Epigenetic signatures of phthalate exposure and potential risks: a DNA methylation analysis using Infinium MethylationEPIC BeadChip

Abstract

Phthalates are common environmental pollutants known to disrupt various regulatory systems and are associated with several health issues, such as impaired immune response, developmental toxicity, hormonal disruption, and type 2 diabetes. Epigenetic modifications, such as DNA methylation, can serve as early indicators of environmental toxicant exposure due to their rapid alteration in response to varying environmental factors without altering the underlying DNA sequence. To investigate the impact of phthalate exposure on human health and the affected regulatory mechanisms, this study analysed a DNA methylation dataset generated using the Illumina Infinium MethylationEPIC BeadChip (EPIC BeadChip) array, along with the concentrations of 15 urinary phthalate metabolites from 389 participants. The results revealed sex-specific differences in phthalate concentrations, with females exhibiting relatively higher levels than males. These differences may reflect a combination of factors, including lifestyle behaviours and potential differences in exposure sources. Furthermore, differentially methylated CpG sites (DMCs) were identified only in the mono-ethylhexyl phthalate (MEHP) dataset, where a total of 53 DMCs were detected, including 11 that were consistently detected across multiple MEHP concentration comparisons. Additionally, the functional analysis showed that these DMCs are primarily involved in protein and nucleotide binding, immune response, ion channel regulation, and membrane-associated pathways. This study provides high-potential phthalate-related methylation markers, their associated genes, and the functions they are involved in. These findings offer valuable insights for the research on environmental toxicants and epigenetics, while supporting clinical applications related to phthalates.

Graphical Abstract

Graphical Abstract.

Introduction

Phthalates are a group of synthetic chemicals commonly used as plasticizers, which increase the flexibility, durability, and longevity of plastics. They are found in a wide range of products in our daily lives and have the potential to accumulate in the human body. Generally, phthalates are classified into two main categories based on their molecular weight: low-molecular-weight phthalates (LMWPs), commonly used in personal care projects, and high-molecular-weight phthalates (HMWPs), typically found in construction materials, medical devices, food packaging, and children’s toys [1, 2]. As a result, humans are primarily exposed to phthalates through ingestion, inhalation, and skin absorption [3]. Once inside the body, phthalates are rapidly metabolized into their respective monoesters and oxidative metabolites [1, 4, 5]. Most phthalates and their metabolites are excreted through urine and faeces within 24–48 h after exposure. However, certain metabolites, such as di-ethylhexyl phthalate (DEHP) and its metabolite, mono-ethylhexyl phthalate (MEHP), may have prolonged retention time in the body, as they are primarily used as plasticizers in plastic production [1, 6].

Phthalates have been linked to various health concerns due to their potential endocrine-disrupting properties. Several studies have reported the association between phthalate exposure and some adverse health effects, including reproductive and developmental toxicity [7], hormone dysregulation [8–10], obesity [3, 9], type 2 diabetes [11–13], immune dysfunction [14–18], respiratory diseases [19–21], neurodevelopmental impacts [22, 23], and even tumour development [24–28]. However, their underlying mechanisms remain largely unclear.

Epigenomics, the genome-wide study of heritable changes in gene expression without altering the underlying DNA sequence, is a crucial approach for understanding the influence of phthalates on human health [29, 30]. Epigenetic modifications can occur rapidly in response to environmental toxicants, reshaping the epigenetic landscape and leading to changes in gene regulation that may contribute to various health problems [31–33]. Consequently, epigenetic modifications could serve as early indicators of environmental exposure and potential health risks [33]. DNA methylation, a key epigenetic mechanism, can be disrupted by phthalates at specific CpG sites, causing health issues [34, 35]. Therefore, investigating genome-wide or high-density DNA methylation provides valuable insights into the molecular impacts of phthalates, facilitating the identification of early biomarkers for environmental toxicant exposure and elucidating potential pathogenic pathways of disease.

However, obtaining the data for whole-genome or high-density DNA methylation from a sufficient number of participants is extremely challenging. The Taiwan Biobank (TWB) [36], a prospective database, collects data from over 200 000 cancer-free Taiwanese participants, all of whom are required to return for follow-up assessments every 2–4 years. From the extracted DNA, various omics data have been generated, including whole-genome sequencing, genome-wide association studies, gene expression profiling, protein expression information, and DNA methylation data using EPIC BeadChip. Moreover, the encrypted personal and clinical information was provided in the dataset, as well as the results from blood and urine tests, including phthalate metabolite concentrations in the urine samples of participants. Based on the valuable data provided by the TWB, we conducted a systematic analysis of the influences of phthalate exposure on DNA methylation in this study.

To understand the influences of phthalate exposure on human regulatory mechanisms and to identify potential biomarkers, comprehensive and systematic analyses were performed by using EPIC BeadChip data from 389 participants in the TWB, along with the concentrations of 15 different phthalate metabolites in their urine samples. Through these comparisons, several CpG sites potentially modulated by phthalate were identified. Furthermore, the biological functions and metabolic pathways affected by the differentially methylated CpG sites (DMCs) were also detected. The results of this study provide valuable information for future research and clinical research.

Materials and methods

Dataset and assessment of phthalate metabolites

The TWB is a government-supported, ongoing, prospective cohort study that includes both a community-based arm and a hospital-based arm. In the community-based arm, men and women aged 30–70 years without a prior cancer diagnosis are recruited from over 30 sites across Taiwan. Participant information, such as age, gender, smoking habits, alcohol consumption, betel nut usage, and dietary patterns, is collected through questionnaires conducted by trained interviewers. Furthermore, the TWB collects biological samples, including DNA, blood plasma, and urine. The DNA methylation dataset used in this study was retrieved from the 2017 TWB [36]. Within the TWB, a total of 389 participants (194 males and 195 females) have both DNA methylation profiles, generated using the Illumina Infinium MethylationEPIC BeadChip, and corresponding urinary phthalate metabolite measurements. This unique dataset, spanning 865 918 CpG sites, enables an integrated analysis of epigenetic modifications in relation to phthalate exposure. The participants, aged 30–70 years, had no history of cancer. The Houseman reference-based deconvolution method was applied to compare and adjust for five cell types (B lymphocytes, natural killer cells, CD4⁺ T cells, CD8⁺ T cells, and monocytes) [37]. Methylation levels (β values), data normalization, and P-values were calculated using the Illumina GenomeStudio software (v2011.1) with the methylation module.

Statistical test for baseline comparison

In the baseline statistical analysis of clinical data between groups, continuous values were assessed with the Mann–Whitney U test, and categorical data were evaluated with the χ² test. A P-value of < .05 is deemed statistically significant. Since smoking is considered a factor that can alter epigenetic landmarks [38], the information on smoking status is further analysed. The participants were classified to three groups—‘currently smoking’, ‘former smoker’, and ‘never smoked’, and converted as 2, 1, and 0 by using one-hot encoding, respectively. Moreover, linear regression was performed using smoking status as a factor to compare and adjust DNA methylation data. To further explore the impact of smoking on our DNA methylation dataset, we conducted a PERMANOVA test using the scikit-bio (skbio) Python package. NMDS plots and heatmaps of the Euclidean distance matrix were also generated to visualize potential clustering patterns based on smoking status.

Quality control and data processing

To increase the data quality, single nucleotide polymorphism (SNP) probes within the dataset were removed to avoid interference with the analysis of methylation data. Additionally, the analyses included only the samples with at least 99% valid values and P-values ≤ .01 while ensuring that each of the CpG sites had a minimum of 95% valid values and a P-value ≤ .01. After applying these filters, the retained CpG sites numbered 860 189 for males and 859 117 for females. Furthermore, several previous studies have reported that the M value is statistically more appropriate for differential analysis of methylation levels than the β value [39, 40]. Thus, to perform downstream analyses, the β values were converted to M values by applying Equation 1.

(1)

ComBat [41] was applied to remove the batch effects.

Differential expression analysis and functional annotation

Based on the grouping in Fig. 1, the analysis begins with comparisons of 15 phthalates between Low_10% and High_10%, then proceeds to compare Low_20% with High_20%, progressively narrowing the differences in phthalate metabolite concentrations between groups until comparisons were made between Low_50% and High_50%. The DMCs at different levels of phthalate metabolite concentrations can be detected in these comparisons. The differential expression analysis was conducted by limma [42], which is a widely utilized method for microarray, transcriptomics, proteomics, and methylation data [43–45]. The false discovery rate (FDR) < 0.05 is considered significant.

Figure 1.

Names of the 10 subgroups based on phthalate metabolite concentrations percentile. The black percentages represent the percentiles of the phthalate metabolite concentrations. The arrows and their corresponding labels represent the data ranges included and the group names. Number of male and female participants in each subgroup is indicated in the parentheses. M, male. F, female.

In order to study the functions of DMC-associated genes, functional annotations of Gene Ontology (GO) [46] were conducted by applying the UniProt ID mapping tool [47, 48]. Additionally, STRING [49] was utilized to identify the interaction networks involving the DMCs. STRING is a database that provides known and predicted protein–protein interactions (PPIs) based on the information from experimental data, computational prediction methods, and text mining of scientific literature. By using STRING, the PPI networks and enriched functions associated with DMCs can be explored.

Results

Sample grouping and statistical analysis

The DNA methylation dataset used in this study was retrieved from the TWB [36] comprising 389 volunteers (194 males and 195 females) and covers 865 918 CpG sites. Participants are aged between 30 and 70 years with no history of cancer. To understand the influence of phthalate, the urine samples from these participants were collected to measure various phthalate metabolites by using liquid chromatography and tandem mass spectrometry. These phthalate metabolites include MEHP, DEHP, dibutyl phthalate (DBP), monobenzyl phthalate (MBzP), mono(2-carboxymethylhexyl) phthalate (MCMHP), mono(2-ethyl-5-carboxypentyl) phthalate (MECPP), mono(2-ethyl-5-hydroxyhexyl) phthalate (MEHHP), mono(2-ethyl-5-oxohexyl) phthalate (MEOHP), monoethyl phthalate (MEP), mono-isobutyl phthalate (MiBP), monomethyl phthalate (MMP), mono-n-butyl phthalate (MnBP), mono-isononyl phthalate (MiNP) as well as the mean of melamine. Moreover, (HWMP), the sum of all HMWPs (MEHP, MEHHP, MEOHP, MECPP, MCMHP, and MiNP), and (LWMP), the total of all LMWPs (MBzP, MnBP, MiBP, MEP, and MMP), were also used to investigate the link between DNA methylation and phthalate molecular weight. However, MiNP was undetectable in ∼40% of the participants’ urine samples; thus, it was excluded from the analyses. Additionally, blood samples were collected from the identical participants for clinical testing to serve as a baseline comparison. The haematology test included WBC (white blood cell count), RBC (red blood cell count), TG (triglyceride), HDL (high-density lipoprotein cholesterol), LDL (low-density lipoprotein cholesterol), AFP (alpha-fetoprotein), BUN (blood urea nitrogen), mALB (microalbumin), AST (aspartate aminotransferase), total cholesterol, total bilirubin, albumin, creatinine, anaemia, and uric acid.

The participants were grouped based on the percentiles of the detected concentrations (ng/mL) of phthalate metabolites to assess the impact of varying levels of phthalate exposure on human health. First, all participants were divided into high-exposure and low-exposure groups based on the median value of phthalate metabolite concentrations. Within each major group, participants were further categorized into five subgroups according to their concentration ranking percentiles (Fig. 1). For instance, Low_10% refers to participants with phthalate metabolite concentrations in the 0–10th percentile, while Low_20% included those in the 0–20th percentile. This pattern continued up to 0–50th percentile in subgroup Low_50%. The same grouping principles were applied to the high-exposure group, where High_50% represents the participants with phthalate metabolite concentrations in the 50–100th percentile, High_40% includes those with concentrations in the 60–100th percentile, and so on, down to High_10%. This categorization allowed for a more nuanced analysis of the data. The specific phthalate concentrations for each 10th percentile are detailed in Table S1.

Significant differences in phthalate metabolite concentrations between sexes

In the dataset, significant differences in the concentrations of phthalate metabolites between males and females were observed in 11 out of the 15 phthalates (Fig. 2). These sex-specific differences in phthalate levels may result from a combination of factors, including lifestyle behaviours and exposure sources such as the use of personal care products [50–52]. Moreover, these differences would contribute to variations in sex hormone levels, including sex hormone-binding globulin, estradiol, and testosterone [9, 10]. Therefore, analyses in this study were conducted separately for males and females to account for these differences.

Figure 2.

Box plots visualizing the concentrations 15 phthalate metabolites with comparisons between sexes. From panels A to O are the results for MEHP, DEHP, MEOHP, MEHHP, MECPP, MCMHP, MBzP, MnBP, MiBP, MEP, MMP, LMWP, HMWP, and DBP and the mean of melamine. The boxes represent the concentrations of phthalate metabolites for males and females. The line in each box shows the median of the concentrations. The x-axis displays the group names as defined in Fig. 1, which also documents the number of participants in each group. The number of stars above the boxes indicates the significance of the P-value, with 0 to 3 stars representing P-value > .05, .01 < P-value ≤ .05, .001 < P-value ≤ .01, and P-value ≤ .001, respectively. The square subplot in each plot is the magnified box plots of the low concentration groups.

Baseline comparisons

The baseline comparisons of clinical data between high and low phthalate metabolite exposure groups are presented in Table 1 and Tables S2–S6. Across all subgroup comparisons, no significant differences in clinical features were observed, except for body mass index in the Low_50% versus High_50% male cohort, which exhibits a slight difference (P-value = .04). These results indicate that the impact of phthalate exposure can be assessed without interference from covariates.

Table 1.

P-values of baseline comparisons between different phthalate metabolite concentration groups

	High_10% vs Low_10%		High_20% vs Low_20%		High_30% vs Low_30%		High_40% vs Low_40%		High_50% vs Low_50%
	Male	Female	Male	Female	Male	Female	Male	Female	Male	Female
N of Higha	20	20	39	39	59	59	78	78	97	97
N of Lowa	20	20	39	39	59	59	78	78	97	98
Age	0.81	0.34	0.14	0.78	0.13	0.36	0.17	0.53	0.28	0.79
Red blood cells	0.57	0.50	0.57	0.90	0.53	0.49	0.43	0.56	0.36	0.52
White blood cells	0.08	0.88	0.18	0.81	0.58	0.38	0.61	0.71	0.34	0.78
Total cholesterol	0.51	0.73	0.09	0.92	0.51	0.91	0.94	0.48	0.68	0.36
Triglycerides	0.40	0.76	0.78	0.76	0.56	0.20	0.32	0.73	0.95	0.59
High-density lipoprotein	0.88	0.89	0.79	0.41	0.71	0.58	0.34	0.85	0.42	0.43
Low-density lipoprotein	0.80	0.79	0.20	1.00	0.72	0.96	0.60	0.36	0.80	0.24
Total bilirubin	0.15	0.66	0.68	0.14	0.41	0.13	0.21	0.78	0.03	0.60
Albumin	0.17	0.56	0.35	0.21	0.53	0.09	0.35	0.11	0.48	0.12
Alpha-fetoprotein	0.98	0.36	0.30	0.42	0.31	0.30	0.07	0.82	0.06	0.28
Blood urea nitrogen	0.84	0.36	0.76	0.38	0.34	0.54	0.51	0.50	0.67	0.14
Creatinine	0.84	0.97	0.82	0.87	0.64	0.70	0.86	0.99	0.53	0.76
Microalbumin	0.23	0.55	0.14	0.49	0.36	0.62	0.78	0.84	0.83	0.85
Hyper-/hypotension	0.13	0.30	0.27	0.63	0.17	0.92	0.35	0.94	0.29	0.57
Diabetes mellitus	0.74	0.48	0.38	0.69	0.42	1.00	0.21	0.94	0.52	0.88
Body mass index	0.33	1.00	0.25	0.99	0.13	0.76	0.06	0.23	0.04	0.40
Aspartate aminotransferase	1.00	1.00	1.00	1.00	0.44	1.00	0.53	1.00	1.00	0.76
Anaemia	1.00	1.00	1.00	0.79	1.00	0.13	1.00	0.13	0.61	0.12
Uric acid	0.33	1.00	1.00	0.67	0.70	1.00	0.19	0.85	0.55	0.56
Smoking	0.37	0.20	0.39	0.21	0.74	0.75	0.51	0.90	0.53	0.85

^aN of High and N of Low indicate the number of participants in the corresponding high- and low-level phthalate exposure groups, respectively.

The statistical model was adjusted for cell-type composition and smoking status to account for potential confounding effects in the differential expression analysis. Details of the adjustment procedure are provided in the ‘Materials and methods’ section. Since smoking is a common factor that can influence epigenetic landmarks [38], its impact on DNA methylation was further assessed using a permutational multivariate analysis of variance (PERMANOVA) test. The results indicate P-values of .288 for males and .150 for females, respectively. Furthermore, no clear clustering was observed in the heatmap or the non-metric multidimensional scaling (NMDS) plot of the Euclidean distance matrix (Figs S1 and S2).

MEHP is the sole phthalate metabolite causing significant differences in DNA methylation sites

Differential expression analysis is one of the most commonly used approaches for identifying the vital genomic features between sample groups. Among all comparative analyses of the phthalates across various concentration differences, DMCs were only identified in the MEHP comparisons (Fig. 3; Figs S3–S17  and Table S7). This may be attributed to MEHP being the primary metabolite of DEHP, which is a widely used phthalate found in numerous products such as polyvinyl chloride plastics, medical devices, and other daily utensils, leading to frequent human exposure. Among the DMCs identified in the MEHP analysis, no overlap between males and females was discovered (Table S7), further demonstrating that the effects of phthalates differ by sex. Although the DMCs differ between males and females, the most significant sites for both groups were identified in the comparison between Low_20% and High_20%. The main reason may be that the sample size for the High_10% vs Low_10% group is too small, while the differences in phthalate metabolite concentrations of other groups are not significant enough. Based on this analysis, the top and bottom 20th percentiles of phthalate concentrations could provide a practical framework for studying the influence of phthalates on human health.

Figure 3.

Volcano plots from the differential expression analysis for MEHP. The left and right columns display the results from the male and female datasets, respectively, while the rows correspond to the comparison of different phthalate concentrations. Each dot represents a CpG site, with a total of 865 918 sites. The x-axis indicates logFC (fold change), and the y-axis shows −log(P-value). Red dots represent DMCs with an FDR ≤ 0.05. The numbers of up-regulated and down-regulated DMCs are marked on the plots.

Most participants can be clustered based on the high or low MEHP concentration levels by using the methylation levels of the DMCs (Fig. 4A–E). The classification accuracies of all comparisons were >70% (Fig. 4F). The accuracies for the Low_20% vs High_20% comparison in females and the Low_40% vs High_40% comparison in males were relatively lower than the other groups, likely due to the identification of only three and two DMCs, respectively. In contrast, the accuracies for other groups ranged from 86% to 97%. Notably, in the male data, Low_20% vs High_20% comparison, which detected the most DMCs, the classification accuracy reached 92.3%. This further supports the previous finding that the comparison between Low_20% and High_20% may be a more effective grouping method for studying the impacts of phthalate metabolite concentrations on human health.

Figure 4.

Clustering of the participants and DMCs. Plots A to D present the clustering heatmaps for Low_10% vs High_10%, Low_20% vs High_20%, Low_30% vs High_30%, and Low_40% vs High_40% in the male dataset, respectively. Plot E shows the results for the Low_20% vs. High_20% comparison in the female dataset. From plots A to E, the titles indicate the comparison groups along with their sample sizes (n). In the heatmaps, each row corresponds to a DMC, while each column represents a participant. The colour gradient from blue to red reflects the methylation levels, ranging from low to high. The colour on the phylogenetic tree along the y-axis distinguishes the DMCs, with green denoting up-regulated and orange indicating down-regulated DMCs. Additionally, the horizontal bar above the heatmap differentiates the participant groups, where blue representing those with low MEHP concentrations and red indicating high MEHP concentrations. Plot F presents the clustering accuracy and the number of DMCs for the comparison groups, corresponding to plots A to E from bottom to top. The orange bars represent the number of DMCs, while the blue bars indicate the clustering accuracy.

Comparison of DMCs across MEHP concentration groups

In total, 53 DMCs were identified from the comparisons of MEHP concentration groups. Among these, 34 are associated with coding genes, 9 with lncRNAs, 3 with both coding genes and lncRNAs, 2 with pseudogenes, and the remaining 13 are not linked to any known genes (Fig. S18). Among the 53 DMCs identified, 11 were consistently detected across multiple MEHP comparison groups (Fig. 5). Notably, all 11 DMCs were derived from the male dataset. Of these, 6 were associated with coding genes (Fig. 5B). The cg11336938, which is associated with MICB (MHC class I polypeptide-related sequence B), is the only DMC detected across all groups. This indicates that the regulatory mechanisms of innate immune responses related to MHC class I are particularly susceptible to the influences of phthalate exposure. Previous studies have also demonstrated that phthalates affect inflammatory responses, CD8 T cells, type I interferon, Th1 responses, and macrophage regulation, resulting in immune system suppression [14–18]. In the female dataset, only one comparison group (Low_20% vs High_20%) identified 3 DMCs, of which 2 are related to lncRNAs, and 1 is not associated with any known gene. These two lncRNAs (SLC22A18 and FLJ22536/CASC15) have both been reported as tumour suppressors [53–56]. This may explain the observed relationship between phthalates and cancer initiation, development, and recurrence reported in previous studies [24–28].

Figure 5.

Distributions of DMCs from male datasets. (A) The upset plot of all DMCs, with the rows of the dot matrix corresponding to the comparison groups and the columns representing the intersections between these groups. The bars show the number of DMCs. (B) Information on cross-group DMCs from the male MEHP comparison. The green, blue, brown, and black text represent the DMCs associated with coding genes, pseudogenes, lncRNAs, or unknown genes, respectively. The left column indicates whether the DMCs are up- or down-regulated. The middle column presents the associated genes and probe IDs of the DMCs. The black dots and lines in the right column indicate the comparison groups in which the DMCs were identified.

The functional analysis reveals the regulations influenced by MEHP

To further understand the regulatory and metabolic effects of MEHP on human health, functional analyses were performed. All the genes associated with the CpG probes were mapped to the GO terms. Additionally, the GO terms were traced back to level 4 (with the root as level 0) for the comparisons, ensuring result comparability and reducing data bias. In the DMCs, 23 sites were removed because they either were not associated with any genes or did not correspond to any GO terms. Among the three GO term categories (biological process, molecular function, and cellular component), the top 20 GO terms that accounted for at least 10% of the DMCs were further plotted in Fig. 6.

Figure 6.

The top 20 GO terms that account for >10% of the DMC list. The green, grey, and blue bars represent the percentages of the corresponding GO terms in the DMC list for molecular function, cellular component, and biological process, respectively.

In the molecular function category, most annotations are related to nucleotide or protein binding, involving various biological mechanisms such as immune response, transport, and transcriptional regulation. In addition, some annotations are associated with ion transporters or channels. These functional annotations align with those in the cellular component category, where the focus is primarily on membrane-associated locations. Regarding biological processes, the majority are related to metabolic processes, with some terms addressing stimulus response, transport, and cell cycle differentiation. These findings correspond to the protein or nucleotide-binding annotations observed in the molecular function category.

The genes associated with the DMCs identified across multiple MEHP comparison groups are also linked to the GO terms, which occupy a high proportion of the dataset (Fig. 6). As shown in Fig. 5B, the GO terms of the six cross-group DMC-associated genes are MICB, COL11A2 (collagen type XI alpha 2 chain), SCNN1A (sodium channel epithelial 1 subunit alpha), ALSCR11 (amyotrophic lateral sclerosis 2 chromosomal region candidate gene 11 protein, also known as C2CD6–C2 calcium dependent domain containing 6), MYOM1 (myomesin-1), and OLFM2 (olfactomedin 2). Both MICB (cg11336938) and COL11A2 (cg13189434) are involved in the mechanisms of innate immune response, which is closely related to the enriched GO terms such as protein or nucleotide-binding-related functions and stimulus response [14–18, 57]. Jaimes et al. [58] and Lu et al. [59] both discovered the inhibitory activity of sodium channels by the phthalates in 2019 and 2022, respectively. However, the mechanisms of this influence remain unclear. Since SCNN1A (cg06372539) is a subunit of the epithelial sodium channel, it may be the target genomic feature affected by MEHP at an early stage. Moreover, cg21834394 is associated with ALS2CR11/C2CD6, which plays an important role in sperm motility by regulating calcium channels in sperm, such as the CatSper channel [60]. In 2021, Hughes et al. [61] demonstrated that increased MEHP levels enhance CatSper activity, and in our methylation dataset, the DMC cg21834394 is also up-regulated at high MEHP concentrations. Furthermore, the influences of phthalate on the development and differentiation of skeletal and smooth muscle, which are associated with some enriched GO terms such as actin binding, cell–cell junction, and cell differentiation, have also been reported in several studies [62–64]. In 2022, Chen et al. [62] discovered that MEHP can inhibit myogenesis in murine skeletal muscle cells. Our analysis further found that MEHP may inhibit myogenesis through MYOM1 (cg12067287), a gene highly expressed in skeletal muscle that helps stabilize the internal muscle structure by linking thick filaments [65]. In addition to skeletal muscle cells, the differentiation and proliferation of smooth muscle are influenced by phthalate [63, 64]. In our finding, OLFM2 (cg19984804), which is involved in the regulation of smooth muscle cell differentiation, may also be one of the target sites affected by MEHP [63].

In order to further understand the interaction network in that the six cross-group DMC-associated genes are involved, STRING analysis revealed a clear network enriched in immune responses (Fig. S19), with MICB, COL11A2, and OLFM2 as the participants. Moreover, MICB (cg11336938) serves as a hub protein with 7 degrees of interaction, highlighting its significance, which aligns with the result that cg11336938 was identified as a DMC across all comparative groups for MEHP concentrations. These results indicate that the immune system may be the most sensitive of all regulation systems in responding to phthalate exposure. Although many biological regulations influenced by phthalate exposure have been identified in previous studies, their mechanisms largely remain unclear. The DMCs with their associated genes discovered in this study may offer insights into the influence of phthalate on human health at an early stage.

Discussion

Due to the rapid response of epigenetic modifications to environmental factors such as phthalates, DNA methylation data offer an accurate approach for the early detection of the impacts of environmental toxicants on human health. In this study, we conducted comprehensive analyses of phthalate metabolite concentrations in relation to DNA methylation data generated using the EPIC BeadChip. Participants were grouped based on their percentile ranges of phthalate metabolite concentrations, and systematic comparisons were conducted between groups that enables a comprehensive assessment of the differences in concentrations from the largest to the smallest. The analysis results revealed significant differences in phthalate metabolite concentrations between males and females, indicating that the levels of phthalate exposure and absorption in daily life may vary by sex, which can serve as a valuable reference for future phthalate studies. Furthermore, in the differential expression analysis, DMCs can only be identified in MEHP data. A total of 53 DMCs were identified, with 50 and 3 detected in the male and female datasets, respectively. The lower number of DMCs in females may be due to their lower MEHP concentrations relative to males (Fig. 2A). Within the DMCs, 11 were detected across multiple MEHP comparison groups in the male dataset. Specifically, of these 11 cross-group DMCs, 1 (cg11336938 associated with MICB) was detected in all concentration comparisons, while the remaining 10 were identified in two comparison groups. Moreover, the functional analysis revealed that most of the DMCs are related to protein or nucleotide binding, immune response, and ion channels. Several of these functions have been shown to be associated with phthalate exposure in previous studies. The DMCs identified in this study may benefit to elucidate more detailed and comprehensive mechanisms of the regulatory networks affected by phthalate exposure.

This study is characterized by the use of data collected from a daily life setting, covering 389 cancer-free adult participants, whereas many previous studies have primarily focused on children or pregnant women [66–69], such as maternal phthalate exposure [68] and its effects on placental physiology [67]. Notably, our study focuses on analysing the concentrations of 15 phthalate metabolites in urine samples, as urine is widely regarded as the preferred medium for assessing recent exposure [70]. However, it is still a challenge to obtain information regarding the routes of phthalate exposure among participants since most phthalate metabolites can be excreted in urine within 2 days [70]. Moreover, this study lacks information of the potential comorbidities related to phthalate toxicity, such as allergic diseases, asthma, or eczema [71]. Further research is needed to elucidate the health effects of different types of phthalate exposure and explore possible interventions, such as modifying dietary habits to reduce intake of HMWPs or substituting personal care products to minimize exposure to LMWPs [72].

The dataset for this study is sourced from the TWB, which includes only Taiwanese individuals. Therefore, it may not fully represent the relationship between phthalate metabolite concentrations and DNA methylation across all ethnic populations. Furthermore, the participants are cancer-free and exhibit no significant differences in various clinical and blood tests, which may explain the relatively lower number of DMCs compared to studies analysing disease samples. While obtaining DNA methylation and urinary phthalate metabolite data from the same sample source is inherently challenging, this study provides a unique and valuable dataset for exploring the early epigenetic effects of phthalate exposure. Although no independent replication dataset was used, we employed multiple concentration-based groupings to enhance result reliability and minimize false positives. Moving forward, we aim to further strengthen our conclusions by seeking or establishing additional datasets for validation. Nevertheless, the DMCs identified in this study will serve as an important foundation for future research on the early effects of phthalate exposure.

In this study, only MEHP had a significant impact on DNA methylation. Although HWMP also contains MEHP, its effect may have been obscured by the presence of other phthalates, leading to no significant association in the analysis (Fig. S6). MEHP is recognized as one of the most toxic and biologically active phthalates and serves as a key precursor for many derivative metabolites [73, 74]. MEHP is also known to activate various nuclear receptors, including peroxisome proliferator-activated receptors [75], which regulate lipid metabolism and adipogenesis, as well as sex hormone receptors [76].

Among the 53 DMCs, 24 sites are not associated with any currently known coding genes; instead, they are related to lncRNA pseudogenes or have no gene association at all. Therefore, the functions affected by these DMCs are not yet clearly understood. These DMCs may also be involved in functional pathways identified from the DMCs associated with coding genes, such as immune response or ion channels; however, further experimental validations are required.

In conclusion, this study utilized Illumina Infinium MethylationEPIC BeadChip array data from 389 cancer-free individuals to investigate the effects of phthalates on human health. In total, 53 DMCs were associated with MEHP exposure, with 11 of these DMCs were consistently detected across multiple concentration comparisons. The DMC-associated genes are primarily involved in immune response, ion channels, and nucleotide/protein binding. These findings provide critical insights into the epigenetic mechanisms underlying phthalate exposure and its potential health implications, serving as a foundation for future research and clinical applications.

Author contributions

Ping-Hsun Wu (Conceptualization [equal], Data curation [equal], Funding acquisition [equal], Investigation [equal], Methodology [equal], Writing—original draft [equal], Writing—review & editing [equal]), Shiau-Ching Chen (Data curation [equal], Formal analysis [equal], Investigation [equal], Methodology [equal], Software [equal], Visualization [equal]), Chun-Jui Chien (Formal analysis [supporting], Methodology [supporting], Visualization [equal]), Johnathan Lin (Methodology [supporting], Writing—review & editing [supporting]), Hsiang-Ying Lee (Funding acquisition [equal], Investigation [supporting]), Yi-Ting Lin (Conceptualization [supporting], Investigation [supporting]), Ting-Chia Weng (Funding acquisition [equal], Investigation [supporting]), Ping-Chi Hsu (Investigation [supporting], Methodology [supporting]), Ming-Tsang Wu (Data curation [supporting], Resources [lead]), and Sung-Huan Yu (Conceptualization [lead], Data curation [lead], Formal analysis [lead], Funding acquisition [lead], Investigation [lead], Methodology [lead], Project administration [lead], Software [lead], Supervision [lead], Validation [lead], Visualization [lead], Writing—original draft [lead], Writing—review & editing [lead])

Conflict of interest

None declared.

Funding

This work was supported by the National Science and Technology Council, Taiwan (grant number NSTC 113-2221-E-110-079), a joint research project of National Sun Yat-sen University and Kaohsiung Medical University (NSYSU-KMUJOINTRESEARCHPROJECT—#NSYSUKMU 114-I01 and NSYSUKMU 111-P21), Pingtung Veterans General Hospital (PTVGH-11446), a joint research collaboration between National Kaohsiung University of Science and Technology and Kaohsiung Medical University (NKUST-KMUJOINTRESEARCHPROJECT-114KK004), and Kaohsiung Medical University Hospital, Taiwan (KMUH110-0M13, KMUH111-1M09, KMUH112-2M08, and KMUH113-3R19). This work was also supported partially by the Research Center for Precision Environmental Medicine, Kaohsiung Medical University, Kaohsiung, Taiwan, from the Featured Areas Research Center Program within the framework of the Higher Education Sprout Project by the Ministry of Education (MOE) in Taiwan and by Kaohsiung Medical University Research Center Grant (KMU-TC114A01).

Data availability

The data used in this study were obtained from the Taiwan Biobank (https://www.biobank.org.tw/english.php) under restricted access. The data are not publicly available due to privacy and ethical restrictions. However, researchers can apply for access to the Taiwan Biobank dataset through their official application process.