Killer immunoglobulin-like receptor (KIR) genes are associated with the risk of episodes of high-level and detectable viremia among HIV controllers

Study subjects and ethical issues

The individual description of the HICs analyzed in this study had already been published (de Azevedo et al., 2017; Côrtes et al., 2018; Caetano et al., 2020). Individuals who fulfilled the criteria of HICs were identified in a cohort of HIV-1 seropositive individuals followed at the Instituto Nacional de Infectologia Evandro Chagas, Fundação Oswaldo Cruz (INI/FIOCRUZ), Rio de Janeiro, Brazil. These individuals were defined as subjects over 18 years old with documented HIV-1 infection for >3 years and RNA viral load below the detection limit (40 or 80 copies/mL, depending on the year of inclusion) in the absence of antiretroviral treatment. As this is a longitudinal study, these individuals were/have been followed up and classified under one of the categories of viral control.

After 12 years of follow up, a cohort of 28 HICs could be composed including HIV-1-infected individuals classified in three categories according to the plasmatic viral load (VL): (1) persistent elite controllers (PECs) if 100% of VL measures were below the limit of detection (< LD; 50-80 copies/mL) for the respective available commercial assays (n = 7); (2) ebbing elite controllers (EECs) if subjects had occasional (≤ 30%) episodes of transient low-level (> LD to 400 copies/mL) viremia (n = 7); and (3) viremic controllers (VCs) if most (≥ 70%) of VL determinations were between 51 and 2,000 copies/mL (n = 14). Occasional VL measurements above the upper limits were accepted for the EECs and VCs. Participants were followed at least once every 6-12 months, starting in November 2008, to perform HIV-RNA VL quantification and CD4⁺ T lymphocytes counts. In each visit, whole blood was collected into an ethylenediaminetetraacetic acid (EDTA)-containing tube, and 1 mL was stored at -20 °C until use. At this moment, from the 28 individuals [median follow-up = 9.02 years (interquartile range (IQR) = 6.46)], 11 are currently being followed up and 17 are no longer study participants due to cART entry (n = 09) or loss of follow up (n = 08).

All participants provided written informed consent, and the ethical committee of Instituto Nacional de Infectologia Evandro Chagas (INI-FIOCRUZ) approved the study (CAAE 1717.0.000.009-07). The corresponding documents can be found as extended data (de Sá & Teixeira, 2021).

Genomic DNA extraction

DNA was extracted from whole blood using the QIAamp DNA Blood Mini Kit (QIAGEN, Hilden, Nordrhein-Westfalen, Germany) according to the manufacturer’s instructions. The DNA concentration was determined using the Thermo Scientific Nanodrop 2000 (Thermo Fisher Scientific, Waltham, Massachusetts, USA), and the filtrates containing the isolated DNA were stored at −20 °C until used for genomic analysis.

HLA typing

High-resolution HLA-B and -C alleles typing was performed by automated nucleotide sequencing (sequencing-based typing — SBT) according to the manufacturer’s instructions on the ABI platform using commercial kits (SECORE Sequencing kit, Invitrogen by Life Technologies, Brown Deer, Wisconsin, USA). HLA-B and -C alleles were assigned using a four-digit designation utilizing uTYPE^® v6.0 SBT software (Invitrogen by Life Technologies, Brown Deer, Wisconsin, USA). The grouping of HLA-B alleles in HLA Bw4 and/or Bw6 epitopes associated specificities followed the Immuno Polymorphism Database (IPD)-International Immunogenetics Project (IMGT)/HLA nomenclature guidelines (Robinson et al., 2013). The grouping of HLA-C genes in C1 (HLA-C*01/*03/*07/*08/*12/*14/*16) and C2 (HLA-C*02/*04/*05/*06/*15/*17/*18) epitope-associated specificities was based in the classification currently used in the literature (Mandelboim et al., 1996; Faridi & Agrawal, 2011; de Sá et al., 2020).

KIR genotyping

Qualitative analysis of KIR genes was performed using a commercial kit based on sequence-specific primer amplification methods — SSP (SSP KIR Genotyping Kit, Invitrogen, Brown Deer, Wisconsin, USA). A total of 14 KIR genes and 2 KIR pseudogenes (2DL1, 2DL2, 2DL3, 2DL4, 2DL5, 2DS1, 2DS2, 2DS3, 2DS4, 2DS5, 3DL1, 3DL2, 3DL3, 3DS1, 2DP1, and 3DP1) were screened using this approach. KIR AA and Bx genotypes designation followed the current working definition, which characterizes these genotypes based on the combinations of haplotype A (absence of all the activating genes, except KIR2DS4) and haplotype B (presence of one or more of the activating genes) (Uhrberg et al., 1997; Fernandes-Cardoso et al., 2016).

Single nucleotide polymorphism selection and genotyping

We selected four single nucleotide polymorphisms (SNPs) in 3 inflammasome genes based on previously published data (Pontillo et al., 2013) and considering the relevance of each gene in the inflammasome pathway: CARD8 (Caspase Activation and Recruitment Domains-8) rs2043211 and rs6509365; NLRP3 (Nucleotide oligomerization domain–Like Receptor family, Pyrin domain-containing protein-3) rs10754558; and IL-1β rs1143634. Besides that, we selected HLA-C rs9264942 C>T (TaqMan^® genotyping assay C_29901957_10) based on the association of this SNP with a lower HIV-1 viral load (Fellay et al., 2007, 2009; Wei et al., 2015). SNP genotyping was performed using commercially available TaqMan assays (Applied Biosystems/AB and Life Technologies) using the ABI7500 Real-Time platform (AB). Allelic discrimination was carried out by means of the Thermo Fisher Connect Software^® version 1.

CCR5 genotyping

DNA samples were PCR amplified to determine the presence of the CCR5 Δ32 mutation. Primers (CCR5-F: GCT GTC TTT GCG TCT CTC CCA GGA and CCR5-R: CTC ACA GCC CTG TGC CTC TTC TTC) were used to amplify a 239 base-pair (bp) fragment covering the Δ32 mutation region. The cycling conditions were: 1 cycle 94°C 5 minutes; 30 cycles 94°C 1 minute, 60°C 30 seconds, 72°C 2 minutes; 1 cycle 72°C 10 minutes. The PCR amplified products (5 ul) were separated by agarose gel electrophoresis and visualized by ethidium bromide staining. A single band of 239 bp indicated the CCR5/CCR5 wild-type genotype, while the heterozygous genotype CCR5/Δ32 was detected by 239 and 207 bp bands, and the homozygous genotype Δ32/Δ32 by a single band of 207 bp.

Statistical analyses

Direct count estimated the frequencies of HLA, KIR, SNPs and CCR5 alleles and genotypes. Kruskal-Wallis ANOVA by Ranks test was used to compare the sociodemographic, clinical, and laboratory characteristics among the different HICs groups for continuous numerical variables. In contrast, for categorical nominal variables, Fisher's exact test was used in the evaluation of frequencies among the different HICs groups. Tests of equal proportions were used to evaluate the allele carriers’ relative frequencies among the HICs groups. For comparative purposes, HLA-B genomic distribution data from the general Brazilian population were extracted from REDOME (National Registry of Bone Marrow Donors, Brazilian Ministry of Health) data (Rede Brasil de Imunogenética, 2018), and for HLA-C we used data available at Allele Frequencies Net Database.

To estimate the hazard of a transitory loss of virological control, defined as the observation of the first VL determination above 2,000 copies/mL, we calculated person-years (pY) at risk for each patient between HIV diagnosis and the occurrence of an event. Individuals were censored either at the time of final observation, death, cART (combined antiretroviral therapy) initiation, or the last follow-up visit/exam, whichever occurred first. The effect of various risk factors on the outcome was assessed using hazard ratios (HR) and corresponding 95% confidence intervals (CI), which were estimated through the Cox proportional hazard model (Therneau & Grambsch, 2000).

To evaluate the occurrence of multiple events of VL blips, herein defined as the number of detectable viral loads counted after HIV diagnosis and until cART initiation, we calculated the incidences/rates of the number of viral loads per pY for each patient, and 95% CI were estimated according to asymptotic standard errors calculated from a Gamma distribution (Lehmann & Casella, 1998). The effect of various risk factors on the outcome was assessed by relative risks (RR) and corresponding 95% CI were estimated employing the Negative Binomial (NB) models. Individual exposure time (in years) was used as an offset in the NB models (Hilbe, 2011).

Two-tailed levels of significance ≤ 0.01, 0.05 and 0.1 were considered “highly significant”, “significant” and “suggestive”, respectively. All statistical analysis was performed using the R statistical software package, version 3.4.1 (R Development Core Team, 2017).

Clinical and epidemiological characteristics of HICs

Table 1 depicts the main clinical and epidemiological characteristics of the HICs cohort distributed according to the viral control groups (de Sá & Teixeira, 2021). Overall, we found a similar frequency of females and males (16 [57.1%] and 12 [42.9%], respectively), while 21 (75%) of the individuals have identified themselves as heterosexual and 6 (21.4%) as men who have sex with men (MSM). Most heterosexual male (75%) and MSM subjects (83.3%) belonged to the VCs group. The ethnic background composition was diverse, with an unequal distribution of White, Brown, and Black participants among groups of HICs (P = 0.04). Participants had a median age of 32 (IQR 10.4) years old, with a similar median age observed among the PECs (33 years old; IQR 6.9), EECs (37 years old; IQR 13.9), and VCs (30 years old; IQR 9.8) groups.

All HICs groups displayed a median CD4⁺ T cells count above 800 cells/μL (IQR 431). Further details about immunological and virological characteristics of most HICs of this study were depicted in previous studies from our group (Bello et al., 2009; Côrtes et al., 2015, 2018; de Azevedo et al., 2017; Caetano et al., 2020).

Distribution of HLA-B, -C, KIR, and CCR5 genes among HICs

Table 2 describes the relative frequencies of HLA-B and -C, KIR, and CCR5∆32 alleles carriers found in the different HICs groups. Due to the dependence between the genotypes, we did not perform this comparative analysis for the SNPs included in this study. However, the complete genetic characterization of the analyzed markers is depicted in Table 3. We observed significant differences between PECs and EECs for protective HLA-B alleles, KIR Bx genotype, KIR2DL5, and KIR2DS5 allelic carrier frequencies. Among EECs and VCs, the frequencies of KIR Bx genotype, KIR2DL3 + C1 pair, KIR2DL2, KIR2DL5, and KIR2DS2 alleles carriers were also significantly distinct. The only genetic marker with significantly different frequencies between PECs and VCs was KIR2DS5. It is noticeable that HLA-B protective alleles (HLA-B*27, B*52, and/or B*57) were found in 71% of the PECs, 43% of the VCs, and only 14% of the EECs.

To check for an enrichment of protective HLA alleles in our cohort, we compared the classical protective HLA-B and -C frequencies here obtained with those from the general Brazilian population, and with HIV-1-infected individuals who progressed to AIDS (Teixeira et al. 2014; Biberg-Salum et al., 2018), considered as HIV non-controllers (HIV-NC) (Table 4). For HLA-B, we noted that the frequency of carriers of these alleles in our HICs cohort (0.464; n = 13) is on average 3.34-fold higher (P < 0.001) than in the overall Brazilian population (0.139; n = 396,740) and 3.22-fold higher (P = 0.014) in the HIV non-controllers group (0.144; n = 29). Similarly, for HLA-C we noted that the frequency of carriers of these alleles in our HICs cohort (0.571; n = 16) is on average 3.60-fold higher (P < 0.001) than in the overall Brazilian population (0.159; n = 42) and 3.50-fold higher (P < 0.001) than in the HIV non-controllers group (0.161; n = 26). Only two PECs carried the CC genotype in the HLA-C rs9264942 polymorphism. The KIR Bx genotype was found in 86% of PECs, 79% of the VCs, and 29% of EECs. The KIR3DL1+Bw4 genotype was detected in only two VCs, and the KIR3DS1+Bw4 genotype in four participants (one PEC, one EEC, and two VCs). All PECs and EECs had the KIR2DL3 allele. The CCR5∆32 mutation was found in four heterozygous (WT/∆32) participants (one PEC, one EEC, and two VCs), resulting in an overall allelic frequency of 7.1%, and no homozygous (∆32/∆32) participants were identified in this cohort. Therefore, we cannot assign any relation between this genetic marker and the distinct patterns of viremia control observed in these HICs.

Episodes of high-level and detectable viremia among HICs

To test the potential relevance of genetic host factors in the risk of a transitory loss of virological control in some HICs, we conducted Cox proportional hazard multivariate analysis to identify factors associated with the occurrence of the high-level viremia (> 2,000 copies/mL) (Table 5). A higher mean hazard of this event was observed for HICs without the KIR2DL3 allele [aHR = 79.098 (3.236-1933.264); P = 0.007], HICs with the HLA-C*08 allele [aHR = 8.379 (1.181-59.472); P = 0.034], HICs with the C/G genotype [aHR = 7.462 (2.294-2505.478); P = 0.042] and G/G genotype [aHR = 75.817 (2.294-2505.478); P = 0.015] in the NLRP3 polymorphism (rs10754558), and HICs without the KIR2DL3 + C1/C2 pair [aHR = 11.78 (1.602-86.619); P = 0.015] than for their counterparts. In the other way, a lower mean hazard of this event was observed for HICs with the following alleles: KIR2DL5 [aHR = 0.131 (0.023-0.738); P = 0.021], KIR2DS1 [aHR = 0.109 (0.014-0.867); P = 0.036], KIR2DS5 [aHR = 0.058 (0.007-0.466); P = 0.007], and KIR3DS1 [aHR = 0.108 (0.013-0.876); P = 0.037] when compared with their counterparts.

The relative risk of occurring any event of detectable viremia (VL blip) was also evaluated by means of the negative binomial (NB) multivariate models (Table 6). Overall, HICs with the KIR2DS5 allele [aRR = 0.294 (0.122-0.705); P = 0.006] had less risk of undergoing VL blips within the same normalized period than those participants without this allele. Conversely, HICs without the KIR2DL3 allele [aHR = 4.987 (1.174-21.196); P = 0.03] had a mean higher risk of experiencing VL blips within the same normalized period than those with this allele. The results of Cox's proportional hazard multivariate analysis and Negative Binomial (NB) models of the HLA-C and inflamassome SNPs are depicted in Tables 7 and 8.

Our results point to a protective role for KIR2DL3 and KIR2DS5 alleles since the absence of the former was associated both with a higher risk of a transitory loss of virological control and to a higher risk of any event of detectable viremia (VL blip); while the presence of the latter was associated to a lower risk of these events. Epidemiological features were introduced in the multivariate models for controlling bias in the statistical analysis. The limited sample size and the study design do not allow us to explore the associations involving these epidemiological features depicted in Tables 5 and 6.

Underling data

Open Science Framework: Killer immunoglobulin-like receptor (KIR) genes are associated with the risk of episodes of high-level and detectable viremia among HIV controllers. https://doi.org/10.17605/OSF.IO/ZGQPA (de Sá & Teixeira, 2021).

This project contains the following underlying data:

Extended data

Open Science Framework: Killer immunoglobulin-like receptor (KIR) genes are associated with the risk of episodes of high-level and detectable viremia among HIV controllers. https://doi.org/10.17605/OSF.IO/ZGQPA (de Sá & Teixeira, 2021).

This project contains the following extended data:

Data are available under the terms of the Creative Commons Zero “No rights reserved” data waiver (CC0 1.0 Public domain dedication).