IL-21 and IFNα therapy rescues terminally differentiated NK cells and limits SIV reservoir in ART-treated macaques

Abstract

Unlike HIV infection, which progresses to AIDS absent suppressive anti-retroviral therapy, nonpathogenic infections in natural hosts, such African green monkeys, are characterized by a lack of gut microbial translocation and robust secondary lymphoid natural killer cell responses resulting in an absence of chronic inflammation and limited SIV dissemination in lymph node B-cell follicles. Here we report, using the pathogenic model of antiretroviral therapy-treated, SIV-infected rhesus macaques that sequential interleukin-21 and interferon alpha therapy generate terminally differentiated blood natural killer cells (NKG2a/c^lowCD16⁺) with potent human leukocyte antigen-E-restricted activity in response to SIV envelope peptides. This is in contrast to control macaques, where less differentiated, interferon gamma-producing natural killer cells predominate. The frequency and activity of terminally differentiated NKG2a/c^lowCD16⁺ natural killer cells correlates with a reduction of replication-competent SIV in lymph node during antiretroviral therapy and time to viral rebound following analytical treatment interruption. These data demonstrate that African green monkey-like natural killer cell differentiation profiles can be rescued in rhesus macaques to promote viral clearance in tissues.

Fig. 1. IL-21 and rIFNα immunotherapies are biologically active in SIV-infected, ART-treated RMs.

a Cartoon schematic of study design as detailed in the Results and Methods. Plasma viral loads (SIV-RNA copies/mL) were longitudinally measured by qRT-PCR with individual b cytokine-treated (ART + IL-21 + IFNα; blue; n = 9 RMs) and c control (ART-only; black; n = 5 RMs) RMs represented by distinct shapes. The dashed horizontal line represents the assay’s limit of detection (30 copies/mL) with undetectable events plotted as 15 copies/mL. d The frequency of immune activation (HLA-DR⁺CD38⁺) was quantified in memory (CD95⁺) CD4⁺ T-cells in PBMCs and e representative flow cytometry plots are given at critical time points (as indicated above) for a cytokine-treated (RQm14) and control RMs (RSp14; 16 of 260 single-replicate stains). f The expression of IFN-stimulated genes (ISGs) was measured by RNA-seq in PBMCs at 2 h post-rIFNα and was calculated as a log₂ fold-change between rIFNα-treated (n = 6) and control RMs (n = 5), which is represented as a double-gradient heatmap. The size of each data point corresponds inversely to the log₁₀-transformed nominal p value with significant (p < 0.05) adjusted p values indicated by a black border. Using DESeq2, data were analyzed with a two-sided (95% CI) Wald test using the Benjamini–Hochberg method for multiple comparisons. By flow cytometry, the frequency of g HLA-E⁺ CD4⁺ T-cells and h NKG2a/c⁺ CD8⁺ T-cells was quantified of CD45⁺ lymphocytes in PBMCs. i NK cells isolated from PBMCs were cultured alone or the presence of K562-HLA-E^*0101 cells either unloaded or loaded with SIVmac_239/251 Env peptides. The SIV-Env-specific, HLA-E-restricted activity was calculated as the log₁₀-transformed percent yield of CD107a surface expression on NK cells by flow cytometry and the horizontal dashed line represents 100% activity. b–d, g–i Treatment phases are indicated with the following background shading: IL-21 (orange), rIFNα (red), and ART (gray). d, g–i Data from individual RMs (staggered open circles) are overlaid against the mean (solid line) ± SEM (shaded region within the dashed lines): control (black; n = 5) and cytokine-treated (blue; n = 8). Data were analyzed with a d, g–i two-sided (95% CI), two-way ANOVA with Bonferroni’s correction for multiple comparisons with cross-sectional comparisons relative to controls.

Fig. 2. Cytokine therapy reduces replication competent virus in lymphoid tissue, which is uniquely correlated with Env-specific NK cell activity.

The content of cell-associated SIV-RNA (log₁₀ copies per 10⁶ cells) was determined by qRT-PCR in snap frozen pellets of a PBMCs, b LN, and c RB prior to ART initiation (d35 p.i.), following two IL-21 cycles (d217 p.i.), and after one subsequent rIFNα cycle amid ongoing ART (d339 p.i. or d374 p.i.), as was cell-associated SIV-DNA (log₁₀ copies per 10⁶ cells) in d PBMCs, e LN, and f RB. a–f Data from individual RMs are overlaid against the mean ± SEM (in red): control (ART-only, black square; n = 5 RMs) and cytokine-treated (ART + IL-21 + IFNα, blue circle; n = 8 RMs). g From cryo-preserved, magnetically isolated LN CD4⁺ cells, the number of infectious units per million (IUPM) cells were measured with a limiting dilution quantitative viral outgrowth assay (QVOA; 3 serial dilutions plated in triplicate) in control (n = 4) and cytokine-treated RMs (n = 6). a–g Treatment phases are indicated with the following background shading: IL-21 (orange), rIFNα (red), and ART (gray). Data were analyzed with two-sided (95%), two-way ANOVA with Bonferroni’s correction with comparisons relative to controls and time. h The SIV reservoir contents in tissue (as indicated below) were correlated against phenotypes of activation (HLA-DR⁺CD38⁺) and proliferation (Ki-67⁺) in CD4⁺ and CD8⁺ T-cells; the frequency of HLA-E⁺ CD4⁺ T-cells and NKG2a/c⁺ CD8⁺ T-cells; and the Env-specific NK cell activity (as indicated at left) in all RMs (n = 13; days 35, 217, and 339/374 p.i. as matched data are available). Per each correlation the two-tailed (95% CI) Spearman’s rank correlation coefficient (rho) is represented as a double-gradient heatmap and the size of each data point corresponds inversely to the log₁₀-transformed Spearman’s p value. The false discovery rates (FDR) were calculated using SAS and significant values (Q < 0.05) are represented by a black border.

Fig. 3. Cytokine therapy promotes the maturation of NKG2a/c^lowCD16⁺ NK cells with enhanced ex vivo innate activity, which correlates with the content of lymphoid replication competent virus.

a The frequency of NK cells (CD45⁺CD20⁻CD3⁻NKG2a/c⁺) of PBMC CD45⁺ lymphocytes was longitudinally measured by flow cytometry, b as was their CXCR5 expression. The frequency of each differentiation stage of NK cells was determined based on the following definition: c Stage 0 (red, NKG2a/c^lowCD16⁻), d Stage 1 (blue, NKG2a/c^highCD16⁻), e Stage 2 (orange, NKG2a/c^highCD16⁺), and f Stage 3 (purple, NKG2a/c^lowCD16⁺). g The mean frequency of each NK cell differentiation stage from above was also re-visualized as a color-coded (as annotated below), parts-of-whole stacked bar plot for the cytokine-treated (n = 8; at left) and control RMs (n = 5; at right) over time (indicated at left). Flow cytometry was used to quantify the ex vivo innate frequency of h IFN-γ⁺ (intracellular) NK cells and i CD107a⁺ (surface) stage 3 (NKG2a/c^lowCD16⁺) NK cells. a–i Data from individual RMs (staggered open circles) are overlaid against the mean (solid line) ± SEM (shaded region within the dashed lines): control (ART-only, black; n = 5) and cytokine-treated (ART + IL-21 + IFNα, blue; n = 8). Treatment phases are indicated with the following background shading: IL-21 (orange), rIFNα (red), and ART (gray). Data were analyzed with two-sided (95% CI), two-way ANOVA with Bonferroni’s correction with cross-sectional comparisons relative to controls. j The frequencies of each differentiation stage of NK cells in PBMCs (as indicated below) were correlated against levels of cell-associated and replication competent SIV content in tissue, and the ex vivo innate and Env-specific NK cell activities in PBMCs (as indicated at left) in all RMs (n = 13; days 35, 77, 217, and 374 p.i. as matched data are available). Per each correlation the two-tailed (95% CI) Spearman’s rank correlation coefficient (rho) is represented as a double-gradient heatmap and the size of each data point corresponds inversely to the log₁₀-transformed Spearman’s p value. The false discovery rates (FDR) were calculated using SAS and significant values (Q < 0.05) are represented by a black border.

Fig. 4. Cytokine therapy delays the rebound of plasma viremia following ATI.

Following ART analytical treatment interruption (ATI), the plasma SIV-RNA copies/mL were measured by qRT-PCR in each a cytokine-treated (PEG-IFNα with prior ART + IL-21 + rIFNα, blue; n = 8 RMs) and b control RMs (prior ART-only, black; n = 5 RMs). The horizontal dashed line (200 copies/mL) represents the threshold for virologic rebound and PEG-IFNα treatments are indicated by the purple arrows above. These kinetics of plasma viremia following ATI were then re-visualized as follows: c The delay in rebound of plasma viremia was represented as a treatment-stratified survival curve, which was analyzed with a Log-rank Mantel–Cox test. d The delay in viral rebound, in days, per each RM was represented as a color and shape-coded symbol overlaid against the mean ± SEM (red), which was analyzed with a two-sided (95% CI) Mann–Whitney U test. e The log₁₀ SIV-RNA copies/mL are given as a longitudinal mean (solid line with closed circles) ± SEM (color-coded shaded region within the dashed lines), which was analyzed with a two-sided (95% CI), two-way ANOVA with Bonferroni’s correction for multiple comparisons across treatments (d13 p = 0.0001), and a mixed-effects linear model was used to analyze the slope between d13 and d20 post ATI (as indicated by the bracket; p = 0.08).

Fig. 5. The formation and activity of NKG2a^lowCD16⁺ NK cells correlate with viral recrudescence following ATI.

a From PBMCs taken 24 h following the first (d6 post ATI) and the fifth PEG-IFNα dose (d37 post ATI), the expression of interferon-stimulated genes was calculated as a cross-sectional log₂ fold-change between cytokine-treated (n = 8) and control (n = 5) RMs, which is represented as a double-gradient heatmap. The size of each data point corresponds inversely to the log₁₀-transformed nominal p value with significant (p < 0.05) adjusted p values indicated by a black border. Using DESeq2, data were analyzed with a two-sided (95% CI) Wald test using the Benjamini–Hochberg method for multiple comparisons. b In PBMCs at d13 post ATI, the distribution of the differentiation subsets was measured by flow cytometry as a frequency of NK cells: Stage 0 (red, NKG2a/c^lowCD16^–), Stage 1 (blue, NKG2a/c^highCD16^–), Stage 2 (orange, NKG2a/c^highCD16⁺), and Stage 3 (purple, NKG2a/c^lowCD16⁺). NK cells isolated from PBMCs at d13 post ATI were utilized to determine c the frequency of ex vivo innate activity (CD107a surface expression) or d the Env-specific activity upon co-culture with K562 cells expressing HLA-E loaded with SIVmac Env peptides. b–d Data from individual cytokine-treated (PEG-IFNα with prior ART + IL-21 + rIFNα, blue; n = 8) and control (prior ART-only, black; n = 5) RMs (open symbols) are overlaid against the mean ± SEM (in red) and were analyzed with b, c two-sided (95% CI), two-way ANOVA with Bonferroni’s correction for cross-sectional comparisons relative to controls or d with a two-sided (95% CI) Mann–Whitney U test. e The delay in the rebound of plasma viremia was correlated against measures of SIV content, NK cell differentiation, and NK cell activity (as indicated at left; n = 13) from the final on-ART measurement (d339–374 p.i.) or during rebound following ATI (d6–13 post ATI). Per each correlation the two-tailed (95% CI) Spearman’s rank correlation coefficient (rho) is represented as a double-gradient heatmap and the size of each data point corresponds inversely to the log₁₀-transformed Spearman’s p value. The false discovery rates (FDR) were calculated using SAS and significant values (Q < 0.05) are represented by a black border. Correlations for which data did not exist during that experimental phase are indicated as “NA”.