Cytomegalovirus Coinfection Is Associated with Increased Vascular-Homing CD57⁺ CD4 T Cells in HIV Infection

Although combination antiretroviral therapy (ART) has dramatically prolonged the life expectancy of people living with HIV infection (PLWH), PLWH are still at higher risk for a number of comorbidities, including cardiovascular disease (CVD) (1, 2). In PLWH, the ubiquitous pathogen CMV is a risk factor for CVD and other comorbidities even when HIV viremia is suppressed by ART (2, 3); CMV is also associated with increased CVD risk in the general population (4). It is becoming increasingly evident that immunologic and inflammatory components drive CVD pathogenesis and severity (5, 6). Therefore, understanding how CMV infection alters immune responses may help identify mechanisms by which CMV contributes to HIV-associated comorbidities, including CVD.

CMV is an immunodominant pathogen associated with elevated inflammation and T cell activation, particularly in PLWH (3, 7–12). A large proportion of circulating CD4 and CD8 T cells are CMV specific (13–15). Many CMV-specific CD4 T cells have a phenotype characterized by expression of CD57 and/or lack of the costimulatory receptor CD28 (16–18) that is associated with both cytotoxicity and senescence. Most of our understanding of CD57 as a marker of cytotoxicity comes from studying CD57⁺ CD8 T cells, which have limited proliferation capacities but maintain cytotoxic potential (19, 20). CD57⁺ CD4 T cells express a number of cytolytic elements, including granzymes A, B, and K and perforin, and exhibit killing capacity comparable to that of cytotoxic CD8 T cells in human and animal models (21–23). Cytotoxic CD4 T cells have been identified in patients with a range of cardiovascular conditions including acute coronary syndrome and myocardial infarction, in which expansions of this population were associated with increased acute mortality and disease recurrence (24–26). Cytotoxic CD57⁺ CD4 T cells are also observed in PLWH (21, 27). The proportion of circulating CD4 T cells that expresses CD57 is higher in untreated PLWH compared with matched HIV-uninfected controls, and CD57 expression does not normalize even in patients receiving at least 6 mo of ART (28). Because nearly all PLWH are coinfected with CMV, we hypothesize that CMV is an important driver of the expansion of these cytotoxic cells.

The mediators of cytotoxic CD4 T cell homing to sites of vascular inflammation in PLWH are not well understood, although several lines of evidence suggest that the vascular-endothelium homing receptor CX3CR1 may be involved. First, genetic polymorphisms in CX3CR1 that limit its surface expression are associated with reduced coronary artery disease in the general population, independent of other risk factors (29, 30). Second, in HIV-uninfected patients with unstable angina, plaque rupture is predicted by numbers of CX3CR1⁺ lymphocytes in the blood and by plasma levels of fractalkine (CX3CL1), the ligand for CX3CR1 expressed by endothelial cells (31, 32). Third, lymphopenia following primary percutaneous coronary intervention is predicted by lymphocyte CX3CR1 expression and coincides with serum CX3CL1 levels in HIV-uninfected patients with ST-elevation myocardial infarction (33). Fourth, in PLWH, plasma levels of CX3CL1 are elevated compared with plasma levels in HIV-uninfected controls (34). Fifth, a higher proportion of peripheral blood CD4 T cells express CX3CR1 in PLWH than in HIV-negative individuals (35),and CX3CR1⁺ CD4 T cells have been linked to increased carotid intima-media thickness (IMT) in PLWH (36). Notably, although CX3CR1 expression on CD8 T cells is associated with cytotoxic potential (20, 37, 38), and CX3CR1⁺ CD8 T cells are expanded in PLWH (20, 35, 39), it is not clear whether CX3CR1-expressing CD4 T cells are also cytotoxic. Furthermore, the low CD28 expression on cytotoxic CD4 T cells suggests that they may use alternate mechanisms of costimulation. Indeed, interaction between CD2 and its ligand LFA-3 is an effective costimulatory pathway for CD28⁻ T cells (40). As all T cells express CD2 and endothelial cells can express LFA-3 (41, 42), the CD2/LFA-3 axis is a strong candidate for providing an alternative method of costimulation for CD28⁻ cytotoxic CD4 T cells, particularly when targeting vascular tissues.

In this article, we aim to clarify the roles of CMV coinfection and CD2/LFA-3 interactions in the accumulation and function of CD57⁺ cytotoxic CD4 T cells in PLWH. We propose a mechanism by which these cells traffic toward vascular endothelium, setting the stage for further endothelial damage. We hypothesize that cytotoxic CD4 T cells are a key link between immune activation and CVD risk in PLWH as well as in the general population. To explore this further, we use atherosclerotic plaque tissue of HIV-uninfected patients undergoing carotid endarterectomy to examine CD57⁺ CD4 T cell infiltration and expression of CX3CL1 and LFA-3 in plaque. We identify a potential pathway of cytotoxic CD4 T cell contribution to CVD, by CX3CR1-directed migration and LFA-3-enhanced degranulation. Our findings may help bridge the gap in understanding how CMV infection and immunity are linked to increased CVD risk in PLWH.

All human experiments were approved by the institutional review boards of University Hospitals Cleveland Medical Center and Moscow State University of Medicine and Dentistry. With informed consent, and in accordance with the Declaration of Helsinki, blood was acquired in Vacutainer tubes containing EDTA (BD) from PLWH receiving combination ART with plasma HIV RNA <40 copies/ml (CMV-seronegative, n = 12; CMV-seropositive, n = 67) and HIV-uninfected controls (n = 17). Participant characteristics are shown in Table I. A subset of the HIV-infected CMV-seropositive donors had been previously characterized (43) for whether they had asymptomatic CMV replication in their seminal plasma (shedders, n = 15) or not (nonshedders, n = 14). We also collected peripheral blood and plaque tissues from donors not known to be infected with HIV undergoing clinically indicated carotid endarterectomy (Cleveland cohort, n = 14; Moscow cohort, n = 10). Characteristics of endarterectomy donors are shown in Table II. PBMCs were purified by centrifugation over a Ficoll-Paque cushion (GE Healthcare). In some cases, T cells were further enriched using the Pan T cell Isolation kit (Miltenyi Biotec) and AutoMACS Pro Separator (Miltenyi Biotec) system. For atherosclerotic plaque analysis, discarded plaque tissue was acquired in sterile saline following clinically indicated carotid endarterectomy procedures and processed as described previously (44). In short, plaques were washed in PBS, weighed, and cut into 2-mm pieces with a scalpel. Pieces were flash frozen in OCT reagent in a cryomold and preserved at −80°C for sectioning and immunostaining. To derive a single-cell suspension, plaque pieces were digested with 0.2 mg/ml DNase I (Roche) and either 0.016 mg/ml Liberase (Roche) or 1.25 mg/ml Collagenase XI (Sigma-Aldrich) by shaking at 250 rpm for 1 h at 37°C, strained through a 40-μm nylon cell filter (BD), and then washed with PBS.

Lymphocytes were identified by forward and side scatter, and phenotype was assessed using the following fluorochrome-conjugated Abs: anti-CD2 (clone RPA-2.10; BD), anti-CD3 (UCHT1; BD), anti-CD4 (SK3; BD), anti-CD8 (SK1; BD), anti-CD27 (M-T271; BD), anti-CD28 (CD28.2; BioLegend), anti-CD45RO (UCHL1; BD), anti-CD57 (HNK-1; BioLegend), anti-CD69 (FN50; BD), anti-CD122 (TU27; BioLegend), anti-CCR7 (3D12; BD), anti–PD-1 (EH12.1; BD), anti–TIM-3 (F38-2E2; BioLegend), anti-TIGIT (MBSA43; eBioscience), anti-CD244/2B4 (2-69; BD), anti–LAG-3 (T47-530; BD), anti–CTLA-4 (BNI3; BioLegend), and anti-CX3CR1 (2A9-1; BioLegend). Cells were stained for 20 min in the dark at room temperature, washed, and fixed in PBS containing 1% paraformaldehyde. The data were acquired on a LSRFortessa flow cytometer (BD). Viable cells were gated using Live/Dead Aqua viability dye (Invitrogen) per the manufacturer’s instructions. For intracellular detection of transcription factors, after surface staining, cells were fixed and permeabilized with Cytofix/Cytoperm (BD) for 20 min on ice, then stained for 40 min on ice with anti–T-bet (eBio4B10; eBioscience) and anti-Eomes (WD1928; eBioscience). For detection of intracellular cytokines, cells were cultured for 6 h in the presence of brefeldin A (GolgiPlug; BD), anti-CD107a (H4A3; BD), and medium control or with 5 μg/ml plate-bound anti-CD3 (HIT3a; BD) alone, with 5 μg/ml soluble anti-CD28 (CD28.2; BD) or with 5 μg/ml plate-bound recombinant human LFA-3 protein (R&D Systems). After Live/Dead and surface staining, cells were fixed and permeabilized with Cytofix/Cytoperm for 20 min on ice and stained for 40 min on ice with anti–IFN-γ, (B27; BD), anti-TNF (MAb11; BD), anti–MIP-1β (D21-1351; BD), and anti–IL-2 (5344.111; BD). For intracellular accumulation of cytolytic molecules and transcription factors after stimulation, cells were treated with 20 ng/ml IL-15 (R&D Systems) or control for 48 h, then harvested, stained with Live/Dead and surface Abs, treated with Cytofix/Cytoperm, stained with anti–granzyme B (GB11; BD), anti-perforin (B-D48; BioLegend), anti–c-myc (9E10; R&D Systems), and anti–Bcl-2 (Bcl-2/100; BD). MitoTracker Green (Thermo Fisher Scientific) labeling was performed following the manufacturer’s instructions. For analysis of STAT5 phosphorylation, cells were treated with 20 ng/ml IL-15 (R&D Systems) or control for 45 min, then fixed in 16% formaldehyde, washed, and treated with 90% methanol for 20 min at −20°C. Cells were then stained with an Ab mixture containing fluorochrome-conjugated anti-STAT5 pY694 (47/STAT5 [pY694]; BD) for 40 min on ice.

Deparaffinized sections were rehydrated and processed at 95°C for 20 min in 10 M sodium citrate, 0.05% Tween 20 (pH 6) for epitope retrieval, followed by blocking with 2% BSA in TBS + Triton X-100 0.025% for 1 h at room temperature. For immunostaining, cells and sections were incubated overnight at 4°C with affinity-purified primary Abs to CX3CL1 (goat polyclonal; R&D Systems) and LFA-3 (1C3; BD), or appropriate isotype controls followed by staining with Alexa Fluor 488– or Cy3-conjugated anti-goat and anti-mouse secondary Abs (Life Technologies), as per the manufacturer’s recommendations, then mounted in Vectashield with DAPI (Vector Laboratories) for fluorescence microscopy.

Sections of carotid atherosclerotic plaques were imaged by epifluorescence microscopy using 20×, and 100× oil immersion objectives. ImageJ software (National Institutes of Health) was used to analyze the acquired digital images of aortic endothelium. Briefly, images of each fluorescence channel were converted to eight-bit monochrome images and the background value was subtracted to 0. The representative numerical fluorescence intensity values were measured for the respective target signal and DAPI. We determined the target to nuclear fluorescence ratio (Fc/n) according to the formula Fc/n = (Fc-Fb)/(Fn-Fb), where Fb is background autofluorescence (45). We also applied iterative deconvolution methods (up to 10 iterations) to enhance and study high-resolution images.

Human aortic endothelial cells (HAoECs; PromoCell) were cultured in manufacturer-specified endothelial cell growth medium for 7 d. Purified T cells were added to the cultures on the seventh day, separated from HAoECs by a 5-μm collagen-coated Transwell membrane (product no. 3496; Corning) and further cultured in RPMI 1640 medium (Life Technologies) supplemented with 10% FBS (Gemini Bio-Products), 1% penicillin/streptomycin (Life Technologies), and 1% l-glutamine (Life Technologies). The T cells in the upper chamber were prestimulated with recombinant human IL-15 (20 ng/ml) for 2 d. The HAoEC monolayer lower chamber was activated with human TNF (10 μg/ml) (R&D Systems) for 7 d. T cells from the upper and lower chambers were harvested separately at indicated time points and examined by flow cytometry. Absolute cell counts from the upper and lower chambers were determined by adding Liquid Counting Beads (BD) prior to analysis by flow cytometry.

Primary HAoECs (PromoCell) were treated for 7 d with recombinant human TNF (R&D Systems). Relative expression of CX3CL1 was measured by real-time PCR (TaqMan Gene Expression Assays; Thermo Fisher Scientific) and in supernatant by ELISA (R&D Systems). Magnetic bead–purified memory CD4 T cells (Memory CD4 T Cell Isolation kit; Miltenyi Biotec) were stimulated with recombinant human IL-15 (247-ILB; R&D Systems) for 2 d. Supernatants were tested for CX3CL1 and TNF protein by ELISA (R&D Systems). Supernatants were then diluted 1:2 and cultured with HAoECs. After 7 d, culture supernatants were harvested and tested by ELISA for CX3CL1.

Comparisons between unrelated groups used nonparametric two-tailed Mann–Whitney U tests. Comparisons among three or more groups were performed with nonparametric Kruskal–Wallis tests with Dunn multiple comparison posttests. Paired group analyses used Wilcoxon matched-pairs signed rank test. Paired comparisons among three or more groups were performed with nonparametric Friedman test. All statistics were performed using Prism 8 software (GraphPad). Boolean-gated cytokine pie charts were compared using SPICE software (National Institutes of Health) with 10,000 permutations per test. Differences were considered significant if the p value <0.05.

Expression of CD57 on T cells is associated with cytotoxicity, and cytotoxic T cells have been linked to CVD risk (20, 21, 24–26, 36). CD57⁺ CD4 T cells have been identified in the peripheral blood during HIV and other infections, as well as in the elderly (28, 46–48). In this study, in a cohort of ART-treated PLWH with CMV coinfection (Table I), we found that CD57⁺ CD4 T cells were confined almost exclusively to the memory T cell (Tmem) compartment (Fig. 1A). Expression of CD57 on CD4 Tmem was highest within the effector memory (CD45RO⁺CCR7⁻CD27⁻) and terminally differentiated (CD45RO⁻CCR7⁻) compartments (Fig. 1B). CD57⁺ CD4 T cells had a robust effector phenotype typified by high T-bet and low eomesodermin (Eomes) expression and elevated levels of granzyme B and perforin (Fig. 1C). CD57⁺ CD4 Tmem also had elevated expression of inhibitory receptors, including PD-1, TIM-3, CD244, and LAG-3, compared with CD57⁻ CD4 Tmem (Fig. 1D).

Expression of CD57 on T cells is associated with shorter lifespan, poor replicative capacity, and immune senescence (19, 28, 49). These observations led us to consider what was driving the accumulation of CD57⁺ CD4 T cells within the Tmem compartment in HIV infection. Surprisingly, similar proportions of CD57⁺ and CD57⁻ Tmem were in cell cycle, as evidenced by intracellular expression of Ki67 (Fig. 1E). Notably, we found equivalent expression of the antiapoptotic molecule Bcl-2 in CD57⁺ and CD57⁻ CD4 Tmem, suggesting that viability was not reduced in CD57⁺ CD4 Tmem (Fig. 1F).

We next asked whether CMV coinfection, which has been shown to drive CD8 T cell expansion, differentiation, and activation in PLWH (12, 50), had a similar effect on CD57⁺ CD4 Tmem. PLWH who were CMV-seronegative had very few CD57⁺ CD4 Tmem compared with CMV-seropositive donors (Fig. 2A) and the few CD57⁺ cells that the seronegative donors did maintain exhibited lower surface expression of the CD57 molecule (Fig. 2B). Among a larger cohort of male CMV-seropositive donors, asymptomatic seminal CMV shedding, which is highly correlated with plasma CMV DNA levels (51), had no discernible effect on the proportion of CD4 Tmem that expressed CD57 (Fig. 2C), suggesting active CMV replication was not necessary for the continued prevalence of CD57⁺ CD4 Tmem. These observations led us to ask whether CD57⁺ CD4 Tmem from CMV-seropositive donors were fundamentally different from those from CMV-seronegative donors. Using 11-parameter mean fluorescence intensity (MFI) data acquired by flow cytometry, we compared CD57⁺ CD4 Tmem in CMV-seronegative and CMV-seropositive donors by principal component analysis and found that CD57⁺ CD4 Tmem were significantly different (p = 0.0018, Mann–Whitney U test) along PC1, which was driven mainly by the expression of CD28 (Fig. 2D) and CD57 (as evident in Fig. 2B).

Interactions of CD2 with its ligand LFA-3 (also called CD58) have been shown to be the strongest costimulatory pathway for T cells that lack CD28 (40), so we next examined the expression of the costimulatory receptors CD28 and CD2 on Tmem. CD57⁺ CD4 Tmem from CMV-seropositive donors were mostly CD28⁻ and had high expression of CD2, whereas in CMV-seronegative donors those few cells that were CD57⁺ retained CD28 and had relatively lower CD2 expression (Fig. 2E). In both seropositive and seronegative populations, CD57⁻ CD4 Tmem had high CD28 and low CD2 expression.

These findings led us to test whether costimulation via the CD2/LFA-3 axis would enhance the function of CD57⁺ CD4 Tmem, which might be exhausted because of elevated inhibitory receptor expression (Fig. 1D). We stimulated PBMCs from CMV-seronegative and -seropositive PLWH by traditional TCR activation (plate-bound anti-CD3 mAb; pbCD3) alone, with CD28 costimulation (soluble anti-CD28 mAb; sCD28), or with LFA-3 costimulation (plate-bound rLFA-3 protein; pbLFA-3) and measured IFN-γ, IL-2, MIP-1β, and TNF expression by intracellular flow cytometry and surface expression of the degranulation marker CD107a (Fig. 3A, Supplemental Fig. 1A). Even though CD57⁻ and CD57⁺ CD4 Tmem from CMV-seropositive donors had very different levels of CD28 expression, the responses of these populations to CD28 costimulation was similar. For all populations, LFA-3 costimulation provided a significantly enhanced response compared with CD28 costimulation (Fig. 3B) or to TCR activation without costimulation (Supplemental Fig. 1A). Consistent with their elevated expression of CD2, CD57⁺ cells gave a more robust response than CD57⁻ cells after LFA-3 costimulation in both donor groups, and CD57⁺ CD4 Tmem from CMV-seropositive donors had significantly more robust response to LFA-3 costimulation than CD57⁺ cells from CMV-seronegative donors (p = 0.03, SPICE pie comparison). For CMV-seropositive donors, the range of responsiveness of CD57⁺ CD4 Tmem was quite broad, suggesting there might be heterogeneity in this population. Much of the CD2/LFA-3 axis–enhanced polyfunctionality was due to increased IFN-γ, IL-2, and TNF production, regardless of CD57 expression (Fig. 3C). Both costimulatory pathways resulted in similar levels of degranulation, as measured by CD107a expression. Interestingly, when compared with anti-CD3 stimulation alone, CD28 costimulation significantly enhanced IL-2 production, even in CD57⁺ cells from CMV-seropositive donors (Supplemental Fig. 1B), suggesting that the low, residual CD28 expression on CD57⁺CD28⁻ CD4 Tmem in CMV-seropositive donors could still provide some costimulatory efficacy. Additionally, our findings suggest that CD57⁺ CD4 Tmem are not functionally exhausted in terms of ability to express cytokine and CD107a despite elevated inhibitory receptor expression.

The cytokine IL-15 activates T cells, enhances mitochondrial activity, and promotes the intracellular accumulation of cytolytic molecules such as granzyme B and perforin (52, 53). IL-15 expression in lymph nodes correlates with CD8 T cell numbers in the periphery in chronic (viremic) HIV infection, suggesting that IL-15 may have a role in the immune response to HIV infection (52). In addition, there is evidence that IL-15 expression is increased in the bone marrow of CMV-seropositive compared with CMV-seronegative donors in HIV-uninfected individuals undergoing hip replacement surgery (54). We therefore wanted to investigate the effects of IL-15 on CD57⁺ CD4 Tmem. In CMV-seropositive PLWH, CD57⁺ CD4 Tmem more often express CD122, the β-chain of the IL-2/IL-15 receptor, than do CD57⁻ CD4 Tmem (Fig. 4A) (although both populations have abundant CD122 expression), suggesting potential responsiveness to IL-15 signals. Stimulation with IL-15 promoted robust phosphorylation of STAT5 within 45 min (Fig. 4B) and, by 2 d, had enhanced expression of the master transcriptional regulator c-myc (Fig. 4C), induced surface expression of the early activation and resident memory marker CD69 (Fig. 4D), and promoted intracellular expression of the cytolytic molecules granzyme B and perforin (Fig. 4E) in both CD57⁻ and CD57⁺ CD4 Tmem populations obtained from HIV-uninfected donors. The enhancement of cytolytic molecule expression by IL-15 was significantly more robust among CD57⁺ cells than among the CD57⁻ populations. IL-15 also promoted the proliferation of both CD57⁻ and CD57⁺ CD4 Tmem (Fig. 4F), suggesting that CD57⁺ CD4 Tmem are not replicatively senescent but rather can proliferate in response to IL-15. Additionally, IL-15 enhanced expression of the prosurvival factor Bcl-2 in both CD57⁻ and CD57⁺ CD4 Tmem (Fig. 4G). These results contrast to the effect of IL-15 on mitochondrial biogenesis—mitochondrial mass was increased by IL-15 only in CD57⁺ CD4 Tmem (Fig. 4H). Thus, CD57⁺ CD4 Tmem are highly responsive to IL-15, which can be both a homeostatic and an inflammatory stimulus.

CD4 Tmem that lack CD28 express both CD57 and the endothelium-homing receptor CX3CR1 (46, 55), and CX3CL1 promotes the migration of CX3CR1⁺ CD4 T cells in vitro (36, 56). In addition, the proportion of CX3CR1⁺ CD4 T cells positively correlates with carotid IMT and IMT progression over time in PLWH (36). CX3CR1⁺ CD4 T cells might therefore provide a cellular link from HIV to CVD risk, which is increased in PLWH (2). Only CD57⁺ CD4 Tmem from CMV-seropositive PLWH donors exhibited CX3CR1 expression (Fig. 5A). Although IL-15 treatment upregulated surface expression of CX3CR1 in both CD57⁻ and CD57⁺ CD4 Tmem from CMV-seropositive PLWH, CD57⁺ cells showed significantly greater CX3CR1 upregulation in response to IL-15 (Fig. 5B). One of the effector molecules whose TCR stimulation-induced production was enhanced by LFA-3 costimulation was TNF, and TNF is a potent inducer of CX3CL1 gene expression and protein secretion by HAoECs in vitro (Supplemental Fig. 2A). Notably, although IL-15 induces TNF production from sorted memory CD4 T cells in vitro (Supplemental Fig. 2B), the amount produced is insufficient to elicit CX3CL1 production from HAoECs in culture (Supplemental Fig. 2C). In contrast, both CD3/CD28 or CD3/LFA-3 stimulation elicited robust release of TNF (Supplemental Fig. 2D) and likely other factors sufficient to induce endothelial cell CX3CL1 expression (Supplemental Fig. 2E). Importantly, none of the stimulations elicited CX3CL1 production by the sorted CD4 Tmem (Supplemental Fig. 2F). We therefore established a short-term Transwell migration assay to see if IL-15 and/or TNF promoted CD57⁺ CD4 Tmem migration (Fig. 5C). T cells from PLWH that were pretreated, or not, with IL-15 (at a concentration sufficient to enhance CX3CR1 expression, Fig. 5B) were incubated with HAoECs that had been pretreated, or not, with TNF (at a concentration sufficient to elicit CX3CL1 secretion, Supplemental Fig. 2A). Examples of CX3CR1 and CD57 staining prior to the migration assay are shown in Supplemental Fig. 3A. After 3 h, CD57⁺ CD4 Tmem in the upper and lower chambers were recovered and quantified. CD4 T cells that had migrated to the lower chamber were enriched for CD57 (Fig. 5D, 5E), and the CD57⁺ cells trended toward lower expression of CX3CR1, consistent with exposure to CX3CL1 (Supplemental Fig. 3B). We next quantified the numbers of CD57⁺ CD4 Tmem recovered from the lower chambers and expressed those numbers as a percentage of the total starting numbers of CD57⁺ CD4 Tmem in each assay well. We found an increase in the number of CD57⁺ CD4 Tmem that migrated when exposed to IL-15, or both TNF and IL-15, compared with medium only control (Fig. 5F).

Intriguingly, the percentage of CD57⁺ CD4 Tmem that expressed CX3CR1⁺ at baseline predicts the amount of induced migration of CD57⁺ CD4 T cells in the IL-15 and TNF cotreated condition (Fig. 5G), suggesting that the CX3CL1/CX3CR1 axis may be important for the migration of CD57⁺ CD4 T cells toward activated endothelium. To test this, we treated the T cells with the CX3CR1 antagonist AZD8797 (57–59) prior to the migration assay. Although we found no significant reductions in the proportion of CD57⁺ CD4 T cells that migrated in any of the experimental groups following AZD8797 treatment (Supplemental Fig. 3C), we did observe an interesting correlation in the IL-15 and TNF cotreated group. For donors with proportional enrichment of CD57⁺ cells in the lower chamber (values greater than median), AZD8797 treatment resulted in a significant inhibition of migration, but for donors without CD57⁺ cell enrichment (values less than or equal to the median), AZD8797 treatment paradoxically resulted in more CD57⁺ CD4 T cell migration (Fig. 5H). In other words, for donors whose CD57⁺ CD4 T cells were more migratory than the overall CD4 T cell pool, AZD8797 treatment impaired the migration, but for other donors, AZD8797 treatment resulted in increased migration by all CD4 T cells, with no selective effect on CD57⁺ cells. Thus, our data show that there is heterogeneity in the susceptibility of CD57⁺ CD4 T cells to CX3CL1 and suggest other factors are likely also important in the migration of CD57⁺ CD4 T cells toward activated endothelial cells, particularly in the absence of CX3CR1 functionality.

CVD is a leading non-AIDS cause of death in PLWH and in the general population (60–62). Elevated CVD risk in PLWH could be due to an overabundance of factors that also promote CVD in the general population, such as persistent inflammation, immune cell activation, and CMV infection (1, 2), and we hypothesize that risk is linked to CD57⁺ CD4 T cells. We therefore examined the CD4 T cell infiltrate in atherosclerotic plaques of HIV-uninfected persons undergoing clinically indicated carotid endarterectomy to characterize the cells that accumulate at sites of vascular damage in the general population. Endarterectomies were performed at two locations (Cleveland, OH, and Moscow, Russia), and experiments were run in parallel (Table II). In both cohorts we observed a profound upregulation of CD69 expression on CD4 T cells recovered from plaques compared with CD69 expression on peripheral blood cells from the same donors (Fig. 6A). CD69 expression on plaque CD4 T cells is reflective of activation and a tissue resident phenotype (63). Notably, IL-15, which promotes CD69 expression on CD4 Tmem in vitro (Fig. 4D), has been demonstrated within plaque tissues (64, 65). Conceivably, CD69 expression on plaque T cells is the result of recent exposure to inflammatory cytokine or TCR signals. CD57⁺ CD4 Tmem within plaques also had evidence of CX3CL1 exposure (Fig. 6B), suggesting these cells were capable of trafficking to sites of endothelium damage in vivo. Immunohistologic examination of the plaque tissue showed expression of CX3CL1 and LFA-3 on the vascular endothelium (Fig. 6C). Thus, our data are consistent with a model in which CX3CR1⁺ CD57⁺ CD28⁻ CD4 Tmem migrate toward CX3CL1-expressing endothelial cells in the nascent plaque and potentially contribute to vascular damage via cytokine and lytic granule release induced by CD2/LFA-3–mediated costimulation.

Identifying targetable factors that promote atherosclerosis is a major goal of current cardiovascular research. It is becoming increasingly clear that cardiovascular complications like myocardial infarction have immune and inflammatory etiologies, and the contributions of immune cells such as CD4 T cells to these processes are coming into focus. In this article, we show that CMV coinfection, a major driver of inflammation and T cell activation and expansion (3, 11–14, 50), is associated with an expanded subpopulation of CD4 Tmem that express CD57 in the peripheral blood of PLWH. CD57⁺ CD4 Tmem are enriched for cytolytic molecules and expression of the vascular endothelium-homing receptor CX3CR1. CD4 Tmem can be isolated from atherosclerotic plaques, suggesting they can traffic to sites of endothelial activation and may contribute to CVD by damaging endothelial cells via targeted release of cytolytic granules or cytokines that can induce endothelial dysfunction.

CD57⁺ CD4 Tmem in CMV-seropositive PLWH lack the costimulatory receptor CD28 but have elevated levels of CD2. Costimulation via the CD2/LFA-3 axis elicits a polyfunctional response that is enhanced in CD57⁺ CD4 Tmem compared with that in CD57⁻ CD4 Tmem, which have lower CD2 expression. We also show that LFA-3 is highly expressed in plaques in close proximity to CX3CL1, providing an environment in which CD57⁺ CD4 Tmem could use CX3CR1/CX3CL1 interactions to migrate into an inflamed vasculature. To explore this, we established a short-term chemoattraction assay and demonstrated that IL-15–treated CD57⁺ CD4 T cells upregulate CX3CR1 expression and exhibit greater migration toward TNF-treated cultured primary endothelial cells that upregulate CX3CL1. Additionally, the proportion of CD57⁺ CD4 Tmem that expressed CX3CR1 in the absence of treatment predicted the magnitude of induced migration. Our data are consistent with a model in which proinflammatory effector molecules promote CVD by activating CD57⁺ CD4 Tmem and endothelial cells. Reciprocally, activated CD57⁺ CD4 Tmem release TNF to further promote endothelial cell activation and CX3CL1 release. CMV infection drives the expansion of CD57⁺ CD4 Tmem and could be a key component in this model.

The elements that CD57⁺ CD4 T cells recognize in the vasculature or in plaque tissues are undefined, although it is likely that many CD57⁺ CD4 Tmem are specific for CMV (7, 48, 66). Latent CMV is harbored within CD34⁺ hematopoietic stem cells; the virus can reactivate as infected cells differentiate along the myeloid lineage into macrophages (67). If monocytes carrying CMV genomes enter the subendothelium and differentiate into CMV Ag-expressing foam cells, these could provide the source of CMV Ags in the context of MHC class II. We recently showed that activated T cell–derived TNF promotes expression of the procoagulant tissue factor on monocytes in vitro (68). Thus, interactions between CMV-specific CD4 T cells and CMV-infected myeloid cells could plausibly contribute mechanistically to clot formation and CVD development. Alternatively, vascular endothelial cells can express MHC class II (42). Because CMV has been shown to infect endothelial cells in vitro (69, 70), and CMV DNA can be found in vascular tissues (71, 72), CMV-infected endothelial cells might be capable of presenting both CMV Ags and LFA-3 costimulatory signals to infiltrating CD57⁺ CD4 Tmem as well as to CMV-specific CD8 T cells.

Interestingly, the nominally homeostatic cytokine IL-15 could be a key player in our observations. IL-15 activates T cells and NK cells and has been postulated to act as a danger signal to “call in the troops” to sites of infection or tissue damage (73). Consistent with this role, IL-15 has been demonstrated within atherosclerotic plaques (64, 65), and IL-15 promotes the proliferation of CD57⁺ CD4 T cells from patients with rheumatoid arthritis (74). Our data in this article suggest that IL-15 in plaques may serve a mechanistic role in atherosclerosis at least in part by activating and enhancing the migration of cytotoxic CD57⁺ CD4 Tmem and possibly by directly eliciting release of TNF and/or other proinflammatory factors. Circulating levels of IL-15 are increased in PLWH (75), and we find increased IL-15 protein in aortas of SIV/simian-HIV–infected macaques compared with uninfected control animals (S. Panigrahi, B. Chen, M. Fang, D. Potashnikova, A.A. Komissarov, A. Lebedeva, G.M. Michaelson, J.M. Wyrick, S.R. Morris, S.F. Sieg, M. Paiardini, F.J. Villinger, K. Harth, V.S. Kashyap, M.J. Cameron, C. M. Cameron, E. Vasilieva, L. Margolis, S.-A. Younes, N.T. Funderburg, D.A. Zidar, M.M. Lederman, and M.L. Freeman, submitted for publication); thus, IL-15 may be an important contributor to the increased CVD risk in PLWH.

In conclusion, we have identified a potential mechanism for the contribution of cytotoxic CD57⁺ CD4 Tmem to the elevated CVD risk in PLWH and the HIV-infected elderly: CD57⁺ CD4 Tmem expressing CX3CR1 are recruited to the inflamed vasculature where they could further damage endothelium via CD2/LFA-3–mediated degranulation. Future studies are needed to confirm and build upon our findings in vitro and in vivo. Food and Drug Administration–approved drugs that target CMV (e.g., valganciclovir, letermovir) and TNF (e.g., adalimumab, etanercept) provide therapeutic strategies that could be applied in the clinic. Indeed, blocking TNF activity improves endothelial function in patients with rheumatoid arthritis (76). As more therapies to target CX3CL1, IL-15, and CD2/LFA-3 interactions become available, it will be important to examine them in the settings of HIV/CMV coinfection and elderly HIV-uninfected persons with atherosclerosis.

We thank Steven Juchnowski and Sadeer Al-Kindi for their help acquiring atherosclerotic plaque tissue specimens, and Daniela Moisi and Dominic Dorazio for their excellent technical assistance.

N.T.F. serves as a consultant for Gilead. The work of L.M. was funded by the Eunice Kennedy Shriver National Institute of Child Health and Human Development Intramural Program. The other authors have no financial conflicts of interest.