Metformin-induced changes of the gut microbiota in healthy young men: results of a non-blinded, one-armed intervention study

Aims/hypothesis Individuals with type 2 diabetes have an altered bacterial composition of their gut microbiota compared with non-diabetic individuals. However, these alterations may be confounded by medication, notably the blood-glucose-lowering biguanide, metformin. We undertook a clinical trial in healthy and previously drug-free men with the primary aim of investigating metformin-induced compositional changes in the non-diabetic state. A secondary aim was to examine whether the pre-treatment gut microbiota was related to gastrointestinal adverse effects during metformin treatment. Methods Twenty-seven healthy young Danish men were included in an 18-week one-armed crossover trial consisting of a pre-intervention period, an intervention period and a post-intervention period, each period lasting 6 weeks. Inclusion criteria were men of age 18–35 years, BMI between 18.5 kg/m2 and 27.5 kg/m2, HbA1c < 39 mmol/mol (5.7%) and plasma creatinine within the normal range. No prescribed medication, including antibiotics, for 2 months prior to recruitment were allowed and no previous gastrointestinal surgery, discounting appendectomy or chronic illness requiring medical treatment. During the intervention the participants were given metformin up to 1 g twice daily. Participants were examined five times in the fasting state with blood sampling and recording of gastrointestinal symptoms. Examinations took place at Frederiksberg Hospital, Denmark before and after the pre-intervention period, halfway through and immediately after the end of intervention and after the wash-out period. Faecal samples were collected at nine evenly distributed time points, and bacterial DNA was extracted and subjected to 16S rRNA gene amplicon sequencing in order to evaluate gut microbiota composition. Subjective gastrointestinal symptoms were reported at each visit. Results Data from participants who completed visit 1 (n=23) are included in analyses. For the primary outcome the relative abundance of 11 bacterial genera significantly changed during the intervention but returned to baseline levels after treatment cessation. In line with previous reports, we observed a reduced abundance of Intestinibacter spp. and Clostridium spp., as well as an increased abundance of Escherichia/Shigella spp. and Bilophila wadsworthia. The relative abundance at baseline of 12 bacterial genera predicted self-reported gastrointestinal adverse effects. Conclusions/interpretation Intake of metformin changes the gut microbiota composition in normoglycaemic young men. The microbiota changes induced by metformin extend and validate previous reports in individuals with type 2 diabetes. Secondary analyses suggest that pre-treatment gut microbiota composition may be a determinant for development of gastrointestinal adverse effects following metformin intake. These results require further investigation and replication in larger prospective studies. Trial registration Clinicaltrialsregister.eu 2015-000199-86 and ClinicalTrials.gov NCT02546050 Funding This project was funded by Danish Diabetes Association and The Novo Nordisk Foundation Electronic supplementary material The online version of this article (10.1007/s00125-019-4848-7) contains peer-reviewed but unedited supplementary material, which is available to authorised users.


ESM Methods
Clinical examination Height was measured without shoes to the nearest 0.5 cm using a wall mounted stadiometer (MZ10023, ADE, Germany). Weight was measured with participants wearing underwear to the nearest 0.1 kg (WB-110MA, Tanita Corporation of America, IL USA). Hip and waist circumference were recorded as the average of duplicate measurements to the nearest 0.5 cm using a non-expandable measuring tape. Waist circumference was measured midway between the lower rib margin and the iliac crest. Hip circumference was measured as the largest circumference between the waist and the thighs. Blood pressure was measured in the inclined position after a minimum 10 min rest using an automated sphygmanometer (UA-779, A&D medical, Japan) and recorded as the last of triplicate measurements. Body composition was assessed using a bioelectrical impedance analyzer (Tanita Body Composition Analyzer BC-420MA).

Biochemical analyses Plasma concentrations of alanine aminotransferase and cobalamin
were analyzed on a Vitros 5,1 FS/5600 (Ortho Clinical Diagnostics, NJ USA) platform, using multipoint enzymatic slide test and immunometric analysis with a CV of 19.7-33.0% and 6.0%, respectively. Creatinine, glucose, cholesterol, HDL cholesterol, and triglyceride were measured on the Vitros 5,1 FS/5600 platform using colorimetric slide test with CV 12.6%, 6.1%, 11.6%, 17.0% and 14.6%, respectively. LDL cholesterol was calculated using the Friedewald formula[1]. HbA1c was measured on a TOSOH G8 (Tosoh Bioscience, CA USA) platform using high performance liquid chromatography with a CV of 7.2%. Insulin concentration in plasma was quantified on a Roche Cobas 411 (Roche Diagnostics GmbH, Germany) platform using an enzyme-linked chemiluminescent immunoassay with a CV of 2.8%. Leukocytes and differential white blood cell count were measured on an Advia platform using coupled flow cytometry and peroxidase methods. Plasma metformin was measured by high performance liquid chromatography followed by tandem mass spectrometry (LC-MS/MS) at visit 3 and 4, respectively, in order to evaluate compliance. In brief, proteins were Following inspection of quality profiles, the first 10 nt of all reads were trimmed, with forward and reverse reads truncated at 220 nt and 155 nt, respectively. Reads containing ambiguous nucleotides or more than two expected errors were discarded. Sequencing error rates were estimated based on a random subset of 210 6 reads. Reads were dereplicated and denoised sample-wise, followed by merging of forward and reverse reads requiring a minimum overlap of 20 nt, with no mismatch allowed. Chimeric sequences were detected and removed, leaving 7,214,117 reads (mean = 34,702, SD = 12,889; minimum = 10,736) in 1,764 unique amplicon sequence variants (ASV) for downstream analyses. Taxonomical assignment of ASVs from kingdom to species was performed against the Silva v128 database, using the dada2 implementation of the naïve Bayesian RDP classifier.

Metformin responsive amplicon sequence variants.
Bacterial amplicon sequence variants (ASV) exhibiting a change in relative abundance during the metformin intervention. Boxes represent interquartile range (IQR), with the inside line representing the median. Whiskers represent values within 1.5IQR of the first and third quartiles. Circles represent individual samples with lines connecting samples from the same individual. The purple band represents the pre-intervention mean and 95% confidence limits averaged across the three pre-intervention time points. Diamonds and connecting lines represent mean values, with yellow and green diamonds respectively representing nominal (p<0.05) and false discovery rate adjusted (q<0.05) significant differences from the averaged pre-intervention mean. The relative abundance at each time point during the intervention was compared to the averaged pre-intervention mean by linear mixed model regression analysis of variance. Only ASVs with a significant change at least one time point following correction for false discovery rate is presented.
Effect of metformin on gut microbiota diversity. Change in microbial richness defined as the number of unique amplicon sequence variants (ASV) [A], overall diversity represented by Shannon's entropy [B] and evenness represented by Pielou's index [C] during and after the metformin intervention. Samples were rarefied to an equal sequencing depth of 10,000 reads. Boxes represent interquartile range (IQR), with the inside line representing the median. Whiskers represent values within 1.5IQR of the first and third quartiles. Circles represent individual samples with lines connecting samples from the same individual. Diamonds and connecting lines represent mean values. The purple band represents the pre-intervention mean and 95% confidence limits averaged across the three pre-intervention time points.

Effect of metformin on gut bacterial community structure and membership
[A] Unconstrained principal coordinate analysis of gut microbial community structure based on Canberra distances and [B] community membership based on Jaccard distances. Points are individual samples projected on the 1st and 2nd principal coordinate axes with lines connecting consecutive samples from the same individual and ellipses representing the 95% confidence intervals of a multivariate normal distribution stratified by study period. [C]