Cohort characteristics. PBMCs from 34 ECs were obtained from the Spanish AIDS Research Network (RIS) cohort of HIV Controllers Study Group (ECRIS). Within this cohort, 21 participants demonstrated DC over HIV infection, maintaining viral loads below 40 HIV-1 RNA copies/mL and stable CD4+ T cell counts for a period ranging from 5 to 20 years (median 7.25 years). In contrast, 13 individuals experienced a loss of immunological control (AC), as evidenced by a progressive and statistically significant decline in CD4+ T cell count over the follow-up period (P < 0.05, determined by simple linear regression). The longitudinal profiles of CD4+ T cell counts and HIV-1 RNA levels for EC participants are depicted in Supplemental Figure 1; supplemental material available online with this article; https://doi.org/10.1172/jci.insight.182660DS1 Two additional cohorts were included in the study: samples from 2 to 4 years of ART (n = 24), and samples from VIR participants (n = 18). Samples from HD (n = 24) served as controls. The clinical characteristics of HIV+ individuals included in the study are summarized in Supplemental Table 1.

No significant differences in sex distribution were observed among the study groups (Supplemental Table 1). However, the median ages of DC (55 years) and AC (58 years) individuals were higher than that of ART (44 years) and VIR (39 years) groups (P < 0.05, Kruskal-Wallis test followed by Dunn’s multiple-comparisons test). These findings are consistent with the increased time since HIV diagnosis observed in the DC and AC groups compared with VIR (PDC = 0.07, PAC < 0.001, Kruskal-Wallis test followed by Dunn’s multiple-comparisons test). Furthermore, significant differences in transmission routes were observed among all groups (P < 0.0001, Fisher’s exact test). CD4+ T cell counts were also higher in DC, AC, and ART groups compared with VIR (P < 0.05, Kruskal-Wallis test followed by Dunn’s multiple-comparisons test). While DC, AC, and ART groups exhibited effective viral suppression (HIV-1 RNA < 40 copies/mL), VIR participants demonstrated viral loads ranging between 60,000 and 800,000 copies/mL (P < 0.0001, Kruskal-Wallis test followed by Dunn’s multiple-comparisons test). Despite having HLA typing data available for only 22 out of 34 participants within the EC group, protective HLA-B alleles (HLA-B57, B52, B27, and B14) (35) were found at comparable frequencies in both AC and DC participants (Supplemental Table 1). These findings suggest similarities in clinical, genetic, and virologic parameters within HIV+ individuals with spontaneous HIV control, while confirming the anticipated differences between these groups and viremic individuals.

NK cell phenotypes with memory-like attributes distinguish individuals with DC. We first studied the overall frequencies of NK cells, the prevalence of the primary NK cell subsets based on CD56 and CD16 expression, and the percentage of memory-like NK cells characterized by the presence of NKG2C and CD57 markers across all study groups (36–38). An example of the gating strategy and a representative flow cytometry plot delineating the distinct NK cell subsets studied are shown in Supplemental Figure 2 and Figure 1A, respectively. All groups exhibited comparable frequencies of total and CD56bright NK cells (Figure 1B). However, higher proportion of CD56dimCD16hi NK cells was observed in ART individuals (Figure 1B). Interestingly, all PWH exhibited higher frequencies of memory-like NKG2C+CD57+ NK cells, although significance was only reached for DC, ART, and VIR, compared with HD (Figure 1C). These cells are more likely to represent expanded adaptive NK cells, as previously reported in CMV (39), and lately, in HIV infection (40).

Phenotypic characterization of NK cells in ECs. The expression of different NK markers was quantified by flow cytometry in different study groups: healthy donors (HD, n = 25), ECs with durable HIV control (DC, n = 21), ECs with aborted immunological control (AC, n = 13), PWH ART-treated individuals (ART, n = 24), and viremic PWH (VIR, n = 18). (A) Representative flow cytometry plots depicting the NK cell subset gating strategy from CD3– cells based on CD56 and CD16 expression (left: CD56+ total, CD56dimCD16hi, and CD56bright) and NKG2C and CD57 expression (right: NKG2C+CD57+ memory-like NK cells). (B) Violin plots depicting the frequency of different NK cell populations identified (left to right: CD56+ total, CD56dimCD16high, and CD56bright). (C) Violin plots depicting the frequency of memory-like NKG2C+CD57+ NK cells. (D) Violin plots depicting the frequency of different NK cell markers in CD56+ total NK cells by study group (left to right: CD158b, CXCR3, KLRG1, NKG2A, NKG2D, NKp30, NKG2C, and CD57). (E) Violin plots depicting the frequency of distinct NK cell receptors in expanded memory-like NKG2C+CD57+ NK cells (frequency >5% and counts >25; left to right: CD158b, CXCR3, KLRG1, NKG2A, NKG2D, and NKp30). Median with range is represented. Statistical comparisons were performed using Kruskal-Wallis 1-way ANOVA followed by Dunn’s multiple-comparison test. *P < 0.05; **P < 0<01; ***P < 0.001; ****P < 0.0001.

Next, we studied whether NK cells from DC presented a different expression pattern of important receptors for NK activity. For these analyses, we included NK receptors with activating or inhibitory potential upon interaction with ligands found on HIV-infected cells (KIR2DL2/L3 [CD158b], KLRG1, NKG2A, NKG2D, NKp30, and NKG2C), the chemokine receptor CXCR3, and the maturation marker CD57. Expression of selected markers in HD were concordant with previous studies (41). We observed higher expression of the negative receptors CD158b and KLRG1 in ART and VIR groups relative to HD, DC, and AC in total CD56+ and CD56dimCD16hi NK cells (Figure 1D and Supplemental Figure 3A). By contrast, higher frequencies of NKG2A+ and NKp30+ NK cells were observed in HD and AC, in both total CD56+ and CD56dimCD16hi NK cell subsets, compared with ART and VIR groups (Figure 1D and Supplemental Figure 3A). While NK cells from DC generally presented similar receptor expression patterns relative to those from HD and AC groups, DC displayed higher expression of the activating receptor NKG2C in all 3 NK cell subsets (Figure 1D and Supplemental Figure 3, A and B). No remarkable differences were observed in the expression of CXCR3, NKG2D, and CD57 receptors among DC and other groups in both total CD56+ and CD56dimCD16hi NK cells (Figure 1D and Supplemental Figure 3A). By contrast, increased CXCR3 expression was observed in ART compared with DC and VIR groups in CD56bright NK cells (Supplemental Figure 3B).

Next, we focused on the memory-like NKG2C+CD57+ NK cell population. For consistency, only samples with expanded NKG2C+CD57+ NK cells (>5% frequency) were analyzed. Comparative analysis revealed no statistical differences in the expression levels of CD158b, CXCR3, NKG2D, and NKp30 across the distinct study cohorts (Figure 1E). Consistent with the other NK cell subsets, a trend toward elevated frequencies of KLRG1+ memory-like NK cells was observed in ART and VIR groups compared with HD, DC, and AC (Figure 1E). Furthermore, within the DC group, heightened frequencies of NKG2A+ memory-like NK cells were found compared with ART (Figure 1E).

Overall, our study reveals 2 important aspects. The first is that distinct phenotypic variations are observed within NK cell populations among different PWH groups, underscoring the heterogeneity of the NK cell repertoire during HIV infection. Second, the expression of the different NK cell receptors within the distinct NK cell subsets observed in DC is similar to those observed in AC, and in HD, except for the memory-like NKG2C+CD57+ subset of cells, which was enriched in NK cells from the DC cohort. Moreover, within the memory-like NK cell subset, the expression of NKG2A was elevated compared with other groups, indicating their potential to regulate their functional responses (42). These identified markers of NK cell populations represent a distinctive hallmark of ECs who spontaneously control HIV for prolonged periods without virological and immunological progression.

NK cells from individuals with DC exhibit reduced natural cytotoxicity, but enhanced ADCC activity against HIV-infected cells. Next, we aimed to directly elucidate the functional signatures of NK cells in all study groups. We focused on the functional response of DC participants, as they represent a model for a functional cure for HIV. We performed NK cytotoxicity assays against the MHC-devoid K562 cell line, using NK cells that were either unstimulated or primed with IL-15 (gating strategy shown in Supplemental Figure 4A). NK cell degranulation (CD107a) and IFN-γ secretion were measured upon stimulation (Supplemental Figure 4B). Although all study groups showed significant IFN-γ production and degranulation activity against target cells upon stimulation, reduced degranulation responses were observed in total CD56+ NK cells from DC compared with HD, ART, and VIR groups (Figure 2, A–C). Similarly, increased IFN-γ production and degranulation were observed in CD56dimCD16hi and CD56bright NK cell subsets upon stimulation in all study groups (Supplemental Figure 5, A and B). Notably, the expanded memory-like NK cell population (>5% frequency) showed, in general, higher ability to produce IFN-γ and CD107 upon stimulation, compared with total CD56+ NK cells (Figure 2, A–F). A tendency toward decreased degranulation responses was observed within the memory-like NK cell population in DC, exhibiting diminished IFN-γ production, degranulation, and polyfunctional responses (IFN-γ and CD107a expression) compared with HD (Figure 2, D–F).

Functional profile of NK cells in ECs. NK cell activation and cytotoxicity subsequent to stimulation were evaluated by study group. The percentages of (A) IFN-γ+, (B) CD107a+, and (C) polyfunctional IFN-γ+CD107a+ within the CD56+ NK cell population were determined in basal conditions, following coculture with K562 target cells, and with additional IL-15 stimulation. Similarly, these metrics were quantified in expanded memory-like NKG2C+CD57+ NK cells (frequency >5%): the percentages of (D) IFN-γ+, (E), CD107a+, and (F) polyfunctional IFN-γ+CD107a+ NK cells after stimulation were evaluated. (G) Violin plots depicting the natural cytotoxicity exhibited by CD56+ total NK cells from the different study groups following coculture with K562 cells. (H) Spearman’s correlations between natural cytotoxic responses and the frequency of distinct NK cell subsets (left to right: CD56dimCD16hi NK cells, CD56bright NK cells, and NKG2C+CD57+ NK cells). (I) Violin plots showing the ADCC activity mediated by CD56+ total NK cells against HIV-expressing cells by study group. (J) Spearman’s correlations between ADCC responses and the frequency of different NK cell populations (left to right: CD56dimCD16hi NK cells, CD56bright NK cells, and NKG2C+CD57+ NK cells). For violin plots, median with range is represented. *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001 by repeated measures 2-way ANOVA followed by Tukey’s multiple-comparison test (A–F) or Kruskal-Wallis test (G and I).

Subsequently, we evaluated the intrinsic natural cytotoxicity and ADCC responses of NK cells. Natural cytotoxicity was assessed by quantifying their killing capacity against MHC-devoid K562 cell targets. Cell killing was measured by a viability dye–based flow cytometric analysis (Supplemental Figure 6A). ADCC activity of NK cells toward HIV-infected cells was evaluated by coculturing ACH-2 cells with isolated NK cells from the different study groups in the presence of plasma from a viremic HIV+ participant containing a pool of antibodies targeting the gp120 protein. Cell killing was determined by the lack of detection of the viral protein p24 in infected cells by flow cytometry (Supplemental Figure 6B).

Notably, NK cells from DC individuals exhibited a significantly reduced cytotoxic response toward K562 target cells (median killing = 4.46%), in contrast with HD (median killing = 15.10%) and ART groups (median killing = 11.30%) (Figure 2G). Specifically, NK cells from DC demonstrated an approximate reduction of 30% in their ability to eliminate target cells compared with HD (Figure 2G). Importantly, and consistent with prior literature (43–45), our findings reveal a positive association between the frequency of CD56dimCD16hi NK cells and natural cytotoxic responses, while a negative correlation is observed with CD56bright frequencies (Figure 2H). Of note, no significant associations were found between natural cytotoxic responses and the frequency of memory-like NK cells (Figure 2H), nor between cytotoxic responses and the expression levels of NKG2D (Supplemental Figure 7A), a receptor for which K562 cells express high levels of ligand.

We next evaluated the adaptive ADCC activity of NK cells toward HIV-infected cells. NK cells from DC performed strong ADCC responses (median ADCC = 19.09%), with an overall capacity similar to the one observed for HD (median = 29.63%), for which higher variation was observed (Figure 2I). Defective ADCC was observed for AC and VIR NK cells (medianAC = 7.58%, medianVIR = 1.18%), which showed significantly reduced activity compared with HD and DC (Figure 2I). Similarly, ADCC from NK cells derived from the ART group was reduced relative to HD (medianART = 12.82%) (Figure 2I). Importantly, the enhanced capacity of NK cells from DC to kill via the ADCC mechanism was not related to the direct abundance of CD56dimCD16hi NK cells, the subset primarily implicated in ADCC responses, nor to the frequencies of memory NK cells (Figure 2J). Furthermore, since CD16 expression levels on the surface of NK cells might directly affect ADCC activity, we evaluated differences between groups and found that DC had similar CD16 expression levels compared to other cohorts, and this expression did not correlate with ADCC (Supplemental Figure 7B).

Overall, all study groups exhibited elevated frequencies of activated NK cells, marked by the secretion of soluble effector molecules, such as IFN-γ and CD107a, upon stimulation. However, DC individuals exhibited a tendency toward a limited production of these molecules compared with other groups, particularly within the memory-like NK cell subset; however, these differences did not reach statistical significance. These functional changes in NK cells from DC translated into lower natural cytotoxicity. In contrast, DC NK cells demonstrated a potent ADCC response against HIV-infected cells, which was not related to the presence of any major NK subset, suggesting a distinctive feature of NK cells from DC individuals.

A distinctive memory-like NK population in DC is associated with enhanced ADCC responses against HIV-infected cells. Next, to uncover novel NK cell subsets linked with the functional responses observed in DC individuals, we explored the landscape of the different NK cell populations by reduction of dimensionality analysis. Unsupervised clustering analysis via the FlowSOM algorithm allowed us to identify 11 different NK cell clusters with an individual pattern of expression of the selected receptors (Figure 3A). Importantly, we identified C2 and C4 as clusters of interest, as they were enriched in individuals with DC (Figure 3B). These clusters comprised cells expressing CD16, an important receptor to mediate ADCC, together with CXCR3, NKG2A, and CD57 (Figure 3C). Cluster C2 exhibited high expression of NKG2D, an activating receptor associated with ADCC responses, whereas cluster C4 demonstrated low expression of this marker (Figure 3C). By contrast, C4 additionally exhibited CD158b and NKG2C expression, representing a unique memory-like NK cell phenotype (Figure 3C and Supplemental Figure 8A).

Association between NK cell phenotype and functional responses in EC. (A) Optimized t-distributed stochastic neighbor embedding (opt-SNE) representation of distinct NK cell clusters, identified through dimensionality reduction based on the expression of the array of distinct NK cell receptors (CD16, CD158b, CXCR3, KLRG1, NKG2A, NKp30, NKG2D, NKG2C, and CD57), by study group (left to right: HD, DC, AC, ART, and VIR). (B) Violin plots showing the frequency of each NK cell cluster in CD56+ total NK cells by study group. (C) Heatmap depicting the normalized median expression of the selected phenotypic NK cell markers within the cell clusters identified in Figure 3A. (D) Correlation matrix depicting Spearman’s correlations between the frequency of NK cell clusters identified by dimensionality reduction based on the expression of phenotypic markers and functional responses. (E) Spearman’s correlations between the frequency of clusters C6 and C7 and natural cytotoxic responses in all study groups. (F and G) Spearman’s correlations between the frequency of C4 and ADCC responses in (F) all study groups and (G) within the DC group. Graphs represent medians and ranges. Each dot represents 1 individual of a specific cohorts, indicated by color code (HD in green, DC in blue, AC in red, ART in orange, VIR in purple). Statistical comparisons were performed using the Kruskal-Wallis test. *P < 0.05; **P < 0.01.

Subsequently, we examined the relationship between functional NK cell responses and the frequency of the distinct NK cell clusters (Figure 3D). A significant negative correlation was observed between cluster C6 and natural cytotoxic responses (r = –0.33, P = 0.02), while a positive correlation was noted between cluster C7 and natural cytotoxicity (r = 0.30, P = 0.03) (Figure 3, D and E). Notably, there were no significant differences in the frequency of these clusters among study groups (Figure 3B). Both clusters prominently expressed CD158b and NKp30, while NKG2C and CD57 markers were either low or absent (Figure 3C). However, cluster C7, which associated with higher natural cytotoxicity, displayed elevated levels of CD16 and KLRG1, distinguishing it from cluster C6, which was defined by the coexpression of CXCR3 and NKG2A (Figure 3C). In addition, cluster C2 exhibited a trend for a positive correlation with ADCC responses, although no statistical significance was achieved (Figure 3D). Importantly, cluster C4, predominantly composed of memory-like NK cells expressing NKG2A and CXCR3, exhibited a robust association with ADCC responses (Figure 3, D and F). Remarkably, this correlation was particularly strong within the DC group (P = 0.0082, r = 0.68) (Figure 3G). Importantly, although both DC and AC showed similar proportions of C4, we demonstrated that only DC had an enhanced ability to mediate ADCC (Figure 2I). We hypothesize that this could be explained, at least in part, by the significantly higher expression of NKG2C observed in C4 from DC, which is directly related to the ability to perform ADCC (Supplemental Figure 8, A and B).

Overall, we identified different NK populations associated with specific functions. A CD16+CD158b+NKp30+NKG2DdimKLRG1dim population correlated with better natural cytotoxicity activity, while a unique memory-like NK cell population was associated with enhanced ADCC responses, especially in DC. This memory-like NK subset is characterized by the expression of CXCR3, indicating its potential to migrate to inflamed tissues, and NKG2A, an inhibitory receptor that identifies early-differentiated NK cells. This identified NK cell population represents a characteristic feature of individuals exhibiting natural HIV control.

Memory-like NK cells coexpressing NKG2A and CXCR3 exhibit heightened ADCC responses against HIV-infected cells. Next, we aimed to validate whether the memory-like NK cell population previously identified and linked to ADCC responses indeed displayed enhanced ADCC activity. Given the unusual coexpression of NKG2C and NKG2A on NK cells, we initially evaluated the expression of these markers in NK cells across all study groups by flow cytometry. As expected, the fraction of circulating NK cells expressing both markers was low (<15% of total CD56+ NK cells) (Supplemental Figure 9). Due to cell number limitations, ADCC assays were only performed on DC samples. We performed a functional ADCC assay using samples from 5 DC individuals. Distinct NK cell populations were isolated by FACS, based on the expression of selected receptors (Figure 4A). The cell numbers obtained after cell sorting are shown in Supplemental Table 2. Briefly, CD16+ cells derived from CD56+ NK cells were classified based on the expression of NKG2A, NKG2C, CD57, and CXCR3. NK cell classification included NKA (NKG2A+NKG2C+CD57+CXCR3+), NKB (NKG2A–NKG2C+CD57+CXCR3+), NKC (NKG2A+NKG2C–CD57+CXCR3+), and NKD (NKG2A–NKG2C–CD57+CXCR3+). CD16– NK cells were designated as NKE. NK cells were then incubated with HIV+ plasma and cocultured with ACH-2 HIV-infected cells for 4 hours. We calculated cell killing as the reduction in the percentage of virally infected cells.

NK cells from DC enriched in specific receptors exhibit an increased ability to kill HIV-infected cells. Distinct NK cell subsets were isolated from n = 5 DC individuals (median CD4+ T cell count = 1030 cells/μL; median viral load <40 copies HIV-1 RNA/mL) using FACS based on selected receptors, and assessed for ADCC responses. (A) Gating strategy of FACS-isolated NK cell populations based on the expression of CD16, NKG2A, NKG2C, CD57, and CXCR3. Five populations were identified: NKA (CD16+NKG2A+NKG2C+CD57+CXCR3+), NKB (CD16+NKG2A–NKG2C+CD57+CXCR3+), NKC (CD16+NKG2A+NKG2C–CD57+CXCR3+), NKD (CD16+NKG2A–NKG2C–CD57+CXCR3+), and NKE (CD16–). (B) Violin plot depicting the ADCC activity mediated by each specific NK cell population based on the marker expression presented in A. Graphs include group medians and ranges. Statistical comparisons were performed using Friedman’s test followed by Dunn’s multiple-comparison test. *P < 0.05.

NKA, characterized by the coexpression of NKG2A and NKG2C, exhibited a more robust ADCC response against HIV-infected cells than any other NK population (median ADCC = 20.78%), especially compared with NK cells lacking CD16 expression (NKE, median ADCC = 1.54%), NKC (median ADCC = 3.49%), and NKD (median ADCC = 1.38%) (P < 0.05) (Figure 4B). In general, NKA exhibited the ability to eliminate approximately 50% more HIV-infected cells than the other NK cell subpopulations (median ADCC for NKB = 8.33%, NKC = 3.49%, NKD = 1.38%). Altogether, our findings suggest that the expression of the receptors CD16, NKG2A, NKG2C, CD57, and CXCR3 on NK cells identifies a memory-like population with enhanced capacity for killing HIV-infected cells through ADCC.

Single-cell transcriptional analyses identifies NKA as a population with a unique transcriptional signature linked to increased NK cell effector functions, migration, and antiviral responses. Finally, to elucidate the molecular expression patterns associated with enhanced ADCC responses in DC individuals, we conducted single-cell RNA-seq analyses. We aimed to characterize and identify differentially expressed genes within NKA, particularly in comparison to conventional adaptive NK cells. For comparative purposes, we analyzed 3 related NK cell populations previously delineated in the functional ADCC analysis (NKA, NKB, and NKC). All 3 populations were characterized by high protein expression of CD16, CXCR3, and CD57. A total of 7226, 7354, and 8180 cells were sequenced for NKA, NKB, and NKC, respectively. NKA exhibited 2618 genes and 28,800 reads per cell, NKB showed 2516 genes and 26,791 reads per cells, and NKC displayed 2584 genes and 24,379 reads per cell. To check whether the data were appropriate for downstream analysis, different types of quality controls were performed, including the quality control of raw data (FastQC), the alignment and quantification of reads to described mRNAs (Cell Ranger), the number of transcripts per cell, and the percentage of counts coming from different transcript types, as previously described (46–48).

To delineate the heterogeneity of ADCC-mediating NK cells, we compared the transcriptional landscape of the distinct NK cell subsets by projecting cells into 2 dimensions by uniform manifold approximation and projection (UMAP) analysis. UMAP analysis unveiled NKB and NKC cell subsets as 2 separate populations, while NKA cells represented an intermediate state between both (Figure 5A). Unsupervised clustering analyses via shared nearest neighbor (SNN) modularity optimization–based clustering algorithm resulted in the identification of 13 distinct NK cell clusters (Figure 5B). Of note, clustering was not significantly influenced by the cell cycle (Supplemental Figure 10A). Differential expression analysis identified conserved markers for each cluster, irrespective of the NK cell population (Supplemental Table 3). For optimization purposes, differential expression analysis only included genes with a minimum log2(fold change) above 0.25 and expressed in at least 10% of cells in either of the 3 populations tested.

Unique transcriptional signatures define ADCC-mediating NK cells from DC. Gene expression analysis by single-cell RNA-seq. (A) UMAP visualization of NK cell populations sorted from DC individuals (NKA in red, NKB in green, NKC in blue). (B) UMAP visualization of 13 distinct NK cell clusters identified from NK cell populations sorted by unsupervised hierarchical clustering. (C) Number of cells per cluster and sample as counts (left) or proportions (right). (D) Heatmap depicting the top 5 genes most differentially expressed (upregulated) in each NK cell cluster. (E) The top panel shows a volcano plot illustrating the differentially expressed genes between NKB and NKA subsets. Genes significantly overexpressed in NKB compared with NKA are highlighted in red, while those underexpressed in NKB relative to NKA are shown in blue. The bottom panel presents representative UMAP plots depicting the expression of genes upregulated in NKA compared with NKB. (F) Significant canonical pathways predicted by Gene Ontology Biological Process analysis (GO-BP) of differentially expressed genes in NKA versus NKB. (G) The top panel presents a volcano plot illustrating the differentially expressed genes between NKC and NKA subsets. Genes significantly overexpressed in NKC relative to NKA are shown in red, while those underexpressed in NKC compared with NKA are depicted in blue. The bottom panel displays representative UMAP plots highlighting the expression of genes upregulated in NKA compared with NKC. (H) Significant canonical pathways predicted by GO-BP of differentially expressed genes in NKA relative to NKC. (I) Differentiation trajectory analysis of NK cell clusters, illustrating the progression and differentiation pathways of each population. Five distinct lineages were identified. The pseudotime inference for each NK cell cluster is presented, with UMAP plots color coded by inferred lineages. The scale indicates the maturation state, ranging from yellow (least mature) to dark blue (most mature).

The NKA population, defined by the coexpression of NKG2C and NKG2A, encompassed clusters C0, C5, C6, and C10 (Figure 5, B and C). Cluster C0, the most prevalent within the NKA population, exhibited a transcriptional profile marked by the upregulation of genes indicative of adaptive NK cell signatures, including IL32 and CD3E, while notably lacking FCER1G (30) (Figure 5D). Additionally, it exhibited MTCYB expression, signifying elevated metabolic demands, and IFITM2 expression, indicating an enhanced antiviral response (Figure 5D). Moreover, cluster C5 displayed a transcriptomic enrichment of genes encoding soluble factors integral to NK cell effector functions, such as CCL3, CCL4, and CCL4L2, as well as XCL2, a chemokine characteristic of naive CD56bright NK cells involved in the activation of the immune system and chemotaxis (49) (Figure 5D). Additionally, cluster C6 demonstrated upregulation of genes implicated in actin cytoskeleton remodeling and lipid metabolism, including TRIO, PITPNC1, CASK, and APBA2 (Figure 5D). Importantly, cluster C10 exhibited an upregulation of genes coding for IFN-induced proteins associated with the NK cell antiviral response, specifically MX1, MX2, ISG15, and IFI6 (Figure 5D). The NKB subset, characterized by high expression of NKG2C, comprised clusters C2, C4, and C7 (Figure 5, B and C). Clusters C2 and C4 presented the canonical transcriptional profile of adaptive NK cells, exhibiting upregulated expression of IL32 and CD3E, while lacking FCER1G and CD7 expression (Figure 5D). Furthermore, these clusters showed enhanced expression of genes associated with differentiation, maturation, and effector functions (THEMIS, RORA, PDE3B, and CADM1) (Figure 5D). Notably, cluster C7 was defined by the upregulation of genes involved in cell adhesion, activation, and differentiation (ANK3, IL7R, GZMK, and NELL2) (Figure 5D). By contrast, the NKC subset, defined by elevated NKG2A expression, comprised clusters C1, C3, C8, and C9 (Figure 5, B and C). Clusters C1 and C3 were characterized by an upregulation of genes involved in migration and cytokine production (SNX9, SGCD, IGF2R, and FCE1RG) along with chemokines (CCL3 and CCL4) (Figure 5D). Additionally, clusters C8 and C9 exhibited upregulated expression of genes regulating NK cell activation (NFKBIA and LINGO2) (Figure 5D).

Despite exhibiting adaptive NK cell traits, a direct comparison of the transcriptomic profile of the NKA subset with the conventional memory-like NKB population unveiled a heightened expression of CD7, a marker typically downregulated in adaptive NK cells (Figure 5E). NKA cells also presented elevated gene expression levels of CCL3, CCL4, and XCL2 chemokines, known for their pivotal role in NK cell activation and recruitment (Figure 5E and Supplemental Figure 10, B and C). Additionally, NKA exhibited heightened expression of ISG15, an IFN-stimulated gene (ISG) pivotal in inducing IFN-γ in the antiviral response, and MTCYB, a regulator of mitochondrial biogenesis and ATP synthesis, implicated in the secretion of cytolytic granules and cytokines (50) (Figure 5E). Of note, compared with NKB, NKA showed downregulation of RORA, a transcription factor involved in ILC2 differentiation (51) (Figure 5E and Supplemental Figure 10, B and C). Furthermore, Gene Ontology (GO) analysis revealed that the NKA subset exhibited increased pathways related with enhanced lymphocyte chemotaxis, migration, and cell regulation (Figure 5F).

Moreover, relative to the NKC subset, the NKA population displayed increased expression of IL32 and THEMIS, which are involved in the regulation of the NK cell cytolytic response and memory NK cell differentiation (Figure 5G and Supplemental Figure 10, D and E). Additionally, compared with NKC, NKA exhibited the upregulation of GZMM and CD3E genes that are implicated in NK cell cytotoxicity (Figure 5G and Supplemental Figure 10, D and E) and compatible with the upregulation of pathways related to cell cytotoxicity, IL-2 production, and response to IFN antiviral response (Figure 5H). In contrast with NKC, NKA exhibited downregulation of CCL3 and CCL4 chemokines, alongside FCER1G, which is associated with an NK-mediated regulation of the CD8+ T cell response in virus control (52) (Figure 5H). Importantly, compared with both NKB and NKC populations, NKA exhibited elevated expression of XCL2, CD7, MTCYB, IFITM2, and ISG15 that are associated with NK cell migration, differentiation, cell metabolism, and antiviral response (Figure 5, E–H).

Finally, we performed a gene expression trajectory analysis to elucidate the continuum of dynamic changes occurring within these distinct NK cell clusters. A total of 5 distinct lineages, defined as ordered sets of cell clusters sharing a common starting or ending point, were identified using Slingshot (Bioconductor) (Supplemental Table 4 and Figure 5I). Additionally, for each identified lineage, pseudotimes were determined, representing the variable indicating the transcriptional progression of each cell toward the terminal state (Figure 5I). Notably, irrespective of their origin, cells from all lineages transitioned through the NKA population, specifically cluster C10, and consistently ended within cluster C4, highly predominant in the NKB population. This observation suggests that all NK cell subsets studied transition into an intermediate population, distinguished by the expression of NKG2A and NKG2C. Furthermore, the final differentiation stage of all these identified lineages constitutes cells with a differentiated and adaptive NK cell signature (Supplemental Table 4 and Figure 5I).

Altogether, our data support the notion that NKA cells expressing NKG2C and NKG2A, which are associated with higher ADCC responses in DC, represent a transitional state between a less differentiated population (NKC) and a more mature and adaptive state (NKB). Importantly, these cells are characterized by upregulation of specific genes associated with enhanced effector functions, cell migration, metabolism, and antiviral responses. Overall, our data support the potential of NKA cells to mediate strong cytotoxic pressure against HIV.