Dried FL flowers in powdered form were procured from the medical supply store of Dongguk University International Hospital (Goyang, Republic of Korea). FL was extracted from powdered FL (1.4 kg) after mixing with ethanol (6 L) at 100°C for 3 h using a reflux condenser. The collected liquid extract was evaporated to dryness by rotary evaporator (EYELA N-1200A, EYELA, Tokyo, Japan), freeze-dried using a lyophilizer (Bondiro, IlshinBioBase, Dongducheon, Republic of Korea), and kept at −80°C until further use. The yield of FL was 8.7% (w/w).

Chromatographic analysis of FL was performed using an HPLC system (1260 infinity, Agilent Technologies, Santa Clara, CA, USA) equipped with a UV detector, an online degasser, and an autosampler. The mobile phase A was water:o-phosphoric acid (99.5:0.5, v/v) and mobile phase B was acetonitrile:water:o-phosphoric acid (50:49.5:0.5, v/v). The separation was carried out through an Eclipse XDB-C18 column (5 μm, 250 × 4.6 mm. Agilent Technologies, USA) at 25°C maintaining a flow rate of mobile phase at 1 ml/min with the following gradient: 0–5 min, 90% A, 5–18 min 90–18% A for 18 min, and 18–90% A for 0.1 min). The detection was performed at 327 nm using chlorogenic acid and caffeic acids (Sigma Aldrich, St. Louis, MO, USA) as standards.

Total flavonoid content was analyzed by aluminum chloride reaction methods. FL extract was reacted with 5% NaNO₂ for 5 min and added to 10% AlCl₃ for 5 min. The reaction was stopped by adding 1 M NaOH and an absorbance value was measured at 570 nm on a microplate reader (Spectramax Plus, Molecular Devices, Sunnyvale, CA, USA). The Catechin (Sigma Aldrich, USA) was used as standard.

C2C12 cells were cultured in DMEM (Welgene, Daegu, Republic of Korea) supplemented with 1% FBS (HyClone, Logan, UT, USA) and 1U penicillin/streptomycin (Life Technologies, Carlsbad, CA, USA). INS-1 cells were cultured in RPMI 1640 (Gibco, Grand Island, NY, USA) with 10% FBS, 10 mM HEPES (Sigma Aldrich, USA), 1 mM sodium pyruvate (Life Technologies, USA), 50 μM 2-mercaptoethanol (Sigma Aldrich, USA), and 1U penicillin/streptomycin. The growing conditions of both cells were 37°C in a humidified atmosphere of 5% CO₂ in air.

After the confluence of C2C12 reached 90%, the medium was removed and replaced with the differentiation medium (DMEM supplemented with 20 mM glucose and 2% horse serum; Gibco, USA) followed by another change on day 2. The medium was changed on day 4 to DMEM supplemented with 2.5 mM glucose (Sigma Aldrich, USA). Two hours after this change, the cells were treated with 0.75 mM MET (Sigma Aldrich, USA) either alone or in combination with FL (50, 100, 200 μg/ml) for 15 h. Finally, the cells were treated with 0.3 mM 2-deoxy-2-[(7-nitro-2,1,3-benzoxadiazol-4-yl)amino]-D-glucose (2-NBDG, Life Technologies, USA) for 1.5 h. The cells were then washed twice with DPBS. The glucose uptake was measured by detecting the fluorescence intensity of 2-NBDG at excitation and emission wave lengths of 485 and 535 nm, respectively, using a microplate reader (Molecular Devices, USA).

After a 24 h culture of the INS-1 cells, the medium was replaced by 20 mM high glucose RPMI 1640 medium. The cells were treated with FL at 50, 100, and 200 μg/ml concentrations in absence or presence of 0.75 mM MET for 2 days. Following this, the cells were washed with modified KPBB-HEPES buffer (134 mM NaCl, 4.8 mM KCl, 1 mM CaCl₂, 1.2 mM MgSO₄, 1.2 mM KH₂PO₄, 5 mM NaHCO₃, 10 mM HEPES, and 1 mg/mL BSA; pH 7.4) and then incubated in the same buffer supplemented with 20 mM glucose for 1.5 h. The supernatant was collected and centrifuged at 12,000 g for 10 min at 4°C and used for determination of insulin with an ELISA assay kit (Mercodia Rat Insulin Kit, Mercodia, Sweden). Cells were lysed in a lysis buffer [1% Triton-X, 20 mM HEPES, pH 7.4, 100 mM KCl, 2 mM EDTA, 1.0 mM PMSF, with protease inhibitors (Roche)]. Total protein of the cell extract was determined using a BCA protein assay kit (Thermo Scientific, Rockford, IL, USA).

Six-weeks-old male LETO and OLETF rats were purchased from Otsuka Pharmaceutical Co. (Tokushima, Japan) and acclimated for 6 weeks under 12 h light/dark cycle at constant temperature (25°C) and humidity (50–60%) with free access to chow diet and water. After acclimatization, OLETF rats were randomly divided into three groups (seven animals/per group) and treated as per the following specifications. MET group: treated with MET (100 mg/kg/day), FL+MET group: treated with FL (200 mg/kg/day) in combination with MET (100 mg/kg/day), and OLETF group: treated with water. The rats in LETO groups (seven animals) were also treated with water. The treatments were given orally (Food and Drug Administration, 2005) using water as the vehicle. The treatments were carried out six times a week for a period for 12 weeks. After the termination of experimental duration, all animals were subjected to overnight (12 h) fasting. The animals were sacrificed under Zoletil (Tiletamine-zolazepam, Virbac, Carros, France)-Rompun (xylazine-hydrochloride, Bayer, Leverkusen, Germany) combination anesthesia (1:1, v/v). Blood samples were collected from the central aorta and rapidly transferred into a BD Vacutainer (BD, Franklin Lakes, NJ, USA). The liver was excised quickly, washed in ice-cold PBS, pH 7.4, dried, weighed and stored at −80°C until used. For future histological analyses, portions of the liver were fixed immediately in 10% formalin (Junsei, Tokyo, Japan; for paraffin sectioning) and FSC 22® frozen section media (Leica Biosystem, Richmond, IL, USA; for frozen sectioning) and stored at room temperature and −20°C, respectively for further analyses. Fresh stool samples were also collected and stored at −80°C. The blood samples were allowed to clot for 2 h at room temperature and then centrifuged at 3,000 g for 15 min. The sera were separated and stored at −80°C for further analyses.

The animal study was approved by the Institutional Animal Care and Use Committee (IACUC-2014-037) of Dongguk University and performed in accordance with the “Guide for the Care and Use of Laboratory Animals” (Institute of Laboratory Animal Resources, Commission on Life Sciences, National Research Council, USA; National Academy Press: Washington D.C., 1996).

The animals were subjected to overnight (12 h) fasting and administered sterilized glucose solution (2 g/kg, Sigma Aldrich, USA) by oral gavage. The glucose levels of the blood samples collected from the tail vein were measured at five different time points (0, 0.5, 1, 1.5, and 2 h) using ACCU-CHEK Active (ACCU-CHEK, Mannheim, Germany). The IPITT was performed 2 days prior to the termination of animal experiment. The animals were subjected to overnight (12 h) fasting, then administrated with biosynthetic human insulin (0.75 U/kg, Eli Lilly and Company, Indianapolis, IN, USA; Um et al., 2004) by intraperitoneal injection. The glucose levels of the blood samples collected from the tail vein were measured at five different time points (0, 0.5, 1, 1.5, and 2 h) as described above for OGTT.

After embedding in paraffin blocks, formalin-fixed liver tissues were sectioned to 6 μm thickness on a microtome (Leica RM2235, Leica, Nussloch, Germany). The sections were placed on silicon coated glass slides (Leica Biosystem, Richmond, IL, USA), dried, deparaffinized with xylene and rehydrated in decreasing ethanol series. The sections were stained with 2.5% hematoxylin (Merck, Darmstadt, Germany) followed by counter-staining with 0.5% eosin (Sigma Aldrich, USA). The sections were dehydrated with an increasing ethanol series and xylene and mounted in a xylene-based mounting media (Histomount, Atlanta, GA, USA). Oil-Red-O staining was performed on liver tissues preserved in frozen section media as stated above. Freshly trimmed hydrated tissue sections were placed on silicon coated glass slides and fixed in 10% formalin for 5 min at room temperature. The sections were rinsed in running tap water for 10 min and immersed in 100% propylene glycon (Samchun, Gyeonggi-do, Republic of Korea) with two changes, then finally stained for 15 min with Oil-Red-O stain solution (Sigma Aldrich, USA; prepared in propylene glycon). The sections were rinsed in 85% propylene glycon for 3 min, rinsed in distilled water, stained in hematoxylin (Sigma Aldrich, USA), and eosin (Sigma Aldrich, USA) (H&E), and rinsed in distilled water for 5 min. The sections were mounted in aqueous glycerol (Samchun, Repubilc of Korea. Stained tissue sections were examined under an Olympus BX61 light microscope (Olympus, Tokyo, Japan) with 200X magnification. Images were captured (100 μm scale bar) using an Olympus DP70 digital camera (Olympus, Japan).

The serum triglyceride (TG), total cholesterol (TC), high-density lipoprotein cholesterol (HDL-C), glutamic oxaloacetic transaminase (GOT), and glutamic pyruvate transaminase (GPT) levels were determined using commercial enzymatic assay kits (Asan Pharmaceutical Co., Seoul, Republic of Korea) according to the kit manufacturer's instructions. The concentration of low-density lipoprotein cholesterol (LDL-C) was calculated using Friedewald formula, TC-HDL-(TG/5) (Friedewald et al., 1972). The serum insulin level was measured using a rat insulin ELISA kit (Mercodia, Uppsala, Sweden) as per kit manufacturer's protocol.

Stored liver tissues (200 mg) were homogenized in 1 ml of distilled water. Stored fecal samples were freeze-dried using a lyophilizer (IlshinBioBase, Dongducheon, Republic of Korea) and dried fecal matter (100 mg) was homogenized in 1 ml distilled water. The homogenate was extracted with 5 mL chloroform-methanol (2:1) mixture and centrifuged at 7,000 g for 5 min. The chloroform layer was taken, dried, and resolved by isopropanol. Liver TG and TC levels were determined using commercial enzymatic assay kits (Asan Pharmaceutical Co., Seoul, Republic of Korea) as per instructions of kit manufacturer.

Liver tissues were homogenized using an Ultra-Turrax homogenizer (IKA, Staufen, Germany) in ice-cold 0.15 M KCl solution containing 1% phosphoric acid (Sigma Aldrich, USA). Thiobarbituric acid (Sigma Aldrich, USA) was added to the homogenate at a final concentration of 0.8% and the mixture was kept in a water bath at 90°C for 45 min. The mixture was placed on ice and centrifuged at 2,000 g for 15 min following which n-butanol (Sigma Aldrich, USA) was added. The supernatant was aspirated carefully and its absorbance was read at 535 nm on a microplate reader (Molecular Devices, USA). The hepatic MDA content was determined from a freshly prepared standard curve of 1,1,3,3,-tetraethoxypropane (Sigma Aldrich, USA) and normalized with the total tissue protein. Protein concentrations were determined using a BCA protein assay kit (Thermo Scientific, USA).

Endotoxin levels in serum and fecal samples were determined using a Limulus Amebocyte Lysate (LAL) kit (ENDOSAFE, Charleston, SC, USA) according to the kit manufacture's protocol.

Stored liver tissues were homogenized in protein extraction solution (iNtRON Biotechnology, Inc., Gyeonggi-do, Republic of Korea) containing protease inhibitor (Sigma Aldrich, USA) and phosphatase inhibitor cocktail (GenDEPOT, Barker, TX, USA). Protein concentrations of the homogenates were determined using a BCA protein assay kit (Thermo Scientific, USA). Forty microgram of proteins was separated on a 10% SDS-PAGE gel provided with a constant voltage of 80 V, then transferred to a PVDF membrane (Amersham, Buckinghamshire, UK). The membranes were blocked with 5% skim milk (Becton Dickinson, Sparks, MD, USA) in Tris-buffered saline (Sigma Aldrich, USA). The membranes were incubated overnight with appropriate primary antibodies at 4°C. The immunoblots were washed three times in TBS buffer and then incubated with appropriate secondary antibodies for 1 h. The immunoblots were washed twice. The immunoblotted proteins were visualized using an ECL Western bolting luminal reagent, and the band intensities were quantified using a LAS-3000 Chemiluminescence detection system (Fujifilm, Tokyo, Japan). Primary antibodies were: anti-phosphorylated(p)-AMPK, anti-AMPK (Cell signaling, Danvers MA, USA), anti-SREBP-1c (Abcam, Cambridge, UK), anti-p-ACC (Cell signaling, USA), anti-HMGCoA reductase (Merck, Darmstadt, Germany), and anti-β-actin (Santa Cruz Biotechnology, Santa Cruz, CA, USA); secondary antibodies used were: HRP-conjugated mouse and rabbit IgGs (Santa Cruz Biotechnology, Santa Cruz, CA, USA).

Genomic DNA from fecal samples was isolated using QIAamp stool DNA mini kit (QIAGEN, Hilden, Germany) following the kit manufacturer's instructions. The PCR of V1–V3 region of 16S rRNA gene sequences was performed using a C1000 Touch thermal cycler (Bio-Rad, Hercules, CA, USA). Amplicons were purified using a QIAquick PCR purification kit (QIAGEN, Germany). An equimolar concentration of each amplicon from different samples was pooled and purified using an Ampure bead kit (Agencourt Bioscience, Beverly, MA, USA) and quantified using a PicoGreen ds DNA assay kit (Invitrogen, Carlsbad, CA, USA). The mixed amplicons were amplified on sequencing beads by emulsion PCR. Sequencing reactions were performed using a Roche/454 GS Junior system (454 Life Sciences, Branford, CT, USA) according to the manufacturer's instructions. Raw sequence data are available in the European Nucleotide Archive (ENA) with accession numbers PRJEB19873.

Obtained sequence data were sorted by a unique barcode in the demultiplexing step, and low quality reads (average quality score <25 or read length <300 bp) were not considered for analysis. Sequences were assigned to the operational taxonomic units (OTUs; 97% identity; Greengenes database: http://greengenes.lbl.gov) followed by the selection of representative sequence using the Quantitative Insights into Microbial Ecology (QIIME) software package (Caporaso et al., 2010). Phylogenetic Investigation of Communities by Reconstruction of Unobserved States (PiCRUSt) was performed to identify functional genes in the sampled microbial community on the basis of the data in the Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway database (Langille et al., 2013). To identify taxa with differentiating relative abundance in the groups, the LEfSe analyses were performed using the web-based program (http://huttenhower.sph.harvard.edu/galaxy; Segata et al., 2011) for which the following conditions were applied: the alpha value of the factorial Kruskal-Wallis test among classes was set to <0.05 and the threshold on the logarithmic LDA score for discriminative feature was set to >2.0.

Data were expressed as means ± SD unless otherwise indicated. Statistical significance was calculated with one-way analysis of variance (ANOVA) followed by the Tukey post-hoc test. Significance was accepted at the level of p < 0.05, p < 0.01, or p < 0.001. Relationship strength between parameters was assessed using the two tailed Pearson's correlation test. The correlation was considered significant only when the absolute value of Pearson's correlation coefficient r was > 0.5.

HPLC analyses revealed that the acquisition rate of ethanol extract of FL was 13.41%, the chlorogenic acid content of FL was 29.8 μg/mg, and caffeic acid content was 2.3 μg/mg (Figure S1). The flavonoid content of FL analyzed by aluminum chloride reaction method was 109.3 μ g/mg.

Although the body weight of OLETF rats did not change significantly in response to the treatment with FL and MET combination (FL+MET; Figure 1A), both the absolute and relative weights of liver of the animals in FL+MET group, but not in MET group, were significantly lower compared to OLETF rats (Figures 1B,C).

As expected, H&E histological staining revealed the normal structural integrity of liver in LETO group and hepatic steatosis (arrows) in OLETF group (Figure 1D). However, the extent of steatosis in the liver of FL+MET group was much lower compared to OLETF group. Oil-Red-O staining revealed the absence of lipid droplets in LETO group but presence of those in OLETF group (arrow, Figure 1E). However, the hepatic lipid deposition in OLETF rats was reduced remarkably in response to the treatment with both MET and FL+MET groups.

The serum TC, LDL-C levels, and the liver TC content decreased significantly in OLETF rats upon treatment with FL+MET, but not MET (Figures 2A,C,D). The serum HDL-C level increased in OLETF rats when exposed to FL+MET or MET (Figure 2B), but not significantly. Treatment of OLETF rats with FL+MET, but not MET, significantly increased the fecal TC content (Figure 2E). The liver TG content was significantly decreased in OLETF rats when treated with either FL+MET or MET (Figure 2F).

To investigate the molecular mechanisms underlying the preventive action of MET and FL+MET against fat deposition, we measured the phosphorylation or expression levels of hepatic AMPK, SREBP-1c, ACC and HMGCoA reductase, the key proteins involved in lipid metabolism and homeostasis (Figure 2G). Our results showed that the phosphorylation of both AMPK and ACC increased substantially and expression of SREBP-1c moderately decreased in the liver of MET-treated OLETF rats upon co-exposure to FL. Treatment of OLETF rats with both FL and FL +MET reduced the expression of hepatic HMGCoA reductase.

We performed the intraperitoneal insulin tolerance test IPITT and OGTT oral glucose tolerance test in the OLETF rats to investigate the effects of MET or FL+MET. The results revealed a time-dependent decrement in fasting serum glucose level up to 1 h post-insulin treatment which in the OLETF rats was most pronounced in FL+MET group followed by MET group (Figure 3A). The area under curve (AUC) of IPITT glucose in OLETF rats decreased significantly in response to treatment with both MET and FL+MET, while AUC of FL+MET group was significantly lower than the MET group (Figure 3B). Our results on OGTT demonstrated a time-dependent decline in fasting serum glucose concentration in all experimental rats throughout the entire 2 h post-glucose administration period, more effectively in both the MET and FL+MET groups (Figure 3C). The AUC of OGTT glucose in OLETF rats insignificantly, but markedly, decreased in response to both MET and FL+MET (Figure 3D). At the time of sacrifice, the serum levels of both fasting insulin and glucose in the MET or FL+MET groups were significantly lower than the OLETF group (Figures 3E,F).

To verify the beneficial impact FL and MET in combination on the glucose homeostasis, we performed in vitro studies using C2C12 myotube and INS-1 cell lines as models. Our results showed that the glucose uptake by C2Cl2 cells treated with 0.75 mM metformin was enhanced significantly upon co-treatment with CL at 50–200 μ g/ml concentrations (Figure 4A). Insulin secretion by INS-1 cells exposed to 50, 100, or 200 μ g/ml of CL increased upon co-treatment with 0.75 mM metformin, significantly at 100 μ g/ml of CL (Figure 4B).

Treatment of OLETF rats with FL+MET significantly lowered the serum levels of GOT and GPT, the marker enzymes for liver damage and disease, and significantly reduced the hepatic level of MDA, a biomarker for oxidative stress. Exposure of OLETF rats to MET distinctly decreased the serum GOT and GPT level, however only GOT showed statistically significant result (p < 0.05), but did not produce any significant effect on the hepatic MDA content (Figures 5A–C).

Treatment of OLETF rat with FL+MET significantly decreased both the serum and fecal endotoxin levels (Figures 5D,E). Although the exposure of OLETF rats to MET obviously decreased both the serum and fecal endotoxin level (Figures 5D,E), only fecal endotoxin showed statistically significant result (p < 0.05).

To profile the effect of FL and FL+MET on gut microbial structure and composition of OLETF rats, 16S rRNA gene sequencing of the stool samples was performed. Our data revealed that the rarefaction per sample was minimal in the LETO group (Figure S2). According to the PCoA of species level composition, the LETO group was clearly separated from the OLETF, FL+MET, and MET groups (Figure 6A). Furthermore, the distance between the OLETF and FL+MET groups was greater than that between the OLETF and MET groups. The microbial composition for each group was further analyzed at the class level. The dominant classes were Bacteroidia, contributing to 63.5, 54.8, 48.9, and 53.3%; followed by Clostridia, contributing to 25.8, 22.1, 31.7, and 23.7% of the total fecal microbial population of LETO, OLETF, FL+MET, and MET groups, respectively (Figure 6B). Like classes, the microbial composition varied greatly among families. The dominant family was S24-7 in Bacteroidales contributing to 46.1, 32.1, 29.9, and 37.1%, followed by Prevotellaceae contributing to 6.2, 13.6, 11.2, and 10.14% of the total fecal microbial population of LETO, OLETF, FL+MET, and MET groups, respectively (Figure 6B). Heatmap analysis of 150 OTUs showed that FL+MET group was more closely related to LETO group than MET group (Figure 6C).

Next, linear discriminant analysis (LDA) effect size (LEfSe) package was used to determine the differentially relative abundant microbial taxa between the animal groups. The cladogram from the LEfSe results revealed that compared with the OLETF group, 10, 9, and 2 taxa were increased while 12, 6, and 9 taxa were decreased in LETO, FL+MET, and MET groups, respectively (Figure 7). Further, compared with OLETF group, 29, 16, and 8 OTUs were increased while 24, 4, and 7 OTUs were decreased in LETO, FL+MET, and MET groups, respectively (Figure S4). Collectively, these results indicate that the gut microbial composition was differentially modulated in OLETF rats in response to the treatment with MET and FL+MET.

The heatmap analyses assessed the association between the disease biomarkers and gut microbiota (Figure 8). OTU 337724, OTU 987, OTU 6943, and OTU 4053 were significantly negatively associated and OTU 1944, and OTU 350242 were significantly positively associated with body weight, liver weight, liver TG, and liver TC. OTU 4053 was strongly associated with the biomarkers (8 out of 12), demonstrating significant negative correlation with the body and liver weights, hepatic TG, serum LDL-C, serum TC, and serum insulin levels as well as significant positive correlation with hepatic TC and serum HDL-C levels. OTU 3543, OTU 6943, and OTU 4053 were significantly negatively correlated and OTU 433676, OTU 356144, and OTU 5387 were significantly positively correlated with the serum insulin level (Figure 8).

We inferred metagenome functional content using PiCRUSt which predicted several metabolic functions of the gut microbiota. There was no significant difference in the enrichment of metabolic pathways among the experimental animal groups (Figure 9A). However, our results showed that the relative abundance of genes associated with the transporter functional categories including ABC transporters increased more than 2% in OLETF rats in response to FL+MET treatment (Figure S5A). The relative abundance of type I diabetes mellitus associated genes among the bacterial genes related to human metabolic and neurodegenerative diseases were significantly lower in FL+MET group vs. OLETF group (Figure 9B). In metabolic function, most genes associated with glycan biosynthesis and metabolism category were reduced maximally in FL+MET group (Figures 9C,D). Further analysis of this category revealed that many other genes, such as, those associated with glycan degradation, lipopolysaccharide (LPS) biosynthesis, and proteins involved in LPS biosynthesis were significantly reduced in OLETF rats upon treatment with either MET or FL+MET (Figure 9D). Moreover, many genes related to streptomycin biosynthesis and the genes associated with the oxidative phosphorylation category in energy metabolism function were significantly lower in both MET and FL+MET groups compared with OLETF group (Figures S5B,C).

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.