Skip to main content
Download PDF

Analysis of strain, sex, and diet-dependent modulation of gut microbiota reveals candidate keystone organisms driving microbial diversity in response to American and ketogenic diets

Microbiome volume 11, Article number: 220 (2023) Cite this article

Abstract

Background

The gut microbiota is modulated by a combination of diet, host genetics, and sex effects. The magnitude of these effects and interactions among them is important to understanding inter-individual variability in gut microbiota. In a previous study, mouse strain-specific responses to American and ketogenic diets were observed along with several QTLs for metabolic traits. In the current study, we searched for genetic variants underlying differences in the gut microbiota in response to American and ketogenic diets, which are high in fat and vary in carbohydrate composition, between C57BL/6 J (B6) and FVB/NJ (FVB) mouse strains.

Results

Genetic mapping of microbial features revealed 18 loci under the QTL model (i.e., marginal effects that are not specific to diet or sex), 12 loci under the QTL by diet model, and 1 locus under the QTL by sex model. Multiple metabolic and microbial features map to the distal part of Chr 1 and Chr 16 along with eigenvectors extracted from principal coordinate analysis of measures of β-diversity. Bilophila, Ruminiclostridium 9, and Rikenella (Chr 1) were identified as sex- and diet-independent QTL candidate keystone organisms, and Parabacteroides (Chr 16) was identified as a diet-specific, candidate keystone organism in confirmatory factor analyses of traits mapping to these regions. For many microbial features, irrespective of which QTL model was used, diet or the interaction between diet and a genotype were the strongest predictors of the abundance of each microbial trait. Sex, while important to the analyses, was not as strong of a predictor for microbial abundances.

Conclusions

These results demonstrate that sex, diet, and genetic background have different magnitudes of effects on inter-individual differences in gut microbiota. Therefore, Precision Nutrition through the integration of genetic variation, microbiota, and sex affecting microbiota variation will be important to predict response to diets varying in carbohydrate composition.

Video Abstract

Introduction

The gut microbiota has emerged as a key component underlying the application of precision nutrition and individualized dietary response. Gut microbiota utilizes nutrients passing through the gastrointestinal tract to perform biological functions, this in turn impacts host digestion, absorption, and metabolism of nutrients [1]. There is a consensus that a relationship exists between the microbes and their host. Although some studies have been performed in humans and livestock species, e.g., quails and hens [2, 3], the impact of inter-individual variability on how diet modulates gut microbiota composition remains underinvestigated [4, 5].

Previous studies from our group have demonstrated strong mouse strain-specific differences in response to American and ketogenic diets [6,7,8,9], especially between the C57BL/6 J (B6) and FVB/NJ (FVB) strains. The composition of gut microbiota is known to be influenced by both host genetics and environmental factors such as diet [9,10,11], which is considered one of the most potent regulators of gut microbial composition. We have recently demonstrated that B6 is particularly susceptible to altered gut microbiota relative to A/J, FVB, and NOD/ShiLtJ [9]. Furthermore, changes to bacterial abundance do not occur uniformly in response to diets varied in macro- and micronutrient composition because of differences in substrate utilization between bacterial taxa [9, 12]. To determine what the composition of the “ideal” microbiome is, it would be pertinent to disentangle the effects of host genetics and host diet from the extra layers of complexity arising from differences in substrate utilization by individual organisms and ultimately identify genes regulating interindividual differences in the composition of the gut microbiota [12, 13]. Until recently, few studies have considered the extent to which the combination of host genetics and diet modulate the abundance of specific bacterial taxa, and even fewer have considered how sex might add an additional layer of complexity to describing inter-individual variation in microbiota composition [14,15,16,17,18].

In this study, an intercross population (F2) was generated between B6 and FVB to investigate the strain-, sex-, and diet-dependent modulation of the gut microbiota. F2s were fed either an American or a ketogenic diet and fecal microbiota was quantified. The results provide evidence for 32 quantitative trait loci (QTL) that affect microbiota composition, but also significant diet and sex differences in the effect size of the QTL. In many cases, these were sex- and diet-independent QTL (i.e., marginal effect QTL that are not specific to diet or sex, y ~ sex + diet + sex:diet + [marker]), and in other cases, these were genotype- and diet-dependent (y ~ sex + diet + marker + [marker:diet]) or genotype- and sex-dependent QTL (y ~ sex + diet + marker + [marker:sex]), which allowed us to characterize how much host genetics, sex, and diet affect specific gut microbiota, and provided insights into factors driving microbial diversity, which has implications for advancing precision nutrition through preclinical studies.

Methods

Animals and diets

B6 females were crossed with FVB males to generate F1 mice and subsequently intercrossed to generate an F2 population. Both parental strains were acquired from The Jackson Laboratory prior to generating the F2 population. F2s were randomized to five mice per cage and screened for their response to American (a powdered meal composed of 35% of energy from fat, 50% from carbohydrates) and ketogenic (a paste composed of 84% of energy from fat, 0% from carbohydrates) diets during a 3-month feeding trial. Neither diet was irradiated prior to use. Detailed diet compositions are provided in Supplementary Table S1.

For the feeding trials, 3–5 week-old mice were randomly assigned to one of the two diet groups and allowed to eat ad libitum. Half of the F2 mice were placed on American diet (102 males, 122 females) and half on ketogenic diet (126 males, 119 females). All protocols in this study were approved by the Texas A&M University Institution Animal Care and Use Committee (IACUC protocol number: 2022–0273) and animals were maintained in accordance with those guidelines at 22 °C under a 12-h light cycle with up to 5 mice per cage, in compliance with the Guide for the Care and Use of Laboratory Animals published by the US National Institutes of Health. Detailed husbandry descriptions are provided in Supplementary Table S2 per ARRIVE guidelines for reporting housing and husbandry conditions. At the end of the feeding trial, mice were euthanized by carbon dioxide asphyxiation, blood was collected, and tissues and feces were harvested and immediately flash frozen in liquid nitrogen.

Microbiota phenotypes

Stool microbiota was analyzed by 16S rRNA V4 sequencing methodology as reported previously [19]. In brief, total stool DNA was extracted using ZymoBIOMICS™ 96 MagBead DNA kit (Zymo Research, Irvine, CA) with an automated epMotion (Eppendorf, Hamburg, Germany) robotic system. About 100 mg of stool samples were placed in the ZR BashingBead™ Lysis Tube and homogenized using FastPerp24 bead beater (Millipore, Hayward, CA) at 6.5 HZ for 2 min. The lysate was centrifuged at ≥ 10,000 × g for 1 min and 200 μl supernatant from lysis tube was transferred to 96 deep-well plate (Eppendorf, Hamburg, Germany) and loaded in an epMotion 5075t robotic system. Using epMotion, 600 μl ZymoBIOMICS™ MagBinding Buffer and 25 μl of ZymoBIOMICS™ MagBinding Beads were added to each well and was mixed well. After mixing, the plate was placed on a magnetic stand and the supernatant was discarded. MagBinding Beads were washed with MagWash 1 and MagWash 2 and the DNA was eluted using 50 μl ZymoBIOMICS™ DNase/RNase free water. The DNA concentration was measured using NanoDrop One (Thermo Scientific, Petaluma, CA).

Mixed template amplicon library for the 16S variable region 4 (V4) was prepared according to the protocol from the Earth Microbiome Project (EMP_ (http://www.earthmicrobiome.org/emp-standard-protocols/) using the extracted stool total DNA and the primer sets (515F and barcoded 806R) [20]. The PCR master mix, primer, and samples were plated using the epMotion. Appropriate NTC, extraction control, and pooled fecal sample were added to each plate. The PCR master mix was prepared consisting of 37.5 µl of GoTaq Green Master Mix (Promega, Madison, WI), 3 µl of 25 mM MgCl2, 1.5 µl of 10 µM forward primer 515F, and 25.5 µL of nuclease-free water. Then, 1.5 µl of 10 µM barcode-specific reverse primer 806R and 6 µl of extracted stool DNA were added. PCR was performed in duplicate of 25µL under the following conditions: denaturation (1 cycle) at 94 °C for 3 min; amplification of 25 cycles at 94 °C for 45 s, 50 °C for 60 s, and 72 °C for 90 s; and a final extension step cycle at 72 °C for 10 min. Amplicon DNA was multiplexed and purified using Wizard SV Gel and PCR Clean-Up System (Promega, Madison, WI). The amplicon library was sequenced using the Illumina MiSEQ platform with 2 × 250 bp paired-end sequencing. Sequences were de-multiplexed and exact amplicon sequence variants (ASV) in the 16S rRNA gene sequence were determined using the open-source software QIIME2-DADA2 pipeline [21]. A total of 11,316,115 sequences with an average of 26,074 ± 13,697 (mean ± SD) sequences per sample were recovered after demultiplexing. Taxonomy was assigned using the SILVA 132 reference database [22] customized for 16 s V4 (515F/806R) region of sequences at the threshold of 99% pairwise identity. ASV belonging to mitochondria and chloroplast were filtered out from the ASV table. We performed a single rarefaction at a sequence depth of 4,500 sequences per sample. α-diversity (Shannon diversity index and observed species) and β-diversity (unweighted UniFrac, weighted UniFrac, Jaccard Index, and Bray Curtis) were calculated from the unfiltered ASV table. Any ASV not seen more than 5 times in at least 5% of the samples were removed for calculating differential bacteria abundance. 16S V4 Sequences are publicly available on the SRA database under the Bioproject ID “PRJNA803237.”

Microbial traits are listed by both ASV ID and maximum taxonomic information for reference in Table 1. The rank of the maximum taxonomic information is described at first mention of all microbial traits within the text and indicated with a taxonomic rank in subsequent figures and tables (i.e., D0, Kingdom; D1, Phylum; D2, Class; D3, Order; D4, Family; D5, Genus; D6, Species).

Table 1 Linkage analysis for microbial traits in the combined, diet-specific, and sex-specific models
Full size table

Metabolic phenotypes

The data analysis and collection methods for fat mass gain and serum HDL cholesterol concentration have been described previously [23]. Briefly, Echo magnetic resonance spectroscopy (MRI) (EchoMRI, Houston, TX, USA) was used to measure the fat and lean mass of all individuals. Using serum obtained from blood collected at the end of the feeding trial, total cholesterol, HDL, and LDL measurements were performed in duplicate using the EnzyChrom AF HDL and LDL/VLDL Assay kit (BioAssay Systems, Hayward, CA, USA).

Genotyping

The genotyping analysis and collection methods have been described previously [23]. Briefly, the F2 population was genotyped on the Mouse Universal Genotyping Array (MUGA) that includes 7854 SNP markers [24]. Markers that were not polymorphic between B6 and FVB were removed from the dataset and uncertain genotype calls for individuals (GenCall score quality metric < 0.7) were set to missing. The remaining markers were used to generate a genetic map to check for problematic markers and/or sample DNAs. After all corrections, 1667 markers were used for the association analyses. Updated MUGA marker annotation was obtained from Dr. Karl Broman (https://kbroman.org/MUGAarrays/new_annotations.html).

Statistical analyses

Linkage analysis

For microbiota phenotypes, a core measurable microbiota (CMM) was defined as those traits present in at least 20% of the individuals. Thresholds ranging from 0.25 to 10% have been applied for differential abundance analyses by others [25,26,27]. For linkage analyses, more stringent thresholds have been applied to define the CMM [28]. With this threshold, we expect to capture the anticipated 1:2:1 ratio among genotypes and most microbial traits in the F2 population. The CMM consists of 134 ASVs. After determining organisms present in the CMM, absolute microbial abundances (counts) were quantile normalized for linkage analyses. Normal quantiles were calculated with the preprocessCore R package from Dr. Ben Bolstad, version 1.46.0 (https://github.com/bmbolstad/preprocessCore).

QTL mapping was performed on metabolic and microbial features (y) in all F2 mice from both sexes and diets, and linear models using ANOVA targeted three types of genetic effects: (1) QTL effects, whereby the effect of a marker SNP is tested after controlling for sex, diet, and sex by diet interaction, which we describe in the formula as y ~ sex + diet + sex:diet + [marker], where the term in brackets is the alternative but not the null model; (2) QTL by diet effects, using y ~ sex + diet + marker + [diet:marker]; and (3); QTL by sex effects, using y ~ sex + diet + marker + [sex:marker]. QTL peaks with a logarithm of the odds (LOD) greater than thresholds determined by 10,000 permutations were considered genome-wide significant (p < 0.05, LOD > 4.00 microbial abundance, 3.90 measures of diversity) or highly significant (p < 0.01, LOD > 5.19 microbial abundance, 4.68 measures of diversity) for all models. The thresholds applied to microbial abundance and measures of diversity reflect the average genome-wide significant thresholds for all ASV present within the CMM and for all measures of diversity analyzed respectively. A LOD drop of 1.5 LOD from the top marker was used to determine the 95% confidence intervals for each QTL. Linear models using ANOVA were used to check for any interactions between sex and/or diet with the top markers of each QTL. The variance explained by the top markers at each QTL in the combined model was calculated by dividing the sum of squares of the model including the top marker by the total sum of squares of the model without QTL. The variance explained by the top markers at each QTL in the interactive models was calculated by dividing the sum of squares of the model including the interaction between diet and the top marker or sex and the top marker by the total sum of squares of the model without QTL.

Several limitations exist in the available literature for microbiome QTL analysis including the current work. Microbial data is zero-inflated compositional data [29] and to-date no appropriate statistical method has been developed to transform and perform QTL analysis that fully addresses zero-inflation and compositional nature of the data. Zero-inflation might be the result of true biological variation or technical variation in current technologies for measuring abundances of organisms [30, 31]. Of note, normalizing quantiles does not force the data into a normal distribution but rather, makes the individual microbial features more similar in statistical properties. Classical approaches to data transformation, for example, the log transformation on individual ASV, similarly fail to achieve a normal distribution and otherwise substantially alter the distribution of only the non-zero data (zero data cannot be transformed). Another common approach to overcome these limitations is binary modeling of the presence or absence of specific organisms followed by linear modeling of specific organisms for only hosts identified with non-zero counts of that organism. However, this two-step approach is limited to single predictors and reduces power without addressing the zero-inflated and compositional nature of the data encountered in the one-step procedure. After careful consideration of these alternatives, we chose to use a combination of permutation and preprocessCore’s quantile normalization across all ASV present in the CMM. Forcing the skewed data into the same distribution means that the data will behave similarly under permutation which allowed us to determine how unusual specific LOD scores were in the permutated data to identify appropriate thresholds of significance. We acknowledge a great deal of variation in methods used to normalize microbial features and even in the definition of the CMM as well as the great need to standardize these methods between investigators. Continued growth of statistical methods for zero-inflated compositional microbial data sets is needed. Our implementation has been made publicly available at https://github.com/annacsalvador/Salvador_Microbiome_2023.

Confirmatory factor analysis and structural equation modeling

Confirmatory factor analysis (CFA) was conducted with the lavaan R package for structural equation modeling (SEM) from Dr. Yves Rosseel, version 0.6.9 (https://www.jstatsoft.org/index.php/jss/article/view/v048i02/2448) [32]. Initial models were selected based on information from individual QTL models and correlations among traits within and between QTL models. All traits were collapsed into four, ordinal quantiles for CFA and diagonally weighted least squares estimator was used based on methods described elsewhere [33]. The final structural models illustrate QTL models for overlapping microbial and metabolic traits and are refined to include only predictors for which pathway coefficients are significantly different from zero, indicating that each of the remaining predictors in the model is significantly associating with one or more endogenous or exogenous variable. Our implementation has been made publicly available at https://github.com/annacsalvador/Salvador_Microbiome_2023.

Candidate gene annotation

All genes within each significant QTL confidence interval were annotated with KEGG pathway identifiers. Candidate genes were furthered characterized by KEGG pathways related to glucose, insulin, fatty acids, adipocytes, cholesterol, obesity, diabetes mellitus, metabolic syndrome, digestion and absorption of carbohydrates, fats, and proteins, the epithelial barrier, and the immune system. A comprehensive list of KEGG pathway queries is provided in Supplementary Table S3. Transcript variants between the parental mouse strains were identified in genes annotated for selected KEGG pathways of interest from the Mouse Genome Informatics Strains, SNPS, and Polymorphisms database. Tissue-specific expression was determined with Mouse ENCODE Transcriptome data, accessed through the National Library of Medicine National Center for Biotechnology database.

Results

Diet is a strong modulator of the gut microbiome

Diet explains a large proportion of variation in the abundance of microbiota at the Phyla level irrespective to genetic background. Diet explains 64.79% of variation in abundance of Actinobacteria, 61.22% of variation in the abundance of Firmicutes, and 25.49% of variation in the abundance of Bacteroidetes (Supplementary Table S4). The relative abundance of Firmicutes in F2 on the ketogenic diet is nearly twice as high as in F2 on the American diet (Fig. 1). This increase in Firmicutes in F2 on the ketogenic diet appears to occur at the expense of the relative abundance of Actinobacteria and Bacteroidetes (Fig. 1).

Fig. 1
figure 1

Relative abundance of microbiota and correlations at the phyla level (by diet). Relative abundance of Firmicutes in F2s on the ketogenic diet is nearly twice as high as the relative abundance of Firmicutes in F2s on the American diet at the expense of Actinobacteria, and Bacteroidetes for which the relative abundances are lower in F2s on the ketogenic diet

Full size image

Principal coordinate analysis (PCoA) for measures of beta diversity revealed two distinct groups segregating at PC1. The Bray–Curtis index PCo1 and PCo2 describe 31.1% and 16% of the variation in ASV respectively (Fig. 2A). The Jaccard Index PCo1 and PCo2 are nearly identical to the Bray–Curtis index and describe 21.3% and 10.6% of variation in ASV respectively (Supplementary Fig. S2). Unweighted UniFraction PCo1 and PCo2 describe 10.7% and 6.2% of variation in ASV respectively (Fig. 2B). Weighted UniFraction PCo1 and PCo2 describe 51.1% and 31.2% of variation in ASV respectively (Fig. 2C). Overlaying diet with the data illustrates two distinct groups roughly segregating PCo1 for all measures of beta diversity. However, alpha diversity, illustrated by the Shannon Diversity Index, does not depend on diet (Fig. 2D). Eigenvectors extracted from the PCo1 and PCo2 from the Bray–Curtis index, Jaccard index, unweighted and weighted UniFractions, as well as values from the Shannon Diversity Index served as additional traits for linkage analysis below.

Fig. 2
figure 2

Beta diversity Principal Coordinate Analysis (PCoA). A Bray Curtis Index. PC1 and PC2 describe 31.1% and 16% of the variation in ASV respectively. B Unweighted UniFraction. PC1 and PC2 describe 10.7% and 6.2% of the variation in ASV respectively. C Weighted UniFraction. PC1 and PC2 describe 51.1% and 31.2% of the variation in ASV respectively. Overlaying diet with the measures of beta diversity illustrates two distinct groups, roughly segregating PC1 for all measures of beta diversity. D Shannon Diversity Index. The Shannon Diversity Index is not different between diet groups

Full size image

Microbial features are modulated by genetic loci

In the sex and diet-independent QTL model, which tests for marginal effect QTL after controlling for sex and diet (see the “Methods” section; y ~ sex + diet + sex:diet + [marker]), 18 distinct QTL were detected for 15 unique microbial abundances (counts), there were 119 additional organisms remaining within the CMM that did not display a genetic linkage (Fig. 3, Table 1).

Fig. 3
figure 3

Microbial QTL. Outer ring to inner ring: Significant QTL in the QTL model (black) for ASV associated with the genus Coriobacteriaceae UCG-002 on Chr 8 at Asvq1, the genus Enterohabdus on Chr 8 at Asvq2, multiple uncultured bacterium from the Muribaculaceae family on Chr 14 at Asvq3 and Chr 17 at Asvq4 and Asvq5, the genus Alistipes on Chr 12 at Asvq6, the genus Rikenella on Chr 1 at Asvq7, the genus Rikenellaceae RC9 Gut Group on Chr 11 at Asvq8, the genus Streptococcus on Chr 8 at Asvq9 and Asvq10, the genus GCA-900066575 on Chr 9 at Asvq11, the genus Lachnoclostridium on Chr 1 at Asvq12 and Chr 13 at Asvq13, the genus Romboutsia on Chr 5 at Asvq14 and Chr 13 at Asvq15, the genus Ruminiclostridium 9 on Chr 1 at Asvq16, and the genus Bilophila on Chr 1 at Asvq17 and Chr 9 at Asvq18. Significant QTL in the diet specific model (blue) for ASV associated with multiple uncultured bacterium from the Muribaculaceae family on Chr 13 at Asvq19 and Asvq20, Chr 14 at Asvq21 and Asvq22, and Chr 17 at Asvq5, the genus Rikenelleceae RC9 Gut Group on Chr 16 at Asvq23, the genus Parabacteroides on Chr 16 at Asvq24, the genus Lactobacillus on Chr 8 at Asvq25, multiple ASV from the Streptococcus genus on Chr 8 at Asvq9, and Asvq10, the uncultured genera from the family of Clostridiales vadinBB60 group on Chr 6 at Asvq26, and the Lachnospiraceae family on Chr 1 at Asvq27. A single significant QTL in the sex specific model (orange) for the genus Alistipes on Chr 13 at Asvq28; Previously identified QTL for metabolic traits and diversity measures in the combined model (black) and diet specific model (blue). Fat mass gain during the feeding trial on Chr 1 at Fmgq1, along with serum HDL cholesterol concentration after the feeding trial on Chr 1 at Hdlq1, Bray–Curtis and Jaccard measures of beta diversity at Bcpc2q and Jpc2q. Unweighted unifraction on Chr 8 at Uufpc2q, and weighted unifraction on Chr 16 at Wufpc1q. Fmgq1, Hdlq1, Bcpc2q, and Jpc2q overlap the same region of the genome as Asvq7, Asvq16, and Asvq17 for uncultured Rikenella, Ruminiclostridium, and uncultured Bilophila. Wufpc1q overlaps the same region of the genome for Asvq23 and Asvq24 for Rikenelleceae RC9 gut group and Parabacteroides

Full size image

Asvq7, for the genus Rikenella overlaps with Asvq16 and Asvq17 for Ruminiclostridium 9 and Bilophila genera, respectively, as well as with the previously identified QTL for fat mass gain (Fmgq1) and serum HDL cholesterol concentration (Hdlq1).

Apart from Coriobacteriaceae (Asvq1), diet appears to explain a more significant proportion of the variation in the abundance of these ASV despite this QTL not being diet specific. This is the only ASV with a QTL for which the top marker explains a greater proportion of the variation in the abundance of the organism than diet itself despite there being a significant effect of diet as well in this model (Table 1). For all the remaining loci identified under the QTL model, diet explains a greater proportion of the variation than the top marker does at each QTL. This is particularly clear for Streptococcus where diet explains nearly 50% of the variation in abundance of two Streptococcus ASVs while the top markers at Asvq9 and Asvq10 explain only 3.14% and 2.92% of the variation, respectively (Table 1).

Microbial features are modulated by diet-specific genetic loci

The analysis for the interaction between QTL and diet detected 12 QTL for 11 unique microbial features, there were 123 organisms remaining within the CMM not displaying diet-specific genetic linkage (Fig. 3, Table 1). Nine of these QTLs were distinct from the ones identified in the QTL model.

Of note, three diet-specific QTLs were identified that are identical or nearly identical to Asvq5, Asvq9, and Asvq10 that were identified in the QTL model for the same organisms. There are only modest differences in the 95% confidence intervals at Asvq9 and Asvq10 in the loci identified in the QTL and QTL by diet models. The top marker is unchanged between the QTL and QTL by diet models at all 3 loci so these loci will continue to be referred to as Asvq5, Asvq9, and Asvq10.

Interestingly, diet alone explains a greater proportion of the variation than the interaction between diet and the top marker at the diet-specific QTL for Muribaculeceae (Asvq19, Asvq20, Asvq21, Asvq22, Asvq5), Rikenellaceae RC9 Gut Group (Asvq23), Streptococcus (Asvq9, Asvq10), and the uncultured geneus of the Lachnospiraceae family (Asvq27) (Table 1). The interaction between genotype and diet appears to explain the greatest proportion of variation at Asvq24 for Parabacteroides (4.11%), Asvq25 for Lactobacillus (4.22%), and Asvq26 for Clostridiales VadinBB60 Group (4.56%) (Table 1).

Microbial features are modulated by sex-specific genetic loci

The QTL by sex analysis detected a single significant QTL for Alistipes on Chr 13 at 18.4 Mb (Asvq28) (Fig. 3, Table 1).

Sex explains 1.66% of variation in the abundance of Alistipes while the interaction between sex and the top marker at Asvq28 explains over 4% of variation in the abundance of this OTU (Table 1). Even in the sex-specific model of Asvq28, diet explains a greater proportion of the variation than either sex or the interaction between sex and the genotype.

Measures of beta diversity are genetically modulated

In the QTL model, three distinct QTL were identified for PCo2 of several measures of beta diversity including Bray–Curtis, Jaccard, and the unweighted UniFraction (Fig. 3, Table 2). The eigenvectors for both Bray–Curtis and the Jaccard index PCo2 map to Chr 1 at 177.5 Mb (Bcpc2q; p < 0.05, CI = 160.6–185.1 Mb; Jpc2q; p < 0.05, CI = 160.6 = 185.1 Mb), overlapping with Asvq7 (151.9–193.3 Mb), Asvq16 (138–186.3 Mb), Asvq17 (144.5–193.3 Mb) and previously identified QTL for fat mass gain (Fmgq1; 180–194.4 Mb) and serum HDL cholesterol concentration (Hdlq1; 160.6–176.1 Mb) [23]. As noted above, PCo2 of Bray–Curtis and the Jaccard index appears to be modestly represented by alpha diversity. The remaining locus identified by the QTL model was for unweighted UniFraction PC2 at 53.7 Mb (Uwufpc2q; p < 0.05, CI = 43.4–62.8 Mb). No QTL were identified for the Shannon Diversity Index.

Table 2 Linkage analysis for microbial metadata
Full size table

In the QTL by diet analysis, one additional QTL was identified for weighted UniFraction PC1 on Chr 16 at 79.6 Mb (Wufpc1q; p < 0.01, 95% CI = 72.8–95.8 Mb), overlapping with Asvq23 and Asvq24 for the Rikenelleceae RC9 Gut Group and Parabacteroides (Asvq23; 72.8–96.5 Mb, Asvq24; 72–96.5 Mb).

Candidate keystone species modulating the microbiome and physiological traits

A structural equation model (SEM) was built to illustrate the magnitude of the effects of each predictor in the models of Bilophila, Rikenella, Ruminiclostridium 9, and Bray–Curtis PCo2, all mapping to the distal part of Chr 1 (Fig. 4A). Genotypes at Bcpc2q were chosen to model all traits mapping to the distal part of Chr 1 as Bcpc2q is contained inside of the confidence interval for the other three loci mapping to the distal region of Chr 1. The initial model was refined until path coefficients were all significantly different from zero, suggesting that each of the remaining predictors in the model is significantly associating with one or more of the other predictors (Table 3). The refined model suggests that the FVB/FVB genotype at Bcpc2q increases abundances of Bilophila, Ruminiclostridium 9, and Rikenella, and these three ASV are driving differences in the Bray–Curtis index PCo2 (Fig. 4A). A covariance pathway is detected among abundances of Bilophila and Ruminiclostridium 9 in addition to their individual, direct effects on the Bray–Curtis index. The inclusion of metabolic traits does not elucidate direct, indirect, or covariance pathways between the metabolic traits and specific organisms mapping to distal Chr 1. However, a covariance pathway is observed between the Bray–Curtis index PCo2 and the amount of fat mass gained.

Fig. 4
figure 4

Graphical representation of SEM. Solid, single-headed arrows indicate the direction of paths and the weight of each arrow is proportional to the path coefficient (r) from the predictor to the variable and the percentage of variation in the variable that is explained by each predictor. Positive effects (green arrows) indicate that the FVB allele increases the trait; negative effects (red arrows) indicate that the FVB allele decreases the trait. Double-headed, blue arrows represent covariate pathways detected in the structural model of microbial features and measures of diversity

Full size image
Table 3 Conditioned linkage analysis and structural modeling of overlapping QTL
Full size table

Similarly, a SEM was built to illustrate the magnitude of effects of each predictor in the models of Parabacteroides, Rikenellaceae RC9 gut group, and weighted UniFraction PCo1, all mapping to distal Chr 16 (Fig. 4B). Genotypes at Wufpc1q were chosen for the model as the 95% confidence interval for this locus is contained within the QTL of all other traits in this structural model. The path coefficients are again, all significantly different from zero (Table 3). The refined model suggests that the interaction between the FVB/FVB genotype and the American diet at Wufpc1q directly decrease abundances of Parabacteroides and the Rikenelleceae RC9 Gut Group. The abundance of Parabacteroides has a direct relationship with the weighted UniFraction PCo1. A covariance pathway is detected among abundances of Parabacteroides and the Rikenelleceae RC9 Gut Group.

Identification of candidate genes at Asvq7, Asvq16, Asvq17, and Bcpc2q

Candidate genes that might elucidate the relationship between Rikenella, Ruminiclostridium 9, Bilophila, and the Bray–Curtis PCo2 were investigated. Bcpc2q is contained inside of the confidence interval for the other three loci mapping to the distal region of Chr 1. Positional candidates at Bcpc2q that overlap with one or more metabolic KEGG pathways are summarized in Table 4. Out of 275 positional candidates at Bcpc2q, 35 genes overlap with one or more KEGG pathways. Eleven out of the total 35 positional candidates harbor known non-synonymous transcript variants diverging between these strains. The presence of these non-synonymous transcript variants makes Aim2, Apoa2, Atp1a4, Cadm3, Cd244a, Cd48, F11r, Fcer1g, Mpz, Ndufs2, Sdhc, and Sell the primary candidate genes of interest in this region. Of these, F11r, Fcer1g, Ndufs2 and Sdhc are expressed in the intestine.

Table 4 KEGG Pathway annotation of positional candidate genes at Mtq7, Mtq8, Mtq13, Mtq14, and Pc1q
Full size table

Identification of candidate genes at Asvq23, Asvq24, and Wufpc1q

Candidate genes were investigated that might elucidate the relationship between Rikenelleceae RC9 Gut Group, Parabacteroides, weighted UniFraction, Asvq23, Asvq24, and Wufpc1q. Positional candidates at Wufpc1q that overlap with one or more metabolic KEGG pathways are summarized in Table 4. Out of 133 positional candidates at Wufpc1q 11 genes overlap with one or more metabolic KEGG pathways. However, none of the genes annotated with KEGG pathway from our query harbor known non-synonymous transcript variants diverging between the two strains.

Discussion

This study provides evidence that abundances of gut microbiota are driven by unique combinations of effects from the host’s genetics, response to high fat diets varied in carbohydrate content, and sex. Many previous studies have compared the effects of control mouse diets to high fat diets where one or two representative ingredients contribute to the total fat, carbohydrate, and protein content of the diet [6]. The American and ketogenic diets used here more accurately recapitulate the diversity of ingredients found in human diets not only in terms of the macronutrient profile of human diets but also the fiber content and lipid profiles [6]. The diverse set of ingredients is particularly important to studies of the microbiota because of differences in substrate utilization between bacterial taxa. There is an abundance of literature supporting the potent effects of diet on the abundance of gut microbiota driven by fiber, carbohydrate, protein, and lipid source and composition [9, 11, 12, 16,17,18, 34,35,36]. Other studies have demonstrated that the effect of abnormal diets on gut microbiota might stifle the underlying effect of single gene mutations because diets are such a potent regulator of microbial abundances [16, 37]. These authors have called for further study of diets varied in macronutrient content and study of more complex genetic models. The current study has demonstrated that high fat diets varied in carbohydrate content continue to be commanding predictors of abundances of gut microbial abundances even in a more complex genetic model. We will highlight below many results for which genetic effects are likely dependent on specific ingredients in one of the two diets and discuss the importance of incorporating human-comparable diets into microbial studies.

Another unique feature of the current study is the incorporation of latent variables harbored in the PCoA of ASV data. Latent variables are those that are not directly observable in a model but can be inferred from other variables and can hold important information for interpreting biological relationships. PCoA of ASV data revealed that PCo1 captures the variance in ASV caused by diet while it was less clear what variance was captured by PCo2. Extraction of eigenvectors from these principal components is one way to incorporate information from the latent variables contained in the PCoA.

In this study, 18 loci diet and sex-independent loci were identified. Out of these loci, Coriobacteriaceae is the least influenced by diet. Coriobacteriaceae is the only ASV for which the top marker explains a greater proportion of the variation in the abundance of the organism than diet, suggesting that there is a genetic predisposition to having higher or lower abundances of Coriobacteriaceae. Coriobacteriaceae has previously been associated with host genetics and QTL regulating immune function and susceptibility to carcinoma and tumor development in mice [38, 39]. Coriobacteriaceae has been described as a dominant species in the mammalian gut and it is positively correlated with hepatic triglyceride concentration and non-HDL cholesterol concentration in mice [40].

A significant proportion of the variation in all other ASV with loci detected by the diet and sex-independent model is still driven by diet, especially for Streptococcus at Asvq9 and Asvq10. Streptococcus belongs to the Firmicutes phylum. Fiber is a particularly important dietary component for modulating abundance of Firmicutes. When animals switch from a low fat/fiber rich plant diet to a high fat/high sugar diet, they experience a significant increase in the Firmicutes phylum along with a decrease in Bacteroidetes [17]. Dramatic shifts were observed in these phyla between American and ketogenic diet F2s irrespective of their genetic backgrounds. Our ketogenic diet is composed of twice as much soluble and insoluble fiber as the American diet, and this likely drives many of the differences in the abundance of OTUs from these phyla. The relative abundance of Firmicutes in F2s exposed to the ketogenic diet is nearly twice as high as F2s exposed to the ketogenic diet. It appears that this increase in Firmicutes in the F2s exposed to the ketogenic diet coincides with a decrease in the relative abundance of Bacteroidetes. Limited evidence suggests that a higher Firmicutes to Bacteroidetes ratio is positively correlated with obesity while a decrease in this ratio has been associated with inflammatory bowel disease [41]. However, controversy surrounds the association of the Firmicutes to Bacteroidetes ratio and health status [41].

Six diet-specific QTL were also identified under the QTL by diet model. Of note, all six diet-specific QTL are for microbial features from either the Firmicutes or Bacteroidetes phyla. These QTL include, Asvq19, Asvq20, Asvq21, Asvq22, and Asvq5 for Muribaculaceae (Bacteroidetes), Asvq23 for Rikenelleceae RC9 Gut Group (Bacteroidetes), Asvq24 for Parabacteroides (Bacteroidetes), Asvq25 for Lactobacillus (Firmicutes), Asvq9 and Asvq10 for the Streptococcus genus (Firmicutes), Asvq26 for the family of Clostridiales vadinBB60 (Firmicutes), and Asvq27 for the Lachnospiraceae family (Firmicutes). This provides further support for previous findings suggesting that the ratio of Firmicutes to Bacteroidetes is relevant to metabolic disease states. Diet is the strongest predictor for these ASV except for the abundances of Parabacteroides, Lactobacillus, and Clostridiales vadin BB60 (all Firmicutes). The gene-by-diet interaction is most prominent for these three exceptions. Diet-specific QTL are the most clinically actionable observations as they identify a subgroup of the population that would be sensitive to a dietary intervention to modify the microbial trait. Loci identified by the sex- and diet-independent QTL and QTL by sex models are informative but do not provide the same type of direct avenue for intervention.

The American diet contains multiple sources of animal proteins, some of which contribute to the total fat content of the diet, while the main protein source in the ketogenic diet is casein. The fat component of the ketogenic diet is composed of equal parts butter and lard with a small portion of corn and menhaden oils, while the fat component of the American diet is a more diverse mixture of primarily butter as well as corn, menhaden, flaxseed, olive oil, and fat derived from the animal proteins. Lard-derived fat has been shown to reduce the abundance of Streptococcus [11]. This may explain, in part, why we observe that FVB alleles on the American diet are associated with higher abundances of two Streptococci ASV.

For the only loci picked up by the QTL by sex model, Asvq28 for Alistipes, the strongest predictor in the model was again diet. Sex specificity for abundance of Alistipes has been established in studies of pre- and post-menopausal women and men. Men were more likely than pre- or post-menopausal women to have higher abundances of Alistipes in their fecal samples [42]. The realized importance of sex as a biological variable has increased attention paid to the role of steroid hormones in development of obesity and Metabolic Syndrome [23]. Plasma testosterone has also been linked to microbial features in men, and the post-menopausal microbiome becomes more similar to the male microbiome over time [42].

QTL for Rikenella (Asvq7), Ruminiclostridium (Asvq16), Bilophila (Asvq17), Bray–Curtis PCo2 (Bcpc2q), fat mass gained during the feeding trial (Fmgq1), and serum HDL cholesterol concentration (Hdlq1) overlap on the distal part of Chr 1 and QTL for the Rikenelleceae RC9 Gut Group (Asvq23), Parabacteroides (Asvq24) and weighted UniFraction PCo1 (Wufpc1q) overlap on Chr 16. American or westernized diets are associated with increased abundances of Bilophila wadsworthia, which coincides with increased LDL cholesterol concentration and links this species of Bilophila to dyslipidemia and increased inflammation [43]. Gut microbiota signatures from overweight and obese patients have been associated with significant decreases in Rikenella and Parabacteroides species as well as increases in Ruminococcus species in the same subjects [44]. Rikenellaceae RC9 gut group has been associated with lipid metabolism in response to high fat diets [15, 45]. Previous associations between these organisms and metabolic traits make the overlapping loci associated with them higher priority for future analyses.

As mentioned previously, gut microbiota utilizes nutrients passing through the gastrointestinal tract. Microbial metabolism of these nutrients produces metabolites and microbial-derived metabolites known to impact metabolic health [1]. These metabolites may represent latent variables linking the genomic region underlying Fmgq1, Hdlq1, Bcpc2q, Asvq7, Asvq16, and Asvq17 and each of their associated traits as well as Wufpc1q, Asvq23, and Asvq24 and their associated traits.

The SEM for traits mapping to distal Chr 1 illustrated direct effects of the FVB/FVB genotype at Bcpc2q increasing abundance of Bilophila and Ruminiclostridium 9. A covariance pathway was detected between Bcpc2q and Rikenella and while the directionality of this relationship was not defined by the model, this suggests that the FVB/FVB genotype at Bcpc2q also increased the abundance of Rikenella. We observed direct effects of Bilophila, Ruminiclostridium 9 and Rikenella on Bray Curtis PC2 in addition to the direct effect of diet. In addition to the direct effects of Bilophila and Ruminiclostridium 9, we identified a covariance pathway between these organisms that likely contributes to the overall relationship of these microbiota with Bray–Curtis PC2. While the microbiota and metabolic traits appear to be independently linked to Bcpc2q, another covariance pathway is observed between Bray–Curtis PCo2 and the amount of fat mass gained during the feeding trial. Taken together, these observations suggest that Bilophila, Ruminiclostridium 9, and Rikenella are driving differences in microbial beta diversity represented in Bray–Curtis PCo2, and the overall composition of the microbiome may be correlated with the amount of fat mass gained during the feeding trial. Species that other species in an ecosystem rely heavily upon are referred to as keystone species and drive diversity within the ecosystem [46, 47]. These results suggest that Bilophila, Ruminiclostridium 9, and Rikenella are candidate keystone species. The covariance pathway observed between the Bray–Curtis index PCo2 and the amount of fat mass gained might reflect a more complex relationship between the overall composition of the gut microbiota and its effects on metabolic features.

The SEM for traits mapping to distal Chr 16 illustrated direct effects of Wufpc1q interacting with diet on the abundances of Rikenellaceae RC9 Gut Group and Parabacteroides as well as a direct effect of diet and Parabacteroides on weighted UniFraction PCo1. Parabacteroides is a diet-specific, candidate keystone species. The observed a covariance pathway between these two organisms suggests Rikenellaceae RC9 Gut Group may be an additional candidate keystone species driving differences in the composition of the microbiota in a diet-specific manner.

Fluctuations in the abundance of these organisms would have dramatic consequences on other organisms in the ecosystem. Bilophila, Ruminicostridium 9 and Rikenella represent candidates for keystone species among the organisms mapping to the distal region of Chr 1. Their direct effects on Bray–Curtis PCo2 detected in the structural equation model suggest abundances of these organisms drive differences in beta diversity. The proposed model lends itself to this speculation if the abundance of Bilophila has consequences for bile acid composition and abundances of other microbiota in the large intestine as described by others [48, 49]. We have previously demonstrated that the FVB/FVB genotype drives higher serum HDL cholesterol concentration at the locus Hdlq1, likely through Apolipoprotein A2 (Apoa2) [23]. Apoa2 is also a primary candidate gene of interest within the confidence intervals for Asvq7, Asvq16, Asvq17, and Bcpc2q. HDL cholesterol is a preferred precursor to bile acid synthesis and secretion [50]. Despite there being no direct relationship observed between abundance of these organisms and metabolic traits in the current model, these basic biological associations leave ample space for future analyses into what is likely a more complex network of latent variables tying together these microbial and metabolic traits. Additional candidate genes of interest that are expressed in the intestines were identified within the Bcpc2q interval (F11r, Fcer1g, Ndufs2 and Sdhc). F11r and Fcer1g are found on KEGG pathways primarily related to the immune system while Ndufs2 and Sdhc are found on the Non-alcoholic fatty liver disease pathway (mmu04932).

Parabacteroides was also identified as a candidate keystone species among the organisms mapping to the distal part of Chr 16. We were unable to narrow the list of positional candidate genes at Wufpc1q harboring non-synonymous transcript variants with the KEGG pathways included in the query. However, the vast majority of genes at Wufpc1q were annotated with the epithelial barrier and immune system pathways such as Tight junction (mmu04530) and Inflammatory bowel disease (mmu05321) and related pathways. Other types of variants were present in genes within the Wufpc1q confidence interval such as, synonymous transcript variants and intronic variants which may be of interest in future analyses. For example, Sod1 harbors an intron variant that diverges between the two strains and has previously been associated with both the ratio of Firmicutes to Bacteroidetes as well as obesity, providing direct evidence for variants in Sod1 regulating microbial diversity and a possible link to metabolic traits like obesity [51, 52]. Of note, Parabacteroides and Rikenelleceae RC9 gut group both belong to the Bacteroidetes phylum. Future work will focus on confirming causal relationships between candidate keystone species and measures of beta diversity.

Our report is limited in part by the choice to generate a unidirectional F1 and subsequently, F2 population. This precludes the ability to identify epistatic interactions between the autosomal genome and either the Y chromosome or the mitochondrial genome because only the B6 mitochondrial genome and FVB Y chromosome is present in our F2 population. However, this makes us more certain that the findings we have reported are not driven by the paternal chromosome nor the mitochondria. Also, the choice of the 16S V4 region per the EMP 16S analysis protocol described in the methods may limit which bacteria are identified. However, we were careful to report taxonomic assignments only up to the genus level with the understanding that taxonomic assignments after the genus level using a small 16S V4 region are not always reliable.

Conclusions

The current experiment identified organisms for which irrespective to genetic background, diet was the strongest predictor of gut microbiota, organisms for which combinations of sex, diet, and genotypes predictor the gut microbiota, as well as a single organism for which genetic background was the strongest predictor for bacterial, Coriobacteriaceae UCG-002. These results demonstrate the effect that sex, diet, and genetic background have on inter-individual differences in gut microbiota. While diet and genotype-dependent QTLs for microbial abundance are the most clinically relevant regarding efforts to advance precision nutrition, diet-dependent observations are likely related to specific ingredients in the diets which makes these observations heavily context dependent and difficult to recapitulate from investigator to investigator when non-human comparable ingredients are used in the preclinical setting. We observed that nearly all microbial QTL, even those that were identified under the QTL and QTL by sex models, were potently influenced by diet. As such, care should be taken to utilize diets composed of diverse ingredients in preclinical trials to better recapitulate the host-microbiome environment in humans. Precision nutrition will be advanced through integration of genetic variation, microbiota variation, and sex in response to diets varied in carbohydrate composition to elucidate the composition of the “ideal” microbiome and personalized interventions to achieve that composition.

Availability of data and materials

16S V4 Sequences are publicly available on the SRA database under the Bioproject ID “PRJNA803237.”

References

  1. Lim MY, You HJ, Yoon HS, et al. The effect of heritability and host genetics on the gut microbiota and metabolic syndrome. Gut. 2017;66(6):1031–8. https://doi.org/10.1136/gutjnl-2015-311326.

    Article  CAS  PubMed  Google Scholar 

  2. Roth C, Sims T, Rodehutscord M, Seifert J, Camarinha-Silva A. The active core microbiota of two high-yielding laying hen breeds fed with different levels of calcium and phosphorus. Front Physiol. 2022;13. https://doi.org/10.3389/fphys.2022.951350

  3. Haas V, Vollmar S, Preuß S, Rodehutscord M, Camarinha-Silva A, Bennewitz J. Composition of the ileum microbiota is a mediator between the host genome and phosphorus utilization and other efficiency traits in Japanese quail (Coturnix japonica). Genet Sel Evol. 2022;54(1):20. https://doi.org/10.1186/s12711-022-00697-8.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  4. Goodrich JK, Waters JL, Poole AC, et al. Human genetics shape the gut microbiome. Cell. 2014;159(4):789–99. https://doi.org/10.1016/j.cell.2014.09.053.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  5. Bonder MJ, Kurilshikov A, Tigchelaar EF, et al. The effect of host genetics on the gut microbiome. Nat Genet. 2016;48(11):1407–12. https://doi.org/10.1038/ng.3663.

    Article  CAS  PubMed  Google Scholar 

  6. Barrington WT, Wulfridge P, Wells AE, et al. Improving metabolic health through precision dietetics in mice. Genetics. 2018;208(1):399–417. https://doi.org/10.1534/genetics.117.300536.

    Article  CAS  PubMed  Google Scholar 

  7. Wells A, Barrington WT, Dearth S, et al. Tissue Level Diet and Sex-by-Diet Interactions Reveal Unique Metabolite and Clustering Profiles Using Untargeted Liquid Chromatography-Mass Spectrometry on Adipose, Skeletal Muscle, and Liver Tissue in C57BL6/J Mice. J Proteome Res Published online. 2018. https://doi.org/10.1021/acs.jproteome.7b00750.

    Article  Google Scholar 

  8. Cuomo D, Porreca I, Ceccarelli M, et al. Transcriptional landscape of mouse-aged ovaries reveals a unique set of non-coding RNAs associated with physiological and environmental ovarian dysfunctions. Cell Death Discov Published online. 2018. https://doi.org/10.1038/s41420-018-0121-y.

  9. Huda MN, Salvador AC, Barrington WT, et al. Gut microbiota and host genetics modulate the effect of diverse diet patterns on metabolic health. Front Nutr. 2022;9. https://doi.org/10.3389/fnut.2022.896348

  10. Snijders AM, Langley SA, Kim YM, et al. Influence of early life exposure, host genetics and diet on the mouse gut microbiome and metabolome. Nat Microbiol. 2016;2. https://doi.org/10.1038/nmicrobiol.2016.221

  11. Singh RK, Chang HW, Yan D, et al. Influence of diet on the gut microbiome and implications for human health. J Transl Med. 2017;15(1). https://doi.org/10.1186/s12967-017-1175-y

  12. Gentile CL, Weir TL. The gut microbiota at the intersection of diet and human health. Science (1979). 2018;362(6416). https://doi.org/10.1126/science.aau5812

  13. Cabrera-Mulero A, Tinahones A, Bandera B, Moreno-Indias I, Macías-González M, Tinahones FJ. Keto microbiota: A powerful contributor to host disease recovery. Rev Endocr Metab Disord. 2019;20(4). https://doi.org/10.1007/s11154-019-09518-8

  14. Do MH, Lee H bin, Lee E, Park HY. The effects of gelatinized wheat starch and high salt diet on gut microbiota and metabolic disorder. Nutrients. 2020;12(2). https://doi.org/10.3390/nu12020301

  15. Sun L, Jia H, Li J, et al. Cecal Gut Microbiota and Metabolites Might Contribute to the Severity of Acute Myocardial Ischemia by Impacting the Intestinal Permeability, Oxidative Stress, and Energy Metabolism. Front Microbiol. 2019;10. https://doi.org/10.3389/fmicb.2019.01745

  16. Wang JH, Shin NR, Lim SK, et al. Diet control more intensively disturbs gut microbiota than genetic background in wild type and ob/ob mice. Front Microbiol. 2019;10(JUN). https://doi.org/10.3389/fmicb.2019.01292

  17. Senghor B, Sokhna C, Ruimy R, Lagier JC. Gut microbiota diversity according to dietary habits and geographical provenance. Hum Microb J. 2018;7–8:1–9. https://doi.org/10.1016/j.humic.2018.01.001.

    Article  Google Scholar 

  18. den Hartigh LJ, Gao Z, Goodspeed L, et al. Obese mice losing weight due to trans-10, cis-12 conjugated linoleic acid supplementation or food restriction harbor distinct gut microbiota. J Nutr. 2018;148(4):562–72. https://doi.org/10.1093/jn/nxy011.

    Article  Google Scholar 

  19. Nazmul Huda M, Winnike JH, Crowell JM, O’Connor A, Bennett BJ. Microbial modulation of host body composition and plasma metabolic profile. Sci Rep. 2020;10(1). https://doi.org/10.1038/s41598-020-63214-1

  20. Caporaso JG, Lauber CL, Walters WA, et al. Ultra-high-throughput microbial community analysis on the Illumina HiSeq and MiSeq platforms. ISME J. 2012;6(8). https://doi.org/10.1038/ismej.2012.8

  21. Bolyen E, Rideout JR, Dillon M, et al. QIIME 2: Reproducible, interactive, scalable, and extensible microbiome data science. PeerJ Published online. 2018. https://doi.org/10.7287/peerj.preprints.27295.

    Article  Google Scholar 

  22. Quast C, Pruesse E, Yilmaz P, et al. The SILVA ribosomal RNA gene database project: Improved data processing and web-based tools. Nucleic Acids Res. 2013;41(D1). https://doi.org/10.1093/nar/gks1219

  23. Salvador AC, Arends D, Barrington WT, Elsaadi AM, Brockmann GA, Threadgill DW. Sex-specific genetic architecture in response to American and ketogenic diets. Int J Obes. 2021;45(6). https://doi.org/10.1038/s41366-021-00785-7

  24. Morgan AP, Fu CP, Kao CY, et al. The mouse universal genotyping array: From substrains to subspecies. G3: Genes, Genomes, Genetics. Published online 2016. https://doi.org/10.1534/g3.115.022087

  25. Reitmeier S, Hitch TCA, Treichel N, et al. Handling of spurious sequences affects the outcome of high-throughput 16S rRNA gene amplicon profiling. ISME Communications. 2021;1(1). https://doi.org/10.1038/s43705-021-00033-z

  26. Nearing JT, Douglas GM, Hayes MG, et al. Microbiome differential abundance methods produce different results across 38 datasets. Nat Commun. 2022;13(1). https://doi.org/10.1038/s41467-022-28034-z

  27. Rohrer SD, Robertson BQ, Chubiz LM, Parker PG. Gut microbiome composition associated with Plasmodium infection in the Eurasian tree sparrow. J Avian Biol. 2023;2023(1–2). https://doi.org/10.1111/jav.03027

  28. Kemis JH, Linke V, Barrett KL, et al. Genetic determinants of gut microbiota composition and bile acid profiles in mice. PLoS Genet. 2019;15(8). https://doi.org/10.1371/journal.pgen.1008073

  29. Gloor GB, Macklaim JM, Pawlowsky-Glahn V, Egozcue JJ. Microbiome datasets are compositional: And this is not optional. Front Microbiol. 2017;8(NOV). https://doi.org/10.3389/fmicb.2017.02224

  30. Kaul A, Mandal S, Davidov O, Peddada SD. Analysis of microbiome data in the presence of excess zeros. Front Microbiol. 2017;8(NOV). https://doi.org/10.3389/fmicb.2017.02114

  31. Silverman JD, Roche K, Mukherjee S, David LA. Naught all zeros in sequence count data are the same. Comput Struct Biotechnol J. 2020;18. https://doi.org/10.1016/j.csbj.2020.09.014

  32. Rosseel Y. Journal of Statistical Software Lavaan: An R Package for Structural Equation Modeling.; 2012. http://www.jstatsoft.org/

  33. Mîndrilă D. Maximum Likelihood (ML) and Diagonally Weighted Least Squares (DWLS) Estimation Procedures: A Comparison of Estimation Bias with Ordinal and Multivariate Non-Normal Data.; 2010.

  34. Leamy LJ, Kelly SA, Nietfeldt J, et al. Host Genetics and Diet, but Not Immunoglobulin A Expression, Converge to Shape Compositional Features of the Gut Microbiome in an Advanced Intercross Population of Mice.; 2014. http://genomebiology.com/2014/15/12/552

  35. David LA, Maurice CF, Carmody RN, et al. Diet rapidly and reproducibly alters the human gut microbiome. Nature. 2014;505(7484):559–63. https://doi.org/10.1038/nature12820.

    Article  CAS  PubMed  Google Scholar 

  36. Hansen NW, Sams A. The microbiotic highway to health—New perspective on food structure, gut microbiota, and host inflammation. Nutrients. 2018;10(11). https://doi.org/10.3390/nu10111590

  37. Zhang C, Zhang M, Wang S, et al. Interactions between gut microbiota, host genetics and diet relevant to development of metabolic syndromes in mice. ISME J. 2010;4(2):232–41. https://doi.org/10.1038/ismej.2009.112.

    Article  CAS  PubMed  Google Scholar 

  38. Benson AK, Kelly SA, Legge R, et al. Individuality in gut microbiota composition is a complex polygenic trait shaped by multiple environmental and host genetic factors. Proc Natl Acad Sci U S A. 2010;107(44):18933–8. https://doi.org/10.1073/pnas.1007028107.

    Article  PubMed  PubMed Central  Google Scholar 

  39. Dabrowska K, Witkiewicz W. Correlations of host genetics and gut microbiome composition. Front Microbiol. 2016;7(AUG). https://doi.org/10.3389/fmicb.2016.01357

  40. Clavel T, Desmarchelier C, Haller D, et al. Intestinal microbiota in metabolic diseases: From bacterial community structure and functions to species of pathophysiological relevance. Gut Microbes. 2014;5(4). https://doi.org/10.4161/gmic.29331

  41. Magne F, Gotteland M, Gauthier L, et al. The firmicutes/bacteroidetes ratio: A relevant marker of gut dysbiosis in obese patients? Nutrients. 2020;12(5). https://doi.org/10.3390/nu12051474

  42. Mayneris-Perxachs J, Arnoriaga-Rodríguez M, Luque-Córdoba D, et al. Gut microbiota steroid sexual dimorphism and its impact on gonadal steroids: Influences of obesity and menopausal status. Microbiome. 2020;8(1). https://doi.org/10.1186/s40168-020-00913-x

  43. Kazemian N, Mahmoudi M, Halperin F, Wu JC, Pakpour S. Gut microbiota and cardiovascular disease: Opportunities and challenges. Microbiome. 2020;8(1). https://doi.org/10.1186/s40168-020-00821-0

  44. Palmas V, Pisanu S, Madau V, et al. Gut microbiota markers associated with obesity and overweight in Italian adults. Sci Rep. 2021;11(1). https://doi.org/10.1038/s41598-021-84928-w

  45. Zhou L, Xiao X, Zhang Q, et al. Improved glucose and lipid metabolism in the early life of female offspring by maternal dietary genistein is associated with alterations in the gut microbiota. Front Endocrinol (Lausanne). 2018;9(SEP). https://doi.org/10.3389/fendo.2018.00516

  46. Fisher CK, Mehta P. Identifying keystone species in the human gut microbiome from metagenomic timeseries using sparse linear regression. PLoS One. 2014;9(7). https://doi.org/10.1371/journal.pone.0102451

  47. Mouquet N, Gravel D, Massol F, Calcagno V. Extending the concept of keystone species to communities and ecosystems. Ecol Lett. 2013;16(1):1–8. https://doi.org/10.1111/ele.12014.

    Article  PubMed  Google Scholar 

  48. Ijssennagger N, van Rooijen KS, Magnúsdóttir S, et al. Ablation of liver Fxr results in an increased colonic mucus barrier in mice. JHEP Reports. 2021;3(5):100344. https://doi.org/10.1016/j.jhepr.2021.100344

  49. Devkota S, Wang Y, Musch MW, et al. Dietary-fat-induced taurocholic acid promotes pathobiont expansion and colitis in Il10-/- mice. Nature. 2012;487(7405):104–8. https://doi.org/10.1038/nature11225.

    Article  CAS  PubMed  PubMed Central  Google Scholar 

  50. Röhrl C, Eigner K, Fruhwürth S, Stangl H. Bile acids reduce endocytosis of High-Density Lipoprotein (HDL) in HepG2 cells. PLoS One. 2014;9(7). https://doi.org/10.1371/journal.pone.0102026

  51. Lewandowski Ł, Kepinska M, Milnerowicz H. Alterations in concentration/activity of superoxide dismutases in context of obesity and selected single nucleotide polymorphisms in genes: Sod1, sod2, sod3. Int J Mol Sci. 2020;21(14):1–32. https://doi.org/10.3390/ijms21145069.

    Article  CAS  Google Scholar 

  52. Sagi H, Shibuya S, Kato T, et al. SOD1 deficiency alters gastrointestinal microbiota and metabolites in mice. Exp Gerontol. 2020;130. https://doi.org/10.1016/j.exger.2019.110795

Download references

Acknowledgements

We thank Dr. Andrew Hillhouse and the Texas A&M Institute for Genome Sciences and Society (TIGSS) for assistance in the TIGSS Molecular Genomics Workspace and TIGSS Rodent Preclinical Phenotyping Core; Dr. William Barrington for initiating the F2 population; Dr. Karl Broman for his discussion and guidance during the QTL analysis as well as for organizing and managing annotation files for the MUGA arrays; Daniel Genung and Aaron Van Wettering for assistance with mouse husbandry and phenotyping; and Ryan McGovern for assistance with preparing samples for genotyping; and all members of the laboratory for their helpful insights.

Funding

This work was supported by National Institutes of Health (NIH) grants RM1HG008529, R01DK130333, and P30ES029067. Funding for BJB and NH was provided by the U.S. Department of Agriculture (USDA), Agricultural Research Service CRIS projects 2032–51530-025-00D. The USDA is an equal opportunity employer and provider.

Author information

Authors and Affiliations

  1. Department of Molecular and Cellular Medicine, Texas A&M Health Science Center, College Station, TX, 77843, USA

    Anna C. Salvador, Ahmed M. Elsaadi, C. Anthony Gacasan & David W. Threadgill

  2. Department of Nutrition, Texas A&M University, College Station, TX, 77843, USA

    Anna C. Salvador & David W. Threadgill

  3. Department of Nutrition, University of California Davis, Sacramento, CA, 95616, USA

    M. Nazmul Huda & Brian J. Bennett

  4. Obesity and Metabolism Unit, Western Human Nutrition Research Center, USDA-ARS, Davis, CA, 95616, USA

    M. Nazmul Huda & Brian J. Bennett

  5. Albrecht Daniel Thaer-Institut, 10115, Berlin, Germany

    Danny Arends & Gudrun A. Brockmann

  6. Department of Applied Sciences, Northumbria University, Newcastle Upon Tyne, UK

    Danny Arends

  7. Department of Genetics, University of North Carolina at Chapel Hill, Chapel Hill, NC, 27599, USA

    William Valdar

  8. Lineberger Comprehensive Cancer Center, University of North Carolina at Chapel Hill, Chapel Hill, NC, 27599, USA

    William Valdar

  9. Department of Biochemistry & Biophysics, Texas A&M University, College Station, TX, 77843, USA

    David W. Threadgill

Authors
  1. Anna C. Salvador
    View author publications

    You can also search for this author in PubMed Google Scholar

  2. M. Nazmul Huda
    View author publications

    You can also search for this author in PubMed Google Scholar

  3. Danny Arends
    View author publications

    You can also search for this author in PubMed Google Scholar

  4. Ahmed M. Elsaadi
    View author publications

    You can also search for this author in PubMed Google Scholar

  5. C. Anthony Gacasan
    View author publications

    You can also search for this author in PubMed Google Scholar

  6. Gudrun A. Brockmann
    View author publications

    You can also search for this author in PubMed Google Scholar

  7. William Valdar
    View author publications

    You can also search for this author in PubMed Google Scholar

  8. Brian J. Bennett
    View author publications

    You can also search for this author in PubMed Google Scholar

  9. David W. Threadgill
    View author publications

    You can also search for this author in PubMed Google Scholar

Contributions

ACS and DWT conceived and designed the study. AME and ACG performed mouse experiments. NH prepared and performed 16S rRNA V4 sequencing analysis. ACS performed the linkage analysis. ACS, DWT, DA, GB contributed to the linkage analysis pipeline. ACS, WV, DWT performed statistical analysis. ACS and DWT designed the figures and wrote the manuscript. All authors revised the manuscript.

Corresponding author

Correspondence to David W. Threadgill.

Ethics declarations

Ethics approval and consent to participate

Not applicable.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.

Additional information

Publisher's Note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Supplementary Information

Additional file 1: Supplementary Figure S1.

Genetic Map. 1,667 markers that were polymorphic between B6 and FVB were used for the association analyses. Supplementary Figure S2. Jaccard Index. PC1 and PC2 describe 21.3% and 10.6% of the variation in ASV respectively. Table S1. Diet compositions. Table S2. ARRIVE criteria used in the study. Table S3. Comprehensive list of all KEGG querie. Table S4. Effect of sex and diet on phyla abundance.

Rights and permissions

Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated in a credit line to the data.

Reprints and permissions

About this article

Check for updates. Verify currency and authenticity via CrossMark

Cite this article

Salvador, A.C., Huda, M.N., Arends, D. et al. Analysis of strain, sex, and diet-dependent modulation of gut microbiota reveals candidate keystone organisms driving microbial diversity in response to American and ketogenic diets. Microbiome 11, 220 (2023). https://doi.org/10.1186/s40168-023-01588-w

Download citation

  • Received:

  • Accepted:

  • Published:

  • DOI: https://doi.org/10.1186/s40168-023-01588-w

Keywords

Microbiome

ISSN: 2049-2618

Contact us