Novel Interactions Between Circulating microRNAs and Gut Microbiota Composition in Human Obesity.

Background: Unbalances in microRNAs (miRNA) and gut microbiota patterns have been proposed as putative factors concerning onset and development of obesity and other metabolic diseases. However, the determinants that mediate the interactions between miRNAs and the gut microbiome impacting on obesity are scarcely understood. Thus, the aim of this article was to investigate possible interactions between circulating miRNAs and gut microbiota composition in obesity. Method: The analyzed sample comprised 78 subjects with obesity [cases, body mass index (BMI): 30 – 40 kg/m 2 ] and 25 eutrophic individuals (controls, BMI £ 25 kg/m 2 ). The expression of 96 miRNAs was investigated in plasma of all individuals using miRCURY LNA miRNA Custom PCR Panels (Exiqon). Bacterial DNA sequencing was performed following the Illumina 16S protocol. The FDR (Benjamini-Hochberg test, q-value) correction was used for multiple comparison analyses. Results: A total of 26 circulating miRNAs and 12 bacterial species were found differentially expressed between cases and controls. Interestingly, an interaction among three miRNAs (miR-130b-3p, miR-185-5p, and miR-21-5p) with Bacteroides eggerthi, and BMI levels was evidenced (r 2 = 0.148, P= 0.004). Those miRNAs that correlated with obesity-associated gut bacteria abundance are known to regulate target genes that participate in metabolism-related pathways, such as fatty acid degradation, carbohydrate digestion and absorption, insulin signaling, and glycerolipid metabolism. Conclusion: This study characterized an interaction between the abundance of 4 bacterial species and 14 circulating miRNAs in relation to body adiposity. Moreover, the current study also suggests that miRNAs may serve as a communication mechanism between the gut microbiome and human hosts. Clinical trial registration: clinicaltrials.gov

consequently, changes in their expressions and functions have been linked to many diseases, including metabolic disorders and obesity (Maurizi et al., 2018;Lorente-Cebrián et al., 2019).
Recent ndings indicate that host miRNAs contribute to the regulation of the gut microbiome, specially involving at least two main processes: (i) host-secreted miRNAs regulate the gut microbiota; and (ii) the gut microbiota affects the host via inducing special miRNA expression (Belcheva, 2017). Indeed, evidences suggest that miRNAs produced by the host's intestinal epithelial cells (IECs) participate in shaping the gut microbiota and affect bacterial growth (Liu and Weiner, 2016). These miRNAs target bacterial mRNA, and then the host controls the gut microbiota via bacterial mRNA degradation or translational inhibition (Liu and Weiner, 2016). On the other hand, it was demonstrated, using Dicer1 knock-out mice, that miRNAs were essential for epithelial cell proliferation, differentiation, nutrient absorption, and that defective miRNA biogenesis was also responsible for impaired intestinal barrier function (McKenna et al., 2010).
Additionally, the gut microbiota regulates miRNA expression in IECs subtypes, and this regulation may alter intestinal homeostasis (Nakata et al., 2017). In this sense, it was demonstrated that the expression of some miRNAs is different among IEC subtypes and the difference depends on microbial patterns (Peck et al., 2017). Similarly, the expression of 16 miRNAs was found to be altered in the caecum of conventionally raised versus germ-free mice (Singh et al., 2012). Recently, it has been reported that the gut microbiota speci cally controlled the miR-181 family expression in white adipocytes during homeostasis to modulate key pathways related to adiposity, insulin sensitivity, and white adipose tissue (WAT) in ammation in mice (Virtue et al., 2019). Furthermore, high-fat diet (HFD) feeding altered the composition of the gut microbiota, leading to aberrant overexpression of miR-181 in WAT adipocytes (Virtue et al., 2019). Altogether, these studies provide clues that gut microbiota regulates host gene expression through modulation of the host miRNA signature, and that host metabolism could be in uenced by this interaction.
According to these ndings, miRNAs appear to play an important role in host-to-microbe interaction and may be considered as molecular targets for novel anti-microbial therapies to be developed. However, very little is known about the interactions between miRNAs and the host microbiome in the context of obesity. Therefore, the aim of this study was to investigate interactions between circulating miRNA patterns and gut microbiota composition in obesity.

Study population
This study was designed in accordance with STROBE guidelines for reporting association studies ( Nutrition Research of the University of Navarra, Spain. Major exclusion criteria included a history of diabetes mellitus (DM), cardiovascular disease or hypertension, pregnant or lactating women, current use of lipid-lowering drugs or medications that affect body weight, and weight change ≥3 kg within three months before the recruitment.
This study followed the ethical principles for medical research in humans from the Helsinki Declaration (Association, 2013). Moreover, the research protocol was properly approved by the Research Ethics Committee of the University of Navarra (ref. 132/2015) and it is registered at clinicaltrials.gov (reg. no. NCT02737267). A written informed consent of each participant was obtained prior to enrollment in the study.
All patients underwent anthropometric and laboratory evaluations, as previously described (Lopez-Legarrea et al., 2013; Ramos-Lopez et al., 2018). The measurements of height (cm), body weight (kg), and waist circumference (WC, cm) were collected in the fasting state by trained nutritionists following validated procedures (Lopez-Legarrea et al., 2013). BMI was calculated as the ratio between weight and squared height (kg/m 2 ). Body composition was quanti ed by dual-energy X-ray absorptiometry according to instructions provided by the supplier (Lunar Prodigy, software version 6.0, Madison, WI, USA). Biochemical measurements including fasting plasma glucose (FPG, mg/dl), total cholesterol (TC, mg/dl), high-density lipoprotein cholesterol (HDL-c, mg/dl), and triglycerides (TG, mg/dl) were determined in an automatic analyzer (Pentra C200, HORIBA Medical), following standardized procedures. Endocrine markers such as insulin, adiponectin, and leptin were quanti ed with commercial ELISA kits (Mercodia Insulin ELISA, Biovendor Human adiponectin ELISA, and Mercodia Leptin).
A validated semiquantitative food frequency questionnaire was used to evaluate habitual consumption (daily, weekly, monthly, or never) of 137 foods during the previous year (de la Fuente-Arrillaga et al., 2010).
Energy and nutrient intakes were further calculated with an ad hoc computer program based on the standard Spanish food composition tables (Moreiras, 2018 MicroRNA expression analysis miRNA isolation and reverse transcription-quantitative PCR Total RNA was extracted from 200µ l EDTA-plasma using the miRNeasy Serum/Plasma Advanced kit (Qiagen. Hilden, Germany) according to the manufacturer's recommendations. RNA spike-in was added to each sample (RNA Spike-In Kit, Qiagen. Hilden, Germany). The purity and concentration of RNA samples were measured using the NanoDrop ND-1000 Spectrophotometer (Thermo Fisher Scienti c. Massachusetts, USA) (Bustin et al., 2009).
Relative expression of the 86 miRNAs was analyzed in plasma from all subjects using the Custom Pick-&-Mix microRNA PCR Panel v5 (Qiagen. Hilden, Germany). Moreover, 9 controls (reference genes + Spike-in controls) and a blank were also included in each plate, as shown in Supplementary Table 1

Quality control and normalization
Quality control was carried out using synthetic spike-in RNAs to analyze the robustness of the RNA isolation process and quality of extracted miRNA. The RNA isolation controls (UniSp2, UniSp4, and UniSp5; Qiagen. Denmark) were added to the thawed plasma before the isolation process, aiming to detect differences in the extraction e ciency. The cDNA synthesis control (UniSp6, Qiagen) and cel-miR-39-3p were added to the reverse transcription reaction to determine the effectiveness of this process.
Furthermore, UniSp3 was included in all plates and used as an inter-plate calibrator and PCR ampli cation control.
Hemolysis was assessed by the ratio between hsa-miR-451a (which is expressed in erythrocytes) and hsa-miR-23a-3p (which is relatively stable in plasma and not affected by hemolysis) as described elsewhere (Blondal et al., 2013). The difference in expression values between these 2 miRNAs provides a good measure of the extent of hemolysis, with values > 5 suggesting erythrocyte miRNA contamination. Only samples without hemolysis (values < 5) were included in the study. (Blondal et al., 2013). The assay cut-off was 35 cycles, and miRNAs expressed in at least 20% of the total sample (Gevaert et al., 2018). All individual samples were run on a prede ned assay panel of 96 speci c human miRNAs (Supplementary  table 1). The miRNAs with complete data were used for the global mean method for normalization of the data, since this approach was found to be the most stable normalizer (Mestdagh et al., 2009). miRNA target prediction and pathway enrichment analysis Potential targets of selected miRNAs were searched using miRWalk 3.0 (http://zmf.umm.uniheidelberg.de/apps/zmf/mirwalk2/, accessed 4th August 2020). To better understand the biological relevance of the miRNAs target genes, a network analysis was executed using PathDIP (accessed 4th August 2020, (Rahmati et al., 2017)). A hypergeometric test was used to calculate the statistical signi cance of the enriched pathways, and P-values were corrected for multiple tests using the Benjamini-Hochberg procedure, which provides a False Discovery Rate (FDR) adjusted-P-value (q-value). Pathways associated with a q-value < 0.05 were considered signi cantly enriched. OTUs, phylum, genus, family, order, class, and species. Brie y, taxa less than 10% of frequency in our population were removed for the analysis and a global normalization was performed using the library size as a correcting factor and log2 data transformation.

Gut Microbiota Analysis
To evaluate alpha diversity, the Shannon index was calculated (Shannon, 1997). To assess beta diversity, permutational multivariate analysis of variance (PERMANOVA) was used to analyze whether the structures of gut microbiota were signi cantly different among groups based on the Jaccard and Bray-Curtis distance matrices (Anderson et al., 2006).

Statistical analysis
Normalized data (RQ expression levels) were initially analyzed, with an estimation and comparison of expression levels between groups. Normal distribution of data was assessed using the Kolmogorov-Smirnov and Shapiro-Wilk tests. Variables with normal distribution are presented as mean ± standard deviation (SD). Variables with skewed distribution were log-transformed prior to analysis and are presented as median (25th -75th percentiles). Categorical data are shown as percentages. Clinical and laboratory characteristics, miRNA expressions, and gut microbiota abundance were compared among groups using Student's t-test or χ 2 tests, as appropriate. Correlations between quantitative variables were assessed using Pearson's correlation tests.
All classical statistical analyses were performed using the SPSS statistical package (v.20.0) for Windows (SPSS Inc, Chicago, IL, USA) and PAST v3.24 (University of Oslo, Norway) for statistical analyses of biodiversity. FDR correction was used to account for multiple comparisons using the Benjamini-Hochberg method (q-value < 0.05). The network visualization of miRNA-microbe was generated using Cytoscape v.3.7.1 (Shannon et al., 2003). One heatmap plot of the correlation values were produced using MORPHEUS web tool (Morpheus, https://software.broadinstitute.org/morpheus).

Results
Clinical and laboratory characteristics of individuals included in the study.
Clinical, laboratorial, and nutritional characteristics of cases with obesity and eutrophic controls are shown in Table 1. There were no differences between cases and controls regarding age, gender, and energy intake. Moreover, both groups had a comparable dietary macronutrient composition. As expected, subjects with obesity presented higher waist circumference, glucose, total cholesterol, and triglyceride levels compared to normal weight individuals. Additionally, cases also presented elevated levels of metabolic markers such as insulin, leptin, TyG, and HOMA-IR indexes and, lower levels of METs compared to controls. Quality control of miRNA expression The RNA spike-in expressions presented low variation in Cq among samples in RNA isolation and cDNA synthesis, demonstrating that extraction, reverse transcription, and qPCR were effective, and none of the samples contained inhibitors. As expected, the expression of UniSp2, UniSp4, UniSp5, and UniSp6 did not differ between groups (cases vs. controls). UniSp5 was expressed in all analyzed samples, demonstrating that miRNAs expressed in low levels was not lost during isolation. The ratio between miR-451a and miR-23a-3p ranged between 5 and − 1, indicating that the samples were not affected by hemolysis. Generally, these results showed a good and similar level of sample quality and reproducibility of the miRNA pro ling processes.
MicroRNAs differentially expressed in plasma of patients with obesity.
Expression of 86 target miRNAs was evaluated in plasma of subjects with obesity and in normal weight individuals. Of these 86 miRNAs, 61 were expressed in at least 20% of the sample with Cq values ≥ 35. Of these 61 miRNAs, 26 were differentially expressed between cases and controls after FDR correction (Table 2). Data are shown as median (25th-75th percentiles) of n-fold values. *P-values were obtained using Student t test using the log-transformed variable. **P-values were corrected using false discovery rate (FDR; q-value).

Gut microbiota pro le in subjects with obesity compared to eutrophic individuals
The effect of obesity on gut microbiota composition was investigated at the genus and species levels.
The levels of eighteen bacterial genera were signi cantly different when comparing obese and normal weight individuals, being nine bacterial genera signi cantly increased in obese subjects when compared to controls (Fig. 1A). Twelve bacterial species were statistically different between obese and normal weight individuals, being ten of them more abundant in subjects with obesity compared to eutrophic individuals ( Fig. 1B and Table 3). Shannon index, which re ects the alpha diversity, was not different between obese and normal weight groups ( Supplementary Fig. 1). However, the beta diversity values of gut microbiota, based on Jaccard index (PERMANOVA, P = 0.025; Supplementary Fig. 2A) and Bray-Curtis dissimilarity (PERMANOVA, P = 0.015; Supplementary Fig. 2B), was signi cantly different between groups.
Crosstalk between host miRNAs and gut microbiota To further investigate the relationships between circulating miRNAs and the gut microbiota composition, interactions between bacteria and miRNAs differentially expressed in obesity were analyzed. At the genus level, of the 18 genera differently expressed in obesity, 9 were signi cantly correlated with the expression of 10 miRNAs out of 26 miRNAs differently expressed in subjects with obesity ( Fig. 2A). Fourteen of these miRNAs were signi cantly associated with 4 bacterial species (Dorea longicatena, Banesiela intestinihominis, Bacteroides eggerthii, and Haemophillus parain uenzae), as illustrated in Fig. 2B and Fig. 2C.

Predicted functions of miRNAs correlated with obesity-associated bacteria
Target gene prediction of the 14 miRNAs that correlated with the 4 bacterial species associated with obesity were investigated (Supplementary Table 2). Of the total 9,584 genes identi ed as potential targets of these miRNAs, 5,381 were found to be regulated by two or more miRNAs; however only 719 were experimentally validated (Supplementary Table 2). After that, functional enrichment analysis of miRNA targets was carried out to explore biological pathways possibly regulated by this set of miRNAs. A total of 248 pathways were signi cantly enriched (q-value < 0.05) for these miRNAs (Supplementary Table 3). However, considering only the experimentally validated target genes, 98 pathways were signi cantly enriched (Supplementary Table 3).
As shown in Fig. 3, H. parain uenzae, D. longicatena, B. intestinihominis, and B. eggerthii correlated with miRNAs associated with pathways related to obesity and metabolic processes, including carbohydrate and lipid turnover, endocrine and in ammatory signaling pathways. More speci cally, the target genes of miRNAs associated with the four bacterial species related to obesity participate in the fatty acid degradation, mineral absorption, carbohydrate digestion and absorption, insulin signaling pathway, and glycerolipid metabolism.
These miRNAs that interacted with obesity-associated bacteria regulate the expression of genes that participate in several metabolism and obesity-related pathways, such as carbohydrate and lipid metabolism, endocrine and in ammatory signaling pathways. Indeed, evidence suggests that the majority of miRNAs do not regulate a speci c or individual target gene, but rather they modulate the expression of large number of genes in networks, demonstrating their importance in the regulation of several metabolic processes (Bartel, 2004;Virtue et al., 2019).
Additionally, an interaction between BMI levels, B. eggerthii abundance, and the expression of three miRNAs (miR-130b-3p, miR-185-5p, and miR-21-5p) was also evidenced. Interestingly, B. eggerthii is one of the intestinal bacteria that metabolize phenolic acids, which are regarded as bene cial for human health (Russell et al., 2013). In a recent study, B. eggerthii abundance was signi cantly higher in children with obesity and positively correlated with body fat percentage, but negatively with insoluble ber intake in Mexican children (López-Contreras et al., 2018). On the other hand, this bacterium was found to be underrepresented after sleeve gastrectomy surgery (Medina et al., 2017).
Of the three miRNAs associated with the abundance of B. eggerthii and BMI levels, miR-185-5p and miR-21-5p were also correlated with D. longicatena abundance. Furthermore, miR-185-5p was described as involved in oxidative stress, obesity, and DM in many studies [reviewed at (Matoušková et al., 2018)].
Moreover, miR-185-5p was identi ed as a regulator of de novo cholesterol biosynthesis and low density lipoprotein uptake (Yang et al., 2014). However, we could not nd in the literature evidences of association between this miRNA and gut microbiota. Regarding gut microbiota composition, our results evidenced that obesity had no signi cant impact in alpha diversity, indicating that microbial species diversity is relatively stable in response to obesity. However, obesity in uenced the beta diversity of human gut microbiota compared to the control group, suggesting that this disease is accompanied by species replacement (changes in species taxa) and species sorting (changes in abundance).
According to a meta-analysis of metagenomic datasets obtained from fecal samples of healthy human adults living in different world regions, Bacteroides and Barnesiella genera are markers of Western populations (Mancabelli et al., 2017). Barnesiella spp (represented mainly by the specie Barnesiella intestinihominis) were identi ed only in populations living in developed countries, suggesting that their presence was promoted by the urbanization/ industrialization process and Western-type diet (Mancabelli et al., 2017).
In agreement with our results, the levels of Dorea genera were previously reported to be higher in overweight children compared to normal weight counterparts (Karvonen et al., 2019). Moreover, this association was stronger for non-white children than for white children, and also stronger for boys than for girls (Karvonen et al., 2019). Interestingly, a recent study in an early-life HFD mouse model found that the this diet increased the relative abundances of Dorea genus (Villamil et al., 2018).
Our investigation has strengths and limitation. The strengths include study and data analyses of a verywell characterized cohort of subjects with obesity and eutrophic subjects was analyzed. Moreover, several quality controls for miRNA extraction, cDNA synthesis, and PCR process were implemented. Additionally, robust bioinformatic analyses were performed to explore the pathways where these miRNAs target genes are participating, explaining the association with obesity. Likewise, we highlighted candidates for potentially linking host miRNAs and gut microbiota, which can be directly validated and explored in model systems.
Even though these methods are powerful, this evaluation has some limitations. First, it is important to note that our study uses 16S rRNA gene sequencing to characterize microbiome taxonomic composition.
Second, the results from bioinformatics are predictions and may not represent the real biological system. Third, our approach identi es correlations and not causal relationships. Even though a hypothesis-driven approach was performed, selection of only miRNAs previously associated with obesity or metabolism makes possible type I or type II errors due to multiple comparisons. These limitations should be considered when interpreting the results. Although limitations exist in the current data, the patterns uncovered here are important for understanding the contribution of miRNAs and gut microbiota in obesity.

Conclusion
This current research characterized a global relationship between microbial community composition and miRNA expression in plasma of subjects with obesity compared to normal weight individuals. Indeed, our study featured an interaction between B. eggerthi abundance and circulating miRNA expression in the control of body adiposity. The current study also adds to the growing body of literature that suggests that miRNAs may serve as a communication mechanism between the gut microbiota and human hosts. Ethics approval and consent to participate. The research was conducted in accordance with the rules of the Helsinki Declaration. The research protocol was properly approved by the Research Ethics Committee of the University of Navarra (ref. 132/2015) and it is registered at clinicaltrials.gov (reg. no. NCT02737267). A written informed consent of each participant was obtained prior to enrollment in the study.
Consent for publication. All authors approved the nal version and agreed to be accountable for all aspects of the work regarding accuracy and integrity aspects. All authors agree to publish this article in the journal of Molecular Medicine. Figure 1 Gut microbiota composition in subjects with and without obesity. A) Bacterial genera. B) Bacterial species. Differences in bacterial abundance at the genus and species levels in cases with obesity and controls with normal weight. Only genera or species whose abundances were signi cantly different (P 0.05) are shown.

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download. Supplementarytable3.xlsx Supplementarytable2.xlsx Supplementarytable1.docx