Epigenome-wide association study of Chinese monozygotic twins identifies DNA methylation loci associated with estimated glomerular filtration rate

Abstract

Background

DNA methylation (DNAm) has been shown in multiple studies to be associated with the estimated glomerular filtration rate (eGFR). However, studies focusing on Chinese populations are lacking. We conducted an epigenome-wide association study to investigate the association between DNAm and eGFR in Chinese monozygotic twins.

Methods

Genome-wide DNAm level was detected using Reduced Representation Bisulfite Sequencing test. Generalized estimation equation (GEE) was used to examine the association between Cytosine-phosphate-Guanines (CpGs) DNAm and eGFR. Inference about Causation from Examination of FAmiliaL CONfounding was employed to infer the causal relationship. The comb-p was used to identify differentially methylated regions (DMRs). GeneMANIA was used to analyze the gene interaction network. The Genomic Regions Enrichment of Annotations Tool enriched biological functions and pathways. Gene expression profiling sequencing was employed to measure mRNA expression levels, and the GEE model was used to investigate the association between gene expression and eGFR. The candidate gene was validated in a community population by calculating the methylation risk score (MRS).

Results

A total of 80 CpGs and 28 DMRs, located at genes such as OLIG2, SYNGR3, LONP1, CDCP1, and SHANK1, achieved genome-wide significance level (FDR < 0.05). The causal effect of DNAm on eGFR was supported by 12 CpGs located at genes such as SYNGR3 and C9orf3. In contrast, the causal effect of eGFR on DNAm is proved by 13 CpGs located at genes such as EPHB3 and MLLT1. Enrichment analysis revealed several important biological functions and pathways related to eGFR, including alpha-2A adrenergic receptor binding pathway and corticotropin-releasing hormone receptor activity pathway. GeneMANIA results showed that SYNGR3 was co-expressed with MLLT1 and had genetic interactions with AFF4 and EDIL3. Gene expression analysis found that SYNGR3 expression was negatively associated with eGFR. Validation analysis showed that the MRS of SYNGR3 was positively associated with low eGFR levels.

Conclusions

We identified a set of CpGs, DMRs, and pathways potentially associated with eGFR, particularly in the SYNGR3 gene. These findings provided new insights into the epigenetic modifications related to the decline in eGFR and chronic kidney disease.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12967-025-06067-4.

Background

Chronic kidney disease (CKD) is characterized by a persistent decline in kidney function lasting for at least three months, regardless of the underlying cause [1]. In addition to possibly advancing to kidney failure and leading to death, CKD is also a major risk factor for common cardiovascular diseases, posing a significant public health burden worldwide [2].

Estimated glomerular filtration rate (eGFR) is a common indicator for assessing renal function, and its decline signals the onset of CKD. The eGFR is likely influenced by both environmental factors and genetic factors [3–5]. Previous studies indicated its heritability was approximately 44% [6]. Genome-wide association studies (GWAS) have successfully identified multiple genetic variants associated with eGFR; however, these variants explain only 5.08% of the variance in eGFR, suggesting that other genetic regulatory mechanisms, such as epigenetics, may play a role in eGFR [7–9].

Epigenetics refers to heritable trait changes that occur without alterations in the DNA sequence [10, 11]. DNA methylation (DNAm) is the earliest and most well-studied form of epigenetic modification and has been reported to be associated with eGFR in several studies [12, 13]. One study identified 19 eGFR-associated Cytosine-phosphate-Guanines (CpGs) and found that TF EBF1 binding sites were enriched for these CpGs. Low EBF1 expression led to glomerular maturation defects and reduced eGFR [14]. Another study showed three CpGs located at AGTR1 and PRKCA were associated with eGFR, which can lead to kidney injury [15]. Differences in DNAm patterns may also occur between races due to differences in environment and genetic background. However, there is little research regarding this topic among the Chinese population. Trait-discordant monozygotic twin's design, which can control twins' genetic information, is an effective tool for performing epigenome-wide association study (EWAS) [16]. In this case, we can more accurately identify DNAm due to different environmental factors. Therefore, it is necessary to perform EWAS in Chinese monozygotic twins.

Furthermore, traditional EWAS does not allow for causal inference due to the limitations of the cross-sectional study design. The Inference about Causation from Examination of FAmiliaL CONfounding (ICE FALCON) has been used to infer causal relationships between DNAm levels and traits based on a twin design [17]. Consequently, we preliminary explored the causal relationship between DNAm and eGFR using ICE FALCON.

Therefore, we performed EWAS in this study to investigate the CpGs, differentially methylated regions (DMRs) and biological pathways associated with eGFR using Chinese monozygotic twins. Additionally, we conducted causal inference to clarify the causal association between eGFR and DNAm. The results were further validated in community populations and integrated with gene expression data.

Methods

Participants

Recruitment of the study population and information collection have been described previously [18]. Specifically, between 2012 and 2013, twins were recruited from residential communities in Qingdao through household registration, medical records, and media announcements. A total of 244 pairs of monozygotic twins were collected. Participants underwent a questionnaire survey, physical examination, and venous blood sample collection after fasting for 10–12 hours. Participants were excluded from the study if they met any of the following criteria: missing serum creatinine data, insufficient DNA sample quality, use of medications that affect eGFR, pregnancy or breastfeeding, inability to complete the survey due to health conditions, or a history of heart failure, kidney failure, or cancer. Based on the above criteria, a total of 67 twin pairs remained for the study. The eGFR was calculated by age, sex, and blood creatinine using the CKD-EPI creatinine equation [19]. Intra-pair eGFR difference ≥ 0.6 ml/min/1.73m² monozygotic twins were selected. A total of 61 pairs of intact monozygotic twins with discordant eGFR were included in the methylation analysis, after excluding one pair with an outlier eGFR value, and five pairs with an eGFR difference of less than 0.6 ml/min/1.73m². In addition, we analyzed the relationship between gene expression levels and eGFR using a randomly selected subsample of 12 monozygotic twin pairs.

Reduced representation bisulfite sequencing data preparation

Total DNA was extracted from venous blood samples and analyzed using Reduced Representation Bisulfite Sequencing (RRBS) to obtain the DNA data. It combines restriction enzyme cleavage and bisulfite treatment, allowing the researchers to efficiently analyze methylation patterns in CpG-rich island regions of the genome. RRBS test is often used in epigenetic research to help reveal the role of DNAm in gene regulation and disease. A total of 551,447 CpGs were finally incorporated. Bismark software was used to map the raw reads to the Genome Reference Consortium Human Build 37 (GRCh37, hg19) [20]. The BiSeq package was then used to smooth the data to determine methylation levels [21]. Coverage was controlled to the 90% quartile, and CpGs with an average methylation β-value below 0.01 or more than 10 missing values were excluded. If one individual in a twin pair is missing a value, the other individual in the pair would be excluded from the analysis of that CpG, which counted as two missing values. 10 refers to the number of individuals, not twin pairs. Ultimately, the β-values for methylation were converted into M-values through a log₂ transformation [22]. As a result, a total of 247,162 CpGs remained after quality control.

Cell-type composition estimation

Different cell types result in different DNAm patterns. The ReFACTor method was used to address the confounding bias of various cell types on whole blood DNAm analysis [23]. ReFACTor is a tool designed to estimate the proportion of different cell types in mixed cell samples, resolving cellular composition through principal component analysis and linear modeling [24]. It does not require external reference data, provides accurate deconvolution of cell types in complex samples, and is widely used in DNAm analysis. In this study, the first five principal components were selected as covariates to be included in the analysis to correct for bias caused by cell type.

Gene expression data

mRNA extraction was performed on whole peripheral blood. Subsequently, RNA-Seq libraries were constructed and sequenced. The resulting sequence data were mapped to the human genome by TopHat₂ [25]. Cufflinks were then used to calculate FPKM values to quantify gene expression levels [26].

Statistical analysis

Epigenome-wide association analysis

The generalized estimation equation (GEE) models were used to estimate the relationship between DNAm levels at each CpG and eGFR, incorporating age, sex, and the first five cellular principal components as covariates in the equation. In order to identify twin pair data, we added a vector to GEE that represents the different twin pair numbers. False discovery rate (FDR) values were used to correct for bias due to multiple testing. Genome-wide significance was defined as FDR < 0.05 [27]. BiomaRt package was used to annotate the CpGs to the nearest gene [28]. The bacon package was used to estimate the inflation factor (lambda value) and to plot quantile–quantile (Q-Q) plot [29].

Causal inference

ICE FALCONE was used to infer whether there was a causal relationship between the methylation level of individual CpG (P < 1 × 10^–5) and eGFR. ICE FALCONE is a causal inference method for twin family data [17]. By controlling for shared family factors, causality could be more accurately identified. The GEE model was used to calculate β_self, β_co-twin, β’_self, and β’_co-twin. β_self represented the overall correlation, including both family confounding and causal proportions, while β_co-twin estimated the family confounding proportion. β’_self and β’_co-twin were derived from the full model. A causal relationship was considered to exist if the absolute value of the difference between the β_co-twin and β’_co-twin was greater than the absolute value of the difference between the β_self and β’_self of the two individuals within the pair. Conversely, there was no causality, and the association was caused by family confounding. We conducted two rounds of causal inference in our study. In the first round, methylation data served as the exposure and eGFR data as the outcome, while in the second round, the roles of exposure and outcome were reversed.

Differentially methylated regions analysis

The comb-p was used to detect DMRs associated with eGFR [30]. Stouffer-Liptak-Kechris (slk) method was used for correcting significantly enriched DMRs, and corrected P < 0.05 were considered relevant methylated regions.

GeneMANIA database analysis

GeneMANIA was used to analyze the interaction network between genes involved in eGFR-related CpGs and DMRs. GeneMANIA was used to explore the interactional network of genes where CpGs are located, determine the priority of genes, and visualize the results. GeneMANIA is a user-friendly and flexible website that leverages extensive genomics and proteomics data to perform functional assessments for given genes and visualize the results, including three main aspects: hypothesizing gene functions, analyzing gene lists, and determining gene prioritization [31].

Genomic region enrichment analysis

The Genomic Regions Enrichment of Annotations Tool (GREAT) was employed to perform genomic region enrichment analyses to find relevant biological pathways [32]. The CpGs (P < 0.05) were uploaded to the GREAT online platform using the default "base plus expansion" association rules. Annotations were performed based on GRCh37. Pathways with FDR < 0.05 were considered statistically significant in the analysis.

Gene expression analysis

The association between mRNA expression level and eGFR was analyzed using the GEE model, with age and gender as covariants. The FDR was used to correct for bias due to multiple testing.

Quantitative methylation analysis of SYNGR3

Based on the annotation results of the top CpGs in the EWAS, the correlation between gene expression levels and eGFR, the causal relationship with eGFR, and the primer design results, we selected SYNGR3 as the validation gene for the community population.

We randomly recruited 74 cases and 148 controls from the community for the case–control study. Cases were defined as eGFR < 90 ml/min/1.73m², and others were classified as controls. We excluded participants who were pregnant or breastfeeding, unable to complete the survey due to physical conditions, or had cardiovascular disease or tumors. Subsequently, we matched participants based on age and sex frequency distributions. Participants first underwent blood sample collection. Participants underwent an epidemiological survey and physical examination when blood samples were collected and stored at − 80 °C for methylation analysis. We designed eGFR primers for SYNGR3 to cover most of the sites. Methylation ratios were determined using MassARRAY EpiTYPER software (Agena Bioscience, San Diego, California) to obtain β-values, which were then converted to M-values using log₂ transformation. A total of 22 CpGs were quantified using the Sequenom MassARRAY platform. Independent samples t-tests and Wilcoxon rank-sum tests compared DNAm levels between groups. The methylation risk score (MRS) for SYNGR3 was calculated based on the significant CpGs. Logistic regression was used to assess the relationship between MRS with low eGFR level, adjusting for age, sex, serum uric acid (SUA), total cholesterol (TC), triglycerides (TG), fasting blood glucose (FBG), and high-density lipoprotein cholesterol (HDL-C), with a significance level of P < 0.05.

Results

The EWAS included 61 monozygotic twins with a median (P_2.5, P_97.5) eGFR of 105.501 (57.589, 125.963) ml/min/1.73m² and a median (P_2.5, P_97.5) absolute value of the within-pair difference in eGFR of 4.413 (0.731, 22.718) ml/min/1.73m². Significant intra-pair correlations were observed for other clinical indices, i.e., body mass index (BMI), smoking, drinking, systolic blood pressure (SBP), and diastolic blood pressure (DBP), suggesting the significant advantage of utilizing a trait-discordant identical twin design. Therefore, these factors were not included as covariates in the GEE model (Table 1).

Table 1.

Basic characteristics of the participants

Characteristics	Values	Intrapair correlation
		r	P-value
Number of twin pairs	61
Gender, pairs (%)
Male	31 (50.820%)	–	–
Female	30 (49.180%)	–	–
Age, mean (SD) (year)	52 (7)	–	–
BMI, mean (SD) (Kg/m2)	25.140 (3.582)	0.637	P < 0.001
Smoking, pairs (%)
Yes	24 (39.344%)	0.931	P < 0.001
No	37 (60.656%)
Drinking, pairs (%)
Yes	21 (34.426%)	0.639	P < 0.001
No	40 (65.574%)
SBP, M (P2.5, P97.5) (mmHg)	130 (104, 181)	0.384	0.002
DBP, M (P2.5, P97.5) (mmHg)	82 (62, 105)	0.281	0.028
SUA, M (P2.5, P97.5) (μmol/L)	283.500 (153.380, 545.970)	0.539	P < 0.001
FBG, M (P2.5, P97.5) (mmol/L)	5.500 (3.608, 10.786)	0.592	P < 0.001
TC, mean (SD) (mmol/L)	4.967 (1.180)	0.603	P < 0.001
TG, M (P2.5, P97.5) (mmol/L)	1.140 (0.212, 5.601)	0.613	P < 0.001
HDL-C, M (P2.5, P97.5) (mmol/L)	1.340 (0.701, 2.701)	0.808	P < 0.001
LDL-C, mean (SD) (mmol/L)	2.886 (0.881)	0.516	P < 0.001
eGFR, M (P2.5, P97.5) (ml/min/1.73m2)	105.501 (57.589, 125.963)	0.922	P < 0.001

Continuous variables are presented as mean (standard deviation, SD) or median (M) (P_2.5, P_97.5); Categorical variables are presented as numbers with percentiles

BMI: body mass index, SBP: systolic blood pressure, DBP: diastolic blood pressure, SUA: serum uric acid, : fasting blood glucose, TC: total cholesterol, TG: triglyceride, HDL-C: high-density lipoprotein cholesterol, LDL-C: low-density lipoprotein cholesterol

Epigenome-wide association analysis

The Manhattan plot of EWAS can be seen in Fig. 1. A total of 80 CpGs reaching genome-wide significance levels (FDR < 0.05) were identified. Five CpGs were located at OLIG2 (chr21: 34,391,958—34,392,030 bp), three CpGs in SYNGR3 (chr16: 2,042,984 −2,042,999 bp), five CpGs in RNA5SP207 (chr6: 41,207,271—41,207,335 bp), two CpGs in DLX2 (chr2: 172,961,045—172,961,057 bp), one CpG in ZNF493 (chr19: 21,626,727 bp) and the remaining 64 CpGs in 29 different genes, including LONP1, EBF3, ZNF696, EYS, NR2F2, etc. The methylation levels of 25 CpGs (in ZNF696, EYS, SF3A2, GRTP1, and PTPLA) were positively correlated with eGFR, suggesting that these CpGs showed a hypermethylation tendency in twins with higher eGFR. In comparison, the DNAm level of the other 55 CpGs (in OLIG2, SYNGR3, LONP1, NR2F2, and LIPI, etc.) were negatively correlated with eGFR, indicating that these CpGs were hypomethylated in twins with higher eGFR (Table 2, Supplementary Table 1). Based on the calculation using the bacon package, the corrected inflation factor value was 1.001, Q-Q plot is shown in Supplementary Fig. 1.

Fig. 1.

Standard Manhattan plot for the EWAS of eGFR. The horizontal coordinate is the chromosome number, and the vertical coordinate is the P-value after the conversion of -log10. The solid black line represents the significance threshold, and the significance site is represented by a red dot

Table 2.

EWAS results for eGFR (Top 20)

Chromosome	Position (bp)	β	P-value	FDR	Gene
chr21	34,392,030	− 0.015	1.59E−14	3.92E−09	OLIG2
chr21	34,391,994	− 0.012	1.07E−10	1.32E−05	OLIG2
chr21	34,391,988	− 0.012	7.80E−10	6.42E−05	OLIG2
chr21	34,391,984	− 0.012	1.86E−09	1.15E−04	OLIG2
chr16	2,042,984	− 0.013	3.05E−09	1.51E−04	SYNGR3
chr6	41,207,313	− 0.017	1.46E−08	5.41E−04	RNA5SP207
chr16	2,042,987	− 0.013	1.53E−08	5.41E−04	SYNGR3
chr6	41,207,285	− 0.016	1.91E−08	5.90E−04	RNA5SP207
chr2	172,961,057	− 0.013	2.68E−08	7.35E−04	DLX2
chr19	21,626,727	− 0.152	1.38E−07	3.29E−03	ZNF493
chr6	41,207,271	− 0.015	1.46E−07	3.29E−03	RNA5SP207
chr19	5,698,881	− 0.105	1.61E−07	3.33E−03	LONP1
chr6	41,207,328	− 0.016	1.95E−07	3.71E−03	RNA5SP207
chr10	131,771,198	− 0.012	4.51E−07	7.96E−03	EBF3
chr8	144,378,531	0.074	5.14E−07	8.47E−03	ZNF696
chr6	66,373,911	0.064	7.64E−07	1.18E−02	EYS
chr15	96,877,513	− 0.012	8.93E−07	1.30E−02	NR2F2
chr10	131,771,193	− 0.012	1.15E−06	1.51E−02	EBF3
chr8	144,378,528	0.072	1.16E−06	1.51E−02	ZNF696
chr6	41,207,335	− 0.015	1.36E−06	1.65E−02	RNA5SP207

Causal inference analysis

Details of the estimated causal relationships between genome-wide significant CpGs and eGFR were provided in Table 3. We found 5 CpGs (in ZNF493, SF3A2, LIPI, SYNGR3, C9orf3) for which a causal effect of DNAm on eGFR was observed. In addition, a bidirectional causal effect of eGFR with DNAm was observed in 7 of these CpGs (in OLIG2, SYNGR3, and EPHB3). Finally, the causal impact of eGFR on DNAm was supported by 6 CpGs located at OLIG2, MLLT1, LITAF, and PRIMA1.

Table 3.

Results of causal inference analysis for eGFR

CpG ID	Chromosome	Position	Gene	Methylation to eGFR					eGFR to methylation
				βself change	Pself change	βco-twin change	Pco-twin change	Ratio	βself change	Pself change	βco-twin change	Pco-twin change	Ratio
1*#	21	34,392,030	OLIG2	− 8.082	4.63E−16	− 11.658	3.18E−16	1.442	− 0.010	7.02E−02	0.019	1.04E−03	1.840
2*#	21	34,391,994	OLIG2	− 9.087	4.60E−25	− 12.446	4.82E−25	1.370	− 0.011	3.62E−02	0.018	1.70E−03	1.580
3*#	21	34,391,988	OLIG2	− 9.134	3.14E−26	− 12.420	2.14E−26	1.360	− 0.011	3.59E−02	0.017	1.89E−03	1.570
4*#	21	34,391,984	OLIG2	− 9.159	1.52E−26	− 12.393	9.56E−27	1.353	− 0.011	4.01E−02	0.017	2.34E−03	1.576
5*#	16	2,042,984	SYNGR3	− 5.979	2.22E−05	− 8.607	8.41E−07	1.439	− 0.003	5.65E−01	0.014	1.27E−02	4.306
6*#	16	2,042,987	SYNGR3	− 5.880	5.95E−06	− 8.232	5.13E−07	1.400	− 0.003	5.92E−01	0.014	1.68E−02	4.429
7*	19	21,626,727	ZNF493	− 0.222	3.27E−02	− 0.449	3.13E−04	2.020	− 0.084	2.62E−01	0.125	9.85E−02	–
8*	19	2,247,678	SF3A2	0.508	1.07E−01	1.556	4.51E−06	3.063	0.010	2.32E−01	− 0.011	1.66E−01	–
9*	21	15,436,207	LIPI	− 0.146	2.15E−02	− 0.367	6.78E−06	2.505	− 0.085	1.87E−01	0.106	8.78E−02	–
10*	16	2,042,999	SYNGR3	− 5.103	1.76E−07	− 6.704	8.74E−08	1.314	− 0.003	6.87E−01	0.013	6.23E−02	–
11*	9	97,547,377	C9orf3	− 0.755	1.19E−29	− 1.015	1.78E−27	1.345	− 0.061	1.94E−01	0.094	7.87E−02	–
12*#	3	184,280,551	EPHB3	− 10.677	2.40E−07	− 15.385	1.18E−11	1.441	− 0.005	2.09E−01	0.013	1.17E−03	2.927
13#	21	34,391,958	OLIG2	− 8.809	3.11E−29	− 11.244	2.72E−32	1.276	− 0.008	1.42E−01	0.013	1.30E−02	1.746
14#	19	6,230,568	MLLT1	0.721	1.09E−19	0.882	2.98E−20	1.222	0.040	2.81E−01	− 0.082	4.38E−02	2.028
15#	16	11,712,567	LITAF	− 0.665	1.65E−11	− 0.855	1.46E−12	1.285	− 0.044	3.68E−01	0.104	3.58E−02	2.392
16#	16	11,712,569	LITAF	− 0.666	1.91E−11	− 0.856	1.86E−12	1.285	− 0.043	3.71E−01	0.104	3.60E−02	2.402
17#	14	94,254,075	PRIMA1	− 5.181	2.13E−22	− 6.055	1.34E−22	1.169	− 0.003	7.20E−01	0.016	4.18E−02	5.192
18#	14	94,254,078	PRIMA1	− 5.389	1.33E−23	− 6.271	1.71E−22	1.164	− 0.003	7.35E−01	0.015	4.56E−02	5.424

^*DNAm has a causal effect on eGFR, ^# eGFR has a causal effect on DNAm, ^*#DNAm and eGFR have bidirectional causal effects

Analysis of differentially methylated regions

A total of 28 DMRs associated with eGFR were detected (Table 4). Among these DMRs, the methylation levels of 6 DMRs (2, 6, 14, 21, 26–27) at MINDY2-DT and SEMA6B were positively associated with eGFR, while 19 DMRs (3–5, 7–8, 10–12, 15–20, 22–25, 28) at HIST2H2BB, NAT8L, SOWAHC, ZNF853, and MAB21L1, etc. were negatively correlated with eGFR. However, the methylation levels of 3 DMRs (1, 9, 13) were inconclusive about eGFR (Fig. 2).

Table 4.

Significantly different methylated region Results

DMR ID	Chromosme	Start	End	Length	Slk-corrected P-value	Gene
1	8	144,378,449	144,379,225	39	1.00E−05	ZNF696
2	2	241,588,115	241,588,439	11	1.20E−04	AQP12B
3	1	149,398,882	149,399,176	30	4.06E−04	HIST2H2BB
4	3	45,077,255	45,077,766	32	2.45E−03	CDCP1
5	21	38,068,339	38,068,932	34	3.07E−03	SIM2
6	7	57,484,214	57,484,725	36	3.12E−03	ZNF716
7	14	94,254,054	94,254,481	14	3.31E−03	PRIMA1
8	9	66,456,253	66,456,627	20	3.87E−03	LNIC01410
9	19	51,171,278	51,171,979	46	9.27E−03	SHANK1
10	5	83,679,737	83,680,038	20	1.09E−02	EDIL3
11	2	177,024,709	177,025,289	24	1.45E−02	HOXD3
12	5	3,602,346	3,602,748	22	1.92E−02	IRX1
13	14	23,305,835	23,306,945	47	2.05E−02	MMP14
14	1	1,004,924	1,005,064	13	2.06E−02	RNF223
15	1	7,764,265	7,764,350	13	2.09E−02	CAMTA1
16	8	144,267,308	144,267,487	13	2.12E−02	GPIHBP1
17	12	81,101,778	81,102,177	17	2.15E−02	MYF6
18	11	19,264,005	19,264,275	14	2.24E−02	E2F8
19	20	39,316,634	39,317,380	32	2.33E−02	MAFB
20	5	140,306,413	140,307,027	42	2.59E−02	PCDHA
21	2	170,624,866	170,625,450	26	2.79E−02	KLHL23
22	3	61,549,308	61,549,450	13	3.04E−02	PTPRG
23	10	134,659,972	134,661,083	26	3.18E−02	TTC40
24	3	6,903,067	6,903,982	26	3.30E−02	GRM7
25	22	40,391,446	40,391,594	11	3.60E−02	FAM83F
26	2	636,551	637,017	31	3.72E−02	TMEM18
27	22	51,042,073	51,042,594	20	4.33E−02	MAPK8IP2
28	16	29,185,813	29,186,189	13	4.57E−02	NPIPB7

Fig. 2.

The plot of the different methylation patterns of the identified DMRs. The y-axis represents the correlation coefficient of each CpG with eGFR, while the x-axis indicates chromosome position. Black dots represent each CpG, and the blue line illustrates the methylation pattern of each DMR. BP base pairs, chr chromosomes

Gene interaction network analysis

We also explored the interactions network with 53 eGFR- and DMRs-related genes through the Gene MANIA tool. The interaction network exhibited 40.57% co-expression, 35.38% physical interaction, 13.92% genetic interaction, and 10.13% prediction. The results showed that SYNGR3 was co-expressed with MLLT1 and had gene interactions with AFF4 and EDIL3 (Fig. 3).

Fig. 3.

Schematic diagram of gene interaction network. Purple: co-expression; red: physical interactions; green: gene interactions; yellow: predicted

Enrichment analysis

Many important pathways potentially related to eGFR were identified, such as alpha-2A adrenergic receptor binding pathway, corticotropin-releasing hormone receptor activity pathway, regulation of renal albumin absorption pathway, nephron tubule development pathway, and negative regulation of glomerular filtration by angiotensin pathway (Table 5).

Table 5.

The GREAT ontology enrichments for regions potentially related to eGFR

Ontology database	Term name	Binom FDR Q-value	Binom region fold enrichment
GO Molecular Function	corticotropin-releasing hormone receptor 2 binding	6.88E−07	4.518
	alpha-2A adrenergic receptor binding	6.92E−07	3.093
	corticotropin-releasing hormone receptor activity	4.51E−05	4.903
	corticotrophin-releasing factor receptor activity	1.38E−03	3.405
	corticotropin-releasing hormone binding	2.07E−03	3.255
GO Biological Process	regulation of metanephric mesenchymal cell migration	5.46E−67	14.139
	regulation of renal albumin absorption	5.86E−51	17.687
	renal system process involved in regulation of blood volume	3.78E−41	5.913
	regulation of glomerular filtration	5.36E−40	6.950
	pattern specification involved in kidney development	2.09E−25	2.714
	regulation of metanephros development	4.13E−21	2.443
	nephron tubule development	1.01E−20	1.828
	renal system process involved in regulation of systemic arterial blood pressure	1.21E−19	2.746
	regulation of kidney development	1.78E−18	1.726
	nephron epithelium development	5.77E−18	1.659
	paramesonephric duct development	6.53E−17	4.988
	nephron epithelium morphogenesis	3.38E−16	1.757
	nephric duct elongation	4.51E−16	9.863
	negative regulation of glomerular filtration by angiotensin	2.23E−15	63.405
PANTHER Pathway	Cortocotropin releasing factor receptor signaling pathway	9.49E−12	2.157
	Adrenaline and noradrenaline biosynthesis	4.86E−08	1.994
	Alpha adrenergic receptor signaling pathway	1.03E−04	1.597
MSigDB Pathway	Regulation of Water Balance by Renal Aquaporins	3.27E−04	1.440

Gene expression analysis

A total of 12 pairs of twins were included in the analysis, with a mean age of 55 years (SD: 6) and a median eGFR of 99.647 ml/min/1.73 m² (P_2.5, P_97.5: 58.205, 114.489). SYNGR3 (β: −11.115) expression was found to be negatively correlated with eGFR, and PRIMA1 (β: 27.664), CAMTA1 (β: 9.402), and DLX2 (β: 33.941) expression was found to be positively correlated with eGFR. No statistically significant results were found for the other genes (Supplementary Table 2).

Quantitative methylation analysis of SYNGR3

Comparison between groups revealed differences in methylation levels of the four CpGs (Supplementary Table 3). The logistic regression results showed that the MRS of SYNGR3 was positively associated with lower eGFR level (β: 0.785, P: 5.28 × 10^–3), indicating that higher methylation levels of SYNGR3 were associated with lower eGFR.

Discussion

In this study, we performed a monozygotic twins-based EWAS to assess epigenetic variations associated with eGFR. We identified 80 CpGs, 34 genes, 28 DMRs, and various biological pathways associated with eGFR. In addition, we found a causal effect of DNAm on eGFR and eGFR on DNAm. The candidate gene SYNGR3 was validated in an independent community cohort. These findings contribute to further epigenetic studies of eGFR changes and provide insights for interventions and treatments for eGFR decline.

We found a negative association between the methylation level of SYNGR3 and eGFR. In an independent community population, our validation result was consistent with this. Furthermore, we demonstrated that the CpGs of SYNGR3 have a causal effect on eGFR using ICE FALCON. Similarly, a genetic study involving millions of samples also found an association between SYNGR3 and eGFR [33]. SYNGR3 was presumed to be an epigenetic regulatory gene due to its typical CpG features [34]. The SYNGR3 encoded synaptogyrins-3, a protein that regulated neurotransmitter release [35]. Synaptogyrins-3 affected renal nerve signaling by influencing the release of neurotransmitters. Renal nerves played an important role in the regulation of renal function, including glomerular filtration, sodium reabsorption, and renin release [36]. We also found a significant causal effect from eGFR on the methylation of SYNGR3. However, the exact mechanism is unclear, and further studies are needed to elucidate it.

In this study, methylation levels of CpGs located at genes such as EYS, MLLT1, and KIF26A were found to be positively associated with eGFR. Multifactorial analysis showed that high EYS expression predicted worse renal function and shorter survival in patients with clear cell renal cell carcinoma [37]. Abnormal expression of MLLT1 during early kidney development enhanced transcription, leading to the occurrence of nephroblastoma [38]. KIF26A encoded an unconventional motor protein that affected cilia formation and function, and its impairment could lead to congenital disabilities such as renal and urinary tract abnormalities [39].

We found that the methylation levels of CpGs in LONP1, NR2F2, BRD1, LITAF, and SHH were negatively correlated with eGFR. NR2F2 reduced renin promoter activity, and the consequent low renin, hypotension, and hyponatremia could lead to decreased glomerular filtration rate and renal function impairment [40]. Increased LONP1 expression has been observed in diabetic nephropathy patients and was closely associated with renal tubulointerstitial fibrosis. When LONP1 expression was inhibited, serum creatinine and renal tubule injury were improved [41]. BRD1 was found to be positively associated with urinary albumin excretion, and high urinary albumin excretion was indicative of impaired kidney function [42]. LITAF induced the secretion of tumor necrosis factor-alpha and other inflammatory mediators, causing renal inflammation and the progression of CKD[43]. SHH has been found to promote kidney damage and fibrosis [44]. In a GWAS on 1.2 million individual sample, SHH and was also found to be negatively correlated with eGFR levels, which supports our results to some extent [45]. However, the association of other genes (e.g., OLIG2, PRIMA1, LIPI, and GPR144) with eGFR has not been extensively investigated, and further studies are needed to determine their roles.

In addition, an EWAS conducted on a Korean population found that the methylation levels of CpGs located in ZNF696, GPR144, and KIF26A were associated with eGFR, consistent with our findings [46]. Another epigenome study involving 1.5 million European individuals also supported our results, identifying associations between eGFR levels and the methylation levels of CpGs in genes such as OLIG2, DLX2, ZNF493, LONP1, EBF3, ZNF696, EYS, NR2F2, SF3A2, CPN2, GRTP1, BRD1, MLLT1, LITAF, PRIMA1, GPR35, EPHB3, DRGX, KIF26A, SHH, ITGB2, CAMTA1, CD8B2, GFM2, and MAFB [33].

Gene MANIA results showed that SYNGR3 was co-expressed with MLLT1 and had gene interactions with AFF4 and EDIL3. SYNGR3 can regulate synaptic function and affect neuronal messaging. MLLT1 shift code deletion was associated with glioblastoma. AFF4 was involved in mediating the genesis and development of neurons. EDIL3 probably affected neural function through extracellular matrix accumulation. This explained possibly the interaction of SYNGR3 with MLLT1, AFF4, and EDIL3 [35, 47, 48].

Using the ICE FALCON method, this study found that eGFR was causally related to CpGs on several genes, such as C9orf3, EPHB3, and MLLT1. The aminopeptidase produced by C9orf3 was an important component of the renin-angiotensin system and caused hypertension by promoting the conversion of angiotensin II [49]. The association of hypertension with decreased eGFR and renal injury was self-evident[50]. Receptor signaling by EPHB3 regulates the cytoarchitecture and spatial organization of adult renal medullary tubular cells via Rho family GTPases and may, therefore, influence tubular reabsorption capacity [51]. In patients with nephroblastoma, MLLT1 was observed to act as a messenger to drive aberrant expression of target genes by mediating phase separation and protein–protein interactions [52].

This study identified 28 DMRs associated with eGFR within genes CDCP1, SHANK1, HOXD3, IRX1, GPIHBP1, PCDHA, PTPRG, etc. For SHANK1 and E2F8, the results of this study were consistent with previous research. The DNAm level of SHANK1 has been shown to be associated with eGFR in African HIV patients, suggesting its potential impact on kidney function [53]. The DNAm level of E2F8 has been found to be related to survival in kidney cancer patients, and E2F8 was also involved in the repair process of acute kidney injury [54, 55]. High CDCP1 expression was found to be negatively correlated with eGFR by co-expression analysis [56]. Overexpression of HOXD3 inhibited the proliferation, invasion, and migration of 786-O and CAKI-1 cells, and it was a key gene for inhibiting the progression of renal cell carcinoma [57]. IRX1, when under-expressed, promoted inflammatory responses, thereby impairing kidney function [58]. GPIHBP1 has been found to induce chronic kidney failure by impairing lipoprotein lipase function [59]. Hypermethylation of PCDHA has been demonstrated to be associated with nephroblastoma [60]. PTPRG exerted an antitumor effect by modulating the immune phenotype of renal cell carcinoma patients [61].

This study has several strengths. First, the use of monozygotic twin design controls for genetic background, family upbringing, and intrauterine environment, thereby enhancing the credibility of the study findings. Second, we conducted causal inference and identified that DNAm has a causal effect on eGFR. Thirdly, using the Chinese population as a sample, this study helps to provide information on the decline in eGFR in Chinese.

However, this study also has some limitations. First, the sample size of this study was relatively small due to the difficulty of obtaining twin samples. However, this study employed a trait-discordant twin design, significantly reducing the required sample size compared to traditional cross-sectional or case–control designs while achieving the same statistical power [62]. Additionally, although every effort has been made to control for confounding factors, some unknown factors may affect the results that cannot be measured.

Conclusions

In conclusion, this study identified multiple CpGs, regions, genes, and signalling pathways associated with eGFR. The findings may provide important clues for further studies on epigenetic modifications of eGFR decline and help to discover new diagnostic markers and therapeutic targets for CKD.

Abbreviations

BMI

Body mass index

CKD

Chronic kidney disease

CpGs

Cytosine-phosphate-Guanines

DBP

Diastolic blood pressure

DMRs

Differentially methylated regions

DNAm

DNA methylation

eGFR

Estimated glomerular filtration rate

EWAS

Epigenome-wide association study

FBG

Fasting blood glucose

GEE

Generalized estimation equation

GRCh37, hg19

Genome Reference Consortium Human Build 37

GREAT

Genomic Regions Enrichment of Annotations Tool

GWAS

Genome-wide association study

HDL-C

High-density lipoprotein cholesterol

ICE FALCON

Inference about Causation from Examination of FAmiliaL CONfounding

LDL-C

Low-density lipoprotein cholesterol

MRS

Methylation risk score

RRBS

Reduced representation bisulfite sequencing

SBP

Systolic blood pressure

SUA

Serum uric acid

TC

Total cholesterol

TG

Triglyceride

Author contributions

XQ, JW, TW, WW conducted the data collection and processing. XQ conceived the study, analyzed the data, and drafted the manuscript. DZ performed critical revision of the manuscript. All authors read and approved the final manuscript.

Funding

The study was funded by a grant from the National Natural Science Foundation of China (Grant No: 31741063).

Availability of data and materials

The raw data will be uploaded to the dbGaP website after the project is completed. Full summary statistics and code have been uploaded to the Github repository (Repository Name: EWAS_eGFR, URL: https://github.com/Enterpriseee/EWAS_eGFR).

Declarations

Ethics approval and consent to participate

The Regional Ethics Committee of the Institutional Review Board of the Qingdao Municipal Centre for Disease Control and Prevention approved the study. The study followed the Declaration of Helsinki, and all participants provided informed consent.

Consent for publication

Not applicable.

Competing interests

The authors declare that they have no competing interests.