This was a descriptive, cross-sectional, and analytical study. We recruited patients with symptoms of intermittent fever or episodes of recurrent fever from the Setor de Atendimento Médico Unificado do Instituto Evandro Chagas (Unified Medical Care Sector of Instituto Evandro Chagas; SOAMU-IEC); patients with a positive diagnosis of HIV in a subclinical state from the Centro de Atenção à Saúde nas Doenças Infecciosas Adquiridas (Health Care Center for Acquired Infectious Diseases; CASA DIA); and voluntary blood donors from the Fundação Centro de Hematologia e Hemoterapia do Estado do Pará (Pará State Hematology and Hemotherapy Center Foundation; HEMOPA). From January 2018 to January 2020, peripheral blood samples were collected weekly.

A total of 576 participants were included, of whom 19 had serology consistent with primary EBV infection (P.I.) (anti-VCA IgM (+), anti-VCA IgG (−)), 90 had a class serological transition profile (S.T.) (anti-VCA IgM (+), anti-VCA IgG (+)), and 467 had serology consistent with past EBV infection (Past I.) (anti-VCA IgM (−), anti-VCA IgG (+)).

Sociodemographic data (gender, age, education and family income), behavioral data (smoker, alcohol drinker, user of illicit drugs, sexual orientation, active sexual life, steady sexual partner, history of sexual relations with sex workers and use of sexual condoms) and clinical data (sleep quality, diagnosis or family history of cancer, symptoms and history of infections) were obtained through a project-specific questionnaire applied during the participant interview and from medical records databases accessed through authorization filed by the institutions where the study was carried out. All access and disclosure of participant data were included in the study’s ethical opinion.

EBV infection was screened by semiquantitative detection of anti-VCA IgM and IgG class antibodies by enzyme immunoassays (Dia. Pro Diagnostic Bioprobes EBV VCA, Italy). Identification of EBV genotypes was performed by nested PCR targeting the EBNA-3C gene using primers described by Sample et al. (23) and Lorenzetti et al. (24) following appropriate recommendations: (1° round) (F: 5′-AGATGGTGAGCCTGACGTG-3′/R: 5′-GCATCC TTCAAAACCTCAGC-3′) (2° round) (F: 5′-AGAAGGG GAGCGGTGTGTTGT-3′/R: 5′-GGCTGTTTTTGACGTCGGC-3′). The reaction mixture and program were as follows: 10 pmol/μL primers, MgCl2 (50 mM), dNTPs (10 mM), and Taq (5 U/μL); cycling 1° round–1 cycle at 95°C/3′; 20 cycles at 94°C/45, 56°C/45, 72°C/45; 1 cycle at 72°C/7; cycling 2° round–1 cycle at 95°C/3′; 35 cycles at 94°C/45, 56°C/45, and 72°C/45; and 1 cycle at 72°C/7. The presence of a 153-bp fragment was considered positive for EBV-1; a 246-bp fragment was considered positive for EBV-2.

To quantify the EBV load, we used blood plasma samples from patients with positive serology for anti-VCA IgM in a real-time PCR estimation matrix following the protocol of the XGEN MASTER EBV kit (Mobius Life Science, Pinhais, PR, Brazil).

Plasma concentrations of the cytokines IL-17A, IFN-γ, TNF, IL-10, IL-6, IL-4 and IL-2 were determined using a Cytometric Bead Array (CBA) with BD FACSCanto™ II and BDTMCBA Human Th1/Th2/Th17 Cytokine kits (BD Biosciences, San Jose, CA, USA). Quantification of CD4(+) T, CD8(+) T, CD4(+)/CD8(+) T (double-positive) and CD4(−)/CD8(−) T (double-negative) lymphocytes was performed by immunophenotyping and flow cytometry using BD FACSCalibur-4 colors and monitoring kits FACSCountTM Reagents and TriTEST™/TruCount (BD Biosciences, San Jose, CA, USA).

Peripheral blood samples were collected from the participants, and DNA was extracted using a QiaAmp DNA Mini Kit (Qiagen, Düsseldorf, Nordrhein-Westfalen, Germany) following the manufacturer’s recommendations. The extracted DNA samples were quantified by spectrofluorimetry using Qubit equipment (Invitrogen, USA) following the manufacturer’s recommendations. The degree of purity was evaluated using a NanoDrop™ 2000/2000c spectrophotometer (Waltham, Massachusetts, USA), in which the elution solution used for extraction of genetic material was used as a reference standard.

We standardized the following profiles as the ideal range for successful amplification: concentrations between 10 and 15 μg/ml. The degree of purity for the locus was represented by the ratios 260/280: 1.8–2.0 and 260/230: 1.8–2.2. Samples outside the established standard were diluted in ultrapure distilled water free of DNase/RNase (Invitrogen, USA) at concentrations above the expected concentrations. Samples with concentrations below the expected concentrations were re-extracted.

HLA genotyping was performed using low/medium resolution PCR-SSO methodology (polymerase chain reaction – sequence-specific oligonucleotide) with Luminex technology (Luminex Corporation, Austin, TX, USA) and a LABType^® kit (One Lambda Inc., Canoga Park, USA). CA, USA).

The target DNA was amplified by conventional PCR using specific primers for each locus provided by the recombinant Taq kit (Invitrogen, USA) and D-mix solution following the supplier’s amplification protocol, in which, for each sample, 13.8 μL of D-mix, 4 μL of the supplied primer and 0.2 μL of TAQ polymerase, totaling 18 μL of preparation for 2 μL of extracted DNA. Sequence amplification followed the following cycles: 1 cycle at 96°C for 3 min; 5 cycles at 96°C for 20 s, 60°C for 20 s and 72°C for 20 s; 30 cycles at 96°C for 10 s, 60°C for 15 s and 72°C for 20 s; and 1 cycle at 72°C for 10 s. To confirm amplification, electrophoresis was performed on a 2.5% agarose gel.

The amplified DNA was denatured and subjected to hybridization with a set of specific fluorescently stained oligonucleotide probes immobilized on polystyrene microspheres. The reagents were standardized in the following proportions: 2.5 μL of denaturation buffer/sample, 5 μL of neutralization buffer/sample, 34 μL of hybridization buffer/sample, and 4 μL of bead mix/sample.

The microspheres were washed with wash buffer (480 μl/sample) and then reacted with 50 μL of streptavidin-phycoerythrin (SAPE) conjugate, which binds to biotinylated amplified DNA. The SAPE mixture was standardized in the following proportions: 0.5 μL of stock SAPE at 10 × /sample and 49.5 μL of SAPE buffer/sample.

In the analytical phase, the fluorescence intensity of phycoerythrin in each microsphere was evaluated using a Luminex analyzer, and the data were archived. Data analysis was performed with HLA Fusion software to determine the HLA gene alleles, and the reactivity pattern of each DNA sample in relation to the set of probes conjugated to the microspheres allowed the establishment of the genotype.

We applied multivariate analysis to determine the separation of the studied groups according to the values of their variables (lymphocyte quantification, cytokine dosage, anti-EBV antibody titer and EBV viral load). A scatterplot was generated to identify the groups and visualize the group separations and approximations.

We calculated the Spearman coefficient through a matrix of general correlations between the immunological and virological variables in each of the three groups.

The frequency of HLA alleles was calculated by direct counting and compared between groups using the G test. For groups with significantly different allelic profiles, the chi-square residual test was applied to determine the probabilistic importance of each of the alleles, followed by calculation of the odds ratio to determine the advantage or disadvantage of significant alleles. We calculated the false positive probability (FPRP) for significant associations, with a predefined threshold value of 0.5. The odds ratio for calculating the statistical power was 1.5. We adopted a range from 0.25 to 0.00001 as the prior probability of association of the alleles with the serological profiles of EBV infection, in accordance with the recommendations of Wacholder et al. (25).

We also applied the G test to compare sociodemographic and clinical data between the groups. To assess significance, logistic regression was performed, with the EBV serological profile used as the dependent variable and the analyzed data used as independent variables.

HWE was measured by selecting as an alternative hypothesis of interest an excess of heterozygotes in the groups; standard Markov chain parameters were adopted (26).

We applied the Mann–Whitney test for two-by-two comparisons of quantitative and semiquantitative data between the groups and between HLA alleles. We opted for nonparametric tests due to the degree of normality of the variables in question, which was estimated by the Lilliefors test.

We adopted a significance level (α) of 95% while considering a probability of significance (p) less than or equal to 0.05 as a criterion for rejecting the null hypotheses for the statistical analyses. The diagrams and graphs were assembled using GraphPad Prism 8.4.3 (San Diego, CA, USA), and statistical analyses were performed using BioEstat 5.3 software (27).

In compliance with resolutions 466/2012 and 347/05 of the National Health Council, which address guidelines and regulatory standards for research involving humans, the project was submitted for ethical consideration and approved by the Ethics Committee in Research with Human Beings of the IEC (Protocol: 3.121.265; CAAE: 73927717.3.0000.0019). All participants were informed about the research objectives, and those who agreed signed a consent form. Individuals under 18 years of age or who were using any specific therapy (antivirals, antiretrovirals or immunosuppressants) were excluded.

Among the sociodemographic and behavioral characteristics, the most common were male sex, age between 18 and 28 years, complete primary education and family income between 1 and 3 years. There were also higher rates of nonsmokers, those who consumed alcohol and those who had no history of using illicit drugs among the participants, as well as heterosexuals and those who had an active sexual life, who had a steady partner, and who had no relationship with sex workers. and who reported using condoms in their relationships (Table 1).

In relation to stress and history of cancer, individuals who did not have difficulty sleeping were more likely to experience stress, but with variable rest times and without a diagnosis of cancer or family history (Table 1).

Regarding the history of infections and clinical aspects, individuals with a history of coinfection with HIV and syphilis prevailed. Fever and sore throat were more common in symptomatic individuals (Table 1).

HIV coinfection varied between 70 and 100% in the serological groups, and it was significantly associated with the primary phase of EBV infection (100% of patients; p = 0.0008). In fact, patients with primary EBV infection were approximately 11 times more likely to be coinfected with HIV than patients in the other groups were (OR: 11.4; CI (95%): 1.51–85.88; p = 0.0064) (Table 1).

In multivariate analysis, we observed an overlap between the serological transition groups and the past EBV infection group; the primary infection group tended to cluster in coordinates further away from the others; the generated analysis represented approximately 47% of the total data variance (Figure 1A). The viral load and CD4(+) and CD8(+) T lymphocyte counts were the most relevant parameters. The primary infection group had a greater viral load than did the serological transition group (P.I. = median: 15.33; IIQ: 6.50–78.50); (S.T. = median: 0; IIQ: 0.0–2.0); p < 0.00001) (Figure 1B) and a lower CD4(+) T lymphocyte count (P.I. = median: 319.50; IIQ: 140.00–443.25); (S.T. = median: 475; IIQ: 245–764); (Past I. = median: 525.50; IIQ: 280.75–938.25); p: 0.0068) (Figure 1C) but a higher CD8(+) T lymphocyte count than did the other groups (P.I. = median: 1024.50; IIQ: 886.25–2136.75); (S.T. = median: 1140; IIQ: 582–1735); (Past I. = median:807; IIQ: 553–1235); p < 0.00001) (Figure 1D).

In the group with primary infection (Figure 2A), the IgM concentration correlated negatively with the viral load (r: −0.568; p: 0.0334) and positively with the CD4(+) T lymphocyte count (r: 0.7004; p: 0.0006), IL-6 level (r: 0.1767; p: 0.0457), TNF level (r: 0.2019; p: 0.0522) and IL-4 level (r: 0.1860; p: 0.0517). The EBV viral load correlated negatively with the CD4(+) T lymphocyte count (r: −0.1874; p: 0.0526) and IL-6 level (r: −0.1915; p: 0.0525) but positively with the IL-4 level (r: 0.6976; p: 0.0427). In addition, the CD4(+) T lymphocyte count correlated positively with the IL-17A (r = 0.1250; p = 0.0494) and TNF (r = 0.1367; p = 0.0513) levels. The CD8(+) T lymphocyte count correlated negatively with the CD4(+) T lymphocyte count (r: −0.1612; p: 0.0497) and IL-4 level (r: −0.1748; p: 0.0475) and positively with the double-negative T lymphocyte count (r: 0.7431; p: 0.0002). The double-positive T lymphocyte count also correlated positively with the IL-17A level (r: 0.6584; p: 0.0016), and cytokine levels correlated positively with each other (r: 0.1051–0.8297).

In the serological transition group (Figure 2B), the viral load correlated negatively with the CD4(+) T lymphocyte count (r = −0.1791; p = 0.0569) and positively with the IL-6 level (r = 0.3312; p = 0.0017). The CD4(+) T lymphocyte count correlated negatively with the IL-6 level (r: −0.6256; p: 0.0021), though the double-positive T lymphocyte count correlated positively with the CD4(+) T lymphocyte count (r: 0.6521; p < 0.0001) and negatively with the IL-6 level (r: −0.2536; p: 0.0177). Additionally, cytokine levels correlated positively (r = 0.4998–0.9343).

In the group with past EBV infection (Figure 2C), the IgG level correlated negatively with the CD4(+) T lymphocyte (r: −0.3256; p < 0.0001) and double-positive T lymphocyte (r: −0.1108; p: 0.0401) counts and positively with the CD8(+) T lymphocyte count (r: 0.1712; p: 0.0016). However, the CD4(+) T lymphocyte count correlated negatively with the IL-6 level (r: −0.1069; p: 0.0476) and positively with TNF (r: 0.1042; p: 0.0535) and IL-4 (r: 0.1053; p: 0.0510) levels and with double-negative T lymphocyte (r: 0.3555; p < 0.0001) and double-positive T lymphocyte (r: 0.4532; p < 0.0001) counts. The CD8(+) T lymphocyte count correlated positively with the double-positive T lymphocyte (r: 0.1238; p: 0.0219) and double-negative T lymphocyte (r: 0.2279; p < 0.0001) counts, and the double-negative T lymphocyte count correlated positively with the IL-2 (r: 0.1128; p: 0.0368), TNF (r: 0.1554; p: 0.0041) and IL-4 (r: 0.1283; p: 0.0175) levels. For the other groups, cytokine levels correlated positively with each other (r = 0.0913–0.8608).

The DRB1 locus was in Hardy–Weinberg equilibrium in all groups studied (P.I.: p = 0.9158; S.T.: p = 0.7626; Past I.: p = 0.8574). We observed relevant differences in the frequency of DRB1 locus alleles between the groups studied (p = 0.0477), with the DRB1*09 allele showing greater probabilistic relevance in the group with primary EBV infection (residues x²: 2.4538), the DRB1*03 allele being more relevant in the serological transition group (residues x²: 2.8528) and the DRB1*16 allele being more relevant in the group with past EBV infection (residues x²: 2.6223) (Figure 3A and Table 2). The rate of heterozygosity was high in the three groups, with no significant differences among them (p = 0.1817) (Table 2).

The DRB1*09 allele was approximately 4-fold more frequent in the group with primary EBV infection than in the other groups (OR: 4.235; CI (95%): 1.21–14.80; p: 0.054) (Figure 3B).

The median EBV viral load was greater in carriers of this allele (DRB1*09 = 47.33; IIQ = 40.33–80.67) (others = median: 9.33; IIQ = 3.75–18.33); p = 0.0485) (Figure 4A). However, the anti-VCA IgM titer was not associated with DRB1*09 (Figure 4B).

The CD8(+) T lymphocyte count was greater (DRB1*09 = median: 2844; IIQ: 2401.5–3054.0) (others = median: 975.5; IIQ: 795.5–1179.5); p = 0.0206) (Figure 5B), as was the double-positive T lymphocyte count (DRB1*09 = median: 27; IIQ: 21–29) (others = median: 3.5; IIQ: 2.75–7.00); p = 0.0093) (Figure 5D).

The levels of TNF (DRB1*09 = median: 11.64; IIQ: 11.07–12.38); (others = median: 9.52; IIQ: 8.72–10.24); p = 0.0161) and IL-4 (DRB1*09 = median: 11.78; IIQ: 10.30–12.04); (others = median: 8.52; IIQ: 6.67–8.85); p = 0.0464) were also higher in the DRB1*09 allele carriers (Figure 6C). Other cytokines were not associated with the allele (Supplementary Data 1).

The DRB1*03 allele was approximately 2-fold more frequent in patients who underwent a serological transition (OR: 2.1560; CI (95%): 1.27–3.67; p: 0.001) (Figure 3C).

Moreover, the EBV viral load was greater in DRB1*03 allele carriers (DRB1*03 = median: 20; IIQ: 0.00–52.98) (others = median: 0; IIQ: 0–0); p: 0.0015) (Figure 4A). In contrast, anti-VCA IgM and IgG titers were not associated with this allele (Figures 4B, C).

The CD4(+) T lymphocyte counts (DRB1*03 = median: 272; IIQ: 83.25–450.25); (others = median: 477; IIQ: 223.0–743.5); p: 0.0112) (Figure 5A) and double-negative T lymphocyte counts (DRB1*03 = median: 50; IIQ: 39.25–90.00); (others = median: 75; IIQ: 45.5–116.5); p: 0.0540) (Figure 5C) were lower in carriers of the DRB1*03 allele, but the CD8(+) T lymphocyte count was greater (DRB1*03 = median: 1749; IIQ: 1426.00–2367.25); (others = median: 1111; IIQ: 620.0–1687.5); p: 0.0260) (Figure 5B).

Additionally, IL-4 dosage levels were lower in carriers of this allele (DRB1*03 = median: 8.89; IIQ: 7.09–11.34); (others = median: 10.10; IIQ: 8.45–14.10); p: 0.0478) (Figure 6A), but the IL-6 level was greater (DRB1*03 = median: 14.96; IIQ: 11.22–26.00); (others = median: 12.46; IIQ: 9.44–16.62); p: 0.0175) (Figure 6B). Other cytokines were not associated with the allele (Supplementary Data 1).

In this study, the DRB1*16 allele was significantly associated with past EBV infection (OR: 0.2724; CI (95%): 0.10–0.76; p: 0.012) (Figure 3C), being up to 7-fold more frequent in this group.

The anti-VCA IgG titer was lower in DRB1*16 allele carriers (DRB1*16 = median: 5.41; IIQ: 3.76–7.63) (others = median: 6.9; IIQ: 4.93–8.39); p = 0.0020) (Figure 4C).

The CD4(+) T lymphocyte count was greater in carriers of this allele (DRB1*16 = median: 541; IIQ: 361–980) (others = median: 480; IIQ: 250.75–891.25); p = 0.0116) (Figure 5A). The counts of other cells were not associated with the allele; however, CD8(+) T lymphocytes tended toward lower counts in carriers of the allele (DRB1*16 = median: 744; IIQ: 595–1106; (others = median: 853; IIQ: 595–1273); p: 0.0602) (Figure 5B).

Our data showed that HIV infection was associated with serological profiles of EBV infection, which was more frequent in the group with primary infection (Table 1). We subsequently evaluated whether this aspect was also associated with alleles of the DRB1 locus. We found that the frequency of the DRB1*09 allele was significantly greater in individuals with HIV who were in the primary phase of EBV infection, at approximately 6 times greater in this group (OR: 6.15; CI (95%): 1.79–21.16; p: 0.0067). In contrast, the frequency of the DRB1*16 allele was greater in individuals with HIV who were in the past phase of EBV infection, at approximately 4 times greater in this group (OR: 4.19; CI (95%): 1.27–13.86; p: 0.0192) (Table 3).

We also evaluated whether these alleles are associated with several pathological markers of HIV infection (viral load, CD4+ T-cell count and CD8+ T lymphocyte count) (Figures 7A–H). We observed that carriers of the DRB1*09 allele in the primary phase of EBV infection had both higher log10 HIV viral loads (DRB1*09 = median: 5.67; IIQ: 5.56–5.76); (others = median: 4.45; IIQ: 3.85–5.33); p: 0.0176) (Figure 7A) and CD8(+) T lymphocyte counts (DRB1*09 = median: 2766.5; IIQ: 2506.5–2949.0); (others = median: 1024.5; IIQ: 834.5–2136.8); p: 0.0285) (Figure 7C) than carriers of other alleles. For individuals with past EBV infection and carriers of the DRB1*16 allele, no significant differences were observed in the quantification of these markers (Figures 7D–F).

Our next step was to assess whether the HIV load is associated with the serological profile of EBV infection and the viral load. We found that the log10 HIV viral load was lower in patients with past EBV infection (P.I. = median: 4.96; IIQ: 4.34–5.67); (S.T. = median: 5.06; IIQ: 4.49–5.52); (Past I. = median: 4.68; IIQ: 3.99–5.21); p = 0.0060) (Figure 7G).

Furthermore, positive linear regression was inferred between viral load in patients with primary EBV infection (R2: 0.22; p: 0.0412). There was also a negative trend in the viral load in patients who underwent serological transition, although the difference was not statistically significant (R2: 0.03; p = 0.1214) (Figure 7H).

The FPRP values for significant findings at different probability levels are shown in Table 4. In most findings, FPRP was notable in the range between 10 and 25% probability of association between alleles with serological profiles of EBV infection, which seems adequate for studies of genetic associations in general (25).

The low power of the statistical test is challenging if we consider the sampling obtained, suggesting additional validations using larger samples. However, the epidemiological aspects of control studies can lead to bias in the intended analyses.