High Frequencies of Polyfunctional CD8⁺ NK Cells in Chronic HIV-1 Infection Are Associated with Slower Disease Progression

ABSTRACT

Natural killer (NK) cells are effector and regulatory innate immune cells and play a critical role in the first line of defense against various viral infections. Although previous reports have indicated the vital contributions of NK cells to HIV-1 immune control, nongenetic NK cell parameters directly associated with slower disease progression have not been defined yet. In a longitudinal, retrospective study of 117 untreated HIV-infected subjects, we show that higher frequencies as well as the absolute numbers of CD8⁺ CD3⁻ lymphocytes are linked to delayed HIV-1 disease progression. We show that the majority of these cells are well-described blood NK cells. In a subsequent cross-sectional study, we demonstrate a significant loss of CD8⁺ NK cells in untreated HIV-infected individuals, which correlated with HIV loads and inversely correlated with CD4⁺ T cell counts. CD8⁺ NK cells had modestly higher frequencies of CD57-expressing cells than CD8⁻ cells, but CD8⁺ and CD8⁻ NK cells showed no differences in the expression of a number of activating and inhibiting NK cell receptors. However, CD8⁺ NK cells exhibited a more functional profile, as detected by cytokine production and degranulation.

IMPORTANCE We demonstrate that the frequency of highly functional CD8⁺ NK cells is inversely associated with HIV-related disease markers and linked with delayed disease progression. These results thus indicate that CD8⁺ NK cells represent a novel NK cell-derived, innate immune correlate with an improved clinical outcome in HIV infection.

INTRODUCTION

Natural killer (NK) cells are traditionally seen as innate immune cells constituting a first line of immune defense against malignant cells and viruses (1, 2). The recognition of virally infected cells is mediated by an arsenal of inhibitory and activating receptors, and the sum of the signals mediated by these determines the functional activity of NK cells (2–4). Different subsets of NK cells have been described in the peripheral blood of humans (3, 5). The majority of peripheral blood NK cells are CD56^dim CD16⁺ cells, whereas lymph node-resident NK cells are predominantly CD56^bright NK cells (3, 5–7). Roughly 30% of peripheral blood NK cells express the CD8α homodimer, and these cells were shown to exhibit better survival during target cell killing (8, 9).

Genetic studies at the population level provided evidence for a protective role of NK cells in human immunodeficiency virus type 1 (HIV-1) infection. For instance, certain KIR3DL1 alleles in the context of HLA-Bw4 were shown to exert a strong protective influence during HIV infection (10). HIV-infected individuals expressing KIR3DS1 in conjunction with HLA-Bw4-80I exhibited a considerably slower progression to AIDS (11). Notably, the results of these epidemiological studies were supported by the findings of subsequent functional NK cell studies and by the observation of NK cell-mediated sieve effects in carriers of KIR3DS1 and HLA-Bw4-80I (12–14). Furthermore, KIR2DL2-expressing NK cells were suggested to mediate significant immune pressure against HIV, as evidenced by selected amino acid polymorphisms in the viral sequence leading to a decreased ability of NK cells to kill virus-infected CD4⁺ T cells (15). These data thus indicate that NK cells are potentially important contributors to the host immune defense against HIV-1. However, to this day the nongenetic NK cell parameters associated with a slower progression to clinical HIV-1 disease are unknown.

NK cells are also subject to profound alterations in chronic HIV-1 infection. A number of reports have demonstrated phenotypic and functional changes in peripheral blood NK cells during HIV-1 infection in humans (16–21). Progressive HIV-1 infection is associated with a decline in cytotoxic CD56^dim CD16⁺ NK cells and an expansion of dysfunctional CD56⁻ CD16⁺ NK cells (22, 23). We have previously shown a decline of less differentiated and functionally more potent CD56^dim CD16⁺ NK cells, which are either CD57⁻ or CD57^dim (20). Furthermore, we demonstrated significant correlations between HIV loads and a decrease of CD4⁺ T cell counts with a loss of CCR7 on CD56^bright NK cells, indicating that an NK cell-derived parameter can robustly correlate with HIV-related disease markers (18). Notably, NK cells from patients able to spontaneously control HIV replication and long-term nonprogressors maintained unchanged properties, especially with regard to their ability to express natural cytotoxicity receptors (24).

Here, we performed a retrospective study with 117 untreated HIV-infected patients and found that the relative numbers and the absolute counts of CD8⁺ CD3⁻ lymphocytes and NK cells were significantly correlated with slower clinical progression to AIDS. Subsequent analysis revealed that the majority of CD8⁺ CD3⁻ cells were CD56-expressing NK cells. The loss of CD8⁺ NK cells was significantly correlated with HIV-1 plasma loads and inversely correlated with CD4⁺ T cell counts as well as the CD4/CD8 T cell ratio. Furthermore, CD8⁺ NK cells exhibited a more functional profile than their CD8⁻ counterparts. These results therefore indicate that the frequency of highly functional CD8⁺ NK cells represents a robust correlate for delayed disease progression to AIDS in chronic HIV infection.

MATERIALS AND METHODS

Study subjects.

The data presented in this study were derived from a cohort of, altogether, 162 HIV-seropositive patients and 15 uninfected control subjects. Of the 162 HIV-infected patients, 117 untreated subjects with CD4⁺ T cell counts above 500 cells/μl were longitudinally followed at the Medizinische Hochschule Hannover (MHH) for up to 90 months, with the median follow-up time being 30.6 months (interquartile range, 18.5 to 50.4 months). A description of this longitudinal patient cohort is provided in Table 1. Peripheral blood samples from 60 untreated HIV-seropositive subjects chosen on the basis of sample availability were analyzed in a cross-sectional study. Twenty-five of these 60 patients were part of the longitudinal study, but none of these samples were taken at the baseline or endpoint. Blood samples for the cross-sectional study were taken at random time points during the patients' visits after study recruitment and were thus taken at various clinical stages. To study the effect of antiretroviral therapy (ART) on CD8 expression on NK cells, samples from 28 patients who had been on ART for at least 1 year and who had suppressed viral loads were analyzed (8/28 of these patients overlapped with the longitudinal cohort). We obtained samples from 21 individuals before and after the initiation of ART (9/21 of these individuals overlapped with the longitudinal cohort). All study subjects were recruited at the HIV outpatient clinic of MHH and gave written, informed consent prior to their participation. The study was approved by the local ethics committee (Votum der Ethikkommission der MHH, approval no. 3150). Plasma HIV-1 RNA levels were measured using a Cobas TaqMan HIV-1 test (Roche Diagnostics), which has a lower limit of detection of 34 copies/ml. NK cells and CD4⁺ and CD8⁺ T cells, as well as CD8⁺ CD3⁻ lymphocyte counts, were routinely determined by a flow cytometry-based assay using a Cyto-Stat tetraCHROME flow cytometry system (Beckman Coulter).

TABLE 1.

Patients' characteristics at baseline and endpoint

Characteristic	Value ata:
	Baseline	Endpoint
No. of lymphocytes/μl	2,207 (1,917–2,653)	1,781 (1,323–2,264)
No. of CD3 T cells/μl	1,759 (1,510–2,132)	1,393 (1,082–1,823)
% CD3 T cells among lymphocytes	81 (77–85)	81 (76–86)
No. of CD4 T cells/μl	606 (546–711)	339 (280–505)
% CD4 T cells among lymphocytes	29 (24–34)	23 (18–28)
CD4/CD8 ratio	0.59 (0.46–0.77)	0.4 (0.29–0.58)
Viral load (no. of copies/ml)	13,247 (2,425–34,600)	23,623 (5,443–95,877)
No. of CD8+ CD3 cells/μl	64 (49–101)	45 (27–75)
% CD8+ CD3 cells among lymphocytes	3 (2–4)	3 (2–4)
No. of NK cells/μl	179 (120–270)	105 (66–174)
% NK cells among lymphocytes	8 (6–11)	7 (4–11)

^aAll values are provided as medians (interquartile ranges).

Blood sample processing.

Peripheral blood mononuclear cells (PBMCs) were isolated from fresh blood using Ficoll (Biochrom) density gradient centrifugation. Cells were washed three times with phosphate-buffered saline (PBS), and aliquots of 10⁷ PBMCs were cryopreserved in heat-inactivated fetal calf serum (FCS) supplemented with 10% dimethyl sulfoxide (Merck).

Phenotypic analysis of NK cells by flow cytometry.

PBMCs were thawed and washed in PBS supplemented with FCS. Staining and flow cytometric analyses were performed as described before (19, 25). Intracellular expression of perforin, granzyme B, and Ki-67 was analyzed ex vivo in unstimulated NK cells using a Fix and Perm kit (Life Technologies) following the manufacturer's protocol. The following monoclonal antibodies were employed in our study: anti-CD19 peridinin chlorophyll protein (PerCP), anti-CD14 PerCP, anti-CD3 PerCP, anti-CD16 allophycocyanin (APC)-H7, anti-CD94 APC, anti-NKG2D APC, purified anti-NKp46, anti-NKp80 biotin, purified anti-CD85j, anti-CD2 biotin, anti-CD57 fluorescein isothiocyanate (FITC), anti-HLA-DR V500, anti-perforin phycoerythrin (PE), anti-granzyme B Alexa Fluor 700, anti-Ki-67 Alexa Fluor 647 (BD Biosciences), anti-CD8 FITC, anti-KIR2DL2/DL3 APC (BioLegend), anti-CD3 ECD, anti-CD62L R-phycoerythrin–Texas Red-X (ECD), anti-NKp30 PE, and anti-NKG2A PE (Beckman Coulter). Secondary staining was performed using streptavidin Alexa Fluor 700 and goat anti-mouse Pacific Blue (Life Technologies). Stained cells were immediately run on an LSR II SORP flow cytometer (BD Biosciences). At least 1,000 gated events had to be acquired for CD8⁺ and CD8⁻ NK cells for samples to be evaluable for phenotypic analyses.

Degranulation and intracellular cytokine staining assay.

Functional NK cell assays were performed as described previously (19, 26). Briefly, 10⁵ sorted NK cells were stimulated with 100 ng/ml interleukin-12 (IL-12), 10 ng/ml IL-15, and/or K562 cells at an effector-to-target cell ratio of 2:1. After 1 h of culture, 10 μg/ml brefeldin A (Sigma), GolgiPlug protein transport inhibitor, and anti-CD107a PE (BD Biosciences) were added and cells were cultured for another 5 h. Anti-gamma interferon (anti-IFN-γ) Pacific Blue (Biolegend), anti-tumor necrosis factor alpha (anti-TNF-α) Alexa Fluor 700, and anti-macrophage inflammatory protein 1β (anti-MIP-1β) PE (BD Biosciences) were used to detect intracellular expression of cytokines after fixation and permeabilization.

Univariate and multivariate survival analyses.

Survival from the date of recruitment to the date of the endpoint, that is, disease progression, was calculated. To define survival comparison groups, patients were ranked according to their absolute and relative CD8⁺ CD3⁻ cell counts or absolute and relative NK cell counts, and subjects belonging to the first tertile were compared to patients belonging to the second and third tertiles. Initially, the study endpoint was defined as either the initiation of ART or a drop of the CD4⁺ T cell count to below 200 cells/μl. Since the patients had different CD4⁺ T cell counts (median, 316 cells/μl; interquartile range, 251 to 433.5 cells/μl) when ART was initiated, we sought to answer the question whether there was a potential ART bias. Thus, further survival analyses were performed using only the initiation of ART as the endpoint. Individuals who did not achieve their endpoint were censored on their last clinic visit date or, at the latest, after 90 months. GraphPad Prism (version 5.0) software was used to perform log-rank tests. Cox proportional hazards regression (backward conditional) was calculated using IBM SPSS Statistics 20 software. Included in the analysis were continuous variables: percentages of CD4⁺ T cells, CD8⁺ T cells, CD8⁺ CD3⁻ cells, and NK cells; the CD4⁺/CD8⁺ T cell ratio; and age. Categorical variables were gender (females versus male), ethnicity (Caucasian, African, Asian), viral copy numbers (per 100,000-copy/ml increment), and the absolute counts of CD4⁺ T cells, CD8⁺ T cells, CD8⁺ CD3⁻ cells, and NK cells (per 100-cell/μl increment).

Further statistical analyses.

For all other statistical analyses as well as graph plotting, GraphPad Prism software was employed. Spearman rank correlation analysis was used to assess the correlations between the frequency of CD8⁺ NK cells and HIV disease parameters. Pestle and Spice software was provided by the NIH (27) and used to visualize the coexpression patterns of surface markers and multifunctional NK cell responses. Unless otherwise indicated, an unpaired, two-tailed t test for comparison of two groups or one-way analysis of variance, followed by the Tukey posttest, for comparison of more than two groups was performed, and P values of less than 0.05 were considered significant.

RESULTS

High frequencies of CD8⁺ CD3⁻ lymphocytes are associated with slower HIV-1 disease progression.

A cohort of 117 untreated, chronically HIV-infected subjects was followed for up to 90 months at the HIV outpatient clinic of MHH. At the time of inclusion, the patients had a baseline CD4⁺ T cell count of over 500 cells/μl. The frequency as well as the absolute numbers of CD8⁺ CD3⁻ cells was routinely determined from all HIV-infected patients at the HIV outpatient clinic of MHH. To assess the association of the frequencies of CD8⁺ CD3⁻ lymphocytes with HIV disease progression, we ranked the cohort according to their absolute or relative CD8⁺ CD3⁻ cell counts. We compared patients belonging to the first tertile with high CD8⁺ CD3⁻ cell counts (>83 cells/μl or >3.9% of lymphocytes, n = 39) to patients belonging to the second and third tertiles with low CD8⁺ CD3⁻ cell counts (<83 cells/μl or <3.9% of lymphocytes, n = 78) at the baseline. We initially defined disease progression as a loss of CD4⁺ T cells to levels below 200 cells/μl or the initiation of ART. We identified a significantly slower disease progression in the cohort with high numbers of CD8⁺ CD3⁻ cells than in the patients with lower numbers of CD8⁺ CD3⁻ cells (Fig. 1A and B; see Table S1A in the supplemental material).

FIG 1.

High absolute number and high frequencies of CD8⁺ CD3⁻ cells are associated with slower HIV disease progression. Patients were ranked according to their absolute or relative numbers of CD8⁺ CD3⁻ cells, and subjects belonging to the first tertile (n = 39) were compared to patients belonging to the second and third tertiles (n = 78). Kaplan-Maier survival analyses were performed to study the effects of either high or low absolute and relative baseline numbers of CD8⁺ CD3⁻ cells on disease progression. Endpoints were defined as ART initiation or a drop of CD4⁺ T cell numbers below 200 cells/μl (A, B) or as ART initiation only (C, D). (E) A comparison of survival of patients with either high or low absolute CD8⁺ CD3⁻ lymphocyte counts after censoring of patients with ART initiation when CD4⁺ T cell counts were >359 cell/μl is shown. (F) The survival of patients with high absolute CD8⁺ CD3⁻ cell counts (n = 37) was compared to that of patients with low CD8⁺ CD3⁻ cell counts (n = 73) after exclusion of 7 elite controllers (EC).

ART was initiated at times at which the patients had various CD4⁺ T cell counts ranging from 202 to 619 cells/μl (median, 251 cells/μl). To examine whether there was a potential bias caused by ART, we repeated the survival analysis after censoring patients with CD4⁺ T cell counts below 200 cells/μl at the endpoint, thus using only the initiation of ART as the endpoint. Again, we found significantly delayed HIV disease progression in patients with high absolute and relative CD8⁺ CD3⁻ lymphocyte counts (Fig. 1C and D; see Table S1A in the supplemental material). Furthermore, we ranked all patients who had reached their endpoint by ART initiation according to their CD4⁺ T cell counts at the endpoint. After censoring subjects belonging to the first tertile (ART initiation when CD4⁺ T cells counts were >359 cells/μl, n = 12), patients with high CD8⁺ CD3⁻ lymphocyte counts still exhibited prolonged survival (Fig. 1E). Since our cohort had an unusually high number of elite controllers (defined as those with undetectable viral loads during at least three visits during a period of 12 months or longer), we sought to address if these could potentially influence the findings of our survival analysis. After exclusion of these 7 patients, we still found the survival difference to be statistically significant (Fig. 1F).

Furthermore, we used Cox proportional hazards regression (backward conditional, upper tertile versus lower two tertiles), adjusting for age, gender, ethnicity, CD4⁺ T cell count, CD8⁺ T cell count, CD4/CD8 ratio, CD8⁺ CD3⁻ cell count and percentage, and viral load. The final Cox model for the combined endpoint of either ART initiation or a drop of the CD4⁺ T cell count to below 200 counts/μl, the ART initiation only endpoint, and a further analysis excluding elite controllers included the percentages of CD4⁺ T cells, CD8⁺ CD3⁻ cell count, and viral load as independent prognostic factors (see Table S1B in the supplemental material). For the combined endpoint, a higher CD4⁺ percentage (odds ratio = 0.93), a higher CD8⁺ CD3⁻ count (odds ratio = 0.46 per increase by 100 cells/μl), and a lower viral load (odds ratio = 1.42 per increase by 100,000 copies per ml) were associated with slower progression to the endpoint. We thus show that high numbers of CD8⁺ CD3⁻ cells are independently associated with an improved clinical outcome, when the above-mentioned factors are controlled for (see Table S1B in the supplemental material).

CD8⁺ CD3⁻ cells are mostly CD56-expressing NK cells.

Having shown that higher frequencies of CD8⁺ CD3⁻ lymphocytes are associated with a prolonged AIDS-free survival, we next sought to characterize these cells. Since CD8 in conjunction with other cell surface antigens is frequently used to define nonhuman primate NK cells (28), we hypothesized that the majority of CD8⁺ CD3⁻ cells are NK cells. To address this question, we analyzed randomly sampled PBMC specimens derived from 37 HIV-infected patients. We also included specimens from 15 HIV-seronegative control subjects. After gating on CD8⁺ CD3⁻ lymphocytes, a subsequent analysis of CD56 and CD16 expression revealed that the majority of these cells were previously well-defined peripheral blood NK cell subsets, namely, CD56^bright cells (4.7% ± 2.2% [standard deviation {SD}] and 6.3% ± 5.4% [SD] for HIV-infected patients and control subjects, respectively), CD56^dim CD16⁺ cells (74% ± 11.6% [SD] and 50% ± 22.9% [SD]), and CD56⁻ CD16⁺ cells (6.7% ± 4.1% [SD] and 19.3% ± 12.2% [SD]) (Fig. 2A and B). We and others previously reported a decrease of CD56^dim CD16⁺ NK cells and an expansion of less functional CD56⁻ CD16⁺ NK cells in chronic HIV-1 infection (17, 19, 20). These NK cell subset alterations were also evident in the distribution of NK cell subsets within CD8⁺ CD3⁻ cells (Fig. 2B).

FIG 2.

CD8⁺ CD3⁻ cells are mostly CD56-expressing NK cells. (A) The gating strategy employed to analyze CD8⁺ CD3⁻ cells is shown. (B) Pie charts comparing the composition of CD8⁺ CD3⁻ lymphocytes in terms of NK cell subsets in 15 seronegative control and 37 HIV-1-seropositive subjects are shown. (C, D) Kaplan-Maier survival analyses were performed to compare the progression-free survival of untreated HIV-seropositive patients. Subjects were ranked according to their absolute (C) or relative (D) NK cell counts. Patients with high NK cell numbers belong to the first tertile, and patients with low NK cell numbers belong to the second and third tertiles. (E, F) Kaplan-Maier survival analyses of untreated HIV-infected patients comparing the first tertile to the second and third tertiles on the basis of absolute counts (E) or relative frequencies (F) of CD8⁺ CD3⁻ cell-subtracted NK cells.

Notably, the portion of CD56⁻ CD16⁻ cells was larger among HIV-seropositive subjects than among uninfected blood donors. Upon closer examination, we found that these cells were mostly negative for NK cell markers, such as NKp46 and NKp80, and the majority of these cells had high levels of HLA-DR expression, strongly suggesting that these were not NK cells (data not shown). Importantly, a high percentage of these cells did not translate into high absolute numbers of CD8⁺ CD3⁻ CD56⁻ C16⁻ cells. For example, one patient with 70% CD56⁻ CD16⁻ cells among CD8⁺ CD3⁻ lymphocytes had very low numbers of CD8⁺ CD3⁻ cells (0.3% of lymphocytes). Thus, a high percentage of CD56⁻ CD16⁻ cells among the CD8⁺ CD3⁻ lymphocytes was, in general, reflective of lower numbers of the parent population.

Since the majority of CD8⁺ CD3⁻ cells were thus conventional blood NK cells, we sought to answer the question whether there were comparable associations of high frequencies of NK cells with slower HIV-1 disease progression. The absolute and relative numbers of NK cells defined as CD3⁻, CD16⁺, and/or CD56⁺ cells were determined for all HIV-seropositive patients as part of the routine clinical analysis. We thus ranked the cohort according to their absolute and relative NK cell numbers and compared the survival of patients belonging to the first tertile with that of patients belonging to the second and third tertiles. Patients with high absolute NK cell counts (>230 cells/μl) were associated with a delayed onset of disease progression compared to that in subjects with lower NK cell numbers (Fig. 2C; see Table S1A in the supplemental material). Elevated relative NK cell frequencies (>10.3%) were associated with a modestly improved clinical outcome (Fig. 2D).

We next sought to determine whether CD8⁺ CD3⁻ cells or overall NK cells primarily correlated with delayed HIV-1 disease progression. To this end, we either subtracted the absolute counts of CD8⁺ CD3⁻ cells from the absolute counts of total NK cells or subtracted the percentage of CD8⁺ CD3⁻ cells from the percentage of total NK cells. With the caveat that there was some variability in the frequency of CD56⁻ CD16⁻ non-NK cells among the CD8⁺ CD3⁻ lymphocytes, this approach allowed us to determine an approximation of the absolute or relative numbers of CD8⁻ NK cells. We subsequently performed a survival analysis comparing HIV-1-seropositive patients with either high (>80 cells/μl, or >4%) or low (<80 cells/μl, or <4%) counts of NK cells devoid of CD8. Interestingly, we found that neither the frequency nor the absolute count of CD8⁺ CD3⁻ cell-subtracted NK cells was associated with slower HIV-1 disease progression (Fig. 2E and F), suggesting that CD8⁺ cells but not CD8⁻ NK cells represent a correlate for an improved HIV-1 disease outcome.

CD8⁺ NK cells correlate with clinical parameters in untreated, chronic HIV-1 infection.

Given that CD8⁺ CD3⁻ cells predominantly consist of CD8⁺ NK cells expressing CD56 and/or CD16 and having shown that a high frequency of these cells correlates with delayed HIV-1 disease progression, we next analyzed the frequency of CD8-expressing cells in total peripheral blood NK cells. The clinical flow cytometric data, which were the basis for our survival analyses, allowed a solid enumeration of lymphocyte subsets. However, as a consequence of the limited number of parameters employed in these analyses (4 or 5 colors), residual contaminations by non-NK cells cannot be ruled out. Whereas these potential contaminations should not affect the overall number of NK cells much, there could be an impact on downstream phenotypic and functional studies. Thus, for all subsequent analyses, we performed 9- or 10-color polychromatic flow cytometry and were able to employ an improved gating strategy. We excluded doublets and dead cells and defined NK cells as CD3⁻, CD14⁻, CD19⁻, CD56⁺, and/or CD16⁺ lymphocytes, as depicted in Fig. S1A in the supplemental material. In addition, possible myeloid cell contamination was ruled out by exclusion of HLA-DR⁺ cells (see Fig. S1A in the supplemental material). With gating on total NK cells, we determined the frequency of CD8⁺ cells in a cross-sectional analysis of 33 untreated and 27 treated HIV-seropositive individuals and 15 uninfected control subjects (Fig. 3A and B). Importantly, the samples from untreated HIV-infected patients were randomly taken and thus represent samples from a variety of clinical stages of HIV immunopathogenesis. We observed substantially lower frequencies of CD8⁺ NK cells in untreated HIV-1-infected individuals than healthy controls (Fig. 3B), and these cells were partially restored after 1 year of ART. To further investigate the effect of antiviral treatment, we performed a follow-up study in previously untreated patients 1 year after the initiation of ART. Both the relative numbers and the absolute counts of CD8⁺ NK cells were significantly increased in patients after they received antiviral treatment (Fig. 3C). In summary, we show that HIV infection leads to a significant decline of CD8⁺ NK cells. Our cross-sectional patient data as well as longitudinal patient data indicate that antiretroviral treatment can restore the number of CD8⁺ NK cells almost to levels comparable to those in uninfected control subjects.

FIG 3.

Frequencies of CD8⁺ NK cells correlate with clinical disease parameters of HIV-1 infection. (A) Representative histogram showing CD8 expression on gated NK cells in healthy controls and in HIV-1-infected patients. (B) The frequencies of CD8⁺ NK cells were compared in 15 control subjects and in either 45 untreated or 28 treated HIV-infected patients. *, P < 0.05; **, P < 0.01; ***, P < 0.001. (C) Analysis of the frequency and absolute counts of CD8⁺ NK cells in 21 HIV-infected patients before and after treatment. A paired Student's t test was performed to determine statistically significant differences. (D, E) Spearman's rank correlation analysis of the frequency of CD8⁺ NK cells in 60 untreated HIV-infected patients compared with the HIV load (D) and CD4⁺ T cell counts (E). (F, G) Spearman's rank correlation analysis of the frequency of CD8⁺ NK cells in 60 untreated HIV-infected subjects compared with the percentages of CD4⁺ T cells (F) and the CD4/CD8 T cell ratio (G).

To further explore the relationship between HIV disease states and the frequency of CD8⁺ NK cells, we investigated the association of CD8⁺ NK cells with CD4⁺ T cell counts and viral loads in untreated HIV-1-seropositive individuals. We observed a modest negative correlation between the frequencies of CD8⁺ NK cells and HIV loads (r = −0.33, P = 0.01) and a significant, positive correlation between the frequencies of CD8⁺ NK cells and absolute CD4⁺ T cells counts (r = 0.33, P = 0.009) (Fig. 3D and E). There was also a significant positive correlation between the frequencies of CD8⁺ NK cells and the frequencies of CD4⁺ T cells (r = 0.26, P = 0.046; Fig. 3F). Furthermore, since the ratio of CD4⁺ to CD8⁺ T cells is an important clinical parameter in HIV-caused immune disease, we performed a correlation analysis between the frequencies of CD8⁺ NK cells and the CD4/CD8 T cell ratio. There was a trend for a positive correlation between the percentages of CD8-expressing cells and the CD4/CD8 T cell ratio, although the result did not reach statistical significance (r = 0.1, P = 0.1; Fig. 3G). Altogether, our data indicate a correlation between CD8⁺ NK cells and multiple surrogate markers of HIV-caused immune disease.

Phenotypic characterization of CD8⁺ NK cells.

The activity of NK cells is regulated by a sophisticated array of germ line-encoded activating or inhibiting receptors (2, 4). Having shown that the frequency of CD8⁺ NK cells represents a correlate for delayed HIV disease progression also reflected in the correlation with various HIV disease markers, we hypothesized that CD8⁺ NK cells possess a repertoire of NK cell receptors different from that in CD8⁻ NK cells. To test our hypothesis, we performed a detailed phenotypic analysis of CD8⁺ and CD8⁻ NK cells. However, neither in untreated HIV-infected patients (Fig. 4A) nor in uninfected control subjects (see Fig. S1B in the supplemental material) were we able to find significant differences in the percentages of cells expressing CD62L, CD69, CD2, CD94, NKG2D, NKp30, NKp46, NKp80, CD85j, NKG2A, and KIR2DL2/DL3. We also measured PD-1 expression on NK cells. PD-1 expression was detectable on NK cells ex vivo in several subjects, but we found no difference in the frequency of cells expressing PD-1 when comparing CD8⁺ to CD8⁻ NK cells (data not shown).

FIG 4.

Phenotypic analyses of CD8⁺ and CD8⁻ NK cells in untreated HIV-infected individuals. (A) The frequencies of CD8⁺ and CD8⁻ NK cells expressing various NK cell receptors were compared. Bars indicate the means for 37 untreated HIV-infected individuals. (B, C) The patterns of coexpression of inhibiting receptors CD85j, KIR2DL2, and NKG2A (B) and activating receptors (C) on CD8⁺ and CD8⁻ NK cell subsets are shown. Pies and bar graphs indicate the means for all 37 untreated HIV-infected subjects. Statistically significant differences are noted. (D) The frequency of CD8⁺ and CD8⁻ NK cells expressing CD57 is shown. Horizontal bars show the mean. (E) The frequencies of CD56^bright, CD56^dim, and CD56^neg NK cell subsets among CD8⁺ and CD8⁻ NK cells are shown. The bar graph shows the means of data derived from 37 untreated HIV-seropositive patients. *, P < 0.05; **, P < 0.01.

The absence of notable phenotypic differences when studying NK cell receptors one at a time prompted us to investigate the coexpression patterns of either inhibiting or activating NK cell receptors. To this end, we performed Boolean gating analysis for the inhibiting receptors CD85j, KIR2DL2/DL3, and NKG2A or activating receptors NKG2D, NKp30, NKp46, and NKp80 on either CD8⁺ or CD8⁻ NK cells. Neither in HIV-infected patients (Fig. 4B and C) nor in uninfected control subjects (data not shown) were we able to find substantial differences in the majority of the coexpression patterns of these receptors in CD8⁺ NK cells from those in CD8⁻ NK cells. Statistically significant differences in frequencies were found for NK cells expressing only NKp30 among activating receptors and only NKG2A among inhibiting receptors, the frequencies of both of which were both higher among CD8⁻ NK cells (Fig. 4B and C).

We and others had previously demonstrated that CD57 expression on NK cells is consistent with a phenotype resembling terminal differentiation (20, 29, 30). To address the differentiation status of these subsets, we thus analyzed the expression of the senescence marker CD57 on CD8⁺ and CD8⁻ NK cells. In HIV-seronegative control subjects, CD8⁺ NK cells exhibited marginally higher frequencies of CD57-expressing cells than their CD8⁻ counterparts (see Fig. S1C in the supplemental material). We also observed a modest difference in HIV-1-seropositive individuals (Fig. 4D), indicating a moderately more differentiated phenotype in CD8⁺ NK cells. Finally, we determined the frequency of CD56^bright, CD56^dim, or CD56⁻ cells among either CD8⁺ or CD8⁻ NK cells and found almost identical NK cell subset distributions (Fig. 4E). These data thus indicate that apart from a moderately more differentiated phenotype of CD8⁺ NK cells, as indicated by higher numbers of CD57-expressing cells, CD8⁺ NK cells exhibit a phenotype remarkably similar to that of CD8⁻ NK cells.

The CD8⁺ subset exhibits a more functional profile than CD8⁻ NK cells.

Increasing levels of CD57 expression on NK cells were shown to be associated with an increased expression of perforin and granzyme B (20). Having shown that CD8⁺ NK cells are more likely to express CD57, we thus hypothesized a higher cytolytic potential within this NK cell subpopulation. To test this hypothesis, we assessed the ex vivo intracellular expression of perforin and granzyme B in untreated HIV-seropositive subjects and in our control group. Indeed, higher frequencies of granzyme B-expressing (granzyme B⁺) and perforin-expressing (perforin⁺) cells were detected among CD8⁺ NK cells than CD8⁻ cells in HIV-1-infected patients as well as in our cohort of HIV-seronegative blood donors (Fig. 5A; see also Fig. S2A in the supplemental material). Since an increase of granzyme B⁺ and perforin⁺ cells is not necessarily reflected in their capacity to degranulate (20), we evaluated the functional capacity of CD8⁺ and CD8⁻ NK cells by performing an intracellular cytokine staining assay. To this end, sorted NK cells were stimulated in the presence of K562 cells as well as IL-12 and IL-15 to achieve a robust activation inducing both cytokine production and degranulation. We measured the degranulation marker CD107a on the cell surface and the intracellular expression of IFN-γ and TNF-α. In addition, we were interested in the ability of NK cells to produce MIP-1β since previous studies indicated a virus-suppressive role for NK cell-derived CC chemokines by competitive blocking of the HIV coreceptor CCR5 (31–33). These experiments were carried out in 20 untreated HIV-seropositive subjects, which were a subset of the aforementioned patients from the cross-sectional study selected on the basis of sample availability as well as viral loads (patients with low levels of viremia as well as patients with high levels of viremia). For all four functions, there was a moderate, yet statistically significant increase in the number of responding cells in the CD8⁺ NK cell subpopulation compared to the number of responding CD8⁻ cells in untreated HIV-positive individuals (Fig. 5B). In uninfected control subjects, we observed a similar increase in the number of cells expressing IFN-γ, MIP-1β, and CD107a among CD8⁺ NK cells (see Fig. S2B in the supplemental material).

FIG 5.

CD8⁺ NK cells display a more functional profile. (A) Summary data comparing the frequencies of CD8⁺ and CD8⁻ NK cells expressing granzyme B or perforin in 35 untreated HIV-infected patients. (B) Summary data on the frequencies of CD8⁺ and CD8⁻ NK cells expressing IFN-γ, TNF-α, MIP-1β, or CD107a from 20 untreated HIV-infected patients after stimulation with K562 cells, IL-12, and IL-15 are shown. Bars show the means. (C) The patterns of CD107a, IFN-γ, MIP-1β, and TNF-α coexpression in CD8⁺ and CD8⁻ NK cells are shown. Pie charts show the means for 20 untreated HIV-seropositive subjects and 9 uninfected control subjects. The bar graph shows the means for 20 untreated HIV-infected blood donors. (D) The frequencies of NK cells expressing all four functional markers (IFN-γ, TNF-α, MIP-1β, and CD107a) were analyzed in 9 healthy control subjects and untreated HIV-infected patients with either high (>10,000 copy/ml, n = 8) or low (<10,000 copy/ml, n = 12) viral loads (VL).

We further studied the functional capacity of CD8⁺ and CD8⁻ NK cell responses by performing Boolean analysis of NK cells expressing IFN-γ, CD107a, TNF-α, and MIP-1β after stimulation. Notably, the frequencies of cells exhibiting all four functions were significantly enhanced in the CD8⁺ NK cell subset compared to that in the CD8⁻ cell subset in both untreated HIV-seropositive subjects and uninfected controls (Fig. 5C and D). In a similar vein, the frequencies of cells performing none of the four functions were higher in the CD8⁻ NK cell subpopulation than in the CD8⁺ cell population (Fig. 5C). Since these data were derived from untreated HIV-1-seropositive patients with varied viral loads, we divided the patients into groups with viral loads below or above 10,000 copies/ml. With the exception of one outlier, polyfunctional cells were not increased in CD8⁺ NK cells compared to CD8⁻ cells in patients with viral loads above 10,000 copies/ml. The majority of individuals with enhanced functional activities of CD8⁺ NK cells were found among HIV-seropositive subjects with viral loads below 10,000 copies/ml (Fig. 5D).

We also sought to determine the proliferative activity of CD8⁺ and CD8⁻ NK cells ex vivo. To this end, we analyzed the intracellular expression of Ki-67, a marker associated with cycling cells. As shown previously (18, 34), a greater number of Ki-67-positive (Ki-67⁺) cells was observed in untreated HIV-infected individuals than uninfected subjects (see Fig. S2C in the supplemental material). However, we were unable to detect differences in the relative frequencies Ki-67⁺ cells between CD8⁺ NK cells and CD8⁻ NK cells in both HIV-infected and uninfected subjects (see Fig. S2C in the supplemental material). This finding suggests that the prevalence of cycling cells among CD8⁺ NK cells and CD8⁻ NK cells is similar.

In summary, our data indicate that CD8⁺ NK cells are characterized by a greater responsiveness to stimulation with major histocompatibility complex (MHC) class I-devoid target cells and cytokines. These cells therefore exhibit an enhanced functional profile, as further shown by the pattern of IFN-γ, CD107a, TNF-α, and MIP-1β coexpression.

DISCUSSION

Accumulating evidence suggests that human NK cells are of pivotal importance in the host defense against viruses, including HIV (35). Whereas epidemiological and genetic data highlight the importance of NK cell contributions to viral control, to the best of our knowledge, direct NK cell correlates with delayed progression to AIDS have not been demonstrated yet.

Here, we performed a retrospective analysis of 117 untreated HIV-infected patients who had been followed longitudinally at the HIV outpatient clinic of MHH. We show that high numbers of CD8⁺ CD3⁻ lymphocytes in chronically infected, untreated HIV-infected patients were associated with delayed HIV disease progression, which we defined as a drop of CD4⁺ T cell counts below 200 cells/μl or ART initiation. Upon closer examination, we could show that the majority of these CD8⁺ CD3⁻ cells were well-described peripheral blood NK cells expressing either CD56 or CD16, or both. Notably, high numbers of bulk NK cells were also associated with a delayed onset of AIDS. We subtracted the absolute numbers and percentages of CD8⁺ CD3⁻ lymphocytes from the respective NK cell numbers, which provided an approximation of the number of CD8⁻ NK cells, with the caveat that the numbers of CD56/CD16 double-negative cells within the CD8⁺ CD3⁻ lymphocyte population varied. The resulting analysis indicated that CD8⁺ NK cells rather than CD8⁻ NK cells represent a correlate with slower disease progression. Nonetheless, further studies employing a consistent gating strategy to define CD8⁻ and CD8⁺ NK cells and their association with disease progression are warranted.

A drawback in our analysis is that the precise time point of first infection with HIV in all of our subjects remained elusive, thus introducing a potential bias mediated by a possible overrepresentation of long-term nonprogressors. In our longitudinal study cohort, we identified 7 elite controllers, who exhibited robust viral control throughout the study, with viral copy numbers being undetectable by conventional methods at most time points assessed (data not shown). None of these patients reached our endpoint (median follow-up time, 53.6 months). The estimated frequency of HIV-1-infected patients with such durable viral control is usually 1 in 300 (36), indicating a possible overrepresentation of elite controllers in our study cohort. Viral control can be mediated by a number of adaptive immune mechanisms, such as HIV-specific CD8⁺ T cells, in particular, those restricted by protective HIV alleles, including HLA-B*57 or HLA-B*27 (37, 38). In addition, the advent of single-cell antibody cloning techniques has led to the discovery of dozens of highly potent and broadly neutralizing antibodies from HIV-seropositive patients (39, 40). The potential impact of these cellular and humoral immune mechanisms and their protective effects in HIV infection could not be accounted for in our study. Thus, although we took various HIV disease parameters into account in our survival analyses, there are additional known factors which we were not able to address and which could confound our study.

Nonetheless, our survival data are corroborated by cross-sectional data derived from a subset of these patients and further samples from HIV-infected individuals on ART and uninfected controls. We have recently demonstrated an inverse correlation between the frequency of CCR7-expressing CD56^bright NK cells and HIV copy numbers in plasma and a positive correlation between the frequency of CCR7-expressing CD56^bright NK cells and CD4⁺ T cells counts (18). Here, we demonstrated that the frequency of CD8⁺ NK cells is also negatively correlated with the HIV load, similar to the findings for CCR7⁺ CD56^bright NK cells, and positively correlated with absolute and relative CD4⁺ T cell frequencies. We thus provide evidence for correlations between CD8⁺ NK cells and multiple HIV-related clinical disease parameters, emphasizing the potential usefulness of CD8⁺ NK cells as a novel NK cell-derived, inverse biomarker for HIV pathogenesis. Similar to the findings for CCR7-expressing CD56^bright NK cells, our cross-sectional as well as longitudinal results strongly suggested that the frequencies of CD8⁺ NK cells could be partially restored by suppressing viral loads by the initiation of ART. Furthermore, we found a modest correlation between CCR7⁺ CD56^bright NK cells and CD8⁺ NK cells (data not shown).

Our observation of CD8⁺ NK cell loss in chronic HIV-1 infection is a confirmation of earlier studies reporting a selective depletion of CD16⁺ CD8⁺ or CD56⁺ CD16⁺ CD8⁺ NK cells in HIV infection (41, 42). However, the mechanism of the selective loss of C8⁺ NK cells remains elusive. Although direct infection of CD16⁺ CD8⁺ NK cells and viral replication in this subset had been suggested (43), it represents an unlikely mechanism to fully explain the dramatic loss of CD8⁺ NK cells, since the expression of CD4 in conjunction with the HIV coreceptor CCR5 or CXCR4 on NK cells is rather rare (44). An alternative explanation could be that CD8⁺ NK cells are more susceptible to apoptosis. More so than other human peripheral blood lymphocyte subpopulations, CD8⁺ NK cells were shown to be particularly sensitive to irradiation of PBMCs, and their loss correlated with the dose of gamma irradiation (45). Furthermore, the ligation of CD8 on the surface of NK cell clones using either soluble MHC class I molecules or HLA-G was shown to induce apoptosis (46). Although this finding was challenged by data indicating that engagement of CD8⁺ on NK cells prevented apoptosis (9), the discrepancy could be explained by the fact that triggering of apoptosis was dependent on Fas, which was present on NK cell clones but absent on freshly isolated NK cells (9). Given that Fas is significantly upregulated on NK cells in chronic HIV infection (18, 34), ligation of CD8 followed by subsequent Fas-mediated apoptosis could represent a viable hypothesis for the selective loss of CD8⁺ NK cells.

Furthermore, our data emphasize that a better understanding of the precise role of CD8 on human NK cells is required. The surface expression of the homodimer CD8αα on NK cells was observed in rats (47) and pigs (48) but was reported to be absent in mice (49) and thus seems to vary across different mammalian species (50). In rhesus macaques, one of the best-studied nonhuman primate models, virtually all NK cells express CD8 (28, 51, 52). In contrast, in chimpanzees, which are phylogenetically substantially closer to humans, peripheral blood NK cells were reported to consist of CD8⁻ as well as CD8⁺ subsets (53). Chimpanzee CD8⁺ NK cells apparently expressed more NKp46 and exhibited a more functional profile, whereas CD8⁻ cells were described to be functionally anergic (53). However, a subsequent study indicated that CD8⁻ NK cells, as defined by Reeves et al., could be contaminated with myeloid dendritic cells, which in turn might account for the phenotypic differences between CD8⁺ and CD8⁻ cells (54). Thus, the overwhelming majority of peripheral blood NK cells in chimpanzees seem to express CD8, which is in contrast to the pattern of expression by human NK cells. This could be an indication that CD8⁻ and CD8⁺ NK cells assume different roles in chimpanzees than in humans. Several studies addressed the role of CD8⁺ NK cells during HIV-1 infection in chimpanzees and found no decline of these cells (53, 55, 56). One likely explanation is arguably the observation that the immune pathologies caused by HIV-1 infection in humans vary substantially from those caused by HIV-1 infection in chimpanzees, since HIV-1-infected chimpanzees do not progress to AIDS, although a seminal study demonstrated the development of AIDS in wild chimpanzees infected with naturally occurring chimpanzee simian immunodeficiency virus (SIV_cpz) (57).

Notably, in our phenotypic analyses we were unable to identify any striking differences between CD8⁺ and CD8⁻ NK cells when comparing the expression of various activating and inhibiting NK cell receptors one at a time. After performing Boolean gating and analyzing every coexpression permutation of four activating or three inhibiting receptors, we identified a difference in the frequency of NK cells expressing NKp30 in the absence of NKG2D, NKp46, and NKp80 and NK cells expressing NKG2A in the absence of KIR2DL2/DL3 and CD85j. Furthermore, we found a modest increase in the frequency of CD57-expressing cells among CD8⁺ NK cells. These findings could be an indication that 10-color polychromatic flow cytometry, as performed in this study, is insufficient for a rigorous delineation of the diversity of NK cell phenotypic states. A recent study employed 37-parameter time of flight mass cytometry to simultaneously assess the expression of 28 NK cell receptors on human PBMCs (58). Subsequent Boolean gating analysis enabled the authors to study NK cells with unprecedented depth and uncovered tens of thousands of distinct patterns of expression of NK cell receptors, indicating that the enormous diversity and heterogeneity of human NK cells had been vastly underappreciated (58). Thus, it remains a likely possibility that a more detailed in-depth analysis of NK cells at a single-cell level could uncover further differences in the patterns of coexpression of various NK cell receptors on CD8⁺ and CD8⁻ NK cells, which could help with elucidation of the functional role of CD8 on human NK cells.

Previous studies reported CD8⁺ NK cells to be more cytolytic than CD8⁻ NK cells (9, 59). In accordance with these studies, we identified the CD8⁺ subset to exhibit a more functional profile, which was evident when analyzing single functions as well as coexpression patterns of four functional markers. We were thus able to confirm previously published results and could further extend these observations by providing additional data on IFN-γ, TNF-α, and MIP-1β. Interestingly, NK cells derived from HIV-infected elite controllers carrying KIR3DL1*h/*y in conjunction with HLA-B*57 were shown to be more responsive to K562 cells and exhibited a more polyfunctional profile, very similar to the findings for CD8⁺ NK cells shown in our study (13). Similar to our data, Kamya et al. showed that the functionality of NK cells was impacted by the viral burden in HIV-1-infected patients, although viral loads were not a major determinant for the functional capacity of the NK cells (13). Thus, it will be of great interest to evaluate in future studies how host genetics shape the phenotypic and functional profiles of NK cells, in particular, with respect to CD8 expression.

In conclusion, we have presented evidence for a role of CD8⁺ NK cells as a novel innate immune correlate for delayed disease progression and as a robust inverse marker for HIV-induced immune disease. Future studies are warranted to address whether the high prevalence of these cells in HIV-infected patients with slower disease progression is merely an epiphenomenon or whether these cells can make an active contribution to viral control, as suggested by their increased functionality.