Male C57BL/6J mice (n = 8 per diet group) from Jackson Laboratories (Bar Harbor, ME) were singly housed in standard microisolator cages at the Veterinary Research Building, College of Veterinary Medicine, Auburn University (Protocol reference number 2014-2547). The room was maintained at an ambient temperature of 22°C ± 2°C on a 12:12 light:dark cycle with zeitgeber time (ZT) 0 representing lights on and ZT12 representing lights off. All experimental procedures were approved by the Auburn University Animal Care and Use Committee. Animals were fed standard rodent chow for 1 week during acclimation to the facility. After which, animals were split into groups receiving either the standard chow diet with tap water (Chow) or a High-Fat Western Diet with tap water (WD). The chow diet (Teklad Global Rodent Diet) contained 24% of calories from protein, 18% from fat, and 58% from carbohydrate. The WD diet was based on the AIN-93G diet and consisted of 44% carbohydrate, 16% protein, and 40% fat, 30% of which was provided from lard, 30% from butterfat, 30% from Crisco, 7% from soybean oil and 3% from corn oil (Reeves, 1997). All dietary groups were given food and water ad libitum.

Promethion Metabolic Mouse Cages (Sable Systems, Las Vegas, NV) were used to house animals for metabolic screening and phenotyping. Animals were transferred from their home cages and singly housed in the metabolic cages at three time points (2, 4 and 12 weeks after diets began). The animals (n=8 per treatment group) were singly housed in the cages for 3 consecutive days with the 1^st day committed to environment acclimatization and the 2^nd and 3^rd days for data collection. All animals were returned to their home cage after completion of metabolic phenotyping. Animal activity was measured by the Promethion XYZ Beambreak Activity Monitor. Food, water, and body weight were measured by Promethion precision MM-1 Load Cell sensors. The amount food and water withdrawn from the hoppers was measured and analyzed. The body mass monitors were plastic tubes that also functioned as in-cage enrichment and nesting devices. Water vapor, CO₂, and O₂ were analyzed by the Promethion GA-3 gas-analyzer to provide detailed respirometry data. Energy expenditure (EE) was calculated in kilocalories (kcal) by utilizing the Weir equation: 60*(0.003941*VO₂ (n) +0.001106*VCO₂ (n)) in which VO₂ is the oxygen uptake and VCO₂ is the carbon dioxide output, both of which are measured in ml/min. Respiratory exchange ratio (RER) was determined by measuring gas exchange within the metabolic cages to identify the substrate being primarily utilized for energy within the body. Specifically, RER is the ratio of CO₂ produced to the volume of O₂ consumed (RER = VCO₂/VO₂) where a RER ~ 0.7 indicates lipid utilization and a RER ~ 1.0 indicates carbohydrate utilization. All metabolic phenotyping data were analyzed using ExpeData software (version 1.8.2; Sable Systems) with Universal Macro Collection (version 10.1.3; Sable Systems).

Collection of fecal matter was performed at the initiation of diet change (Day 0) as well at four other time points following WD feeding: 2 days, 2, 8, and 12 weeks. These timepoints were selected to evaluate microbiome composition prior to WD feeding and to allow for early (2 day and 2 weeks) and late (8 weeks and 12 weeks) profiling of the microbiome prior to the estimated timeline for diet induced obesity (16 weeks). Animals (n=8) were removed from their home cages and placed in sterile microisolator cages without bedding for 3 hours. Food and water were provided to the animals during this time. Mice were then returned to resident cages and feces were collected from the sterile cages for DNA extraction and shotgun metagenomic sequencing. Six random mice fecal samples from each treatment group were used to create the 3 samples for microbiome analysis. The comprehensive timeline of the experimental procedures and sampling periods are depicted in Figure S6 (SMART - Servier Medical ART).

Upon completion of the study, all animals were fasted and then sacrificed via CO₂ asphyxiation and quickly decapitated by guillotine to allow trunk blood collection. Tissues including the liver and visceral (epididymal and retroperitoneal) and subcutaneous (inguinal) white adipose depots were excised and weighed. Final blood glucose was measured by an Accu-Chek blood glucose meter. Serum insulin levels were determined by an insulin ELISA assay (Crystal Chem, Inc., Downers Grove, IL) and data were analyzed for insulin resistance using the HOMA-IR score (HOMA-IR = 26 * fasting serum insulin * fasting blood glucose)/405).

Immediately following collection, DNA was extracted from fecal samples using the Omega E.N.Z.A. Stool DNA kit according to manufacturer’s guidelines. Extracted DNA samples were pooled (2/sample) and shotgun metagenomic sequencing was performed by Hudson Alpha (Huntsville, AL). Whole genome sequencing was performed using an Illumina HiSeq v4 with a 2 x 125 paired-end sequencing 200 million reads. Metagenomic sequences were evaluated for quality using FastX toolkit (FASTX-Toolkit). Raw sequences can be found on the NCBI repository under the BioProject PRJNA730805 ¹ . Low quality sequences (Q-score less than 30) and sequencing adapters were removed using Trimmomatic (Bolger et al., 2014). Sequences were then uploaded to the Metagenomic Rapid Annotation Server (MG-RAST) version 4.0.3 for taxonomic and functional annotation (Meyer et al., 2008). Briefly, sequences were paired, filtered for quality, dereplicated, filtered for host-specific sequences (Mus musculus, UMD v3.0), and annotated. Annotated profiles of each sample are publicly available at the MG-RAST repository (MG-RAST Project ID: mgp81921)  ² . Taxonomic classifications were annotated using the GenBank repository with the minimum cutoff parameters of 1x10^-5 e-value and alignment length of 15. Density plots, calculated by total annotated hits, were used to set stringent uniform percent identity thresholds while maintaining an accurate measurement of the microbial populations. The percent identity was set to a minimum of 80% identity for bacterial and functional annotations and 70% identity for bacteriophage (Order Caudovirales) annotations. At these parameters, we were able to provide a conservative estimate of taxa while excluding ambiguous sequences. Current databases are not complete for viral genera classification due to variations in taxonomic classification strategies. Therefore, species within the order Caudovirales, as annotated by MG-RAST, were cross-referenced with the International Committee on Taxonomy of Viruses taxonomic database or other documentation detailing classification of the bacteriophage (International Committee on Taxonomy of Viruses. and King, 2011) ³ ( Table S8 ). From this, we were able to update viral species into current taxonomic genera in order to describe the types of bacteriophage present and how they fluctuate following dietary change. Descriptions of genera present in our samples are outlined in Appendix 1 . Taxonomic classification hits were then normalized based on the total hit count. Rarefaction curves depicting alpha diversity were generated using the MG-RAST server.

Final body and tissue weights along with serum measures were analyzed using a one-way ANOVA with a Newman-Keuls post-hoc test. The percent body weight change and 24-hour cycle data were assessed by a repeated measures two-way ANOVA so that animals in one diet group could be compared with animals in ;another diet group across dietary weeks or circadian time points The above statistical analyses were performed using SigmaPlot with significance determined at p < 0.05. The National Mouse Metabolic Phenotyping Centers (MMPC) Energy Expenditure analysis page ⁴ was used for multiple linear regression analysis (ANCOVA) to assess body weight as a covariate on energy expenditure with significance determined at p < 0.05. Significance for all measures was determined at p < 0.05 and all data are presented as Mean ± SE.

Three pooled samples for each diet group at each collection point were used to calculate changes of relative abundance in the microbiome. Relative abundance was used to calculate means and standard deviations of each treatment groups at each time point using the statistical program GraphPad Prism v4. Using the R studio statistical platform (R Core Team, 2014), t-tests were performed to identify significant difference in relative abundance of microbial taxa. Non-metric multidimensional scaling (nMDS) ordination was generated in R studio using the vegan package (Dixon, 2003). To generate the nMDS, raw bacterial hits were used to compute a sample dissimilarity matrix using the Bray-Curtis dissimilarity index. This matrix was then used to compute an ordination of the samples in two dimensions. The vegan package was also used to calculate Shannon’s Diversity Index scores. Then, the Pielou’s Evenness Index was calculated by dividing the Shannon’s Diversity Index score by the log of unique species amount. Mann-Kendall Trend tests were performed on diversity and evenness scores separately using the randtest package (Pohlert, 2018). Pearson correlation coefficients were calculated using relative abundance and metabolic readouts at 12 weeks following dietary exposure in R studio using the package psych (Revelle, 2017). Pearson correlation plots were generated in R studio using the package ggcorplot (Kassambara, 2015) using a correlation coefficient threshold of an absolute value of 0.6.

Body weight was tracked over the course of the 12 week experiment ( Figure 1 ). Prior to the experiment, the average body mass of the chow group was 20.9 +/- 1.8 g and the WD group was 20.8 +/- 1.7 g. By week 2 on different diets, animals fed the WD had a significantly higher percent body weight change compared to chow (p < 0.01) and persisting throughout the duration of the 12 week experiment (p < 0.001). The change in body weight was mainly due to an increase in body fat as the WD-fed group had significantly heavier visceral and subcutaneous (Sub-Q) fat pad weights at the 12 week endpoint of the experiment ( Table 1 , visceral: p < 0.001; Sub-Q: p < 0.01). There was no effect of WD on fasting blood glucose by 12 weeks, yet serum insulin levels were significantly elevated in the WD-fed group ( Table 1 , p < 0.01). Insulin resistance, as calculated using the Homeostatic Model Assessment of Insulin Resistance (HOMA-IR) method, revealed insulin resistance in the WD-fed animals compared to the Chow-fed animals also at 12 weeks ( Table 1 , p < 0.01).

To examine how diet impacted average energy expenditure (EE) within the light and dark phases, we measured EE at each ZT over a 24-hour cycle. Although diurnal rhythmicity was observed in both dietary groups, the WD induced significant disruptions evident as early as 2 weeks. The WD-fed group exhibited elevated average EE when compared to Chow-fed group during the inactive (day) period from ZT4-ZT11 ( Figure 2A , p < 0.01). At 4 weeks, the WD-fed group continued to demonstrate elevated average EE during the day, specifically at ZT3 and ZT8-11 ( Figure 2B , p < 0.05), as well as ZT20 ( Figure 2B , p < 0.01). By 12 weeks, the WD-fed group exhibited significantly elevated average EE for the entirety of the day cycle ( Figure 2C , p < 0.01). Additionally, we observed a dip in the average EE in from ZT18-22, which was not as pronounced in the WD-fed group. Consequently, the EE of the WD-fed group was significantly elevated during that time ( Figure 2 , p < 0.05).

Across all three experimental time points, a diurnal rhythm was observed in the respiratory exchange ratio (RER) of Chow-fed mice: greater metabolism of lipids during the inactive, day phase and more carbohydrate utilization during the active, night phase ( Figures 3A–C ). At week 2 and persisting through week 12, this rhythm of metabolic flexibility was significantly dampened in the WD-fed group with near constant lipid utilization across the time points. At 2 weeks, WD-fed animals demonstrated a significantly elevated RER from ZT4-ZT7 and a significantly decreased RER from ZT13-17 and ZT22-ZT24 ( Figure 3A , p < 0.05). Four weeks after diets began, RER in the WD-fed group was significantly decreased from Chow-fed animals starting at ZT12 and continuing to ZT24 ( Figure 3B , p < 0.05). Lastly, after 12 weeks of dietary exposure, we observed results similar to week 4 during ZT13-ZT18 with the WD-fed animals exhibiting a significantly decreased RER ( Figure 3C , p <0.05). These data suggest that the WD significantly impacted metabolic diurnal rhythms as well as metabolic flexibility. These effects were observed as early as 2 weeks and persisted through 4 and 12 weeks of WD feeding.

We observed a typical murine diurnal rhythm in our animals with elevated activity during the night phase and decreased activity during the day ( Figures S1A–C ). Diet did not appear to have an effect on activity at the 2 week time point ( Figure S1A ). Similarly, 4 weeks of WD consumption did not induce large-scale changes in diurnal activity ( Figure S1B ). However, at the 12 week experimental time point, we observed a significant reduction activity in the WD-fed group from ZT13-ZT15 ( Figure S1C , p < 0.05). However, this did not result in a significant decrease in activity when averaged across the entire night cycle.

After 2 weeks on the diets, the WD-fed group consumed significantly more food and water by weight than the Chow-fed group during the day ( Figure S2A , p < 0.05). Kilocalorie consumption was found to be significantly greater in the WD-fed group during all three time points ( Figure S2A , p < 0.05). At the 4 week time point, the WD-fed group consumed more grams of food during the day, but less food during the night than the Chow-fed groups ( Figure S2B , p < 0.01). This translated into greater kilocalorie consumption in the WD-fed animals during the day and total ( Figure S2B , p < 0.01). Twelve weeks after diets began, the WD-fed group did not consume more food by weight than the Chow-fed animals. However, kcal consumption was significantly elevated in the WD-fed group compared to Chow-fed group for the day, night and 24-hour total data points due to caloric density of the food ( Figure S2C , p < 0.05). There was not a significant difference in water consumption between the Chow-fed and WD-fed groups at 4 or 12 weeks.

To identify WD-induced changes in the enteric microbiota, we used shotgun metagenomic sequencing of freshly isolated feces at each indicated timepoint. First, to ensure that we had adequate depth of sequencing, estimates of alpha diversity and depth of sampling were assessed using rarefaction curves ( Figures S3A–E ). Each sample at each time point reached a plateau, (indicating more sequences than OTUs) signifying adequate depth of sampling and alpha diversity.

We next assessed global changes in the enteric microbiome in the WD-fed vs. Chow-fed mice. To achieve this, we analyzed the composition of the enteric microbiota over time by means of non-metric dimensional scaling (nMDS) ordination plots based on the Bray-Curtis dissimilarity index ( Figure 4 ). This method provides a similar visualization of the data as a principle component analysis (PCA) without the biases associated with a PCA concerning low abundance organisms (Zhu and Yu, 2009). The fit or stress of the nMDS denotes how well the ordination consolidates the observed distances among the samples where a value less than 0.3 would indicate a good fit. The fit associated with our analysis was 0.156. In Figure 4A , the nMDS ordination plot factored by diet revealed one predominant central cluster (Chow diet) with the other diet (WD) scattered in a radiating pattern, indicating increasing dissimilarity. Thus, the composition of the microbiome was much more stable in the chow group over time, than in the WD group. To examine how dissimilarity progressed over time between the two diet groups, an nMDS ordination plot factored by diet and time was constructed ( Figure 4B ). This analysis highlighted the progression of increasing changes within the intestinal microbiome over time in those animals fed a WD ( Figure 4B , Figure S5 ). This trend of increasing dissimilarity continued throughout the duration of the experiment in which greater body weight and metabolic changes overtime were observed between the WD- and Chow-fed mice. We also observed a significant decrease in bacterial diversity and evenness in the WD-fed group compared to Chow-fed group, primarily at 12 weeks (p = 0.006; p = 0.008) ( Figures S4A, B and Tables S6, S7 ). Using Pearson’s correlation we also noted that bacterial diversity and evenness were inversely related to parameters that increased during the development of obesity including HOMA-IR, body weight, visceral fat weight, and Day EE ( Figure S4E and Table S10 ). In summary, decreases in enteric bacterial diversity and evenness, indicators of dysbiosis, were observed in mice fed a WD and correlated strongly with the metabolic phenotype of obesity.

Our next goal was to determine which microbial populations most significantly contributed to the dissimilarity seen in WD-associated microbiota. Relative abundance plots illustrate the composition of the microbiota in each group over time ( Figure 5A and Table S1 ) (Shreiner et al., 2015). Prominent in the Chow-fed group were members of four well known bacterial phyla: Firmicutes (60 ± 8.5%), Bacteroidetes (33.460 ± 7.1%), Proteobacteria (1.675 ± 0.2%), and Actinobacteria (1.012 ± 0.2%), at day 0, consistent with previous reports (David et al., 2014; Ziętak et al., 2016). The microbial composition remained relatively stable throughout the 12 week experiment for mice fed a Chow diet. On the other hand, shifts in relative abundance profiles of WD-fed mice samples were detected as early as day 2 ( Figure 5A and Table S1 ). For example, Proteobacteria showed a significant reduction in abundance in the WD-fed group vs. Chow-fed group (p = 0.02), while Verrucomicrobia began to increase after 2 days from 1.467 ± 2.4% to 9.182 ± 4.6% in the WD-fed group. The increase in Verrucomicrobia abundance was transient and returned to pre-diet abundance levels by 12 weeks. Notably, significant increases were observed in Firmicutes (p = 0.04) at 12 weeks in the WD group ( Figure 5A and Table S1 ).

To more precisely resolve shifts in bacterial communities, we next evaluated class-level and family-level community profiles ( Figures 5B, C and Tables S2, S3 ). The composition of the bacterial classes and families in the Chow-fed group were relatively stable throughout the experiment ( Figures 5B, C and Tables S2, S3 ). In contrast, we observed significant changes in bacterial abundance/composition in the WD-fed group over the experimental timeline.

Most notable among the WD-induced changes was an increase in class Bacilli at 12 weeks (p = 0.04). Bacilli have been previously shown to be elevated in the intestinal microbiota in obese individuals (Ziętak et al., 2016). Enterococcaceae, specifically the Genera Enterococcus, appeared to be the major contributor to this increase at 12 weeks in the WD group (p = 0.04). Additionally, the increase seen in the class Bacilli was also due to a bloom in Streptococcaceae abundance at 2 weeks (p = 0.005), 8 weeks (p =0.01) and 12 weeks (p =0.02). Other notable increases in bacterial composition included Verrucomicrobiae. Starting as early as 2 days, we observed an increase in the class Verrucomicrobiae in the WD-fed group that peaked at 2 weeks and returned to roughly baseline levels by 12 weeks ( Figure 5B and Table S2 ). This increase could be attributed to an increase in the family Verrucomicrobiaceae ( Figure 5C and Table S3 ), primarily from the genera Akkermansia (data not shown).

We also observed decreases in class Clostridia starting at 2 weeks and becoming significant at 12 weeks (p = 0.01) ( Figure 5B and Table S2 ). These decreases were accounted for primarily by members of the family Clostridiaceae beginning at 2 days on WD (p = 0.008) ( Figure 5C and Table S3 ). Other significant decreases in the WD-fed group were observed in class Bacteroidia at 2 weeks (p = 0.04) and 12 weeks (p = 0.0007) and many members of the phylum Proteobacteria. Together, the reduction of overall diversity and evenness within the intestinal microbiota could be attributed to over-growth of Bacilli and Verrucomicrobiae family members along with reductions in Clostridiaceae and several other bacterial constituents on WD.

We next wanted to determine how other constituents of the microbiome, namely bacteriophage populations, were impacted by the WD. First, we assessed the degree of the dissimilarity by analyzing bacteriophage diversity and evenness ( Figures S4B, D and Table S6 ). Similar to the trends observed in bacterial populations, we observed a reduction in the diversity and evenness of the enteric phageome over the course of the experiment in WD-fed mice. The Mann-Kendall trend analysis revealed a downward trend over the course of the experiment in both diversity (S = -65, p = 0.008) and evenness (S = -75, p = 0.003) in WD-fed groups ( Table S7 ). Moreover, the reduction in the diversity and evenness of the enteric phageome preceded similar changes in bacteria following dietary exposure ( Figures S4A, B ). Bacteriophage diversity and evenness were also inversely related to metabolic parameters that increased during obesity development, specifically HOMA-IR, body weight, visceral fat weight, and Day EE ( Figure S4E and Table S11 ). Thus, changes in bacteriophage populations preceded changes in bacterial composition, but both had similar overall trends of decreasing diversity on WD. Also, both changes correlated with changes in metabolic phenotype.

We next evaluated the relative abundances of bacteriophage at the family and genus levels. It is important to note here that phage taxonomy classifications are not as complete as those for bacteria, especially at the genus level (Kuhn, 2020). Thus, many of the bacteriophages in our analysis remain unclassified beyond the family annotation. The enteric phageome of Chow-fed mice remained relatively consistent throughout the experiment at the family and genus levels ( Figures 6A, B ). In contrast, mice fed a WD demonstrated an increase in Siphoviridae at 2 weeks (p = 0.0001) coupled with a reduction in Myoviridae at 2 weeks (p = 0.001) and 8 weeks (p = 0.05), and Podoviridae as early as day 2 ( Figure 6A and Table S4 ). The increase in Siphoviridae in the WD-fed group could be largely attributed to increases in P335-like viruses starting at 2 days, which were maintained at elevated abundance through 12 weeks ( Figure 6B and Table S5 ), with significant increases at 2 weeks (p = 1.78 x 10^-5) and 8 weeks (p =0.0001). Additionally, we observed an increase in phiFL-like viruses at 12 weeks. phiFL-like viruses are primarily temperate bacteriophage that target members of the Enterococcaceae family (Adriaenssens et al., 2015), and this increase could be attributed to an increase in host species containing a prophage ( Table S8 ).

Myoviridae family abundance decreased significantly as early as 2 weeks and continued to decline in WD-fed mice at 12 weeks (2.297 ± 0.5%) ( Figure 6A and Table S4 ). This reduction was due in part to a significant depletion of the bacteriophage genus phiCD119-like viruses at 2 weeks (p = 0.004) in WD-fed mice ( Figure 6B and Table S5 ). Though not significant, phiCD119-like viruses remained reduced in relative abundance in the WD-fed group at 8 weeks and 12 weeks. Taken together, WD feeding increased the abundance of the Siphoviridae family and reduced abundance in the Myoviridae family compared to Chow-fed mice, further, WD-induced alterations in the enteric phageome occurred as early as 2 days on diet ( Figure 6B and Table S5 ), preceding changes in their bacterial host abundance ( Figure 5B ).

To obtain a more comprehensive picture of how diet affects the relationship between specific phage genera and bacterial families, we calculated Pearson correlation coefficients between the Caudovirales genera and the 55 most abundant bacterial families at the final time point of 12 weeks ( Figure 7 and Table S9 ). The correlation coefficients were sorted into negatively and positively correlated groups which resulted in: 1) a large group of bacteria that correlated positively with a large group of phages (both highlighted with green brackets) and 2) a smaller group of bacteria and bacteriophage that were positively correlated with each other (orange brackets). We observed members within the designated groups had the same trend in response to diet. For example, members within the green group were often reported as reduced in abundance following WD-feeding and the members of the orange group were often seen to bloom following WD-feeding. The two groups also had conversely negative correlations with members of the other (green vs. orange) group. This correlation grouping resulted in separation of bacterial families, most of which clustered with other families belonging to the same class. Given these findings, we next determined the frequency of putative phage-host relationships between the positively correlated groups.

Among the organisms within the green brackets, roughly 48% had positive correlations between bacteriophage genera and their putative bacterial hosts ( Figure 7 and Tables S8, S9 ). Six of these were very strong correlations of 0.6 or higher. These include bacteriophage genera who target bacterial families Pasteurellaceae (phiPLPE-like viruses) (Leblanc et al., 2009; Comeau et al., 2012), Burkholderiaceae (Bpp-1- and phiE125-like viruses) (Summer et al., 2007), Clostridiaceae (phiCD119-like viruses) (Revathi et al., 2011; Sekulovic et al., 2014; Hargreaves and Clokie, 2015) and Enterobacteriaceae (Che8- and RB49-like viruses) (International Committee on Taxonomy of Viruses (ICTV); Petrov et al., 2010; Hatfull, 2012). Many of the bacteriophage in these genera were temperate. phiCD119-like viruses, for example, are temperate bacteriophage that target Clostridium species as their host (International Committee on Taxonomy of Viruses (ICTV); Revathi et al., 2011; Sekulovic et al., 2014; Hargreaves and Clokie, 2015). Temperate bacteriophage, such as phiCD119-like viruses, would correlate directly with bacterial host abundance if they are present as integrated prophage. Also in the organisms within the green brackets were 7 bacteriophage genera that negatively correlated with their putative host (correlation coefficients between 0.4 and 0.6; Table S8, S9 ). Interestingly, bacteriophage genera with the strongest negative correlations in this grouping targeted either members of Bacillaceae (McLaughlin et al., 1986; Grose et al., 2014), Streptococcaceae (Garneau et al., 2008; Guglielmotti et al., 2009) or Staphylococcaceae (Goerke et al., 2009) ( Table S8 ). For example, Streptococcaceae-targeting bacteriophage genera 1706- and Sfi21-like viruses were strongly negatively correlated with their host, Streptococcaceae (International Committee on Taxonomy of Viruses (ICTV); Garneau et al., 2008; Guglielmotti et al., 2009).

While the orange group contained fewer bacterial families and bacteriophage genera than the green group ( Figure 7 and Table S8, S9 ), the majority of bacteriophage and putative hosts in this group showed strong positive correlations. These include Streptococcaceae-targeting P68- (Deghorain and Van Melderen, 2012), Sfi11- (Deveau et al., 2008; Guglielmotti et al., 2009; Ali et al., 2014), and P335-like viruses (Labrie and Moineau, 2000; Samson and Moineau, 2010; International Committee on Taxonomy of Viruses and King, 2011), Staphylococcaceae-targeting P68- (Nelson et al., 2003) and phiETA-like viruses (Bae et al., 2006; Daniel et al., 2007; International Committee on Taxonomy of Viruses and King, 2011; Bueno et al., 2012), Bacillaceae-targeting TP21-like viruses (Klumpp et al., 2010) and Enterococcaceae-targeting phiFL-like viruses (Adriaenssens et al., 2015). For example, P335-like viruses were seen to strongly correlate with their bacterial host, Streptococcaceae. This genus is comprised of an equivalent amount of virulent and temperate bacteriophage species. Thus, the elevation in abundance of this bacteriophage genera cannot be solely due to replication of prophage via host replication. Taken together, phage abundance positively correlated with the abundance of their putative host bacteria in many cases, but not all.

Next, we examined how specific bacterial families correlated with metabolic phenotypes within the WD-fed group. Interestingly, bacterial families clustered into similar groupings as those generated by correlations of bacteria and bacteriophage (again designated by orange or green brackets, Figure 8A and Table S10 ). Bacterial families grouped in the orange cluster positively correlated with metabolic parameters shown to be indicative of metabolic disease, namely HOMA-IR, Percent Body Weight Change, Visceral Fat Weight, and Day EE. Conversely, this orange group of bacterial families showed a strong inverse relationship with Night RER. As mentioned previously, a lower night RER (close to 0.7) during the active cycle would indicate primarily lipid utilization, characteristic of metabolic inflexibility. These results indicate that bacterial families within the orange group positively correlate with key pathophysiological metabolic parameters associated with obesity. Thus, our findings suggest that bacterial families within the orange group represent an obesity-responsive group of bacteria.

Correlation profiles showing a negative relationship with metabolic parameters associated with obesity were apparent for the bacterial families within the green grouping. Bacterial families within the green grouping negatively correlated with HOMA-IR, Percent Body Weight Change, Visceral fat weight, and Day EE and were most often positively correlated with Night RER. Hence, bacterial families grouped in green negatively correlate with key pathophysiological metabolic parameters associated with obesity.

Because strong relationships were identified between bacterial families and metabolic phenotypes, we sought to examine the relationships between bacteriophage populations in the intestinal microbiome and the metabolic phenotype observed in the WD-fed animals. Many of the bacteriophage genera that showed strong correlations to metabolic readouts correlated similarly in terms of green and orange grouped bacteriophage genera ( Figure 8B and Table S11 ). Orange labeled bacteriophage genera positively correlated with HOMA-IR, Percent Body Weight Change, Visceral Fat Weight, and Day EE. As these particular bacteriophage genera increased in relative abundance, phenotypic markers for body weight gain increased following dietary exposure. In contrast, green labeled bacteriophage genera had a negative relationship with the metabolic parameters mentioned above, revealing that as these metabolic parameters increased, there was a decrease in the relative abundance of the bacteriophage genera in the green group ( Figure 8B and Table S11 ). Because of the linked relationship between the bacteriophage genera and their host, correlations between bacteriophage and metabolic parameters must be considered within the framework of bacterial host dynamics.