The present study was approved by the Ethics Committee for Research with Humans of the School of Physical Education and Sport of Ribeirão Preto, University of São Paulo (EEFERP-USP). The registration number is CAAE: 79582817.0.0000.5656. The study was also registered in the Brazilian Registry of Clinical Trials under RBR-3g38dx. Participants were recruited in the Physical Education Program for the Elderly at EEFERP – USP. The inclusion criteria were women between 50 and 70 years with a medical clearance certificate to perform physical exercise and who were sedentary at the baseline. The exclusion criteria included diagnostic limitations to physical training or physical test, current smokers, caffeine consumption in the last 24 hours before any blood draw, and more than 25% combined training absence.

Forty-eight post-menopause women participated in the study. They were divided into two groups according to the fasting blood glucose levels determined by the American Diabetes Society (2020): Pre-diabetes Group (PG), composed of 24 women with a fasting blood glucose level higher than 100mg/dL and the Normal Glucose Group (NGG), composed of 24 women with a fasting blood glucose level lower than 100mg/dL (9).

The anthropometric evaluation was assessed before and after 14 weeks of combined training. The height was measured using a vertical stadiometer - Balmak - EST-223 and weight using the G-Tech digital scale - Balgl200, G-Tech. Subsequently, the body mass index was calculated using the formula (weight/(height*height)) (17). Waist and hip circumferences were measured following the recommendations of Malachias et al. (17). Participants’ systolic and diastolic blood pressure was measured using an Automatic Digital Upper Arm Blood Pressure Monitor Meter (SBH, 2010 - OMRON^®, model HEM-7113) (17).

The biochemical analyses included uric acid, total cholesterol, high-density lipoproteins (HDL-C), low-density lipoproteins cholesterol (LDL-C), triglycerides (TG), blood glucose, and insulin. We also calculated the HORMAR IR (Blood (mmol) x Insulin (Ul/mL) ÷ 22. 5).

To investigate the physical benefits of 14 weeks of combined training, the elbow flexion and extension, sitting and standing up from a chair, agility dynamics, and the six-minute walk test was adopted. Also, the 6MWT (m) methodology was used to measure aerobic capacity (18).

We also determined the total serum concentration of metals (Hg202, Al27, As75, Cd111, Pb208, Ni60, Li7) through inductively coupled plasma mass spectrometry (ICP-MS), following the protocol described by Batista et al. (19).

Combined training was performed for 14 weeks, three times a week, lasting 60 minutes. The training was divided into strength exercises (30 minutes) and aerobic training (treadmill or bicycle, with a total duration of 30 minutes) (14).

The training periodization followed the first two weeks. The participants performed two sets of 15 to 17 maximum repetitions and 50% of the aerobic training in the reserve heart rate. The participants started to do 10 to 12 maximum repetitions in the remaining weeks, and aerobic training became 70% of the reserve heart rate. The participants-maintained values ​​between three and six, a score presented on Borg’s subjective effort scale. The score we adopted on a scale from 0 to 10 represents moderate intensity. For the selection of exercises for strength training, the need to train the main muscles was considered, totaling five exercises for the muscles of the upper limbs and three exercises for the muscles of the lower limbs. Our group previously detailed the protocol (14).

The DNA methylation was analyzed at the baseline and after the 14 weeks of training. Salting-out technique (20) was used to extract DNA from leukocytes. Biodrop spectrophotometry was used to assess DNA quality and concentration using. 1% agarose gel test was used to guarantee genomic DNA integrity. The run time was 90 minutes at 80W (21).

Following the manufacturer’s recommendations, the bisulfite method was applied using the DNA EZ methylation kit (Zymo Research, CA, USA). Illumina BeadChip EPIC Infinium Methylation BeadChip 850k was used for the methylation experiment. All recommendations for the Infinium^® HD Assay Methylation Protocol Guide were followed (22).

The raw data files were exported from iScan. A Samplesheet file was created to associate the collected information with raw data (23). The RStudio software (version 4.0.2) was used to process the methylation data. The data was organized as the participants’ codes, age, anthropometric, physical, biochemical, and metal information. The Sentrix Position column informs which row and column each sample was located in. The Sentrix_ID column provides the corresponding code of the BeadChip (23). The Bioconductor ChAMP data package was used for data normalization through the MyNorm function (24).

Differentially methylated regions (DMRs) were analyzed using the champ.DMR() function to compare PG vs. NGG groups. The bumphunter method was used to aggregate all probes into small groups. The random permutation method was applied to estimate DMRs with p < 0.05. Based on text mining and literature review, all DMRs were functionally annotated (24).

After identifying the DMRs inter and intra-group, the singular value decomposition (SVD) analysis method was used to evaluate the number and nature of significant components, correlating with various components of interest (24). Methylation data were normalized via minfi’s ssNoob (25). DNAmAge was calculated using the Methylclock package through the DNAmAge function (26). We also investigated the epigenetic clock of participants from both groups for intrinsic epigenetic age acceleration (IEAA) and extrinsic epigenetic age acceleration (EEAA). The epigenetic clock is a biomarker for age-related epigenetic changes as well as disease-specific changes (27). This clock has been recently developed by using specific DNA methylation sites to describe biological age at the level of the cell, tissue, and organ (28). The multi-tissue algorithm by Horvath generates age prediction, which highly correlates with chronological age above r=0.9 (29). In this study, DNAm Age was calculated using Horvath’s epigenetic clock algorithm using 353 CpGs (3).

In SVD analysis, the champ algorithm first does SVD deconvolution on the beta dataset. Then, the Random Theory Matrix method in the “isva” package is used to calculate the numbers of the latent variable and the top components of the SVD result. This analysis is comparable to a Principal Components Analysis (24)

Two researchers verified the resulting data to maintain the values’ reliability and accuracy. The Statistica 7.0 software (StatSoft Statistica^®) was used to perform statistical comparisons. The Shapiro-Wilk test tested the normality. The Levene test assessed the equality of variance between groups. Then, repeated-measures ANOVA was performed, considering p <0.05. Fisher’s post hoc compared all means and controlled the error rate at the significance level. Then, a Spearman correlation was performed for the non-parametric data and a Pearson correlation for the parametric data to identify possible correlations. The Student’s t-test compared means in the descriptive analysis.

The mean age of participants in the PG was 61.3 ± 5.2 years, while in the NGG was 59.9 ± 4.8 years ( Table 1 ). The 14 weeks of combined training had an effect of time in systolic blood pressure (F=24.304; p<0.001), diastolic blood pressure (F=6.180; p=0.016), total cholesterol (F=7.191); p=0.010), HDL-C (F=8.168; p=0.006), TG (F=11.176; p=0.001), and glucose (F=25.281; p<0.001). No undesirable effects were noted, but we did not observe group-time effect or group-time interaction (Table 1).

We observed the effect of time on elbow flexion and extension (F=17.028; p<0.001) for upper limb strength; sitting and standing up from a chair (F=20.264; p<0.001) for lower limb; agility test (F=51.248; p<0.001); and six-minute walk test (F=23.872; p<0.001) for aerobic fitness. The effect of time in both groups demonstrated an improvement in physical skills after 14 weeks of combined training. No group-time effect or group-time interaction was observed (Table 2).

Analyzing metal levels and combined training together, we observed the effect of time decreasing mercury (Hg) (F=5.028; p=0.030) and lead (Pb) (F=12.100; p=0.001) in both groups. No effect or group-time interaction was observed ( Table 3 ). There was a positive correlation between the arterial blood pressure and Pb²⁰⁸ in the PG_pre (PAS: r=0.389, p=0.003; PAD: r=0.337, p=0.010). After the intervention, the correlation was not significant. The epigenetic clock was not different between groups and did not change after the 14-week physical intervention (Figure 1).

Before the physical exercise training, we found 12 DMRs between the PG_pre and the NGG_pre (Table 4). After the 14 weeks of exercise, five regions remained Differentially Methylated between the two groups (Table 5).

After the 14-week of exercise, we found 118 DMRs in the PG_post compared to the PG_pre (Table 6), and we performed a Singular Value Decomposition Analysis (SVD) to verify if the covariables of interest could explain the DNA methylation patterns (Figure 2).

We also performed a Linear Regression of significant traits (Pb and glucose) and the DNA methylation targets of the PG group. We found three DMRs for Pb and 25 DMRs for glucose (Figure 3). BRCA1 was hypomethylated after a 14-week physical exercise intervention and was associated with glucose in the PG group. There was also an interesting overlap between the DMRs related to Pb and glucose enriched in the gene SLFN12. Notably, this gene region was differently methylated in the PG_pre group than NGG_pre and associated with Pb in the PG. After the intervention, a decrease in Pb was observed, and the methylation patterns of SLFN12 in PG_post became similar to NGG_post.,,

One of the aims of this study was to investigate the effect of combined training on epigenetic patterns in the prediabetes (PG) and normal glucose group (NGG). Combined training is an alternative program involving aerobic and strength exercises in the same training session (30). Its effects are widely recognized in the literature for the numerous benefits to the cardiorespiratory system, prevention of oxidative stress, and critical metabolic alterations in the glycemic balance (31). According to Bellazi et al. (12), the regulating factor of blood glucose concentration is achieved due to the power of action of the three variables: insulin, adequate meals, and physical exercise (32). Indeed, after 14 weeks of training, we found a decrease in glycemic values in the PG group, and in both groups, we observed decreased levels of total cholesterol and TG levels, indicating an improvement in metabolic indicators of cardiovascular health.

Moreover, the combined training improved both groups’ systolic and diastolic blood pressure. A study by Schroeder et al. (13) observed that the combined training protocol was the only one that promoted significant changes in blood pressure values, with positive reductions in diastolic pressure (33). They demonstrated that the combined training had positive cumulative results in all cardiovascular outcomes (33).

Combined training also demonstrated benefits in functional capacity in our study, providing a time effect for the PG and NGG groups. These findings corroborated with Medeiros et al. (14), demonstrating an increase in maximal and functional strength after 12 weeks of training in 58 women aged 50-75 years (34).

Regarding heavy metal concentration, we observed a decrease in mercury and lead levels after 14 weeks of combined training. Another finding of our study was the positive correlation between lead and systolic and diastolic blood pressure values ​​in the PG_pre group. Lead is widespread in the environment, and even low blood levels of this toxic metal may raise the risk of cardiovascular diseases, stroke, and chronic kidney disease. It is estimated that in adult populations lead exposures account for 5% of the population attributable risk for high blood pressure. Further, single toxicant exposures provide some evidence for developmental nephrotoxicity demonstrated by altered estimated glomerular filtration rate (eGFR) and other kidney biomarkers, increasing a risk for nephropathy in individuals with T2D (35). In a cross-sectional study carried out by Asgary et al. (15), the authors demonstrated that lead, mercury, and cadmium levels were higher in patients with coronary artery disease (36).

Our findings demonstrated that genes in differentially methylated regions of the PG group were related to inflammation and insulin signaling. Physical activity has improved glycemic control in adults with type 2 diabetes (27). Moreover, it can induce metabolic adaptations and changes in DNA methylation due to energy requirements (14). These changes in DNA methylation allowed for increased insulin sensitivity, expression of genes involved in energy metabolism, myogenesis, contractile properties, and oxidative stress (14). In the Figure 4 is shown some metabolic pathways functionally enriched from the obtained DMRs.

Figure 5 shows hypermethylated pathways that are potentially altered after 14-week physical exercise intervention. Here we described specific hypermethylated genes’ role in PG observed in our study. HLA-DPB1 plays a central role in the immune system by presenting peptides derived from extracellular proteins. Also, it is linked with inflammation and systemic insulin resistance (37). RWDD3 gene also plays an immune function. The isoforms RWDD3 positively regulate the NF-kappa-β signaling pathway by enhancing the NF-kappa-β inhibitor, promoting its stabilization, and regulating the hypoxia-inducible factor-1 alpha (38). Other genes are found hypermethylated in PG group-related metabolism and genomic stability. The gene TYW3/CRYZ is related to levels of resistin. Previous studies reported associations between elevated circulating resistin levels and increased risk of cardiovascular disease (28). PM20D1 is an enzyme that condenses fatty acids and amino acids. The PM20D1 locus shows genome-wide association studies (GWAS) with body mass index (BMI) and weaker associations with obesity-related conditions like type 2 diabetes and HDL-C levels (39). Aurora C kinases (AURKC) are essential kinases for cell division via regulating mitosis, especially the process of chromosomal segregation, associated with malignant cell transformation and genomic instability (16). PDE3A binding to SLFN12 results in cell killing and optimizing cancer therapeutics (40). WFIKKN2 is a follistatin domain-containing protein that binds GDF8/GDF11 proteins with high affinity, both of which have been implicated in DM development (41). The hypermethylation of these genes in individuals of the PG group could indicate a compromised metabolic function which could be associated with a higher risk of T2DM (42).

After the 14-week physical exercise intervention, 118 DMRs were detected. Three of these genes were hypermethylated: CALD1, RNF121, and MSI2. The CALD1 was previously related to diabetes nephropathy (43), while the role of RNF121 in the VEGF‐induced endothelial cell proliferation inhibition has been associated with diabetic retinopathy and age-related macular degeneration (44). The MSI2 hypomethylation has already been described as an epigenetic biomarker for hyperglycemic individuals by the Korean Genome Epidemiology Study (KoGES) (45).

The additional 115 genes were hypomethylated and involved in many biological functions (Figure 6). Regarding the effects on general metabolism, the FDPS (Farnesyl pyrophosphate synthase) is a crucial enzyme for several classes of essential metabolites, including sterols, dolichols, carotenoids, and ubiquinones. The BHLHE40 Class E basic helix-loop-helix protein 40 is also a target that may influence the general metabolism due to its role as a transcriptional repressor in regulating the circadian rhythm and negatively regulating the activity of the clock genes and clock-controlled genes (46).

The mechanisms related to metabolism also included targets related to macronutrients, such as lipids and carbohydrates, which are important in the etiology of hyperglycemia (47). The NR1H2 (Oxysterols receptor LXR-beta; Nuclear receptor) gene was differentially methylated and can regulate cholesterol uptake through the ubiquitination of LDLR, VLDLR, and LRP8. Indeed, the PLD3 - Phospholipase D3 is an important key in lipids metabolism and may be involved in the Amyloid Precursor Protein (APP), already been related to Alzheimer´s Disease (AD). Three genes were differentially methylated in carbohydrate metabolism: IDI1, AGL, and SIDT1. ID1 has already been related to nephropathy, a condition connected to advanced stages of T2D (48). The AGL gene encodes the glycogen debrancher enzyme involved in glycogen degradation (49). The SIDT1 gene is associated with insulin secretion, and its expression was decreased in individuals with hyperglycemia (50).

Four differentially methylated genes were associated with energetic metabolism: ATP6V1A, ATP5A1, NDUFA4, and SLC25A19. The ATP6V1A is responsible for acidifying a variety of intracellular molecules, leading to HIF1A hydroxylation and catalyzing AMP formation (51). The ATP5A1 produces ATP from ADP (52). NDUFA4 is a cytochrome c oxidase subunit NDUFA4 and catalyzes oxygen reduction in water (53). The SLC25A19 mitochondrial thiamine pyrophosphate carrier is a transporter mediating thiamine pyrophosphate (ThPP) entrance into mitochondria (54).

The most interesting genes involved with the immune function were the COPS4 signalosome complex subunit 4, involved in phosphorylation of p53/TP53, c-jun/JUN, IkappaBalpha/NFKBIA, which leads to modulation of the inflammasome (55); theWDR62 and DMP-1 related to oxidative stress; and the SLFN11, interferon (IFN)-induced target (16).

The tumor suppression mechanisms, insulin resistance, and metabolic diseases have a common path (56). The CMTM6 depletion significantly decreases tumor-specific T cell activity (57). Indeed, the hypermethylation of two CpG sites upstream of the CASP8AP2 promoter influenced gene expression and treatment outcomes in lymphoblastic leukemia (58). In the present study, CASP8AP2 was hypomethylated after the intervention, suggesting that physical exercise training may decrease the risk of developing cancer in this population. The VWA5A plays a role in tumorigenesis as a tumor suppressor (59). BRCA1 is another important tumor suppressor, and studies have shown that individuals with hyperglycemia had this gene naturally more methylated than normoglycemic individuals (60).

Physical exercise can promote many alterations at a cellular level. Liu et al., 2020, demonstrated that miRNAs could epigenetically modulate the myogenesis and adipogenesis processes after a 12-week physical exercise intervention in older individuals (61). In the present study, important differentially methylated regions in the PG were related to myogenesis and adipogenesis after the intervention. The DMRT2 directly regulates early activation of the myogenic determination. It is required to initiate or maintain the proper organization of the sclerotome, dermomyotome, and myotome (by similarity). Mutations in the MYOF gene can cause muscle weakness, which is involved in membrane regeneration and repair. Reduced KLF14 expression increases pre-adipocyte proliferation and decreases lipogenesis. In mice, the deletion of KLF14 partially reproduces the human phenotype of insulin resistance, dyslipidemia, and T2D (62). The deregulation of the KLF14/PLK1 cascade plays a crucial role in thrombin-induced endothelial dysfunction in T2DM patients (63).

There is substantial epidemiological evidence that exposure to low levels of various environmental chemicals can influence the development of chronic metabolic diseases, including diabetes (64). It has been well demonstrated that lead shows potential genotoxicity to various research subjects under different conditions (65, 66). A study with human lymphoblastoid TK6 cells found that lead exposure results in DNA damage via promoting oxidative stress and the promoter methylation of DNA repair genes, such as BRCA1 (67).

The BRCA1 plays a critical role in DNA repair. Mutations in the gene that lead to loss of protein function can accumulate damaged DNA within the cell tumor-suppressor gene, which is strongly linked to an increased risk of breast cancer development (68). Recent studies demonstrated that BRCA1 is expressed in skeletal muscle and is a critical regulator of metabolic function in cultured human myotubes (69). Also, deletion of BRCA1 expression in vascular cells results in the reduced catabolic breakdown of free fatty acids, and it can influence the metabolic phenotype of cultured cells and potentially the cells’ ability to respond to nutrient overload conditions (70).

Schlafens (SLFN) are intriguing proteins that play different roles in regulating cell proliferation, cell differentiation, immune cell growth and maturation, and inhibiting viral replication (17, 71).. However, it is still poorly understood how SLFNs interfere with the cell cycle machinery. It is a particularly tempting target for further research, as this gene is known to be down-regulated during T cell activation and up-regulated by type I interferons (IFNs) (18). It has also been suggested in cancer therapy since its expression was positively correlated with the growth inhibitory action of topoisomerase inhibitors on human cancer cells (19).

Acute exercise seems to be associated with active demethylation pathways. The mechanism proposed for active demethylation involves ten-eleven translocation (TET) proteins, suggesting that metabolites produced during exercise could alter the activity of TET proteins (20).

This study’s strengths were demonstrating the modulation of biochemical and physical parameters after a physical exercise intervention between different groups. In addition to verifying differently methylated regions in the PG group compared to the NGG group using a large-scale DNA methylation array (Infinium-EPIC beadchips) and the effect of combined training. The limitations were that analyzes of RNA expressions of the differentially methylated genes were not performed.