Experimental Environment Is a Determinant of Gene Methylation and One-Carbon Metabolism in Obese Adult Mice

Abstract

Background

Diet-induced obesity (DIO) leads to insulin resistance (IR) and alters gene expression through epigenetic mechanisms, including DNA methylation. Here, we aimed to investigate whether experimental environment is an important variable in determining DNA methylation and one-carbon metabolism in DIO mice fed a multi-vitamin-mineral mixture (MVM).

Methods

Three experiments with identical design were conducted in three independent animal facilities (i.e., experimental environments or locations). In each location, 12-week-old male C57BL/6J mice were randomly assigned to two dietary groups: high-fat (HF) and HF-MVM for an average of 10 weeks. Global and gene-specific methylation of adipose function related genes in epididymal white adipose tissue (eWAT), and insulin signaling genes in the liver were analyzed using bisulfite pyrosequencing. Hepatic 1-C metabolites were measured and the ratio of S-adenosylmethionine (SAM) and S-adenosylhomocysteine (SAH) was used as an indicator of methylation potential.

Results

Experimental location affected global methylation patterns in the eWAT, but not in the liver. At the gene-specific level, experimental location, MVM, and their interaction altered the methylation of genes related to adipose function (Srebf1, Acaca, Fasn, Pparg, and Rbp4) in the eWAT and insulin signaling (Pi3kr1 and Akt1) in the liver (p < 0.05). The SAM/SAH ratio was correlated with gene-specific methylation at some CpG sites of Srebf1, Pi3kr1, Acaca, Fasn, Pparg, Rbp4, and Akt1) genes (p < 0.05).

Conclusion

The experimental environment is a significant determinant of the effects of micronutrient supplement on 1-C metabolism and the methylation of genes associated with IR in tissues of DIO adult male mice.

Introduction

The global increase in obesity and its related metabolic consequences, such as insulin resistance (IR) and type 2 diabetes, poses a significant threat to public health [1]. Obesity is a multifactorial condition resulting from the interaction between genetic predispositions and the environment [1]. The environment is a significant modifier of the epigenome, which consists of chemical modifications to DNA or chromatin structures that alter gene expression without changing the DNA sequence itself [2, 3].

Until recently, the epigenome was thought to be established during early development and then remain tightly regulated throughout the lifespan; however, it is now well-recognized that environmental factors change epigenetic patterns postnatally [4]. Experimental variations among animal facilities (i.e., experimental environment) including the size and color of the cage [5], neighboring rearing cages of the animals [6], and sex of the personnel [7], affect the phenotypes and epigenome of the animals, potentially by inducing varying levels of stress [8]. These factors are likely to be modifiers of the regulatory function of the epigenome on gene activity via several epigenetic mechanisms, including DNA methylation. However, gene activity can be modified by various mechanisms beyond the epigenetic environment, primarily through changes in transcription, translation, and post-translational modifications of proteins. In mammals, DNA methylation predominantly involves the addition of a methyl group to the cytosine within CpG dinucleotides, playing a role in gene expression regulation [2]. Micronutrients, such as vitamins and minerals, play an essential role in the one-carbon (1-C) metabolism, supplying S-adenosyl methionine (SAM), the universal methyl donor for DNA methylation [9]. Elevated levels of S-adenosylhomocysteine (SAH), the byproduct of SAM-mediated methylation reactions, can also influence this process. The genome-wide and locus-specific changes in DNA methylation have been associated with diet-induced obesity (DIO) and IR [10].

We recently reported that adding a multi-vitamin-mineral mix (MVM), consisting of vitamin A, B1, B6, B12, zinc (Zn), and selenium (Se), to a high-fat (HF) diet fed to adult mice reduced weight gain, IR and modified expression of genes in lipogenesis, adipogenesis, and adipokine synthesis pathways in white adipose tissue and the insulin signaling pathway in the liver [11]. The effect of the MVM was accompanied by a shift in 1-C metabolism in the liver toward favoring DNA methylation [11]. Given the growing recognition that preclinical studies are often not reproducible when conducted by investigators in other centers [12, 13], we designed a study to examine the effects of variability from the experimental environment on the 1-C metabolites and DNA methylation of the genes targeted in our previous studies [11]. To do so, we conducted the same experiment in three different animal facilities at the University of Toronto. Since environmental factors can influence the epigenome, we hypothesized that they represent an unaccounted variable affecting the outcomes of studies examining the effects of MVM on gene methylation and 1-C metabolism in obese adult mice. This could, in turn, explain the lack of reproducibility often observed in preclinical experiments.

Methods

Animals, Diets, and Experimental Procedures

The experiments were conducted between December 2020 and October 2022 in three independent animal housing facilities (experimental locations): the Department of Comparative Medicine (DCM) at the University of Toronto, the Centre for Phenogenomics (TCP) located at Mount Sinai Hospital, and the Terrence Donnelly Centre for Cellular & Biomolecular Research (CCBR) located at University of Toronto. Ethical approvals were obtained from the University of Toronto Office of Research Ethics (protocol No. 20012670 and No. 20012824) and the Animal Care Committee of the Centre for Phenogenomics (protocol No. 25-0372H).

In each experiment, 10-week-old male C57BL/6J mice with DIO (Stock No: 380050), obtained from Jackson Lab (Bar Harbor, ME, USA), were provided housing conditions included a 12:12-h light-dark cycle (lights on at 0700) with a temperature maintained at 22°C ± 1°C, as well as ad libitum access to food and water. At 12 weeks of age (after 2 weeks of acclimatization), all mice were randomly allocated to one of two dietary groups: a HF diet (60% energy from fat) as the control group, or a HF diet supplemented with a multi-vitamin-mineral mix (MVM), providing a total of 5-fold the recommended for the AIN-93M diet amounts of vitamins A, B1, B6, B12, zinc (Zn), and 2-fold selenium (Se). Mice were maintained on their corresponding diets for either 9 weeks (DCM, n = 12/group, 24 mice in total), 10 weeks (TCP, n = 15/group, 30 mice in total), or 12 weeks (CCBR, n = 12/group, 24 mice in total).

The two experimental diets, HF and HF-MVM diets, were purchased from Research Diets (Research Diets Inc., New Brunswick, NJ, USA). They were isocaloric and had similar macronutrient, vitamin, and mineral compositions, differing only in vitamins A, B1, B6, B12, Zn, and Se (online suppl. Table 1; for all online suppl. material, see https://doi.org/10.1159/000546727). The selection of micronutrient amounts added to the HF-MVM diet was based on previous relevant preclinical studies [11, 14, 15] and randomized control trials with similar methodologies [16–18], which showed positive effects on metabolic control. The dose of MVM was calculated relevant to the rodent nutrient requirements (AIN-93M, National Research Council, a total of 5-fold the recommended for the AIN-93M diet amounts of vitamins A, B1, B6, B12, Zn, and 2-fold Se) and then converted to human equivalents using nutrient density scaling [19] to ensure that the added amounts fell within a range below the current upper tolerance level (UL).

At the end of each experiment, following a 6-h daytime water-only fasting period, mice were anesthetized with isoflurane inhalation, euthanized via cervical dislocation, and tissues were collected. Epididymal white adipose tissue (eWAT) and liver samples were weighed, snap-frozen, and stored at −80°C for future analyses. Gene expression related to lipogenesis (Srebf1, Acaca, and Fasn), adipogenesis (Pparg), adipokine synthesis (Rbp4) pathways in eWAT, and the insulin signaling pathway (Pi3r1 and Akt1) in the liver of DIO mice from three experimental locations was measured by qPCR in our research laboratory at the University of Toronto. The detailed methods and results previously documented [20] and are summarized in the Table 1. The genes of interest were selected based on their roles in metabolic syndrome [11] and their altered expression in response to MVM supplementation and/or experimental location, as previously reported [11, 20].

Table 1.

Relative mRNA expression of genes in the three locations, partially adapted from reference [20]

Gene	Treatment	Location			p values
		DCM	TCP	CCBR	MVM	location	MVM*Location
eWAT
Srebf1	HF	1.06a,b±0.16	1.04a,b±0.11	1.01a,b±0.05	0.444	0.065	0.038
	HF-MVM	1.17a±0.12	1.05a,b±0.06	0.73b*±0.1	​	​	​
Fasn	HF	1.04a,b±0.15	1.07a,b±0.2	1.13a,b±0.22	0.071	0.016	0.007
	HF-MVM	1.53a,*±0.15	0.56b,*±0.05	0.71b±0.13	​	​	​
Acaca	HF	1.01a,b±0.06	1.17a±0.28	1.05a,b±0.12	0.039	0.173	0.070
	HF-MVM	1.22a±0.09	0.75a,b±0.13	0.61b,*±0.15	​	​	​
Rbp4	HF	1.06a,b±0.21	1.12a,b±0.20	1.11a,b±0.19	0.049	0.006	0.001
	HF-MVM	1.77a,*±0.23	0.47b,*±0.06	0.45b,*±0.05	​	​	​
Pparg	HF	1.01±0.06	1.08±0.16	1.02±0.06	0.334	0.685	0.244
	HF-MVM	1.34*±0.09	1.00±0.09	0.88±0.10	​	​	​
Liver
Pik3r1	HF	1.03±0.01	1.05±0.06	1.02±0.08	0.216	0.700	0.652
	HF-MVM	1.23±0.03	1.05±0.09	1.16±0.06	​	​	​
Akt1	HF	1.01±0.07	1.00±0.02	1.00±0.03	0.983	0.076	0.196
	HF-MVM	1.23±0.08	0.93±0.03	0.93±0.05	​	​	​

Values are mean ± SEM, n = 5–8/group. A two-way ANOVA was conducted with MVM (HF or HF-MVM) and Location (DCM, TCP, and CCBR) as main factors and an MVM*Location interaction term. A Tukey’s post hoc analysis adjusted for multiple comparisons followed all significant effects

^a,bSignificantly different at p < 0.05 by Tukey’s post hoc analysis. A t test was used to compare the difference between HF and HF-MVM groups within each Location.

Significant differences (p < 0.05) are indicated by an asterisk.

*p < 0.05.

Hepatic 1-C Metabolites

The liver was the primary targeted tissue for measuring 1-C metabolites, as it serves as the central organ for producing 1-C units and maintaining the homeostasis of the 1-C metabolism among non-proliferative adult tissues [21]. Approximately 100 mg of snap-frozen liver samples were weighed and sent to the Center of Metabolomics Research, Baylor Scott & White Research Institute (Dallas, TX, USA) for analysis. The concentrations of 1-C metabolites, including S-adenosylmethionine (SAM), SAH, cystathionine, methionine, betaine, and choline, were quantified using liquid chromatography with tandem mass spectrometry (LC-MS/MS) [22]. The SAM to SAH ratio was calculated and used as an indicator of DNA methylation potential [23].

DNA Methylation Analyses

All DNA methylation analyses were conducted at the Center for Applied Genomics, SickKids Research Institute, Toronto, Canada. DNA extraction from tissues (eWAT and liver) was performed using the DNeasy Blood and Tissue Kit following the kit protocol (Cat# 69504, Qiagen, Valencia, CA, USA). Gene-specific methylation was measured using a quantitative pyrosequencing assay of sodium bisulfite converted DNA (1–2 μg) using the EpiTect Plus DNA Bisulfite Kit (Cat# 59124, Qiagen, Valencia, CA, USA) as previously described [24]. Long interspersed nuclear element-1 (LINE-1) methylation was used as a proxy for assessing global change in DNA methylation levels in both eWAT and the liver [24]. In each sample, methylation levels (percentages) of targeted CpG sites were analyzed and subsequently averaged within the promoter region of the targeted genes [24], namely, Srebf1, Acaca, Fasn, Pparg, and Rbp4 in the eWAT and Pi3r1 and Akt1 in the liver. These genes were selected based on their roles in obesity and IR and a comprehensive rationale of selection was documented in our previous publication [11]. The sequencing reactions were carried out using the PyroMark Q24 pyrosequencer and accompanying software (Cat# 9001514, Qiagen, Valencia, CA, USA). Online supplementary Table 2 contains information regarding PCR primers and sequencing primers used in the pyrosequencing assays for LINE-1 and the selected genes. Additionally, online supplementary Figures 1–7 provide details on the specific locations of the targeted CpG sites. ALGGEN-PROMO bioinformatic program was used to identify potential transcription factor binding sites at targeted CpG sites [25].

Statistical Analysis

Data analysis was conducted using R Studio (Version 1.4.1106, Posit Software, Boston, MA, USA). Robust regression analysis was employed to identify and exclude outliers from the dataset. A two-way ANOVA was conducted with MVM supplementation (with or without) and experimental location (DCM, TCP, or CCBR) as primary factors and included an MVM × experimental location interaction term. Tukey’s post hoc analysis was conducted to identify mean differences, adjusting for multiple comparisons. Student t tests were used to compare the HF and HF-MVM groups within each experimental location. Additionally, correlation analyses between DNA methylation status and gene expression as well as DNA methylation status and 1-C metabolites were performed by using Pearson’s correlation coefficient. All values were reported as mean ± standard error of the mean (SEM).

Results

Effects of Experimental Location and MVM on Gene Expression in the eWAT and the Liver

In the eWAT, the experimental location alone influenced the expression of two genes (Fasn and Rbp4), while MVM affected two (Acaca and Rbp4) out of the five genes (p < 0.05, Table 1) [20]. Additionally, an interaction between MVM and experimental location was observed for six of the eight genes involved in lipogenesis (Srebf1 and Fasn) and adipokine synthesis (Rbp4) (p < 0.05, Table 1). In contrast, no significant effects of MVM or its interaction with experimental location were detected in the liver.

Effects of Experimental Location and MVM on Global DNA Methylation (LINE-1) in the eWAT and the Liver

In the eWAT, experimental location affected methylation of CpG site 4 (p < 0.05, Fig. 1a). However, MVM did not affect global methylation status, neither across five CpG sites nor the average methylation percentage (p > 0.05, Fig. 1a). No interaction between the effect of MVM and experimental location was observed (p > 0.05, Fig. 1a). In the liver, neither experimental location nor MVM supplementation nor their interaction affected global methylation (p > 0.05, Fig. 1b).

Fig. 1.

Percentage of global methylation (LINE-1) and average methylation in the eWAT (a) and the liver (b) of mice fed HF or HF-MVM diet. Two-way ANOVA was coducted, n = 5–6 animals/group.

Effects of Experimental Location and MVM on DNA Methylation in Genes Related to the Lipogenesis Pathway in the eWAT

Srebf1

Experimental location had a significant impact on CpG 3, 4, and 5 (p < 0.05, Fig. 2a). At CpG 3 and 4, the HF-MVM mice in the DCM had higher methylation than mice in the other two facilities (Fig. 2a). MVM, but not experimental location, affected the methylation percentage at CpG 1 (p = 0.002, Fig. 2a) of the gene Srebf1. Compared to the HF group, it reduced the methylation by 50% in the TCP (p < 0.001). At CpG 2, MVM induced higher methylation in both the TCP and the CCBR (p < 0.05, Fig. 2a). The effect of MVM at CpG 3 also depended on the experimental location (interaction p < 0.001). MVM increased the methylation percentage by 50% in the DCM. However, no effect of MVM, experimental location, or their interaction was observed from CpG 5–9, as well as in the average methylation percentage of all investigated CpGs (p > 0.05).

Fig. 2.

Gene specific methylation (%) and average methylation of targeted CpG sites in genes related in lipogenesis pathway (Srebf1 (a), Acaca (b), and Fasn (c)) in the eWAT of adult male mice fed HF or HF-MVM diet. Two-way ANOVA followed by Tukey’s post-hoc test (p < 0.05). Letters indicate differences amongst groups; T test was conducted between HF and HF-MVM mice within each experimental location, n = 5–6 animals/group. *p < 0.05, **p < 0.01.

The mRNA expression of Srebf1 was negatively correlated with the methylation of CpG 8 (p < 0.05, Fig. 3a). Additionally, there was a positive trend between mRNA expression of the gene and methylation of CpG 3 (p = 0.07, Fig. 3b).

Fig. 3.

a, b The correlations between mRNA expression and the methylation percentage of CpGs in the Srebf1 of eWAT in DIO mice in three experimental locations (△: DCM-HF; ▲: DCM-HF-MVM; □: TCP-HF; ■: TCP-HF-MVM; ○: CCBR-HF; ●: CCBR-HF-MVM).

Acaca

For the gene Acaca, experimental location affected the average methylation percentage of the gene, showing higher methylation in the DCM compared to the TCP location (p < 0.05, Fig. 2b). Moreover, the effect of MVM on the average methylation was dependent on the experimental location (interaction p = 0.017), with HF-MVM having a higher percentage of methylation compared to HF in the DCM location (p < 0.05). However, no significant MVM effects were observed across the three targeted CpG sites (p > 0.05, Fig. 2b).

Fasn

Experimental location, but not MVM, affected the methylation of CpG 4 (p < 0.05) and the average methylation percentage (p < 0.001, Fig. 2c) in Fasn. Both HF and HF-MVM mice in DCM had higher methylation levels than the HF-MVM mice in the CCBR at the average methylation level (p < 0.05, Fig. 2c). No interaction between MVM and experimental location was observed at any of the CpGs investigated in Fasn (p > 0.05).

The mRNA expression of Fasn was positively correlated with the methylation of CpG 4 (p < 0.05, Fig. 4a) and the average methylation percentage across nine CpGs (p < 0.01, Fig. 4c). Furthermore, a positive trend was observed between Fasn expression and the methylation of CpG 8 (p = 0.059, Fig. 4b).

Fig. 4.

a–c Correlations between mRNA expression and the methylation percentage of CpGs in the Fasn of eWAT in DIO mice in three experimental locations (△: DCM-HF; ▲: DCM-HF-MVM; □: TCP-HF; ■: TCP-HF-MVM; ○: CCBR-HF; ●: CCBR-HF-MVM).

Effects of Experimental Location and MVM on DNA Methylation in Genes Related to Adipogenesis (Pparg) in the eWAT

Experimental location had an effect in the first two sites, with the lowest levels of methylation at CpG 1 in the TCP and the highest levels at CpG 2 in the DCM, compared to the other two facilities (p < 0.05). Moreover, the addition of MVM to the HF diet resulted in a decreased methylation at CpG 2 (p = 0.006) and 3 (p = 0.02, Fig. 5). However, no interaction between MVM and experimental location was observed (p > 0.05).

Fig. 5.

Gene specific methylation (%) in 5 CpG sites and average methylation in Pparg in the eWAT of adult male mice fed HF or HF-MVM diet. Two-way ANOVA followed by Tukey’s post hoc test (p < 0.05). Letters indicate differences amongst groups; T test was conducted between HF and HF-MVM mice within each experimental location. N = 5–6 animals/group. *p < 0.05, **p < 0.01.

Effects of Experimental Location and MVM on DNA Methylation in Genes Related to Adipokine (Rbp4) in the eWAT

Experimental location affected CpG 1, 2, 6, 7, and the average methylation across the three experiments (p < 0.05, Fig. 6). Adding MVM to the HF diet also increased methylation at CpG sites and the average methylation level within the Rbp4 gene (p < 0.05), except for CpG 4 (p = 0.080) and CpG 6 (p = 0.061), which showed an increasing trend (Fig. 6). No significant interaction between MVM and experimental location was observed across all CpGs in Rbp4 (p < 0.05). At CpG 5, HF-MVM and HF mice in the CCBR had the highest and the lowest methylation levels, respectively, among all groups across three experimental locations (p < 0.05). The CpG 6 in Rbp4 of HF-MVM mice in the CCBR had the highest methylation but in both HF and HF-MVM mice from the TCP had the lowest (p < 0.05). Additionally, at CpG 7, MVM-fed mice in the DCM had the highest methylation, and HF mice in the TCP had the lowest methylation percentage (p < 0.05). For CpG 8, HF-MVM mice had the highest methylation in the CCBR, over 25% higher than HF mice in the TCP and the CCBR.

Fig. 6.

Gene specific methylation (%) in 8 CpG sites and average methylation in Rbp4 in the eWAT of adult male mice fed HF or HF-MVM diet. Two-way ANOVA followed by Tukey’s post hoc test (p < 0.05). Letters indicate differences amongst groups; T test was conducted between HF and HF-MVM mice within each experimental location. N = 5–6 animals/group. *p < 0.05.

Effects of Experimental Location and MVM on DNA Methylation in Genes Related to Insulin Signaling Pathway (Pi3kr1 and Akt1) in the Liver

Pi3kr1

The effect of adding MVM to the HF diet was dependent on the experimental location at the CpG 4 (interaction p < 0.001), where MVM decreased the methylation by over 50% in location the DCM (p < 0.05) but increased it by a similar magnitude in the TCP (p < 0.05), with no effect in the CCBR (p > 0.05). Experimental location had a main effect on CpG 1 (p = 0.02), 4 (p < 0.001), and CpG 6 (p = 0.02, Fig. 7a). At CpG 1 in the CCBR and CpG 2 in TCP, HF-MVM mice had increased methylation compared to their HF counterparts (p < 0.05, Fig. 7a).

Fig. 7.

Gene specific methylation (%) and average methylation in CpG sites in genes related to insulin signaling pathway (Pi3kr1 (a) and Akt1 (b)) in the liver of adult male mice fed HF or HF-MVM diet. Two-way ANOVA followed by Tukey’s post hoc test (p < 0.05). Letters indicate differences amongst groups; T test was conducted between HF and HF-MVM mice within each experimental location, n = 5–6 animals/group. *p < 0.05, **p < 0.01.

Akt1

For Akt1, experimental location had a main effect on CpG 4 (p = 0.008), 6 (p = 0.019), and the average of all the CpGs (p = 0.002, Fig. 7b). MVM addition to the HF diet increased the methylation of the CpG 5 (p = 0.032, Fig. 7b). Its effect also interacted with experimental location at CpG 2 (p = 0.034) and 6 (p = 0.007), showing a more than 70% increase in methylation in HF-MVM compared to HF mice in the TCP (p < 0.05), and a decrease of about 50% in methylation levels in HF-MVM compared to HF at the DCM (p < 0.05).

Effects of Experimental Location and MVM on Hepatic 1-C Metabolites

There was a significant location effect observed across all the metabolites (p < 0.01 for all, Table 2). Both HF and HF-MVM mice had higher levels of SAM in TCP than both groups in the CCBR as well as HF mice in the DCM (p < 0.05, Table 2). Mice in the DCM had the highest levels of methionine, more than 1.5 times the concentrations of methionine compared to the mice in the CCBR, which itself had more than twice the levels observed in the mice in the TCP (p < 0.05, Table 1). As for choline, mice in the TCP had half the amount compared to those in the DCM and the CCBR (p < 0.05, Table 2).

Table 2.

Hepatic 1-C metabolites in mice across the three experimental locations

Metabolites	Treatment	Location			p values
		DCM	TCP	CCBR	MVM	location	MVM*Location
SAM, nmol/g tissue	HF	43.25c±4.04	68.88a±1.71	48.60c±3.15	0.309	<0.001	0.030
	HF-MVM	55.60b,c,*±2.90	66.38a,b±3.03	47.90c±2.38	​	​	​
SAH, nmol/g tissue	HF	59.13a±4.59	23.88c±1.49	25.59c±1.94	0.005	<0.001	0.001
	HF-MVM	40.00b,*±3.59	22.45c±1.77	25.25c±2.13	​	​	​
SAM/SAH	HF	0.80c±0.13	3.00a±0.16	1.94b±0.10	0.059	0.001	0.109
	HF-MVM	1.54c,*±0.18	2.96a±0.24	2.04b±0.18	​	​	​
Methionine, nmol/g tissue	HF	157.78a±8.55	44.10c±3.28	87.95b±6.20	0.336	<0.001	0.259
	HF-MVM	161.59a±4.56	39.99c±2.89	103.13b±8.25	​	​	​
Cystathionine, nmol/g tissue	HF	21.53a±2.73	10.23c±0.40	8.70c±1.00	0.031	<0.001	0.109
	HF-MVM	15.56b±1.79	9.09c±0.82	8.03c±0.53	​	​	​
Betaine, nmol/g tissue	HF	405.06a,b±26.50	525.19ab±29.05	394.08b±31.85	0.040	0.002	0.369
	HF-MVM	541.59a,b,*±36.3	572.18a±56.58	432.17a,b±38.93	​	​	​
Choline, nmol/g tissue	HF	413.50a±41.51	209.42b±24.59	366.25a±42.91	0.647	<0.001	0.885
	HF-MVM	394.55a±43.16	182.63b±18.99	372.00a±36.07	​	​	​

Values are mean ± SEM, n = 9–13/group. A two-way ANOVA was conducted with MVM (HF or HF-MVM) and location (DCM, TCP, and CCBR) as main factors and an MVM × location interaction term. A Tukey’s post hoc analysis adjusted for multiple comparisons followed all significant effects.

SAM, S-Adenosyl methionine; SAH, S-Adenosyl-homocysteine.

^a,b,cSignificantly different at p < 0.05 by Tukey’s post hoc analysis. A t test was used to compare the difference between HF and HF-MVM groups within each location.

Significant differences (p < 0.05) are indicated by an asterisk.

MVM had main effects on the hepatic levels of SAH (p < 0.01), cystathionine (p < 0.05), and betaine (p < 0.05, Table 2) across the three experiments. HF mice in the DCM had the highest hepatic SAH and cystathionine levels, 50% higher than those in MVM-fed mice in the DCM, and more than twice the levels in mice in the TCP and the CCBR (p < 0.05, Table 2). For betaine, HF-MVM mice in the TCP had the highest level, and the HF mice in the CCBR had the lowest (p < 0.05). Moreover, there was a trend indicating the effects of the MVM on the SAM/SAH (p = 0.059, Table 2) across the three experimental locations. Mice in the TCP had the highest SAM/SAM, followed by those in the CCBR and the DCM (p < 0.05, Table 2). Additionally, the effect of MVM on both SAM and SAH depended on the experimental location (interaction: p < 0.05, Table 2), although no significant interactions were observed for SAM/SAH ratios (p = 0.109, Table 2) or any other measured metabolites (p > 0.05, Table 2).

Correlations between SAM/SAH, SAM, or SAH, and Global/Gene-Specific DNA Methylation in Both Tissues

At the global level, there were no correlations between the SAM/SAH ratio and the DNA methylation in either the eWAT or liver (p > 0.05, data not presented). However, SAM correlated positively with methylation of CpG 1 (r = 0.4145, p = 0.0120), CpG 2 (r = 0.3678, p = 0.0273), and CpG 3 (r = 0.3655, p = 0.0284), as well as with the average methylation (r = 0.3745, p = 0.0244) in the eWAT.

At the gene-specific level, both SAM/SAH and SAH had significant correlations with the methylation of at least one CpG site in all investigated genes in both tissues (Table 3). For SAM/SAH, there were positive correlations between SAM/SAH ratio and specific CpGs in Srebf1 (CpG 6) in the eWAT and Pi3kr1 (CpG1) in the liver, and negative correlations between SAM/SAH ratio and specific CpGs in Acaca (average methylation of all CpGs), Fasn (CpG 9 and average), Pparg (CpG 1), Rbp4 (CpG 7) in the eWAT, and Akt1 (CpG 6) in the liver (p < 0.05). For SAM alone, in eWAT, only CpG1 and 2 of Pparg had negative correlations with SAM (p < 0.05 Table 3). In the liver, different CpGs from Pi3kr1 and Akt1 showed both positive (CpG1 in Pi3kr1 and 2 in Akt1) and negative correlations (CpG 3 in Pi3kr1 and CpG 6 in Akt1) with this metabolite (p < 0.05). For SAH alone, all correlations observed between SAH and CpGs in all genes were significantly positive in both tissues (p < 0.05, Table 3).

Table 3.

Correlations between SAM/SAH, SAM, or SAH and DNA methylation percentage of CpGs investigated in all investigated genes in DIO mice across the three experimental locations

	SAM/SAH	SAM	SAH
Srebf1 (eWAT)
CpG 4
r	–	–	0.3778
p value	NS	NS	0.0302
CpG 6
r	0.4545	–	0.3655
p value	0.0069	NS	0.0335
CpG 7
r	NS	–	0.3437
p value	NS	NS	0.0446
Average of all
r	–	–	0.3558
p value	NS	NS	0.0359
Acaca (eWAT)
Average of all
r	−0.3481	–	–
p value	0.0404	NS	NS
Fasn (eWAT)
CpG 3
r	–	–	0.3724
p value	NS	NS	0.0267
CpG 9
r	−0.4591	–	0.5542
p value	0.0063	NS	<0.001
Average of all
r	−0.5009	–	0.6939
p value	0.0026	NS	<0.001
Pparg (eWAT)
CpG 1
r	−0.3337	–	–
p value	0.0501	NS	NS
CpG 2
r	−0.3723	–	0.3272
p value	0.0277	NS	0.0551
CpG 4
r	–	0.3631	–
p value	NS	0.0296	NS
Rbp4 (eWAT)
CpG 7
r	−0.4312	–	0.3678
p value	0.0137	NS	0.0379
Pi3kr1 (liver)
CpG 1
r	0.4340	0.3476	–
p value	0.0103	0.0408	NS
CpG 3
r	–	−0.3696	–
p value	NS	0.0315	NS
CpG 4
r	–	–	0.5800
p value	NS	NS	0.0006
CpG 6
r	–	–	0.4331
p value	NS	NS	0.0118
Akt1 (liver)
CpG 2
r	–	0.3885	–
p value	NS	0.0232	NS
CpG 6
r	−0.5066	−0.4249	0.3706
p value	0.0036	0.0154	0.0402

NS, not significant.

Discussion

The results show that the experimental environment is a significant determinant of the methylation of genes associated with IR in the adipose and liver of DIO adult male mice fed a MVM diet. While the environment was a primary factor influencing methylation patterns, it also interacted with the effects of MVM, collectively affected the methylation and expression of genes linked to DIO and IR in these tissues. Alterations in 1-C metabolites suggest that both experimental environment and MVM supplementation modulate the 1-C cycle, potentially contributing to epigenetic modifications. The study is also the first to show that environmental factors interact with dietary interventions in the 1-C cycle, ultimately influencing gene methylation and function.

The effects of both experimental location and MVM on gene methylation are supported by results from both global and gene specific levels. At the global level, experimental location affected global methylation patterns in the eWAT, but not in the liver. However, MVM did not affect DNA methylation in either the eWAT or the liver. The differences in DNA methylation patterns between the adipose tissue and the liver may reflect their distinct functional roles, developmental pathways, and responses to environmental cues, all of which are crucial for maintaining metabolic homeostasis and overall health. At the gene specific level, the effects of MVM and experimental location increased or decreased methylation in different CpG sites in the same genes and these site-specific methylation changes were not always reflected in their average CpG methylation percentages. Nonetheless, methylation of individual CpGs is a distinct and well-established epigenetic mechanism of regulation of gene expression and function [26, 27].

In this study, the targeted CpGs were in the promoter region, downstream of the transcription start site (TSS) in genes involved in lipogenesis (Srebf1, Acaca, and Fasn), adipogenesis (Pparg), and adipokine synthesis (Rbp4) in the eWAT as well as insulin signaling pathways (Pi3kr1 and Akt1) in the liver. However, significant correlations between gene expression and methylation percentages were observed only in Srebf1 and Fasn. This suggests that mRNA expression of the other five genes may not be directly influenced by DNA methylation within our targeted CpG regions within the promoter region, but rather by CpGs located upstream of the TSS or outside of the promoter region, or by other epigenetic mechanisms regulating gene expression.

In the eWAT, both experimental location and MVM affected the methylation status in Srebf1, Acaca, and Fasn, genes that encode corresponding enzymes functioning in lipogenesis in the eWAT. In obesity and diabetes, they are often downregulated in white adipose tissues in both humans and rodents [28]. Sterol regulatory element-binding transcription factor 1 (SREBF1) initiates lipogenesis, primarily through regulating SREBP1 mRNA expression and proteolytic processing. In this study, experimental location alone or its interaction with MVM significantly affected Srebf1 methylation at multiple CpG sites (Fig. 2a). For instance, mice in the DCM exhibited the highest methylation levels at CpG 3 and CpG 4 but the one of the lowest at CpG 5, indicating a CpG-specific epigenetic effect of experimental location-MVM interaction. The expression of the gene was negatively correlated with methylation at CpG 8 but positively correlated with methylation at CpG 3 (Fig. 3), suggesting a complex regulation of gene expression in a site-specific manner and challenging the simplicity of the traditional understanding of an inverse relationship between gene methylation and expression. Notably, both CpG 3 and 8 contain a common binding site for the transcription factor E2F, which plays a role in adipose function and plasticity [29]. Similarly, Acaca methylation was influenced by experimental location and the experimental location-MVM interaction at three CpG sites (Fig. 2b). In the DCM, but not in the other two locations, MVM supplementation increased overall Acaca methylation in obese mice, which accompanied by reduced obesity-related metabolic symptoms [11]. In contrast, Fasn methylation was affected by experimental location but not by MVM. Two CpG sites, along with the average methylation level of the targeted CpGs, were significantly altered by experimental location (Fig. 2c). Interestingly, higher Fasn methylation correlated with increased gene expression (Fig. 4), likely due to the positioning of the targeted CpGs downstream of the TSS, a region known to promote transcription elongation and enhance gene expression [30].

Pparg, a master regulator of adipogenesis, lipid metabolism, and an insulin sensitizer in adipose tissue [31], was affected by both experimental location and MVM in serval CpGs. Location the CCBR had the highest methylation level at CpG 1 but the one of the lowest at CpG 2. MVM reduced methylation at CpG sites 1–3 across all three experimental locations (Fig. 5). This reduction was accompanied by increased Pparg mRNA expression and improved insulin resistance in DIO mice in the DCM [11]. Similarly, Rbp4, which encodes an adipokine linked to insulin resistance [32], displayed methylation changes influenced by both experimental location and MVM, though no significant interaction was observed. Additionally, Rbp4 gene expression correlated with eWAT mass across all three experimental locations [20].

In the liver, the insulin signaling genes Pi3kr1 and Akt1 were affected by both experimental location and MVM at serval CpGs (Fig. 7). For Pi3kr1, the effect of MVM was dependent on experimental location, with adding MVM increased methylation at CpG 1 in the CCBR and at CpG 2 and CpG 4 in TCP but decreased CpG 4 methylation in DCM (Fig. 7a). Similarly, Akt1 methylation was influenced by both experimental location and its interaction with MVM (Fig. 7b). MVM increased methylation at CpG 2 in TCP and CpG 5 in both the DCM and the CCBR while reducing CpG 6 methylation in DCM, inversely associated with increased Akt1 expression in the DCM [11], suggesting that alterations in DNA methylation may contribute to PI3K/AKT signaling dysfunction in obesity and insulin resistance.

A potential mechanism underlying the effects of experimental location and MVM on gene methylation involves key 1-C metabolites, particularly SAM, SAH, and their ratio (SAM/SAH), which serve as indicators of methylation potential [23]. Both experimental location and MVM influenced these metabolites (Table 2), and multiple correlations were observed between DNA methylation and SAM, SAH, and SAM/SAH levels. Specifically, SAH correlated positively with CpG methylation in multiple genes across both tissues (Table 3), consistent with others showing that SAH alone is a sufficient marker for reflecting the DNA methylation in tissues [33]. Additionally, the SAM/SAH ratio was positively correlated with gene-specific methylation at some CpG sites in Srebf1 and Pikr1 and negatively correlated at others (Acaca, Fasn, Pparg, Rbp4, and Akt1). In eWAT, Pparg-specific methylation and global methylation were positively correlated with SAM levels, whereas Rbp4 methylation was negatively correlated. These findings suggest that experimental location and MVM influence gene methylation at least in part through their effects on key 1-C metabolites and may extend beyond the precursor and substrate availability.

The strength of the study arises from its conduct across three distinct animal facilities under the same leading investigator from the Anderson laboratory. Despite using standardized diet formulations across experimental locations, notable differences in animal housing conditions such as cage size, room occupancy, location of cages within the room and the rack, sex of experimenter/facility personnel, seasonal variations (online suppl. Table 3), and the duration of the dietary intervention, all can affect animal stress levels [20], potentially explaining the observed differences between experimental locations. Chronic stress can disrupt the delicate balance of the 1-C cycle leading to alterations in the availability of essential methyl donors like folate and B12 [34], thus impacting SAM levels and methylation processes. Indeed, the average effect size between experimental locations on the SAM/SAH ratio, an indicator of methylation potential, is 1.45, whereas the average effect size of MVM on SAM/SAH – 0.56. This reflects the statistical significance observed with experimental location (p < 0.001), indicating a stronger effect of experimental environment on the SAM/SAH ratio than the effect of MVM. While this does not identify the primary contributor to the results in each experimental location, it shows that subtle differences among environmental environment, even when conducted by the same principal investigator and team, alter the reproducibility of preclinical data.

Several limitations should be acknowledged. First, this study establishes associations between DNA methylation and 1-C metabolites across different experimental locations and MVM treatments, but causal relationships require validation through gain- and loss-of-function techniques. Second, only HFD-fed mice were included based on prior findings in DIO models [20]; future studies should assess whether similar effects of experimental location and dietary interaction apply to lean mice. Third, only male mice were examined, despite evidence that DNA methylation exhibits sexual dimorphism [35]. Investigating the effects of experimental environment and MVM in female mice will be important to discover sex-specific responses.

Although the primary environmental factors modifying study outcomes have not been identified, the results add plausibility to experimental environment is a significant modifier of the effects of micronutrient supplements in obese subjects. Thus, the results are of significance to the design of dietary interventions in obese adults with characteristics of metabolic syndrome. Blood measures of micronutrients, including vitamins A, B1, B6, and B12, as well as the trace minerals Se and Zn, have been reported to be suboptimal in these individuals [36–38]. However, the effects of micronutrient supplementation on metabolic health in obese adults are inconsistent [39]. Our findings provide a possible explanation for the variability in out-comes of the human trials conducted to date, suggesting the living conditions of selected participants need attention. In conclusion, the experimental environment is a significant determinant of the effects of micronutrient supplement on 1-C metabolism and the methylation of genes associated with IR in tissues of DIO adult male mice.

Acknowledgments

We would like to thank Sanaa Choufani and Youliang Lou from Rosanna Weksberg’s laboratory (The Hospital for Sick Children) for technical assistance with the methylation assay and Teodoro Bottiglieri and his laboratory (Baylor Scott & White Research Institute) for measuring 1-C metabolites.

Statement of Ethics

The study was conducted according to the guidelines of the Declaration of Helsinki and approved by the Ethics Committee of the University of Toronto (protocol #20012670 on 8 December 2020, and #20012824 on 2 August 2022) and by the Animal Care Committee of the Centre for Phenogenomics (protocol #25-0372H on 8 November 2021).

Conflict of Interest Statement

The authors declare that they have no conflict of interest.

Funding Sources

This research was funded by the Canadian Institute of Health Research, Institute of Nutrition, Metabolism and Diabetes (CIHR-INMD), Reference MOP-130286; Natural Sciences and Engineering Research Council of Canada (NSERC), Reference RGPIN-2016-06639.

Author Contributions

Conceptualization: Z.Y., R.K., and G.H.A.; formal analysis, writing – original draft, and investigation: Z.Y; funding acquisition, resources, and supervision: G.H.A.; methodology: Z.Y. and R.K.; writing – review and editing: Z.Y., R.K., E.K., C.E.C., and G.H.A. All authors have read and agreed to the published version of the manuscript.

Funding Statement

This research was funded by the Canadian Institute of Health Research, Institute of Nutrition, Metabolism and Diabetes (CIHR-INMD), Reference MOP-130286; Natural Sciences and Engineering Research Council of Canada (NSERC), Reference RGPIN-2016-06639.

Data Availability Statement

All data generated or analyzed during this study are included in this article and its supplementary material files. Further inquiries can be directed to the corresponding author.