Gut microbiota and metabolic marker alteration following dietary isoflavone‐photoperiod interaction

Abstract Introduction The interaction between isoflavones and the gut microbiota has been highlighted as a potential regulator of obesity and diabetes. In this study, we examined the interaction between isoflavones and a shortened activity photoperiod on the gut microbiome. Methods Male mice were exposed to a diet containing no isoflavones (NIF) or a regular diet (RD) containing the usual isoflavones level found in a standard vivarium chow. These groups were further divided into regular (12L:12D) or short active (16L:8D) photoperiod, which mimics seasonal changes observed at high latitudes. White adipose tissue and genes involved in lipid metabolism and adipogenesis processes were analysed. Bacterial genomic DNA was isolated from fecal boli, and 16S ribosomal RNA sequencing was performed. Results NIF diet increased body weight and adipocyte size when compared to mice on RD. The lack of isoflavones and photoperiod alteration also caused dysregulation of lipoprotein lipase (Lpl), glucose transporter type 4 (Glut‐4) and peroxisome proliferator‐activated receptor gamma (Pparg) genes. Using 16S ribosomal RNA sequencing, we found that mice fed the NIF diet had a greater proportion of Firmicutes than Bacteroidetes when compared to animals on the RD. These alterations were accompanied by changes in the endocrine profile, with lower thyroid‐stimulating hormone levels in the NIF group compared to the RD. Interestingly, the NIF group displayed increased locomotion as compared to the RD group. Conclusion Together, these data show an interaction between the gut bacterial communities, photoperiod length and isoflavone compounds, which may be essential for understanding and improving metabolic health.


| INTRODUC TI ON
Obesity is an epidemic affecting one in three Americans. 1 While the risk of developing obesity is higher in females, males are more prone to developing obesity-related diseases, including nonalcoholic fatty liver disease, insulin resistance, type 2 diabetes, and cardiovascular pathology. 2 Growing evidence indicates that soy isoflavones derived have beneficial roles in reducing obesity and diabetes markers. 3,4 However, the precise mechanisms involved remain unknown.
Isoflavones are metabolized into active compounds by the host gut microbiota. 5 While estrogens and the gut bacterial communities are known to play independent roles in obesity, new research suggests that estrogenic metabolites and the microbiota may work together to regulate metabolism. 5,6 For example, certain microbiota convert isoflavones to biologically active compounds, and in return, isoflavones may permit specific microbiota to thrive, resulting in their proliferation and marked growth. Thus, microbes are essential for the regulation of food's nutritional and energetic benefits, and their diversity is highly correlated with energy balance. 7 The gut microbiota composition and diversity differ between lean and obese profiles. For example, obese mice have a higher relative abundance of bacteria within the Firmicutes phylum compared to the Bacteroidetes phylum. 8 Furthermore, if the fecal matter containing the microbiota of obese mice is cross-transplanted into germ-free lean mice, the lean animals become more obese than the ones transplanted with fecal matter from lean mice. 8 Together, these reports strongly suggest that certain gut microbial profiles are associated with obesity. 9,10 While many other factors influence the gut microbiome, circannual rhythm is one that not only influences the microbiota composition, [11][12][13] but also impacts metabolism. 14 Fluctuations in environmental cues, mainly driven by the length of the active vs inactive periods, have been linked to changes in the gut microbiota. [11][12][13] In winter, when active periods are shorter, the number of human subjects that meet the criteria for metabolic syndrome is significantly higher. 15 In the same study, the authors concluded that short active periods exacerbate metabolic pathology.
Photoperiod interaction with dietary isoflavones may play a critical role in regulating most physiological functions by closely influencing the gut microbiota. 16 While several studies have established the respective effects of diet and photoperiod on the gut microbiome, a gap in knowledge remains as to how these two environmental cues may ultimately interact to affect the gut microbiome. In this study, we manipulated isoflavone consumption and photoperiod duration to examine how these external cues interact to dictate the temporal stability of the gut microbiome and its relationship with obesity and metabolism.

| Animals
Male C57BL/6J mice, aged 8 weeks old upon arrival, were obtained from The Jackson Laboratory (Stock No. 000664). All animals were housed in groups of four animals per cage and maintained on a 12-hour light and 12-hour dark light cycle (lights off 1300 hours) under controlled temperature conditions (22 ± 1°C) with ad libitum access to standard rodent chow and water throughout the experiment. Behavioural testing was conducted between 0600 and 1000 hours, and animals were acclimated to the behavioural room at 0600 hours for an hour before testing at 0700 hours. Cage changes, which coincides with fecal boli collection, were performed between 0700 and 0900 hours. Handling

| Diet
Following 2 weeks of acclimation to the facility and standard rodent chow (2018 Teklad Global 18% Protein Rodent Diet; Envigo), animals were randomly assigned to each experimental group (RD or NIF, 12L:12D or 16L:8D photoperiods). Mice in the first group assignment (diet) were randomly designated to the standard chow (maintained on the RD) or switched to an isocaloric isoflavone-free (NIF) chow (AIN-93G Purified Rodent Diet with corn oil replacing soybean oil; Dyets). The AIN-93 diet did not differ in macronutrient, mineral and vitamin composition from the Teklad Global 18% Protein Rodent Diet. 17 Both diets were previously characterized by our laboratory for isoflavone content using high performance liquid chromatography based analysis according. 16,18,19 The standard chow contains a total of 199.4 μg/g of isoflavone equivalents, whereas the NIF diet contains 0 μg/g of isoflavone equivalents. 18 All experimental groups were acclimated to these diets for 3 weeks prior to photoperiod modification.

| Photoperiod
Following the diet acclimation, mice were randomly assigned into a 12-hour light on, 12-hour light off (12L:12D) photoperiod (lights on at 1300 hours), or to a shortened activity period, 16-hour lights on, 8-hour lights off (16L:8D); lights on at 2100 hours). All animals assigned to the 16L:8D group were switched to an identical neighbouring room where lights were adjusted accordingly (see Figure S1 for timeline).

| Tissue collection
Trunk blood was collected following deep anesthetization with CO 2 inhalation and rapid decapitation. Blood was allowed to clot on ice and spun at 2000 g for 10 minutes prior to serum collection. The serum was then stored at −80°C until assayed.

| White adipose tissue histology
Epididymal white adipose tissue (eWAT) and livers were collected.
Samples were stored at −80°C until processed. While frozen, eWAT was divided in half. One half of the frozen tissue was fixed in 10% buffered formalin at 4°C for 48 hours. 20 The other half was immediately immersed in Ribozol for RNA extraction (see below RNA section). Once fixed, a total of four randomly selected samples per group were paraffin-embedded and sliced at 5 µm using a microtome (Microm HM 355S; Thermo Scientific). Three consecutive sections per paraffin block were cut and mounted per slide. A total of three sections per animal were processed and analysed (n = 4 animals/ group). Standard haematoxylin and eosin staining was performed.

| Adipocyte analysis
Adipocyte areas were measured using ImageJ software and the Adiposoft plugin 21,22 by an investigator blinded to treatment group.
Three random fields per sample were analysed. The average area per field of all measured adipocytes was used to calculate the adipocyte area. An average of 3000 cells per field was analysed (n = 4/group).
The frequency of adipocytes was calculated by combining all areas of measured cells per respective group and then distributing across four bins from 1000 to 4000 μm.

| RNA extraction and isolation
Total RNA from the white adipose and liver tissue was extracted

| Quantitative real-time PCR
The mRNA expression of peroxisome proliferator-activated receptor gamma (Pparg), nuclear factor erythroid 2 p45-related factor 2 (Nrf2), and the control gene TATA-box binding protein (Tbp) was assessed in each liver (n = 6/group) sample. For the white adipose tissue was determined using the delta delta C t (ΔΔC t ) method, 23,24 normalizing each sample to Tbp.

| Open field arena (OFA)
We assessed locomotion at three time points: prior to any experimen- TA B L E 1 White adipose tissue qRT-PCR genes acclimate for at least 60 minutes before the behavioural assessment.
Each animal was individually placed in the centre of the apparatus and allowed to freely move for 30 minutes. Behaviours were video recorded and tracked using the ANY-maze software (Stoelting Co.).
All OFAs were illuminated with approximately 175 lux. 25 Fecal boli were counted at the end of the 30 minutes. 70% ethanol was used to clean the apparatus between animals after fecal boli deposits were visually counted and numbers recorded. The total distance (locomotor activity) and immobile time were recorded. not exploring any of the arms was not included in the calculations.

| Elevated plus maze
The maze was cleaned with 70% ethanol solution after each session and allowed to dry between the sessions. No significant differences were detected in any of the EPM measurements (data not shown).

| Fecal boli
Fecal content was collected (n = 12/group) at four different time points, as depicted in Figure S1. The first collection (baseline) took 12955-4; Qiagen) per the manufacturer's protocol.

| Microbiome analysis
The V4 region of the 16S rRNA gene was PCR amplified from the extracted gDNA using the 515F and 806R primers (https://doi. DADA2 was also used to assign taxonomy to each amplicon sequence variant (ASV) using the Silva non-redundant database (v132).
Reads associated with ASVs that were present in less than 5% of the samples were removed from the analysis. Additionally, reads with either no phylum level classification or were associated with the Cyanobacteria phylum were removed due to low sample prevalence.
The final reads per ASV data for each fecal sample is provided in Table S1. Relative abundance, alpha diversity metrics, distance matrix calculations, and ordination analyses were performed using the Phyloseq R package (v1.30.0). 26 We used the analysis of variance of distance matrices test (Adonis) to detect significant differences in ordination clusters using the vegan R package (v2.5-6). Furthermore, ASV relative abundance data were used as input for predictive modelling using the random forest algorithm (ranger R package v0.12.1) in order to extract ASVs that were predictive of diet and light groups.

Accession number
Primer sequence

TA B L E 2 Liver qRT-PCR genes
Models were trained using 75% of the mouse samples either postdiet change (time points 2-4) or post-light change (time points 3 and 4). Models were tested on the remaining 25% of the data. All microbiome data and statistical analyses were conducted within the R programming language (v3.6.1). 27

| Serum hormones
Previously frozen serum was thawed on ice and assayed for serum

| Statistics (non-microbiome)
Non-microbiome-related statistical analyses were performed on GraphPad PRISM 8. Data are presented as mean ± standard error of the mean (SEM). Adipocyte area, hormones, behaviours, qRT-PCR, and fecal boli quantification were analysed using 2-way analysis of variance (ANOVA) followed by Sidak's multiple comparison test when significant interaction (P < .05) between diet and photoperiod were detected. Pre-photoperiod body weight was analysed using a 2-way ANOVA with repeated measures and post-photoperiod using 3-way ANOVA with repeated measures, both followed by appropriate post hoc tests (further described in results). Threshold for statistical significance was set at P < .05.

| Weight gain following isoflavone-free diet and photoperiod interaction
We tracked the body weight of every animal from their arrival to the end of the protocol. For analysis purposes, we split the weight measurements into pre-and post-photoperiod alterations ( Figure 1A,B).
A 2-way ANOVA with repeated measures of the pre-photoperiod data showed that the lack of isoflavones led to increased weight gain when compared to animals on the regular diet (P < .05). Sidak's multiple comparisons test revealed a significant change at days 4 (P < .005) and 6 (P < .005). The weight gain observed in animals on the IF diet was exacerbated by the longer inactive period ( Figure 1B). in animals on the NIF diet when compared those on the RD ( Figure 1E).

| Isoflavone content and length of photoperiod alter white adipose tissue metabolism markers
To investigate the potential cause of the adipose tissue remodelling differences were detected on CD36 mRNA transcripts ( Figure 2D).

| Isoflavone content and length of photoperiod alter gut microbiota composition
To assess whether the NIF diet and photoperiod length induced  Figure S2.
We next used machine learning to extract genera that were most important in differentiating between the RD and NIF groups. We found that our random forest model was able to predict diet group with 100% accuracy. The top 10 most predictive ASVs are shown in Figure S3. We found that much of the variation between the RD and NIF diet groups was attributed to ASVs within the Lachnospiraceae and Clostridiales family of bacteria. We also found that the single most predictive ASV belonged to the Muribaculaceae family which was highly abundant in the RD group (~5%-30% relative abundance), and was nearly ablated from the gut in the NIF group (<5% relative abundance  Figure S4). The relative abundance of the top 30 genera is available in Figure S5, for means and Figure S6 for individual animals.
Although we did not observe a major effect on microbial composition due to altered photoperiod within the RD or NIF diet groups (Adonis, P > .05, time points 3 and/or 4) (Figure 3), we were able to extract ASVs that were predictive of photoperiod group with 100% accuracy in the RD group and 91.7% accuracy for the NIF group. The top 10 predictive ASVs of photoperiod group within each diet group are shown in Figure S7. Overall, the majority of photoperiod-predictive ASVs were low in relative abundance (<2% relative abundance) and often absent in the samples. We additionally did not observe overt differences between photoperiod groups among the top 30 most abundant genera in our study ( Figures S5 and S6).

| Behaviour (OFA, EZM, fecal boli)
We assessed locomotion effects using the Open Field Apparatus (OFA, Figure S9). This test was performed at three different time

| Photoperiod-NIF interaction on endocrine profile (prolactin, GH, and TSH)
A multiplex ELISA analysis was performed to test the effect of shorter active photoperiod and isoflavones on the endocrine profile.
Thyroid-stimulating hormone levels were lower in the NIF diet when

| Liver Nfr2 and PPar-gamma transcripts
In view of the striking difference in adipocyte size and body weight, we then interrogated the effect of isoflavone and photoperiod duration on liver metabolism. We focused on the nuclear factor erythroid 2 p45-related factor 2 (Nrf2) due to its involvement in adipogenesis, adipocyte differentiation, and metabolic  metabolism, and ultimately, differentially regulate its energy balance.

| D ISCUSS I ON
Our results show that the gut microbiota composition is strongly altered by isoflavone consumption and the length of the photoperiod.
These alterations were accompanied by robust metabolic changes suggesting that isoflavones' anti-lipogenic effects may be driven by the alteration in the composition of the gut microbiota. 28 A diet free of isoflavones induced significant modifications in the bacterial phyla in vivo, as well as increased indicators of metabolic imbalance.
Previous work by our laboratory and others indicates that isoflavones have a substantial effect on body weight. 18,29,30 These studies show that animals maintained on an isoflavone-rich diet are lighter than those on an isoflavone-free (NIF) diet. This effect seems to be more prominent when the isoflavone concentration is higher. 29 In the present study, we were able to replicate the bodyweight results, as the mice in the NIF diet showed increased weight gain when compared to animals on a regular diet (RD). Our previous work showed that the RD contains considerable amounts of isoflavones (see Russell et al for exact amounts 18 ). In contrast, isoflavones were undetectable in the NIF diet.
Isoflavones are commonly found in soy products, and they are known to mimic estrogen actions. 18,31,32 They are plant derived nonsteroidal chemicals, which can bind both ERα and ERβ because of their conformational similarity to estradiol. 31,32 Isoflavones are also found in standard rodent chow. 18 Studies across different species have shown that dietary isoflavones play a beneficial role in reducing obesity and diabetes markers. 3,4 Estrogens have a dichotomous effect on body weight and adipose deposition. While estrogens are known to decrease body and adipose tissue weight, they have been shown to increase adipocyte deposition in the aging population, 33 as well as in pubertal and pregnant women. 34 However, several studies have shown that the lack of estrogen receptors, or the depletion of estrogen itself, increases adiposity and body weight. [35][36][37] In the current study, the lack of the estrogenic compound isoflavones, supports the idea that both estrogens and isoflavones are modulators of body weight; however, the mechanism by which this occurs is unknown. Nevertheless, several studies have confirmed the decrease in lipogenesis and weight, and increase in lipolysis caused by isoflavones. 29,38 This effect may involve the alteration of glucose metabolism by conformational alteration of the membrane-associated GLUT4 transporter. 39 Interestingly, we observed an increase in Glut-4 levels in the white adipose tissue mRNA transcript levels driven by the NIF diet, perhaps indicating a potential compensational increase of this gene to overcome an insufficient signalling mechanism. We also detected a marked increase in adipocyte size in the animals on the NIF diet compared to the RD group.
This finding is consistent with previous studies indicating that treatment with isoflavones decreases adipocyte size. 30 highlighting genistein as a ligand of Pparg. 42 In the current study, we detected a decrease of Pparg in the white adipose tissue in the NIF groups when compared to the regular diet groups, consistent with the increase in body weight and adipocyte area. These data suggest that the increase in body weight and adipocyte area may be due to a compromised lipid metabolism via Pparg. Several studies supporting these findings have shown that activation of PPars can improve lipid metabolism and insulin sensitization. 41,43 Interestingly, the levels of Pparg mRNA transcripts were higher in the animals exposed to the shorter active photoperiod, when compared the control 12L:12D, suggesting adipocyte dysfunction similar to recent studies in a rat model of obesity. 44 The lack of isoflavones also dysregulated the levels of lipoprotein lipase (Lpl) in white adipose tissue. Dysregulation of this lipase has been shown to result in obesity and insulin resistance. 45 We detected an increase in the Lpl RNA transcript levels in the animals exposed to the NIF diet when compared to animals in the regular diet. However, the animals in the regular diet that were in the longer inactive photoperiod had lower levels of Lpl transcripts. It is possible that the increase in Lpl in the NIF group could be leading the increase in body weight, as overexpression of Lpl in mice leads to obesity and insulin resistance. 45 Another gene linked to obesity and type 2 diabetes is CD36. CD36 is a protein that is upregulated in the subcutaneous tissue of obese men and women. There was a significant statistical interaction between diet and the length of the photoperiod in the current study. However, similar to previous studies in which no effects were observed in visceral adipose tissue, no further effects were observed, suggesting limited dynamic changes in non-subcutaneous adipose tissue. 46 Another indication of the effects of isoflavones on metabolism was the change in the expression of Nrf2 and Pparg in the liver. Nrf2 gene expression is involved in the regulation of adipogenesis and insulin resistance. Its involvement with the modulation of antioxidant signalling makes it a potential target for the prevention of metabolic syndrome and cardiovascular disease. 47 Additionally, adipocyte differentiation is compromised in the absence of Nrf2. 48 Our data suggest that the increase in liver Nrf2 transcripts may be involved in the adipocyte size anomaly following the removal of isoflavones.
We also detected a slight but significant increase in liver Pparg transcripts driven by more extended inactive periods. There is exten- Most studies have found a strong association between the gut microbiota composition and several diseases, including obesity. 28 In humans, for example, the composition of the gut microbiota is altered in obese and diabetic patients, 50 as well as in those with eating disorders. 51 The host-microbiome symbiotic relationship is critical for the maintenance of homeostasis and promoting heath. 52 For instance, the gut microbiota is required to ferment isoflavones, commonly found in soy food, to more biologically active compounds, which have a relatively higher affinity for ERβ in humans. 53 54,56 Additionally, the observed change in ratio of these two phyla was accompanied by a drop in gut diversity levels once isoflavones were removed from the diet. This gut diversity change has also been previously linked to an isoflavone-free diet. 57 Together, these results support the roles for the Firmicutes to Bacteroidetes abundance ratio and diversity metrics as a potential descriptive biomarkers of obesity. 50,58 In the laboratory, many of the physiological changes related to In the current study, we interrogated the changes in the gut microbi- In addition to the increase in locomotion, mice in the NIF group showed a decreased number of fecal boli and immobile time, suggesting a reduced response to anxiogenic environments. We have shown that ovariectomized female rats treated with estradiol and isoflavone-containing diet displayed higher levels of anxiety-like behaviours when compared to those placed on a NIF diet. 19 Based on this and previous studies, there is a possibility that isoflavones and estrogens may be interacting to ultimately regulate behaviours.
In the above-discussed study, Russell et al suggest that isoflavones differentially restrict the effects of estradiol, since the anxiogenic effects of estradiol are only present when isoflavones are on board.
This study is not without limitations. Firstly, animals were cohoused in groups of three. Each cage may have its own microenvironments with similar microbiota induced by cagemate allocoprophagy. 63 Indeed, we observed that animals from the same cage were more similar to each other (data not shown) than mice from other cages.
Despite this observation, we were still able to distinguish microbiome differences between study groups that corroborated with previous literature. 29,[38][39][40]64 Secondly, training accurate machine learning models requires a large number of training observations. Despite the limited number of samples in our study, we were able to discern predictive ASVs that should serve as fodder for future hypothesis-driven research. Our random forest model was able to predict that much of the variation between the RD and NIF diet groups was attributed to ASVs within the Lachnospiraceae and Clostridiales family of bacteria. Interestingly, the single most predictive ASV belonged to the Muribaculaceae family. This family was highly abundant in the RD group, and was nearly ablated from the gut in the NIF. Similar findings have also been observed in models of induced obesity and type 2 diabetes using the leptin-deficient mouse (ob/ob) as well as high-fat diet. 58 Therefore, the findings that were revealed via machine learning may be helpful when predicting potential metabolic disorders.
In summary, isoflavones have substantial effects on metabolism.
Due to isoflavones' potential benefits, several studies have investigated their ability to tackle obesity and type 2 diabetes, hyperlipidemia, hypercholesterolemia, and cardiovascular diseases. In the current study, we show that diet changes the composition of the gut microbiome, perhaps driving the observed metabolic changes.
Of note, it is imperative to consider changes in the gut microbiota composition when working with experimental protocols that involve dietary modifications. Based on our and others' data, behavioural and physiological deviations observed in a given experiment could be indeed driven by alterations in the gut microbial community following the diet modification. Finally, it is a safe practice to be conscientious of the composition of the standard diet used in the facilities where animals are housed, since some of the dietary compounds can change seasonally and consequently could interact with the drugs or manipulations studied, ultimately altering the gut microbiome and expected outcomes.

ACK N OWLED G EM ENTS
We are grateful to the following for technical support: Dr. Dennis

DATA AVA I L A B I L I T Y S TAT E M E N T
The data that support the findings of this study are openly available