Six-weeks-old, adult male C57BL/6J mice, weighing 25 ± 3 g were procured from CSIR-Indian Institute of Integrative Medicine, Jammu, India. The animals were caged into individually ventilated cages (IVC) under standard laboratory temperature and humidity conditions and allowed free access to standard mouse chow pellet feed and water ad libitum. Both animal handling procedures and protocols were approved by the Institutional Animal Ethical Committee (IAEC) and carried out as per the guidelines of Committee for the Purpose of Control and Supervision of Experiments on Animals (CPCSEA).

C57BL/6J mice were randomly separated into three groups (n = 6).

Figure 1A illustrates the overall treatment regimen. Mice in control and WD group receive equal volume of CMC-Na solution orally to normalize the handling stress. The fecal samples of mice were collected every fourth week. The body weight and food intake of mice were measured weekly. At the end of the study, the animals were euthanized by CO₂ asphyxiation and the liver tissues were excised after the dissection of animals and stored for further analysis.

Serum triglyceride (TG), total cholesterol (TC), low-density lipoprotein cholesterol (LDL-C), and high-density lipoprotein cholesterol (HDL-C) levels were determined by colorimetry based automated bioanalyzer using designated colorimetric kits (ERBA, Transasia, India), as per manufacturer’s protocol.

The serum concentrations of insulin and inflammatory cytokines (IL-6 and TNF-α) were assessed using designated ELISA kits (RayBiotech, GA, USA), as per manufacturer’s instructions.

To determine the fecal bacterial endotoxin concentration, 200 mg of fecal samples were homogenized and further sonicated in 1 ml of PBS. The supernatants were collected and subjected to endotoxin estimation using commercially available chromogenic endotoxin ELISA kit (MyBioSource, CA, USA) following manufacturer’s instructions.

Formalin-fixed liver tissues were processed for histopathology as described by Chhimwal et al. (2020) and then sliced into 4 μM thick section to perform hematoxylin and eosin (H&E) and Picrosirius red staining following procedure reported previously (Chhimwal et al., 2020). Further, snap frozen liver specimens were embedded in OCT medium and sliced at 10 μm thickness with the help of cryotome (Lieca, Wetzlar, Germany). Followed by fixation in 10% buffered formalin for 10 min, and rinsed in water for 5 min. Fixed sections were then stained with freshly prepared 0.4% ORO staining solution for 10 min at RT followed by counter stain with Mayer’s hematoxylin and again washed with running water for 5 min. Stained liver sections were then envisioned under bright field in EVOS FL Auto 2 microscope (Thermo Fisher Scientific, MA, USA). MAFLD scoring was done according to previously described method (Chhimwal et al., 2021).

Total RNA was isolated from liver tissue using TRIzol TRI Reagent (Sigma, MO, USA). Briefly, 50 mg of liver tissue was homogenized in one ml of TRI reagent and the supernatant was collected by centrifugation at 12,000 g for 5 min at 4°C. To separate the aqueous layer containing nucleic acids, 0.2 ml chloroform was added per ml of collected supernatant and centrifuged (12,000 g; 15 min; 4°C). The upper aqueous layer was collected and an equal volume of 70% ethanol was added to it. The mixture was then passed through the column provided in RNeasy Mini Kit (Qiagen, Hilden, Germany) and further RNA was isolated using the manufacturer’s protocol. The quality and quantity of RNA samples were checked by NanoDrop (Thermo Fisher Scientific, MA, USA). RT-qPCR reactions were executed using RT-qPCR ROX Mix Kit supplied by Thermo Fisher Scientific. GAPDH was used as reference gene. Primers sequences (forward and reverse) are listed in Supplementary Table 1. All the data (analyzed by 2^–ΔΔCt method) were adjusted to reference gene and displayed as fold change versus the control group.

Total bacterial DNA was isolated from all 72 samples using QIAamp PowerFecal Pro DNA Kit (Qiagen, Hilden, Germany) according to manufacturer’s instructions. The quality and quantity of DNA samples were checked by Qubit 2.0 fluorometer (Thermo Fisher Scientific, MA, USA). The amplicon PCR was carried out using the primers 341F (5′-CCTACGGGNGGCWGCAG-3′) and 805R (5′-GACTACHVGGGTATCTAATCC-3′) to amplify the V3–V4 region of the bacterial 16S rRNA gene. Following the PCR purification, the Nextera XT indexed adapters (Illumina, CA, USA) were added to the ends of the primers. KAPA HiFi HotStart ReadyMix (2X) (Roche, Basel, Switzerland) was used for all PCR reactions. The purified libraries were then subjected to quality check on Qubit 2.0 fluorometer (Thermo Fisher Scientific, MA, USA) and Agilent 2100 Bioanalyzer system (Agilent, Singapore). The libraries were then sequenced and 250 bp paired-end reads were generated on the Illumina platform using MiSeq Reagent Kit v3 (600 cycle) (Illumina, CA, USA). By sequencing all the samples in a single MiSeq run, any potential batch effects were eliminated.

Computing and data processing was performed using the computing resource on Google cloud platform. Trimming of raw reads was done by remove primers using cutadapt (Martin, 2011) and low-quality bases (<Q30). These high-quality reads were considered for denoising, merging of paired-end reads, removal of chimeric sequences, and to pick unique amplicon sequence variants (ASVs) using the DADA2 plugin within qiime2 (Caporaso et al., 2010; Callahan et al., 2016). For taxonomic assignments of these ASVs, Silva database (sequences along with taxonomy, version silva-138-99) were downloaded (Quast et al., 2013). Further, unclassified ASVs from the same family, genus or species were differentiated using consecutive numbering (1, 2, 3, and so on). Selected sequences, extracted using V3–V4 primers, were used to train the naïve Bayes classifier and this trained classifier was used to predict the complete taxonomy of each ASV. This training and testing were done using feature-classifier fit-classifier-naive-bayes, and feature-classifier classify-sklearn plugins within qiime2. Furthermore, functional potential of these microbial communities was evaluated using Phylogenetic Investigation of Communities by Reconstruction of Unobserved States (PICRUSt, version 2; maximum nearest sequenced taxon index = 2) (Langille et al., 2013). Abundance of each ASV and predicted KEGG (kyoto encyclopedia of genes and genomes) metabolic pathways were used for downstream statistical analysis (Kanehisa and Goto, 2000).

One-way or two-way (wherever applicable) ANOVA was performed to analyze data using GraphPad Prism statistical analysis software (version 8.0.2), in order to determine the statistical significance of the differences among groups. Data are represented as mean ± SD (n = 6) and p-value < 0.05 was considered as significant. R statistical interface was used to perform all microbial community ecology analyses. Because of low sequencing depth in 2 out of total of 72 samples, 70 samples were considered for downstream analysis. ASV counts were used to access the microbial richness, diversity and ASVs relative abundances were used to calculate distances between each sample using vegan and ape packages in R (Paradis and Schliep, 2019). Bray–Curtis distances were used for the ordination analysis. Adonis function in R was used to calculate permutational multivariate analysis of variance (PERMANOVA) for each variable. To identify significantly discriminating taxa and pathways, “indicator species analysis” was performed using labdsv package and variance partition analysis performed using Variance Partition package in R (Hoffman and Schadt, 2016; Roberts et al., 2019). Indicator species analysis (ISA) is one of the most used statistical methods in microbial ecology. The benefit of using ISA is that it provides the features (taxa/pathways) which are significantly discriminating across groups (p-value) and at the same time their prevalence within the group (IndVal score). Cumulative abundances of selected taxa and functions were shown in different plots. Kruskal–Wallis test was applied to check the statistical significance of discriminating features (taxa/pathways). To understand the co-occurrence of microbial taxa among all three experimental groups, compositional corrected correlations were performed using Compositionality Corrected by REnormalization and PErmutation (CCREPE) function in R (Schwager et al., 2019). All strong (r > 0.5) significant correlations (p < 0.05) were used to construct the co-occurrence network plots in Cytoscape (Shannon et al., 2003). Neighborhood Connectivity was used to access the overall network topology.

Effects of WD on pathophysiology of animal and induction of MAFLD has been evaluated through different parameters. ORO staining in Figure 1B shows that WD-fed mice had significant micro and macro-vesicular fat deposition in the hepatocytes. The H&E staining of liver sections revealed ballooning degeneration along with multifocal necrotic regions and invasive mononuclear cells indicating hepatic inflammation (p < 0.001; Figure 1C). In comparison to the control group, WD feeding resulted in mild to moderate fibrous tissue deposition surrounding steatotic cells and vessels. We also observed a significant increase in feed intake in WD-fed mice over the course of 16 weeks. This resulted in an average 20 g of weight gain in WD mice, which is more than 30% of the weight of control mice (p < 0.001) (Figures 1D, E). Further, we observed that the blood glucose and serum insulin levels were significantly elevated (p < 0.001) in WD fed mice (Figures 1F, G).

Treatment with phloretin substantially reduced the fat deposition in the liver while preserving normal hepatic architecture (Figure 1B). In comparison to the WD group, phloretin administration was observed to significantly (p < 0.001) decrease the extent of degenerative alterations in the liver (Figures 1B, C). Administration of phloretin also lowered the fibrotic changes in the liver as compared to the WD group. Furthermore, as compared to the WD group, we observed a 40% reduction (p < 0.001) in body weight gain of phloretin treated mice, however, no significant change was observed in weekly food intake (Figures 1D, E). Phloretin treatment also resulted in a significant (p < 0.01, p < 0.001, respectively) reduction in blood glucose and serum insulin levels thereby maintaining homeostasis model assessment-estimated insulin resistance (HOMA-IR) (Figures 1F–H).

Western diet feeding resulted in dyslipidemia causing significant increase in serum triglycerides (TG) and cholesterol (TC) levels. Precisely, following 16 weeks of WD feeding, a ∼3-fold (p < 0.001) increase in TG and ∼1.5-fold increase in both TC (p < 0.01) and LDL-C (p < 0.001) levels were observed (Figures 2A–C). Abnormal lipid metabolism is a defining feature of MAFLD, which is accompanied by alterations in the lipid metabolism pathway. Thus, we investigated the expression of genes encoding for fatty acid (FA) metabolism to govern the effect of phloretin on hepatic lipid metabolism. Our findings showed that the hepatic expression levels of lipogenic genes SREBP-1c, ChREBP, CD36, and FASN are elevated in response to WD, which contributes to de novo lipid synthesis (Figure 2E). In comparison to the control group, a 4.8-fold (p < 0.001) and a 3.4-fold (p < 0.001) elevation in the mRNA expression of SREBP-1c and ChREBP transcripts, respectively, as well as a 5.7-fold (p < 0.001) increase in the FASN gene expression was detected in the liver of WD group mice. Apart from an increase in SREBP-1c, ChREBP, and FASN, a 3.5-fold (p < 0.001) rise in the mRNA expression of CD36 was also observed in WD fed mice. Further, in WD fed mice, the mRNA expression of PPAR-γ and PPAR-α showed a small but no significant alteration (Figure 2F).

Steatosis-induced inflammation plays a critical role in the progression of disease to more severe outcomes. Serum ELISA has been performed to evaluate the levels of inflammatory cytokine markers. Results revealed that exposure to WD augmented the circulating inflammatory cytokines viz. IL-6 (>3-fold; p < 0.05) and TNF-α (∼5-fold; p < 0.001) levels (Figures 2G, H). Further, the gene expression analysis revealed significant (p < 0.001; Figure 2I) elevation in IL-6 and TNF-α expression after chronic exposure to WD. Additionally, Fecal endotoxin levels play critical role in assisting the inflammation and disease progression. The fecal endotoxin level was estimated in each group every fourth week. At week 4, a slightly elevated but not significant increase was observed in endotoxin level in feces of mice from WD and WD + P groups. However, continuous exposure to WD led to a substantial increase in the endotoxin release at 8th, 12th, and 16th week (p < 0.05, p < 0.001, p < 0.001, respectively) as compared to the control group (Figure 2J).

Phloretin treatment was found to improve lipid homeostasis and help in subverting the dyslipidemia in WD-fed mice. As illustrated in Figure 2, phloretin treatment of WD-fed mice prevented increase in TG (p < 0.001), TC (p < 0.01), and LDL-C (p < 0.001) levels (Figures 2A–C). However, no statistically significant difference was observed in the levels of HDL-C across all the groups (Figure 2D). Moreover, phloretin treatment reduced the expression of lipogenic marker genes. Importantly, we observed a 57 and 56% reduction (p < 0.001) in the expression of SREBP-1c and ChREBP, respectively on treatment with phloretin. Similarly, phloretin treatment resulted in a 2.7-fold (p < 0.001) and ∼1.6-fold (p < 0.001) downregulation of FASN and CD36 genes, respectively (Figure 2E). Interestingly, phloretin supplement further upregulated the mRNA expression of PPAR-γ and PPAR-α by 1.7-fold (p < 0.01) and 1.3-fold (p < 0.05) respectively, facilitating hepatic FAs oxidation (Figure 2F).

On the contrary to the WD exposure, phloretin treatment significantly reduced the levels of inflammatory cytokine markets IL-6 and TNF-α their normal levels (Figures 2G, H). Phloretin, in instance, was found to cause a 2.3-fold (p < 0.05) and >3-fold (p < 0.01) decline in serum IL-6 and TNF-α levels, respectively. The gene expression of IL-6 and TNF-α further demonstrated the ability of phloretin to diminish hepatic inflammation and consequently prevent the progression of NASH in mice (p < 0.001; Figure 2I). Furthermore, the endotoxin release initially increased in the feces of mice treated with phloretin, but prolonged exposure to phloretin decreased the endotoxin level to normal (Figure 2J). The result indicated the protective efficacy of phloretin against microbial toxic byproducts. These results suggested the potential of phloretin in mitigation of WD-induced imparities leading to MAFLD progression in mice.

To evaluate the role of gut-microbes in WD-induced disease progression and role of phloretin as counter strategies 16S rRNA sequencing and analysis of metagenome was performed. Post 16S rRNA sequencing, data was analyzed and microbial diversity was compared across the treatment groups at different time-points (Figure 3). Analysis of comparative microbial diversity and richness within groups at different time intervals revealed differences in diversity, as showed by the Shannon index and observed ASVs (p < 0.05). Significant differences were observed in diversity among WD groups versus control group, stratified across time points, prominent change was observed after 8 weeks (Figure 3A). Phloretin treatment reversed these changes at week 4, 8, and 12 but not at week 16. However, phloretin treatment maintained the overall diversity index similar to the control group (Figure 3B). Furthermore, there was a statistically significant difference (p < 0.05) in taxonomic richness between control and WD fed groups at week 8, 12, and 16. However, significant change in the richness (p < 0.05) between the WD and WD + P groups was observed only at week 12 (Figure 3C). Phloretin did not effectively alter the microbial richness at each time point; however, it managed to reverse the overall richness (Figure 3D). Overall, these results indicated that phloretin treatment resists WD induced changes in microbial diversity and richness in mice and indicates its prophylactic effects to maintain the diversity similar to healthy control group. Detailed Shannon and Simpson indices and observed ASVs of individual samples were given in Supplementary Table 2.

Distance matrices suggest that WD fed mice exhibited distant diversity as compared to control and phloretin treatment group, which was comparatively less distant to the control group (Figure 3E). Additionally, separate comparative PCoA analysis at each time point exhibited a prominent difference in microbial diversity of WD as compared to control and WD + P (PERMANOVA r² ranging between 0.57 and 0.65) (Figure 3F). Results indicate that Bray–Curtis distance was higher between WD and control group across the timeline. On the other hand, the diversity of the phloretin treatment group was comparatively less distant to control group. These outcomes substantiated that phloretin treatment restores the changes in microbial community composition.

In order to evaluate the phylogenetic comparison of microbial communities, taxonomic signatures were identified in each experimental group at different timepoints. Relative abundance of different phylum and mean relative abundance of top 20 bacterial families were observed (Figure 4). After fourth week, relative abundance of Firmicutes was decreased and Desulfobacterota increased in the WD group as compared to control. In phloretin treatment group, no such observation was recorded for Desulfobacterota. In the eighth week, increase in Campilobacterota and decreased abundance of Desulfobacterota was observed in the phloretin treatment group similar to control but not in the WD group. After 12th week, Campilobacterota had largely decreased abundance in WD group as compared to the control and phloretin treatment groups. In 16th week, abundance of Firmicutes was decreased in the WD and phloretin treated groups as compared to the control. Large increase in abundance of Desulfobacterota was observed in the WD group, which was negligible in control and phloretin treated groups comparatively. Cumulatively, with increase in time abundance of Campilobacterota was increased and Desulfobacterota was decreased in the phloretin treatment group similar to control but dissimilar to the WD fed mice. Additionally, Cyanobacteria was found to be more abundant in control and phloretin treated mice as compared to control. Interestingly, abundance of Proteobacteria was found to be increased with time in the control group but decreased in the WD and phloretin treatment group (Figure 4A).

To delve deeper, mean relative abundance of top 20 bacterial families were analyzed. Abundance of Muribaculaceae, Tannerellaceae, Desulfovibrionaceae, and Lachnospiraceae, were increased in WD fed mice in comparison to control and phloretin treatment groups. Similarly, Anaerovoraceae, Peptostreptococcaceae, and Erysipelotrichaceae were least abundant or absent in the control group as compared to WD and phloretin treatment group. Ruminococcaceae was less abundant whereas Selenomonadaceae was completely absent in WD fed mice as compared to control group. Interestingly, abundance of Helicobacteraceae was more abundant in control as compared to the other groups. Although, it is highly unlikely that WDs with comparable nutrients would lead to decrease in Helicobacteraceae abundance observed in our study, but it is still a potential limitation of the present study. Finally, decreased abundance of Lactobacillaceae was observed after WD intake, which improves with the phloretin treatment (Figure 4B and Supplementary Figure 1). At the same time, pathogenic anaerobes like Peptostreptococcaceae were also found abundant in the WD fed group, which was least abundant or absent in the control group. The phloretin treatment group witnessed less abundance of these bacterial families and increased abundance of non-pathogenic Lactobacillaceae family (Figure 4B and Supplementary Figure 1). We also checked for the relative abundance of ASVs that had >1% occurrence and was plotted to observe the strain level difference amongst samples. The Bacteroides-5 genus was observed to be more prevalent in WD group as compared to the control and phloretin treated groups, with an increase in dominance occurring over time. In comparison, ASVs belonging to the species Bacteroides vulgatus remained more dominant in the both control and phloretin treated group. Supplementary Figure 2 illustrates the dynamic of the gut bacterial population among the groups over a period of 16 weeks and demonstrate the diet and treatment induced variations in taxa. Taxonomic signature analysis upholds the role of WD in alteration of microbial diversity dynamics and phloretin was found to restore these changes.

Variance Partition analysis was performed to assess differentially abundant taxa in all experimental groups using diet pattern and time as independent variables. The diet pattern appeared to attribute more fraction of the variance as compared to time (weeks) (Supplementary Table 3). Heatmap signature of selected taxa (with a variable fraction value >0.25 of diet pattern) revealed the changes in abundance of observed ASVs at family, genus and species level across the treatment groups (Figure 4C). Hierarchical clustering heatmap revealed association of the selected taxa with the progression of MAFLD and phloretin treatment. On a larger extent, ASVs belonging to the genus Blautia, Terrisporobacter, Alloprevotella, Muribaculaceae, Eubacterium, Clostridium, Parabacteroides, Allobaculum, Odoribacter, and family Desulfovibrionaceae, Paludibacteraceae, was highly abundant in WD fed mice indicating a positive association with the progression of MAFLD. These taxa have been shown less abundant in control and WD + P groups, which suggests negative correlation with phloretin treatment. Interestingly, Lactobacillus johnsonii species were also abundant in the WD group. In contrast, several ASVs belonging to genus Lactobacillus, Quinella, Prevotellaceae, Alloprevotella-1, Odoribacter-1, Bacteroides-1, and species B. vulgatus and Bacteroides caecimuris-1 showed negative association with WD intake and they started restoring to their normal abundance status in phloretin treatment group. Additionally, taxa from genus Ruminococcus, Eubacterium coppostanoligenes 1, Hungatella, Bacteroides-5 and some species like Bacteroides nordii, and Bacteroides uniformis was highly abundant in WD + P group and was less abundant in both control and WD groups. Interestingly, Bacteroides thetaiotaomicron was also found abundant in WD + P group in the early phase but was least abundant in later phases of phloretin treatment. Findings indicates strong positive association of these taxa with phloretin treatment. These associations strongly suggests that WD intake changes the gut microbiota with progression of MAFLD and phloretin treatment not only reversed these changes but also modulated some other important taxa signatures.

To delve deeper into the disease associated changes of microbial taxa signatures, cumulative relative abundance of taxonomic markers and distribution of selected taxonomic markers in ordination plot for each experimental group was evaluated. Taxonomic signatures were identified using ISA. Top 10 bacterial taxa (at families, genera and species level) were selected for each experimental group based on significant discrimination across groups (p-value) and their prevalence within the group (the maximum IndVal scores) (Supplementary Table 4 and Figure 5). Clear demarcation between cumulative relative abundance (sum of relative abundances of all 10 representative taxa) was observed across the experimental groups. Results indicate that highly abundant (>35 at week 16) signature marker taxa for control group (belonging to the genus Lactobacillus, Alloprevotella-1, Bacteroides-2, Prevotellaceae, Odoribacter-1, Alistipes-8, and Quinella) was relatively least abundant in the WD group (<8 at week 16) and was found to be restored after phloretin treatment (>15 at week 16), relative abundance at the start of the study was <10 in all the experimental groups (Figure 5A). Similarly, results suggested relatively high abundance of signature taxa of WD group (belonging to the family Desulfovibrionaceae, Paludibacteraceae and genus Clostridium, Parabacteroides, Muribaculaceae, Odoribacter, Alloprevotella, and species L. johnsonii) was present (>20 at week 16) as compared to the relative abundance of these taxa in control and WD + P group (<5 at week 16) (Figure 5B). Moreover, relative abundance of marker taxa of WD + P group (belonging to family Lachnospiraceae-7, genus Prevotellaceae, Hungatella, Bacteroides, and species B. nordii, B. uniformis, Ruminococcus torques 1 and 2) was decreased with time in WD + P and control groups (>25 at week 8 and <20 at week 16 for WD + P and >5 at week and <5 at week 16 in control) as compared to the WD group, where a time-dependent increment was observed (<10 at week 4 and >15 at week 16) (Figure 5C).

Distribution of selected taxonomic markers in ordination plot was done using PCoA analysis (Figure 5D). Colored ordination plots clearly indicate the higher abundance of ASVs from Parabacteroides and Paludibacteraceae in the WD group as compared to the control and WD + P groups (Figures 5G, H). Additionally, ASVs from Lactobacillus, Alloprevotella were highly abundant in control group and relatively more abundant in the WD + P group as compared to the WD group (Figures 5E, F). B. nordii was more abundant in the WD + P group as compared to the WD and control groups and B. uniformis was also found to be highly abundant in the WD + P group and week 12 and 16 of control groups as compared to the WD group (Figures 5I, J). It is important to note that all of these observations were largely time dependent, indicating a clear role of phloretin in reversing the effect of gut dysbiosis associated changes in microbial diversity in MAFLD. The samples from each group were then analyzed to create a Spearman’s rank correlation matrix. The selected taxa from each group were used to calculate the distance matrix by Euclidean method. At the end of the fourth week, there was negligible distance between samples indicating all the group sharing similar microbial environment. Furthermore, in the 8th, 12th, and 16th weeks, the control and WD + P groups formed comparable clusters, whereas the WD group formed a distinct cluster (Supplementary Figure 3). The results indicated that the samples in WD group were more distant from the samples in the control and WD + P groups.

In order to examine the association between taxa in different groups, a co-occurrence network plot was created with all strong (r > 0.5) significant correlations (p < 0.05), and only those taxa were picked up that were positively associated to each other in each group. The neighborhood connectivity was found higher in the control group as compared to the WD and WD + P group. In particular, ASVs belonging to genus Odoribacter-1 was found highly connected with other ASVs from genus Lactobacillus, Alloprevotella-1, Bacteroides-2, Prevotellaceae, Alistipes-8, Quinella, and species Ruminococcus flavefaciens with r value of 1 (Figure 6A). However, absence of Odoribacter-1 and R. flavefaciens in the co-occurrence network of WD and WD + P groups lead to a decrease in neighborhood connectivity. Further, network density of taxa belonging to genus Blautia-3, Bacteroides-5, Muribaculaceae-25, and Muribaculaceae-18 was also higher in the control group with network strength of <1. In contrast, in the WD group the neighborhood connectivity was dense only with taxa of genus Bacteroides-5, Clostridium_sensu_stricto-1, and species Lactobacillus murinus, L. johnsonii with r value of <1 (Figure 6B). However, the neighborhood connectivity was less frequent in the WD + P group as compared to control group but the correlation strength was higher in both control and WD + P group as compared to WD group indicating strong connectivity among bacterial species (Figure 6C). All of these observations indicated loss of neighborhood connectivity as well as network strength among species during the progression of MAFLD. However, phloretin treatment was ineffective in restoring connection, but it did maintain the strength of the network among species.

Western diet and phloretin supplementation both have been shown to affect the composition of the gut microbiota, and has potential to cause alterations in bacterial metabolic processes. KEGG pathway signatures were identified in whole microbiome and ISA was performed as a statistical test to determine significantly discriminating metabolic pathways in each group at different time points. A subset of level 3 metabolic pathways that are substantially linked with the course of MAFLD and have a statistically significant (p < 0.05) alteration was chosen for each experimental group. WD supplementation resulted in time-dependent enrichment in some of the lipid metabolism pathways, i.e., primary and secondary bile acid biosynthesis, FA elongation in mitochondria, biosynthesis of unsaturated FAs, and glycerophospholipid metabolism. Phloretin supplementation, on the other hand, has been shown to deplete these pathways (Figure 7A). Taking carbohydrate metabolism into account, fructose and mannose metabolism and tricarboxylic acid (TCA) cycle pathways were more abundant in the representative genomes of the taxa possessing a positive association with WD group. In contrast, their abundance appears to be decreased in the phloretin treatment group as compared to WD group. However, galactose metabolism, inositol phosphate metabolism and ascorbate and aldarate metabolism pathways were significantly upregulated by phloretin treatment as compared to WD group (Figure 7B). Vitamin metabolism pathways were also altered in the WD group with time dependent depletion of vitamin B6 and retinol metabolism pathways occurring simultaneously with enrichment of thiamin, biotin, and riboflavin metabolism and the induction of folate biosynthesis. However, it has been demonstrated that phloretin supplementation can help to reverse these alterations (Figure 7C). Amino acids metabolism pathways (excluding D arginine and D ornithine metabolism) were highly enriched in the representative genome from the WD group but less abundant in both control and WD + P groups (Figure 7D). WD group also showed enrichment with pathways involved in lipopolysaccharide biosynthesis at week fourth but depleted with time, as shown by related taxa in glycan biosynthesis and metabolism pathways. However, phloretin supplementation inhibited these pathways as the time progressed further (Figure 7E). Phloretin also facilitates drugs and xenobiotic metabolism by cytochrome P450 pathways in WD fed mice (Figure 7F). Moreover, the heatmap signature demonstrated the abundance of these marker pathways in each sample from each group during time intervals varying from 4th to 16th week (Figure 7G). Since these findings were merely correlations, we can only speculate that altered metabolic pathways were to account for the MAFLD phenotype; nevertheless, more research in this area is required.