Timing of standard chow exposure determines the variability of mouse phenotypic outcomes and gut microbiota profile

Abstract

Standard chow diets influence reproducibility in animal model experiments because chows have different nutrient compositions, which can independently influence phenotypes. However, there is little evidence of the role of timing in the extent of variability caused by chow exposure. Here we measured the impact of different diets (5V5M, 5V0G, 2920X and 5058) and timing of exposure (adult exposure (AE), lifetime exposure (LE) and developmental exposure (DE)) on growth and development, metabolic health indicators and gut bacterial microbiota profiles across genetically identical C57BL/6J mice. Diet drove differences in macro- and micronutrient intake for all exposure models. AE had no effect on phenotypic outcomes. However, LE mice exhibited significant sex-dependent diet effects on growth, body weight and body composition. LE effects were mostly absent in the DE model, where mice were exposed to chow differences only from conception to weaning. Both AE and LE models exhibited similar diet-driven beta diversity profiles for the gut bacterial microbiota, with 5058 diet driving the most distinct profile. However, compared with AE, LE effects on beta diversity were sex dependent, and LE mice exhibited nine times more differentially abundant bacterial genera, the majority of which were inversely affected by 2920X and 5058 diets. Our findings demonstrate that LE to different chow diets has the greatest impact on the reproducibility of several experimental measures commonly used in preclinical mouse model studies. Importantly, weaning mice from different diets onto the same diet for maturation may be an effective way to reduce unwanted phenotypic variability among experimental models.

Standard rodent (chow) diets are grain- and cereal-based feeds¹ that contain adequate nutrition but vary greatly in macro- and micronutrient composition². While ingredient lists are usually provided, the sources and proportions of the ingredients can vary from diet to diet and batch to batch and are usually proprietary^2,3. This means that the absolute nutrient content of standard chows is often unknown unless directly measured. For micronutrients in particular, the amounts contributed by vitamin and mineral mixes are provided, but the amounts contributed by whole food ingredients are not, meaning the true content is often higher than expected⁴. Since chow diets are formulated to benefit different aspects of mouse models (such as growth, breeding and maintenance), there is no universally recommended standard chow across research institutions, resulting in a broad range of standard chow usage among studies. Differences in nutrient intake and other naturally occurring bioactive compounds can independently influence phenotypic outcomes^5–8; therefore, the use of different chows between labs or model development steps is a major problem for the reproducibility of research findings. This is especially problematic for studying preclinical animal models of disease, which often require accurate repeated measures of the isolated phenotypic effects of genotype, surgery or pharmaceutical drug exposure. For example, in mice with acetaminophen-induced liver injury, differences in chow diet determined the extent of liver damage⁹. In another mouse study, standard chow composition determined host immunity response to influenza¹⁰.

Previous studies also show how chow diet differences alone can drive significant differences in mouse model phenotypes. A hexokinase II knockdown study demonstrated significant effects of chow on mouse body weight, heart weight and hexokinase phenotypes¹¹. Another example comparing the effects of standard chow with those of a purified diet (AIN-76A) found that the chow diet drove differences in food intake and plasma essential amino acid levels in inbred C57BL/6 mice¹². Lastly, differences in chow diets with variable fiber sources were found to induce differences in the distribution of gut bacterial communities in outbred CD-1 mice¹³.

Epigenetic changes are often implicated as mechanistic drivers of the phenotypic variability caused by diet. Although this has not been extensively studied for chow diets, several studies using mouse models demonstrate the effects of isolated nutrients. For example, a high-fat diet (HFD) induced histone modifications in mouse adipose tissue that were then linked to metabolic syndrome-like phenotypes¹⁴. A separate study showed that HFD induced epigenetic modifications in the mouse brain that caused learning and memory deficits¹⁵.

Most studies on the effects of chow diet have focused on adult exposure (AE), but research in animal models and human populations shows that developmental windows are particularly susceptible to dietary changes¹⁶. Data demonstrating the effect of diet timing on outcome within the same study are limited, but a recent study in Swiss Webster mice found that postnatal exposure to HFD at different stages of life had differing effects on diabetic outcomes¹⁷. HFD exposure across the lifespan (postnatal day (PND) 2–325) resulted in the greatest impact, while developmental exposure (DE) between birth and weaning (PND 2–21) had the second greatest impact¹⁷. Surprisingly, slightly longer DE, including pre- and postweaning (PND 2–35), had the least impact¹⁷.

Here, we measured the phenotypic impacts of exposing mice at different life stages to commonly used standard chows (5V5M, 5V0G, 2920X and 5058). Three different exposure windows were compared using a multigenerational parallel treatment scheme with inbred mice to limit effects to those within a single genetically and environmentally controlled population of mice: (1) AE, that is, parents; (2) lifetime exposure (LE), that is, offspring; and (3) DE, that is, siblings of LE offspring. Females and males were tested to address sex effects on the most common preclinical mouse model strain background, C57BL/6J. Differences in the macro- and micronutrient composition of chow diets were measured, and the impacts of diet and timing of exposure on nutrient intake, growth and development, metabolic health indicators and gut bacterial microbiota composition were assessed.

Results

Measured values for micronutrients are variable among standard chow diets and exceed recommended levels

To assess differences in nutrient exposure from the chow diets, we measured nutrient levels in the four chows fed to the mice in this study (Supplementary Table 1). We compared these with the manufacturer’s estimates on the diet specification sheets (Supplementary Table 2). Macronutrient composition and the levels of two micronutrients with important roles in development (vitamin D and folic acid) were assessed. We found that total measured calories, carbohydrate, fat and protein content were relatively similar among the four diets except for 2920X, which contained the lowest total calories and fat and the highest carbohydrate content (Fig. 1a,b). This was consistent with the manufacturer’s estimates (Fig. 1a,b). Measured fiber content was variable among the diets, with 2920X containing the highest total and soluble fiber and 5V0G containing the highest crude fiber (both ~1.3× higher than the lowest, 5V5M) (Fig. 1c). Only crude fiber levels were provided in the manufacturer estimate; the pattern across diets was similar to the measured values (Fig. 1c).

Fig. 1 |. Standard chow diet composition and dietary treatment scheme.

a–e, The macro and micronutrient estimates provided by the manufacturer were compared with measured values for total calories (a); macronutrients (fat, carbohydrates and protein) (b); fiber (only crude fiber content was provided by the manufacturer) (c); vitamin D (only vitamin D3 levels were provided by the manufacturer) (d); and folic acid (e). f, Standard chow dietary treatment scheme for AE, LE and DE models. AE model: C57BL/6J mice at 6 weeks of age were equally divided across four standard chow models (5V5M, 5V0G, 2920X and 5058) for 3 weeks of chow diet acclimation. After 3 weeks, a subset of AE females was euthanized for micronutrient measures. The remaining AE females and males were mated in breeder trios based on chow diet assignment. AE males were euthanized after 3 weeks of mating, and females were euthanized at weaning. LE and DE models: at weaning, pups were caged by litter and sex, and each litter was divided between LE and DE. LE mice were weaned onto the diet assignment of their parents, while DE mice were weaned onto the 5V5M diet and remained on this diet until the end of the study.

For each diet, vitamin D and folic acid concentrations were more variable than the macronutrients and less consistent with the manufacturer’s estimates, which are based on added micronutrients and do not account for contributions from other ingredients. For vitamin D, 1 IU/g of feed is considered adequate and potentially even a considerable excess for laboratory mouse chow^18,19. Total vitamin D content (D2 + D3) measured for our standard chows ranged from 2.3 IU/g (2920X) to 4.2 IU/g (5V0G). This was considerably higher (~2–4×) than the recommended amounts and higher than the manufacturer’s estimates of 1.5 IU/g (2920X) to 3.4 IU/g (5058) (Fig. 1d). Vitamin D2 contributed to this outcome since the manufacturer’s estimate provided only vitamin D3 values. However, our measured vitamin D3 values also differed from the manufacturer’s estimate (Fig. 1d). The recommended dietary folic acid intake for mice is 2 μg/g of diet²⁰. Measured folic acid values were lower than the manufacturer’s estimates and consistent with the recommended value (~2.0–2.3 μg/g), except for 5V0G. The measured folic acid level in 5V0G was similar to the manufacturer’s value and 4× higher than the other diets and the recommended value, making it more similar to a folic acid-supplemented diet (10 mg/kg (ref. 21)) (Fig. 1e).

Standard chow diets impact nutrient exposure differently depending on the diet and timing of exposure

We first assessed the effect of AE to different standard chow diets. Female and male C57BL/6J mice were fed one of the four standard chows that differed in nutrient composition (Fig. 1f, Supplementary Tables 1 and 2 and Supplementary Data 1). After 3 weeks of acclimation to the diet, females were randomly split into two groups: group 1 was euthanized to measure nutrient status after diet acclimation but before mating, and group 2 remained on the diet while mating with males to measure nutrient status and phenotypic outcomes after breeding (Fig. 1f).

Nutrient exposure is determined in part by nutrient intake, which is a combination of dietary content and food intake. Food intake in the AE model was relatively similar among the diets, except for 5058, which showed a ~1.6× higher food intake than 5V5M (the lowest food intake) in females before mating (Fig. 2a,b and Supplementary Fig. 1a,b). Less variability was observed in males before mating and during breeding when males and females were combined (Fig. 2a,b and Supplementary Fig. 1a,b). Intake of total calories, macronutrients and fiber mirrored the food intake variability among the diets, with 5058 exhibiting the highest nutrient intake in females before mating (Fig. 2c–e). By contrast, micronutrient intake more closely mirrored the diet composition (Fig. 2f,g). Vitamin D intake was more variable in females than in males before mating, with ~2.3× difference between the highest (5058) and lowest (2920X) intake (Fig. 2f). Folic acid intake was highly variable before and after breeding, with a ~4× difference between the highest (5V0G) and lowest (5V5M) after breeding (Fig. 2g). These differences in vitamin D and folic acid intake did not drive significant differences in serum vitamin D (25(OH) D) or folate concentrations (Supplementary Fig. 2).

Fig. 2 |. Standard chow diets impact food intake and nutrient intake for the AE model.

a,b, Food intake was measured weekly for each cage, and average daily values were calculated as g/mouse/day for females and males during diet acclimation (a) and breeding (b). Sample sizes in order from left to right for diet acclimation: n(F) = 9, 9, 9 and 9, n(M) = 2, 2, 2 and 2 and breeding n = 6, 6, 6 and 6. c–g, Average nutrient intake was calculated on the basis of food intake and measured macro- and micronutrient values, and reported as total calories (c); macronutrients (fat, carbohydrates and protein) (d); fiber (e); vitamin D (f); and folic acid (g). Sample sizes in order from left to right for diet acclimation: n(F) = 9, 9, 9 and 9, n(M) = 2, 2, 2 and 2 and breeding n = 6, 6, 6 and 6. Diet effects were calculated by one-way ANOVA or Kruskal–Wallis test and significant when *P < 0.05. n.d., P-value not determined due to small sample size. F, female; M, male.

TS: edit ‘Male’ to ‘male’Offspring generated from breeders in the AE model were used to assess the effects of exposure timing (DE and LE (development + adult)) on the response to the different chow diets. LE mice were exposed to one of four standard chow diets from conception to adulthood (Fig. 1f). DE mice were siblings of LE mice that were exposed to different diets only from conception to weaning and then placed on the same 5V5M diet until adulthood (Fig. 1f).

Consistent with findings in the AE model, food intake in the LE model significantly differed among the diets, with 5058 mice eating the most and 5V5M the least (Fig. 3a and Supplementary Fig. 3a). However, in contrast to the AE model, LE males and females were both affected. Less variability in food intake among the diets was observed in the DE model, although 5058 remained the diet with the highest intake and 2920X intake was seemingly reduced compared with other diets (Fig. 3b and Supplementary Fig. 3b). The variability in nutrient intake among the diets in the LE model was also very similar to the AE model, except both males and females were affected (Fig. 3c–g). Intake of both macro- and micronutrients in the DE model (where mice were fed the same 5V5M diet from weaning to adulthood) closely mirrored the patterns of food intake, with the highest nutrient intakes for 5058–5V5M mice and the lowest for 2920X-5V5M mice (Fig. 3c–g). In contrast to the AE model, chow diet differences in the LE model drove significantly different serum vitamin D (25(OH)D) concentrations (Supplementary Fig. 4a). However, serum 25(OH)D differences did not reflect vitamin D intake (Fig. 3f). Instead, 2920X, which had the lowest vitamin D intake, exhibited the highest mean serum 25(OH)D levels and 5V5M exhibited the lowest (Supplementary Fig. 4a). Different chow diets did not drive significant differences in serum vitamin D for the DE model nor serum folate concentrations for the LE or DE model (Supplementary Fig. 4b–d).

Fig. 3 |. Standard chow diets impact food intake and nutrient intake for LE and DE models postweaning.

a,b, Food intake was measured weekly for each cage, and average daily values were calculated as g/mouse/day for females and males for LE (a) and DE (b) models. Sample sizes in order from left to right: for LE n(F) = 8, 6, 6 and 6 and n(M) = 12, 6, 10 and 5 and for DE n(F) = 4, 4 and 4 and n(M) = 4, 8 and 3. c–g, Average nutrient intake was calculated on the basis of food intake and measured macro- and micronutrient values, and reported as total calories (c) ; macronutrients (fat, carbohydrates and protein) (d); fiber (e); vitamin D (f); and folate (g). LE and DE models were analyzed separately. Sample sizes in order from left to right: for LE n(F) = 8, 6, 6 and 6 and n(M) = 12, 6, 10 and 5 and for DE n(F) = 4, 4 and 4 and n(M) = 4, 8 and 3. The error bars represent the standard error of the mean (s.e.m.). Diet effects were calculated by one-way ANOVA or Kruskal–Wallis test and significant when *P < 0.05. Different letters or letter combinations indicate groups that differ significantly.

LE to different standard chow diets drives the greatest variability in offspring growth and adiposity

Given the substantial differences in macro- and micronutrient intake among the four diets tested, we assessed the effects of the different diets on phenotypes in the AE, LE and DE models. Despite the large differences in nutrient intake among chow diets in the AE model, up to 11 weeks of adult chow diet exposure had no significant effect on metabolic health indicators in females, including body weight (Fig. 4a and Supplementary Fig. 5a), perigonadal white adipose tissue (PWAT) weight (Fig. 4b) or fasting blood glucose (Fig. 4c). Males were similarly unaffected (Fig. 4d–f and Supplementary Fig. 5b), although statistical testing was not possible due to the small sample size (n = 2 per diet). The AE model also showed no significant effects of diet on breeding success metrics (Table 1). Thus, AE to these chow diets during breeding probably had minimal effects on these phenotypes.

Fig. 4 |. No effect of standard chow diets on AE model phenotypes.

a, The change in female body weight from diet start to diet end. b, Female PWAT weight and PWAT weight relative to body weight determined after euthanasia. c, Fasting blood glucose was measured in live females following a 12 h fast. d–f, Similarly, body weight (d), PWAT (e) and fasted blood glucose (f) were assessed in male mice. Sample sizes in order from left to right: n(F) = 4, 4, 4 and 4 and n(M) = 2, 2, 2 and 2. The error bars represent the standard error of the mean. Diet effects were calculated by one-way ANOVA or Kruskal–Wallis test and were considered significant when *P < 0.05.

Table 1 |. No effect of standard chow diets on AE model breeding success.

			Average litter size ± s.e.m.
Diet	No. of litters	Fecundity (%)	At birth	At weaning	Litters surviving to weaning (%)	Pups surviving to weaning (mean % ± s.e.m.)	Males at weaning (mean % ± s.e.m.)
5V5M	4	100	6.8 ± 0.9	5.3 ± 0.3	100	81.7 ± 0.1	57.5 ± 8.5
5V0G	4	100	7.0 ± 0.4	6.7 ± 0.7	75	71.4 ± 0.2	50.0 ± 9.6
2920X	4	100	7.3 ± 0.9	7.0 ± 0.9	100	96.4 ± 0.0	65.2 ± 16.7
5058	4	100	6.5 ± 1.0	6.3 ± 0.3	75	65.6 ± 0.2	43.7 ± 20.6
Diet effects		–	P = 0.92	P = 0.43	–	P = 0.58	P = 0.75

Litter size at weaning does not include litter sizes of zero, which were litters lost due to death or illness between birth and weaning. Diet effects were calculated by one-way ANOVA or Kruskal–Wallis test and significant when *P < 0.05.

During development, when LE and DE mice were siblings housed together, there was a significant effect of chow diet on male and female body weight as early as PND 5, where mice on the 5V5M diet showed 1.2× lower body weight compared with mice on the other three diets (Fig. 5a). By weaning, 5V5M-fed mice still weighed the least but were similar to mice fed 2920X, while mice fed 5V0G and 5058 exhibited higher body weights (Fig. 5b). By adulthood, in the LE model (where mice remained on the different chows postweaning), 5V5M-fed mice still weighed the least, but a significant diet effect was detected only for male mice and was characterized by higher body weights for 5058-fed mice (Fig. 5c). By contrast, adult mice in the DE model (where mice were on different diets only until weaning) exhibited no significant differences in body weight (Fig. 5d). Crown–rump length (CRL), a measure of growth, showed similar (but not significant) pattern of differences as body weight in the LE and DE model (Fig. 5c,d), except for adult LE females, where 5058-fed mice exhibited significantly shorter CRL (Fig. 5c). Thus, exposure to different standard chow diets induces variability in mouse model developmental growth phenotypes in a sex-dependent manner; however, placing mice on the same diet at weaning reduces this variability.

Fig. 5 |. Standard chow diets alter body weight and CRL in LE model but not DE model.

a–d, Body weight and CRL measures were measured on live offspring at PND 5 (a) and weaning (PND 25–33) (b) when LE and DE mice were caged together, and adult (12 weeks) time points for LE (c) and DE (d), where LE and DE models were caged separately. Sample sizes in order from left to right for neonatal and weaning: n(F) = 8, 10, 10 and 10 and n(M) = 12, 10, 18 and 8. Sample sizes in order from left to right for adulthood: for LE n(F) = 8, 6, 6 and 6 and n(M) = 12, 6, 10 and 5 and for DE n(F) = 4, 4 and 4 and n(M) = 4, 8 and 3. The error bars represent the standard error of the mean. Diet effects were calculated by one-way ANOVA or Kruskal–Wallis test and significant when *P < 0.05. Different letters or letter combinations indicate groups that differ significantly.

To determine whether diet-induced differences in developmental growth phenotypes are associated with differences in adult phenotypes that are known metabolic health indicators, we measured the effects of diets on adult body composition, fat depot size and glycemic status in the LE and DE models. In the LE model, 5058-fed females had 1.6× higher total fat mass than 2920X-fed females, but no difference in lean mass (Fig. 6a). By contrast, 5058-fed males had no difference in fat mass but had significantly higher lean mass compared with males fed all other diets (Fig. 6a), which was consistent with the body weight differences (Fig. 5c). In the DE model, we detected no significant effects of preweaning diet on fat mass in females or males (Fig. 6b). However, consistent with the LE model, 5058-fed males in the DE model (5058–5V5M males) exhibited significantly higher total lean mass compared with males fed other diets (Fig. 6b), which was consistent with the body weight differences (Fig. 5d).

Fig. 6 |. Standard chow diets alter adult body composition and metabolic health outcomes for LE and DE models.

a–d, EchoMRI measured body composition for 8–9-week-old female and male offspring. Total fat and total lean mass were measured for LE (a) and DE (b) models. Fat-to-lean-mass ratios were calculated for LE (c) and DE (d) models. e,f, Perigonadal fat pad weights (PWAT) were collected for LE males and females (12–14 weeks and 15–16 weeks, respectively) (e) and DE males and females (12–14 weeks and 15–16 weeks, respectively) (f). g,h, Fasted whole-blood glucose was measured in live LE males and females (12–14 weeks and 15–16 weeks, respectively) (g) and live DE males and females (12–14 weeks and 15–16 weeks, respectively) (h). Sample sizes in order from left to right: for LE n(F) = 8, 6, 6 and 6 and n(M) = 12, 6, 10 and 5 and for DE n(F) = 4, 4 and 4 and n(M) = 4, 8 and 3. LE and DE models were analyzed separately. The error bars represent the standard error of the mean. Diet effects were calculated by one-way ANOVA or Kruskal–Wallis test and considered significant when *P < 0.05. Different letters or letter combinations indicate groups that differ significantly.

Total fat-to-lean-mass ratios are considered a predictor of metabolic disease, and lower ratios are considered protective while higher ratios suggest an increased risk of cardiac events and death²². In the LE model, diet drove significant differences in fat-to-lean-mass ratios in females, with 2920X-fed mice exhibiting ~1.7× lower ratios than mice fed the other diets (Fig. 6c). The DE model did not exhibit any diet-induced differences in fat-to-lean-mass ratios, although 2920X-fed females had the lowest ratios (Fig. 6d).

The perigonadal fat pad is usually the largest white adipose tissue depot in the mouse, and larger size is associated with impaired metabolic health²³. In the LE model, 5058- and 5V0G-fed females exhibited ~2× larger PWAT weights than 2920X-fed mice; these differences remained significant even after normalizing to body weight (Fig. 6e). These effects partly mirrored total body fat mass differences (Fig. 6a). There were no significant diet-induced differences in PWAT measurements for LE males (Fig. 6e) or DE males or females (Fig. 6f). There were also no diet effects on glycemic status in the LE model (Fig. 6g) or the DE model (Fig. 6h).

Consistent with growth phenotypes, these findings show that chow diet differences can drive variability in metabolic health phenotypes that is sex specific and may be ameliorated by weaning mice onto the same diet.

Mice with LE to different standard chows are more susceptible to altered composition of the gut bacterial microbiota

To investigate the impact of different standard chow diets on the composition of the gut bacterial microbiota, we performed 16S ribosomal RNA sequencing of DNA from cecal contents. Since mice in the LE model exhibited the greatest phenotypic response, we assessed microbiota changes in this group compared with the parental AE model, which exhibited no detectable phenotypic response.

Alpha diversity describes microbial richness and how evenly distributed microorganisms are within a sample; the Shannon index is used as a measure of alpha diversity²⁴. In the AE model, females exhibited significant diet-induced differences in microbial richness, with 5058-fed females exhibiting the highest number of observed sequence variants and 2920X-fed mice exhibiting the lowest (Fig. 7a). AE males were not similarly affected (Fig. 7a), and there were no diet effects on the AE model’s Shannon index for both male and female mice (Fig. 7b). By contrast, LE mice exhibited no diet effects on microbial richness (Fig. 7c), but diet significantly altered the Shannon index, with 5058- and 5V0G-fed female mice exhibiting significantly lower Shannon indices than 2920X-fed mice (Fig. 7d).

Fig. 7 |. Standard chow diet alters the alpha diversity of gut bacterial microbiota in the AE and LE models.

a,b, Measures of alpha diversity of gut bacterial microbiota in the AE model with observed sequence variants (a) and Shannon index (b). c,d, Measures of alpha diversity of gut bacterial microbiota in the LE model with observed sequence variants (c) and Shannon index (d). Box plot elements include median (center line), upper 75th quartile and lower 25th quartile and expected variation of the data (whiskers). The open circles indicate female samples, and the closed circles indicate male samples. Sample sizes for 5V5M, 5V0G, 2920X, and 5058: for AE n(F) = 4, 4, 4 and 4 and n(M) = 2, 2, 2 and 2 and for LE n(F) = 6, 6, 6 and 6 and n(M) = 5, 6, 6 and 4. Main diet effects were calculated by linear regression and considered significant when *P < 0.05. Different letters or letter combinations indicate groups that differ significantly.

Beta diversity describes the variability in microorganisms across samples within a population²⁴. The effects of chow diets on the beta diversity of the gut bacterial microbiota were determined using principal coordinate analysis (PCoA) and permutational multivariate analysis of variance (PERMANOVA) with Bray–Curtis distances. In the AE model, diet and sex drove significant differences in beta diversity, with PC1 explaining 25.1% of the variability and PC2 explaining 13.1% (Fig. 8a). Among the four diets, 5058-fed mice clustered furthest away from the other three (Fig. 8a). There was no sex × diet interaction effect, and when the samples were stratified by sex, there was a similar clustering of diets but no detectable diet effect, probably due to the small sample size (Fig. 8b,c). In the LE model, diet drove significant differences in beta diversity that were similar to the AE model, with similar amounts of variance explained on PC1 and PC2 (23.2% and 11.6%, respectively) and with distinct clustering of 5058-fed mice (Fig. 8d). However, there was also a significant sex and sex × diet interaction effect detected in the LE model (Fig. 8d). When stratified by sex, females and males exhibited a significant diet effect, with 5058-fed mice remaining distinct from mice fed the other diets (Fig. 8e,f).

Fig. 8 |. Standard chow diet alters the beta diversity of gut microbiota for AE and LE models.

a–c, Principal coordinates analysis (PCoA) in the AE model, using two-way PERMANOVA and Bray–Curtis distances (a) and using one-way PERMANOVA and Bray–Curtis distances (female only, b; male only, c). d–f, PCoA in the LE model, using two-way PERMANOVA and Bray–Curtis distances (d) and using one-way PERMANOVA and Bray–Curtis distance (female only, e; male only, f). Sample sizes for 5V5M, 5V0G, 2920X and 5058: for AE n(F) = 4, 4, 4 and 4 and n(M) = 2, 2, 2 and 2 and for LE n(F) = 6, 6, 6 and 6 and n(M) = 5, 6, 6 and 4. Main diet or sex effects were calculated by linear regression and considered significant when *P < 0.05.

Finally, we investigated the impact of diet on the differential relative abundance of individual bacterial genera using a multifactor model (diet + sex) that was validated using Analysis of Compositions of Microbiomes with Bias Correction 2 (ANCOM-BC2)²⁵ to identify diet effects after adjustment for sex effects. In the AE model, chow diet significantly altered the abundance of three bacterial genera (Clostridium sensu stricto 1, Ruminococcus and Tuzzerella) similarly for males and females (Fig. 9a–c). ANCOM-BC2 confirmed these effects (Supplementary Data 2). Genera altered by AE were also differentially abundant in the LE model, but with differences between the sexes (Fig. 9d–f). 5058-fed mice exhibited an increased abundance of Clostridium sensu stricto 1 in AE males and females, but only females were affected in the LE model (Fig. 9a,d). 5058- and 5V0G-fed mice also exhibited an increased abundance of Ruminococcus in AE males and females, but only males were affected in the LE model (Fig. 9b,e). On the other hand, 5058-fed mice exhibited the lowest abundance of Tuzzerella in males and females in both models, but overall abundance was substantially lower for all diets in the LE model versus the AE model (Fig. 9c,f). ANCOM-BC2 confirmed the differential abundance of all genera except Ruminococcus (Fig. 9b,e), which yielded P < 0.05 but failed to pass bias correction (Supplementary Data 2 and 3).

Fig. 9 |. Differentially abundant bacterial genera affected in both AE and LE models after adjusting for sex.

a–f, The diet-dependent abundance of Clostridium sensu stricto 1 (a and d), Ruminococcus (b and e) and Tuzzerella (c and f) in the AE model (a–c) and the LE model (d–f) after multifactor analysis of microbial differential abundance with diet as the primary effect and sex as a covariate. Box plot elements include median (center line), upper 75th quartile and lower 25th quartile and expected variation of the data (whiskers). The open circles indicate female samples, and the closed circles indicate male samples. Sample sizes for 5V5M, 5V0G, 2920X and 5058: for AE n(F) = 4, 4, 4 and 4 and n(M) = 2, 2, 2 and 2 and for LE n(F) = 6, 6, 6 and 6 and n(M) = 5, 6, 6 and 4. FDR shown for main diet effects after adjustment for sex and significant when FDR <0.05.

The LE model exhibited significant diet effects on the abundance of 24 additional bacterial genera (27 total; Supplementary Table 3). ANCOM-BC2 confirmed all except three genera (Muribaculaceae, Eubacterium xylanophilum group and Eubacterium ventriosum group), which failed to pass correction for multiple testing (Supplementary Data 3). Interestingly, females showed larger differences in genus abundance with 5058- and 2920X-fed females exhibiting the greatest effect sizes (Supplementary Table 3). Two-way hierarchical clustering of all 27 genera confirmed that samples clustered in part by diet, with the most distinct clusters observed for 5058- and 2920X-fed females in the LE model (Supplementary Fig. 6). This distinct clustering is seemingly driven by five genera with greater abundance in 5058-fed mice and lower abundance in 2920X-fed mice (Eubacterium xylanophilum group, Clostridium sensu stricto 1, Lactobacillus, Bifidobacterium and Eubacterium nodatum group) (Supplementary Table 3 and Supplementary Fig. 6) and 12 bacterial genera with lower abundance in 5058-fed mice but higher abundance in 2920X-fed mice (Anaeroplasma, Tuzzerella, Eubacterium ventriosum group, Acetatifactor, Blautia, Clostridia vadinBB60 group, Alistipes, Colidextribacter, Oscillibacter, Intestinimonas, Lachnospiraceae FCS020 group and Butyricicoccaceae UCG-009) (Supplementary Table 3 and Supplementary Fig. 6).

Discussion

In this study, we demonstrate that the timing of exposure to different chow diets determines the extent of diet impact on mouse model phenotypes even among genetically identical animals. While a few previous studies have measured the phenotypic impact of AE to standard chow^11,13,26, this is the first study to quantitatively measure the differences in nutrient intake among chow diets and test the impact of exposure during different life stages, including AE, LE and DE. Furthermore, including both sexes in our study provides new evidence of sex-dependent diet effects, which have been poorly assessed in previous literature. Ultimately, this study highlights the important roles of the chow diet, timing of exposure and sex in the reproducibility of preclinical mouse model phenotypes. Chow diet effects on the gut bacterial microbiota were also investigated. Surprisingly, despite exhibiting no substantial effects on phenotypic responses, the AE model exhibited a significant but limited microbial response. LE mice exhibited a broader microbial response that encompassed effects observed in the AE model. Lastly, we found that weaning mice from different diets onto the same 5V5M diet (DE model) for maturation to adulthood resulted in more similar adult phenotypes compared with LE. This implies that phenotypic variability introduced by breeding mice on different chows is reduced by using a postweaning acclimation period.

Mouse diet users typically rely on the manufacturer’s estimates to determine the extent of differences in diet composition and, thus, nutrient exposures. However, while our experimental measures found that the macronutrient contents were as estimated by the manufacturer, most of the diets had higher levels of vitamin D and lower levels of folic acid than estimated by the manufacturer. Furthermore, we found that the different standard chows were eaten at different rates (food intake measures) resulting in macro- and micronutrient exposures that were substantially different from the manufacturer’s estimates. Strikingly, despite similar caloric and macronutrient content between 5058 and 5V5M diets, 5058-fed mice in both the AE and LE models ate substantially more chow and thus had substantially higher macronutrient and fiber intake levels while 5V5M-fed mice ate the least. These higher intake levels may have contributed in part to the larger body weights observed in 5058-fed mice and smaller body weights of 5V5M-fed mice in the LE model.

While most chow diet studies investigate the effects of calorie and macronutrient differences, our study adds new information about vitamin D and folic acid content, two micronutrients that are widely variable among chow diets. During development, vitamin D supports a healthy in utero environment and regulates offspring muscle mass, bone health, adipose storage and brain development²⁷. Folate has critical roles in DNA synthesis and 1-carbon metabolism but is most well recognized for regulating offspring neural tube formation²⁸. The vitamin D content of all four diets exceeded recommended levels^18,19 while one diet (5V0G) also contained >4× more folic acid than the recommended amount²⁰, putting it close to experimentally supplemented diets²¹. Despite the high variability in micronutrient intake across the diets for all exposure models, only the LE model (which had the longest exposure period ~15–18 weeks) exhibited a significant difference in circulating levels of vitamin D (25(OH)D). Here, the highest circulating levels of vitamin D were observed in 2920X-fed LE mice despite them having the lowest vitamin D intake. In accordance with the role of vitamin D in regulating fat storage²⁹, 2920X-fed LE females exhibited the lowest adiposity despite similar total body weight and CRL compared with other diet groups. 2920X-fed mice also exhibited distinct effects on the makeup of the cecum bacterial microbiota, which could be linked to their vitamin D status based on recent studies showing that vitamin D supplementation alters the microbiome composition (in particular, Firmicutes and Bacteroidota^30–32). Firmicutes were highest in abundance in 2920X-fed mice, despite the 2920X diet containing the lowest levels of vitamin D. This is inconsistent with a recent human study that found Firmicutes was positively correlated with vitamin D intake³³. These findings warrant further studies to investigate the role of vitamin D supplementation levels in mouse standard chows (estimated to vary up to 100-fold³⁴) in driving phenotypic variability. This would probably require the use of purified mouse diets where the exact amount of vitamin D supplementation could be measured and tested. Meanwhile, circulating folate concentrations were high but did not significantly differ among the diets, making folate less likely to contribute substantially to the phenotypes observed here.

This study focused on macro- and micronutrient differences among the diets. However, it should be noted that the chows tested vary for many different ingredients (Supplementary Data 1) that contain bioactive compounds that could also, in part, contribute to the phenotypic variability detected. For example, phytoestrogens³⁵ are plant-derived endocrine disruptors that structurally mimic 17β-estradiol and can interfere with estrogen signaling^2,36. Of greatest concern to the developmental mouse model community is that phytoestrogens can affect reproductive and developmental outcomes, including uterine growth^37,38 and reproductive cycling³⁹. The most common phytoestrogens in standard chow diets are isoflavones, which are present at high concentrations in soybeans and alfalfa^2,36,40. To address this issue, some chow diet manufacturers have removed alfalfa meal⁴⁰, while others have removed both alfalfa meal and soy products from diet formulas. Relevant to this study, previous studies have reported soy-induced weight gain^41,42 and greater microbiota diversity in mice exposed to soy-based diets⁴³. While all four chows tested here are primarily composed of wheat and corn, they all contain soybean oil while 5058 also contains soybean meal (Supplementary Data 1). Estimates of isoflavone content (genistein, daidzein and glycitein) among the four diets were reported to range from nondetectable to a maximum of 50 ppm (5V5M and 5V0G) or 20 ppm (2920X). Although phytoestrogen estimates were not provided for the 5058 diet, it is plausible that the addition of soybean meal to this diet may have contributed to its distinct effects on body weight⁴¹ and lean mass⁴¹ (increased in adult males) and composition of the gut bacterial microbiota⁴³.

This study adds new insight into the role of timing of exposure on phenotypic variability. Mice in the LE model demonstrated heightened sensitivity to nutrient variability compared with AE and DE models. For example, 5V5M-fed mice exhibited restricted body weight across the lifetime of the LE model. On the other hand, 5058-fed males in the LE model revealed connections between chow diet-driven increases in food intake, nutrient intake, adult body weight and adult lean mass. 2920X-fed females in the LE model exhibited restricted PWAT and growth and reduced total body fat and fat-to-lean mass ratios. These effects were not observed in the AE or DE models. Thus, diet timing influences the extent of diet impact on phenotype reproducibility. Chow diet selection and use are critical in studies on developmental growth phenotypes and body composition phenotypes commonly used as indicators of metabolic health in mice. Importantly, a normalization diet postweaning may be an effective way to reduce unwanted diet-induced phenotypic variability in mouse models. While this study characterized the impact of dietary exposure throughout the lifetime context tested here (from conception to PND 84 (young adulthood)), we recognize that this is a limited time window, and future work extending this window to assess exposure across the lifespan (including normative aging stages) is important.

This is the first study to demonstrate that the timing of chow diet exposure and sex influence variability in the gut bacterial microbiota composition. Nutrients are well known to have a major role in the establishment and maintenance of the gut microbiome⁴⁴. Here, we showed that differences in chow diet drive changes in the gut’s bacterial composition regardless of whether exposure was during development and adulthood (LE) or only in adulthood (AE). However, while the diet effects on beta diversity and differential abundance of bacterial genera in the AE model were mostly replicated in the LE model, the LE model exhibited 9× more differentially abundant genera that were sex dependent. Among the diets, 5058 exhibited the most distinct effect on beta diversity, while both 5058 and 2920X stood out as the extremes in diet effects on bacterial genera abundance. Although we were unable to measure the phenotypic consequences of these microbial differences, several have predicted pathogenic effects (Clostridium sensu stricto 1⁴⁵, Ruminococcus⁴⁶, Tuzzerella⁴⁷ and Acetatifactor⁴⁸), while others are predicted to have protective effects on a range of outcomes including gut and metabolic health (Bifidobacterium⁴⁹, Lactobacillus⁵⁰, Eubacterium nodatum group, Alistipes⁵¹, Blautia⁵² and Eubacterium xylanophilum group⁵³).

Fundamental sex effects on body composition, feeding behavior, nutritional need and the gut microbiome are well established for a number of rodent models⁵⁴, but not in the context of standard chow exposure. Here, the effects of standard chow diet were strikingly similar between males and females for measures taken before weaning; however, all timing- and diet-dependent adult/postweaning phenotypes were sex specific. This was especially evident for gut microbiota outcomes, with significant sex-dependent diet effects on beta diversity and differential abundance of bacterial genera. Despite the study offering only a limited assessment of males in the AE model (due to low sample sizes), the extensive data provided for the other models further underscore the need to include both sexes in biomedical research and in studies of factors affecting experimental reproducibility.

Taken together, our findings highlight the importance of standard chow selection and the timing of its use in the rigor and reproducibility of experimental mouse model research. Further research is required to pinpoint the specific nutrients that are most influential in driving unwanted mouse phenotypic variability. Nonetheless, our findings build on a growing body of previous reports that strongly support the need for developing specific diet reporting guidelines expanded beyond what has already been initiated via valuable resources such as the Animal Research: Reporting of In Vivo Experiments (ARRIVE) guidelines⁵⁵.

Methods

Animal husbandry, dietary treatment and breeding

All animals were handled in accordance with the Guide for the Care and Use of Laboratory Animals under an approved and registered animal use protocol at the University of North Carolina (UNC) at Chapel Hill. Throughout the study, all mice were housed in ventilated microisolator cages with corncob bedding and maintained at a vivarium temperature of 21–23 °C with a 12 h light cycle and ad libitum access to sterilized water and standard chow. Animals were monitored daily for stress and adverse health. Any animals that demonstrated stress, adverse health and/or did not survive until euthanasia were excluded from the study following established animal care guidelines.

Standard chow selection.

Four standard chows were selected for this study: 5V5M (PicoLab Select Mouse 50 IF/9F; Lab Diet), 5V0G (Select Mouse 50 IF/9F Auto; Lab Diet), 2920X (Teklad Global Soy Protein-Free Extruded Rodent Diet; Teklad) and 5058 (PicoLab Mouse Diet 20; Lab Diet). Diets were selected on the basis of the high prevalence of use at UNC and >18 donor institutions that submit mice to the UNC Mutant Mouse Resource and Research Center strain repository⁵⁶.

AE model.

Sample sizes are provided in Supplementary Table 4. Six-week-old virgin female and male C57BL/6J mice were simultaneously procured from Jackson Laboratories (stock no. 000664) and transferred with random assignment (separated by sex) into cages with one of four standard chow diet groups (5V5M, 5V0G, 2920X and 5058) (Fig. 1f). After 3 weeks of acclimation to dietary treatment, a subset of females was euthanized to collect premating micronutrient measurements (Fig. 1f). The remaining mice were mated in trio breeding for 3 weeks. Bred males (sires) were euthanized after breeding, and bred females (dams) were euthanized at weaning (Fig. 1f). Time on a diet is provided in Supplementary Table 5. All mice were fasted 12 h before euthanasia by CO₂.

LE and DE models.

Sample sizes are provided in Supplementary Table 6. At weaning, pups were caged by sex, and siblings were assigned to either remain on their preweaning diet for LE or switch to a common postweaning diet (5V5M) for DE (Fig. 1f). At the end of treatment, all mice were fasted for 12 h before euthanasia by CO₂. Time on a diet is provided in Supplementary Table 5.

Nutrient intake

Food intake was measured weekly for each individual cage, and average daily values were calculated per mouse (average ((food weight at week end − food weight at week start)/(number of days)/(number of mice in the cage))). All four diets were provided in pelleted form and of similar consistency; cages were carefully monitored for food loss due to grinding or pellet degradation and none of the diets exhibited this effect. Each diet was vacuum-sealed separately and stored at −80 °C. Diets were shipped overnight on dry ice to be measured by Eurofins Food Chemistry Testing (Supplementary Table 1). Measured values of total calories, macronutrients (fat, carbohydrates and protein), fiber (total dietary fiber and crude fiber) and micronutrients (vitamin D2, vitamin D3 and folic acid) were converted into units of measures comparable to the manufacturer’s estimates.

Phenotype measurements

For the AE model, body weights were measured every 3 weeks after the start of treatment (premating) and then once at the end of treatment. For the LE and DE models, body weights were measured biweekly, starting at PND 5 until the end of treatment. Minimal handling was performed at sensitive periods (for example, during gestation and birth to PND 4). CRL was measured using carbon fiber calipers (cat. no. 36934–152, VWR) at PND 5, weaning (PND 25–33) and adulthood (PND 84). Body composition was measured at 8–9 weeks of age by EchoMRI (Animal Metabolism Phenotyping Core, UNC at Chapel Hill). Fasting blood glucose was measured at the end of treatment after a 12 h fast using a single tail nick bleed to measure blood glucose levels via glucometer (cat. no. 68623221868, Accu-Chek Performa, Roche). PWAT was dissected and weighed at the end of treatment immediately following euthanasia.

Serum measurements (vitamin D and folate)

Serum was isolated from whole blood collected by cardiac puncture and centrifuged 10 min at 2,000g at 4 °C. Serum was flash-frozen in liquid nitrogen and stored at −80 °C until use. Vitamin D (25(OH)D) levels were measured by ELISA following the manufacturer’s protocol (Mouse/Rat 25-OH Vitamin D ELISA Assay Kit (VID21-K01), Eagle Biosciences). Serum folate concentrations were determined using the Lactobacillus casei microbiological assay as previously described⁵⁷.

Statistical analyses for phenotype data

Unless otherwise stated, statistical analyses were performed using JMP Pro 17 software (SAS). Data were tested for normality (Shapiro–Wilk test), and normally distributed data were analyzed using one-way analysis of variance (ANOVA). Nonnormally distributed data were analyzed using the Kruskal–Wallis test. Tukey’s post-hoc was used to identify which diets differed significantly.

Bacterial microbiota profiling

Cecum was collected after euthanasia from AE females and males (n = 4 per diet and n = 2 per diet, respectively) and LE females and males (n = 6 per diet and n = 5–6 per diet, respectively), flash-frozen in liquid nitrogen and stored at −80 °C. Samples close to the population median for body weight, body composition and cecum weight in each diet group were selected from three to four litters per diet.

Frozen samples were shipped overnight on dry ice to the University of Missouri (MU) Metagenomics Center for DNA extraction, preparation and plating. 16S rRNA library preparation and sequencing were performed at the MU Genomics Technology Core; data cleanup and quality control, sequence alignments and annotations were carried out at the MU Bioinformatics and Analytics Core, as previously described¹³. In brief, bacterial 16S rRNA amplicons were built via amplification of the V4 region of the 16S rRNA gene using the U515F/806R primer pair⁵⁸ and processed for paired-end sequencing by MiSeq (Illumina) as previously described^13,58,59. Read pairs were rejected if either read did not match the 5′ primer (using an error rate of 0.1). Cutadapt⁶⁰ (version 2.6) was used to trim primers. Obtained sequence data were demultiplexed using the QIIME2^61,62 DADA2⁶² plugin (version 1.10.0) as follows: forward and reverse reads were truncated to 150 bases; forward and reverse reads with expected errors >2.0 were discarded; and chimeras were detected using the consensus method and removed. Output sequences were classified as amplicon sequence variants (ASVs). Taxonomy was assigned using the SILVA.v132⁶³ reference database with the classify-sklearn procedure. The ASV dataset (Supplementary Data 4) was randomly subsampled to a uniform sequence depth of 42,577 reads per sample for rarefaction. The rarefied table (Supplementary Data 5) contained 2,980,390 reads (42,577/sample) and 205 operational taxonomic units (OTUs).

Alpha and beta diversity metrics were calculated in PAST4.13⁶⁴ using rarefied data. Linear regression models and Tukey’s post-hoc tests were used to measure main diet effects on observed sequence variants and Shannon index using JMP Pro 17 software (SAS). PERMANOVA using Bray–Curtis distances was performed to test for diet and sex effects, and data were quarter-root transformed to generate the PCoA. Microbial differential abundance was calculated in MicrobiomeAnalyst 2.0 (available at https://microbiomeanalyst.ca/) using the Marker Data Profiling tool. Features with identical values (that is, zero) across all samples and features appearing in only one sample were excluded, resulting in a total of 2,980,202 reads (average of 42,574 reads per sample) and 205 ASVs analyzed. Multifactor comparisons with the linear model at the genus level were performed for female and male samples combined to evaluate diet as the primary effect and sex as a covariate. Filtering was applied to features with low count (<4), low prevalence (<20%) and low variance (<10% based on interquantile range). False discovery rate (FDR) <0.05 after Bonferroni⁶⁵ and Benjamini–Hochberg⁶⁶ corrections were considered statistically significant. Confirmatory differential abundance analysis was performed using ANCOM-BC2. ANCOM-BC2 testing was performed using fixed effects for diet and sex. Taxa exhibiting significant differential abundance were determined by a Benjamini–Hochberg⁶⁶-corrected P value <0.05 (Supplementary Data 2 and 3). Structural zeros were identified on the basis of the presence or absence of taxa across the groups defined by the diet variable.

Two-way hierarchical clustering of differentially abundant microorganisms at the genus level was performed in JMP Pro 17 software (SAS). AE and LE groups were analyzed separately.