Treated HIV Infection Alters Phenotype but Not HIV-Specific Function of Peripheral Blood Natural Killer Cells

Abstract

Natural killer (NK) cells are the predominant antiviral cells of the innate immune system, and may play an important role in acquisition and disease progression of HIV. While untreated HIV infection is associated with distinct alterations in the peripheral blood NK cell repertoire, less is known about how NK phenotype is altered in the setting of long-term viral suppression with antiretroviral therapy (ART), as well as how NK memory can impact functional responses. As such, we sought to identify changes in NK cell phenotype and function using high-dimensional mass cytometry to simultaneously analyze both surface and functional marker expression of peripheral blood NK cells in a cohort of ART-suppressed, HIV+ patients and HIV- healthy controls. We found that the NK cell repertoire following IL-2 treatment was altered in individuals with treated HIV infection compared to healthy controls, with increased expression of markers including NKG2C and CD2, and decreased expression of CD244 and NKp30. Using co-culture assays with autologous, in vitro HIV-infected CD4 T cells, we identified a subset of NK cells with enhanced responsiveness to HIV-1-infected cells, but no differences in the magnitude of anti-HIV NK cell responses between the HIV+ and HIV- groups. In addition, by profiling of NK cell receptors on responding cells, we found similar phenotypes of HIV-responsive NK cell subsets in both groups. Lastly, we identified clusters of NK cells that are altered in individuals with treated HIV infection compared to healthy controls, but found that these clusters are distinct from those that respond to HIV in vitro. As such, we conclude that while chronic, treated HIV infection induces a reshaping of the IL-2-stimulated peripheral blood NK cell repertoire, it does so in a way that does not make the repertoire more HIV-specific.

Keywords: CyTOF; HIV-1; NK cells; innate immunity; memory.

FIGURE 1

Profiling of NK cell repertoire using CyTOF in both healthy, HIV– and ART-suppressed, HIV+ donors demonstrates alterations in NK cell surface receptor expression. (A) Multidimensional scaling (MDS) plot showing separation of NK cells from HIV+ (n = 10) and HIV– donors (n = 10). Only markers whose contributions are greater than 0.25 in both MDS1 and MDS2 are displayed in the marker loadings. (B) A generalized linear model with bootstrap resampling was used to find receptors predictive of either HIV+ (right), or HIV– (left) donor NK cells. Log-odds are logarithm of ratios of the probability that a cell belongs to either group. For each marker, the 95% confidence interval is represented by the line surrounding the point estimate; a larger absolute log-odds value of the parameter indicates that the marker is a stronger predictor. (C) Mean signal intensity (MSI) of NKG2C, CD2, NKp46, and PD-1 (top 4 predictors of the HIV+ group; top), and CD244, NKp30, DNAM-1, and NKG2A (top 4 predictors of the HIV– group; bottom). (D) UMAP visualization of all NK cells from the HIV+ and HIV– groups, colored by expression of CD56, CD16, NKG2C, and CD2 (top 2 predictors of the HIV+ group), and CD244 and NKp30 (top 2 predictors of the HIV– group). Scales show asinh-transformed channel values.

FIGURE 2

The magnitude of the HIV-specific response does not differ between HIV- and HIV+ donors. (A) Schematic of experimental set-up for NK co-culture assays. (B) Representative flow cytometry plots of CD107a, IFN-γ, MIP-1β, and TNF-α production (by frequency of positive cells), after 4 h co-culture without (top) or in the presence of HIV-infected autologous CD4⁺ T cells (bottom). (C) Summary data of CD107a, IFN-γ, MIP-1β and TNF-α production (by frequency of positive cells), after 4 h co-culture with HIV-infected autologous CD4⁺ T cells, in NK cells from HIV– (n = 10) and HIV+ (n = 10) donors. ns, not significant, by t-test.

FIGURE 3

Phenotype of NK cells responding to HIV are similar between HIV- and HIV+ donors, and include many of the markers that are altered in ART-treated HIV infection. (A) A generalized linear model with bootstrap resampling was used to identify NK cell receptors predictive of either functionally responding (right), or non-responding (left) NK cells, in HIV– (n = 10) and HIV+ (n = 10) donors. Log-odds are logarithm of ratios of the probability that a cell belongs to either group. For each marker, the 95% confidence interval is represented by the line surrounding the point estimate; a larger absolute log-odds value of the parameter indicates that the marker is a stronger predictor. (B) Gating on markers that predict functionally responding cells (NKp30^hi CD96^hi TIGIT^hi CD62L^- CD16^- NKp46^-), with a minimum log-odds threshold of 0.2 in both groups. (C) Comparison of functional activity in the gated population, compared to bulk NK cells. Summary data of percentage positive functionally responding (positive for any of CD107a, IFN-γ, MIP-1β, and TNF-α) in HIV– (n = 10) and HIV + groups (n = 10). ****p < 0.001, by paired t-test. (D) Comparison of the frequency of the gated population, as a percentage of total NK cells, between the HIV– (n = 10) and HIV+ (n = 10) groups. ns, not significant, by t-test.

FIGURE 4

The predominant changes in the NK cell repertoire of HIV+ individuals occur in NK cell compartments that do not respond to HIV in in vitro restimulation. (A) UMAP visualization of all NK cells in co-culture with autologous HIV-infected cells in HIV– (n = 10) and HIV+ (n = 10) donors, colored by metacluster identity generated by ConsensusClusterPlus metaclustering. (B) UMAP visualization of all NK cells from the HIV+ and HIV– groups, colored by expression of functional markers CD107a, IFN-γ, MIP-1β, and TNF-α. Scales show asinh-transformed channel values. (C) Heatmap of scaled mean expression of all markers profiled, for each cluster 1 to 10. The abundance of each cluster (% of total cells) is given on the right of the heatmap. Functional markers (CD107a, IFN-γ, MIP-1β, and TNF-α) are on the left. (D) Heatmap of the relative abundance of each cluster between the HIV- (left) and HIV+ (right) groups. Each individual column represents a single donor. The heat represents the proportion of each metacluster in each donor, with yellow showing over-representation and blue showing under-representation. These proportions were first scaled with an arcsine-square-root transformation and then z-score normalized in each cluster. Clusters with a statistically significant (p < 0.05) difference in abundance between HIV– and HIV+ groups are highlighted in green; adjusted p-values (FDR) are shown beside it.