The phase 2a trial (ClinicalTrials.gov identifier: NCT01705223) enrolled a total of 150 HIV-uninfected healthy men and women to evaluate the safety and immunogenicity of a HIV-1 clade B′/C DNA vaccine (DNA/HIV-1) followed by an rTV/HIV-1 booster. Trial participants received three DNA/HIV-1 primes at weeks 0, 4, and 8, followed by a single subcutaneous dose of rTV/HIV-1 (Beijing Biological Preparations Institute Co., Ltd., Beijing, China). For the vaccinia sub-study, specimens from 15 participants who received the entire vaccination series were used. The Ethics Committee of the Chinese CDC approved this phase 2a trial (Ethics Committee Project Identifier: X111012202). All participants signed a written informed consent form prior to the initiation of the study procedures.

The DNA/HIV-1 vaccine consisted of two plasmids encoding the gp140 and the gag/pol/nef genes of clade B′ HIV-1 and clade C HIV-1, while the rTV/HIV-1 vector encoded the gag, pol, and gp140 genes of the circulating HIV-1 recombinant form CRF/07_BC/CN54 (17).

Blood was drawn from each volunteer into EDTA-2K Vacutainer tubes (BD Biosciences, Franklin Lakes, NJ, USA) on the day of the rTV/HIV-1 boost (day 0, pre-vaccination) and on days 3, 7, 14, 28, and 90 thereafter. Peripheral blood mononuclear cells (PBMCs) were isolated from these samples using standard Ficoll-Paque Plus (GE Healthcare, Chicago, IL, USA) density gradient centrifugation and frozen in dimethyl sulfoxide with 10% fetal bovine serum (FBS) in liquid nitrogen. The assays performed at the different sample collection time points are shown in Figure 1.

Plasma samples were heated to 56°C for 60 min to inactivate complement and serially diluted with Dulbecco’s modified Eagles’ medium (DMEM) supplemented with 2% FBS. Diluted serum (100 µl) was distributed into each of three corresponding wells in a 96-well plate, 50 µl of the rTV-Fluc virus was added to each well, and the plates were incubated at 37°C/5% CO₂ for 1 h. Following incubation, 100 µl of Vero cell suspension was added to each well. The plates were placed on a shaker for 1 min and incubated at 37°C/5% CO₂ for up to 48 h. After incubation, 100 µl of the supernatant was aspirated and 100 µl of D-luciferin substrate (Caliper, Waltham, MA, USA) was added into each well. The plate was incubated in the dark at room temperature for 2 min, after which luminescence was measured using a GLOMAX 96 microplate luminometer (Promega, Madison, WI, USA). The 50% neutralization titer (NT₅₀) for each serum sample was defined as the serum dilution at which the relative light unit (RLU) was reduced by 50% compared with the virus-containing control wells after subtraction of the background RLU in the cell-containing control wells. Titers >30 (the assay lower limit of detection) were considered positive (30). Anti-vaccinia nAbs were measured by the firefly luciferase-based assay as described previously (30). “Responders” (n = 10) were defined as participants whose nAb titer was >1:30 in at least three time points post-boost, and “non-responders” (n = 5) were defined as participants whose nAb titer did not exceed 1:30 in at least three time points post-boost.

Complete blood counts were conducted on fresh blood samples to obtain cell counts for each blood cell type and hemoglobin concentrations. For cell phenotyping, 1 million PBMCs from each sample were surface stained with the antibody cocktails shown in Supplementary Table S1 and with Live/Dead staining reagents (Life Technologies, Carlsbad, CA, USA) to assess cell viability. Staining was performed at 4°C in the dark for 30 min. For the immunostaining of intracellular proteins, cells were fixed and permeabilized with Fix/Perm (eBioscience, San Diego, CA, USA) for 30 min at 4°C in the dark prior to antibody incubation.

For analysis of natural killer (NK) cell-mediated antibody-dependent cellular cytotoxicity (ADCC), an established assay was used (31). In this assay, the target cells were coated with anti-P815 antibodies to serve as the Fc target cells for NK cell-mediated ADCC. The cells were then co-cultured with PBMCs as effector cells and NK cell function was assessed. Briefly, 0.2 million K562 cells (human chronic myelogenous leukemia cells, maintained in our laboratory) were cultured with P815-specific antibodies (Accurate Chemical & Scientific, Carle Place, NY, USA) for 1 h at 37°C in 5% CO₂. Antibody-coated P815 cells were washed twice with ice-cold RPMI medium containing 10% FBS and used as the target cells. The target cells were co-cultured with 1 million PBMCs (effector cells) and then incubated with anti-CD107a antibodies and Golgi-Stop (BD Biosciences) for 5 h at 37°C in 5% CO₂. Following culture, the samples were processed according to standard protocols and stained intracellularly with Alexa Fluor 700-conjugated anti-interferon gamma (IFNγ) and fluorescein isothiocyanate (FITC)-conjugated anti-tumor necrosis factor alpha (TNFα) antibodies.

For the quantification of T-cell function, 1 million PBMCs were infected with the parent rTV vector (no HIV-1 inserts) at a multiplicity of infection (MOI) of 1 or stimulated with phorbol myristate acetate (PMA) as a positive control. Brefeldin A (BFA) was added to each well, after which all the samples were incubated for 6 h at 37°C in 5% CO₂. The cells were permeabilized and stained intracellularly using phycoerythrin (PE)-conjugated anti-interleukin 2 (IL-2), Alexa Fluor 700-conjugated anti-IFNγ, and allophycocyanin (APC)-conjugated anti-TNFα antibodies. Events were acquired on a BD LSRFortessa instrument (BD Biosciences) and analyzed using FlowJo software (TreeStar, Ashland, OR, USA). The antibodies used for immunostaining are listed in Supplementary Table S1.

Vaccinia-specific IFNγ-secreting T cells were detected by enzyme-linked immunospot (ELISpot) assay. Plates were coated with purified anti-human IFNγ at a concentration of 5 μg/ml, incubated at 4°C overnight, and washed and blocked for 2 h at room temperature. PBMCs (2 × 10⁵) were added to the wells in duplicate. The cells were stimulated with vaccinia virus at a MOI of 1. The positive control was stimulated with PMA at 50 ng/ml and ionomycin at 1 μg/ml; the negative control was stimulated with medium alone. The PBMCs were incubated at 37°C and 5% CO₂ for 24 h and then lysed with sterile water. The plates were washed three times with PBST (phosphate-buffered saline with Tween 20) prior to a 1-h incubation with biotinylated anti-human IFNγ antibody, followed by the addition of streptavidin–horseradish peroxidase (HRP) at 37°C for 1 h. The plates were washed again and developed with 100 µl of a 3-amino-9-ethylcarbazole (AEC) substrate solution for approximately 30 min. The reaction was stopped by washing the plate with distilled water. IFNγ spots were analyzed by an automated ELISpot plate reader (ImmunoSpot, Cleveland, OH, USA). Spot-forming cells (SFCs) were quantitated as the average number of spots in duplicate wells per 10⁶ PBMCs.

After isolation, 1 × 10⁶ PBMCs were immediately lysed in 1 ml of TRIzol (Life Technologies) and stored at −80°C until sample processing. For processing, the samples were thawed and the RNA extracted following the manufacturer’s instructions. The quality of the RNA was assessed using a Nanodrop 2000 Spectrometer (Thermo Scientific, Waltham, MA, USA) and by visualizing the integrity of the 28S and 18S bands on an Agilent Bioanalyzer 2100 instrument (Agilent Technologies, Santa Clara, CA, USA).

Qualified total RNA was further purified using a RNeasy Micro Kit (QIAGEN, Hilden, Germany). Contaminating genomic DNA was removed using an RNase-Free DNase Set (QIAGEN). The purified RNA was stored at −80°C until microarray analysis.

To obtain biotin-labeled complementary RNA (cRNA), total RNA was amplified, labeled, and purified using the GeneChip 3′IVT Express Kit (Affymetrix, Santa Clara, CA, USA) following the manufacturer’s instructions. After hybridization on Human PrimeView Arrays for 16 h at 45°C and rotation at 60 Work4Perpm in a Hybridization Oven 640 (Affymetrix), the slides were washed and stained with a Fluidics Station 450 system (Affymetrix). Scanning was performed on a seventh-generation GeneChip Scanner 3000 (Affymetrix). Affymetrix GCOS software was used to perform image analysis and generate raw intensity data.

All microarray data analysis was performed in R (version 3.5.2) (32). Initially, data quality was assessed by determining the background level, 3′ labeling bias, RNA quality, and pairwise correlation among the samples. For the PrimeView chip, the customized CDF file (version 22, ENTREZG) downloaded from the BrainArray website was performed in probe set mapping. Normalization was performed with the RMA algorithm, which includes global background adjustment and quantile normalization. Interquartile range (IQR) was applied for raw data filtering using the genefilter package (version 1.64.0) (33), and the threshold was set to remove genes with an IQR less than the 50th percentile of the IQRs across all genes. The sample gene intensities were log₂ transformed in all subsequent analyses. Genes were annotated using information from NCBI (34) (modified April 25, 2017). These data have been deposited in the NCBI Gene Expression Omnibus and are accessible through GEO Series accession number GSE118976.

For this analysis, the samples used were from a subset of participants (n = 15, age range = 26–51 years, mean ± SD = 38.46 ± 8.04) in the phase 2a DNA/HIV-1 prime, rTV/HIV-1 booster trial. The study design is illustrated in Figure 1.

We began by assessing the kinetics of the vaccinia-specific cellular and humoral responses to rTV/HIV-1 vaccination. rTV/HIV-1 vaccination induced high vaccinia-specific IFNγ T-cell responses on day 14; these responses declined over time, but remained detectable through day 90 (Figure 2A). The vaccinia-specific CD8⁺ T-cell responses waned more rapidly than did the vaccinia-specific CD4⁺ T-cell responses on days 28 and 90 (Supplementary Figures S1A, B). Moreover, the vaccination also elicited the vaccinia-specific IFNγ NK cell responses on day 14 (Supplementary Figure S1C). Anti-vaccinia nAbs were detectable on day 14 and were maintained at a similar level through day 28 (Figure 2B). As substantial variability was observed across participants’ vaccinia-specific T-cell responses and nAb responses, we investigated correlations at the individual level. Moderate positive correlations between the nAb and T-cell responses were observed on days 14 and 28 (Figure 2C).

In some participants, nAb responses were not detected, enabling their characterization into Rs and NRs according to the vaccine-induced nAb titer (see Materials and Methods). In order to ascertain that nAb responsiveness was not led by failures of inoculation, vaccine “take” acted as an alternative measure to evaluate whether the vaccine established infection successfully in Rs and NRs because vaccinia vaccination induces barely detectable viremia (48, 49). Vaccine “take,” corresponding to the development of a lesion at the inoculation site, provides evidence of local viral replication and is generally held to be a correlate of vaccinia-specific immunity and clinical protection against smallpox (50). It was thus surprising that Rs and NRs alike all developed vaccine takes, with no significant difference in skin lesion size between the two groups (Figure 2D). This finding suggested that the vaccinia virus established infection similarly in Rs versus NRs and prompted us to further investigate the reasons for NRs exhibiting vaccine take, yet failing to mount a nAb response.

We first employed the weighted gene co-expression network analysis (WGCNA) algorithm, which defines transcriptional modules based on gene co-expression and identifies gene clusters associated with particular immune response readouts, to examine the associations between the gene modules and vaccine-induced immune responses, including vaccinia-specific T-cell responses, anti-vaccinia nAbs, and the frequencies of different immune cell types circulating in the PBMC compartment. The dataset (9,317 gene transcripts) from Rs clustered into 11 distinct modules and identified “turquoise,” “red,” and “magenta” modules positively associated with CD8⁺ T cells and/or nAb titer, along with “brown” and “cyan” modules that were negatively correlated with nAb titer in Rs (Supplementary Figure S2A), and 12 differential modules were clustered in NRs, but no module was significantly associated with nAb response (Supplementary Figure S2B); therefore, subsequent analysis only focused on interpreting the Rs dataset.

Overrepresentation analyses revealed that the top genes (in terms of strong gene–module relationships and/or high gene–immune response correlations; see Materials and Methods) in the brown, cyan, magenta, turquoise, and red modules were related to the AP-1 TF network, inflammation signatures, B cells, T-cell activation, and T-cell mitotic cycle, respectively (Figure 3). These findings imply that the AP-1 TF network and inflammation genetic signatures were suppressed in Rs, but T cells and B cells were active. Nevertheless, these co-expression networks were only constructed in Rs in order to characterize the landscape of transcriptional response in both Rs and NRs. The BTM-based gene sets were utilized to compare potential differences between the Rs and NRs and to identify the association of nAb and gene profiling (27). The NRs had almost no transcriptional response to rTV/HIV-1 vaccination, with the exception of a small day 3 response such as the activation of immune sensing modules and the downregulation of translation- and transcription-related modules (Figure 4A). In contrast, the Rs had a broad transcriptional response to rTV/HIV-1 vaccination, spanning the activation of modules related to cell cycle, metabolism, nuclear pore transport, translation, and T-cell activation/differentiation. Especially, the Rs showed persistent downregulation of multiple modules throughout the entire time course, including the “AP-1 transcription factor network (M20)” and “proinflammatory cytokines and chemokines (M29),” which was consistent with the results from the nAb titer metric pre-ranked GSEA on gene data on days 14 and 28 (Figure 4B). Through the pre-ranked GSEA, any gene set positively enriched can be considered to be positively associated with nAb titer, and vice versa. Consistent with the data presented in Figure 4A, the modules with the highest enrichment scores in the Rs were almost entirely related to T-cell activation (Figure 4B). In contrast, the modules with the largest negative enrichment scores in the NRs involved monocyte, neutrophil, and inflammation enrichment (Figure 4B). To visualize the individual genes that were present in modules with the largest positive and largest negative enrichment scores, we generated a Circos plot (Figure 4B). The small arc on the left shows the enrichment-adjusted p-values for the NRs and Rs at different time points, and the larger arc on the right shows whether the genes in each module were identified as DEGs. The links between the modules and genes indicate the relationships. The plot shows that the inflammation-related genes (NALP3 and PTX3) and the innate sensing gene (TREM1) were downregulated in Rs, while genes related to T-cell status and differentiation (CD8B, ETS1, CXCR6, CCR7, GZMM, etc.) were upregulated.

The results suggest that nAb titers were positively associated with the T-cell response, but inversely correlated with inflammation.

Considering that the AP-1 TF network was suppressed over the entire time course in the Rs and was enriched in the brown module (which was inversely associated with nAb), we constructed transcriptional regulatory networks in Rs and NRs, with the aim of identifying master regulators (i.e., TFs) that potentially drive the development of a nAb response. Overall, 149 TFs were identified as differentially regulated in Rs (81 upregulated and 68 downregulated) (Figure 5A). The kinetics of TF changes varied across TFs. Some were persistently upregulated from day 3 through day 28, whereas other NRs exhibited transient upregulation, predominantly on day 3 (Supplementary Figure S3). As combinatorial interactions of multiple TFs can dictate the gene expression outcomes, a TF-coordinated interaction network was generated for each time point using the differentially expressed TFs in Rs. Across all time points (Figure 5B), AP-1 complex TFs (ATF3/FOSB/JUN) were downregulated and dominated the network. While these network hubs were downregulated, ATF1 and BATF were persistently upregulated; CREB1, TCF20, SMAD4, and CTCF were highly expressed at later time points (days 14 and 28). However, the expression patterns of these TFs in Rs differed from those in NRs on day 7 (Figure 5C), in which AP-1 complex genes were highly upregulated in NRs, albeit with some degree of individual variability. Moreover, while the pro-inflammatory cytokines IL1β, IL6, and TNFα were significantly upregulated in NRs, they were all downregulated in Rs (Figure 5C). Aside from these later time points, we also found that AP-1 gene expressions were detected within 1 day post-rTV/HIV-1 vaccination (hour 4 and day 1), and both Rs and NRs showed a slight upregulation of JUN, FOS, and other AP-1 complex genes (Figure 5D), but with inter-individual variabilities.

These results imply that the AP-1 pathway was initiated in both Rs and NRs, but AP-1 pathway activation was persistent up to day 7 post-immunization in NRs, which demonstrated an aforementioned adverse correlation with nAb.

Vaccinia inoculation initiated the activation of the AP-1 pathway (51), which is a major pathway mediating the inflammatory response. Our gene profiling data also pointed to a clear association between inflammation and nAb. Neutrophils and monocytes, two immune cell subsets that secrete pro-inflammatory cytokines, can be considered as surrogates of inflammation alternatively. Across all participants, a moderate inverse correlation (R = −0.57, p = 0.066) was found between the percentage of neutrophils in whole blood on day 7 and anti-vaccinia nAb titer on day 14 (Figure 6A). Cell type profiling was assessed in fresh whole blood to calculate the percentages of neutrophils (Figure 6B) and monocytes (Figure 6C) in circulation. As shown in Figure 6B, NRs started with a similar percentage of neutrophils at baseline to Rs, but relatively elevated at hour 4 through day 28 and revealed near significances on days 7 and 28. Moreover, the percentages of monocytes (Figure 6C) remained similar in Rs and NRs from baseline through day 3, then followed by a rapid elevation on day 7 in NRs with near significance (p = 0.079). We also measured the inflammation index of both groups within the sampling time course. This measurement was calculated based on the expressions of inflammatory pathway-related genes (see Materials and Methods). The inflammatory index of the Rs was slightly higher than that of the NRs at baseline, whereas the opposite pattern was observed post-rTV/HIV-1 boost. The highest difference between the NRs and Rs occurred on day 7 (0.551 vs. 0.358, p = 0.076) (Figure 6D).