.058], p_FDR = 0.003), but this association did not remain significant in Model 2 (β = 0.017, 95% CI [−0.006, 0.040], p_FDR = 0.148).

CMV seropositivity was associated with greater depressive symptom burden (Model 1 β = 0.187, 95% CI [0.138, 0.236], p_FDR < 0.001; Model 2 β = 0.079, 95% CI [0.027, 0.131], p_FDR = 0.003). These findings suggest that CMV-related T cell aging corresponds to higher depressive symptom severity across 3 waves over four years, with effects more pronounced for the CD4⁺ compartment than CD8⁺ compartment.

3.4. Mediation pathways

Structural equation modeling (SEM) was used to test whether CMV-associated T cell aging profiles might plausibly mediate the associations between CMV serostatus and brain or depressive symptom outcomes. In Sample 1, no T cell measure showed a significant indirect effect after FDR correction (Figure 3A and Supplementary Table S6). The indirect effect of CD4⁺ TEMRA percentage on parahippocampal GMV was nominally significant (indirect = −0.382, 95 % CI [−0.753, −0.012], p_FDR = 0.077) but did not survive multiple-comparison correction. Indirect effects through the CD4⁺ TEMRA:naïve ratio (indirect = −0.349, 95 % CI [−0.736, 0.038], p_FDR = 0.116) and CD4⁺ naïve cells (indirect = 0.042, 95 % CI [−0.071, 0.155], p_FDR = 0.601) were also nonsignificant. Because no significant CD8⁺ association with regional GMV was consistent across covariate models, CD8⁺ mediation models were not fitted in this sample.

Figure 3. Mediation of the association between CMV seropositivity and parahippocampal GMV (Sample 1) and depressive symptoms (Sample 2) through CD4⁺ T cell subsets.

Note. Structural equation models (SEM) were used to test whether CMV seropositivity was associated with parahippocampal gray matter volume (GMV; panel A) or greater depressive symptom severity (CES-D scores; panel B) through alterations in CD4⁺ T cell composition.

In Sample 2, significant indirect effects were observed for CD4⁺ subsets linking CMV seropositivity to depressive symptom severity (Table 3 and Figure 3B). The indirect effects were 0.050 for the CD4⁺ TEMRA:naïve ratio (95 % CI [0.034, 0.066], p_FDR < 0.001) and 0.039 for CD4⁺ naïve cells (95 % CI [0.027, 0.051], p_FDR < 0.001), indicating partial mediation of the CMV to CES-D association. The corresponding direct effects of CMV on CES-D scores remained significant for both mediators (CD4⁺ TEMRA:naïve ratio direct = 0.139; CD4⁺ naïve direct = 0.149; both p_FDR < 0.001), indicating partial mediation of the CMV-CES-D association. No significant indirect effect was found for CD4⁺ TEMRA cells (indirect = 0.009, p_FDR = 0.317); its direct effect (direct = 0.179, p_FDR < 0.001) accounted for nearly the entire total effect (total = 0.188), indicating that this pathway does not meaningfully mediate the CMV-CES-D association.

Table 3.

Structural equation model estimates of direct, indirect, and total effects linking CMV serostatus, CD4⁺ T cell subsets, and depressive symptoms (CES-D scores) in Sample 2.

Exposure	Outcome	Mediator	Effect	Estimate	95% CI	p	p FDR
CMV serostatus	CES-D	CD4+ TEMRA:naïve ratio	Indirect	0.050	[0.034, 0.066]	<0.001	<0.001***
			Direct	0.139	[0.090, 0.188]	<0.001	<0.001***
			Total	0.189	[0.142, 0.236]	<0.001	<0.001***
		CD4+ naïve	Indirect	0.039	[0.027, 0.051]	<0.001	<0.001***
			Direct	0.149	[0.102, 0.196]	<0.001	<0.001***
			Total	0.189	[0.142, 0.236]	<0.001	<0.001***
		CD4+ TEMRA	Indirect	0.009	[−0.009, 0.027]	0.317	0.317
			Direct	0.179	[0.130, 0.228]	<0.001	<0.001***
			Total	0.188	[0.141, 0.235]	<0.001	<0.001***

Note. Estimates reflect standardized effects from structural equation models (SEM) using robust maximum likelihood (MLR) with clustering by participant. Indirect, direct, and total effects correspond to paths a×b, c′, and a×b + c′, respectively. P-values are FDR-corrected across mediators.

Asterisks indicate statistical significance: *** p_FDR < 0.001.

Abbreviations: 95% CI = 95% confidence interval; p_FDR= false discovery rate–adjusted p-value; CMV = cytomegalovirus; CES-D = center for epidemiologic studies depression scale; TEMRA = terminally differentiated effector memory T cells re-expressing CD45RA.

In contrast, none of the CD8⁺ subsets showed significant indirect effects, suggesting that CMV-related increases in depressive symptoms were more strongly linked to CD4⁺ rather than CD8⁺ T cell differentiation. Full model results for the Sample 1 CD4⁺ GMV mediation and Sample 2 CD8⁺ CES-D mediation analyses are provided in Supplementary Table S6.

4. Discussion

In this dual-sample study, we provide convergent evidence that CMV-associated T cell immunosenescence is linked to smaller parahippocampal gray matter volumes and elevated depressive symptoms. Specifically, in our neuroimaging sample (Sample 1), higher CD4⁺ TEMRA percentages were associated with lower parahippocampal volume in both covariate models, and a higher CD4⁺ TEMRA:naïve ratio was associated with lower parahippocampal volume in the minimally adjusted model. In the population cohort (Sample 2), a higher CD4⁺ TEMRA:naïve ratio, indicating a skew toward more terminally differentiated and fewer naïve T cells, was associated with greater depressive symptom severity across three assessments spanning 2016, 2018, and 2020, and statistically mediated the association between CMV serostatus and depressive symptoms. Together, these findings identify potentially related associations among CMV serostatus, T cell differentiation, brain structure, and depressive symptoms. Although none of the T cell measures significantly mediated the association between CMV serostatus and parahippocampal GMV after FDR correction, the indirect effect through CD4⁺ TEMRA cells was suggestive (p_FDR = 0.077). Given the modest neuroimaging sample, this proposed pathway remains inconclusive and requires confirmation in a larger study.

Our findings align with a growing body of evidence implicating CMV infection and immune aging in depression. Epidemiological studies show that CMV seropositivity is more common in depression and predicts incident depressive symptoms independent of baseline inflammation (Simanek et al., 2014; Zheng and Savitz, 2022), while neuroimaging work found that CMV-positive individuals with depression exhibit regional gray matter reductions, including orbitofrontal and temporal alterations, as well as disrupted white-matter and resting-state functional connectivity (Zheng et al., 2021a; Zheng et al., 2020; Zheng et al., 2021b). These studies examined CMV serostatus, whereas the present analysis examined T cell differentiation measures and was not intended as a direct replication of the same spatial pattern. Although several orbitofrontal and other cortical regions showed model-specific associations, the parahippocampal association was the most consistent across covariate models. These findings do not establish that the parahippocampal region is uniquely vulnerable, and their anatomical specificity requires confirmation in larger studies. Our observation that a more differentiated CD4⁺ T cell profile is associated with reduced parahippocampal volume extends prior work by implicating a potential immunological mechanism in medial temporal lobe structural variation.

Previous studies in schizophrenia and bipolar disorder have similarly shown that higher anti-CMV antibody titers are associated with structural alterations within medial temporal regions, including smaller hippocampal volumes and poorer memory performance (Houenou et al., 2014) and that CMV-positive men with severe mental illness exhibit reductions in dentate gyrus surface (Andreou et al., 2021; Andreou et al., 2022). Building on this, our results indicate that accelerated T cell aging (Ford et al., 2020; Klenerman and Oxenius, 2016) may help explain how CMV-related immune differentiation may contribute to vulnerability across interconnected medial temporal structures. This interpretation is consistent with conclusions from a recent multidisciplinary Royal Society expert meeting, which highlighted CMV-associated immunosenescence as a plausible pathway underlying a wide range of indirect health effects, including neuropsychiatric outcomes, even in the absence of classical CMV end-organ or tissue damage (Yates et al., 2025). Importantly, large-scale population data show that chronic social stressors (e.g., discrimination, trauma, financial strain) are associated with the same immunosenescent T cell profile observed here (i.e., more TEMRA cells and fewer naïve T cells) and that these associations are substantially attenuated or eliminated after accounting for CMV infection (Klopack et al., 2022), suggesting that CMV infection may act as a mediator or amplifier of stress-related dysregulation of the adaptive immune system. Taken together, these convergent findings situate CMV-associated T cell aging as a clinically relevant and potentially modifiable biological pathway linking latent infection and neuroanatomical vulnerability relevant to depression.

Biologically, CMV’s influence on the T cell compartment is well established: lifelong chronic infection repeatedly stimulates antiviral T cells, driving clonal expansion of TEMRA cells and contraction of the naïve T cell pool across both CD4⁺ and CD8⁺ lineages (Ford et al., 2020; Klenerman and Oxenius, 2016). These senescent-like T cells often acquire a pro-inflammatory secretory profile (e.g., IFN-γ, TNF-α) and lose key costimulatory molecules (CD28⁻CD27⁻), creating an immune milieu characterized by chronic, low-grade inflammation and reduced adaptability (Klenerman and Oxenius, 2016; Nicoli et al., 2022). Medial temporal lobe regions such as the hippocampus and parahippocampus are particularly sensitive to inflammatory cytokines (Ironside et al., 2020; Koo and Duman, 2008; Monje et al., 2003), which can impair neurogenesis (Koo and Duman, 2008; Monje et al., 2003). We therefore speculate that an expanded reservoir of CMV-associated effector memory T cells may serve as a persistent source of peripheral inflammatory signals that gradually erode the structural integrity of these regions. This interpretation is supported by our previous postmortem evidence showing that CMV seropositivity is associated with increased CNS expression of inflammation-related genes and more activated microglial morphology and linked to mood disorders and suicide (Zheng et al., 2023). Preclinical work further demonstrates that T cells can directly influence neural integrity: cytotoxic CD8⁺ T cells infiltrate the aged mouse brain and induce white-matter injury via granzyme-mediated axonal damage (Groh et al., 2025; Groh et al., 2021; Kaya et al., 2022), and chronic stress can reprogram naïve CD4⁺ T cells to secrete neuroactive metabolites, including xanthine, that cross the blood–brain barrier and act on oligodendrocytes in limbic regions (Fan et al., 2019); circulating xanthine is elevated in patients with major depressive disorder (Ali-Sisto et al., 2016). Together, these findings support a model in which CMV-expanded TEMRA cells, through cytokine release or other soluble factors, compromise neural homeostasis over time. Given its rich vascularization and central role in limbic stress circuitry, the parahippocampus may be particularly susceptible to such immune-mediated insults.

Notably, only CD4⁺ and not CD8⁺ T cell composition showed significant indirect effect between the CMV–depression relationship, with the mediation driven principally by naïve CD4⁺ T cell contraction rather than TEMRA expansion. This pattern is consistent with the unique and non-redundant role of CD4⁺ T cells in maintaining medial temporal lobe integrity. Foundational preclinical studies have shown that repopulation of immunodeficient mice with CD4⁺, but not CD8⁺, T cells restores hippocampal neurogenesis and normalizes brain-derived neurotrophic factor (BDNF) expression (Wolf et al., 2009), while adoptive transfer of CD4⁺ but not CD8⁺ T cells rescues depressive- and anxiety-like behavior in T cell-deficient animals (Rattazzi et al., 2013). A mechanistically coherent explanation for these CD4⁺-specific effects centers on the capacity of naïve CD4⁺ T cells to generate peripheral regulatory T cells (Tregs), which suppress microglial activation and sustain hippocampal integrity through IL-10 and TGF-β signaling — a pool that contracts as naïve CD4⁺ cells are displaced by CMV-associated TEMRA expansion. Consistent with this, experimental Treg depletion reduces hippocampal serotonin levels and aggravates depressive-like behavior in rodents (Kim et al., 2012), while clinical studies demonstrate that Treg deficits correlate with MDD severity (Rachayon et al., 2024). Together, these observations suggest that CMV-associated erosion of the naïve CD4⁺ pool may impair parahippocampal structural integrity and mood regulation primarily through loss of Treg-mediated neuroprotection, a mechanism not shared by the CD8⁺ lineage. Future studies directly measuring Treg frequency, hippocampal BDNF, and antigen-specific CD4⁺ responses within the same individuals will be needed to confirm this proposed pathway.

Despite its strengths, several limitations warrant consideration. First, because neuroimaging and longitudinal depressive symptoms were examined in different cohorts, no single sample tested the complete proposed pathway linking CMV serostatus, T cell differentiation, brain structure, and depressive symptoms. The dual-sample findings should therefore be interpreted as complementary associations that motivate, but do not establish, an integrated mechanistic pathway. Second, the mediation models support indirect pathways linking CMV seropositivity, T cell differentiation, and depressive symptoms, but unmeasured or bidirectional influences cannot be ruled out. Residual confounding from health behaviors or socioenvironmental exposures remains possible—for example, chronic psychological stress can promote both TEMRA expansion and depression. Other latent herpesviruses (e.g., Epstein–Barr virus, varicella zoster), which also shape the T cell repertoire and have been linked to neuropsychiatric outcomes, were not assessed. Third, although our T cell measures capture bulk differentiation states rather than CMV-specific memory T cells, prior work shows that CMV drives exceptionally large antigen-specific clonal expansions of both CD4⁺ and CD8⁺ memory pools (Sylwester et al., 2005) and is a major non-genetic determinant of immune variation (Brodin et al., 2015). Thus, the TEMRA:naïve patterns observed here are still very likely influenced by CMV. However, because we did not measure CMV-specific T cell responses directly, future studies using antigen-specific MHC tetramer- or peptide-based assays will be essential to confirm specificity and clarify the downstream pathways linking CMV-associated immune remodeling to neural and affective outcomes. Finally, the LIBR neuroimaging cohort is modest in size, which may limit power to detect meaningful associations. In addition, the neuroimaging analysis was an exploratory screen across 34 bilateral cortical regions rather than a test of predefined regions of interest. Although FDR correction was applied, the modest sample size limited power for extensive regional screening and increased uncertainty in the estimated associations. The parahippocampal findings were more consistent across covariate models than findings in other regions. Therefore, the parahippocampal findings require independent replication, and findings outside this region should be considered preliminary. In contrast, the HRS cohort provides substantial statistical power but relies on self-reported depressive symptoms rather than structured diagnostic assessments, and its focus on older adults may limit generalizability to younger populations, in whom CMV imprinting and T cell aging may follow different trajectories. Consequently, these results require replication, generalization, and mechanistic validation before firm conclusions can be drawn about causality or clinical application.

5. Conclusion

Our findings suggest that CMV-related T cell aging may constitute a potential biological mechanism linking chronic viral infection and susceptibility to neurobehavioral change. Consistent associations between CD4⁺ T cell differentiation, parahippocampal structure, and depressive symptoms highlight the peripheral immune aging process as a potential upstream driver of mood-relevant brain alterations. Notably, recent large-scale population studies reported that recombinant zoster vaccination (by reducing reactivation of another persistent herpesvirus) lowers dementia risk (Eyting et al., 2025; Taquet et al., 2024), supporting the broader principle that latent herpesvirus-associated immune activation is a modifiable risk factor for brain health and neuropsychiatric burden. These converging lines of evidence point to CMV infection and immune senescence as actionable targets, and suggest that antiviral, vaccine-based, or immune-rejuvenation approaches may hold therapeutic potential for mitigating risk across a range of neuropsychiatric conditions.

Models were estimated using SEM (lavaan, robust maximum likelihood). Panel A (Sample 1) was adjusted for age, sex, BMI, and total intracranial volume; panel B (Sample 2) was adjusted for age, sex, BMI, and survey wave, with clustering by participant. FDR correction was applied across mediators within each panel. Solid arrows indicate significant paths (p_FDR < 0.05), while dashed arrows indicate paths that were not significant. Panel A shows no indirect effects surviving FDR correction for the CD4⁺ TEMRA:naïve ratio or either component subset; the indirect effect through CD4⁺ TEMRA percentage was nominally significant (p = 0.043) but did not survive correction (p_FDR = 0.077). Panel B shows significant indirect effects for both the CD4⁺ TEMRA:naïve ratio and CD4⁺ naïve cells, indicating that reduced naïve and skewed TEMRA:naïve balance partially mediated the CMV to depression association. The indirect path via CD4⁺ TEMRA cells was not significant. Abbreviations: p_FDR= false discovery rate–adjusted p-value; CMV = cytomegalovirus; TEMRA = terminally differentiated effector memory T cells re-expressing CD45RA; GMV = gray matter volume; CES-D = center for epidemiologic studies depression scale.

Highlights.

The Role of CD4+ T-cell Differentiation in Cytomegalovirus-associated Parahippocampal Volume Reduction and Depressive Symptoms: A Dual-Sample Study

• Higher CD4⁺ TEMRA levels were associated with smaller parahippocampal volume.

• CD4⁺ TEMRA:naïve ratio was associated with depressive symptoms in 5,759 adults.

• CD4⁺ subsets statistically mediated the CMV–depression association.

• Dual-sample design links immunophenotyping, neuroimaging, and longitudinal symptoms

• CMV-associated T-cell differentiation maybe a candidate pathway to depression risk

Data availability

Data from the Health and Retirement Study (HRS) are publicly available to qualified researchers through the HRS data repository (https://hrsdata.isr.umich.edu/), with access granted upon application via https://hrs.isr.umich.edu/about. Data from the Laureate Institute for Brain Research (LIBR) cohort are not publicly available due to institutional considerations but can be made available from the corresponding author upon reasonable request. Analysis code used in this study is available from the corresponding author upon reasonable request.