Persistence of a skewed repertoire of NK cells in people with HIV-1 on long-term ART

Abstract

HIV-1 infection greatly alters the NK cell phenotypic and functional repertoire. This is highlighted by the expansion of a rare population of FcRγ^– NK cells exhibiting characteristics of traditional immunologic memory in people with HIV (PWH). While current antiretroviral therapy (ART) effectively controls HIV-1 viremia and disease progression, its impact on HIV-1-associated NK cell abnormalities remains unclear. To address this, we performed a longitudinal analysis detailing conventional and memory-like NK cell characteristics in n = 60 PWH during the first four years of ART. Throughout this regimen, a skewed repertoire of cytokine unresponsive FcRγ^– memory-like NK cells persisted and accompanied an overall increase in NK surface expression of CD57 and KLRG1, suggestive of progression toward immune senescence. These traits were linked to elevated serum inflammatory biomarkers and increasing antibody titers to human cytomegalovirus (HCMV), with HCMV viremia detected in approximately one-third of PWH at years one through four of ART. Interestingly, 40% of PWH displayed atypical NK cell subsets, representing intermediate stages of NK-poiesis based on single-cell multiomic trajectory analysis. Our findings indicate that NK cell irregularities persist in PWH despite long-term ART, underscoring the need to better understand the causative mechanisms that prevent full restoration of immune health in PWH.

Introduction

As innate cytotoxic effectors, natural killer (NK) cells are characterized by their ability to lyse cells showing generic signals of stress, transformation, or infection without prior sensitization (1–4). However, they also function as helper cells, linking innate and antigen-specific adaptive immunity. Their supportive role in shaping adaptive immune responses is highlighted through their reciprocal crosstalk with dendritic cells (DCs) in the periphery and secondary lymphoid tissues, ultimately leading to enhanced DC-mediated priming of CD4⁺ Th1 and CD8⁺ cytotoxic T cell responses (5, 6). As immune modulators, NK cells also function as regulators of inflammation and immune homeostasis through the production of cytokines such as IL-13 and IL-10 (7). The capacity of NK cells to respond with great flexibility to innate environmental cues is crucial for the subsequent induction and regulation of effective adaptive T cell responses to viruses, including HIV-1 (8).

Diversity within the NK cell repertoire is derived from the expression of unique combinations of germline-encoded activating and inhibitory receptors. The interplay between these competing signals is not only complex—dictating the response potential and activation threshold of each individual NK cell—but also highly dynamic (9, 10). That is, additive pathogen exposures promote alterations in the NK receptor repertoire over time, undoubtedly impacting the quality of immune responses to subsequent infections (11). In spite of this inherent complexity, NK cells are classically divided into two basic subsets based on relative surface expression of CD56 (the 140-kDa isoform of neural cell adhesion molecule) (12, 13) and CD16 (or FcγRIIIA, the low-affinity receptor for IgG) (14–16). The high cytokine-producing CD56^brightCD16⁻ subset accounts for a maximum of 10%, and the cytolytic CD56^dimCD16⁺ population represents approximately 90%, of circulating NK cells (7, 14, 17–19). These populations are easily distinguished by flow cytometry, as well as their distinct phenotypic and functional properties, which lend credence to the linear model of differentiation, whereby CD56^bright NK cells are the immediate immature precursors of the CD56^dim population (14, 20, 21).

NK cells also are capable of acquiring attributes of traditional immunologic memory, with adaptive and memory-like NK cells developing in either an antigen-dependent or -independent manner (22–29). Initial studies described IL-18 as being uniquely capable of driving the differentiation of resting CD56^dimCD16⁺ NK cells into a functionally stable helper subset. That is, IL-18 promotes a CD16⁻/CD83⁺/CD25⁺/CCR7⁺ phenotype and primes lymph node-homing NK helper cells for immediate responsiveness to secondary exposures to stimulatory factors (5, 30). Subsequent studies reported that combined exposure to IL-12, IL-15, and IL-18 induces a pool of cytokine-induced memory-like NK cells, which are characterized by enhanced recall responses to cytokines or tumor targets (22, 26).

A second population of memory-like NK cells, originally identified in association with human cytomegalovirus (HCMV) infection and reactivation, is distinguished by a lack of the intracellular signaling protein FcRγ and enhanced expression of NKG2C (24, 28, 31). This FcRγ^– memory-like population is distinct from adaptive NKG2C⁺ NK cells, though, with their expansions driven by interactions between CD16 and the Fc portion of antibodies or NKG2C and HLA-E, respectively (28, 32–35). As FcRγ^– NK cells are unresponsive to innate cytokine signaling, including that provided by IL-18, their capacity to provide immune help—to shape the response to a novel immunogen—is thus limited (36). However, they are highly responsive to the adaptive immune signaling provided by antibodies, demonstrating superior antibody-dependent cellular cytotoxicity that is presumably mediated through altered CD16 signaling via the zeta chain of CD3 (24, 28, 29, 31). Notably, FcRγ^– memory-like NK cells are more highly expanded in HIV-1 seropositive, compared to uninfected HIV-1 seronegative, individuals (29, 36–38). Since HIV-1 infection is associated with innate immune dysfunction (37, 39), the inflated expansion of FcRγ^– memory-like NK cells may be an important contributor to the chronic immune activation noted in people with HIV (PWH) resulting from subclinical reactivation of HCMV, a notion supported by higher HCMV antibody titers in HIV-1 infection (29, 40).

The effect of HIV-1 infection extends beyond the expansion of the FcRγ^– population, as NK cell subset distribution and function are sequentially deregulated in chronic infection (41–44). Chronic HIV-1 viremia promotes the expansion of highly dysfunctional CD56^–CD16⁺ NK cells, decreased expression of natural cytotoxicity receptors (NCRs), and increased levels of inhibitory NK cell receptors, with the exception of NKG2A (41–49). These phenotypic alterations are accompanied by limitations in NK cell helper and cytolytic functions (42–44, 50, 51). On average, 24 months of suppressive antiretroviral therapy (ART) are required before surface expression of CD56 and NKG2A normalize (42, 45, 52). A shift in the balance of inhibitory and activating receptors also occurs, which correlates with a recovery in functionality (44). However, it is highly plausible that HIV-1 infection has an enduring effect on NK cells, as PWH on effective ART show signs of residual immune dysfunction (53–55), which includes persistence of NK cell activation (29, 56).

Here we demonstrate in a longitudinal study that despite effective ART, NK cells continue to progress along a spectrum of differentiation. In other words, increasing time on ART is not associated with additional improvements in NK cell phenotype and function, with steadily increasing levels of CD57 and KLRG1, as well as a prominent FcRγ^– population that remains distinct from conventional NK cells. Interestingly, stimulation with IL-18 and IL-12 leads to downregulation of surface expression of KLRG1—a phenomenon that is noticeably absent in FcRγ^– NK cells. We also reveal a striking disruption in subset distribution within the NK cell compartment, with unusually large proportions of CD56^brightCD16⁺ and CD56^dimCD16⁻ NK cells maintained, representing intermediate stages of NK-poiesis based on single-cell multiomic trajectory analysis.

Materials and Methods

Study participants

Participants of two well established HIV-1 cohorts were included in this study. PWH (n = 14; median age 57 years), who self-identify as men who have sex with men (MSM), were randomly selected from the Pittsburgh clinical research site of the MACS/WIHS Combined Cohort Study (MWCCS) (57). These PWH recorded a plasma HIV-1 load less than 20 copies/ml at the time of their study visit, with a median virally controlled treatment duration of 12.1 years (range: 1.8–22.7) (Table 1). Age-matched people without HIV (PWOH; n = 14; median age 58 years), who self-identify as MSM, were also selected from the Pittsburgh MWCCS.

Table 1.

Characteristics of MWCCS participants

Age at study visit, median (Q1, Q3), years
PWH	57 (55, 64)
PWOH	58 (55, 64)
Race/ethnicity
White, non-Hispanic*	13 (93%)
Black, non-Hispanic*	1 (7%)
Highest plasma HIV-1 RNA, median (Q1, Q3), log10copies/ml	4.8 (4.4, 5.7)
Nadir CD4+ T cell count, median (Q1, Q3), cells/mm3	244 (144, 325)
CD4+ T cell count at study visit, median (Q1, Q3), cells/mm3	620 (473, 853)
Plasma HIV-1 RNA at study visit, copies/ml	< 20
Time to treatment, median (Q1, Q3), years	2.9 (1.4, 6.3)
Time on suppressive ART, median (Q1, Q3), years	12.1 (8.7, 17.4)
ART regimen at study visit
NNRTI + NRTI	14%
PI + NRTI	7%
InSTI + NRTI	71%
Other	7%

^*The race/ethnicity data applies to both PWH and PWOH; InSTI, integrase strand transfer inhibitor; NNRTI, non-nucleoside reverse transcriptase inhibitor; NRTI, nucleoside/nucleotide reverse transcriptase inhibitor; PI, protease inhibitor; PWH, people with HIV; PWOH, people without HIV; Q1, first/lower quartile; Q3, third/upper quartile

The n = 60 participants included in our present study (Table 2) were selected from the previously analyzed AIDS Clinical Trials Group (ACTG) A5321 longitudinal study (a total of n = 101 study participants) (58), which examined HIV-1 persistence and immunologic measures pre‐ART (year 0), as well as year 1, 4, and years 6 through 15 on-ART. All participants had documented long‐term plasma HIV‐1 RNA suppression (from week 48 of ART) and no reported ART interruptions of greater than 21 days through study entry to A5321. A5321 study participants had initiated ART in ACTG ART‐naïve studies and were followed in the ACTG observational study A5001 prior to enrollment in A5321. Further selection criteria of these ACTG study participants were based on clinical sample availability and having cell‐associated HIV‐1 DNA data at both year 1 and year 4 of ART (58), as well as plasma HIV-1 RNA data at year 4 by single-copy assay (59). The selected samples were within 52 weeks of the previously tested on‐ART timepoints of year 1 (TP1) and year 4 (TP2); if prior to ART year 1, the sample was within 16 weeks. In addition, early on‐ART samples (range: 1–8 weeks on-ART; TP0) were identified for a subset of ACTG participants (n = 20).

Table 2.

Characteristics of PWH of the ACTG A5321 study

Age at initiation of ART, median (Q1, Q3), years	39 (34, 46)
Female	20%
Race/ethnicity
White, non-Hispanic	33 (55%)
Black, non-Hispanic	11 (18%)
Hispanic (regardless of race)	15 (25%)
American Indian, Alaskan Native	1 (2%)
Pre-therapy plasma HIV-1 RNA, median (Q1, Q3), log10copies/ml	4.6 (4.2, 4.9)
Year 4 of ART single copy assay plasma HIV-1 RNA, median (Q1, Q3), copies/ml	0.5 (0.5, 2.1)
Year 1 of ART HIV-1 DNA, median, (Q1, Q3), log10copies/106 CD4+ T cells)	3.2 (2.8, 3.5)
Year 4 of ART HIV-1 DNA, median, (Q1, Q3), log10copies/106 CD4+ T cells)	2.9 (2.5, 3.2)
Pre-therapy CD4+ T cell count, median (Q1, Q3), cells/mm3	284 (175, 365)
Year 4 of ART CD4+ T cell count, median (Q1, Q3), cells/mm3	607 (470, 783)
Pre-therapy CD4:CD8 T cell ratio, median (Q1, Q3)	0.3 (0.2, 0.4)
Year 4 of ART CD4:CD8 T cell ratio, median (Q1, Q3)	0.8 (0.5, 1.0)
ART regimen, initial, at time of last sample collection
NNRTI + NRTI	62%, 65%
PI + NRTI	35%, 33%
InSTI + NRTI	0%, 2%
Other	3%, 0%

InSTI, integrase strand transfer inhibitor; NNRTI, non-nucleoside reverse transcriptase inhibitor; NRTI, nucleoside/nucleotide reverse transcriptase inhibitor; PI, protease inhibitor; PWH, people with HIV; Q1, first/lower quartile; Q3, third/upper quartile

Real-time quantitative PCR for HCMV and EBV DNA

Plasma samples were first centrifuged (Thermo Scientific^™ Sorvall^™ ST8R) at 800 x g for 2 minutes at 4°C. Prior to extraction, 250 μl of the spun plasma was combined with 1,950 μl of Guanidine thiocyanate (GuSCN)* Triton buffer (NUCLISENS^® easyMAG^®) and incubated at room temperature for 15 minutes. To measure proper extraction and efficiency of the quantitated PCR reaction, 10 μl of phocine herpesvirus (PhHV-1), an in-house propagated virus, was added as an internal control. A 20 μl TaqMan^™ PCR was performed by mixing 5 μl of viral DNA with TaqMan^™ Gene Expression Master Mix (Applied Biosystems^™ by Thermo Scientific^™), in addition to the appropriate forward and reverse primers (60–63) (Table 3) and TaqMan^™ Real-time PCR was performed using the ViiA 7 A&B Applied Biosystems^™ instrument (Life Technologies) and the following cycling conditions: 50°C for 2 minutes, 95°C for 10 minutes, 40 cycles of 95°C for 15 seconds, and 60°C for 1 minute. A no template control was included in each assay to control for PCR cross-contamination; each sample was assayed in duplicate and controls in triplicate. QuantStudio^™ Real-time PCR Software (Applied Biosystems^™ by Thermo Scientific^™) was used for PCR data analysis, and viral copy numbers were reported using plasmid controls previously extrapolated and verified from quantitated DNA controls, with copy numbers purchased from Advanced Biotechnology, Inc.

Table 3.

Real-time PCR forward and reverse primers

Target	Forward primer	Reverse primer	Reference(s)
HCMV	5’-CGATCAAGAACGCGATAACG-3’	5’-ACCGTCGATGGCAGGTCAT-3’	Sanghavi, et al. Ding, et al.
EBV	5’-AAACCTCAGGACCTACGCTGC-3’	5’-AGACACCGTCCTCACCAC-3’	Jebbink, et al.
PhHV-1	5’-GGGCGAATCACAGATTGAATC-3’	5’-GCGGTTCCAAACGTACCAA-3’	Ding, et al. Niesters

Quantitation of HCMV and EBV IgG antibody titers

The HCMV and EBV antibody titers of study participants were determined by quantitative CMV and EBV EBNA-1 IgG ELISA (GenWay), respectively, following the manufacturer’s protocol. Frozen, longitudinal plasma samples were thawed and analyzed in batches that included all samples for a participant. Plasma samples were diluted 1:101 with the provided sample diluent, and absorbance was read at 450 nm using the BioTek ELx800^™. The ready-to-use standards and controls of the ELISA kits were defined and expressed in arbitrary units (U/ml), resulting in an exact and reproducible quantitative evaluation. The absorptions of the standards and controls were graphed against their concentrations. From the resulting reference curve, the concentration values for each sample were then extracted in relation to their absorptions.

Peripheral blood mononuclear cell (PBMC) cultures

Cryopreserved PBMCs were thawed and cultured in Iscove’s Modified Dulbecco’s Medium (Gibco^®), containing 10% fetal bovine serum (Atlanta Biologicals) and 0.5% gentamicin (Gibco^®), in 24-well plates (Costar^®) at a density of 1.5×10⁶ cells/well. The phenotypic and functional profiles, as well as frequencies of FcRγ⁻ NK cells, were determined by flow cytometry (see Flow cytometry) on NK cell subsets following a 24h culture in media alone or in the presence of recombinant human (rh) IL-18 (hereinafter referred to as IL-18; 500 ng/ml; R&D Systems^®) and rhIL-12p70 (hereinafter referred to as IL-12; 50 ng/ml; R&D Systems^®). PBMCs were cultured and analyzed in batches that included all samples for a participant.

Antibody-mediated polyfunctional responses

NK cells, purified from fresh or cryopreserved PBMCs by magnetic bead negative selection using a human NK cell enrichment kit (EasySep^™), were plated in serum free AIM V^® medium (Gibco^®) in 96-well round bottom plates (Costar^®) at a density of 2×10⁵ cells/well. Raji cells were incubated for 30 minutes at 37°C with 10 μg/ml rituximab, washed two times with 1x PBS, and subsequently resuspended in serum free AIM V^® medium at a concentration of 2×105 cells/ml. 100 μl of the Raji cell suspension was plated per well of NK cells, resulting in an effector (NK) to target (Raji) cell ratio of 10:1. After incubating the co-cultures at 37°C for 6h, cells were harvested and analyzed for their expression of CD107a, TNFα, IFNγ, and FcRγ (see Flow cytometry).

Flow cytometry

Cells were pre-exposed for 15 minutes to 50 μg/ml of unfractionated murine IgG (Sigma-Aldrich) to block nonspecific Fc receptor binding before immunostaining. The LIVE/DEAD^™ Fixable Aqua Dead Cell Stain (Life Technologies) was used for viability exclusion, and the following antibodies were used for immunostaining: CD3-APC-H7 (clone SK7, BD Pharmingen^™), CD56-PE-Cy7 (clone N901, Beckman Coulter), CD57-BV421 (clone NK-1, BD Horizon^™), CD16-PerCP-Cy^™5.5 (clone 3G8, BD Pharmingen^™), NKG2A-APC (clone Z199, Beckman Coulter), NKG2C-PE (clone 134591, R&D Systems^®), NKp46-PE (clone BAB281, Beckman Coulter), PD-1-BV421 (clone EH12.1, BD Horizon^™), and KLRG1-APC (clone 13F12F2, eBioscience^™). Staining was done in FACS buffer consisting of 1x PBS solution (GE Life Sciences), 0.5% bovine serum albumin, and 0.1% sodium azide (Sigma). For intracellular protein expression, NK cells were fixed with BD Cytofix/Cytoperm^™ (BD Biosciences), permeabilized using BD Perm/Wash^™ (BD Biosciences), and labeled with anti-FcRγ-FITC (Milli-Mark^®), anti-IFNγ-Alexa Flour^® 700 (clone B27, BD Pharmingen^™), and/or anti-TNFα-PE (clone 6401.1111, BD FastImmune^™). For CD107a mobilization assays, cells were exposed to anti-CD107a-APC (clone H4A3, BD Pharmingen^™) in the presence of 0.1% monensin by volume (BD GolgiStop^™), followed by the viability exclusion, surface marker, and intracellular immunostaining procedures described above. Samples were stored in FACS buffer until data acquisition using a BD LSRFortessa^™ flow cytometer. Data were analyzed using FlowJo version 10.5.3 (Tree Star), with expression levels based on comparison to fluorescence minus one samples or unstimulated controls for intracellular cytokine staining (64).

Single-cell multiomic analysis

From the n = 60 PWH of the ACTG A5321 cohort included in our present study, n = 4 were sampled based on being representative of the pool of participants and together having the broad repertoire of NK cell subsets based on flow cytometry results. NK cells from the on-ART timepoint of 4 years (TP2) were purified from cryopreserved PBMCs by magnetic bead negative selection using a human NK cell enrichment kit (EasySep^™). The resulting enriched populations were stained with BD^® AbSeq antibodies against CD3 (clone SK7), CD56 (clone NCAM16.2), and CD16 (clone 3G8), in addition to donor-specific oligomer-conjugated multiplexing antibodies (BD Biosciences). Barcoded cells were then mixed in equal parts before loading onto a BD Rhapsody^™ cartridge per the manufacturer’s instructions. Single cell capture was performed using the BD Rhapsody^™ Express Single-Cell Analysis System, and quality control metrics were verified using the BD Rhapsody^™ Scanner. Prepared cDNA libraries of the whole transcriptome, generated according to the manufacturer’s instructions, were submitted to the Emory Yerkes National Primate Research Center Genomics Core, and sequencing data was acquired on an Illumina NovaSeq6000. Initial cell calling, quality filtering, alignment, and annotation were performed using the BD Rhapsody^™ WTA Analysis Pipeline on the Seven Bridges Genomics Platform. The data were then imported into R Studio version 4.2.1, and the Seurat package version 4.3.0 was used to perform further quality control (65). After removing multiplets and cells with mitochondrial reads > 25%, N = 10,871 cells were analyzed and assessed for differential gene expression. Seurat objects were converted and input into the Monocle3 package version 1.3.1 for graph-based trajectory inference (66–68). The Seurat package was also used to integrate data from all four participants for combined analysis. The ComplexHeatmap package was used to create a heatmap for each analyzed participant of the top markers differentially expressed between NK cell subsets with unsupervised clustering of genes and scaling by rows (69). The EnhancedVolcano and SCpubr packages were used to generate volcano plots and uniform manifold approximation projection (UMAP) plots, respectively (70, 71).

Statistics

Data were analyzed using GraphPad Prism version 8.0.2. Normality was determined by the Shapiro-Wilk test, and data not following a normal distribution were analyzed by the non-parametric Wilcoxon matched-pairs signed rank test. The paired Student’s t-test was used to determine statistical significance between two related groups (e.g., TP1 vs TP2; unstimulated vs IL-18+IL-12) and for direct comparisons of the FcRγ⁻ and FcRγ⁺ NK cell subsets in PWH with an FcRγ⁻ NK cell frequency greater than 10%. The unpaired Student’s t-test was used to compare the means of two independent or unrelated groups. For analyses of repeated measures in which values were missing, data were analyzed by fitting a mixed model (i.e., mixed effects analysis), with correction for multiple comparisons by Tukey’s HSD post-hoc test. The linear relationship between FcRγ⁻ NK cell frequency and IFNγ expression was determined by Pearson correlation analysis. Rank-based Spearman correlations examined associations of flow cytometry–based outcome measures with soluble markers of inflammation and antigen-specific adaptive immune responses. Measures below the assay limit were analyzed as the lowest rank. Detailed methodologies of the outcome measures related to the ACTG A5321 cohort were previously described (58, 59, 72, 73). For the single cell multiomic analysis, differential gene expression was determined using the Wilcoxon rank sum test with Bonferroni correction via Seurat:FindMarkers.

Study approval

All participants provided written informed consent prior to inclusion in this study, which was approved by ethics committees at each participating ACTG site and by the Institutional Review Board at the University of Pittsburgh.

Data Availability

Study data are available upon request from the Statistical and Data Management Center (SDAC) of the ACTG; the request may be made by emailing sdac.data@sdac.harvard.edu.

Results

Time on ART is associated with increasingly differentiated NK cells

Based on the phenotypic and functional abnormalities observed in our cross-sectional study of chronic HIV-1 infection (36), we expanded this work to a longitudinal setting in collaboration with the ACTG, detailing the impact of chronic HIV-1 infection and long-term ART on the phenotype and functional status of NK cell subsets. We began by drawing comparisons between total NK cells derived from age-matched people without HIV (PWOH) and virally suppressed PWH of the MWCCS, who self-identify as MSM (Table 1). In accordance with total NK cells from PWH moving toward a more mature phenotype, HIV-1 infection was associated with a decrease in surface expression of the NCR NKp46 (Fig. 1A), as well as a higher frequency of NK cells expressing PD-1 and NKG2C (Fig. 1B, 1C). Confirming a prior report of restoration of NK cell expression of NKG2A in the absence of HIV-1 viremia (52), PWOH and virally suppressed PWH had similar proportions of NKG2A⁺ NK cells (Fig. 1D). The modest reduction in the NKG2A to NKG2C ratio observed in PWH was, therefore, a consequence of elevated NKG2C expression (Fig. 1C–1E). Levels of CD57 and KLRG1, both of which signal a highly differentiated and senescent state (74–80), were comparable between PWOH and PWH (Fig. 1F, 1G). As a novel finding, surface expression of KLRG1 was reduced on NK cells after a 24h culture in the presence of IL-18 and IL-12—the extent of which was comparable between age-matched PWOH and PWH (Fig. 1H). These observations serve as a reference point for how treated chronic HIV-1 infection impacts the phenotype of total NK cells since we did not have access to pre-ART samples from the ACTG cohort and corroborate, to an extent, a recovery of NK cell phenotype and function with effective ART, as reported previously (42, 44, 45, 52, 81). However, differences in the general phenotype of total NK cells between PWOH and PWH were still detected despite a median duration of ART-mediated viral suppression of 12 years.

Figure 1. HIV-1 infection skews toward a differentiated NK cell phenotype.

Flow cytometry analyses of enriched NK cells from age-matched PWOH and PWH of the MWCCS. Gating was performed on live cells by forward and side scatter areas and single cells. NK cells were selected from live cells based on a CD3^–CD56⁺ gate. (A-G) This cross-sectional analysis highlights the phenotypic and functional impact of HIV-1 infection on total NK cells. In particular, NK cells from PWH demonstrated decreased levels of NKp46 (A), yet they were more likely to express PD-1 (B) and NKG2C (C), together characteristic of an increasingly differentiated population of NK cells. (H) Comparisons of KLRG1 surface expression on NK cells, at baseline (—) and following a 24h exposure to IL-18 and IL-12 (18/12), showed that NK cells from both PWOH and PWH downmodulated expression of KLRG1 in response to cytokine activation. Statistical significance was determined by the unpaired (A-G) or paired (H) Student’s t-test (****p < 0.0001; ***p < 0.001; **p < 0.01; *p < 0.05).

Next, PBMC samples were selected from PWH of the ACTG cohort study A5321 at three timepoints: ~4 weeks on-ART (TP0, n = 20); 1 year on-ART (TP1, n = 60); and 4 years on-ART (TP2, n = 60) (Table 2). Selection criteria of these ACTG study participants were based on clinical sample availability and having cell‐associated HIV‐1 DNA data at both year 1 and year 4 of ART (58), as well as plasma HIV-1 RNA data at year 4 by single-copy assay (59). An evaluation of the phenotypic and functional changes occurring longitudinally in total NK cells over the course of treatment with ART revealed that NK cells gained expression of CD57 between TP1 and TP2 (Fig. 2A), as was similarly noted by Ahmad et al. (81). The frequency of KLRG1⁺ NK cells also increased with each subsequent timepoint (Fig. 2B). With the acquisition of these markers was a reduction in the expression of NKG2A (Fig. 2C), an inhibitory receptor of immature NK cells (17), and the relative intensity of NKp46 (Fig. 2D), consistent with an increasingly differentiated phenotype. Furthermore, increasing time on ART did not alter IFNγ responses, as the expression of IFNγ by total NK cells following co-stimulation with IL-18 and IL-12 was similar across all three timepoints (Fig. 2E). Although ART is known to promote a reversal of NK cell phenotypic and functional abnormalities (42, 44, 45, 52, 81), our findings from this longitudinal analysis suggest that NK cells in treated HIV-1 infection continue to progress along a path of differentiation.

Figure 2. NK cells become more differentiated over the course of treatment with ART.

Evaluation by flow cytometry of PBMCs from PWH of the ACTG A5321 study at ~4 weeks on-ART (TP0, n = 20), 1 year on-ART (TP1, n = 60), and 4 years on-ART (TP2, n = 60). Gating was performed on live cells by forward and side scatter areas and single cells. NK cells were selected from live cells based on a CD3^–CD56⁺ gate. Shown is a comparison between TP1 and TP2 for all participants (left), the mean of differences between TP2 and TP1, with a value greater than 0 indicating an increase in expression from TP1 to TP2 (middle), and the subanalysis performed on those with PBMCs available from all three timepoints (n = 20) (right). Connecting lines indicate matched data from individual patients across timepoints. (A, B) NK cells gained expression of CD57 and KLRG1, both of which are markers of terminal differentiation, with the highest levels observed at TP2. (C, D) Concomitant with the acquisition of this differentiated phenotype was a reduction in the expression of NKG2A and NKp46. (E) Total NK cells responded to IL-18+IL-12 by producing IFNγ. However, they did not gain responsiveness over time. Statistical significance was determined by the paired Student’s t-test (left) or mixed-effects analysis, with correction for multiple comparisons by Tukey’s HSD post-hoc test (right) (****p < 0.0001; **p < 0.01; *p < 0.05).

Persistence of a skewed NK cell repertoire in treated HIV-1 infection

Previous studies have described a relatively rare population of memory-like NK cells, characterized by lack of expression of the intracellular signaling adapter protein FcRγ (FcRγ⁻), that is selectively expanded in the peripheral blood of individuals with chronic HIV-1 infection (29, 36–38). By contrast, the FcRγ⁻ population comprises less than 3% of NK cells in approximately two-thirds of PWOH (24, 36). These cells are specialized for antibody-induced responses but lack the flexibility to respond to innate cytokine signaling (36). To determine if the differential expression signatures between FcRγ⁻ and FcRγ⁺ NK cells remained consistent in a longitudinal setting of treated HIV-1 infection, we performed a detailed characterization of the two subsets in PWH of the ACTG A5321 cohort. First and foremost, four years of ART did not affect the frequency of FcRγ⁻ NK cells (Fig. 3A). Confirming the findings from our previous cross-sectional study (36), the FcRγ⁻ population was enriched for cells with higher expression of CD57 (not shown), NKG2C (Fig. 3B), and PD-1 (Fig. 3C), but lower levels of NKG2A (Fig. 3D) and NKp46 (not shown). On the other hand, the frequency of KLRG1-expressing NK cells was comparable between the two subsets (Fig. 3E). Our current longitudinal analysis builds upon this framework by demonstrating that the phenotypic differences between FcRγ⁻ and FcRγ⁺ NK cells did not diminish with increasing time on ART (Fig. 3F–3I). For example, both populations gained expression of KLRG1 (Fig. 3I). This was accompanied by decreased PD-1 expression (Fig. 3G) and slightly lower levels of NKG2A at TP2, although not statistically significant (Fig. 3H). Therefore, long-term ART does not promote a decrease in the FcRγ⁻ population nor longitudinal alterations in their phenotypic profile relative to conventional NK cells.

Figure 3. The phenotypic differences between FcRγ⁻ and FcRγ⁺ NK cells do not diminish in response to ART.

Evaluation by flow cytometry of PBMCs from PWH of the ACTG A5321 study with an FcRγ⁻ NK cell frequency >10% (n = 44) at 1 year (TP1) and 4 years (TP2) post-ART initiation. Gating was performed on live cells by forward and side scatter areas and single cells. NK cells were selected from live cells based on a CD3^–CD56⁺ gate. (A) The proportion of FcRγ⁻ NK cells remained stable over the course of treatment with ART. (B-E) Direct comparisons were drawn between FcRγ⁻ and FcRγ⁺ NK cells within each timepoint. Connecting lines indicate matched data from individual patients across FcRγ⁻ and FcRγ⁺ NK cell subsets. The FcRγ⁻ population was enriched for cells with higher levels of NKG2C (B) and PD-1 (C) but lower levels of NKG2A (D). FcRγ⁻ and FcRγ⁺ NK cells expressed comparable levels of KLRG1 (E). (F-I) Comparisons were drawn between TP1 and TP2 within each subset. A mean of differences value greater than 0 indicated an increase in expression from TP1 to TP2. Both populations followed the same trends over time for each of the markers analyzed, e.g., levels of PD-1 decreased (G) and KLRG1 increased (I) from TP1 to TP2 in both FcRγ⁻ and FcRγ⁺ NK cells. Statistical significance was determined by the paired Student’s t-test (****p < 0.0001; ***p < 0.001; **p < 0.01; *p < 0.05).

As expected, FcRγ⁻ NK cells also demonstrated decreased responsiveness to IL-18 relative to conventional NK cells (Fig. 4). Although comparable frequencies of FcRγ⁻ and FcRγ⁺ NK cells expressed KLRG1 at both timepoints (Fig. 3E), conventional NK cells downregulated this marker upon stimulation with IL-18 and IL-12 (Fig. 4A, 4B), as was observed with total NK cells (Fig. 1H). By contrast, a key finding was that KLRG1 expression trended toward an increase in the FcRγ⁻ subset at TP1 (Fig. 4A) and TP2 (Fig. 4B). Additionally, the differentiation of CD56^dim NK cells into a helper phenotype is characterized by their dramatic downmodulation of CD16 expression (36); however, at both on-ART timepoints, cytokine co-stimulated FcRγ⁻ NK cells maintained CD16 expression relative to unstimulated controls, which was in stark contrast to the FcRγ⁺ subset (Fig. 4C). FcRγ⁻ NK cells also failed to produce IFNγ in response to innate cytokine stimulation (Fig. 4D). Similar to what was noted in total NK cells (Fig. 2E), neither subset was marked by an increased frequency of IFNγ-expressing cells at TP2 (Fig. 4E), and the proportion of FcRγ⁻ NK cells inversely correlated with the frequency of NK cells positive for IFNγ by intracellular staining (Fig. 4F). By contrast, the FcRγ⁻ population demonstrated superior polyfunctional responses to antibody-mediated signaling via CD16 relative to conventional NK cells (Fig. 4G). These data confirm that FcRγ⁻ NK cells specialize in antibody-dependent reactivity but are minimally responsive to IL-18. Furthermore, this decreased responsiveness persists despite effective treatment with suppressive ART.

Figure 4. Persistence of reduced IL-18 responsiveness in the FcRγ⁻ population.

Evaluation by flow cytometry of PBMCs from PWH of the ACTG A5321 study with an FcRγ⁻ NK cell frequency >10% (n = 44) at 1 year (TP1) and 4 years (TP2) post-ART initiation. Cultures were exposed to media alone or IL-18+IL-12 for 24h. Gating was performed on live cells by forward and side scatter areas and single cells. NK cells were selected from live cells based on a CD3^–CD56⁺ gate. (A, B) Comparisons of KLRG1 surface levels on NK cells, at baseline (—) and following a 24h exposure to IL-18+IL-12 (18/12). Although conventional NK cells downregulated KLRG1 following IL-18+IL-12 stimulation, its expression increased modestly within the FcRγ⁻ subset at TP1 (A) and TP2 (B). (C) Unlike the FcRγ⁺ subset, cytokine co-stimulated FcRγ⁻ NK cells maintained CD16 expression relative to unstimulated controls at both timepoints. The graph depicts relative CD16 expression, in which the CD16 MFI of cytokine-treated cells was divided by the CD16 MFI of unstimulated cells. (D) The lack of responsiveness of the FcRγ⁻ population to innate stimuli was magnified by their dismal production of IFNγ following a 24h culture with IL-18+IL12. (E) Expression of IFNγ was similar between TP1 and TP2 in both FcRγ⁻ and FcRγ⁺ NK cells. (F) The proportion of FcRγ⁻ NK cells inversely correlated with the frequency of NK cells positive for IFNγ by intracellular staining. (G) The FcRγ⁻ population demonstrated superior polyfunctional responses to antibody-mediated signaling via CD16 compared to conventional NK cells (n = 6 PWH). Connecting lines indicate matched data from individual patients across FcRγ⁻ and FcRγ⁺ NK cell subsets. Statistical significance was determined using the paired Student’s t-test (A-E, G) or Pearson correlation (F) (****p < 0.0001; ***p < 0.001; **p < 0.01; *p < 0.05).

Contribution of inflammation and immune dysfunction to FcRγ⁻ NK cell expansion

Although HCMV infection has been implicated in the induction of FcRγ⁻ NK cells, the mechanisms contributing to their inflated expansion in HIV-1 infection are unclear. We hypothesized that more frequent HCMV reactivations in PWH would promote an increase in HCMV antibody titers, in turn selectively driving expansions of FcRγ⁻ NK cells, and together supporting a sustained inflammatory state characterized by an increasing level of immune dysfunction. Indeed, active HCMV viremia was detected in 35% and 30% of PWH at years 1 and 4 of ART (Table 4). Elevated levels of immune activation markers sCD163 and sCD14 (Fig. S1A, S1B) and inflammation markers IL-6 and hsCRP (Fig. S1C, S1D) also persisted despite consistent ART-mediated viral suppression. By contrast, active viremia of Epstein-Barr virus (EBV), another common herpesvirus, was only detected in 7% and 5% of PWH at years 1 and 4 of ART (Table 4). The importance of HCMV in the context of chronic HIV-1 infection was further underscored by the magnitude of HCMV-specific immunity, with HCMV antibody titers increasing between TP1 and TP2 (Fig. S1E), and T cell responses targeting one HCMV protein exceeding total T cell responses to the entire HIV-1 proteome, as previously reported within this cohort of PWH (72).

Table 4.

HCMV and EBV DNA viremia

		TP1 (1y, n = 60)	TP2 (4y, n = 60)
HCMV DNA	Detectable	35% (n = 21)*	30% (n = 18)
copies/ml	Median (Q1, Q3)#	15.2 (11.9, 26.0)	15.4 (7.8, 26.7)
EBV DNA	Detectable	7% (n = 4)	5% (n = 3)
copies/ml	Median (Q1, Q3)#	283.1 (179.6, 385.1)	366.3 (148.8, 27,896.2)

^*Of the 21 participants with detectable HCMV DNA at TP1, 38% (n = 8) were also positive at TP2

^#Median, Q1, and Q3 reflect only participants with detectable viremia

We next tested for relationships between proportions of FcRγ⁻ NK cells and antigen-specific adaptive immune responses. Modest positive correlations were observed (albeit not significant) with HCMV-specific IgG antibody titers (Fig. S1F) and T cell responses (Fig. S1G), whereas no relationship was identifiable between FcRγ⁻ NK cell frequency and immune responses to EBV (Fig. S1H).

Abnormal NK cell subset distribution due to interrupted NK-poiesis

As described previously (12–16), NK cells are classically divided into two major populations, CD56^brightCD16⁻ and CD56^dimCD16⁺, with the former typically accounting for less than 10% of NK cells in the periphery (Fig. 5A (PWOH), 5B). Interestingly, we noted extremely pronounced CD56^bright populations in PWH, comprising upwards of 40 to 50% of total circulating NK cells. This abnormal accumulation of the CD56^bright subset was particularly apparent at TP1. Although the relative proportions of CD56^bright NK cells began to normalize by TP2, they continued to account for a considerable fraction of NK cells in the peripheral blood relative to PWOH (Fig. 5A– 5C). Furthermore, we observed unusually large CD56^brightCD16⁺ and CD56^dimCD16⁻ subsets in 24 of the 60 PWH of the ACTG A5321 study (Fig. 5C–5E). Compared to PWOH, the relative proportion of the CD56^dimCD16⁺ NK cell subset was significantly lower in PWH as a result of elevated frequencies of the CD56^brightCD16^+/– and CD56^dimCD16⁻ populations (Fig. 5D). Again, the most compelling perturbations in NK cell subset distribution were observed at TP1. By TP2, the relative frequency of the CD56^dimCD16⁺ population experienced a resurgence, which was accompanied by a decline in the relative proportions of the remaining three subsets (Fig. 5E), suggesting a positive effect of ART on the normalization of NK cell subset distribution. These observations led us to hypothesize that the abnormal NK cell subset distribution observed in HIV-1 may arise from sustained inflammation and that direct effects exerted by the virus during initial infection contribute to interrupted NK-poiesis, arresting NK cells in transitional states and resulting in the accumulation of unconventional subsets.

Figure 5. An abnormal NK cell subset distribution in HIV-1 infection.

(A) The CD56^bright population comprised upwards of 50% of total circulating NK cells, whereas they normally account for <10% of NK cells in the periphery, as exemplified by the PWOH pie graph on the left. (B, C) Representative FACS plots, illustrating the expected distribution at baseline of CD56^dim and CD56^bright NK cells (B) and unusually large CD56^brightCD16⁺ and CD56^dimCD16⁻ populations at the three on-ART timepoints (C). (D, E) Relative to PWOH, PWH of the ACTG A5321 study experienced an aberrant NK cell subset distribution, which was particularly pronounced at TP1 and began to normalize by TP2; n = 20 PWOH, n = 20 PWH at TP0, n = 60 PWH at TP1 and TP2. Statistical significance was determined by the unpaired Student’s t-test (D) or mixed-effects analysis, with correction for multiple comparisons by Tukey’s HSD post-hoc test (E) (****p < 0.0001; ***p < 0.001; **p < 0.01; *p < 0.05). The quadrants are defined as follows: Q1, CD56^brightCD16⁻; Q2, CD56^brightCD16⁺; Q3, CD56^dimCD16⁺; Q4, CD56^dimCD16⁻.

To begin exploring this hypothesis, we compared the transcriptomes of these NK cell subsets using a multiomic single-cell analysis platform, which enabled in silico separation of NK cells based on surface CD56 and CD16 protein expression. Each subset had a distinct pattern of gene expression, with the CD56^brightCD16⁻ and CD56^brightCD16⁺ cells appearing most similar to each other (Fig. S2). Graph-based trajectory inference revealed steadily increasing pseudotime values from CD56^brightCD16⁻ through CD56^brightCD16⁺ to the CD56^dimCD16⁺ subset, consistent with sequential progression through these phenotypes (Fig. 6A, 6B). Importantly, the bimodal distribution seen in the CD56^dimCD16⁻ subset suggested that this population was a combination of both immature and highly differentiated cells (Fig. 6B), a notion also supported by CD56^dimCD16⁻ NK cells falling within distinct UMAP cluster regions that represent cells with either high or low IL7R transcript expression (Fig. S3A, S3B). Furthermore, the set of transcripts identified as changing significantly over pseudotime included multiple genes associated with stages of NK cell differentiation (Fig. 6C, S3C). For example, the CD56^brightCD16⁻ subset expressed the highest levels of genes indicative of a lesser differentiation status such as the transcription factor Tcf1 (TCF7) (82, 83), IL7RA (IL7R) (83, 84), and a subunit of the IL-12 receptor (IL12RB2) (31, 85, 86). Conversely, the CD56^dimCD16⁺ group expressed the highest levels of genes typical of the classical mature CD56^dim NK cell subset (Fig. 6C, S3C), including those associated with cytotoxicity (PRF1, GZMB, NKG7) (82, 84, 87), and peripheral homing (CX3CR1, CCL5) (17, 18, 84, 88, 89).

Figure 6. Multiomic analysis corroborates HIV-1 interruption of NK-poiesis.

scRNA-seq analysis was used for transcriptome comparison and graph-based trajectory inference across NK cell subsets at the on-ART timepoint of 4 years (TP2; n = 4 PWH). These samples were chosen in particular based on being representative of the pool of participants and together having the broad repertoire of NK cell subsets based on flow cytometry results. NK cells were grouped by surface CD56 and CD16 protein expression in silico via AbSeq staining. (A) Trajectory graph (black line) overlaid on a UMAP plot of all NK cells from a representative participant with an abnormal distribution. (B) Mean pseudotime values from the same participant, indicated by the vertical lines, increased from the CD56^brightCD16⁻ to CD56^dimCD16⁺ subsets. The CD56^dimCD16⁻ population had a bimodal distribution consisting of “young” and “old” cells. (C) Expression of genes associated with NK cell development changed significantly throughout the trajectory for all 4 participants and were further analyzed at the subset level. CD56^brightCD16⁻ cells had the highest expression of TCF7, which encodes for Tcf1—a transcription factor that guides NK cells through early stages of development. The CD56^dimCD16⁺ population had the highest expression of markers classically associated with this mature population, including CX3CR1, GZMB, and PRF1. Displayed genes are scaled, and all had adjusted p values < 0.0001 for comparisons across pseudotime. The quadrants are defined as follows: Q1, CD56^brightCD16⁻; Q2, CD56^brightCD16⁺; Q3, CD56^dimCD16⁺; Q4, CD56^dimCD16⁻.

We then analyzed the individual NK cell subsets for protein expression of markers related to differentiation status. The four distinct subsets appeared to cluster based on surface expression of CD56, with the CD56^brightCD16⁺ and the CD56^dimCD16⁻ populations being phenotypically more closely related to the classical CD56^bright and CD56^dim subsets, respectively (Fig. 7). With KLRG1, which is understood to be marker of senescence (78–80), we saw a progressive increase in the frequency of NK cells expressing this marker from the CD56^brightCD16⁻ through the CD56^dimCD16⁺ subsets (Fig. 7A, 7E). The same pattern was observed for CD57 (Fig. 7B, 7F), a marker of terminal differentiation (74, 75, 81), and NKG2C (Fig. 7C, 7G), an activating receptor enriched on highly differentiated memory-like NK cells (Fig. 3B). By contrast, the CD56^dim subsets had the lowest expression of NKG2A (Fig. 7D, 7H), an inhibitory receptor found on immature NK cells (17). These phenotypic profiles support the notion that NK cells progress through the stages of differentiation in a stepwise fashion and that HIV-1 infection interrupts NK-poiesis, resulting in abnormal collections of NK cells in intermediate states.

Figure 7. Protein expression of markers of differentiation further illustrates HIV-1 interruption of NK-poiesis.

(A-H) Evaluation by flow cytometry of PBMCs from PWH of the ACTG A5321 study with the noted aberrant NK cell subset distributions at ~4 weeks on-ART (TP0, n = 8), 1 year on-ART (TP1, n = 24), and 4 years on-ART (TP2, n = 24). Gating was performed on live cells by forward and side scatter areas and single cells. NK cells were selected from live cells based on a CD3^–CD56⁺ gate. While CD56^dim NK cells had the highest levels of KLRG1 (A, E), CD57 (B, F), and NKG2C (C, G), the CD56^bright populations had the highest NKG2A expression (D, H), illustrating the differences in their differentiation status. Statistical significance was determined by mixed-effects analysis, with correction for multiple comparisons by Tukey’s HSD post-hoc test. Comparisons were made between subsets within each TP (A-D) (****p < 0.0001; ***p < 0.001; **p < 0.01; *p < 0.05) or between TPs within each subset (E-H) (TP0 vs TP1 and TP1 vs TP2: ****p < 0.0001, ***p < 0.001, **p < 0.01, *p < 0.05; TP0 vs TP2: ^###p < 0.001, ^##p < 0.01, ^#p < 0.05).

Discussion

Our findings illustrate that chronic HIV-1 infection alters the composition of the NK cell compartment and repertoire, the impact of which remains evident despite four years of suppressive ART. First, higher proportions of NK cells from PWH express PD-1 and NKG2C while displaying reduced surface density of the NCR NKp46 (Fig. 1), suggestive of an increasingly mature phenotype. Throughout the course of treated chronic HIV-1 infection, NK cells continue to progress along a spectrum of differentiation, notably gaining expression of CD57 and KLRG1 (Fig. 2A, 2B). A previous study has reported that terminal differentiation of CD56^dimCD16⁺ NK cells is enhanced after ART, as measured by CD57 expression (81). Indeed, the frequency of CD57-expressing NK cells increases longitudinally in the context of ART-treated chronic HIV-1 infection (Fig. 2A). As we did not have access to pre-ART PBMCs, a weakness of our study is the inability to track the changes that occur in the NK cell receptor repertoire between acquisition of HIV-1 infection and initiation of ART. However, we do note comparable proportions of CD57⁺ NK cells between age-matched PWOH and PWH, who self-identify as MSM (Fig. 1F), suggesting that HIV-1 infection, together with ART, is not solely responsible for the acceleration of immune aging, but rather that these immunologic effects are secondary to—or heavily confounded by—HCMV, as MSM have a similar rate of HCMV seropositivity (36).

We also demonstrate that HIV-1 infection has a profound effect on NK cell subset distribution, with unusually large CD56^brightCD16⁺ and CD56^dimCD16⁻ subsets observed in more than one-third of PWH. These atypical NK cell compartments are most pronounced within the first year of ART (Fig. 5), reinforcing the notion that recovery of NK cell phenotype and function occurs only after 24 months of suppressive ART (42, 45, 52). The CD56^dimCD16⁻ population is a rare NK cell subset that typically accounts for a minute fraction of fresh PBMCs (<1%). The CD56^bright population likewise comprises less than 1% of fresh PBMCs, or less than 10% of NK cells in the periphery, and is predominantly negative for expression of CD16 (90). By contrast, we show that in some PWH, CD56^bright NK cells account for upwards of 50% of total circulating NK cells, with the relative frequency of the CD56^brightCD16⁺ population reaching a staggering 44% (Fig. 5). This phenomenon has also been reported in the setting of graph-versus-host disease following hematopoietic stem cell transplantation (HSCT), whereby the accumulating CD56^bright NK cells act in a regulatory fashion (83). It is conceivable that a similar temporal effect may be occurring with ART initiation, where these expanded NK cells function to regulate immune events that can otherwise lead to immune reconstitution inflammatory syndrome (91).

Takahashi et al. reported that cultures containing an inflammatory cytokine milieu induce the differentiation of CD56^brightCD16⁻ NK cells into the CD56^brightCD16⁺ population. Additionally, cytokine stimulation promotes the development of the rare CD56^dimCD16⁻ subset into CD56^brightCD16⁻ and further into CD56^brightCD16⁺ NK cells (90, 92). We also know that the classical CD56^dimCD16⁺ NK cells are capable of differentiating into CD16⁻ NK helper cells in response to IL-18 (5, 36). This effectively forms a circuitous clockwise differentiation loop, a concept supported by our sequencing data and trajectory analysis (Fig. 6, S2, S3), whereby CD56^brightCD16⁻ NK cells first gain expression of CD16. This CD56^brightCD16⁺ population then differentiates into classical CD56^dimCD16⁺ NK cells. Under the appropriate conditions, this mature population loses expression of CD16 and further develops into the CD56^brightCD16⁻ subset. Notably, the bimodal distribution of the mean pseudotime sequencing values noted within the CD56^dimCD16⁻ NK cell fraction suggests a collection of both immature and highly differentiated NK cell populations (92) (Fig. 6B). Given that these fluctuations in the NK cell compartment are driven by combinations of inflammatory cytokines, it is conceivable that the aberrant NK cell subset distribution observed in select PWH may arise from undue systemic inflammation. We also posit that HIV-1 infection interrupts NK-poiesis, arresting NK cells in transitional states and/or prematurely driving NK cells into the periphery from the bone marrow at different stages of development, thereby resulting in the accumulation of unconventional NK cell subsets, similar to what has been reported during lymphocyte reconstitution following HSCT (92). Together, our data suggest that over time ART partially reverses some HIV-1-induced NK cell abnormalities, but a skewed NK cell repertoire persists during long-term treatment of HIV-1 infection, with direct implications for NK cell functionality.

On a related note, while we found that surface expression of the inhibitory molecule KLRG1 on NK cells is dampened with combined exposure to IL-18 and IL-12, this novel finding does not apply to the FcRγ⁻ NK cell population, which continues to maintain high expression of KLRG1 (Fig. 4A, 4B). This suggests that metabolic dysfunction, induced by HIV-1 infection, contributes to their limited responsiveness to innate stimuli (93), as accumulating evidence links impaired cellular metabolism to NK cell dysfunction (94, 95). This notion is further supported by reports in which hepatitis C virus infection provokes NK cells to increase surface expression of KLRG1, which is associated with defective phosphorylation of Akt at Ser473 and impaired functionality (76). Moreover, engagement of KLRG1 appears to curb the effector functions of KLRG1^hi NK cells by enhancing pre-existing AMPK activity, thus trapping them in a quiescent state (96). Similarly, persistent antigen stimulation in chronic viral infections leads to an increase in KLRG1 expression on virus-specific CD8⁺ T cells, negatively influencing their function and proliferative capacity (78). Together, these studies provide the scientific premise for our hypothesis that the metabolism of FcRγ⁻ NK cells from PWH is negatively skewed, resulting in an exhausted phenotype marked by decreased functional plasticity. A paucity of information exists on the metabolic changes in human NK cells during viral infections, including the influence of cellular metabolism on their memory-like functions; however, targeting NK cell metabolism may be key to restoring proper NK cell function in HIV-1 infection.

The lack of any significant direct correlations between frequencies of FcRγ⁻ NK cells and HCMV-specific adaptive immune responses, despite their common trigger, likely reflects complex and competing influences from a multitude of factors, with cumulative effects throughout chronic HCMV/HIV-1 co-infection obscuring relevant impacts of direct interactions. Our data do suggest, though, that immunosenescence and inflammation contribute, at least in part, to inflated proportions of FcRγ⁻ NK cells in PWH. For one, long-term ART fails to reverse the inflammation induced by HIV-1 infection (Fig. S1). This is supported by previous studies, including one by Hearps et al., in which levels of sCD163 and CXCL10 are elevated in viremic HIV-1 infection and remain elevated in virologically suppressed PWH relative to age-matched PWOH, instead resembling those observed in elderly controls, suggesting an acceleration of immune aging (97). Additionally, the magnitude of HCMV-specific immunity (72) (Fig. S1E–S1G), combined with the rate of active HCMV viremia (Table 4), points to an inability of the immune response to keep pace with HCMV reactivations, further feeding into the cyclic loop of immune activation, senescence, and inflammation.

This chronic immune activation drives immunosenescence, which subsequently triggers additional inflammation and immune dysfunction (98). HIV-1 infection, in fact, is characterized by an accumulation of late-differentiated T cells with a senescent phenotype and heightened secretion of pro-inflammatory cytokines (99–101). In particular, HCMV-specific memory CD8⁺ T cells lacking CD28 are highly expanded in PWH (102, 103), and this is accompanied by elevated HCMV antibody titers (29, 37, 104), with serum HCMV IgG levels correlating with inflammatory markers (105–107). Furthermore, an inverse relationship exists in HIV-1 infection between the extent of NKG2C⁺CD57⁺ NK cell expansion and the fraction of HCMV-specific CD8⁺ T cells expressing CD28. Parallel expansions of highly differentiated HCMV-associated adaptive and/or memory-like NK cell populations with limited functional plasticity and progressively senescent CD28⁻ HCMV-specific T cells imply increased HCMV reactivation to immunogenic levels, reflecting a level of underlying immune dysfunction in HIV-1 infection (35, 102).

It is plausible that FcRγ⁻ memory-like NK cell differentiation and persistence occur as compensatory mechanisms for poor protective T cell responses in PWH (37), but their expansion to high levels indicates stress on the overall ability of the immune system to control HCMV reactivations (102). Similarly, the rapid and premature delivery of new NK cell recruits from the bone marrow to counteract this added immunologic burden may, in part, explain the appearance of NK cells with immature or intermediate phenotypic signatures. As immune cells do not act independently of one another, but rather in a coordinated fashion, dysfunction in one cellular compartment likely skews the delicate balance of the entire immune system, resulting in additional but interrelated layers of dysfunction. Ultimately, we posit that NK cell irregularities are both a cause and contributor of a chronic inflammatory state that promotes diseases of aging, including malignancies, cardiovascular disease, and neurocognitive disorders, in PWH despite highly effective suppressive ART. It will be critical to identify the factors contributing to the lack of complete immune restoration in PWH while on effective ART in order to provide the best opportunities for success in future curative treatment approaches.

Key Points.

• NK cells progress along a spectrum of differentiation in PWH on long-term ART

• Conventional but not FcRγ^– NK cells downregulate KLRG1 upon IL-18+IL-12 exposure

• PWH display atypical NK cell subsets, representing intermediate stages of NK-poiesis