Clinical characteristics of HIV-1-infected patients with high levels of plasma interferon-γ: a multicenter observational study

The study complied with the principles of the Declaration of Helsinki regarding the investigations in human subjects and was performed after an ethics approval H26-NHO (AIDS)-03 had been received from Central Institutional Review Board Committee of the National Hospital Organization of Japan. Written consent was obtained from all study participants. The study cohort comprised 200 HIV-1-infected patients over 20 years of age who were regularly seen at one of the four hospitals of the National Hospital Organization. Patients were either naïve to ART, with CD4⁺ T cell count > 200 cells/μL, or had achieved viral suppression after having been on ART for over a year. Patients who developed fever or had other acute diseases were excluded from the study cohort.

Plasma was separated from the whole blood and kept frozen in − 80 °C until use. The levels of plasma IFN-γ and IL-6 were measured by the enzyme-linked immunosorbent assay (ELISA) (ThermoFisher Scientific, Waltham, MA) according to the supplier’s protocol. A cutoff level of 5 pg/mL was selected to delineate groups with high and low IFN-γ levels as previously described [7]. The cutoff level for IL-6 was set to 2 pg/mL, close to the upper quartile (1.9 pg/mL). When repeated blood collections were possible for patients with IFN-γ ≥5 pg/mL, samples were collected three times with at least 1 month in between each collection to confirm the reproducibility of high plasma IFN-γ levels.

Participants were divided into the groups with high and low IFN-γ and IL-6 levels based on cutoff values indicated above. The initial measurement of IFN-γ was used to divide the patients into the groups. Sex, age, assumed route of infection, use of ART, and comorbidities were compared at the time of entry to the study. The patients were defined to have diabetes mellitus, dyslipidemia, hypertension, chronic kidney disease, and osteoporosis if they were diagnosed with these conditions according to the Japan Diabetes Society criteria, Japan Atherosclerosis Society guideline, Japanese Society of Hypertension guideline, Japanese Society of Nephrology guideline, and Japanese Society for Bone and Mineral Research criteria [12‐14], respectively. Patients that received treatment for these conditions were also included. History of malignancy was defined as having history of any cancer that had been proven by biopsy. Chronic hepatitis B was defined as persistence of hepatitis B surface antigen for 6 months or more. Chronic hepatitis C was defined as the presence of detectable hepatitis C virus RNA in the serum. Only those comorbidities that were identified in over 5% of the study patients at entry were considered for analysis. Clinical categories, number of CD4⁺ T cells (CD4⁺ cell count), plasma HIV-1-RNA levels, and comorbidities at the time of HIV-1 diagnosis were also collected retrospectively from the medical records and compared between the groups. The change in CD4⁺ cell count following ART was calculated based on the cell count prior to the initiation of ART. Changes in CD4⁺ cell count were measured from 1 year after ART initiation for up to 10 years.

The results of the drug resistance test for HIV-1 were collected retrospectively for the purpose of the study. Resistance testing was performed as described previously [15]. Briefly, viral RNA was extracted from a plasma sample. The regions of HIV-1 protease and reverse transcriptase were amplified by using the reverse transcription polymerase chain reaction (PCR) method followed by nested PCR. After the purification of the amplified PCR products, the sequences were obtained by direct sequencing. Sequences used in the study (protease amino acid residues 1–99 and reverse transcriptase amino acid residues 1–240) were registered in GenBank (Additional file 1: Supplementary Materials and Methods). Nucleic acid sequences were aligned using Clustal W program version 2.1 (http://​www.​clustal.​org/​clustal2/​) and GENETYX-MAC version 18.0.6 (GENETYX, Tokyo, Japan). HIV-1 subtype B (HXB2, accession no. K03455) and HIV-2 (accession no. KX174313) were used for subtyping and as outgroup, respectively. Phylogenetic tree was constructed by using the neighbor-joining method (MEGA6 program), and the bootstrapping method was used to perform statistical evaluations.

Clinical characteristics of the patients are shown in Table 1. The median age was 42 years, and 98% of the patients (n = 196) were men. Twenty-seven percent of the patients (n = 53) were over the age of 50. The majority of the patients infected with HIV-1 (99%, n = 198) were Japanese; the remaining two patients were from Latin America and Oceania, respectively. Six percent of the patients (n = 12) were naïve to ART.

Plasma IFN-γ and IL-6 levels measured by ELISA are shown in Fig. 1. High IFN-γ and IL-6 levels were observed among the patients both on ART (Fig. 1a and c) and naïve to ART (Fig. 1b and d). Based on the cutoff value of 5 pg/mL, 41 patients were categorized as high IFN-γ group (21, 95% confidence interval: 15–27%), and the remaining 159 patients were categorized as low IFN-γ group. Furthermore, 42 patients (21, 95% confidence interval: 16–27%) had plasma IL-6 ≥ 2 pg/mL; among them, 17 patients also had IFN-γ ≥ 5 pg/mL. The distributions of patients between high and low cytokine groups were different for IFN-γ and IL-6 (P < 0.0001).

Reproducibility of the test results was examined by measuring plasma concentrations of IFN-γ three times in 35 out of 41 patients in the high IFN-γ group. IFN-γ concentration was ≥5 pg/mL in the two follow-up measurements for all patients who had IFN-γ ≥ 10 pg/mL at the initial measurement (n = 24, 69%). Similarly, IFN-γ concentration was ≥5 pg/mL in the two follow-up measurements for 6 of the 11 patients whose IFN-γ was 5–10 pg/mL at the initial measurement. For the remaining five patients, IFN-γ concentration was 2–5 pg/mL in at least one of the two follow-up measurements. Thus, the high levels of plasma IFN-γ were consistent and not transient across the three replicate measurements in most patients (n = 30, 86%).

Patients were divided into the two groups based on the initial levels of IFN-γ and IL-6, and their characteristics were compared. In order to match the characteristic background, the patients on ART were included for comparison (n = 188). Table 2 shows patient characteristics for each group. The median age of the patients at the time of study entry was statistically different (P = 0.0349), and the proportion of the patients below the age of 50 was higher in the high IFN-γ group (P = 0.0051). The proportion of patients who were on a protease inhibitor was significantly lower in the high IFN-γ group (26%) than in the low IFN-γ group (44%, P = 0.0449). The duration of ART in the high IL-6 group was shorter than that in the low IL-6 group. Patients in the high IFN-γ group were less likely to have dyslipidemia (15%, n = 6) compared with those in the low IFN-γ group (33%, n = 49; P = 0.0476). In addition, patients in the high IL-6 group had a higher rate of diabetes mellitus and higher levels of serum C-reactive protein than those in the low IL-6 group (P = 0.0266 and 0.0073, respectively). These observations indicated that patients’ characteristics were different in the groups with high and low IFN-γ and IL-6 levels.

The rates of comorbidities were also evaluated at the time of HIV-1 diagnosis (Table 3). The rate of dyslipidemia at diagnosis was lower than that at entry (9% versus 29%; P < 0.0001). Similar results regarding the rates of hypertension (2% versus 11%; P = 0.0006) and chronic renal disease (2% versus 11%; P = 0.0002, respectively) were observed at diagnosis and at entry. Dyslipidemia rates between the high and low IFN-γ groups and the rates of diabetes mellitus between the high and low IL-6 groups (P = 0.5327 and P = 0.0635, respectively) were not significantly different. These observations indicate that comorbidities tended to appear during the course of HIV infection.

Next, Spearman’s rank correlation coefficients among the factors related to high IFN-γ levels (age < 50, abacavir use, protease inhibitor use, and with dyslipidemia at entry) were calculated (Table 4). In contrary to the results illustrated in Table 2, a significant association was observed between high IFN-γ and abacavir use. A significant relationship was observed in all other comparisons except those between protease inhibitor use and abacavir use as well as between protease inhibitor use and age < 50, suggesting that these factors may interact with each other.

The recovery of CD4⁺ cell counts following ART was compared between the groups (Fig. 2). Patients that did not receive ART (n = 12) or for which CD4⁺ cell count just before initiating ART was unknown (n = 7) were excluded from this analysis. At the initiation of ART (baseline), CD4⁺ cell counts were not significantly different between the groups, with 268 and 210 cells/μL in the high and low IFN-γ groups (P = 0.0961), and 229 and 217 cell/μL in the high and low IL-6 groups (P = 0.9112), respectively. Figure 2 shows the median CD4⁺ cell count increase ± interquartile range (IQR) from the baseline. The difference between the groups was not significant for 3 years on ART. However, there was a significant difference indicating lower recovery of CD4⁺ cell counts in the high IFN-γ group after 4 years of ART (Fig. 2a). A similar reduction was not observed in the high IL-6 group (Fig. 2b). Cutoff values were evaluated by the comparisons with CD4⁺ cell count recovery after 4–7 years of ART. Cutoff values varied from 0.5, which was close to the median value of plasma IFN-γ concentration (0.54) in this study, to 10, so the participants were divided into two groups using these cutoff values. Figure 3 indicates P values and differences in the increase of CD4⁺ cell count for intergroup comparisons. The lowest P values were located within 5.1–5.7 pg/mL of IFN-γ concentration in all four assays (Fig. 3a). Maximum differences were also located within 4.5–5.7 pg/mL (Fig. 3b). These observations suggested that 5 pg/mL was potentially the optimal cutoff value for IFN-γ concentration and higher cutoff value may show nearly the same results. No significant differences were observed in any cutoff values at 1–3 years of ART for plasma IFN-γ concentration and at 1–10 years of ART for plasma IL-6 concentration.

Phylogenetic analysis of HIV-1 variants was performed in 113 patients by comparing 197-bp and 720-bp long sequences of genes encoding protease and reverse transcriptase, respectively (Fig. 4). Among the 45 patients with high levels of plasma IFN-γ and/or IL-6, 39 were distributed in a scattered manner within B subtype and non-B subtype. The sequences from the remaining six patients formed a cluster with a bootstrapping value of 84.8%. Sequences from four out of these six patients formed a cluster with a bootstrapping value of 92.3%; they exhibited similarity to the sequence previously described by Mori et al., which was a novel HIV-1 variant with rapid disease progression during primary HIV-1 infection [16]. Specifically, there was an insertion of the amino acid sequence Q[S/N]RPE in the p6 region of gag in addition to the amino acid sequence QNME in the C′-terminus of integrase. These observations suggested a possibility that viral factors might be involved in the maintenance of high plasma IFN-γ levels in these patients.