We previously reported that pegylated IFN-α2a (Peg–IFN-α2a) added to antiretroviral therapy (ART)–suppressed, HIV-infected subjects resulted in plasma HIV control and integrated HIV DNA decrease. We now evaluated whether innate NK cell activity or PBMC transcriptional profiles were associated with decreases in HIV measures. Human peripheral blood was analyzed prior to Peg–IFN-α2a administration (ART, baseline), after 5 wk of ART+Peg–IFN-α2a, and after 12 wk of Peg–IFN-α2a monotherapy (primary endpoint). After 5 wk of ART+Peg–IFN-α2a, immune subset frequencies were preserved, and induction of IFN-stimulated genes was noted in all subjects except for a subset in which the lack of IFN-stimulated gene induction was associated with increased expression of microRNAs. Viral control during Peg–IFN-α2a monotherapy was associated with 1) higher levels of NK cell activity and IFN-γ–induced protein 10 (IP-10) on ART (preimmunotherapy) and 2) downmodulation of NK cell KIR2DL1 and KIR2DL2/DL3 expression, transcriptional enrichment of expression of genes associated with NK cells in HIV controller subjects, and higher ex vivo IFN-α–induced NK cytotoxicity after 5 wk of ART+Peg–IFN-α2a. Integrated HIV DNA decline after immunotherapy was also associated with gene expression patterns indicative of cell-mediated activation and NK cytotoxicity. Overall, an increase in innate activity and NK cell cytotoxicity were identified as correlates of Peg–IFN-α2a–mediated HIV control.
Interferon-α is a type I IFN produced by leukocytes as part of the host’s TLR-mediated antiviral response (1). Type I IFNs modulate cellular antiviral immune responses in vivo either directly by activation of antiviral host restriction factors (2) or indirectly via stimulation of innate NK cell–mediated responses (3–6). Clinical trials with IFN-α support modulation of cell-mediated responses as an outcome of immunotherapy, leading to increased perforin expression in NK and CD8+ T cells (7, 8), increases in CD16+CD56+ NK cell numbers (9), and activation of CD56+ NK cells (10). Activation of innate host restriction factors and NK responses have been associated with control of HIV and lysis of autologous HIV-infected CD4+ T cell targets ex vivo (11, 12), suggesting that IFN-α immunotherapy may activate similar mechanisms in vivo to control HIV infection.
Several human clinical trials in which IFN-α immunotherapy was administered without antiretroviral therapy (ART) in HIV-infected viremic individuals support a predominantly anti-HIV effect without advancement of disease progression (13–23). In the absence of ART, temporal reductions of HIV plasma viral load following administration of pegylated IFN-α2a (Peg–IFN-α2a) were also associated with activation of baseline levels of host restriction factors, indicating a role for IFN-induced gene induction as part of the anti-HIV mechanism of action (13). Short-term treatment with Peg–IFN-α2a added to ART in acute infection also led to a decrease in HIV reservoirs (24).
We and others have tested the potential for Peg–IFN-α2a immunotherapy to reduce the size of the HIV reservoir (measured as the integrated HIV DNA levels) in chronic ART-suppressed HIV or hepatitis C virus (HCV)/HIV–infected subjects (25–27). Together, these studies suggest that IFN-α can suppress plasma viral load and act to decrease CD4+ T cell integrated HIV DNA levels, as reported in our NCT00594880 clinical study (25). However, the mechanisms underlying the in vivo responses described for our NCT00594880 clinical study remain undefined. We now describe innate activity, NK cytotoxicity, and gene expression as correlates of retained plasma viral load suppression and CD4+ T cell integrated HIV DNA reduction after ART interruption and Peg–IFN-α2a monotherapy.
Materials and Methods
Fresh whole blood samples were obtained from 20 HIV-infected subjects on suppressive ART (with >6 mo at ≥400 CD4+ T cells/mm3 [nadir ≥ 200 cells/mm3] and undetectable HIV viral load measurement of <50 copies/ml at screening) participating in an open-label longitudinal study (NCT00594880) investigating the antiviral activity of 90 or 180 μg/wk Peg–IFN-α2a when administered with ART for 5 wk, followed by ART interruption and Peg–IFN-α2a monotherapy for 12 wk as primary endpoint. The details and clinical findings of this study have been published elsewhere (25).
Briefly, and as described in our prior publication (25), 9 of 20 subjects maintained plasma HIV viral load <400 copies/ml by the 12-wk study primary endpoint (primary endpoint responders [R]). Of the remaining 11/20 subjects, seven exhibited virologic failure (viral load >400 copies/ml) prior to the 12-wk study primary endpoint, and four discontinued the study because of moderate depression scores (n = 3) and grade 3 neutropenia (n = 1); according to the study design, all of these subjects were considered to have achieved primary endpoint failure (primary endpoint nonresponders [N]). In addition, and as described in our prior publication (25), responders at primary endpoint had a significant drop in the levels of integrated HIV DNA in PBMC, as detected by Alu-Gag PCR between baseline (ART) and primary endpoint (Peg–IFN-α2a). Levels of integrated viral DNA did not change in nonresponders.
In the current study, results are described for the 20 subjects participating in NCT00594880 over three time points. All available peripheral blood samples were used as obtained prior to Peg–IFN-α2a initiation (baseline, ART only), after 5 wk of ART+Peg–IFN-α2a, and after 12 wk of Peg–IFN-α2a monotherapy (primary endpoint).
Written informed consent was obtained from patients according to the directives of the institutional review boards at the Wistar Institute, University of Pennsylvania, Philadelphia FIGHT, and Drexel University. All study participants were adults.
Fresh blood samples were used for clinical assessment, immune subset characterization by flow cytometry, and for plasma and PBMC isolation. PBMC were isolated as previously described using a standard Ficoll-hypaque density gradient centrifugation (28). Plasma was cryopreserved and used for cytokine assessment. All available PBMC were either used fresh for ex vivo assessment of constitutive and/or IFN-α–induced innate functionality (flow cytometry–based assays for STAT-1 phosphorylation and CD107a expression and a standard 51Cr release assay) or after cryopreservation for host gene expression studies in PBMC.
Phenotypic characterization of innate and adaptive cell subsets
Immunophenotypic characterization of NK cells, dendritic cells (DC), and T cell subsets was performed by same-day whole blood five-color staining as previously described (28, 29) using the following combinations of directly fluorochrome-conjugated anti-human cell-surface mAbs: 1) CD56–FITC, CD3-PerCP-Cy5.5, CD94–allophycocyanin, and CD16-allophycocyanin-Cy7; 2) CD56-FITC, CD25-PE, CD3-PerCP-Cy5.5, HLA-DR-allophycocyanin, and CD16-allophycocyanin-Cy7; 3) CD158a (killer cell Ig-like receptor, two Ig domains and long cytoplasmic tail 1 [KIR2DL1]–FITC, CD158b (KIR2DL, two domains, long cytoplasmic tail 2/3 [KIR2DL2/DL3]–PE, CD3-PerCP-Cy5.5, CD94-allophycocyanin, and CD56-allophycocyanin-Cy7; 4) CD1c (blood DC Ag 1 [BDCA1]–FITC, CD303 (BDCA2)–PE, CD304 (BDCA4)–PE, CD11c-PerCP-Cy5.5, CD197 (CCR type 7 [CCR7])–allophycocyanin, and CD19-allophycocyanin-Cy7; 5) CD95-FITC, CD25-PE, CD3-PerCP-Cy5.5, CD38-allophycocyanin, and CD4-allophycocyanin-Cy7; and 6) CD3-PerCP-Cy5.5, HLA-DR-allophycocyanin, and CD4-allophycocyanin-Cy7. Isotypes used included IgG1k-FITC, IgG1k-PE, IgG1k-PerCP-Cy5.5, IgG1k-allophycocyanin, and IgG1k-allophycocyanin-Cy7. All Abs were from Becton Dickinson (BD) Biosciences (San Diego, CA) except BDCA1-FITC and BDCA2-PE, which were from Miltenyi Biotec (San Diego, CA). Stainings 1–3 allowed for the assessment of markers of activation/inhibition (HLA-DR, CD94, CD25, and KIRs) on NK cell subsets (30, 31), identified as the following: 1) CD3−CD56bright and CD3−CD56dim with or without CD16, and 2) CD3−CD56−CD16+, CD3−CD56+CD16+, and CD3−CD56+CD16−. Staining 4 allowed for the assessment of maturation markers (CCR7) on DC subsets (32), identified as BDCA2+BDCA4+ (plasmacytoid DC [PDC]) and CD19−BDCA1+CD11c+ (myeloid DC [MDC]). Finally, stainings 5 and 6 allowed for the assessment of markers of activation (i.e., CD38, CD95, CD25, and HLA-DR) on T cells (CD3+CD4+ and CD3+CD4−). Cells were analyzed on an LSRII cytometer (BD Biosciences) by collecting >200,000 events, and data were analyzed using FlowJo software (Version 8.8.4; Tree Star, Ashland, OR). Gating was originally done on singlets and then on “live lymphocyte” (for NK and T cells) or “all live cell” (for DC) gates defined by size and granularity in forward scatter and side scatter. Thresholds were set by isotype-matched negative controls and unstained cells. Results were expressed as mean fluorescence intensity (MFI), percentages, and cells/mm3.
Assessment of the in vitro role of IFN-α on STAT-1 phosphorylation within PBMC subsets
Freshly isolated PBMC (2 × 106/ml) were stained for 1) CD3-FITC, CD14-FITC, CD19-allophycocyanin, CD20-allophycocyanin, CD16–Pacific Blue, and CD56-PECy7; 2) CD14-FITC, BDCA2-allophycocyanin, BDCA4-allophycocyanin, and CD3–Pacific Blue; or 3) corresponding isotypes (IgG1k-FITC, IgG2ak-FITC, IgG1-allophycocyanin, IgG1k–Pacific Blue, and IgG1k-PECy7) for 30 min at 4°C, washed with 1× PBS at 1500 rpm for 5 min, and resuspended in warm 1× PBS. PBMC were then treated for 10 min at 37°C with medium, IFN-α (5000 U/ml; PBL, Piscataway, NJ), or IFN-γ (10 ng/ml; R&D Systems, Minneapolis, MN). Cells were then fixed with paraformaldehyde (5% final concentration) for 10 min at 37°C, washed, and permeabilized with PhosFlow buffer (BD Biosciences) for 30 min at room temperature (RT). PBMC were then washed in FACS washing buffer at 2200 rpm for 10 min, stained with an Ab against p–STAT-1 (p–STAT-1–PE) or corresponding isotype IgG2ak-PE for 1 h at RT, washed with FACS washing buffer, and analyzed in the CyAn cytometer as described above. Staining 1 allowed for the assessment of NK cell subsets (identified as Lin3−CD56+CD16+, Lin3−CD56+CD16−, or Lin3−CD56−CD16+, with Lin3 consisting of CD3, CD14, CD19, and CD20), whereas staining 2 allowed for the identification of monocytes (identified as CD3−CD14+) and PDC (identified as CD3−CD14−BDCA2+BDCA4+). STAT-1 phosphorylation was expressed as MFI of p–STAT-1 (constitutive or in vitro IFN-α– or IFN-γ–induced defined as in vitro IFN-α– or IFN-γ–induced MFI of p–STAT-1/constitutive MFI of p–STAT-1) for all the above described cell subsets. All Abs were from BD Biosciences except BDCA2-allophycocyanin, BDCA4-allophycocyanin, and IgG1-allophycocyanin, which were purchased from Miltenyi Biotec.
Assessment of constitutive CD107a expression on NK cells
Fresh PBMC (1 × 106) were incubated with CD107a-PE or corresponding isotype (IgG1k-PE) for 4 h at 37°C, blocked with 10% serum for 10 min at RT, and stained for 30 min on ice with CD3-FITC, CD14-FITC, CD19-FITC, CD20-FITC, CD56-PE-Cy7, and CD16–Pacific Blue, or corresponding isotypes IgG1k-FITC, IgG2ak-FITC, IgG1k-PE-Cy7, and IgG1k–Pacific Blue. All Abs were from BD Biosciences. PBMC were washed with 2 ml of FACS washing buffer at 1500 rpm at 4°C for 5 min, lysed with 1 ml of BD FACS Lyse for 10 min at 37°C, washed again with 2 ml of FACS washing buffer, and resuspended in 100 μl of FACS washing buffer. Cells were analyzed on a CyAn cytometer as described above. This staining allowed for the assessment of markers of CD107a (marker of degranulation) and NKp46 (activating receptor) on NK cell subsets (identified as Lin3−CD56bright and Lin3−CD56dim with or without CD16, with Lin3 consisting of CD3, CD14, CD19, and CD20).
51Cr release assay for NK cytotoxicity
NK function was assessed measuring constitutive and IFN-α (5000 U/ml)–induced NK cell–mediated cytotoxicity by standard 51Cr release assay, as previously described, using fresh PBMC preparations as effector cells against the tumor-derived erythroblastoid cell line K562 (29, 33). Briefly, viable fresh PBMC preparations from whole blood (effector cells) were treated for 18 h at 37°C with medium alone or IFN-α (5000 U/ml; PBL). Erythroblastoid MHC-null K562 cells, which served as targets, were labeled with Na251CrO4 (∼50 μCi) for 1.30 h at 37°C, washed, and resuspended at a concentration of 1 × 105 cells/ml in culture medium. Effectors and labeled K562 targets were cultured in triplicate to yield the desired E:T ratios in 0.2-ml volume (usually 50:1, 25:1, 12.5:1, and 6.25:1) in round-bottom 96-well plates and incubated for 4 h. Percent lysis was determined by the following formula: [(experimental counts − spontaneous released counts)/(total counts − spontaneous released counts)] × 100. Results were expressed as area under the curve (AUC) for E:T ratios of 50:1, 25:1, 12.5:1, and 6.25:1 for both constitutive and in vitro IFN-α–induced NK function.
Cytokine assessment in plasma
Cryopreserved plasma was used for cytokine profile definition by using the Human Cytokine 30-Plex Panel (Invitrogen Multiplex Bead Immunoassay Kit) with the Luminex 100 or 200 dual-laser detection system. This panel allowed for the quantitative determination of epidermal growth factor (EGF), eotaxin, basic fibroblast growth factor, G-CSF, GM-CSF, hepatocyte growth factor (HGF), IFN-α, IFN-γ, IL-1Ra, IL-1β, IL-2, IL-2R, IL-4, IL-5, IL-6, IL-7, IL-8, IL-10, IL-12 (p40/p70), IL-13, IL-15, IL-17, IP-10, MCP-1, monokine induced by IFN-γ (MIG), macrophage inflammatory protein 1α (MIP-1α), MIP-1β, RANTES, TNF-α, and vascular endothelial growth factor (VEGF).
Isolation of total RNA and DNA from cryopreserved PBMC was performed using Trireagent (Sigma-Aldrich, St. Louis, MO) according to the manufacturer’s instructions. For the gene expression microarrays, amplified cRNA was generated from 100 ng of RNA using the TargetAmp Nano-g Biotin-aRNA Labeling Kit (Epicentre, Madison WI) and then hybridized to the HumanHT-12 v4 Expression BeadChips (Illumina, San Diego, CA). An additional 200 ng of RNA was used for microRNA (miRNA) assays using TaqMan OpenArray Human miRNA Panels (Life Technologies, Grand Island, NY) with Megaplex Reverse Transcription Primers and PreAmp Human Pools Set v3.0 (Life Technologies); 100 ng was used for each of the two primers, A and B.
Raw gene expression microarray data were quantile-normalized and log2-scaled. Noninformative probes, either expressed at background level or showing <1.2-fold change between all samples pairs, were removed prior to further analysis. For miRNA data preprocessing, cycle threshold (Ct) values were converted to expression levels. The small nucleolar RNAs (RNU) RNU44 and RNU48 were used as endogenous controls (housekeeping genes) to normalize the expression levels of the samples and compute relative amounts for each miRNA (ΔCt). First, the average Ct of the RNUs (RNUavg) was calculated. Ct values were then restricted to <24, as suggested by the manufacturer, and the maximum ΔCt value was designated as ΔCt24 (where ΔCt24 = 24 − RNUavg). The ΔCt values were then converted to absolute expression levels by calculating 2ΔCt24 – ΔCt. ΔCt values exceeding ΔCt24 were considered unreliable and were floored to the ΔCt24 value for the comparative analyses. miRNAs with ΔCt values at the ΔCt24 level across all samples were filtered out. Expression levels were log2-scaled for further analysis.
Microarray data analysis
Expression level comparisons between ART and 5 wk of ART+Peg–IFN-α2a were done using paired t tests with correction for multiple testing done according to Storey et al. (34). Genes that passed a false discovery rate (FDR) < 5% were called significant and used for hierarchical clustering of samples with standardized Euclidean distance and average linkage. The significance of the difference of changes due to 5 wk of combined administration of ART+Peg–IFN-α2a between R and N was calculated using unpaired t tests, and genes that passed FDR < 5% and miRNAs with p < 0.05 were considered. Ingenuity Pathway Analysis (IPA, www.qiagen.com/ingenuity; QIAGEN, Redwood City, CA) was used for information linking miRNAs and their targets where 1) information was experimentally confirmed and 2) information was predicted by a TargetScan algorithm with high/moderate confidence.
Genes that were significantly detected above the microarray background were annotated using the Interferome database (35), and those known to be stimulated at least 10-fold by IFN-α were considered as IFN-stimulated genes (ISGs). The top 30 of those ISGs were then used to illustrate differences in expression among ISG responders/primary endpoint responders (RR), ISG responders/primary endpoint nonresponders (RN), and ISG nonresponders/primary endpoint nonresponders (NN).
Unpaired t tests were used for the assessment of a gene signature on ART that can distinguish R from N. Principal component analysis was then performed using expression of genes that passed a p < 0.001 threshold.
Association of gene expression changes after 5 wk of ART+Peg–IFN-α2a (n = 11) and by primary endpoint (Peg–IFN-α2a, n = 6) with integrated HIV DNA changes was tested using Pearson correlation, and genes with p < 0.05 were considered for enrichment analysis. Enrichment analysis was done using IPA, and significantly enriched biological functions that passed a p < 0.001 threshold were considered. Functions with predicted activation (z-score > 1 calculated by IPA based on the direction of correlation of member genes) were then manually categorized into major classes with maximum z-score assigned to each of the classes. Expression of the genes involved in the functional categories was shown on an expression heatmap, with shared genes that belong to multiple categories illustrated within a category with the fewest genes.
In addition, gene set enrichment analysis (GSEA) was also performed on the genes ranked by the significance and direction of change between RR and NN and baseline expression of R and N, as well as correlation of week 5 and week 8 changes with HIV DNA changes. Results that passed FDR < 15% were considered significant.
Analysis of the microarray data was done using MATLAB 7.10.0. The data were submitted to the Gene Expression Omnibus (https://www.ncbi.nlm.nih.gov/geo/) under accession number GSE107549.
Phenotypic and functional data are reported as means with SD. The effect of 5 wk of ART+Peg–IFN-α2a on study variables was assessed by paired t tests, and two-sided p values <0.05 were considered statistically significant. Comparisons between groups were performed by nonpaired t tests with a cutoff of p <0.05. Analysis was done by using R.2.5.1.
Study schema and patient groups
Study schema, time points used for analysis, and response group definitions used in the current study are shown in Fig. 1. The number of patients used in the current study for analysis of the different sets of data (immune variables, integrated DNA, and gene expression) was dependent on sample availability and is shown along with the different response groups used in the current study in Supplemental Table I.
Innate cell subsets and inflammatory protein changes after 5 wk of dual treatment with Peg–IFN-α2a and ART
Consistent with the expected leukopenic effects of Peg–IFN-α2a immunotherapy, 5 wk of ART+Peg–IFN-α2a resulted in reductions in whole blood cell count, neutrophil count, and CD4+ T cell count (p = 0.0012, p = 0.0074, and p = 0.0009, respectively; Fig. 2A–C). The reduction in CD4+ T cell count did not reflect a selective loss of CD4+ T cell percentage within PBMC (Fig. 2D). As with CD4+ T cell frequencies, no change was detected in T cell activation or in the major NK or DC subsets (Fig. 2E–I). However, a clear reduction in STAT-1 phosphorylation was noted in all leukocytes (NK, DC, and monocytes) after immunotherapy when challenged ex vivo with exogenous IFN-α (Supplemental Fig. 1A, 1B), in contrast to a retention of IFN-γ–induced STAT-1 phosphorylation observed in monocytes and PDC (Supplemental Fig. 1B). With regard to soluble plasma cytokine changes, we detected an increase in plasma levels of IL-8 (p = 0.0093) and MCP-1 (p = 0.0005) after 5 wk of ART+Peg–IFN-α2a (Fig. 2J, 2K). None of the significant changes detected in clinical parameters, leukocyte cell subset distribution, or ex vivo IFN-α–induced STAT-1 phosphorylation after 5 wk of ART+Peg–IFN-α2a were associated with HIV plasma viral load control or changes in integrated proviral DNA, as measured between R and N (data not shown).
Gene expression analysis identifies gene signatures that inform clinical outcomes
Gene expression (mRNA and miRNAs) was analyzed in PBMC samples obtained on ART and after 5 wk of ART+Peg–IFN-α2a. A total of 1436 probes were found to be significantly differentially expressed between these two time points (FDR < 5%). Hierarchical clustering of the samples based on these 1436 probes separated them into two main arms based on the presence or absence of immunotherapy (i.e., ART versus 5 wk of ART+Peg–IFN-α2a; Fig. 3A). Overall, there was no separation between R and N with regard to overall gene expression within ART or after 5 wk of ART+Peg–IFN-α2a, although in four subjects (008, 009, 013, and 029, indicated with asterisks in Fig. 3A), the samples after 5 wk of ART+Peg–IFN-α2a clustered with their corresponding samples on ART, suggesting that these four subjects had no detectable gene modulation after ART+Peg–IFN-α2a treatment.
To further investigate the lack of response observed in these four subjects, the average expression fold change after 5 wk of ART+Peg–IFN-α2a treatment (i.e., ART+Peg–IFN-α2a ISG expression/ART ISG expression) was calculated for the top 30 known IFN-α–stimulated genes [shown in other studies to be stimulated at least 10-fold, as shown in the Interferome database (35)].The following groups were defined based on ISG expression change and primary endpoint outcome: 1) subjects who showed a response (as defined by modulation of ISG) to 5 wk of ART+Peg–IFN-α2a treatment and a response at primary endpoint (RR), 2) subjects who showed a response (as defined by modulation of ISG) to 5 wk of ART+Peg–IFN-α-2a treatment but did not show a response at primary endpoint (RN), and 3) subjects who did not show a response (as defined by modulation of ISGs) to 5 wk of ART+Peg–IFN-α-2a treatment and did not show a response at primary endpoint (NN). The results clearly showed that induction of ISG expression alone was not able to segregate persons that controlled viral load under monotherapy at primary endpoint, yet the lack of ISG induction was only observed within persons failing to control plasma viral load at primary endpoint (Fig. 3B).
Apart from the 30-gene ISG induction criteria, a comparison of the changes in the expression levels for genes and miRNAs after 5 wk of ART+Peg–IFN-α2a treatment between the RR and RN or between the RR and NN groups was performed. No significant differences in the response were found between the RR and RN groups (FDR > 95% for all genes, only five genes with nominal p < 0.001). In contrast, comparison between the RR and NN groups identified 111 gene probes with significant differences (FDR < 5%; Supplemental Table IIA) with the majority (85 probes corresponding to 77 unique genes) being upregulated after 5 wk of ART+Peg–IFN-α2a treatment in the RR group (Fig. 3C). Of the 77 genes found to be significantly upregulated after 5 wk of ART+Peg–IFN-α2a treatment in the RR group when compared with the NN group, 46 genes (60%) were previously described to be stimulated at least 10-fold by IFN-α, based on information from the Interferome database (35), representing an enrichment of 34-fold over the total prevalence of such genes (1.7%) among all the genes detected by microarrays.
In addition, 12 significantly differentially changed miRNAs (nominal p < 0.05) were found to be more upregulated in the NN group, with nine of them having prior evidence of targeting at least one gene from the upregulated list in the RR group (noted below). A combined heatmap for the 77 genes and the 12 miRNAs is shown in Fig. 3C, with all predicted or experimentally confirmed target genes for the miRNAs highlighted. Among the identified matches between the expression of a target gene and the putative regulation by the corresponding miRNA, high-confidence matches were noted for IFN-induced protein 44 like (IFI44L) (36), Ral guanine nucleotide dissociation stimulator like 1 (RGL1) (37), plant homeodomain finger protein 11 (PHF11) (38), mitochondrial ribosomal protein L17 (MRPL17) (39), and miR-19b (40, 41), as well as for membrane associated ring-CH-type finger 1 (MARCH1) (42), miR-155 (43, 44), IFN-α and β receptor subunit 1 (IFNAR1) (45), and miR-370 (43, 46). In addition, experimentally confirmed matches were found for TLR7 (1) and miR-17 (47, 48), IFN-induced protein with tetratricopeptide repeats 5 (IFIT5) (49, 50) and let7e (43, 51), and ankyrin repeat and FYVE domain containing 1 (ANKFY1) (52, 53) and miR-155 (43).
We also tested by GSEA whether other relevant gene sets are differentially induced between the RR and NN groups, including a gene set that was recently reported to characterize a state of activation of NK cells from HIV-1 controllers (HIC) (54). We found that the genes that were upregulated in the RR as compared with the NN group were significantly associated with the NK genes reported to be upregulated in HIC patients when compared with progressor patients (normalized enrichment score [NES] = 1.6, p = 0.0062, FDR = 0.89%; Fig. 3D). Briefly, the 21 genes detected to be enriched in RR when compared with NN were CC3a receptor 1 (C3AR1); MX dynamin like GTPase 2 (MX2); placenta specific 8 (PLAC8); hematopoietic SH2 domain containing (HSH2D); tripartite motif containing 22 (TRIM22); IFI44L; nucleic acid binding protein 1 (NABP1); GTPase, IMAP family member 8 (GIMAP8); sphingosine-1-phosphate receptor 1 (S1PR1); zinc finger protein 143 (ZNF143); cytokine inducible SH2 containing protein (CISH); DExD/H-box helicase 58 (DDX58); tetratricopeptide repeat and ankyrin repeat containing 1 (TRANK1); GIMAP4; C-X3-C motif chemokine receptor 1 (CX3CR1); GIMAP6; GIMAP7; tubulin δ 1 (TUBD1); grancalcin (GCA); tRNA nucleotidyl transferase 1 (TRNT1); and ZNF518A. Of these genes, four were significantly upregulated in the RR group when compared with the NN group (IFI44L [p = 0.02], MX2 [p = 0.02], PLAC8 [p = 0.04], and C3AR1 [p = 0.01]).
In addition, analysis of the baseline preimmunotherapy gene expression between groups controlling or not controlling plasma viral load upon primary endpoint identified a 30-gene signature with significant differential expression (p < 0.001; Supplemental Fig. 2A). Principal component analysis using the expression of these genes confirmed a separation of the RR group using preimmunotherapy gene expression alone (Supplemental Fig. 2B). Baseline gene expression differentiating R from N was also explored by GSEA. Results revealed one gene set (GSE18791) (55) to be significant in association with genes that were more upregulated in R when compared with N (NES = 2.1, p < 0.001, FDR = 4.7%). This gene set was identified as a regulatory network underlying the antiviral state transition during the first 18 h following the in vitro infection of DC with Newcastle disease virus (55). Interestingly, we also identified an association of genes that were upregulated in R compared with N with the gene set of HALLMARK_INTERFERON_ALPHA_RESPONSE (NES = 1.87, p < 0.001), although with FDR = 43.8%.
In summary, microarray data analysis allowed for the identification of gene signatures at baseline (on ART) and after 5 wk of ART+Peg–IFN-α2a with regard to R and N outcomes.
Enrichment in expression of genes that are associated with activation of NK and cell-mediated responses in subjects with decreased levels of integrated HIV DNA following Peg–IFN-α2a immunotherapy
Results from our NCT00594880 study (25) indicated that R, when compared with N, exhibited a significant drop in the levels of integrated HIV DNA in PBMC between baseline (ART) and primary endpoint (Peg–IFN-α2a). We evaluated whether these changes were associated with specific gene expression patterns. A total of 1260 probes (992 unique genes) identified after 5 wk of ART+Peg–IFN-α2a and 880 probes (703 unique genes) identified at primary endpoint (Peg–IFN-α2a) were analyzed for functional enrichments using IPA. Significantly enriched functions (p < 0.001) were then manually categorized into major classes. Only enriched categories with predicted activation state (z-score > 1 calculated by IPA based on direction of correlation of member genes) were considered. As demonstrated in Fig. 4A and Supplemental Table IIB and IIC this analysis showed that subjects who experienced reduction in integrated HIV DNA exhibited an enrichment in genes associated with leukocyte proliferation and survival, leukocyte chemotaxis and recruitment, leukocyte activation, cytotoxicity/cell-mediated response, NK cytotoxicity, and Ab-dependent cytotoxicity (ADCC). In contrast, subjects without decreases in integrated HIV DNA showed enrichment for genes associated with leukocyte cell death and cancer/neoplasia. No enrichment for ISG groupings was detected in association with changes in integrated HIV levels. Importantly, a similar pattern in gene enrichment was observed in data from two independent time points between subjects analyzed either after 5 wk of ART+Peg–IFN-α2a or during Peg–IFN-α2a monotherapy (Fig. 4 and Supplemental Table IIB, IIC showing the same genes across both time points). Interestingly, GSEA after 5 wk of ART+Peg–IFN-α2a or during Peg–IFN-α2a monotherapy further supported an association between immune responses and reduction in HIV. Briefly, subjects who experienced reduction in integrated HIV DNA had an enrichment in genes associated with immune functions such as NK cell–mediated immunity (NES = −1.68, p = 0.01, FDR = 6%), regulation of leukocyte mediated cytotoxicity (NES = −1.46, p = 0.04, FDR = 18%), and T cell proliferation (NES = −1.93, p = 0.001, FDR = 0.4%). In addition, GSEA REACTOME pathways analysis showed an enrichment in these subjects of genes associated with IFN signaling (NES = −2.21, p < 0.001, FDR = 0.01%), HIV life cycle (NES = −1.85, p < 0.001, FDR = 2%), antiviral mechanism of ISGs (NES = −1.56, p = 0.01, FDR = 11%), and apoptosis (NES = −2.1, p < 0.001, FDR = 0.07%).
Overall, these data indicate that reductions in cell-associated integrated HIV DNA following Peg–IFN-α2a administration are associated with activation of NK and cell-mediated gene expression programs.
Increase in innate activity and cytotoxic responses before and after Peg–IFN-α2a immunotherapy in subjects controlling HIV during Peg–IFN-α2a monotherapy
We analyzed whether innate variables measured on ART and after 5 wk of ART+Peg–IFN-α2a were related to plasma viral control after 12 wk of Peg–IFN-α2a monotherapy.
Consistent with baseline values as indicative of viral control outcomes postimmunotherapy, R had a higher baseline frequency of CD3+CD4− T cells expressing CD38 (p = 0.0459) and HLA-DR (p = 0.0079), higher baseline frequencies of NK cells bearing inhibitory receptors (i.e., CD3−CD56brightKIR2DL2/DL3+, p = 0.0397), and higher baseline plasma levels of IP-10 (p = 0.0395; Fig. 5A–D) when compared with N.
After 5 wk of ART+Peg–IFN-α2a, R also showed changes not observed in N, such as a decrease in the frequencies of CD25+HLA-DR+ NK cells (i.e., CD3−CD56dimCD16−CD25+HLA-DR+, p = 0.017; Fig. 5E) and of NK cells bearing inhibitory markers (i.e., CD3−CD56dimKIR2DL1+ [p = 0.0235], CD3−CD56brightKIR2DL2/DL3+ [p = 0.0184]; Fig. 5F, 5G). In addition, R also showed an increase in the frequencies of mature MDC (i.e., CD19−BDCA1+CD11c+CCR7+, p = 0.0318; Fig. 5H). NK cytotoxicity was tested ex vivo at baseline and after 5 wk of ART+Peg–IFN-α2a, showing no detectable effect of treatment on constitutive or in vitro IFN-α–induced cytotoxicity (Fig. 6A, 6B). However, subjects controlling viral rebound during Peg–IFN-α2a monotherapy (R) had a higher ex vivo IFN-α–induced NK cytotoxic response (Fig. 6C, 6D). Supporting higher NK degranulation in vivo, we also detected a trend for higher constitutive expression of CD107a in CD56dim NK cells (i.e., Lin3−CD56dimCD16−CD107a+, p = 0.099; Fig. 6E) in R when compared with N.
Overall, we identify lower KIR expression and higher NK cytotoxicity as correlates of HIV plasma viral load control upon Peg–IFN-α2a monotherapy.
In this study, we show that increases in innate activity, NK cytotoxicity, and gene expression are correlates of integrated HIV DNA decrease and plasma viral load control after ART interruption and continued Peg–IFN-α2a immunotherapy. Specifically, we identify modulations in KIR expression, NK cytotoxicity, and mature MDC as innate cell changes that are associated with HIV control after ART interruption and Peg–IFN-α2a monotherapy. We also introduce a baseline gene signature able to identify subjects more likely to control viremia after Peg–IFN-α2a monotherapy, as well as a gene signature able to identify subjects that fail to modulate ISGs after adding Peg–IFN-α2a to ART or to control viremia after Peg–IFN-α2a monotherapy. Interestingly, we describe that subjects failing to modulate ISG gene expression may include active mRNA expression patterns actively countering gene expression after Peg–IFN-α2a. Although future independent validation of described gene signatures is needed, our data following Peg–IFN-α2a treatment are consistent with prior observations with panobinostat suggesting that induction of ISGs together with changes in NK activation correlate with decreases in integrated HIV DNA after treatment with panobinostat in well controlled HIV-infected subjects on ART (56). Furthermore, we do find a significant association between genes that were upregulated in PBMC from subjects controlling viral load under Peg–IFN-α2a monotherapy and genes described to be upregulated in isolated NK cells from HIC (54).
Combined administration of ART+Peg–IFN-α2a for 5 wk resulted in a clear induction of IFN-mediated genes in the majority of subjects; however, the levels of antiviral IFN-induced genes alone were not enough to segregate between R and N. The fact that not all subjects that induced upregulation of these genes were R highlights a potential difference between the dominant antiviral correlates of IFN modulation on or off ART, as induction of ISGs from preimmunotherapy levels was noted as a correlate of temporal viral control of Peg–IFN-α2a administered to viremic subjects (13). By contrast, our study indicates that when Peg–IFN-α2a is administered in suppressed subjects on ART, the activation of innate cellular activity may be a greater correlate to control of HIV than ISG induction. Importantly, two independent NK cell changes (lower inhibitory KIR expression in NK cells and higher NK cytotoxicity) were correlated with control of plasma viral load, whereas two independent assessments by IPA gene expression supported NK cell activity (among other cell-mediated activity) as being significantly correlated with a decrease in integrated HIV DNA by primary endpoint.
Our data do not exclude a role for ISGs in HIV control after Peg–IFN-α2a on ART, as all R did increase ISGs. However, the observation that four of the nine N had gene expression profiles indicative of a refractory response indicates the possibility of identifying subjects that will not control HIV upon Peg–IFN-α2a monotherapy. Importantly, the lack of response in these subjects includes evidence for increases in miRNAs together with the downregulation of their corresponding gene targets. Among these miRNA gene targets were genes associated with cancer [e.g., IFI44L (36), RGL1 (37), MARCH1 (42), and MRPL17 (39)] and antiviral activity [e.g., IFIT5 (57) and TLR7 (1)]. Taken together with the upregulation of miRNAs (e.g., miR-155, let-7e, miR-370, miR-192, and miR-1275) that target IFN-inducible antiviral effector molecules (43), it is interpreted that these subjects may exemplify a response profile to Peg–IFN-α2a that would predict a return of plasma viral load, as was observed upon ART interruption.
Previous data by Hua et al. (58) in HIV-1/HCV-coinfected patients showed an increase in NK subsets after treatment with Peg–IFN-α2a as well as an association between reduction in viral reservoir and NK cells coexpressing activation markers NKG2D and NKp30. In our study, 5 wk of ART+Peg–IFN-α2a did not result in changes in the major NK subsets (besides an increase in CD56bright NK cells) or in activated NK, suggesting the lack of a direct effect of treatment in these subsets. Furthermore, we observed in subjects able to control plasma viral load a decrease in CD25+HLA-DR+CD56dimCD16− NK cells, supporting an association between levels of NK activation and HIV burden. Importantly, the observed differences between our study and the study by Hua et al. likely reflect differences in study design (e.g., patient population studied: HIV-1/HCV in the study by Hua et al. versus HIV in our study; and treatment duration: average of 11 mo in the study by Hua et al. versus 5 wk in our study). Of interest, the most uniform change noted across all subjects receiving Peg–IFN-α2a was a reduction in the ex vivo IFN-α–induced p–STAT-1, suggesting a lower cellular capacity to signal through the type I IFN receptor. Interestingly, constitutive levels of p–STAT-1 in circulating leukocytes were not higher after Peg–IFN-α2a immunotherapy, indicating that drug levels in circulation can maintain gene expression changes without sustained higher levels in constitutive p–STAT-1. We also document that the reduction in type I IFN receptor response did not affect type II IFN receptor responses within the same cells, indicating a receptor-level mechanism of unresponsiveness. The decrease in type I IFN receptor activity, together with the ex vivo data showing higher IFN-induced NK cytotoxicity as a correlate of HIV control, suggests that despite noted decreases in IFN-induced p–STAT-1, the retained induction of NK responses was retained within subjects controlling plasma viral load.
Hansen et al. (59) have also shown that IFN-α can increase T and NK cells’ cytotoxicity by boosting the IL-15–induced signaling and cytotoxic activity of these cells. In our study, no increased levels of IL-15 were observed in plasma after 5 wk of ART+Peg–IFN-α2a. This could be explained because soluble IL-15 can generally not be detected in physiological fluids, as it binds with a very high affinity to IL-15Rα at the surface of APCs. IL-15Rα gene expression was also not detected at a significant level over the background in microarrays in our study. However, GSEA exploring whether IL-15–responsive genes are differentially induced between RR and NN groups using the datasets GSE70214 (60), GSE120904 (database reported by Cezari et al. at the National Center for Biotechnology Information on October 16, 2018), GSE22886 (61), GSE22919 (62), and GSE7764 (63), which have been described to be modulated by IL-15, revealed three of these gene sets to be significant (p < 0.05 and FDR < 5%) in association with both upregulated and downregulated genes (GSE70214, GSE120904, and GSE22886; data not shown). These results suggest that IFN-α may have increased NK cells’ cytotoxicity by boosting IL-15 effects. Future studies will need to further investigate the connection between Peg–IFN-α2a and IL-15.
In this study, we show an increase in cytotoxic NK response against K562 targets at 5 wk of ART+Peg–IFN-α2a among subjects able to control plasma viral load. These results support the interpretation that NK cells contribute to a reduction in viral control via direct cell-mediated lysis of infected cells. Consistent with higher cytotoxicity, we also detected a decrease in CD3−CD56dimKIR2DL1+ and CD3−CD56brightKIR2DL2/DL3+ NK cells in subjects able to control viral load. Although it cannot be excluded that this loss of KIR expression (which is skewed to more differentiated NK cells) could also imply a reduction in more differentiated NK cells in favor of less-differentiated NK cell subsets or an expansion of KIR-negative NK cells, we interpret that lower inhibitory receptor activity may contribute to an increase in NK-mediated killing. Previous studies have shown that treatment with IFN-α can increase NK cells’ cytokine secretion, viral suppressive capacity, and cytotoxic function (12, 64–67). Gene analysis further supported this interpretation via an enrichment of genes associated with leukocyte activation, proliferation, survival, chemotaxis, and recruitment, as well as cytotoxicity/cell-mediated response and NK ADCC in persons with reductions in integrated HIV DNA. Although our data support higher NK cytotoxic function, they do not address induction of viral suppressive activity. Our study also does not address an additional role for NK cells in viral control, as indicated in nonpathogenic SIV infection, in which NK cells migrate into lymph node follicles in association with reductions in viral levels within lymph nodes (68).
Finally, unpaired t tests allowed for the identification of a signature that consisted of the top 30 most significantly differentially expressed genes at baseline (Supplemental Fig. 2) and was able to segregate R from N. This signature consisted mainly of genes that have been reported to be associated with gene suppression [e.g., DICER1 (69, 70) and Wolf–Hirschhorn Syndrome candidate 1 (WHSC1L1) (71)], negative regulation of innate immune responses [e.g., G-patch domain containing 3 (GPATCH3) (72)], or cancer [e.g., annexin A2 pseudogene 1 (ANXA2P1) (73), Rho associated coiled-coil containing protein kinase 1 (ROCK1) (74), and hemogen (HEMGN) (75)]. The presence in both R and N of high levels of expression of genes coding for proteins that participate in transcriptional repression (i.e., WHSC1L1 (71) and DICER1 (69, 70), respectively) underlies uncharacterized associations between gene expression and IFN-α–mediated viral control. Furthermore, no clear segregation was observed between subjects that were able to upregulate ISGs and those that did not, as shown by an upregulation in the RN and NN groups of the same genes. Future studies will be needed to validate and reconfirm this baseline gene signature on an independent validation set.
As no T cell responses were measured in our study, our data do not address the effects of Peg–IFN-α2a on T cell–mediated viral control. Paradoxically, recent data from humanized murine models infected with HIV and treated temporally with ART suggest a detrimental role for type I IFNs by acting against CD8+ T cell–mediated anti-HIV responses (76, 77). However, the lack of NK effector responses in humanized mice (78) and the absence of qualification of the effects of IFN-β versus IFN-α on the impairment of CD8+ T cell responses (79, 80) limit the inference from this animal model data to human responses using Peg–IFN-α2a as described in this study.
Our study has limitations. First, as a pilot study, the sample size is limited, and the uncovered associations and gene signatures described in this study need to be further validated in larger, independent studies. Second, our data do not address an ART interruption without Peg–IFN-α2a monotherapy because, as described in our prior publication (25), the U.S. Food and Drug Administration mandated the removal of that control group. Third, the gene signatures described are obtained in PBMC as opposed to independent cell subsets, and the relationship between gene expression and protein levels remains to be determined. Fourth, as no information on treatment history is available on study subjects, the role of time to ART initiation in the observed baseline differences between groups remains to be determined. Finally, all measures described were performed in peripheral blood, whereas no measures in tissue were performed.
Overall, this study provides, to our knowledge, the first proof-of-concept that in ART-suppressed subjects, immunotherapy with type I IFNs resulting in a reduction of HIV measures is associated with activation of NK responses. A randomized trial to formally test whether reductions in HIV measures on ART are the result of IFN-α immunotherapy is now underway (NCT02227277).
We thank the HIV-1 patients who participated in the study and the providers. We acknowledge support for this work by Griffin Reynolds, Natalie Opsitnick, Charity Calloway, Maxwell Pistilli, Skip Maino, Jocelin Joseph, Celia Chang, Sonali Majumdar, Sandy Widura, Shashi Bala, and Tran Nguyen.
This work was supported by National Institutes of Health Grants U01AI065279 and UM1 AI126620 (to L.J.M.). Additional support was provided by The Philadelphia Foundation (Robert I. Jacobs Fund), a Kean Family Professorship, Ken Nimblett and the Summerhill Trust, AIDS funds from the Commonwealth of Pennsylvania, and from the Commonwealth Universal Research Enhancement Program, Pennsylvania Department of Health, the Penn Center for AIDS Research (Grant P30 AI 045008), and Wistar Cancer Center Grant P30 CA10815. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.
The microarray data presented in this article have been submitted to the Gene Expression Omnibus (http://www.ncbi.nlm.nih.gov/geo/) under accession number GSE107549.
The online version of this article contains supplemental material.
Abbreviations used in this article:
area under the curve
blood DC Ag 1
CC3a receptor 1
CCR type 7
false discovery rate
gene set enrichment analysis
hepatitis C virus
IFN-induced protein 44 like
Ingenuity Pathway Analysis
killer cell Ig-like receptor, two Ig domains and long cytoplasmic tail 1
KIR2DL, two domains, long cytoplasmic tail 2/3
mean fluorescence intensity
primary endpoint nonresponder
normalized enrichment score
ISG nonresponder/primary endpoint nonresponder
primary endpoint responder
ISG responder/primary endpoint nonresponder
ISG responder/primary endpoint nonresponder
The authors have no financial conflicts of interest.