We evaluated 30 individuals (age: 35.23 ± 8.26 years and BMI: 23.91 ± 3.4 kg/m²), from the general population of the Autonomous City of Buenos Aires (BA) and its metropolitan area, Argentina. The recruitment was performed in the Endocrinology Service of Hospital de Clínicas “José de San Martín” at Universidad de Buenos Aires, where clinic and anthropometric measurements were performed, blood samples were collected, and stool samples were received. Fecal samples were collected in household collection tubes delivered by the researchers responsible for the project. All participants signed written informed consent approved by the Ethics Committee of Hospital Posadas, Buenos Aires (according to the bioethical principles of the Declaration of Helsinki), approved by the Teaching, Research and Bioethics Committee of the Hospital Posadas (Ref CEPIC 119 EUPeS1/16).

Clinical and biochemical data were evaluated as well as anthropometric measurements of the participants. Demographic characteristics, presence of environmental factors (type of delivery, feeding, consumption of probiotics, alcohol, smoking, physical activity), medication used for treatment according to the case, and family history of chronic inflammatory pathologies, and the preexistence of emerging diseases such as Obesity, T2DM, Arterial Hypertension, Dyslipidemia among others, were established by anamnesis.

During the interview participants were instructed in the collection method for stool sampling by receiving a standardized written protocol for the collection of approximately 5 g of stool in a sterile wide-mouth tube in DMSO/EDTA/saturated sodium chloride buffer (DESS buffer) (Gray et al., 2013). DESS Buffer has been shown to be highly effective in long-term DNA preservation at room temperature (Beknazarova et al., 2017). Tubes had a small spoon attached to the lid that facilitates the collection of the samples, which were suggested to be collected within 24 h prior to the next interview. Stool samples received were kept at -80°C until use. For each sample, DNA extraction was performed from 200 mg of feces using two different commercial kits following the instructions of the manufacturers: QIAamp DNA Stool Mini Kit (QIAGEN^®) and Quick-DNA Soil (Zymo Research^®). Nucleic acid concentration and purity were determined by 1% agarose gel electrophoresis and spectrophotometry on a NanoDrop ND-1000 (NanoDrop Technologies, Wilmington, DE, United States).

To amplify 16S rRNA gene fragments, 30 ng of purified DNA was used and the hypervariable regions V3 and V4 of the bacterial 16S gene were amplified using Bakt_341F/Bakt_805R primers contained Illumina overhang adapter sequences. Sequencing was performed using a MiSeq sequencer (Illumina^®). Libraries were sequenced in paired-end mode, guaranteeing sequences of ∼550 bp long and ∼100.000 sequences per sample.

De-multiplexed reads were quality trimmed using Trimmomatic (V0.36) (Bolger et al., 2014). Sequences generated were analyzed using quantitative insights into microbial ecology (QIIME) version 1.9.1 software package (Caporaso et al., 2010) for the identification of operative taxonomic units (OTUs) as well as taxonomic allocation and statistical analysis. To this end, the sequences obtained were compared with those deposited in Greengenes 13_8 database (DeSantis et al., 2006). Chimeric sequences were filtered using VSEARCH (Rognes et al., 2016). Phylum, Class, Family, and Gender were assigned to each reading with an open_reference OTU picking process (with sortmerna_coverage of 80%) (Kopylova et al., 2012). Additionally, to compare this dataset with other studies that targeted different 16S regions, closed-reference strategy was used. In order to normalize the subsequent bioinformatic analysis raw sequences from different dataset were used. In particular, we compare our population with the following 16S rRNA gene sequence data set: Healthy Cohort study-Production Phase 2(PP2) of human microbiome project (HMP) SRP002860 (The Human Microbiome Project Consortium, 2012); Rosario-Argentina (Carbonetto et al., 2016); Hadza population from Tanzania (Schnorr et al., 2014); Bologna-Italia (Schnorr et al., 2014) and Santiago-Chile (Schnorr et al., 2014; Fujio-Vejar et al., 2017). The sequences obtained in this study were forwarded and made available in Qiita database (ID 11839) and SRA: PRJNA503303. To account for multiple comparisons, p-values were adjusted by Benjamini–Hochberg FDR correction, applied separately at each taxonomic rank for a given comparison (Benjamini and Hochberg, 1995). To study microbial communities in different sample groups we used Unifrac algorithm (Lozupone and Knight, 2015), specially designed to describe and compare microbial communities based on massive sequencing of phylogenetic marker genes. In order to compare this dataset with other studies that targeted different 16S regions closed_reference strategy was used. Differences on beta diversity were assessed using ADONIS. Ellipses drawings for weighted UniFrac and unweighted UniFrac were performed using cov.dellipse from metRology R package.

Putative functional genes were predicted using PICRUSt (Phylogenetic Investigation of Communities by Reconstruction of Unobserved States) (Langille et al., 2013), which applies 16S rRNA gene to predict the abundance of functional genes by matching sample OTUs with reference genomes. In order to compare the relative abundance of the different taxa between groups, we performed linear discriminant analysis (LDA) effect implemented in LEfSe (Segata et al., 2011).

Buenos Aires and its metropolitan area constitute a megalopolis, being the second most populated agglomeration in South America and the southern hemisphere with 12.806.866 of inhabitants. According to the last census (2010), the average age of the Autonomous City of Buenos Aires (BA) is 39.66 years. As reported by the “Third National Survey of Risk Factors for Non-transmissible Diseases” (NSRF) of the National Ministry of Health performed in 2013 (Tercera encuesta nacional de factores de riesgo para enfermedades no transmisibles, 2013), the average daily servings of fruits or vegetables consumed per person in the general population of BA was 2.1, being below the amount recommended by WHO (five daily servings of fruits and/or vegetables). The highest averages of daily fruit or vegetable consumption were recorded in the oldest age group (2.4), in the highest education level (2.1) and higher income level. In addition, the prevalence of overweight individuals was 32% (IC 95% 27.6 – 36.8), the prevalence of high cholesterol among those who were once controlled (population 18 years and older) was 27.9% (IC 95% 24.0 – 32.1), and 8.2% marked the prevalence of diabetes in the total adult population (IC 95% 6.4 – 10.5).

We evaluated 30 individuals (Female/male: 15/15; age: 35.23 ± 8.26 years and BMI: 23.91 ± 3.4 kg/m²), from the general population of BA (Table 1).

Considering that several methods could be used for stool collection and storage, we standardized sample collection and storage protocol prior processing. In our population based study, participants were instructed in a home collection method for stool sampling, proved to be as effective as freshly collected native stool (Schultze et al., 2014). This strategy comes from the preference of most participants who do not agree to collect the stool samples elsewhere. Therefore, each participant received a written protocol for easy stool collection, consisting in introducing approximately 5 g of stool in a sterile wide-mouth tube with DESS buffer (Gray et al., 2013). This solution preserves the samples both at room temperature and in sub-zero conditions, avoiding freezing and cell damage. Additionally, DESS buffer is a non-proprietary solution that has been used for DNA preservation (Gaither et al., 2010) as it guarantees long term storage at room temperature (up to 24 weeks), and has been used to preserve bacterial DNA from coral mucus swabs for up to 4.5 months (May, 2011). Additionally, given that it is known that freezing and thawing the samples could generate OTUs identification bias (Fouhy et al., 2015), this buffer represents an efficient alternative strategy for adequate sampling and preservation until processing.

DNA isolation protocol and the targeted 16S rRNA gene region sequence chosen for analysis have been shown to be critical for gut microbiota assessment (Voigt et al., 2015), limiting the ability to compare different human microbiome datasets. In other to evaluate possible biases in the 16S profiles related to DNA extraction method, we compare QIAamp DNA Stool Mini (QG) and Quick-DNA Soil (ZR) commercial kits. DNA extraction was performed from 200 mg of feces using both QG and ZR for each sample. Fecal DNA extraction using QC and ZR yielded an average of 49.59 ± 21.41 and 24.85 ± 20.75 μg/mL DNA, respectively. The hypervariable regions V3-V4 of bacterial 16S gene was sequenced using MiSeq-Illumina system, and an average of 47526 ± 4718 sequences per sample were obtained. Rarefaction plots reach an asymptotic state, indicating that the sequence depth was sufficient to represent the bacterial community richness and diversity by both methods (data not shown). In this sense, when we compare alpha diversity (Chao1 index) based on a two-sample t-test, there are not significant differences between QG and ZR kits (p = 0.069). We also found no significant differences when comparing beta diversity between different purification kits, p = 0.97 for weighted UniFrac and unweighted UniFrac p = 0.294 (ADONIS).

Comparing the principal phyla detected using both DNA extraction Kits, we found that the choice of DNA extraction method has an impact on the observed community structure. Using the QG kit, the dominant phyla were Bacteroidetes (47.7 ± 8.9%), Firmicutes (37.2 ± 8.45%), Proteobacteria (8.5 ± 6.7%), Verrucomicrobia (2.5 ± 2.8%) and Actinobacteria (1.3 ± 2.2%), while the principal phyla found using ZR kit were Firmicutes (45.5 ± 8.8%), and Bacteroidetes (40.0 ± 7.8%), followed by Proteobacteria (6.1 ± 5.4%), Verrucomicrobia (2.5 ± 3.0%), and Actinobacteria (3.1 ± 2.2%). In this sense, QG protocol, without the bead-beating step, resulted in a significantly low abundance in Gram-positive phyla, Firmicutes (p = 0.015); Actinobacteria (p = 9.29 × 10^-5) and overestimated Bacteroidetes phyla (p = 0.023) FDR adjusted p-values. These results, in accordance with previous studies (Guo and Zhang, 2013; Wesolowska-Andersen et al., 2014), show that DNA extraction method has an impact on the observed community structure. In particular, the first step of DNA extraction bead-beating disruption and/or lysis of the bacterial membranes can be expected to be biased for specific bacterial taxa due to differences in cell wall structure and integrity.

Since several studies have proposed the Firmicutes to Bacteroidetes ratio (F/B:ratio) as a gauge of short-chain fatty acids (SCFA) production and the overall gut microbiota balance any bias in these parameters must be taken into account (Rodríguez-Carrio et al., 2017; Garcia-Mantrana et al., 2018). In particular, QG protocol showed an increase in the abundance of Bacteroidetes phylum, main acetate and propionate producers, and a decrease in Firmicutes butyrate-producers (Macfarlane and Macfarlane, 2003). It is important to remark that these changes impact notably the F/B: ratio, being 0.77 for QG kit vs. 1.13 for ZR one. However, it is not possible to determine which DNA extraction method best represents the overall microbial community.

Our data is consistent with previous observations showing that the regular intake of yogurt and other fermented products promotes a healthy gut microbiota (Lisko et al., 2017; Marco et al., 2017; Suzuki et al., 2017). In this sense we found significant differences in alpha diversity Chao1 index (p = 0.012) when comparing microbial diversity between subjects that habitually consume probiotics with those who do not (considering both extraction kits).

To further analyze our results, we compared the microbiota of BA with other westernized populations (Santiago de Chile, Rosario-Argentina, United States-Human-microbiome-project, Bologna-Italy) and the Hadza population of hunter-gatherers (Table 2). To understand differences in the composition of microbial communities’ diversity among environments was compared. Beta diversity (considering weighted and unweighted UniFrac distances) was analyzed in order to compare the differences between microbial compositions among different populations (Figure 1). Differences on beta diversity values between habitats, countries, cities, 16S region amplified, purification kit employed, and platform were evaluated (ADONIS). Although all the comparisons were statistically significant (p < 0.001), the R²-value indicates that approximately 43% (weighted) and 23% (unweighted) of the variation effect is produced by the city of origin, being the most important variable studied. The unweighted UniFrac (a qualitative measure that estimates the distance between two communities) clustered the westernized populations together, leaving the hunter-gatherer population from Hadza out. In particular, Santiago de Chile’s population turns out to be the closest to BA’s.

It is relevant to highlight that the microbiota of BA subjects stands out for the presence of Verrucomicrobia of the genus Akkermansia (1.2%). The mucus-degrading bacterium Akkermansia muciniphila is the only identified member of this genus and was also found in the Chilean population (Fujio-Vejar et al., 2017). This microorganism has been proposed by others as a hallmark of a healthy gut due to its anti-inflammatory and immunostimulant properties and its ability to improve gut barrier function, insulin sensitivity and endotoxemia (Collado et al., 2007; Bland, 2016). However, some studies report the induction of pro-inflammatory responses triggered by Akkermansia contributing to an increase of inflammation during infection in Multiple Sclerosis patients (Cekanaviciute et al., 2017). As proposed by Fujio-Vejar et al., these results could serve as a baseline for the characterization of dysbiosis associated with various diseases (Fujio-Vejar et al., 2017; Figure 2).

Putative functional genes were predicted using PICRUSt which applies 16S rRNA gene to predict the abundance of functional KEEG pathways. The LDA effective size (LEfSe) was performed to identify KEEG pathways with statistically differential abundance between hunter-gatherers and westernized population. The results show that metabolism related pathways were more abundant in westernized populations than in hunter-gatherers. In particular, carbohydrate metabolism (amino sugar and nucleotide sugar metabolism), amino acid metabolism (alanine, aspartate and glutamate metabolism), lipid metabolism, biosynthesis of secondary metabolites, and sulfur metabolism showed a LDA score greater than 2.0 (Figure 3).