Epigenome-wide association study identifies DNA methylation sites associated with target organ damage in older African Americans

Figures & data

Table 1. Characteristics of participants in GENOA^a (N = 961)

	Mean (SD) or N (%)	Heritability (h^2, 95% CI)
Female	685 (71.3%)
Age (years)	62.7 (9.7)
Age at phase I (years)	57.7 (10.3)
Time between phases I and II	5.2 (1.3)
BMI (kg/m^b)	31.8 (6.8)
Systolic blood pressure (mmHg)	137.8 (21.1)
Diastolic blood pressure (mmHg)	79.3 (11.1)
Antihypertensive medication use
No	284 (29.6%)
Yes	662 (68.9%)
Missing	15 (1.6%)
Smoking status
Current	118 (12.3%)
Former	257 (26.7%)
Never	571 (59.4%)
Missing	15 (1.6%)
Diabetes status^b
No	653 (68.0%)
Yes	289 (30.1%)
Missing	19 (2.0%)
Estimated glomerular filtration rate (eGFR, mL/min/1.73m^b)	89.9 (21.2)	0.31 (0.11–0.50)
Urine albumin-to-creatinine ratio (UACR, mg/g)	48.4 (192.9)	0.49 (0.30–0.68)
Left ventricular mass index (LVMI, g/height^b.7)	39.1 (10.3)	0.52 (0.35–0.70)
Relative wall thickness (RWT)	0.32 (0.05)	0.30 (0.12–0.46)
White matter hyperintensity (WMH, cm^3)	9.49 (8.0)	0.21 (0–0.43)
Total intracranial volume (cm^3)	1373 (134.7)

SD: standard deviation; CI: confidence interval; BMI: body mass index.

Sample sizes for target organ damage measures: eGFR (n = 940), UACR (n = 943), LVMI (n = 910), RWT (n = 915), WMH (n = 539).

^aMeasured at phase II unless stated otherwise.

^bDefined as glucose level ≥ 126 mg/dL or self-reported physician-diagnosed diabetes and current anti-diabetes medication use.

Table 2. Statistically significant CpG sites associated with TOD traits at FDR q < 0.1 using univariate and multivariate models in GENOA

	Model 1 – univariate
Outcome measure	CpG site	Gene	Chr	Relation to CpG site	Relation to gene	p-value	FDR q	Beta (SE)
RWT	cg03042953	SSBP3	1	Island	Body	9.53 × 10^−8	0.072	1.361 (0.253)
LVMI	cg21134922		5			1.59 × 10^−8	0.013	0.663 (0.116)
UACR	cg04816311	C7orf50	7	North Shore	Body	5.52 × 10^−8	0.044	0.513 (0.094)
					Model 1 – multivariate ^a
					Beta (SE) for each TOD trait
CpG site	Gene	Chr	Relation to CpG site	Relation to gene	eGFR	UACR	LVMI	RWT	p-value	FDR q
cg21134922		5			−0.035 (0.106)	0.124 (0.118)	0.663 (0.117)	0.004 (0.113)	6.54 × 10^−7	0.044
cg04816311	C7orf50	7	North Shore	Body	0.178 (0.089)	0.492 (0.097)	0.390 (0.098)	0.093 (0.094)	1.06 × 10^−7	0.042
cg09155024		10			0.188 (0.057)	−0.070 (0.063)	−0.042 (0.064)	0.269 (0.060)	3.35 × 10^−7	0.044
cg10254690	OAT	10	Island	TSS1500^b	−0.154 (0.064)	−0.044 (0.071)	−0.359 (0.070)	−0.062 (0.068)	2.75 × 10^−7	0.065
cg07660512		12			−0.527 (0.162)	−0.795 (0.179)	−0.003 (0.182)	−0.242 (0.173)	2.62 × 10^−7	0.044
cg12661888	IFT43	14		Body	−0.306 (0.072)	0.259 (0.080)	−0.117 (0.081)	0.019 (0.077)	4.68 × 10^−7	0.053
cg02264946	CATSPERD	19		Body	0.236 (0.125)	0.203 (0.137)	−0.489 (0.137)	0.341 (0.132)	6.16 × 10^−8	0.042

Chr: chromosome; FDR: false discovery rate; UACR: urinary albumin to creatinine ratio; LVMI: left ventricular mass index; RWT: relative wall thickness; eGFR: estimated glomerular filtration rate.

Model 1 is adjusted for age at phase II, sex, time between measurement, smoking status, and 4 genetic principal components.

^aMultivariate model includes eGFR, UACR, LVMI, and RWT as outcome measures.

^bWithin 1500 kb of the gene start site (promoter region).

Table 3. Adjusted multivariate models for statistically significant CpG sites from Model 1 in GENOA

		Model 2 – multivariate^a						Model 3 – multivariate^a
		Beta (SE) for each TOD trait						Beta (SE) for each TOD trait
CpG site	Gene	eGFR	UACR	LVMI	RWT	p-value	FDR q	eGFR	UACR	LVMI	RWT	p-value	FDR q
cg21134922		−0.022 (0.011)	0.093 (0.013)	0.611 (0.012)	0.001 (0.012)	8.54 × 10^−7	0.096	0.004 (0.011)	0.076 (0.012)	0.421 (0.010)	−0.022 (0.012)	6.52 × 10^−4	0.912
cg04816311	C7orf50	0.179 (0.008)	0.428 (0.009)	0.312 (0.009)	0.058 (0.009)	1.74 × 10^−6	0.172	0.141 (0.008)	0.243 (0.008)	0.148 (0.007)	−0.025 (0.008)	1.49 × 10^−2	0.990
cg09155024		0.176 (0.003)	−0.096 (0.004)	−0.063 (0.004)	0.249 (0.004)	7.49 × 10^−7	0.096	0.164 (0.003)	−0.123 (0.003)	0.018 (0.003)	0.242 (0.004)	3.03 × 10^−6	0.289
cg10254690	OAT	−0.164 (0.004)	−0.038 (0.005)	−0.340 (0.004)	−0.060 (0.004)	6.88 × 10^−7	0.096	−0.166 (0.004)	−0.224 (0.004)	−0.039 (0.004)	−0.008 (0.004)	3.08 × 10^−4	0.878
cg07660512		−0.559 (0.026)	−0.800 (0.029)	0.020 (0.029)	−0.268 (0.029)	3.49 × 10^−8	0.028	−0.548 (0.025)	−0.703 (0.027)	0.114 (0.024)	−0.227 (0.028)	8.00 × 10^−8	0.036
cg12661888	IFT43	−0.312 (0.005)	0.236 (0.006)	−0.137 (0.006)	−0.002 (0.006)	4.96 × 10^−7	0.096	−0.319 (0.005)	0.225 (0.005)	−0.122 (0.005)	−0.006 (0.006)	6.21 × 10^−7	0.098
cg02264946	CATSPERD	0.219 (0.015)	0.261 (0.017)	−0.416 (0.017)	0.380 (0.017)	8.48 × 10^−8	0.034	0.224 (0.015)	0.280 (0.015)	−0.338 (0.014)	0.396 (0.017)	9.19 × 10^−7	0.036

FDR: false discovery rate; SE: standard error; eGFR: estimated glomerular filtration rate; UACR: urinary albumin to creatinine ratio; LVMI: left ventricular mass index; RWT: relative wall thickness.

Model 2 is adjusted for age at phase II, sex, time between measurement, smoking status, 4 genetic principal components, systolic blood pressure, diastolic blood pressure and antihypertensive medication use.

Model 3 is adjusted for Model 2 variables plus body mass index and type 2 diabetes status.

Results with a significant FDR q < 0.1 are in bold font.

^aMultivariate model includes eGFR, UACR, LVMI, and RWT as outcome measures.

Table 4. Associations between statistically significant CpG sites from multivariate Model 1 and nearby gene expression (± 250 kb, p-value < 0.1) in GENOA

CpG site	Gene	Protein type	Beta	SE	p-value
cg04816311	AC073957.1	LncRNA	−0.156	0.050	0.002
	AC073094.1	LncRNA	−0.183	0.060	0.003
	AC091729.2	LncRNA	−0.127	0.057	0.026
	UNCX	Protein-coding	−0.108	0.058	0.063
cg10254690	OAT	Protein-coding	−0.163	0.057	0.005
cg07660512	MORN3	Protein-coding	−0.096	0.044	0.031
	AC079360.1	LncRNA	−0.060	0.028	0.035
	HPD	Protein-coding	−0.085	0.043	0.050
cg02264946	AC011444.1	LncRNA	−0.095	0.044	0.030
	SAFB2	Protein-coding	−0.147	0.073	0.045

SE: standard error.

Table 5. Mendelian randomization results showing the inverse-variance weighted effects of multiple SNPs used as instrumental variables in the association of CpG sites and TOD traits in GENOA

TOD trait	Chr	CpG site	Independent SNPs with CpG-SNP FDR q < 0.1	Causal effect estimate ^a	Causal effect estimate SE	Lower bound of causal estimate	Upper bound of causal estimate	p-value
Estimated glomerular filtration rate (eGFR), mL/min/1.73 m^2	5	cg21134922	6	8.718	6.2	−3.432	20.869	0.16
	7	cg04816311	15	8.844	3.584	1.819	15.868	0.014
	10	cg10254690	10	−5.752	2.594	−10.836	−0.669	0.027
	12	cg07660512	13	−14.912	7.321	−29.262	−0.563	0.042
Urinary albumin-to-creatinine ratio (UACR), mg/g	5	cg21134922	6	0.075	0.485	−0.876	1.026	0.877
	7	cg04816311	15	0.285	0.254	−0.213	0.783	0.261
	10	cg10254690	10	0.359	0.222	−0.076	0.793	0.106
	12	cg07660512	13	−0.57	0.571	−1.69	0.55	0.319
Left ventricular mass index (LVMI), g/height^2.7	5	cg21134922	6	6.292	3.299	−0.173	12.757	0.056
	7	cg04816311	15	1.816	1.919	−1.946	5.578	0.344
	10	cg10254690	10	−2.485	1.378	−5.186	0.216	0.071
	12	cg07660512	13	−5.694	3.689	−12.925	1.537	0.123
Relative wall thickness (RWT)	5	cg21134922	6	0.004	0.016	−0.027	0.036	0.780
	7	cg04816311	15	0.015	0.009	−0.002	0.031	0.084
	10	cg10254690	10	−0.002	0.007	−0.015	0.011	0.798
	12	cg07660512	13	0	0.02	−0.04	0.039	0.994

Chr: chromosome; SE: standard error.

^aThe causal effect estimate is the pooled estimate calculated using inverse-variance weighted models from the Mendelian randomization analyses and is interpreted as the effect per 1 unit change in DNA methylation (M-value) using genetic variants on target organ damage trait.