IFNL4 Genotype Does Not Associate with CD4 T-Cell Recovery in People Living with Human Immunodeficiency Virus

Introduction

Approximately 15%–30% of people living with HIV infection who have a CD4 count <200 cells/mm³ at the time of treatment initiation fail to restore a CD4 count more than 350 cells/mm³ despite years of virologic suppression with antiretroviral therapy (ART), termed the immune non-responder (INR) phenotype. ^{1 –3} Such individuals are at increased risk for all-cause mortality, which may relate to higher levels of chronic immune activation and perturbed cellular homeostasis. ^{2,4 –6} The mechanisms for INR are not fully understood, and there are currently no interventions to improve CD4 T cell restoration in INRs. ²

Observations from other chronic infections, such as hepatitis C virus (HCV) infection, may provide insight into the pathogenesis of INR. For example, individuals with at least one ΔG allele (an exonic deletion allele) at the type-III interferon (IFN) lambda (IFNL4) rs368234815 locus, enabling production of full-length IFNL4, ⁷ are less likely to clear acute HCV infection, are less likely to respond to IFN-based therapy for HCV, have higher levels of chronic inflammation in chronic HCV infection, and develop hepatic fibrosis more slowly during chronic HCV infection. ^{7 –13} Conversely, individuals with rs368234815 TT/TT genotype, resulting in an inactive truncated IFNL4, ⁷ are more likely to clear HCV during acute infection, more likely to respond to IFN-based therapy, and develop hepatic fibrosis more rapidly during chronic infection. ^{7,10 –13} The influence of IFNLs on clinical outcomes for other acute and chronic viral infections has since been demonstrated in mouse and human studies. ^{14 –20}

Prior work identified an association of the IFNL4 genotype with HIV long-term non-progression and increased HIV-specific CD8 T cell activation, ²¹ an association not replicated in analysis of a subsequent cohort. ²² To our knowledge, the association of IFNL4 genotype and INR in people with HIV receiving suppressive ART has not been assessed. As IFNL4 may contribute to immune activation, we hypothesized that individuals carrying at least one rs368234815 ΔG allele would be more likely to be INRs than individuals homozygous for the TT allele. To address this hypothesis, we analyzed whether IFNL4 rs12979860 genotype correlates with INR in a group of well-characterized participants living with HIV from prior AIDS Clinical Trials Group (ACTG) trials in which immune responders and INRs were clearly defined. We used rs12979860, as it is in strong linkage disequilibrium with rs368234815, and data for rs12979860 were available based on direct genotyping or imputation from prior genome-wide studies.

Materials and Methods

In this study, we used data from four separate genotyping phases of specimens from ACTG studies in a combined dataset that comprises HIV treatment-naive participants at least 18 years of age enrolled in randomized treatment trials in the Unites States. Participants had enrolled into ACTG protocols 384, ²³ 388, ²⁴ A5014, A5095, ²⁵ A5142, ²⁶ A5202, ²⁷ and A5257. ²⁸ Informed consent for genetic research was obtained under ACTG protocol A5128. This project was approved by the Institutional Review Board at the Medical University of South Carolina.

For the purpose of this study, and to enrich for the INR phenotype and maximize the ability to detect a potential association, we defined an INR group of ART-naive participants who initiated ART in ACTG studies, had a pre-ART CD4 T-cell count <200 cells/mm³, achieved virologic suppression for at least 96 weeks, but still had fewer than 200 CD4 T cells cells/mm³ at week 96. The immune responder group was defined as participants with the same entry criteria who achieved a CD4 T cell count >350 cells/mm³ at week 96. Participants with CD4 T cell counts between 200 and 350 cells/mm³ at week 96 were excluded from analysis. Sustained virologic suppression was defined as having plasma HIV-1 RNA <200 copies/mL within 32 weeks of ART initiation, and at every subsequent study visit through week 96.

We identified 881 participants who met the criteria just mentioned, 596 of whom had genome-wide genotype data available, which had been generated and imputed, as described elsewhere. ^29,30 Of these 596 participants, 426 had available data for the rs12979860 single nucleotide polymorphism (SNP). In addition, we considered the first 10 principal components (PCs) of the full genotype data to address potential confounding due to unrecognized population stratification and to minimize the risk of identifying spurious associations.

The association of the rs12979860 SNP with INR was analyzed as the primary outcome by using logistic regression models. Clinical covariates of age, birth sex, and ethnicity were analyzed for association. No violation of Hardy–Weinberg Equilibrium was detected for rs12979860 (p = .28). As previously reported, ⁷ and confirmed by using National Cancer Institute LDPair analysis, rs12979860 and rs368234815 are in close linkage disequilibrium (R ²  = 0.83 for a combined group of African and African American people and R ² = 0.98 for people with European ancestry per https://ldlink.nci.nih.gov/?tab=ldpair, using rs12979860 and rs11322783, which was merged with rs368234815 October 12, 2018).

To assess whether the sample size was sufficient to detect an association, we employed a power analysis, based on simulation, to consider the impact of IFNL4 genotype observed in association with HCV outcomes. Specifically, we generated a hypothetical genotype dataset consisting of European ancestry, African American, and Hispanic populations by using the corresponding previously reported allele frequencies in a landmark HCV association study. ¹⁰ For each population, we generated corresponding phenotype data, assuming the odds ratio for the IFNL4 genotype impacting INR and HCV phenotypes might be similar. In this process, we kept the ratio of European American, African American, and Hispanic population the same as that of the ACTG dataset. We then applied the same association analysis procedure to the simulated data and repeated the process 1,000 times. Finally, we checked the proportion in which p-values for the association were ≤0.05 among these 1,000 iterations. This proportion was interpreted as the statistical power to detect the signal for the given sample size.

Results

The 426 participants whose immune response phenotype could be clearly defined and who had SNP data available included 338 responders and 88 INR. The group of participants was racially and ethnically diverse (174 white, 143 black, 100 Hispanic, 9 other), and 81% of participants were male (344 male, 82 female). Allelic frequency of all individuals, as well as based on race and gender, is shown in Table 1.

No association between rs12979860 and the INR was identified based on logistic regression models, considering rs12979860 CC genotype versus CT or TT genotype (Table 2). This conclusion remained valid when the first 3 or 10 PCs of the full genotype data were included in the logistic regression model to take into account population stratification (Table 2). Similarly, we obtained the same conclusion when age, sex, and ethnicity were included in the logistic model, in addition to the first 3 or 10 PCs (Table 2). Finally, when the data were analyzed as CC versus CT versus TT, no association with the INR was found (Table 2).

Our power analysis suggested that the study had at least 80% statistical power to detect a significant signal with a sample size of 60. This implies that our sample size of 426 participants should be ample to determine the existence of a significant association.

Discussion

This study failed to identify an association between the HIV INR phenotype and an SNP at the IFNL4 locus. The study was adequately powered to detect a difference based on the racial composition of our study participants and the odds ratio seen for the IFNL4 SNP in the previous HCV work by Ge et al., ¹⁰ if we presume the effect of the IFNL4 SNP on the INR phenotype to be similar.

Our rationale to examine the association of a polymorphism in IFNL4 with the INR phenotype stems from the observation that INR is associated with an elevated IFN signature and that the IFNL4 protein can impact IFN signaling and type-I IFN receptor levels. To this end, CD4 T cells from INRs have previously been shown to have higher IFN-stimulated gene expression, ³¹ suggesting a potentially higher state of endogenous IFN activity. Higher IFN gamma induced protein-10 (IP-10) production by monocytes has been inversely associated with CD4 T cell recovery, ³² and circulating IP-10 levels are higher in INR than people with immune reconstitution. ^{33 –35} The IFN can induce apoptosis of CD4 T-cells and thymocytes and decrease thymocyte proliferation, ^36,37 thereby limiting CD4 T cell recovery. INRs also have more activated CD4 and CD8 T cells but decreased HIV-specific CD8 T cells, ³⁸ which may be a consequence of increased IFN signaling. ³⁹ Indeed, blocking IFN signaling decreased T cell activation and increased HIV-specific CD8 T cells. ⁴⁰

The effect of an IFNL4 variant on other HIV-associated phenotypes has been previously explored. People with the IFNL4 rs368234815 ΔG allele, resulting in capacity to synthesize an active IFNL4 protein, are at increased risk for acquiring HIV by sexual activity ⁴¹ and by intravenous drug use. ⁴² People not taking ART, and who have the IFNL4 ΔG allele, have also been shown to be at increased risk of AIDS-defining illness such as tuberculosis, Pneumocystis jiroveci pneumonia, and cytomegalovirus retinitis. ^14,43 In contrast, those with an IFNL4 rs368234815 ΔG/ΔG genotype have been shown to be more likely to become long-term non-progressors in association with lower IP-10 and increased HIV-specific CD8 T cells. ²¹

Signaling mediated by IFNL4 upregulates USP18-ISG15, resulting in IFNAR1 and IFNAR2 desensitization. ^44,45 Thus, it is tempting to speculate that this desensitization may prevent IFN-mediated defenses at the time of HIV exposure, increasing the risk of acquisition. Similarly, this desensitization may dampen innate immune defense against pathogens, rendering people more susceptible to AIDS-defining illnesses. On the other hand, decreased immune activation and IFN-mediated thymic dysfunction may also facilitate CD4 maintenance ²¹ ; thus, the absence of IFNL4 signaling may result in more immune activation and the INR. Whatever impact IFNL4 has on IFN signaling pathways in the setting of chronic-treated HIV infection, however, it is insufficient to impact CD4 recovery based on our findings.

Limitations of the study include a reduction in sample size due to excluding participants with reconstituted CD4 counts between 200 and 350 cells/mm³, a choice made to maximize the odds of identifying an association by defining two clearly distinct clinical groups. We did not include some parameters such as duration of infection before ART, which may affect the INR phenotype, but which was unavailable in this dataset. We did not access biologic specimens to measure IFNL4 protein levels or to assess IFN signaling pathways. Finally, our statistical analysis assumed a modulating effect size for IFNL4 on the INR phenotype similar to that observed from HCV studies, which likely overestimates power.

We conclude that the IFNL4 genotype is not associated with the INR phenotype. Additional studies are needed to further understand the pathogenesis of the INR phenotype, the relationship to endogenous IFN signaling, and therapeutic approaches to augment immune recovery.