The animal study was carried out in accordance with the EU directive 2010/63/EU and the Danish Animal Experimentation Act (LBK 1306 from 23/11/2007 with 2011 amendments), and was approved by the Animal Experiments Inspectorate, Ministry of Food, Denmark (license number: 2016-15-0201- 01078).
Göttingen Minipigs diet study
Twelve male Göttingen Minipigs (Ellegaard Göttingen Minipigs, Dalmose, Denmark) were castrated at 7 weeks of age and thereafter fed with either control diet (Minipig Expanded, Special Diets Service, Witham, United Kingdom) (n = 5), choline-deficient amino acid defined high fat diet (CDAHFD, Special Diets Service, Witham, United Kingdom) containing 1% cholesterol, 0.35% cholic acid, 30% fat (cocoa butter and milk fat), 0.1% methionine, no choline, and either 20% fructose or sucrose (CDAHFD-F n = 3, CDAHFD-S n = 4) (diet specification: supplementary Table S1). These diets were given for 8 weeks. The CDAHFD diets had the same energy content, and the sugar type did not significant influence the findings in the CDAHFD groups or the development of NASH , also suggested in the literature , therefore they were analyzed as one group: CDAHFD (n = 7). The pigs had ad libitum access to fresh drinking water and were fed twice daily with an access to food for 1 hour. The control group was fed according to internal standards (3–4% of their body weight) where the CDAHFD group was offered 75% of this amount of food (weight basis), i.e. an isocaloric amount relative to control . The CDAHFD group had decreased appetite in the beginning. However, this was corrected over a period of 3–4 weeks by adding different flavors and ending with banana cream flavor. The minipigs were group-housed according to their respective diets and had straw as with a room temperature maintained at 22–24 °C and lights were on between 6 a.m. and 7 p.m. The minipigs had access to a heating lamp and their body weights were monitored. At euthanasia the pigs were fasted overnight and deeply anesthetized using intramuscular injection of a mixture containing 0.25 mg/kg of butorphanol and 1.25 mg/kg of each of the following: tiletamine, zolazepam, ketamine, and xylazine before being exsanguinated. For detailed description of the NASH phenotype of these minipigs see Pedersen and colleagues (2020) .
Tissue and faecal sampling
Colon samples were collected at euthanasia from the colon spiral junction of the minipigs using tools that were sterilized between each animal by the use of 70% ethanol. Tissue samples from the liver were collected and snap frozen on dry ice before being transferred to - 80 °C for later processing. For further details see Pedersen and colleagues (2020) .
Liver fibrosis was determined as part of the phenotyping previously published by Pedersen and colleagues (2020) , using quantitative image analysis on picrosirius red stained liver sections.
At end study, as part of the phenotyping previously published by Pedersen and colleagues (2020) , a hematological analysis was on ethylenediaminetetraacetic acid (EDTA) full blood (Advia 2120i Hematology System, Siemens, Ballerup, Denmark) and a clinical chemistry analysis was performed on plasma (Advia 1800 Chemistry System). In EDTA plasma glucose, fructosamine, alanine aminotransferase (ALT) and glutamate dehydrogenase (GLDH) were measured, in addition to serum BA on a Cobas 6000® autoanalyzer (Roche Diagnos- tics GmbH, Mannheim, Germany). Insulin and glucagon were determined by Luminescent Oxygen Channeling Immunoassay as described by Pedersen and colleagues (2020)  using GLU 1F120 mAb conjugated acceptor beads and biotinylated GLU 2F7 mAb for glucagon.
Serum and faecal metabolite analysis
All serum and faecal samples were randomized prior to the sample preparation. Aliquots of 30 µL blood serum were subjected for protein precipitation using 400 µL methanol, containing internal standards (Sigma-Aldrich, Germany). After vortex mixing, incubating on ice for 30 min and centrifugation at 9400 x g for 3 min, 350 µL of supernatants were collected. The supernatants were then evaporated under gentle nitrogen flow to dryness after which a two-step derivatization was performed. First, 25 µL of 20 mg/mL MOX reagent in pyridine was added and samples were incubated at 45 °C for 60 minutes. Secondly, 25 µL MSTFA was added, and samples were incubated once again for 60 min at 45 °C. Retention index mixture (10 µg/mL n-alkanes) was added to each sample before the analysis on Agilent 7890B gas chromatograph (GC) coupled to 7200 triple quad time of flight mass spectrometer (Q-TOF/MS) instrument. Initial helium flow was set to 1.2 mL/min, increasing to 2.4 mL/min after 16 minutes. Oven temperature program was kept at 50 °C for 5 minutes, with 20 °C/min increase to 270 °C, and then 40 °C/min to final temperature 300 °C (5 min). Samples with injection volume of 1 µL and 100:1 split ratio were injected using PTV injector set to 70 °C, heated to 300 °C at 120 °C/min. Zebron ZB-SemiVolatiles column (20 m length, 0.18 mm inner diameter, 0.18 µm film thickness) (Phenomenex Inc., USA) was used to achieve a chromatographic separation. Electron ionization (EI) source was set to 250 °C, 70 eV and 35 µA emission with 3 minutes solvent delay. Quadrupole was kept at 150 °C having 1.5 mL/min N2 collision gas flow. The data was acquired at 55–650 amu mass range and 200 ms/spectrum acquisition time. Six-points calibration curves at 0.1–80 µg/mL range contained compounds of interest from Sigma-Aldrich, Germany.
SCFAs extraction in faecal material was processed by adding 1 mL of 5 mM aqueous NaOH containing internal standard to 100 mg aliquots. Sample was homogenized with a micro pestle and mixed for 10 min at a shaker (300 rpm). After shaking, the sample was centrifuged for 20 min at 13200 x g at 4 °C. 300 µl of MQ water, 500 µL propanol/pyridine mixture solvent (v/v = 3:2), 100 µL of propyl-chloroformate were added to 500 µL of faecal water obtained after centrifugation. The sample was vortexed and ultrasonicated for 1 min. After adding 300 µL of hexane, the sample was vortexed and centrifuged for 5 min at 2000 x g. 300 µL from the hexane layer was collected in a glass vial. 200 µL of retention index standards in hexane were added before analysis. Another aliquot of 100 mg faecal sample for dry weights determination were freeze-dried overnight at - 50 °C. Acquisition of BCFAs was done using an Agilent 7890A GC to an Agilent 5975C MS equipped with an electron ionization (EI) source (230 °C). GC separation was achieved using a DB-5MS capillary column, 30 m × 0.25 mm i.d. × 0.25 µm film thickness (Agilent Technologies, Atlanta, GA, USA). The oven temperature was as follows: 45 °C (4 min); 10 °C/min to 70 °C; 3 °C/min to 85 °C; 5 °C/min to 110 °C; 30 °C/min to 300 °C (5 min). 1 µL of samples were injected in splitl mode (split ratio 5:1) and carried out by carrier gas (helium) at 260 °C with a constant flow of 1.0 mL/min. The data was acquired in scan mode and mass range was between 50 and 300 amu (additional details is supplied in supplementary section).
Imidazole propionate analysis
The same protocol as in Koh and colleagues (2018)  was used was used to determine imidazole propionate plasma concentrations. In brief, imidazole propionate was quantified using ultra-performance liquid chromatography coupled to tandem mass spectrometry and serum samples were.
extracted using 3 volumes of ice-cold ice-cold acetonitrile in 1.5 ml polypropylene also containing internal standards. Derivatization to 5% hydrochloric acid in butanol was performed, samples were injected onto a C18 column (2.1 × 100 mm with 1.7 mm particles; Waters, Milford, MA) and separated using a gradient consisting of water with 0.1% formic acid and acetonitrile with 0.1% formic acid .
50 mg of snap-frozen liver tissue were used for RNA isolation as described in . Concentration, purity and integrity of the RNA samples were analyzed as described in . All samples had good concentration, acceptable purity and a RQI (RNA quality index) between 8.9–10 and were all accepted for further processing.
Subsequently, cDNA synthesis was done in duplicate for each RNA sample as described in . Two negative controls (without reverse transcriptase) were processed in parallel. The cDNA samples were diluted 16 times prior to quantitative real-time PCR (qPCR) and stored at - 80 °C.
Profiled genes were chosen based on imidazole propionate and GCGR signaling in the liver : RHEB, Mechanistic Target Of Rapamycin Kinase (MTOR), IRS1, IRS2, Glucagon Receptor (GCGR) and Gluconeogenic Enzyme Glucose-6-phosphatase (G6PC). Primers were designed using Primer 3 or primer-Blast (gene sequences: supplementary Table S2).
QPCR was performed in a CFX96™ Real-Time System (Bio-Rad) using SsoAdvanced Universal SYBR® green supermix (Bio-Rad) following manufacturer’s recommendations. PCR cycling condition were: 30 seconds 95 °C followed by 40 cycles of 10 seconds denaturation at 95 °C and 30 seconds 60 °C annealing/extension ending with a melting curve (for more detailed description of this section see supplementary section).
Measurement of pH
Measurement of pH in minipig colon samples was performed by homogenizing the samples in sterilized water in the ratio 1:3 and following using a calibrated pH meter (Consort multi-parameter analyzer C3041, Turnhout, Belgium).
DNA extraction, sequencing and pre-processing of raw data
Library Preparation and Sequencing: The bacterial community composition was determined by Illumina NextSeq-based high-throughput sequencing (HTS) of the 16S rRNA gene V3-region, according to Krych and colleagues. (2018)  and is described in detail in the supplementary materials and methods. Briefly, the amplified fragments with adapters and tags were purified and normalized using custom made beads, pooled, and subjected to 150 bp pair-ended Illumina NextSeq (V3 region 16S rRNA) sequencing. The raw dataset containing pair-end reads with corresponding quality scores were merged and trimmed, followed by de-replication, purging of chimeric reads, and constructing high quality (97% similarity level) Operational Toxonomic Unit (OTU) that and taxonomically assigned using sintax coupled to the EZtaxon 16S rRNA gene reference database. The sequencing dept. was on average 68,194 read per sample before filtering going down to 62,426 after filtering.
Sequencing data pre-processing: The dataset was purged for OTU’s which were detected below 0.005% across all samples, and normalization was accomplished using MetagenomeSeq v 1.32.0 based on Cumulative Sum Scaling algorithm .
Statistics, calculations and bioinformatic analysis
In general, to evaluate the difference between the two groups, students t-test or Mann-Whitney test were used and a p-value< 0.05 was considered statistically significant. The data set was evaluated for normality using QQ plots and Shapiro-Wilk test. The analysis was conducted in PRISM v.9.0 (GraphPad Software, San Diego, California USA, www.graphpad.com). Stepwise backwards regression with linear regression model (based on Akaike’s Informative Criterion) was used to predict glucagon level (dependent variable: glucagon, independent variable: Imidazole propionate and total amino acid level) and hepatic fibrosis level (dependent variable: hepatic fibrosis, independent variables: Ruminococcus genus and total bile acids, random = group). Fligner-Killeen’s test was used to test for homogeneity of variance. The analysis was conducted in R open-source statistical software v4.0.3.
Total serum non-branched chain amino acid concentration included measurable amino acids alanine, proline, glycine, serine, threonine, and aspartic acid (values below limit of detection (LOD) were set to LOD/2) and all other non-branched chain amino acid were below detection limit and set to LOD/2 (supplementary Table S3). For calculation of glucagon-alanine index the following formula was used: glucagon-alanine index = fasting serum glucagon [pmol/L] * fasting serum alanine [µmol/L] . Raw data from gene expression analysis was processed in Genex 6 (MultiD Analyses AB, Gothenburg, Sweden), log2 transformed data was extracted and analyzed using Student’s t-test including fdr adjustment for multiple testing using Benjamin, Krieger, and Yekutieli in PRISM v9.0.1. For conducting bioinformatic analysis R open-source statistical software v4.0.3 using packages Vegan v2.5.7 , PhyloSeq v1.34.0 , MetagenomeSeq v1.32.0 , GUniFrac v1.1 , and DAtest v2.7.17  were used. Bacterial differences between groups were detected using DAtest package (lli function) and Mann-Whitney. Differential clustering was evaluated using ANOSIM. For visualization of data PRISM v9.0.1 and GGplot2 v3.3.3  were used.