Effects of highly active antiretroviral therapy initiation on epigenomic DNA methylation in persons living with HIV

Abstract

Introduction: Highly active antiretroviral therapy (HAART) helps improve some measures of accelerated epigenetic aging in persons living with HIV (PLWH), but its overall impact on the epigenome is not fully understood. Methods: In this study, we analyzed the DNA methylation profiles of PLWH (n = 187) shortly before and approximately 2-3 years after they started HAART, as well as matched seronegative (SN) controls (n = 187), taken at two time intervals. Our aim was to identify specific CpGs and biologic pathways associated with HIV infection and initiation of HAART. Additionally, we attempted to identify epigenetic changes associated with HAART initiation that were independent of HIV-associated changes, using matched HIV seronegative (SN) controls (matched on age, hepatitis C status, and interval between visits) to identify CpGs that did not differ between PLWH and SN pre-HAART but were significantly associated with HAART initiation while being unrelated to HIV viral load. Epigenome-wide association studies (EWAS) on >850,000 CpG sites were performed using pre- and post-HAART samples from PLWH. The results were then annotated using the Genomic Regions Enrichment of Annotations Tool (GREAT). Results: When only pre- and post-HAART visits in PLWH were compared, gene ontologies related to immune function and diseases related to immune function were significant, though with less significance for PLWH with detectable HIV viral loads (>50 copies/mL) at the post-HAART visit. To specifically elucidate the effects of HAART separately from HIV-induced methylation changes, we performed EWAS of HAART while also controlling for HIV viral load, and found gene ontologies associated with transplant rejection, transplant-related diseases, and other immunologic signatures. Additionally, we performed a more focused analysis that examined CpGs reaching genome-wide significance (p < 1 × 10^-7) from the viral load-controlled EWAS that did not differ between all PLWH and matched SN controls pre-HAART. These CpGs were found to be near genes that play a role in retroviral drug metabolism, diffuse large B cell lymphoma proliferation, and gastric cancer metastasis. Discussion: Overall, this study provides insight into potential biological functions associated with DNA methylation changes induced by HAART initiation in persons living with HIV.

Keywords: DNA methylation; EWAS; HIV; antiretroviral therapy; bioinformatics; epigenetics.

FIGURE 1

Manhattan Plot of EWAS over the course of HAART initiation in Undetected PLWH Group. Plotted are the HAART-related CpGs in the Undetected PLWH group (plasma HIV VL < or = 50 copies/mL at the post-HAART visit, n = 120). Top CpGs are labeled with the names of proximal genes, with the x-axis representing the location of the CpG on the chromosome and the y-axis representing the signed -log10 p-value. The positive y-values (signed -log10 p-value) are CpGs that gain methylation with HAART initiation (blue), while the negative y-values are CpGs that lose methylation with HAART initiation (yellow).

FIGURE 2

Manhattan Plot of EWAS over the course of HAART Initiation in Detected PLWH Group. Plotted are the HAART-related CpGs in the Detected (plasma HIV VL > 50 copies/mL at the post-HAART visit, n = 67) PLWH group. Top CpGs are labeled with proximal genes, the x-axis represents the location of the CpG on the chromosome, and the y-axis represents the signed -log10 p-value. The positive y-values (signed -log10 p-value) are CpGs that gain methylation with HAART initiation (blue), while the negative y-values are CpGs that lose methylation with HAART initiation (gold).

FIGURE 3

Comparison of epigenome-wide CpG methylation in PLWH pre and post-HAART initiation to SN controls. The CpG methylation data across all sites were split into four groups: Undetected (solid blue) and Detected (solid orange) PLWH, as described in the text, and SN controls for Undetected (dotted blue) and Detected (dotted orange). “Control (Undetected)” refers to Undetected PLWH matched to SN controls and “Control (Detected)” refers to Detected PLWH matched to SN controls. Visit 1/Pre-HAART and Visit 2/post-HAART mean methylation values were plotted for each group along with error bars representing ±1 SE. Welch’s t-tests were conducted between undetected PLWH and their matched SN controls (p-value = 0.015) as well as between detected PLWH and their matched SN controls (p-value 1.3 × 10⁻⁶).

FIGURE 4

Genomic Regions Enrichment of Annotations Tool for Undetected PLWH groups. GREAT results for HAART-related CpGs in Undetected PLWH groups. The top 2000 positively-correlated CpGs (increased methylation) and the top 2000 negatively-correlated CpGs (decreased methylation) were taken from the EWAS of HAART initiation to use for the GREAT analysis. The graph with the red background shows the enrichment of the set of negatively-correlated CpGs, while the blue background is the enrichment of the set of positively-correlated CpGs. The significant annotations (y-axis) are plotted against the -log10 FDR-adjusted hypergeometric p-value (x-axis). The ontology databases are grouped by color.

FIGURE 5

Genomic Regions Enrichment of Annotations Tool for Detected PLWH groups. GREAT results for HAART-related CpGs in Detected PLWH groups. The top 2000 positively-correlated CpGs (increased methylation) and the top 2000 negatively-correlated CpGs (decreased methylation) were taken from the EWAS of HAART initiation to use for the GREAT analysis. The graph with the red background shows the enrichment of the set of negatively-correlated CpGs, while the blue background is the enrichment of the set of positively-correlated CpGs. The significant annotations (y-axis) are plotted against the -log10 FDR-adjusted hypergeometric p-value (x-axis). The ontology databases are grouped by color.

FIGURE 6

Genomic Regions Enrichment of Annotations Tool for annotations shared between Undetected and Detected PLWH groups. GREAT results for shared annotations between the Undetected and Detected HAART-related CpG GREAT results. Depicted are the Detected group negatively correlated HAART-related CpGs (red background), Undetected group negatively correlated HAART-related CpGs (green background), and Undetected group positively correlated HAART-related CpGs (blue background). The criteria for selecting these annotations were annotations that were shared between at least three sets of GREAT results that were also within the top two most significant results from each set. The significant annotations (y-axis) are plotted against the -log10 FDR-adjusted hypergeometric p-value (x-axis). The ontology databases are grouped by color.

FIGURE 7

Manhattan Plot of EWAS over the course of HAART Initiation in PLWH, controlling for HIV viral load. Plotted are the HAART-related CpGs in 177 PLWH, controlling for the log10-transformed HIV viral load value of each sample at each visit. Top CpGs are labeled with proximal genes, the x-axis represents the location of the CpG on the chromosome, and the y-axis represents the signed -log10 p-value. The positive y-values (signed -log10 p-value) are CpGs that gain methylation with HAART initiation (blue), while the negative y-values are CpGs that lose methylation with HAART initiation (gold).

FIGURE 8

Genomic Regions Enrichment of Annotations Tool for VL-controlled HAART Initiation EWAS CpGs. GREAT results for significant CpGs from VL-controlled HAART-related EWAS in PLWH (n = 177). The significant annotations (y-axis) are plotted against the -log10 FDR-adjusted hypergeometric p-value (x-axis). The ontology databases are grouped by color. The graph with the red background is the enrichment of the set of negatively-correlated CpGs (decreased methylation), while the blue background is the enrichment of the set of positively-correlated CpGs (increased methylation).

FIGURE 9

Methylation beta values across visits for two significant CpGs at the intersection of visit 1/pre-HAART equivalence and visit 2/post-HAART differences. Individual (thin lines) and mean (heavy lines) methylation values at visit 1 (pre-HAART in PLWH) and visit 2 (post-HAART in PLWH) are shown for (A) cg01009486 and (B) cg15391869. Undetected PLWH (blue), detected PLWH (orange), and SN (black) mean lines are shown on top of the mean lines for individual CpGs. Mean methylation was calculated at pre-HAART and post-HAART initiation by averaging the samples across all CpGs. Both CpGs shown here were found to be significantly affected by HAART initiation in the VL-controlled HAART EWAS that consisted of PLWH samples. Both CpGs were also able to reject the null equivalence hypothesis at an FDR-adjusted p-value < 0.05 and were unable to reject the null significance hypothesis at an FDR-adjusted p-value of < 0.05 in the TOST procedure.