Rapid metformin effects on multiple intestinal functions in high-fat high-sucrose fed mice

1 Université de Lyon, Laboratoire CarMeN, INSERM U1060, INRA U1397, INSA Lyon, Université Claude Bernard Lyon1, 69600 Oullins, France. 2 Present address: Department of Molecular and Clinical Medicine/Wallenberg Laboratory, Institute of Medicine, University of Gothenburg and Sahlgrenska University Hospital, Gothenburg, Sweden. 3 Sorbonne Université, Centre de Recherche Saint -Antoine, INSERM U938, 75012 Paris, France.


INTRODUCTION
The oral antidiabetic drug metformin (1,1-dimethylbiguanide hydrochloride) is the first line therapy of type 2 diabetes mellitus (T2DM). Despite its use for more than 60 years, the mechanisms contributing to its effects on blood glucose levels are still the subject of debates 1,2 . Clinical studies have demonstrated that metformin acts primarily by the reduction of exaggerated endogenous glucose production in diabetic patients, especially through a decrease of hepatic gluconeogenesis rate 3 . Animal studies and in vitro experiments in hepatocytes revealed an inhibition of complex I of the mitochondrial electron transfer chain and activation of AMP-activated protein kinase (AMPK), leading to the repression of gluconeogenic gene expression 1,2 . Additional mechanisms, independent from AMPK, were also proposed, such as changes in cell redox state or inhibition of mitochondrial glycerol-3-phosphate dehydrogenase 1,2 . Several works have suggested that metformin may act primarily by targeting intestinal tract as it can accumulate and reach high concentrations in the small intestine 4 . Furthermore, administration of metformin per os appeared to reduce blood glucose more efficiently than the intravenous route 5 , and a gut-restricted formulation was shown to efficiently decrease glucose levels in T2DM patients 6 . It was also shown that metformin is able to inhibit intestinal glucose absorption in the proximal small intestine 7 , and to increase glucose uptake and utilization by the enterocytes 8 . Additional modifications of intestinal functions have been reported, including increased production of the incretin hormone, glucagon-like peptide 1 (GLP1) 9 , stimulation of Goblet cells specialized in mucin production 10 and bile acid pool modifications 11 . These effects can be related to a direct action of the drug on intestinal cells or to consequences of a modification of the gut microbiota. Indeed, it is currently well accepted that metformin is able to alter the overall structure and the functions of the gut microbiota 12,13 and it was shown that metformin can stimulate directly the growth of specific bacterial species, notably Akkermansia muciniphila 13,14 . However, the link between the modulation of microbiota and the metabolic actions of metformin is not yet fully established 15 .
Therefore, there are still some uncertainties regarding the mechanisms of action of metformin in the intestinal tract, and it is also difficult to evaluate if the observed changes are direct effects of the drug or consequences of modifications of other parameters, including the overall metabolic status of the host upon treatment. In the present study, we investigated the changes in a number of biological parameters in different regions of the intestinal tract after a short-term metformin treatment in high-fat high-sucrose (HFS) fed mice. In addition to its effects on the expression of key intestinal genes in different segments (duodenum, jejunum, ileum, colon), we evaluated the impact of the drug on the luminal microbiota composition of each gut segment and we performed a bile acid profiling in the caecum of the animals. Mice were fed HFS diet for 8 days and were treated with metformin by daily gavage. We found rapid multi-level effects of metformin, leading in one week to a restoration of most of the perturbations induced by HFS diet on gene expression, associated with modifications of the bile acid profile and of the luminal microbiota composition all along the gut.

Short metformin treatment prevented metabolic disturbances in HFS fed mice.
HFS feeding for a period of 8 days induced metabolic disturbances in adult C57BL/6J male mice, as evidenced by a significant rise in fasting glucose levels and a deterioration of glucose tolerance during ipGTT (Fig. 1). There was also an increase in body weight upon HFS feeding (mean gain of 1.7±0.2 g in HFS vs 0.2±0.1 in SD group). The administration of metformin was able to counteract these alterations, as reflected by the maintenance of normal body weight gain (mean gain of 0.1±0.1 g) and the preservation of glucose tolerance, although the effect of metformin on fasting glucose levels did not reach statistical significance (Fig. 1).

Short metformin treatment restored the expression of important genes in intestinal regions.
The different intestinal segments express specific patterns of genes allowing a spatial organization of a number of functions along the intestinal tract, such as nutrient digestion and absorption, chylomicron production, bile acid uptake, gut hormone secretion, or immune response and microbe defense. Therefore, we analyzed for each intestinal segment the expression of a specific set of genes after 8 days of HFS feeding with or without metformin. Regarding the genes coding for the apical sugar transporters (Fig. 2), namely the sodium-glucose cotransporter SGLT1(encoded by Slc5a1) and the fructose transporter GLUT5 (Slc2a5), we found that HFS diet increased GLUT5 and SGLT1 expression in the duodenum, the main region of sugar absorption, while metformin counteracted these effects and significantly reduced GLUT5 expression in all segments of the intestine. We also observed that metformin was able to reduce Glut2 expression in the duodenum (data not shown).
Jejunum and ileum are the favored sites for lipid absorption. As shown in Figure 3, HFS diet feeding significantly increased the gene expression of the main actors of apical fatty acid uptake (FAT/CD36 and FATP4), intracellular fatty acid transport (FA2PB2) and chylomicron synthesis (MTTP) in these two regions, while metformin restored their expression to basal levels (Fig. 3). In contrast, the ileal expression of the apical sodiumdependent bile acid transporter (ASBT, coded by Slc10a2) and of the organic solute transporter-alpha (OSTα, coded by SLC51B, which exports bile acid across the enterocyte basolateral membrane), was not significantly affected by HFS diet or by metformin (Fig.   3). The mRNA level of the cholesterol transporter NPC1L1 was strongly decreased upon HFS diet in the jejunum and not modified by metformin treatment (Fig.3).
Results regarding gut hormones and peptides are shown in Figure 4. We analyzed the expression of the glucagon gene (Gcg), encoding GLP-1, in different regions of the gut. HFS diet was associated with an increase in Gcg expression in duodenum and jejunum and metformin treatment tended to restore its expression to basal level in duodenum and to slightly decrease it in the colon (Fig.4). In the other regions, we did not find significant effect of metformin on Gcg gene expression. In contrast, the mRNA levels of gastric inhibitory polypeptide (GIP) and cholecystokinin (CCK) in duodenum, of neuropeptide Y (NPY) in jejunum and of fibroblast growth factor 15 (FGF15) in ileum, which were all increased upon HFS feeding, were restored or down-regulated in the presence of metformin (Fig. 4). Finally, the expression of markers of inflammation (IL1b and TNFa) in the colon was not significantly affected after 8 days of HFS diet or metformin treatment ( Fig. 4).

Metformin modified the bile acid pool
Bile acid profiling was performed on the caecum at the end of the treatment periods. As shown in Figure 5, 8 days of HFS diet significantly increased total bile acid levels. Even though metformin did not modify total bile acid levels compared the HFS group, the treatment tended to increase primary over secondary bile acid ratio and decreased the hydrophobicity index (Fig. 5).
More specifically, bile acid analysis revealed an increase in cholic acid (CA) levels and a reduction in its main secondary metabolite deoxycholic acid (DCA) upon metformin treatment, whereas the amount of 7-sulfocholic acid was not affected (MET vs HFS) (Fig.5). The murine bile acids, beta-muricholic acid (β-MCA) and omega-muricholic acid (w-MCA) as well as chenodeoxycholic acid (CDCA) were increased by HFS diet, but were not significantly affected by metformin compared to HFS (Fig. 5). Interestingly, the secondary metabolites of CDCA were significantly affected by metformin. We observed a significant increase in ursodeoxycholic acid (UDCA) and tauroursodeoxycholic acid (TUDCA) while lithocholic acid (LCA) levels were decreased by the treatment (Fig. 5). Of note, the relative abundance of conjugated bile acids (tauro-, or sulfo-conjugated) was not significantly affected by metformin (data not shown).

Metformin modified the luminal microbiota composition in all the regions of the intestine.
The number of operational taxonomic units (OTUs) measured in the luminal effluents in the duodenum, the jejunum and the ileum, was similar between groups, with a trend for a lower number of OTUs in the HFS diet group (p = 0.049 vs SD group), not restored by metformin treatment (Sup. Fig. S1). Despite the fact that the bacterial richness appeared globally not modified after 8 days of HFS diet or metformin treatment, we observed a significant impact of metformin at the phylum level in the different regions of the intestinal tract (Fig. 6). While animals fed the HFS diet displayed a slight increase in the proportion of Firmicutes in all segments, especially the duodenum, metformin treatment induced a spectacular increase in the abundance of Verrucomicrobia all along the intestinal tract (Fig. 6). Analysis performed at the genus level revealed that 60% of the sequenced bacteria belonging to this phylum corresponded to Akkermansia muciniphila, as shown in Figure 7. Furthermore, additional analyses indicated that HFS diet and metformin treatment were associated with important modifications at the family and genus levels ( Fig. 8 and Sup. Fig. S2). We observed an increase abundance of Clostridium (belonging to the Clostridiaceae family) in all the region of small intestine upon HFS feeding, which was significantly prevented by metformin (Sup. Fig. S2). This effect of metformin treatment on Clostridium was also found in the colon, despite lower abundance of these bacteria in this region (Sup. Fig. S2). Bacteria belonging to the Lachnospiraceae family showed similar trend as Clostridium, especially the genus Dorea, which was increased by HFS diet and restored by metformin in all intestinal segments (Sup. Fig S2).
Furthermore, in the lumen of the 3 sections of the small intestine, bacteria belonging to the Peptostreptococcacae family also showed similar trend to Clostridium, with metformin reducing their abundance (Sup. Fig. S3).
In contrast, metformin treatment tended to increase Adlercreutzia abundance in the jejunum (Sup. Fig. S3). We also observed a positive effect of metformin on Propionibacterium (from the Actinobacteria phylum) in the duodenum (Sup. Fig. S3).
Finally, there was a marked reduction of the Muribaculaceae family (S24-7) in response to HFS diet in the colon and metformin tended to further decrease the abundance of these bacteria (Sup. Fig. S3).

DISCUSSION
The mechanism of action of the antidiabetic drug metformin is still a matter of discussions although it is now widely accepted that the gut plays an important role 1 . Despite major advances, there are still uncertainties regarding the effects of metformin in the intestinal tract, with one major difficulty consisting in evaluating which of the observed changes are direct effects of the drug. Several reports have evidenced modifications of intestinal functions within hours after metformin administration, such as on glucose transport 16,17 or after only few days, like on gut bile acid profile or on faecal microbiota 18  Beside A. muciniphila, additional specific bacteria were recently proposed as potential mediator of metformin action, but not found in the present study, such as Lactobacillus members, that are increased after 1 day of metformin treatment in rat duodenum 16

, and
Bacteroides fragilis which is decreased after 3 days of metformin in the stools of T2DM patients and that may regulate FXR signalling through the production of the bile acid GUDCA 18 .
The involvement of bile acids in the mechanism of action of metformin has recently emerged 11,18,22 . However, the effects of metformin on bile acid pool composition in humans are still rather controversial, with one study showing modifications after 3 days of treatment 18 while another did not find significant difference after 4 months 13 . Less is known in animal models and we found here that metformin did not significantly modify total bile acid levels in the caecum of the HFS diet fed mice, but altered their composition.
The most striking effect of metformin was a marked decrease in the levels of secondary bile acids DCA and LCA whereas UDCA and TUDCA were increased. DCA and LCA can be produced by gut bacterial 7a-dehydroxylation (7aDeOH) of the primary bile acids CA and CDCA. The predominant intestinal species exhibiting 7aDeOH activity belong to the genus Clostridium 32,33 . The fact that metformin significantly decreased their abundance in all the sections of the gut, may potentially support the reduction in DCA and LCA levels.
Ursodeoxycholic acid (UDCA) and its taurine conjugated TUDCA are of clinical interest due to their multiple beneficial effects on human health. CDCA can be converted to UDCA in the colon through epimerization by microbial 7α-and 7β-hydroxysteroid dehydrogenases. The bacteria bearing 7α/β-hydroxysteroid dehydrogenase activities are less characterized, but members of the genera Clostridium possess these activities 34 .
UDCA is then transported from the intestine to the liver through the enterohepatic circulation where it is conjugated with taurine or glycine to produce TUDCA or GUDCA, which are transported back into the gut. We found increased levels of UDCA and TUDCA in the caecum of metformin treated mice, while GUDCA was not detectable (in agreement with the fact that there are very low levels of glycoconjugated bile acids in rodent). Our data in mice agreed therefore with a recent study showing that TUDCA and GUDCA are the most induced bile acids in the stool of T2DM patients treated for 3 days with metformin 18 . These authors also showed that UDCA, TUDCA and GUDCA are inhibitors of the nuclear receptor FXR 18 . Interestingly, DCA and LCA, which were increased in the present study, are also potential inhibitors of FXR 35 . Another study demonstrated that metformin could also directly inhibit FXR activity via AMPK activation 36 . Supporting an inhibition of FXR activity in metformin treated mice, we found a significant reduction in the expression level of FGF15, a major target gene of FXR in the ileum. Therefore, the contribution of FXR signalling in the mechanism of action of metformin, directly or via bile acid modifications in the gut appeared, to be an important mechanism to take in to consideration.
Altogether the presented data demonstrate that metformin administrated during 8 days is able to counteract important perturbations induced by HFS diet in the different regions of the intestinal tract, including altered expression of key genes of nutrient absorption and increased abundances of some gut bacteria, which have been associated with metabolic disturbances (f-Lachnospiraceae, f-Petostreptococcaceae, g-Clostidium).
In addition, metformin promotes a strong increase in the abundance of A. muciniphila all along the gut, and beneficial modification of secondary bile acid profile in the caecum, with reduction in DCA and LCA levels and increased abundance of UDCA and TUDCA, potentially leading to FRX inhibition. It will be now important to determine how these different events are interconnected and triggered in a complementary way by metformin. Body weight was monitored at day 0 and day 8. For glucose tolerance test (ipGTT), mice were fasted for 6 hours then received an intraperitoneal injection of glucose (2 g/kg body weight). Blood glucose was monitored at different time points during 90 minutes at the tip of the tail, using a glucometer (Accu-Check, Roche).

Intestinal tissue sampling.
Animals were killed by cervical dislocation after 6h of fasting and 5h after the last gavage with metformin. Different parts of the intestinal tract (duodenum, jejunum, ileum and the colon from medium part to anus) were sampled. The caecum was also removed and the intestinal segments were cleaned by flushing ice-cold PBS and dipped in liquid nitrogen.

Gene expression analyses in intestinal segments.
Total RNA was extracted with TRI Reagent Solution (Sigma). Target mRNA levels were measured by reverse transcription followed by real-time PCR using a Rotor-Gene (Qiagen). A standard curve was systematically generated with different amounts of purified target cDNA, and each assay was performed in duplicate and normalized using TATA-binding protein (TBP) mRNA level, as previously reported 37 . The list of the target genes with the PCR primers used for the qPCR assays is in Supplementary Table S1.

Bile acid profiling in ceacum.
Bile acid molecular species concentrations were measured by HPLC coupled to tandem mass spectrometry (HPLC-MS/MS) as previously described 38 . Results were expressed in nmol/g of dried caecum.