We previously reported that at 4 weeks post-partum, dams consuming obesogenic diet had higher plasma triglyceride and insulin concentrations, increased body weight and adiposity relative to those on control diet (10). There was no significant difference in blood glucose concentrations between diet groups (10). To analyse impacts of diet and exercise on the maternal gut microbiota, we performed 16S rRNA amplicon sequencing on fecal samples collected from dams at this time (control diet sedentary: C_S, control diet exercised: C_Ex, obesogenic diet sedentary: O_S, and obesogenic diet exercised: O_Ex). We first examined several α-diversity measures across maternal diet groups. Both species evenness [F_{(1, 28)} = 11.1, p = 0.002; C_S: 0.69 ± 0.01, C_Ex: 0.69 ± 0.01, O_S: 0.65 ± 0.02, and O_Ex: 0.63 ± 0.02] and Shannon's diversity index [F_{(1, 28)} = 8.1, p = 0.008; C_S: 4.60 ± 0.13, C_Ex: 4.70 ± 0.10, O_S: 4.20 ± 0.14, and O_Ex: 4.20 ± 0.20] were significantly lower in dams fed obesogenic diet relative to control dams (Figure 1A). There was no significant difference in the number of OTUs between diet groups [F_{(1, 28)} = 2.6, p = 0.12; C_S: 738.4 ± 63.5, C_Ex: 829.3 ± 47.2, O_S: 649.2 ± 47.6, and O_Ex: 718.0 ± 88.1; Figure 1A).

To determine whether the overall gut microbial composition of obesogenic and control diet fed dams were different, we examined different β-diversity measures. Dams fed obesogenic diet clustered differently to control dams at the OTU level as observed by NMDS (Figure 1B). PERMANOVA analyses likewise confirmed significant differences in β-diversity between diet groups (t = 2.6, df = 1, 30, p = 0.001). Maternal diet did not alter sample dispersions (variation of Bray-Curtis similarities; PERMDISP: t = 1.0, df = 1, 30, p = 0.35).

We next examined sedentary dams to identify individual microbial taxa that differed between diet groups at the OTU level. LEfSe analysis identified 292 differentially abundant OTUs, while DESeq2 analysis identified 167 differentially abundant OTUs between diet groups. A total of 131 differentially abundant OTUs was consistent across both analyses, with 60 OTUs enriched and 71 OTUs decreased in O_S dams when compared to C_S dams. Several OTUs affiliated with genera Lactobacillus, Alistipes, and unclassified genera within Porphyromonadaceae and Lachnospiraceae families were decreased in O_S dams. On the other hand, OTUs affiliated with Blautia, Ruminococcus and Bacteroides were more abundant in O_S dams (Figures 1C,D). LEfSe analysis at the higher taxonomic levels identified Lactobacillus, Blautia and Bacteroides genera as differentially abundant between O_S and C_S dams (LDA score>4), which was consistent with the OTU results.

Distance based linear models (DistLM) were used to identify associations between the overall maternal gut microbial composition and several metabolic parameters. The metabolic parameters tested were: fat mass, blood glucose concentrations, plasma leptin, insulin and triglyceride (TG) concentrations at endpoint, and change in body weight over the experiment. The analysis was performed between metabolic parameters and a Bray-Curtis resemblance matrix of the overall gut microbiome at the OTU level. All the considered parameters, except for blood glucose, were significantly associated with the overall gut microbial composition at the OTU level (Table 1).

Further, we performed DistLM analyses to investigate whether specific microbial taxa were associated with these metabolic parameters. Here the analysis was performed between a Euclidean distance resemblance matrix of each parameter and the relative abundances of the top 150 OTUs. Several OTUs differentially abundant with diet were associated with the tested parameters. Lactobacillus_OTU34 (decreased 6-fold in O_S dams), Blautia_OTU5 (enriched 8-fold in O_S dams) and Bacteroides_OTU8 (enriched 3-fold in O_S dams) were associated with all tested metabolic parameters except blood glucose and insulin. Ruminococcus_OTU27 (enriched more than 7-fold in O_S dams) was associated with insulin and leptin. Only Lachnospiraceae unclassified_OTU47 (decreased 8-fold in O_S dams) was associated with blood glucose.

Dams had access to running wheels from 10 days prior to mating until delivery, and exercise levels did not differ between diet groups (10). At the time of sampling, maternal exercise had no discernible metabolic impacts in obesogenic or control diet dams at 4 weeks post-partum (10), however, it is possible that some impacts were present in the gut microbiota. We identified no significant effect of maternal exercise on α-diversity measures (species evenness, number of OTUs, or Shannon's diversity index; Figure 1A) in the dams.

To determine whether maternal exercise had an impact on the overall gut microbial composition, PERMANOVA analysis was conducted between exercised and sedentary dams in each diet group, separately (C_Ex vs. C_S, O_Ex vs. O_S). Pairwise PERMANOVA revealed C_Ex dams had significantly different gut microbial compositions compared to C_S dams (t = 1.2, df = 14, p = 0.041), in contrast, there was no significant difference between O_Ex and O_S dams (t = 0.98, df = 14, p = 0.50). The microbial communities of sedentary dams were more dispersed than those of exercised dams (PERMDISP: t = 3.2, df = 1, 30, p = 0.009). We also tested whether average distance run per day was associated with the gut microbial composition, and found no significant correlation (Table 1).

We then applied LEfSe and DESeq2 analyses between sedentary and exercised dams in each diet group, separately. Consistent across both analyses, one differentially abundant OTU was identified between C_Ex and C_S dams, while no differentially abundant OTUs were identified between O_Ex and O_S dams. Anaerostipes_OTU70 was decreased in C_Ex dams relative to C_S dams (DESeq2: log₂ fold change = −8.6, p = 0.0018; LEfSe: LDA>3, p = 0.018). At the higher taxonomic levels, LEfSe identified genera that were not affiliated with OTUs differentially abundant with maternal exercise such as Clostridium_XlVa (decreased in C_Ex relative to C_S; LDA>4) and Clostridium_XVIII (decreased in O_Ex relative to O_S; LDA>3).

The gut microbiota of male and female offspring at postnatal day (PND) 19 was examined using 16S rRNA amplicon sequencing. Maternal obesogenic diet was associated with higher blood glucose concentrations, final body weight and adiposity in offspring at PND19 (10). Maternal diet and exercise had sex-specific effects on blood glucose concentrations, plasma insulin and TG concentrations in the offspring (10). In male offspring, maternal exercise was associated with significantly lower blood glucose concentrations (O_Ex vs. O_S offspring). Insulin concentrations were lower in male offspring of both C_Ex and O_Ex dams, with no significant effect of maternal diet. In contrast, insulin concentrations were higher in female offspring of O_S vs. C_S dams, with no significant effect of maternal exercise. Plasma TG concentrations were significantly lower in male offspring of C_Ex vs. C_S dams. There were no significant effects of maternal exercise on blood glucose concentrations, plasma insulin or TG concentrations in female offspring of obesogenic and control diet dams (10). Here, we found no sex-specific differences in α-diversity measures (three-way ANOVA: offspring sex, maternal diet and maternal exercise). Thus, data from both sexes was aggregated to examine differences in α-diversity across diet and exercise. A two-way ANOVA indicated a significant interaction between maternal diet and exercise on the number of OTUs [F_{(1, 89)} = 7.4, p = 0.008; C_S: 330.5 ± 10.3, C_Ex: 255.0 ± 19.3, O_S: 203.0 ± 13.5, and O_Ex: 209.5 ± 15.8] and Shannon's diversity index [F_{(1, 89)} = 4.6, p = 0.034; C_S: 3.60 ± 0.08, C_Ex: 2.90 ± 0.13, O_S: 2.90 ± 0.13, and O_Ex: 2.80 ± 0.18] in offspring. Simple main effects analysis indicated that maternal exercise only significantly decreased both α-diversity measures in offspring of control dams (p ≤ 0.001 for both), as shown in Figure 2A. Maternal exercise, across diet groups, significantly decreased species evenness [F_{(1, 89)} = 5.8, p = 0.018; C_S: 0.61 ± 0.01, C_Ex: 0.53 ± 0.02, O_S: 0.55 ± 0.02, and O_Ex: 0.53 ± 0.03], with no significant effect of maternal diet (Figure 2A).

No sex-specific differences in gut microbial composition were identified globally (PERMANOVA: t = 0.78, df = 1, 91, p = 0.66), within each sub-group (PERMANOVA: C_S: t = 0.74, df = 1, 23, p = 0.72, C_Ex: t = 0.99, df = 1, 20, p = 0.39, O_S: t = 0.84, df = 1, 27, p = 0.54, and O_Ex: t = 0.62, df = 1, 15, p = 0.80), or in abundance of individual microbial taxa within each sub-group (LEfSe and DESeq2 analyses) between male and female offspring. Thus, data from both sexes was aggregated to examine differences in β-diversity across diet and exercise. The microbial compositions of offspring from obesogenic diet dams differed significantly when compared to offspring from control dams, as visualized by NMDS (Figure 2B), and confirmed using PERMANOVA analyses (t = 3.5, df = 1, 91, p = 0.001). Microbial communities in offspring from control dams were less dispersed than in offspring from obesogenic diet dams (PERMDISP: t = 4.5, df = 1, 91, p = 0.001).

LEfSe and DESeq2 analyses were employed on microbial abundance data from offspring of sedentary dams across both diet groups. Of the 394 differentially abundant OTUs detected by LEfSe, 82 were consistent with DESeq2 analysis. More OTUs were decreased (53 OTUs) than enriched (29 OTUs) in O_S relative to C_S offspring. OTUs affiliated with Clostridium sensu stricto, Clostridium_XlVa, Bacteroides, Blautia and Lactobacillus were enriched, while Alistipes, Oscillibacter and Alloprevotella OTUs were decreased in O_S offspring (Figures 2C,D). Consistent with the OTU results, LEfSe found Clostridium sensu stricto and Lactobacillus genera were enriched, while Alloprevotella was decreased in O_S vs. C_S offspring (LDA score>4).

The impact of maternal exercise on β-diversity was calculated between offspring of exercised and sedentary dams in each diet group, separately. PERMANOVA analyses revealed significantly different gut microbial compositions in offspring from C_Ex vs. C_S dams (t = 2.4, df = 45, p = 0.001). There was no significant difference in β-diversity between offspring from O_Ex and O_S dams (t = 1.1, df = 44, p = 0.20). The dispersion of microbial communities was similar between offspring from exercised and sedentary dams (PERMDISP: t = 0.46, df = 1, 91, p = 0.67).

We then investigated whether maternal exercise altered the abundance of specific microbial taxa in the offspring. LEfSe and DESeq2 identified 88 differentially abundant OTUs between C_Ex and C_S offspring. In contrast, no differentially abundant OTUs were identified between offspring of O_Ex and O_S dams. OTUs affiliated with genera Romboutsia, Clostridium sensu stricto and Proteus were enriched, whereas OTUs affiliated with Roseburia, Alistipes and unclassified genera within Porphyromonadaceae and Lachnospiraceae families were decreased in C_Ex offspring (Figure 2E). Of the 88 differentially abundant OTUs between C_Ex and C_S offspring, 30 were also differentially abundant between O_S vs. C_S offspring, as shown in Figure 2F. In both O_S and C_Ex offspring, OTUs affiliated with Clostridium sensu stricto and Clostridium_XlVa were enriched, while OTUs affiliated with Alloprevotella, Alistipes and Oscillibacter were decreased relative to C_S offspring (Figure 2F).

DistLM was used to analyse the association between the overall gut microbial composition and metabolic parameters in the offspring. The analysis was performed against metabolic parameters: final body weight, visceral fat, and blood glucose concentrations, as well as plasma leptin, and insulin and TG concentrations. The overall offspring gut microbial composition (Bray-Curtis resemblance matrix) was significantly associated with final body weight, visceral fat, glucose and leptin (Table 2).

The top 150 OTUs whose relative abundance was differentially affected by maternal diet and exercise were analyzed for associations with final body weight, visceral fat and blood glucose (Euclidean distance resemblance matrix) using DistLM. OTUs enriched in O_S offspring such as Lactobacillus_OTU7, Blautia_OTU54 and Bacteroides_OTU9 were associated with all the tested metabolic parameters. We also identified associations for OTUs altered in both O_S offspring (O_S vs. C_S) and C_Ex offspring (C_Ex vs. C_S) with metabolic parameters. Clostridium sensu stricto_OTU2 and Clostridium XlVa_OTU25 (enriched in O_S and C_Ex offspring) were associated with final body weight. Alloprevotella_OTU14, Alistipes_OTU49, and Oscillibacter_OTU55 (decreased in O_S and C_Ex offspring) were associated with all the tested metabolic parameters.

In order to infer changes in microbial metabolic pathways in the offspring gut microbiota resulting from maternal diet or exercise, functional content was predicted from amplicon data using PICRUSt. All offspring samples had low nearest sequenced taxon index (NSTI) values (average NSTI: 0.058 ± 0.003) indicating high prediction accuracy. Several predicted pathways were significantly enriched in offspring from obesogenic vs. control diet dams (Bonferroni corrected ANOVA with maternal exercise status nested as subclass, p < 0.05), among them, fructose and mannose metabolism (Table 3). In contrast, pathways related to indole alkaloid biosynthesis, α-linolenic acid metabolism and carotenoid metabolism were significantly decreased (Table 3). No significantly different pathways were identified after FDR or Bonferroni correction for maternal exercise (ANOVA with maternal diet type nested as subclass). We then analyzed offspring from control diet dams separately (offspring of C_Ex vs. C_S dams). Pathways related to vasopressin regulated water reabsorption (fold change = −3.0, p = 0.00064, q = 0.067), indole alkaloid biosynthesis (fold change = −2.2, p = 0.0011, q = 0.067), and betalain biosynthesis (fold change = −2.2, p = 0.0011, q = 0.067) were decreased, while those related to phosphotransferase system (fold change = 1.5, p = 0.00083, q = 0.067), and Staphylococcus aureus infection (fold change = 1.9, p = 0.0013, q = 0.067) were increased in C_Ex offspring.

This study is based on a cohort of female Sprague-Dawley rats and their offspring previously generated in our laboratory (10). Briefly, 6-week old female rats were assigned either standard laboratory chow (11 kJ/g, 13% fat, 22% protein, and 65% carbohydrate by energy; Gordon's Stockfeeds, NSW, Australia) or obesogenic diet ad libitum: two types of high-fat pellet (SF03-020: 20 kJ/g, 43% fat, 16% protein, and 41% carbohydrate by energy, and SF03-002: 22.8 kJ/g, 59% fat, 14% protein, and 27% carbohydrate by energy; Specialty feeds, Australia) supplemented with a selection of three different western foods (selected from cakes, potato chips, biscuits, meat pie, pasta with lard, and oats mixed with condensed milk). After 6 weeks, half the rats from each diet group were randomly assigned to voluntary exercise through introduction of a running wheel, while the other half remained sedentary with a locked wheel. Female rats were mated (with males fed standard chow) after 10 days of exercise. After delivery, dams were transferred to cages with no running wheel in order to prevent any interference of exercise with nurturing of pups, and to better model likely post parturition behavior. Dams were maintained on their pre-pregnancy diet throughout pregnancy and lactation.

The average litter size and male to female ratios were not significantly different across the groups. At postnatal day (PND) 1, litters were adjusted to 12 pups per mother (10); pups were allocated to groups so that litters were equally represented. Feces from the distal colon were collected from pups and dams at 3 and 4 weeks postpartum, respectively, at the time of tissue sampling for anthropometric and blood analyses. Fecal samples were immediately flash frozen in liquid nitrogen following collection and stored at -80°C. For the purposes of this study a subset of dams and pups (range 2–4 pups per dam with equal male to female ratios) were used.

DNA was extracted from the feces of dams (C_S: n = 9, C_Ex: n = 7, O_S: n = 10, and O_Ex: n = 7) using the PowerFecal DNA isolation kit (Catalog number: 12830-50, MO BIO Laboratories, Inc., Carlsbad, CA, USA), and offspring of both sexes (male: C_S: n = 12, C_Ex: n = 11, O_S: n = 13, and O_Ex: n = 11, and female: C_S: n = 13, C_Ex: n = 11, O_S: n = 16, and O_Ex: n = 10) using the PowerSoil DNA Isolation Kit (Catalog number: 12888-100, MO BIO Laboratories, Inc., Carlsbad, CA, USA). Extractions were performed according to the manufacturer's instruction. DNA concentration and quality was measured using a DeNovix DS-11 Spectrophotometer (DeNovix, Inc., Delaware, USA).

The composition of gut microbial communities was analyzed by Illumina amplicon sequencing of the 16S rRNA gene (performed by Ramaciotti Center for Genomics, UNSW Sydney) using DNA from dams (2 × 300 base pair MiSeq chemistry, V1-V3 region, 27F-519R primer pair) and offspring (2 × 250 base pair MiSeq chemistry, V4 region, 515F-806R primer pair). Sequence data were processed using Mothur, version 1.39.1 (55), which included removal of ambiguous bases and homopolymers longer than 15 base pairs, alignment with SILVA database, chimera checking with UCHIME, classification against the latest RDP Ribosomal Database training set (version 16_022016), and removal of singletons. Sequences were clustered into operational taxonomic units (OTU) at 97% nucleotide identity to generate an OTU table with the taxonomy and number of sequences per OTU in each sample. Commands were derived using the Mothur MiSeq SOP (56) and modified as required. The generated OTU table, combined with previously published anthropometric and biochemical measurements (10), were used as input in follow-up analyses.

Mother and offspring samples were rarefied down to 24,626 and 12,291 reads, respectively. OTU tables were standardized by dividing feature read counts by total number of reads in each sample. Standardized data were then square root transformed. All statistical analyses explored sex-specific differences in the offspring. One dam was identified as an outlier by Non-metric multidimensional scaling (NMDS). As 88% of the gut microbial composition in this dam was dominated by two OTUs (OTUs 1 and 2) from the genus Romboutsia, the dam and her offspring (2 male and 2 female offspring) were excluded from subsequent analyses.

Alpha diversity (number of OTUs, species evenness and Shannon's diversity index) analyses were performed using Calypso (57). Statistical analyses such as two and three-way ANOVA on diversity measures were performed using SPSS software (SPSS statistics version 22, SPSS Inc., an IBM Company). Simple main effects (with Bonferroni adjustment) was performed in the case of a significant interaction between variables on α-diversity. Results are expressed as mean ± SEM and were considered significant if p ≤ 0.05. GraphPad Prism 7 was used to plot diversity results. In bar graphs, symbols above horizontal line indicate main effects: ^**p ≤ 0.01 maternal diet effect and ^#p ≤ 0.05 maternal exercise effect, while symbols above bar indicate simple main effects: ^##p ≤ 0.01 maternal exercise effect.

NMDS plots, Permutational Multivariate Analysis of Variance (PERMANOVA) and Permutational Analysis of Multivariate Dispersions (PERMDISP) were generated using a Bray-Curtis resemblance matrix on PRIMER (Primer-E Ltd., Plymouth, United Kingdom) (58). To identify OTUs differentially abundant between groups we used two different statistical analyses. Linear Discriminant Analysis (LDA) Effect Size (LEfSe) (59) was performed using the Galaxy web application (60) using default settings. The R package Phyloseq (61) was used for the negative binomial Wald test in DESeq2 (62). P-values were adjusted for multiple testing using Benjamini Hochberg false discovery rate correction in DESeq2. LEfSe was performed across all taxonomic levels (phylum to OTU level) whereas DESeq2 was performed at the OTU level. Statistical significance was defined as p ≤ 0.05 in both statistical methods. Only OTUs identified by both LEfSe and DESeq2, and those with relative abundances consistently represented across more than 50% of rats in a group were discussed.

Distance-based linear models (DistLM) analysis was performed using PRIMER to examine associations between the gut microbiota and metabolic parameters. When the analysis was performed between the overall gut microbial composition and metabolic parameters, a Bray-Curtis resemblance matrix of the overall gut microbial composition at the OTU level was used. The relative abundances of the top 150 OTUs and Euclidean distance resemblance matrix of each metabolic parameter was used to find associations between specific OTUs and metabolic parameters. In both cases DistLM was performed using a forward-stepping selection procedure and Akaike Information Criterion selection criterion (999 permutations). Phylogenetic Investigation of Communities by Reconstruction of Unobserved States (PICRUSt) was performed using Galaxy web to generate a profile of putative functions (through metagenomic prediction) from the 16S rRNA OTU data (63). Taxonomic classification was performed against Greengenes 13.5, and pathway counts were compared across groups using nested ANOVA with Bonferroni correction. Subgroup comparisons were conducted using Wilcoxon test with FDR (FDR corrected p-values represented as q).

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.