Fecal microbiome analysis in patients with metabolic syndrome and type 2 diabetes

Recruitment and volunteers

The volunteers for this study were part of the IBEROBDIA project (“Obesity and Diabetes in Iberoamerica: Risk Factors and New Pathogenic and Predictive Biomarkers”), funded by the Iberoamerican Program of Science and Technology (CyTED) (918PTE0540) and the Spanish State Research Agency (PCI2018-093284). This project was approved by the ethics committee of the Galician Health System (SERGAS, Xunta de Galicia), under code 2018/270. All participants were fully informed about the study and provided written informed consent. The volunteers’ data were processed in accordance with Organic Law 3/2018 of December 5 on the protection of personal data and the guarantee of digital rights. Participants were recruited through various media, including press, radio, and posters, in different locations across the autonomous community of Galicia, Spain.

The inclusion criteria for the study were Spanish adults aged 40 to 70 years, BMI of 18.5–24.9 kg/m² (normal weight) or $\ge$27 kg/m², and Finnish Diabetes Risk Score (FINDRISC) of $\ge$12, indicating a moderate risk of developing T2D. The exclusion criteria were consumption of prebiotic or probiotic supplements in the last 2 months, a prior diagnosis of T2D, the presence of other chronic diseases, pregnancy, antibiotic treatment within 2 months prior to the study, chronic medication use (including treatments for hypertension, high cholesterol, contraceptives, and proton pump inhibitors), drug consumption, and a low risk of developing diabetes according to the FINDRISC. Because chronic medication intake is known to influence the gut microbiota composition, it was a key exclusion criterion. Likewise, volunteers could not have a prior diagnosis of either T2D or MS.

Study design

This was a cross-sectional observational study conducted as part of the IBEROBDIA project, as described in the previous section. The study design is illustrated in the flowchart shown in Fig. 1. Following the provision of written informed consent, the volunteers were categorized based on BMI, sex, age, and their risk of developing T2D. Those who met the inclusion criteria were then invited to have their glucose levels checked and undergo an oral glucose tolerance test. On the day of the test, the volunteers were required to bring a fecal sample that had been collected at home for analysis.

Based on the collected data, the participants were classified into three groups: individuals with normal weight and normal glucose values, individuals with a BMI of $\ge$27 kg/m² and normal glucose values, and individuals with a BMI of $\ge$27 kg/m² and altered glucose values. A second classification was then introduced based on BMI and the presence of MS, following the criteria established by the International Diabetes Federation in 2016 (Alberti, Zimmet & Shaw, 2016). Using these criteria, the groups were reorganized as follows: normal weight with MS, BMI of $\ge$27 kg/m² with MS, and BMI of $\ge$27 kg/m² without MS. Although individuals with a BMI of $\ge$27 kg/m² are technically classified as overweight, they were included in the obesity group because their other characteristics—such as BP, biochemical markers, and glucose levels—closely resembled those of individuals in the obesity category with a BMI of $\ge$30 kg/m².

Anthropometry

Height was measured using a portable stadiometer (Leicester Height Measure, Manchester, UK), and body composition was assessed using the In Body 127 scale (Microcaya S.L., Bilbao, Spain). Waist circumference was measured with a flexible measuring tape while volunteers wore light clothing without overgarments. Participants were then classified according to their BMI based on the criteria established by the Spanish Obesity Society (Salas-Salvadó et al., 2007).

Clinical parameters

The study volunteers were recruited early in the morning and instructed to fast for at least 8 h. Upon arrival at the laboratory, BP was measured following the procedure recommended by the American College of Cardiology/American Heart Association (Whelton et al., 2017). This protocol includes a 15-min rest before measurement, ensuring the patient has emptied their bladder, and avoiding conversation during both the rest period and the measurement, among other recommendations (Whelton et al., 2018).

For the measurements the Minimus^® II manual sphygmomanometer (Riester, Jungingen, Germany) was used alongside with the Duplex^® 2.0 phonendoscope (Riester, Jungingen, Germany). BP measurements were taken in duplicate, and BP was categorized into four levels based on the average values recorded, following the guidelines of the American College of Cardiology/American Heart Association: normal, elevated, and stage 1 or 2 hypertension (see Table 1).

After the BP measurements, a blood sample was taken from the volunteers to determine their plasma levels of glucose, high-density lipoprotein cholesterol, low-density lipoprotein cholesterol, very-low-density lipoprotein cholesterol, total cholesterol, triglycerides, insulin, and glycated hemoglobin. A capillary whole blood glucose measurement was then performed using a portable reflectance meter (OneTouch Select^® Plus; Johnson & Johnson, S.A., Madrid, Spain). Volunteers with capillary glucose values of <126 mg/dL underwent an oral glucose tolerance test, which involved consuming a 75-g oral glucose load.

Criteria for volunteers’ classification: glucose levels and presence of MS

The classification of volunteers based on glucose levels was conducted using the standard diagnostic criteria proposed by the International Diabetes Federation and the World Health Organization (Magliano, Boyko & 10th Edition Scientific Committee, IDA, 2021). These criteria also incorporate recommendations from the American Diabetes Association, which defines “prediabetes” as HbA1c values between 5.7% and 6.4% and impaired fasting glucose as fasting plasma glucose levels between 100 and 125 mg/dL. These parameters were included in the classification criteria for the volunteers.

For the classification of volunteers with or without MS, we followed the International Diabetes Federation worldwide definition of MS (Alberti, Zimmet & Shaw, 2016), which includes:

• • Central obesity: Waist circumference—ethnicity specific (for Europeans: $\ge$94 cm for men and $\ge$80 cm for women). If BMI > 30 kg/m², central obesity can be assumed.

• • Plus, any of the following:

  • –

    Raised triglycerides: $\ge$150 mg/dL

  • –

    Reduced high-density lipoprotein cholesterol: <40 mg/dL in men, <50 mg/dL in women

  • –

    Raised BP: systolic BP of $\ge$130 mmHg or diastolic BP of $\ge$85 mmHg

  • –

    Raised plasma glucose: fasting plasma glucose of $\ge$100 mg/dL or a diagnosis of T2D

Fecal DNA extraction and quantification

The commercial DNeasy PowerSoil Kit (Qiagen, Hilden, Germany) was used to extract bacterial DNA from the samples following the manufacturer’s instructions. The Qubit^TM 4 fluorometer (Invitrogen, Thermo Fisher Scientific, Waltham, MA, USA) and the DNA HS Assay Kit (Invitrogen, Thermo Fisher Scientific, Waltham, MA, USA) were used to quantify the extracted DNA from the samples. DNA samples were stored at $-20$ °C until further analysis.

16S rRNA amplicon sequencing

For 16S rRNA amplicon sequencing, 2 μL of DNA extracted from each sample was used to construct the libraries, and sequencing was performed using the Ion GeneStudio^TM S5 system (Life Technologies, Carlsbad, CA, USA). The 16S hypervariable regions were amplified with two primer sets, V2-4-8 and V3-6,7-9, and libraries were prepared using the Ion 16S^TM Metagenomics Kit (Life Technologies, Carlsbad, CA, USA) and the Ion Xpress^TM Plus Fragment Library Kit (Life Technologies, Carlsbad, CA, USA). Libraries containing equal amounts of polymerase chain reaction products combined with a barcode were prepared using the Ion Xpress^TM Kit barcode adapters (Life Technologies, Carlsbad, CA, USA) and quantified using the Ion Universal Library Quantification Kit (Life Technologies, Carlsbad, CA, USA). Next, 10 pM of each library was pooled and loaded onto an Ion OneTouch^TM 2 system (Life Technologies, Carlsbad, CA, USA), which automatically performed template preparation and enrichment. Positive ion sphere particles for the template were enriched using Dynabeads^TM MyOne^TM Streptavidin C1 magnetic beads (Invitrogen, Carlsbad, CA, USA) with an Ion OneTouch^TM ES instrument. Finally, an Ion 520^TM chip (Life Technologies, Carlsbad, CA, USA) was loaded with the samples on an Ion GeneStudio^TM S5 System sequencer, using the Ion 520^TM and Ion 530^TM loading reagents supplied in the OT2 Kit (Life Technologies, Carlsbad, CA, USA). Portions of this text were previously published as part of our prior work (Sinisterra-Loaiza et al., 2023).

Bioinformatics analyses

For all 16S rRNA sequence data, the quality of the raw reads was visualized using FastQC v0.11.7 (Andrews, 2010) to ensure an average quality score of at least 25. The reads were then imported into R (v4.2.0) (R Core Team, 2013) and assembled into amplicon sequence variants (ASVs) using the DADA2 package (v1.26) (Callahan et al., 2016). To asses the length distribution of all sequences in aggregate terms a second quality check of the samples was performed using the quality plots provided by DADA2.

For sequence processing, the default parameters provided by the author of DADA2 were left unchanged. However, following the author’s recommendations for sequences from Ion Technologies, several parameters were modified. In the initial filtering phase, a trimLeft value of 15 was applied. During the inference of real biological sequences phase, a homopolymer gap penalty of −1 and a band size of 32 were set. A total of 20,892 unique variants reached the chimera elimination phase. Of these, 9.9% were removed after being identified as chimeras by the algorithm, leaving 18,831 unique sequences. However, these chimeras accounted for only 1.5% of the total abundance of variants. The non-chimeric assembled ASVs were then taxonomically assigned (from phylum to species) using the Silva reference database (v138.1) (McLaren & Callahan, 2021).

The ASVs table, taxonomic table, and clinical data of the patients were combined into a phyloseq object (v1.42.0) (McMurdie & Holmes, 2013) for downstream bioinformatics analysis. Once the phyloseq object was constructed, agglomeration was performed at different taxonomic levels, followed by an initial filtering phase. During this phase, taxa that were not present more than three times in at least 20% of the samples were eliminated. Additionally, taxa with fewer than 20 counts were removed to ensure data reliability.

The phyloseq package was used to calculate bacterial diversity, including both alpha-diversity and beta-diversity. Alpha-diversity was evaluated using the Shannon index and the Simpson index, while beta-diversity was assessed using the Bray–Curtis index and the Chao index. To visually examine group dissimilarities, principal coordinate analysis (PCoA) was applied. Alpha-diversity was analyzed using non-parametric statistical tests, specifically the Wilcoxon test for comparisons between two groups and the Kruskal–Wallis test for comparisons among three or more groups, implemented through the ggpubr package (v0.5.0) (Kassambara & Kassambara, 2020). Beta-diversity was assessed using permutational analysis of variance with 1,000 permutations, performed with the adonis2 function from the R vegan package (Oksanen et al., 2022).

The LinDA method (Zhou et al., 2022) was used to measure the differences in abundance between the various groups. This method was specifically designed for differential abundance analysis of microbiome compositional data. When calculating differences in the abundance of genera and species, BMI was used as a covariate to ensure that the model accounted for data from both normal-weight and obese patients, thereby minimizing potential bias. The MicrobiomeStat package (Zhang & Chen, 2022) was used to perform this analysis.

Microbial composition and diversity in MS and T2D

In the initial phase, a series of exploratory analyses were conducted to examine the overall microbial composition and diversity in relation to MS and T2D. These analyses considered both the entire population and different pathological conditions. Figure 2A illustrates the abundance of various taxa across different taxonomic levels. The relative abundance of up to 10 taxa is represented individually, with the remaining taxa aggregated. At the phylum level, Firmicutes and Bacteroidota were the most abundant, followed by Actinobacteriota and Proteobacteria. At the family level, Lachnospiraceae and Bacteroidaceae dominated the microbiota of this cohort, followed by Ruminococcaceae and Prevotellaceae, with abundance gradually decreasing across the remaining families. Finally, at the genus level, the microbiota was primarily characterized by the abundance of the Bacteroides genus, followed by a substantial number of genera with more homogeneous abundances.

Subsequently, the ratios of different phyla were analyzed in relation to MS and T2D (all data available in File S1). In patients with MS (Fig. 2B), the relative abundance of the Firmicutes phylum was higher than in healthy individuals, along with an increase in Bacteroidetes. The proportions of the remaining phyla remained similar between both groups. For T2D, the relative abundance of Firmicutes and Actinobacteriota decreased as the disease progressed, while an inverse trend was observed for Bacteroidetes. Additionally, the relative abundance of Proteobacteria was slightly higher in diabetic individuals than in healthy controls. The proportions of the other phyla remained relatively stable across groups, as shown in Fig. 2C.

Alpha diversity, assessed using the Shannon and Simpson indices (Fig. 2D), was first evaluated to detect potential differences in microbial diversity between patients with MS and healthy controls. The results showed no statistically significant differences, suggesting that the overall microbiota diversity in patients with MS is similar to that of healthy individuals. For beta diversity, the Bray–Curtis and Chao distance metrics were used to assess differences in microbial community composition between the two groups. The PCoA plot in Fig. 2D reveals a significant separation between patients with MS and healthy controls, with associated $p$-values of 0.002 and 0.01. This indicates a marked compositional shift in the microbiota between these groups despite the similarity in alpha diversity.

The second experiment aimed to compare microbiota diversity among healthy individuals, prediabetic patients, and diabetic patients. Alpha diversity, evaluated again using the Shannon and Simpson indices (Fig. 2E), did not reveal significant differences between these groups, indicating comparable microbial diversity across the spectrum. The PCoA plots in Fig. 2E illustrate no discernible differences in microbiota composition among healthy individuals, prediabetic patients, and diabetic patients. Additionally, statistical analysis did not yield significant $p$-values, suggesting that microbiota compositional changes between these groups are not statistically robust.

Analysis of taxon abundances in MS

Subsequently, we aimed to examine variations among taxa across patient groups and identify specific patterns within these taxa. To achieve this, an abundance difference analysis was conducted using the LinDA method. Additionally, the analysis was performed at both the genus and species levels to assess differences based on taxonomic depth. Detailed results of these analyses are available in File S1.

Figure 3A presents the results of the comparison between healthy individuals and patients with MS. Upon careful examination of the figure, it became evident that genera such as Tyzerella, Streptococcus, Blautia, Family XIIIAD3011 group, Dorea, and Paludicola, among others, were enriched in patients with MS, with log₂ fold changes (log₂FCs) ranging from approximately 1.5 to 3.0. By contrast, healthy individuals exhibited a higher relative abundance of genera such as [Eubacterium] eligens group and Prevotellaceae NK3B31 group, with log₂FCs lower than −2.

At the species level (Fig. 3B), the representation of the genus Blautia in patients with MS remained and became more pronounced, with four species—Blautia faecis, Blautia massiliensis, Blautia obeum, and Blautia wexlerae—showing higher relative abundance. The genus Dorea also maintained significance, represented by Dorea formicigenerans. These species were accompanied by others such as Bifidobacterium bifidum, Intestinibacter bartlettii, Roseburia intestinalis, and Ruminococcus callidus, with log₂FCs ranging from approximately 1.5 to 3.0. On the other hand, healthy individuals appeared to be enriched in the species Parabacteroides merdae, which exhibited a log₂FC of approximately −2.

Analysis of taxon abundances in T2D

Next, to better understand the role of the microbiota in T2D, differences in taxon abundances were analyzed across three comparisons: healthy individuals vs. prediabetics, healthy individuals vs. diabetics, and prediabetics vs. diabetics. As in the previous case, detailed results of these analyses are available in File S1.

In the comparison between healthy individuals and those with prediabetes, Fig. 4A illustrates that the genus Dorea ( $p$ = $7.80535\times {10}^{-9}$, log₂FC = 10) was highly enriched in prediabetic patients. Additionally, genera such as Anaerostipes, Coprococcus, Fusicatenibacter, Subdoligranulum, and [Eubacterium] hallii group, among others, showed increased abundance in prediabetic individuals, with log₂FCs between 7.5 and 10. By contrast, genera such as Ligilactobacillus, Acidaminococcus, Bifidobacterium, Faecalibacterium, Collinsella, Defluviitaleaceae UCG-011, and GCA-900066575, among others, were enriched in healthy individuals, with log₂FCs ranging from −5 to −15.

At the species level, four genera enriched in prediabetic patients persisted as the only species showing higher relative abundance in this group. These species were Dorea longicatena, Anaerostipes hadrus, Fusicatenibacter saccharivorans, and Coprococcus comes, with log₂FCs ranging from approximately 5 to 8. In healthy individuals, species from three genera were enriched, represented by Lachnoclostridium edouardi, Collinsella aerofaciens, and Bifidobacterium bifidum, with log₂FCs ranging from approximately −5 to −8.

In the comparison between healthy individuals and diabetics, shown in Fig. 4B, the genus Dorea ( $p$ = $2.16796\times {10}^{-5}$, and a log₂FC = 7.5) appeared to be more dominant in diabetic patients, followed by Anaerostipes and Subdoligranulum. Meanwhile, in healthy individuals, genera such as Ligilactobacillus, Acidaminococcus, Bifidobacterium, Faecalibacterium, GCA-900066575, Defluviitaleaceae UCG-011, [Eubacterium] ventriosum group, and Collinsella, among others, exhibited greater prevalence, with log₂FCs ranging from approximately −5 to −12.5.

At the species level, a similar pattern to that observed at the genus level was evident. The only species enriched in diabetic patients was Dorea longicatena, with a log₂FC of approximately 5. In contrast, in healthy individuals, species such as Alistipes putredinis, Bifidobacterium adolescentis, Bifidobacterium longum, Blautia faecis, Butyricicoccus faecihominis, and Collinsella aerofaciens were enriched, with log₂FCs ranging from approximately −5 to −10.

Finally, Fig. 4C, illustrates the differences between prediabetic and diabetic patients. In this case, genera such as Prevotella, Dialister, and Sutterella were more dominant in diabetic patients, with log₂FCs ranging from 5.0 to 7.5. On the other hand, genera such as CAG-56, Erysipelotrichaceae UCG-003, Fusicatenibacter, Lachnospiraceae FCS020 group, Roseburia, [Eubacterium] hallii group, [Eubacterium] ventriosum group, and Dorea, among others, were more prevalent in prediabetic individuals, with log₂FCs ranging from −2.5 to −5.0.

At the species level, diabetic patients exhibited an enriched abundance of Butyricimonas faecihominis, Dialister invisus, Lachnoclostridium edouardi, and Odoribacter splanchnicus, with log₂FCs ranging from approximately 2.5 to 5.0. By contrast, species such as Blautia faecis and Dorea longicatena were enriched in prediabetic patients, with log₂FCs ranging from approximately −2 to −4.