In Vitro Gut Microbiome Response to Carbohydrate Supplementation is Minimally Affected by a Sudden Change in Diet

Background: Interactions between diet, stress and the gut microbiome are of interest as a means to modulate health and performance. Here, in vitro fermentation was used to explore the effects of a sudden change in diet, 21 days sole sustenance on the Meal, Ready-to-Eat (MRE) U.S. military combat ration, on inter-species competition and functional potential of the human gut microbiota. Human fecal samples collected before and after MRE intervention or consuming a habitual diet (HAB) were introduced to nutrient-rich media supplemented with starch for in vitro fermentation under ascending colon conditions. 16S rRNA amplicon and Whole Genome Sequencing (WGS) were used to measure community composition and functional potential. Specic statistical analyses were implemented to detect changes in relative abundance from taxa, genes and pathways. Results: Differential changes in relative abundance of ve taxa, Dorea spp., Akkermansia muciniphila, Prevotella spp., Desulfovibrio spp., and Dialister spp., and four Carbohydrate-Active Enzymes, specically GH13_14, over the 24 h fermentation were observed as a function of the diet intervention and correlated to specic taxa of interest. Conclusions: These ndings suggest that consuming MRE for 21 days minimally effects changes in gut microbiota structure in response to carbohydrate, but may induce alterations in metabolic capacity. Additionally, these ndings demonstrate the potential of starch as a candidate supplemental strategy to functionally modulate specic gut commensals during stress-induced states. These results taxa The ndings demonstrate the value of combining human microbiome studies, in vitro fermentation, and powerful next generation sequencing techniques like 16S and WGS, to effectively gain a more complete understanding of the effects of stress on competitive nutrient:microbiome:interactions and to identify potential strategies toward modulating gut commensal metabolic competition during stress states. (one-way ANCOVA). R and SigmaStat software was used for analysis.

microbiota therapeutic interventions using diet as the mechanism to understand the cardioprotective effects on heart failure [4]. Haak et al. reviewed the potential bene ts of targeting the microbiota to treat sepsis [5]. Our group recently reported that a sudden diet shift from consuming a habitual diet to consuming the US military Meal, Ready-to-Eat (MRE) combat ration altered fecal microbiota composition, resulting in lower relative abundance of multiple genera of lactic acid bacteria (e.g. Lactobacillus, Lactococcus, Leuconostoc) and increased relative abundance of several saccharolytic genera (Streptococcus and Clostridium) [6]. An in vitro fermentation experiment was also used to assess the potential use of carbohydrate, namely resistant starch (RS2) for restoring Lactobacillus following MRE consumption. That approach allowed inter-species competition in samples collected before and after the MRE intervention and at the same time points in a control group to be studied over the time scale of hours, which is not feasible in in vivo human studies which commonly rely on daily stool samples. Ruminococcus bromii, a keystone taxa and resistant starch degrader, increased in relative abundance during the MRE diet in the presence of RS2 while the ability of Lactobacillus to compete in presence of RS2 appeared to be reduced [7]. However, those results were limited in that only a few selected taxa were measured, the identity of which Lactobacillus species affected could not be elucidated, and differences in functional capacity of the community could not be examined nor the whole community metabolic response to RS both compositionally and functionally. The results did provide initial insight into how in vitro studies can complement human study results and reveal microbial community functional understanding in response to stress.
Herein we report a comprehensive genomic analysis employing both 16S rRNA gene amplicon and whole genome sequencing (WGS) of samples collected during the in vitro fermentation experiment described in Pantoja-Feliciano et al 2019 [7] to reveal in uence of a sudden change in diet on whole bacterial community composition and functional potential. To explore stress-induced microbial community responses, carbohydrate content, speci cally RS, in medium was increased ve-fold to allow the study of nutrient:microbiome interactions that cannot easily be explored in vivo [7].

Results:
Stressor-induced changes to microbial composition: Random Forest analysis was used to identify a subset of the community driving differences in community composition (Figure 3). For this analysis, features that were present in less than half of the samples in all diet-day combinations at 0 h (start of fermentation) were removed and the fermentation time points data (0, 5, 10 and 24 h) were combined, resulting in an analysis as a function of the diet and study day.15% of the genera (30 out 201 total) were identi ed for all the groups, but a differentiation by taxa and diet groups was not obtained. For that, a linear mixed model analysis using a multivariate ANOVA with repeated measures on the identi ed 30 taxa was employed to detect abundance variations as a function of the MRE-diet intervention (Diet*StudyDay*Fermentation Time interaction).
Out of the remaining taxa derived from Random Forest analysis, ve taxa, Dorea spp., Akkermansia muciniphila, Prevotella spp., Desulfovibrio spp., and Dialister spp., were identi ed as having a statistically signi cant interaction between diet, study day, and fermentation residence time (p-values 0.0061, 0.0147, 0.0004, 0.0198, 0.0002 respectively), (Figure 4). Dorea spp. notably increases in the MRE day 21 group after 10 h of exposure to starch-supplemented medium relative to the other groups ( Figure 4A). Akkermansia muciniphila was diminished in relative abundance in MRE day 21 compared to HAB diet day 0 and 21 and MRE day 0 throughout the fermentation; however, the rate of change over the course of fermentation differed in MRE day 21 compared to MRE day 0 (p<0.001) as determined by an equality of slopes test ( Figure 4B, Figure S4B). A similar case was observed for Prevotella spp. where abundance at inoculation was higher but decreased as the fermentation proceeded ( Figure 4C); however, differences between MRE day 0 and MRE day 21 were not detected at later time-points, and after 24 h residence time, this organism was completely diminished in both groups. Desulfovibrio spp. abundance was higher at the beginning of the fermentation in the MRE day 21 group relative to the other groups but decreased more rapidly during fermentation (p>0.001) such that no difference was observed between groups after 24 h ( Figure 4D, Figure S4D). Dialister in MRE 21 was signi cantly different than that of HAB 0, HAB 21, or MRE 0 at multiple time-points, and an equality of slopes test further indicated differential growth over the course of fermentation (p<0.001) ( Figure S4E). Because changes between HAB 0 and HAB 21 were also observed, we cannot conclusively state that Dialister's growth in MRE 21 is due to the all-MRE diet. ( Figure  4E). See supplementary Tables S1 and S2 for more details about the p-values and completed analysis.

Stressor-induced changes to Microbial Functional Potential
Whole genome sequencing was employed to complement 16S analyses and explore the in uence of MRE-diet intervention at a functional level. We obtained a total number of raw reads of 245 357 634, including both paired reads; a total number of QC reads of 245 000 244; a mean raw reads per sample of 3 774 732; and a mean QC reads per sample of 3 769 234.
Though the diet*study day interaction only resulted in subtle alterations in community composition as assessed by 16S sequencing, genomes for strains that have nearly identical ribosomal RNA sequences have been shown to possess different functional capabilities [8]. Using the assembly free program HUMAnN2 [9], we assessed community wide function and observed similar clustering patterns as those seen with 16S. PCoA of Bray-Curtis distances in gene family abundances ( Figure S5A To parse whether there were ner scale differences for speci c strains or functions, sequences from all samples were co-assembled, binned into metagenome assembled genomes (MAGs), and functionally annotated. After binning, we assembled 120 MAGs at >50% completion and <10% contamination including 57 MAGs at >90% completion and <5% contamination and 63 MAGs at >50% completion, <10% contamination. PCoA of Bray-Curtis distances for MAG abundances showed similar clustering by diet and fermentation residence time to the 16S PCoA ( Figure S6A). The rst principal coordinate was associated with fermentation time (PERMANOVA, R 2 = 0.27, p = 0.001) and the second with diet (PERMANOVA, R 2 = 0.23, p = 0.001). There was a small, signi cant effect of Diet*Date (PERMANOVA, R 2 = 0.03, p = 0.001). Thus, at the MAG level, there was not a large effect of MRE diet when comparing Day 0 to Day 21.
Due to the inclusion of starch supplementation during fermentation, we assessed whether speci c functions for complex carbohydrate breakdown (Carbohydrate-active enzymes, CAZymes) were affected by Diet*Date. Hidden Markov models were used to identify and classify CAZymes in the metagenome assembly. PCoA of Bray-Curtis distances for CAZyme abundances showed similar clustering patterns to the pathway PCoA ( Figure S6B). The rst principal component was primarily associated with fermentation time (PERMANOVA, R 2 = 0.57, p = 0.001) with small effects of Diet (PERMANOVA, R 2 = 0.06, p = 0.001) and Diet*Date (PERMANOVA, R 2 = 0.03, p = 0.001). To uncover CAZymes which were important for classifying the Diet*Date categories, we employed Random Forests and linear mixed models ( Figure 5). Four CAZymes passed the signi cance threshold: GH13_14 (pullulanase), GT76 (Dol-P-Man: α-1, 6-mannosyltransferase), CBM83 (starch binding), and CBM27 (mannan binding) (Table S3 and Figure 5B). Table S4 shows the p-values corresponding to the pairwise multiple comparison analysis Tukey HSD for each fermentation time point in the different groups, supporting Figure 5B. In the case of GH13_14, MRE day 21 group is signi cantly different from the other groups at 10 and 24 h after fermentation. GH13_14 was of particular interest because these enzymes catalyze the cleavage of branched RS2 breakdown products. MAG and taxonomic breakdown of GH13_14 by Diet*Date indicated that the increased abundance in MRE Day 21 samples was due to a Coproccocus comes MAG ( Figure  6A). Another interesting case was the CAZyme GT76 and its prevalence in the MRE day 21 group associated with Lachnospira eligens ( Figure 6B). Thus, the MRE diet did result in subtle functional differences at the ne-scale CAZyme level. Similar analysis was employed to nd pathways that were important for differentiating Diet*Date categories, but only minor effects of Diet*Date were evident ( Figure S7A, Table S5 and Figure S7B).

Discussion
This study used an in vitro fermentation system that simulated the conditions of the gut, to examine the effects of starch supplementation on gut microbial community composition and functional capacity in samples collected from volunteers that consumed two different diets, the Meal, Ready-to-Eat (MRE) U.S. military combat ration and a habitual diet (HAB) for 21 days. Only subtle changes in gut microbiota structure and metabolic capacity in response to RS2 were observed as a result of MRE diet consumption suggesting that the MRE diet does not substantially in uence competitive dynamics within the gut microbiome for the model substrate.
Several studies have identi ed changes in microbial communities due to RS consumption. Martinez et al has previously shown fecal microbiota composition changes after a RS2 diet in a human study, speci cally a signi cant increase in the Ruminococcus bromii and Eubacterium rectale proportions [10]. A more recent study reported gut microbiota changes in mice when introducing a RS diet. Researchers observed a signi cant increase in members of the Proteobacteria and Verrucromicrobia phyla correlated to an observed increase in anxiety-like behaviors within the animals [11]. These are contrary to our results in which Desulfovibrio and Akkermansia muciniphila decreased upon exposure to RS-supplemented medium. This discrepancy may be due to differences in study design, comparing results from human samples to an animal model results, fermentation conditions and interindividual variability of the volunteers.
Although ~ 200 taxa were identi ed by 16S rRNA sequencing, only 5 bacterial groups (out of 30 after Random Forest) showed differential changes in relative abundance during the fermentations as a function of the diet intervention. These groups, Prevotella spp., Desulfovibrio spp., Dorea spp., Dialister spp, and Akkermansia muciniphila, are different than those identi ed in the previously reported qPCR analysis from the equivalent sample set [7]. Reasons for the inconsistency are unknown, but may include the use of 16S and metagenomic analysis in this study rather than using qPCR to targeted speci c species as was done in the previous study. The ve taxa identi ed by 16S in the present analysis, however, represent gut commensals that have a range of clinical and physiological relevance. For two of them, Dorea spp. and Dialister spp., inter-species competition for RS were signi cantly altered following 21d on the MRE diet. Dorea spp are members of the Clostridium cluster XIVa and are a dominant species in the human gut [12]. Dorea spp utilizes dietary carbohydrates such as simple sugars (eg, glucose, lactose, maltose), inulin and fructo-oligosaccharides to produce metabolic products including acetate, formate, lactate and ethanol [13]. Although some species are unable to directly metabolize starch, it is associated with starch absorption in mice, perhaps through cross-feeding on small hydrolysis products (e.g., maltose, glucose) from initial starch degradation by starch-degrading taxa [14], including possibly Prevotella whose abundance was increased due to the MRE-diet relative to the HAB diet. Dorea is considered part of a healthy gut microbiota, although it has also been shown to have increased abundance in Multiple Sclerosis and IBD patients [15]. Dialister is composed of a small number of species, with only Dialister succinatiphilus has been isolated from feces [16]. Dialister relative abundance is reportedly increased by consuming walnuts [17], almonds [18] and whole grains [19]. These effects may be bene cial to health as Dialister produces propionate [20] and increases in relative abundance has been associated with reduction of IL-6 levels in response to whole grains [19] and lower incidence of depression [21]. However, increased abundance has also been associated with obesity and inability to lose weight [22], gastrointestinal disease (IBS, UC, CD) [23] and Alzheimer's [24]. Taken together, it is not clear whether the effect of the MRE diet on Dorea and Dialister responses to RS2 we observed in this study can be considered bene cial for the microbial community or host. For three other taxa, Prevotella spp., Desulfovibrio spp., and Akkermansia muciniphila, relative abundance differed in the MRE-day 21 samples at the 0h time point, but changes over time during the fermentation did not differ as a function of diet. This suggests an effect of the MRE diet on those taxa, but not the response of those taxa to RS2.
Though that result does not match the previous 16S compositional analysis from the human study [6], this is likely due to only a sub-population (n = 5 volunteers per group) being used in the in vitro experiment rather than all 30 study volunteers. The subtle distinctions within the ndings are di cult to draw conclusions associated with these taxa. Prevotella spp. is thought to be bene cial due to its prevalence in a high ber diets and has also been shown at the family level to increase following high RS diet [25] and during in vitro fermentation studies [26], [27]. A. muciniphila has been previously associated with bene cial health outcomes [28]; [29] and has been shown to metabolically respond to high RS diet in rats [30] and humans [25], while Desulfovibrio spp. has been associated with intestinal in ammation [31]. These taxa warrant further investigation as bacterial targets for RS supplementation.
WGS analysis identi ed two CAZymes, GH13_14 and GT76, that differentially changed in response to RS following MRE consumption relative to other samples. Enrichment of the extracellular glycan-active enzyme glycoside hydrolase (GH13_14) was due to Coprococcus comes, a member of the Clostridium cluster XIVa [32]. GH13_14 is a pullulanase common in human gut lactobacilli. As a butyrate producer, Coprococcus comes is generally thought to be bene cial. It has also been negatively correlated in type 1 diabetes patients [33]. Maier et al has shown CAZymes and transport systems related to C. comes have increased in abundance in response to an RS diet [25]. Lachnospira eligens was related to the CAZyme GT76. L. eligens utilizes pectin and polygalacturonic acid, with acetate, formate, ethanol, and CO2 as major end products [34]. It has been associated with the glycosyltransferase GT76, a α-1, 6mannosyltransferase that uses dolichol-P-mannose as a sugar donor. GT's are enzymes that forms glycosidic bonds and are involved with biosynthesis of di-, oligo-, and polysaccharides (www.cazy.com). Both C. comes and L eligens were not directly identi ed in the 16S rRNA analysis at the species level but the genus Coprococcus and Lachnospira were part of the 30 most abundant taxa. These groups were not signi cant in the random forest analysis, which may suggest that not all species within each genus respond in the same way and also highlights the value of a higher level of resolution provided by WGS compositional and functional analysis. Otherwise, these taxa signi cantly contributed to CAZyme alterations and community carbohydrate metabolism in the presence of RS2. For CAZymes CBM83 and CBM27, the 3-way interactions analysis are being driven by an unexpected response in HAB d21. Therefore, it is not clear that these responses are truly a result of MRE consumption.
The study was limited by use of only a subset of volunteers within the in vitro fermentation studies pooling the samples limited individualized study outcomes and a lack of correlative SCFA, metabolomics and proteomics analysis to corroborate bioinformatics ndings. However, the data does demonstrate sudden changes in diet had a functional effect on community competition for RS and that certain potentially bene cial taxa respond to RS supplementation differentially as a function of diet and stress.

Conclusion
In this study, we used in vitro fermentation to explore the effects of an acute stressor, a sudden change in diet from habitual to sole sustenance on MREs, on inter-species competition dynamics of gut microbiome in response to starch supplementation. The main nding of this study was that MRE consumption does not appear to substantially impact the effects of RS2 on the gut microbiome. Rather, only minimal alterations in community composition and functional potential as measured by CAZyme relative abundance were observed. These results did demonstrate community metabolic capacity and competition for substrates can be altered even when taxa abundance are not signi cantly different in the absence of that substrate. The ndings demonstrate the value of combining human microbiome studies, in vitro fermentation, and powerful next generation sequencing techniques like 16S and WGS, to effectively gain a more complete understanding of the effects of stress on competitive nutrient:microbiome:interactions and to identify potential strategies toward modulating gut commensal metabolic competition during stress states.

Participants
Fecal samples were collected from ten individuals participating in a randomized controlled trial designed to determine the effects of subsisting on a MRE-only diet on gut microbiota composition and intestinal permeability [6]. For more information about the characteristics of the participants and their diets see Supplemental Table 1 from Pantoja-Feliciano et al 2019. Study details and primary ndings have been previously reported . Brie y, the full study population included 64 adults without obesity, 18-62 year who were randomly assigned to follow their normal habitual diet for 21d (HAB) or consume a provided diet containing only the MRE rations for 21d. Study exclusion criteria included: use of antibiotics or colonoscopy within 3 mo of enrollment, vegetarian diet, history of gastrointestinal (GI) disease, infrequent bowel movements (< 4x/wk), and habitual use of medications affecting GI function (e.g. laxatives, anti-diarrheals). All participants were instructed to discontinue use of probiotic, prebiotic, or other dietary supplements ≥ 2 wk before study participation. Study involvement was voluntary, and written informed consent was obtained prior to enrollment. The study was reviewed and approved by the US Army Research Institute of Environmental Medicine Human Institutional Review Board (Natick, MA). Investigators adhered to the policies regarding the protection of human subjects as prescribed in Army Regulation 70 − 25, and the research was conducted in adherence with the provisions of 32 CFR Part 219.
The trial is registered on www.clinicaltrials.gov as NCT02423551.

Fecal Samples
Fecal samples were collected at baseline (day 0) and at the end of the 21d MRE intervention period (day 21). Samples were collected into provided 650 mL collection containers to which an anaerobic sachet (GasPak EZ Anaerobe Container System; Becton, Dickinson and Co., Franklin Lakes, NJ) was immediately added. The sealed container was then kept on ice or in a refrigerator until processing [35]. Fecal slurry (20%) was prepared within 12 h of donation by addition of 0.1M phosphate buffer pH 7.2 supplemented with 15% w/v glycerol and 0.08% L-cysteine (Sigma-Aldrich; St. Louis, MO), to fresh feces in a 4:1 ratio, followed by homogenizing for two minutes in a Seward Ltd. Model 400 stomacher (Davie, FL). The slurry was anaerobically divided into aliquots and stored at -80°C until needed.
Fermentation System Protocol (Scheme 1) All chemicals were obtained from Sigma-Aldrich unless otherwise indicated. The fermentation parameters are outlined in Pantoja-Feliciano et al [7]. Brie y, fermentation medium was prepared based on Macfarlane et al.[36] with the following modi cations: addition of resazurin (1 ug/L) and supplemented with a 5-fold increase in potato resistant starch (15g/L, RS). After mixing well, the nutrientrich medium was added to fermentation vessels (125 mL/vessel) autoclaved, equilibrated overnight under constant headspace ush with oxygen-free N 2 (20psig, 5 mL/min) and adjusted to emulate the ascending colon (pH 5.5). Fecal samples collected from ten individuals participating in the parent study on day 0 and day 21 (HAB, n = 5; MRE, n = 5) were pooled in equal proportions and vessels inoculated with 10% ( v / v ) fecal slurry for nal 2% inocula ( w / v ). Pooling promotes a highly diverse community and allows incorporation of low abundant, keystone species that may be limited using individualized samples to generate more generalizable insight. Parallel control vessels were inoculated with cell-free phosphate buffer/glycerol. Aliquots were temporally removed from each vessel at 0, 5, 10, 24 and 48h after exposure to RS2-supplemented medium and stored at -80 o C for DNA extraction and sequencing analysis. Fermentations were run in triplicate, at the same time, as experimental replicates.
16S rRNA gene amplicon sequencing DNA from fecal samples was extracted using the QIAMP Power Fecal DNA Extraction Kit, QIAGEN, Inc.

Whole-genome sequencing
Metagenomic assembly and binning was completed with metaWRAP pipeline modules [44]. Default parameters were used unless noted. The module metawrap read-qc was used to quality lter reads for each sample. Paired-end reads for all samples were combined and co-assembled with the metawrap assembly module using MEGAHIT [45,46]. Assembled contigs were binned with the metawrap binning module using MetaBAT2, MaxBin2, and CONCOCT programs [47][48][49]. Bins were consolidated and re ned with the metawrap bin_re nement module with a minimum completion of 50% and maximum contamination of 10%. Bin abundances across samples were quanti ed with the metawrap quant_bins module which uses Salmon [50]. To improve assemblies, the re ned bins were reassembled with the metawrap reassemble_bins module with a minimum completion of 50% and maximum contamination of 10%. Reassembled bins were functionally annotated with the metawrap annotate module which uses Prokka [51]. To search the metagenome assembly for CAZymes, the run_dbcan.py (https://github.com/linnabrown/run_dbcan) script was used. This script uses hidden Markov models (HMM) to search for CAZyme boundaries according to the dbCAN CAZyme domain HMM database [52,53]. Finally, bin taxonomy was assigned according to The Genome Taxonomy Database (GTDB) [54]. First, the reassembled bins were converted into contig databases with Anvi'o (anvi-script-reformat fasta, anvi-gen-contigs-database programs) [55]. Single-copy core gene taxonomy search databases were setup with the anvi-scg-databases program and taxonomy was estimated using the anvi-run-scg-taxonomy and anvi-estimate-genome-taxonomy programs. For functional and taxonomic analyses independent of metagenome assembly, HUMAnN2 [9] and MetaPhlan2 [56] were used to annotate gene families/pathways and taxonomy, respectively.
The combination of both techniques (16S and WGS) can contribute to comprehending the differences within and between individuals/samples [57]. By employing 16S analysis, community composition changes were explored while with whole genome sequencing, functional capacity of the community in terms of genes and pathways were examined. The advantage of using 16S for the taxonomic identi cation is the large and comprehensive availability of reference databases leading to more accurate results [58]. Both techniques can be used as a complement of each other as they provide powerful combined information.

Data analysis
Custom R [59] scripts were used for analysis and visualization of 16S and whole genome data. For principal coordinates analysis (PCoA), the vegan package (https://github.com/vegandevs/vegan/) was used to compute Bray-Curtis distances between samples and this distance matrix was input into PCoA with the labdsv package (http://ecology.msu.montana.edu/labdsv/R). Experimental factor signi cance and proportion of variance explained was determined by PERMANOVA with the adonis function using the Bray-Curtis distance matrix. For Random Forest analysis, features that were present in less than half of the samples were removed. The data was baselined by subtracting feature abundance for each bioreactor vessel at time zero from feature abundances at subsequent time points. To determine importance and signi cance for Diet*Date features, the randomForest function (https://www.stat.berkeley.edu/~breiman/RandomForests/) was used with ntree = 10001 using data for fermentation time points 0, 5, 10 and 24 hours. RColorBrewer (https://cran.rproject.org/web/packages/RColorBrewer/index.html) and ggplot2 [60] were used for visualizations.

Statistical Analysis
Relative abundance data for the top 30 organisms, 30 pathways, and 30 CAZymes found to be important for differentiating diet*date by Random Forest analysis in R software were arcsine square root transformed to normalize distributions and analyzed using repeated measured ANOVA. Models included with "Diet" (MRE and HAB) and "Study day" (0 and 21) as between-groups factors and "Fermentation time" (0, 5, 10, 24 and 48h) as a within-subjects factor. Validity of repeated measures results for each organism was assessed by testing for sphericity. For organism data that passed the test for sphericity, pvalues of main effects as well as two-and three-way interactions were generated using an unadjusted univariate F-test; in cases where the test for sphericity failed, a Greenhouse-Geisser (G-G) epsilon adjusted F-test was used to generate p-values.
For all features demonstrating a statistically signi cant diet*study day*fermentation time interaction (p < 0.05), pairwise comparisons between groups were tested within each time point separately using ANOVA with Tukeys HSD. To assess whether signi cant differences were due to relative abundance at 0h, features were subjected to analysis of covariance with fermentation time as the covariate and tested for equality of slopes between groups (p < 0.05) (one-way ANCOVA). R and SigmaStat software was used for analysis.    Table S1. Pairwise multiple comparison (Tukeys HSD Analysis) for the 5 organism obtained after the Linear mixed model analysis within each time-point, is represent by symbols: a (*) symbol indicates a difference between study days 0 and 21 for the same diet, and a (^) symbol indicates a difference between MRE and habitual (HAB) diets for the same study day. One symbol indicates p ≤ 0.05, two symbols indicates p ≤ 0.01, and three symbols indicates p ≤ 0.001. Exact pvalues from the test are also reported in Table S2. symbol indicates a difference between study days 0 and 21 for the same diet, and a (^) symbol indicates a difference between MRE and habitual (HAB) diets for the same study day. One symbol indicates p ≤ 0.05, two symbols indicates p ≤ 0.01, and three symbols indicates p ≤ 0.001.

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download.