Abstract
Background
People living with human immunodeficiency virus (HIV) infection have higher risk for chronic kidney disease (CKD), defined by a reduced estimated glomerular filtration rate (eGFR). Previous studies have implicated epigenetic changes related to CKD; however, the mechanism of HIV-related CKD has not been thoroughly investigated.
Methods
We conducted an epigenome-wide association study of eGFR among 567 HIV-positive and 117 HIV-negative male participants in the Veterans Aging Cohort Study to identify epigenetic signatures of kidney function.
Results
By surveying more than 400 000 cytosine guanine dinucleotide (CpG) sites measured from peripheral blood mononuclear cells, we identified 15 sites that were significantly associated with eGFR (false discovery rate Q value < 0.05) among HIV-positive participants. The most significant CpG sites, located at MAD1L1, TSNARE1/BAI1, and LTV1, were all negatively associated with eGFR (cg06329547, P = 5.25 × 10–9; cg23281907, P = 1.37 × 10–8; cg18368637, P = 5.17 × 10–8). We also replicated previously reported eGFR-associated CpG sites including cg17944885 (P = 2.5 × 10–5) located between ZNF788 and ZNF20 on chromosome 19 in the pooled population.
Conclusions
In this study we uncovered novel epigenetic associations with kidney function among people living with HIV and suggest potential epigenetic mechanisms linked with HIV-related CKD risk.
Keywords: VACS, EWAS, eGFR, renal function, HIV infection
In a large epigenome-wide association study of kidney function among human immunodeficiency virus (HIV)–positive and HIV-negative participants, we identified DNA methylation sites linked to chronic kidney disease risk, which improves the understanding of chronic disease risk among people living with HIV.
Human immunodeficiency virus (HIV) infection continues to be a major cause of morbidity and mortality around the world [1]. In 2016, approximately 36.7 million people worldwide were living with HIV, 1.8 million cases of incident HIV infection occurred, and 1.0 million people died from HIV-related diseases [2]. Although antiretroviral therapy (ART) is effective at suppressing viral replication and improves life expectancy, people with HIV (PWH) [2] are at increased risk for chronic inflammation, premature aging, and metabolic disorders [3] that are associated with renal impairment [4].
Chronic kidney disease (CKD), which emerged as a common complication of both HIV infection and its treatment, has been a critical cause of shortened life span in PWH [3]. The pathogenesis of HIV-related CKD is often multifactorial, linked to direct exposure to HIV viremia, superinfections, and the systemic immune response to infection [3, 5], as well as to traditional CKD risk factors such as increased body mass index (BMI), hypertension, diabetes, and cigarette smoking [6–9]. ART regimens may partially explain the higher burden of kidney dysfunction in PWH [10]. Apart from these observations, the underlying molecular mechanisms and pathophysiologic pathways of developing HIV-related CKD remain largely unknown. To further improve health outcomes for PWH, we need to better understand the molecular mechanisms that contribute to the progression and onset of HIV-related CKD [11] and identify useful biomarkers.
In addition to genetic factors, including common and rare genetic variants, epigenetic modifications have a role in disease susceptibility [12]. DNA methylation (DNAm), the epigenetic modification most studied at the population level, has been associated with kidney disease traits [13]. Previous epigenome-wide association studies (EWAS) identified numerous DNAm sites associated with estimated glomerular filtration rate (eGFR) [14–17], which is widely used to assess the filtration function of the kidneys and to diagnose CKD. However, the results were not consistent across studies, which included populations with CKD of varying causes and degrees of severity. Epigenetic associations with eGFR and CKD have not been investigated among PWH.
HIV infection has also been associated with epigenetic variants and DNAm. Several EWAS have identified and replicated DNAm sites associated with HIV infection and viremia [18]. A recent study also reported epigenetic associations with diabetes among PWH [19]. Epigenetic clock, an emerging aging biomarker based on age-related DNAm [20], has been linked with HIV infection and suggested accelerated epigenetic aging among PWH [21–23].
Characterizing DNAm patterns related to CKD caused by HIV infection and antiretroviral therapy may offer insights into its pathogenesis and the potential for prevention or treatment. We utilized the EWAS approach to identify differential methylation related to the risk for HIV-associated CKD. We examined previously identified eGFR-associated DNAm sites and discovered novel epigenetic associations with eGFR among PWH.
METHODS
Samples and Phenotypes
The phenotypic and epigenetic data were from the Veterans Aging Cohort Study (VACS), which is an observational, prospective study of veterans in care at the Department of Veterans Affairs medical centers across the United States [18]. VACS was approved by the human research protection program of Yale University and the institutional research board committee at the Connecticut veteran healthcare system, West Haven campus. All VACS participants provided written consent.
We included 567 HIV-positive and 117 HIV-negative male participants who had both phenotypic and epigenome-wide DNAm data available in the VACS. A questionnaire was completed by each participant at the baseline in order to collect clinical information, including the presence of chronic health conditions, information on cigarette smoking, use of medication, and antiretroviral treatment [23]. Total white blood cell counts and CD4 counts were enumerated at the time of peripheral whole blood sample collection [18, 23]. The eGFR was calculated from standardized creatinine levels, sex, race, and age using the CKD epidemiology collaboration equation [24]. All individuals with an eGFR less than 60 mL/min/1.73 m2 are classified as having CKD regardless of other evidence of kidney damage.
Epigenomic Data Generation, Processing, and Quality Control
Genomic DNA was extracted from whole blood samples using PAXGene collection tubes (QIAGEN, Hilden, Germany) and FlexiGene DNA extraction kits (QIAGEN). The blood samples for epigenetic analysis and estimation of eGFR were obtained at the same time point. The Illumina Infinium HumanMethylation450 Beadchip (450K; Illumina, San Diego, CA), which targets more than 480 000 DNAm sites (ie, cytosine guanine dinucleotide [CpG]) in the human genome, was utilized for epigenome-wide profiling at the Yale Center for Genomic Analysis. All samples were randomly placed on each array and across arrays to reduce batch effects.
Quality control procedures were performed to exclude problematic samples and CpG sites from analysis. The following steps were taken: (1) intensity values with a detection P value ≥ .001 were set to missing for each CpG site, and 927 CpG sites missing in more than 5% of the samples (ie, site-level missing rate >5%) were removed; (2) 35 605 probes were removed because they were within 10 base pairs from a single-nucleotide polymorphism (SNP); (3) 24 749 CpG sites were removed as they mapped to multiple genomic locations; and (4) all samples passed a call rate >95%. Following Illumina’s control probe scaling procedure, all raw intensity values were quantile normalized using limma package in R. Normalized intensity values were then used to generate a methylation score (β value = methylated allele intensity/(unmethylated allele intensity + methylated allele intensity + 100) for each CpG site, ranging from 0 representing unmethylation to 1 for complete methylation.
DNAm age was calculated using a web-based calculator (https://dnamage.genetics.ucla.edu/) with preselected age-related CpG sites and an algorithm developed by Horvath [25]. A t test was performed to assess the mean difference in DNAm age (ΔDNAm age = DNAm age – chronological age) between the HIV-positive and HIV-negative groups. Linear regressions were modeled to examine the association between DNAm age and chronological age, as well as eGFR.
Heterogeneous cell-type proportions across individuals is a well-established confounder in epigenetic epidemiological studies [26]. We calculated proportions of 6 cell types (CD4+ T cells, CD8+ T cells, natural killer T cells, B cells, monocytes, and granulocytes) in blood using an algorithm developed by Houseman et al [27] (R minfi package). The proportions of these 6 leukocyte subtypes were projected based on the top 100 cell-type–specific DNAm sites from a reference panel of known proportions. These estimated cell-type proportions were subsequently adjusted in the epigenetic association analyses.
Assessment and Adjustment of Potential Confounding Factors
We conducted multivariate linear regression to examine quantitative variables such as DNAm age, race, smoking status, diabetes, antihypertension drug usage, and 6 calculated cell types as predictors of eGFR, controlling for chronological age [28]. Variables with a P value < .05 were subsequently adjusted as covariates in the EWAS of eGFR. Previously established predictors of eGFR (eg, smoking status, diabetes) were controlled in the EWAS model regardless of their P values.
Principle Components Analysis
To adjust for population stratification, we performed a principal components (PC) analysis using an analytical approach developed by Barfield et al [29]. We created a pruned dataset that kept only CpG sites within 50 base pairs of a SNP to approximate the genome-wide genetic variants. We calculated the PCs to measure population stratification using the prcomp function in R package Factoextra. Then, we included the top 10 PCs as covariates in the EWAS of eGFR.
Statistical Methods
For autosomal analysis among PWH, we used linear mixed regression models for the effect of methylation status at individual CpG sites on eGFR with random effect for chip. The final adjusted models for the EWAS included age, race, BMI, average systolic blood pressure (SBP), hepatitis C virus (HCV) infection status, smoking status (current smoking vs noncurrent smoking), diabetes status, antihypertensive medication usage, calculated cell-type proportions, and top 10 PCs. For comparison, we performed epigenetic association analyses among the HIV-positive and HIV-negative groups separately using the same covariates. Also, we estimated the interaction between HIV status and CpG sites by running the same regression model with an additional interaction term (HIV *β value) in the combined study population. To correct for multiple testing, we applied a false discovery rate (FDR) Q value of 0.05 and the conservative Bonferroni-corrected P value of .05 (nominal P value < approximately 10–7 given 480 000 CpG sites) as significance thresholds. We compared our findings with those from 2 previously reported populations without HIV infection [15, 17]. We also separately examined DNAm of 11 232 CpG sites on the X chromosome within the pooled sample using the statistic model as previously described [30].
RESULTS
After quality control, the analysis dataset included 412 583 unique CpG sites in 567 HIV-positive and 117 HIV-negative participants. Characteristics of these groups are summarized in Table 1. All participants were males, with an average age of 52.0 ± 8.0 years. More participants in the HIV-positive group self-reported black race than in the HIV-negative group (85% vs 62%). HIV-positive participants had a higher rate of HCV infection (58% vs 28%) but lower BMI (25.5 vs 30.7 kg/m2) and lower prevalence of diabetes than HIV-negative participants. The eGFR was, on average, 3.4 mL/min/1.73 m2 higher in the HIV-positive group (Table 1 and Supplementary Figure 1), although this result was not statistically significant (P = .24).
Table 1.
Demographic and Clinical Characteristics of the Veterans Aging Cohort Study Participants
| Characteristic | Total (N = 684) | HIV Positive (n = 567) | HIV Negative (n = 117) | P Value |
|---|---|---|---|---|
| Estimated glomerular filtration rate (mL/min/1.73 m2) | 98.4 ± 32.9 | 99.0 ± 34.0 | 95.6 ± 26.9 | .243 |
| Chronic kidney disease | 56 (8%) | 47 (8%) | 9 (8%) | .964 |
| Age (years) | 52.0 ± 8.0 | 52.0 ± 7.8 | 53.2 ± 8.8 | .169 |
| Body mass index (kg/m2) | 26.4 ± 5.5 | 25.5 ± 4.7 | 30.7 ± 7.1 | <.0001 |
| Race/Ethnicity | … | … | … | <.0001 |
| White | 88 (13%) | 58 (10%) | 30 (26%) | … |
| Black | 554 (81%) | 481 (85%) | 73 (62%) | … |
| Hispanic | 24 (3.5%) | 15 (3%) | 9 (8%) | … |
| Other | 18 (2.6%) | 13 (2%) | 5 (4%) | … |
| Smoking status | … | … | … | .028 |
| None | 142 (20%) | 115 (20%) | 27 (23%) | … |
| Current | 376 (55%) | 324 (57%) | 52 (44%) | … |
| Past | 166 (24%) | 128 (23%) | 38 (23%) | ... |
| Antihypertensive medication | 497 (73%) | 402 (71%) | 95 (81%) | .207 |
| Hepatitis C virus | 362 (53%) | 329 (58%) | 33 (28%) | <.001 |
| Diabetes | 156 (23%) | 115 (20%) | 41 (35%) | .001 |
| Systolic BP | 129.3 ± 14.5 | 129.3 ± 14.7 | 129.3 ± 13.1 | .107 |
| Diastolic BP | 79.1 ± 9.5 | 79.4 ± 9.4 | 77.7 ± 10.3 | .169 |
| CD4 | ... | 425 ± 278 | ... | ... |
| Nadir CD4 counts | ... | 217 ± 7 | ... | ... |
| Antiretroviral therapya | ... | 472 (83%) | ... | ... |
| Nonnucleoside reverse transcriptase inhibitor–based regimen | ... | 205 (36.2%) | ... | ... |
| Protease inhibitor–based regimen | ... | 273 (48.1%) | ... | ... |
| Ritonavir | ... | 93 (16.4%) | ... | ... |
| Tenofovir | ... | 124 (21.9%) | ... | ... |
| HIV viral load (copies/mL) | ... | 75 (75–3400)b | ... | ... |
| Undetectable viral load (<200 copies/mL) | ... | 329 (58%) | ... | ... |
| DNA methylation age | 53.1 ± 8.6 | 53.6 ± 8.6 | 51.1 ± 8.2 | .003 |
Continuous variables were summarized as mean ± standard deviation, and categorical variables were summarized as count (%).
Abbreviations: BP, blood pressure; HIV, human immunodeficiency virus.
aNonnucleoside reverse transcriptase inhibitors–based regimen includes efavirenz, nevirapine, rilpivirine, doravirine; protease inhibitor–based regimen includes indinavir, nelfinavir, lopinavir, fosamprenavir, tipranavir, darunavir, atazanavir.
bMedian (interquartile range).
Race, BMI, antihypertensive medication use, SBP, and cell-type proportions were significantly associated with eGFR (Table 2). Because of previous evidence of eGFR association and epigenetic modification, we also controlled for diabetes, current smoking, and HIV infection status as potential confounders in the EWAS of eGFR.
Table 2.
Summary Statistics of Associations Between Covariates and Estimated Glomerular Filtration Rate Controlled for Age
| Variable | Effecta | Standard Error | P Value |
|---|---|---|---|
| Ageb | –0.50 | 0.16 | <.001 |
| Black race | 13.4 | 3.71 | <.001 |
| Body mass index | –0.35 | 0.04 | .129 |
| Current smoking | 2.3 | 3.11 | .467 |
| Antihypertensive drug use | –8.5 | 2.89 | .003 |
| HIV | 2.2 | 3.40 | .521 |
| Hepatitis C virus | 5.2 | 2.55 | .043 |
| Diabetes | –2.0 | 3.01 | .508 |
| CD4 | <0.01 | <0.01 | .573 |
| CD4 nadir | <0.01 | <0.01 | .615 |
| HIV viral load | <0.01 | <0.01 | .851 |
| Systolic blood pressure | –0.2 | 0.09 | .035 |
| Diastolic blood pressure | –0.1 | 0.13 | .506 |
| Cell-type proportion | |||
| CD8T cell | 48.2 | 15.77 | .002 |
| CD4T cell | 19.0 | 19.79 | .338 |
| Natural killer cell | 29.4 | 23.07 | .203 |
| B cell | 110.6 | 25.71 | <.001 |
| Monocyte | –28.0 | 34.12 | .413 |
| Granulocyte | –41.2 | 9.98 | <.001 |
| Antiretroviral therapy | –0.9 | 2.74 | .751 |
| DNA methylation age | 0.1 | 0.21 | .649 |
Abbreviation: HIV, human immunodeficiency virus.
aEffect is the beta coefficient from the linear regression model.
bThis regression model did not control for chronological age.
DNAm age was correlated with chronological age (r = 0.72, P < 2.2 × 10–6) (Table 1). DNAm age was, on average, 1 year older than chronological age among the pooled study population (P = .0002). The difference between chronological age and DNAm age (ie, ΔDNAm age) was significantly associated with HIV infection (P = 8.87 × 10–9), but the significance diminished after controlling for diabetes, current smoking, HCV infection, and cell-type proportions (P = .34). The ΔDNAm age was not significantly associated with eGFR before and after adjusting for all potential confounders in the pooled sample (P = .17 and .67).
Supplementary Figure 2a (Manhattan plot) and Table 3 present 15 CpG sites significantly associated with eGFR among PWH at the FDR <0.05 level after adjusting for covariates. A quantile–quantile (QQ) plot (Supplementary Figure 1b) comparing the observed P values to the expected P values for this EWAS indicates a low level of global inflation (inflation factor = 1.02); therefore, we did not further control for inflation. Manhattan plots and QQ plots of the EWAS among HIV-negative group and the pooled study population are shown in Supplementary Figure 2c–2f. Supplementary Figure 2d and 2f indicate a low level of global inflation for both the HIV-negative group and the whole study population (inflation factor = 1.00 and 1.03). The test statistics and annotations of 15 epigenome-wide significant CpG sites among PWH are summarized in Table 3. Thirteen of these 15 significant CpG sites were negatively associated with eGFR (ie, hypermethylation was associated with reduced renal function). However, the significance of these sites was absent among the HIV-negative group (Table 3). Thirteen of the 15 eGFR-associated CpG sites identified within the HIV-positive group were also statistically significant among the pooled samples (Table 3). At an alpha level of 0.05, only 2 of 13 CpG sites had significant interactions with HIV (cg06329547, P = .04; cg07796977, P = .04; cg07857040, P = .03).
Table 3.
Top-ranked Cytosine Guanine Dinucleotide Sites Exhibiting Differential DNA Methylation for Estimated Glomerular Filtration Rate
| Cytosine Guanine Dinucleotide Sites | Chromosome | Position | Gene | HIV Positive | HIV Negative | Total | Interaction P Value | ||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| β | SE | P Value | β | SE | P Value | β | SE | P Value | |||||
| cg06329547 | 7 | 1937739 | MAD1L1 | –260.6 | 44.4 | 8.28 × 10–9 | 358.6 | 482.5 | .46 | –253.8 | 42.9 | 5.25 × 10–9 | .04 |
| cg23281907 | 8 | 143509556 | TSNARE1/BAI1 | –228.5 | 40.6 | 3.1 × 10–8 | –133.3 | 148.9 | .37 | –219.9 | 38.2 | 1.37 × 10–8 | .19 |
| cg07796977 | 8 | 962467 | DLGAP2 | –456.2 | 81.1 | 3.12 × 10–8 | 488.6 | 319.7 | .13 | –414.7 | 77.4 | 1.20 × 10–7 | .04 |
| cg17355919 | 5 | 1331538 | CLPTM1L | –438.2 | 84.3 | 2.93 × 10–7 | –65.4 | 244.6 | .79 | –403.2 | 79.0 | 4.40 × 10–7 | .21 |
| cg14088090 | 1 | 43738485 | TMEM125 | –408.0 | 78.8 | 3.3 × 10–7 | –72.8 | 75.3 | .34 | –211.7 | 55.8 | 1.62 × 10–4 | .18 |
| cg10657250 | 4 | 156136570 | NPY2R | –531.0 | 103.3 | 3.94 × 10–7 | 389.4 | 250.6 | .12 | –459.8 | 95.0 | 1.62 × 10–6 | .10 |
| cg08198773 | 8 | 1697536 | DLGAP2/CLNB | –490.7 | 97.2 | 6.2 × 10–7 | –77.6 | 213.9 | .72 | –429.5 | 86.8 | 9.55 × 10–7 | .09 |
| cg18368637 | 6 | 144179905 | LTV1 | –181.4 | 36.4 | 8.93 × 10–7 | –166.0 | 95.0 | .09 | –185.8 | 33.7 | 5.17 × 10–8 | .17 |
| cg00296643 | 2 | 234371504 | DGKD | –416.9 | 84.1 | 9.93 × 10–7 | 219.0 | 373.7 | .56 | –378.5 | 79.5 | 2.39 × 10–6 | .28 |
| cg03658275 | 6 | 129285274 | LAMA2 | –162.0 | 33.2 | 1.39 × 10–6 | –88.0 | 58.8 | .14 | –150.7 | 28.9 | 2.53 × 10–7 | .09 |
| cg03012642 | 2 | 21346892 | APOB | –439.0 | 90.0 | 1.44 × 10–6 | 49.5 | 390.8 | .90 | –432.6 | 85.2 | 5.12 × 10–7 | .15 |
| cg11796565 | 19 | 13205681 | NFIX | 89.8 | 18.5 | 1.52 × 10–6 | 28.4 | 37.7 | .45 | 79.9 | 16.6 | 1.74 × 10–6 | .07 |
| cg17344906 | 19 | 13202507 | NFIX | 115.3 | 23.7 | 1.61 × 10–6 | 46.9 | 49.0 | .34 | 101.1 | 21.1 | 2.06 × 10–6 | .36 |
| cg17461641 | 1 | 164023978 | … | –322.7 | 66.6 | 1.68 × 10–6 | –232.8 | 124.4 | .07 | –309.3 | 59.0 | 2.19 × 10–7 | .03 |
| cg07857040 | 16 | 1582219 | IFT140 | –349.4 | 72.3 | 1.77 × 10–6 | 7.1 | 165.8 | .97 | –285.3 | 65.2 | 1.40 × 10–5 | .10 |
The epigenetic association model was adjusted for age, race, body mass index, average systolic blood pressure, HIV-infection, hepatitis C virus, smoking, diabetes, antihypertensive medication use, cell-types proportions, and top 10 principle components.
False discovery rate Q < 0.05. Interaction P value is the P value for the cytosine guanine dinucleotide × HIV status term in the multiple regression model.
Abbreviation: HIV, human immunodeficiency virus; SE, standard error.
Using a more stringent Bonferroni-corrected cutoff (corrected P < .05, nominal P < approximately 10–7), 3 CpG sites, cg06329547, cg23281907, and cg07796977, remained significant. These 3 CpG sites (Supplementary Figure 3a–3c) are located within genic regions of mitotic arrest deficient like 1 (MAD1L1), T-SNARE domain containing 1/ brain-specific angiogenesis inhibitor 1 (TSNARE1/BAI1), and DLG associated protein 2 (DLGAP2), respectively. CpG site cg06329547 (MAD1L1) was hypermethylated among patients with lower eGFR, with a 1% increase in β value associated with a 2.61 unit decrease in eGFR (95% confidence interval [CI], –3.48, –1.74; P = 5.25 × 10–9). CpG site cg23281907, located between TSNARE1 and BAI1, was inversely associated with eGFR, with a 1% increase in β value associated with a 2.29 unit decrease in eGFR (95% CI, –3.08, –1.49; P = 1.37 × 10–8). CpG site cg07796977 (DLGAP2) was also inversely associated with eGFR, with a 1% increase in β value associated with a 4.56 unit decrease in eGFR (95% CI, –6.51, –2.97; P = of 3.12 × 10–8). Such significant associations were absent in the HIV-negative group. The regional plots (Supplementary Figure 3a–3c) showed all tested CpG sites in the neighboring regions of these 3 sites and additional association signals with eGFR within these regions. There were 35 (6.2%), 14 (4.8%), and 13 (4.2%) CpG sites with P values < .05 in regions for cg06329547, cg23281907, and cg07796977, respectively. No statistically significant CpG sites were identified in the X chromosome.
We attempted to replicate eGFR-associated CpG sites identified in previous EWAS. Smyth et al [15] reported 23 genes with at least 1 CpG site significantly associated with eGFR. However, Chu et al [17] reported a lack of replication of Smyth’s findings and 18 novel eGFR-associated sites in a much larger EWAS of eGFR, including 2264 participants from the Atherosclerosis Risk in Communities Study and 2595 participants from the Framingham Heart Study. Only 1 site reported by Smyth et al, cg17500228 in gene exocyst complex component 3 (EXOC3), showed an association with P < .05 (Smyth, P = 1.57 × 10–14; Chu, P = .0002; VACS, P = .0095). We evaluated 17 CpG sites reported by Chu et al (1 CpG site was excluded after quality control) in our study of HIV-positive and HIV-negative participants. We replicated cg17944885 located between zinc finger family member 788 (ZNF788) and zinc finger protein 20 (ZNF20; P = 2.5 × 10–5) after multiple testing correction in the pooled population (Table 4). There were 5 additional sites with P values < .05 (cg23591762, cg16428517, cg19942083, cg06158227, and cg27660627). All 17 associations in our study showed directionality consistent with those reported by Chu et al [17].
Table 4.
Replication Analysis of Estimated Glomerular Filtration Rate–associated DNA Methylation Identified in Non–human Immunodeficiency Virus Samples
| Cytosine Guanine Dinucleotide Sites | Chromosome | Position | Gene | Meta-analysis of ARIC and FHS | VACS | VACS | VACS | |||
|---|---|---|---|---|---|---|---|---|---|---|
| HIV Positive | HIV Negative | Total | ||||||||
| P Value | β | P Value | β | P Value | β | P Value | ||||
| cg17944885 | 19 | 12225735 | ZNF788/ZNF20 | 1.228 × 10–23 | –169.00 | 5 × 10–4 | –165.10 | .050 | –176.9 | 2.5 × 10–5 |
| cg23597162 | 7 | 28102341 | JAZF1 | 2.82 × 10–19 | –135.07 | 8 × 10–4 | –181.78 | .004 | –130.9 | 1 × 10–4 |
| cg04036920 | 11 | 33562503 | C11orf41 | 2.387 × 10–15 | –38.29 | .309 | 10.97 | .894 | –42.0 | .211 |
| cg00501876 | 3 | 39193251 | CSRNP1 | 3.623 × 10–15 | 13.92 | .827 | –240.05 | .045 | –20.7 | .703 |
| cg00994936 | 19 | 1423902 | DAZAP1 | 3.259 × 10–11 | 76.82 | .131 | 10.07 | .917 | 70.7 | .112 |
| cg16428517 | 16 | 3317428 | MEFV/ZNF263 | 1.166 × 10–10 | –193.11 | .016 | –248.22 | .058 | –224.2 | .001 |
| cg11950754 | 1 | 53782077 | LRP8 | 2.006 × 10–10 | –23.88 | .630 | –20.63 | .775 | –30.1 | .460 |
| cg12065228 | 1 | 19652788 | PQLC2 | 2.196 × 10–10 | 33.58 | .242 | 33.21 | .602 | 33.9 | .186 |
| cg10750182 | 10 | 73497514 | C10orf105 | 4.962 × 10–10 | –4.99 | .943 | –297.96 | .056 | –23.2 | .706 |
| cg09022230 | 7 | 5457225 | TNRC18 | 5.222 × 10–10 | 14.70 | .744 | 28.57 | .727 | 13.6 | .727 |
| cg19942083 | 12 | 7070562 | PTPN6/PHB2 | 7.174 × 10–10 | 75.67 | .017 | 45.80 | .595 | 80.9 | .005 |
| cg06158227 | 15 | 43662311 | TUBGCP4 | 8.482 × 10–10 | –230.16 | .008 | –3.75 | .978 | –196.4 | .007 |
| cg27660627 | 16 | 89461803 | ANKRD11 | 9.939 × 10–10 | –175.70 | .002 | 45.92 | .618 | –125.6 | .010 |
| cg12116137 | 17 | 1576449 | PRPF8 | 5.277 × 10–9 | –8.32 | .755 | 91.34 | .028 | 5.8 | .796 |
| cg04460609 | 4 | 16532808 | LDB2 | 7.854 × 10–9 | –27.98 | .401 | –64.95 | .343 | –31.5 | .283 |
| cg19497511 | 2 | 238609807 | LRRFIP1 | 3.451 × 10–8 | –14.49 | .772 | 16.16 | .888 | –2.7 | .951 |
| cg22515589 | 17 | 79426432 | BAHCC1 | 6.437 × 10–8 | 25.79 | .748 | 43.31 | .713 | 26.9 | .682 |
Abbreviations: ARIC, Atherosclerosis Risk in Communities Study; FHS, Framingham Study; HIV, human immunodeficiency virus; VACS, Veterans Aging Cohort Study.
Discussion
HIV type 1 viremia and its corresponding treatment are known to be responsible for the development of CKD and decreased eGFR [4]. Previous reports demonstrated that HIV infection was associated with DNAm age [31]. We hypothesized that epigenetic aging may contribute to aging-related pathologies of CKD. DNAm age, a marker of biological aging, was significantly older than the chronological age (P = .0002). However, the difference between DNAm age and chronological age was not associated with either HIV or eGFR after adjustment for covariates, suggesting that epigenetic aging might not be an independent path between HIV and eGFR.
For the EWAS of eGFR among PWH, we identified 15 significant CpG sites, including 3 significant sites (located in or close to MAD1L1, TSNARE1/BAI1, and DLGAP2) that passed Bonferroni correction. MAD1L1 functions as a homodimer and codes mitotic-arrest deficient 1 (MAD1) protein, which inhibits maturation and expansion of T-lymphocytes, and is targeted by human T-cell leukemia virus type 1 during virus transformation [32]. Defects in MAD1L1 are involved in the development and/or progression of various types of cancer including kidney carcinomas [33]. Reduction of MAD1 may play a role in modulating specific immune responses to HIV, which might be involved in the mechanisms of HIV-associated immune-mediated kidney diseases. Interestingly, both MAD1L1 and TSNARE1 were estimated to be associated with susceptibility to schizophrenia and bipolar disorder [34, 35]. Schizophrenia has been associated with a nearly 25% increase in the risk of developing CKD [36]. Both MAD1L1 and TSNARE1 might be key effectors in the development of schizophrenia-related CKD, which might have pathogenetic pathways that are similar to those of other types of CKD. BAI1 appears to be a mediator of the p53 signal in suppression of glioblastoma [37]. Coherently, the tumor suppressor protein p53 strongly alters HIV type 1 replication [38].
Though the cause–consequence relationships among methylation of these genes, kidney function, and neurological diseases are unclear, it is possible that some of the observed epigenetic associations are consequences of CKD rather than predictors. On the other hand, DNAm of several identified genes, including laminin subunit alpha 2 (LAMA2), apolipoprotein B (APOB), and neuropeptide Y receptor Y2 (NPY2R), are associated with metabolic diseases [39–41] that can contribute to renal impairment.
Similar to the recent study by Chu et al [17], we had limited success in replicating the CKD-associated CpG sites reported by Smyth et al, except for cg17500228 in gene EXOC3. EXOC3 was previously established to be a key effector in HIV-1 Nef protein-mediated enhancement of nanotube formation [42]. It has been suggested that the Nef protein plays a pathogenetic role in HIV-associated nephropathy [3]. Of 17 significant sites implicated in the EWAS by Chu et al, cg17944885 (ZNF788/ZNF20) showed a significant association with eGFR in our study. Five other candidate sites showed associations with a P value < .05 in our study, lending support to protein tyrosine phosphatase, nonreceptor type 6 (PTPN6) and ankyrin repeat domain 11 (ANKRD11) as promising candidate genes for further experimental evaluation [17]. In our replication analysis, the lack of significant association but with consistent directionality of previously implicated CpG sites could be due to lack of statistical power (particularly among the HIV-negative group), differences in the severity or causes of CKD, differences in confounding, or a combination of these factors.
A potential limitation of our study is that we measured DNAm in peripheral blood rather than in kidney tissues. Direct evaluation of DNAm in kidney tissues is not feasible in human population–based studies [15]. Recently, similar trait-associated methylation of CpG sites in blood and target tissues has been found, supporting the use of DNAm in blood as an indicator of methylation in other tissues [17]. The 450K array platform restricts the epigenome-wide coverage compared to the newer array (eg, EPIC850K array) or whole methylome sequencing and is limited to measure 5-mC, not other types of modifications such as 5-hmC. Additionally, the etiology among HIV-associated CKD can be heterogeneous, which may increase false negatives in EWAS.
Our study is a first step in investigating the relationships among HIV infection, DNAm, and CKD. The VACS provided a unique opportunity to investigate the epigenomic association with eGFR among PWH. We replicated previously reported associations of eGFR with methylation of EXOC3, PTPN6, and ANKRD11, which supports the belief that CKD in PWH may have similar causes as in the HIV-negative population. The identified epigenetic associations of MAD1L1, TSNARE1/BAI1, and DLGAP2 suggest that there may be some unique mechanism for HIV-related CKD. Further studies to examine the role of these epigenetic changes in the complex mechanism for development and progression of HIV-related CKD are warranted.
Supplementary Data
Supplementary materials are available at Clinical Infectious Diseases online. Consisting of data provided by the authors to benefit the reader, the posted materials are not copyedited and are the sole responsibility of the authors, so questions or comments should be addressed to the corresponding author.
Notes
Acknowledgments. The authors thank all veterans who participated in this project.
Financial support. This work was supported in part by the Emory University Center for AIDS Research (P30AI050409) and National Institutes of Health grants (K01HL134147, R03DA039745, and K12DA000167).
Potential conflicts of interest. All authors: No reported conflicts of interest. All authors have submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest. Conflicts that the editors consider relevant to the content of the manuscript have been disclosed.
References
- 1. Wang H, Wolock TM, Carter A, et al. . Estimates of global, regional, and national incidence, prevalence, and mortality of HIV, 1980–2015: the Global Burden of Disease Study 2015. Lancet HIV 2016; 3:e361–87. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2. UNAIDS fact sheet 2017. UNAIDS 2017 Available at: http://www.unaids.org/sites/default/files/media_asset/UNAIDS_FactSheet_en.pdf. Accessed 02 July 2018.
- 3. Cohen SD, Kopp JB, Kimmel PL. Kidney diseases associated with human immunodeficiency virus infection. N Engl J Med 2017; 377:2363–74. [DOI] [PubMed] [Google Scholar]
- 4. Nadkarni GN, Konstantinidis I, Wyatt CM. HIV and the aging kidney. Curr Opin HIV AIDS 2014; 9:340–5. [DOI] [PubMed] [Google Scholar]
- 5. Nobakht E, Cohen SD, Rosenberg AZ, Kimmel PL. HIV-associated immune complex kidney disease. Nat Rev Nephrol 2016; 12:291–300. [DOI] [PubMed] [Google Scholar]
- 6. De Wit S, Sabin CA, Weber R, et al. ; Data Collection on Adverse Events of Anti-HIV Drugs (D:A:D) Study Incidence and risk factors for new-onset diabetes in HIV-infected patients: the data collection on adverse events of anti-HIV drugs (D:A:D) Study. Diabetes Care 2008; 31:1224–9. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7. Seaberg EC, Munoz A, Lu M, et al. . Association between highly active antiretroviral therapy and hypertension in a large cohort of men followed from 1984 to 2003. Aids 2005; 19:953–60. [DOI] [PubMed] [Google Scholar]
- 8. Míguez-Burbano MJ, Wyatt C, Lewis JE, Rodríguez A, Duncan R. Ignoring the obvious missing piece of chronic kidney disease in HIV: cigarette smoking. J Assoc Nurses AIDS Care 2010; 21:16–24. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9. Nishijima T, Kawasaki Y, Mutoh Y, et al. . Prevalence and factors associated with chronic kidney disease and end-stage renal disease in HIV-1-infected Asian patients in Tokyo. Sci Rep 2017; 7:14565. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10. Mocroft A, Kirk O, Reiss P, et al. ; EuroSIDA Study Group Estimated glomerular filtration rate, chronic kidney disease and antiretroviral drug use in HIV-positive patients. AIDS 2010; 24:1667–78. [DOI] [PubMed] [Google Scholar]
- 11. Keller BJ, Martini S, Sedor JR, Kretzler M. A systems view of genetics in chronic kidney disease. Kidney Int 2012; 81:14–21. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12. McCaughan JA, McKnight AJ, Courtney AE, Maxwell AP. Epigenetics: time to translate into transplantation. Transplantation 2012; 94:1–7. [DOI] [PubMed] [Google Scholar]
- 13. Geisel J, Schorr H, Heine GH, et al. . Decreased p66Shc promoter methylation in patients with end-stage renal disease. Clin Chem Lab Med 2007; 45:1764–70. [DOI] [PubMed] [Google Scholar]
- 14. Wing MR, Devaney JM, Joffe MM, et al. ; Chronic Renal Insufficiency Cohort (CRIC) Study DNA methylation profile associated with rapid decline in kidney function: findings from the CRIC Study. Nephrol Dial Transplant 2014; 29:864–72. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15. Smyth LJ, McKay GJ, Maxwell AP, McKnight AJ. DNA hypermethylation and DNA hypomethylation is present at different loci in chronic kidney disease. Epigenetics 2014; 9:366–76. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16. Sapienza C, Lee J, Powell J, et al. . DNA methylation profiling identifies epigenetic differences between diabetes patients with ESRD and diabetes patients without nephropathy. Epigenetics 2014; 6:20–8. [DOI] [PubMed] [Google Scholar]
- 17. Chu AY, Tin A, Schlosser P, et al. . Epigenome-wide association studies identify DNA methylation associated with kidney function. Nat Commun 2017; 8:1286. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18. Zhang X, Justice AC, Hu Y, et al. . Epigenome-wide differential DNA methylation between HIV-infected and uninfected individuals. Epigenetics 2016: 1–11. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19. Mathur R, Hui Q, Huang Y, et al. . DNA methylation markers of type 2 diabetes mellitus among male veterans with or without human immunodeficiency virus infection. 2019. [DOI] [PMC free article] [PubMed]
- 20. Horvath S. DNA methylation age of human tissues and cell types. Genome Biol 2013; 14: R115. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21. Horvath S, Levine AJ. HIV-1 Infection accelerates age according to the epigenetic clock. J Infect Dis 2015; 212: 1563–73. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22. Rickabaugh TM, Baxter RM, Sehl M, et al. . Acceleration of age-associated methylation patterns in HIV-1-infected adults. PloS One 2015; 10: e0119201. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23. Nelson KN, Hui Q, Rimland D, et al. . Identification of HIV infection-related DNA methylation sites and advanced epigenetic aging in HIV-positive, treatment-naive U.S. veterans. AIDS 2017; 31: 571–5. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24. Levey AS, Stevens LA, Schmid CH, et al. . A new equation to estimate glomerular filtration rate. Ann Intern Med 2009; 150: 604–12. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25. Horvath S. DNA methylation age of human tissues and cell types. Genome Biol 2013; 14: R115–R. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26. Adalsteinsson BT, Gudnason H, Aspelund T, et al. . Heterogeneity in white blood cells has potential to confound DNA methylation measurements. PLoS One 2012; 7: e46705. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27. Houseman EA, Kelsey KT, Wiencke JK, Marsit CJ. Cell-composition effects in the analysis of DNA methylation array data: a mathematical perspective. BMC Bioinformatics 2015; 16: 95. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28. Bekaert B, Kamalandua A, Zapico SC, Van de Voorde W, Decorte R. Improved age determination of blood and teeth samples using a selected set of DNA methylation markers. Epigenetics 2015; 10: 922–30. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29. Barfield RT, Almli LM, Kilaru V, et al. . Accounting for population stratification in DNA methylation studies. Genetic Epidemiol 2014; 38: 231–41. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30. Klebaner D, Huang Y, Hui Q, et al. . X chromosome-wide analysis identifies DNA methylation sites influenced by cigarette smoking. Clinical Epigenet 2016; 8: 20. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 31. Horvath S, Levine AJ. HIV-1 infection accelerates age according to the epigenetic clock. J Infect Dis 2015; 212: 1563–73. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 32. Jin DY, Spencer F, Jeang KT. Human T cell leukemia virus type 1 oncoprotein Tax targets the human mitotic checkpoint protein MAD1. Cell 1998; 93: 81–91. [DOI] [PubMed] [Google Scholar]
- 33. Pinto M, Vieira J, Ribeiro FR, et al. . Overexpression of the mitotic checkpoint genes BUB1 and BUBR1 is associated with genomic complexity in clear cell kidney carcinomas. Cellular Oncol 2008; 30: 389–95. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 34. Ruderfer DM, Fanous AH, Ripke S, et al. . Polygenic dissection of diagnosis and clinical dimensions of bipolar disorder and schizophrenia. Molecular psychiatry 2014; 19: 1017–24. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 35. Sleiman P, Wang D, Glessner J, et al. . GWAS meta analysis identifies TSNARE1 as a novel Schizophrenia / Bipolar susceptibility locus. Sci Rep 2013; 3: 3075. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 36. Tzeng N-S, Hsu Y-H, Ho S-Y, et al. . Is schizophrenia associated with an increased risk of chronic kidney disease? A nationwide matched-cohort study. BMJ Open 2015; 5. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 37. Mathema VB, Na-Bangchang K. Regulatory roles of brain-specific angiogenesis inhibitor 1(BAI1) protein in inflammation, tumorigenesis and phagocytosis: A brief review. Crit Rev Oncol Hematol 2017; 111: 81–6. [DOI] [PubMed] [Google Scholar]
- 38. Duan L, Ozaki I, Oakes JW, Taylor JP, Khalili K, Pomerantz RJ. The tumor suppressor protein p53 strongly alters human immunodeficiency virus type 1 replication. J Virol 1994; 68: 4302–13. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 39. Andrade RC, Nevado J, de Faria Domingues de Lima MA, et al. . Segmental uniparental isodisomy of chromosome 6 causing transient diabetes mellitus and merosin-deficient congenital muscular dystrophy. Am J Med Genet A 2014; 164a: 2908–13. [DOI] [PubMed] [Google Scholar]
- 40. Kristiansson K, Perola M, Tikkanen E, et al. . Genome-wide screen for metabolic syndrome susceptibility Loci reveals strong lipid gene contribution but no evidence for common genetic basis for clustering of metabolic syndrome traits. Circ Cardiovasc Genet 2012; 5: 242–9. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 41. Wei Z, Zhang K, Wen G, et al. . Heredity and cardiometabolic risk: naturally occurring polymorphisms in the human neuropeptide Y(2) receptor promoter disrupt multiple transcriptional response motifs. J Hypertens 2013; 31: 123–33. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 42. Mukerji J, Olivieri KC, Misra V, Agopian KA, Gabuzda D. Proteomic analysis of HIV-1 Nef cellular binding partners reveals a role for exocyst complex proteins in mediating enhancement of intercellular nanotube formation. Retrovirology 2012; 9: 33. [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.