This was a case–control pilot study. The Ethics Committee of Ribeirão Preto Medical School approved the study protocol at the University of São Paulo, Brazil (process CAAE: 57459816.1.0000.5440). Informed consent was obtained from all individual participants, and the inclusion criteria are shown in Figure 1. The study was conducted according to the Declaration of Helsinki. Six women, aged 30.8 ± 2.2, were subdivided into two groups (Figure 1).

Adult women without declared diseases were included in the study. Non-inclusion criteria included the use of antibacterial therapy, anti-inflammatory medicine, antifungal and anti-allergic medicines, probiotics and prebiotics, lactation, smoking, drinking, chemotherapy, a history of gastrointestinal surgery, gastrointestinal disease, diabetes mellitus, hypothyroidism or hyperthyroidism, the practice of physical exercise and dietary interventions already in course, and the use of new medical therapy that could influence the intestinal microbiota. The included patients were healthy and had no severe diseases diagnosed during the study.

The groups were divided based on the recommended reference values for vitamin C ingestion, resulting in three participants in each group. Data collection commenced on 19 November 2013, and concluded on 29 June 2015. During this period, each woman underwent the 24-h dietary recall three times. It is important to note that the 24-h recall method was used to assess the participants’ vitamin C consumption. This approach allowed us to evaluate their habitual vitamin C intake within the context of their daily dietary patterns. Blood collection and anthropometric measurements are only taken. Anthropometry, body composition, and resting metabolic rate were assessed following the protocol previously described by our group (5). Three 24-h dietary recalls were performed to evaluate the participants’ food consumption, and the software Dietwin® was used to calculate it. The subjects were evaluated after 12 h of fasting. Total blood was collected to assess glycemia, lipidogram, and gene expression analysis. Stool samples were collected to analyze the microbiota relative abundance.

DNA was extracted using the kit QIAamp® Fast DNA Stool Mini. The primer sequences used for each phylum were Firmicutes (F-ATGTGGTTTAATTCGAAGCA, R-AGCTGACGACAACCATGCAC), Bacteroidetes (F-GTTTAATTCGATGATACGCGAG, R-TTAASCCGACACCTCACGG), Actinobacteria (F-TGTAGCGGTGGAATGCGC, R-AATTAAGCCACATGCTCCGCT), Proteobacteria (F-CATGACGTTACCCGCAGAAGAAG, R-CTCTACGAGACTCAAGCTTGC), Verrucomicrobia (F-TCAKGTCAGTATGGCCCTTAT, R-CAGTTTTYAGGATTTCCTCCGCC), and rRNA 16S universal (F-AACAGGATTAGATACCCTGGTAG, R-GGTTCTTCGCGTTGCATC). The primers were at 10 μM. We used 20 ng of DNA in each well, ultra-pure water, and a master mix (SYBR Green, Promega®) with a final volume of 10 μL. After the reactions were performed, the results were analyzed using the Step One Plus® software (Thermo Fisher Scientific). We set the limit manually at 0.1. We use the 2^-ΔCt method to calculate the fold change and determine the relative abundance (6).

DNA was extracted from whole peripheral blood using the Promega (Madison, Wisconsin, EUA) automated system with magnetic beads following the manufacturer’s instructions (7). Integrity was evaluated by agarose gel. Quality was evaluated using NanoDrop (Thermo Fisher, Waltham, Massachusetts, EUA). DNA from total leucocytes were firstly bisulfite converted using the EZ DNA Methylation Kit (Zymo Research Corporation, Irvine, CA, United States), then 500 ng were hybridized in the Infinium Human Methylation 450 k (IHM450K) BeadChip (Illumina, Inc., San Diego, CA, United States) (8). After the protocol, images were obtained using the iScan system.

The analysis was performed using R version 3.6.2 and ChAMP, a user-friendly R package available on Bioconductor. Raw intensity data from the IHM450K idats files were loaded on the R package, ChAMP (9). After the normalizations of the myNorm function, 409,819 probes continued in the pipeline for further processing, the missing data were removed, and the KNN imputation was performed (10). White blood cell type estimation was performed with the Houseman method (11). The intensities of the database were obtained from purified cells, and the IHM450K contains 473 CpG sites as a reference for the cell types. The obtained proportions were used to correct beta values using the Refbase function from the Champ package. Single value decomposition (SVD) was performed to check the principal components of the data variation. Limma package was used to determine differentially methylated positions (DMPs) (12), and a p-value of <0.05 and an FDR of <0.05 were considered significant. Linear regressions were performed to verify the association with clinical parameters (12). We performed Gene Functional Enrichment using the STRING database, we used as input the gene symbol from the mapped differentially methylated sites (13).

The collected blood sample was used for RNA extraction to evaluate the gene expression of the target genes: TNF-α, NFKB, IL-6, TLR4, and PI3kR1. The phenol–chloroform method was used for RNA extraction. Reverse transcription (RT) was performed using the high-capacity kit (Applied Biosystems) to obtain cDNA. Quantitative PCR (qPCR) was performed using the TaqMan (Thermo Fisher Scientific) inventoried assays for the target genes. The results were evaluated by the 2-ddct method, using ACTB and GAPDH as endogenous genes and women without obesity as the normalizing group.

Shapiro–Wilk tests were used to verify normality, t-tests were used to verify differences between groups, and Spearman’s correlation was performed between vitamin C and phyla. For all analyses, we adopted a p-value of <0.05. Analysis was performed using the Statistica 7.0 software (StatSoft Statistica®).

Six women (30.8 ± 2.2 years old) were included in this sample, and the group was divided by the recommendation for vitamin C ingestion (14). There were three women with high vitamin C (HVC) (30.7 ± 0.9 years old) and three with low vitamin C (LVC) (31.0 ± 2.9 years old) reference values. Table 1 shows all the collected clinical data. The two groups were similar for all evaluated parameters.

The HVC group had a higher relative abundance of Actinobacteria (Table 1). A strong positive correlation existed between the relative abundance of vitamin C and Actinobacteria (r = 0.84, p = 0.018). Table 1 shows the individual percentual relative abundance of the analyzed phyla. As expected, Bacteroides and Firmicutes were the most abundant groups in all women’s intestinal microbiota.

Regarding the quality control steps, the SVD analysis revealed that vitamin C was the most important variable in the metafile and better explained the data variation (p < 0.01). It was significant in the first principal component (PC1). Regarding bacteria relative abundance, only Actinobacteria were significant (p < 0.05) and influenced data variability. Figure 2A shows the clusterization of samples before normalization. The occurrence of the two clusters confirmed that the two groups had significant DNA methylation differences. The distance between samples was calculated based on the identified principal components previously obtained by SVD analysis. The generated outputs are part of a series of quality control analyses in the Champ pipeline. The analysis of QC parameters revealed that all samples were suitable for downstream analysis.

The association analysis retrieved that the associated sites were located in 239 genes for vitamin C and 5,818 genes for Actinobacteria; from those, there were 207 common genes (Figure 2B, upper Venn diagram). There were 40 genes that did not show any DMPs and are not shown in the subsequent Venn diagram (Figure 2B, lower Venn diagram). These 207 genes were classified into three categories according to their methylation status (Figure 2B, lower Venn diagram).

Based on the functional enrichment findings, we performed gene expression validation on some targets of interest (Figure 2C). We observed that the HVC group had, as expected, lower expression of the classical inflammation targets: TNF-ɑ, TLR4, NFKB, and IL-6 (GO:0070944, GO:0050829, GO:0002443) and higher expression of the PI3K target (enriched in the pathways GO:0043552 and GO:0014068).

The common genes (n = 167) that had differentially methylated CpGs were functionally enriched. The analysis was performed separately for the genes that had hypomethylated and hypermethylated sites (Tables 2, 3). The genes that had hypomethylated CpGs had targets involved in the respective processes: immune function, glucose homeostasis, and general cellular metabolism. The hypermethylated sites were also enriched in immune function-related processes, and interestingly, more immune responses against pathogens were detected.

Regarding cell fraction estimation, the group with HVC consumption had 30% more granulocytes (HVC: 0.68 ± 0.06; LVC:0.48 ± 0.02; p = 0.0100) (Figure 3). Furthermore, vitamin C and granulocytes had a strong positive correlation (r = 0.97, p = 0.001). Vitamin C and the relative abundance of Actinobacteria were also positively correlated (r = 0.84, p = 0.036).