Vertical selection for nuclear and mitochondrial genomes shapes gut microbiota and modifies risks for complex diseases

Abstract

Here we postulate that the heritability of complex disease traits previously ascribed solely to the inheritance of the nuclear and mitochondrial genomes is broadened to encompass a third component of the holobiome, the microbiome. To test this, we expanded on the selectively bred low capacity runner/high capacity runner (LCR/HCR) rat exercise model system into four distinct rat holobiont model frameworks including matched and mismatched host nuclear and mitochondrial genomes. Vertical selection of varying nuclear and mitochondrial genomes resulted in differential acquisition of the microbiome within each of these holobiont models. Polygenic disease risk of these novel models were assessed and subsequently correlated with patterns of acquisition and contributions of their microbiomes in controlled laboratory settings. Nuclear-mitochondrial-microbiotal interactions were not for exercise as a reporter of health, but significantly noted for increased adiposity, increased blood pressure, compromised cardiac function, and loss of long-term memory as reporters of disease susceptibility. These findings provide evidence for coselection of the microbiome with nuclear and mitochondrial genomes as an important feature impacting the heritability of complex diseases.

INTRODUCTION

Humans are “holobionts” or host macro-organisms in symbiotic relationships with commensal micro-organisms that reside within the host. Microorganisms, bacteria in particular, are abundant in the host distal gut collectively termed as gut microbiota (GM). In recent years, it has become quite clear that gut commensal microbiota contribute to the overall health and well-being of the invertebrate and mammalian host (28). The gut acquires bacteria by vertical transmission from the mother to offspring or horizontal transmission throughout life via social interactions and shared environments. Regardless of the route or timing of acquisition, human epidemiological studies suggest that distinct microbiota can contribute to the heritability for disease risk phenotypes (13). This cosegregation between microbiota and disease risk lends support to the view that host allelic fixation is perhaps an important factor determining GM composition and function (30). Over the past decade, there is growing evidence that the holobiont acts at an independent level of selection within evolution (36). The concept of evolution of the hologenome specifies that, in mammals, host nuclear genome, mitochondrial genome, and microbiome are an ensemble that form a unit of genetic selection that can propagate from one generation to the next to maintain structure and function. Thus, the hologenome is an entity that describes, in part, a level of biological organization for understanding the coevolution of host and symbiont genes (40).

Since the introduction of endosymbiotic theory from Lynn Margulis in the late 1960s, the critical need for experimental models to study the microbiome-mitochondrion relationship and its evolution-based influence on health and disease is clear (4). In mammals, mitochondrial DNA is unique in that it is a small (15–17 kb) circle of double-stranded DNA that codes for 13 proteins essential for mitochondrial respiration and transmits down the maternal line of inheritance. Increasing evidence supports the view that the GM produces metabolites that influence mitochondrial function and biogenesis within a cell type to increase ATP production. The first indication that mutations in mitochondrial genes significantly affect GM composition comes from a study in mice. Hirose et al. (14) generated single mutations in ATP8 synthase gene (location: m.7778 G>T) encoded by the mitochondrial DNA (mtDNA) from a female mitochondrial donor strain onto C57BL/6J genomic background by repeated backcrossing (conplastic strain C57BL/6J-mt^FVB/NJ). Then, they compared gut microbial communities from two different conplastic strains (C57BL/6J-mt^FVB/NJ and C57BL/6J-mt^NZB/BlnJ) of mice to the original C57BL/6J strain. The results show that a strain carrying a mutation in the mitochondrial ATP8 synthase gene demonstrated a higher Firmicutes-to-Bacteroidetes ratio along with phenotypes related to metabolic and inflammatory disorders. The authors suggest that the mtATP8 mutation causes a modulation of oxidative-phosphorylation and glycolysis pathways that drives a shift in the microbial composition of the gut as a consequence of host metabolism. These results further suggest that the GM play a synergistic role in regulating the mitochondrial bioenergetics of the host and has the potential to regulate host genomics, ultimately leading to an increased risk for complex diseases.

Quantitating the extent to which the hologenome coevolves to influence the fitness of the host is hindered by the lack of good experimental designs that can address this question and dissect the individual contributions. Using laboratory rat models that have undergone several generations of two-way artificial selective breeding, we designed this study to test the hypothesis that the host genome is indeed partly responsible for the composition of the GM. The rat models termed high capacity runner (HCR) and low capacity runner (LCR) rats serve as a contrasting, genetically heterogeneous animal model resource for the study of health and disease respectively (19, 20). The basis of the HCR and LCR contrasting model system is to test the hypothesis that variation in energy transfer is the major determinant in the divide for health and disease (energy transfer hypothesis). The HCR rats perform eightfold higher for treadmill running capacity, have increased mitochondrial function, are protected from obesity and metabolic syndrome (32, 45), and have a longer lifespan than the LCR rats (21). Recent papers show that HCR and LCR rats divide for diversity and composition of GM with diet and aging (7, 34, 35), yet the purpose for this segregation is unclear.

For the intention of this study, we expand on the selectively bred LCR and HCR rat exercise model system into a genetically defined rat model framework for studying holobionts in controlled laboratory settings. To test our first hypothesis that vertical selection of the host nuclear genome via the fixation process is sufficient to restructure the GM, we developed four new strains (Fig. 1). 1) We developed standard inbred models by brother-sister mating >20 generations to experimentally compare LCR vs. HCR hologenomes. In addition, 2) we developed two reciprocal conplastic strains, one with nuclear DNA (nDNA) from HCR and mtDNA from LCR, and the other with nDNA from LCR and mtDNA from HCR. We further hypothesized that the predominantly vertical transmission of microbiome-mitochondrion from mother to offspring in the LCR and HCR inbred and conplastic strains will associate with the low and high inheritance of cardiometabolic disease risks exhibited within the selectively bred models, whereas the conplastic strains with mismatched matrilineal contributions will disrupt these connections.

Fig. 1.

Schematic diagram for development of inbred and conplastic strains. HCR, high-capacity runner; LCR, low-capacity runner; BC, backcross.

Our results demonstrate that the GM are dependent on the host genomes and closely associate with multiple disease risk traits including increased adiposity, blood pressure, maladaptive response to cardiac heart function and stress. This study provides a first report in strong support of the theory that the host genomes operate as a permissive determinant of the gut microbiome. We provide additional phenotypic data within these novel holobiont rat models demonstrating that coevolved host and microbiome genomes interact and affect host metabolic, cardiovascular disease risks, behavior, and cognitive function. The animal models generated through this study are useful genetic tools for further dissection of the relationships between the triad constituting the hologenome- the nuclear genome, the mitochondrial genome, and the microbiome.

METHODS

Animals.

Rats were bred and maintained on a low-salt diet (0.3% NaCl, Harlan Teklad diet 7034; Madison, WI). All procedures were approved by the University of Toledo Institutional Animal Care and Use Committee in accordance with the National Institutes of Health Guide for the Care and Use of Laboratory Animals. Breeding stock for the inbred and conplastic strains were the extreme representatives from generation 23 of the selectively bred LCR and HCR lines and developed at The University of Toledo College of Medicine and Life Sciences, Toledo, Ohio. Per the definition of an inbred strain by the rat and mouse strain nomenclature committee, 20 generations meets the required number of generations for designating a rat line as an inbred rat strain (MGI-Guidelines for Nomenclature of Mouse and Rat Strains, http://www.informatics.jax.org/mgihome/nomen/strains.shtml). Mismatched mitochondrial strains were developed by taking advantage of the maternal inheritance of mitochondrial genome. In brief, a female HCR progenitor rat was crossed with a male LCR progenitor. The resultant F1 female offspring was backcrossed with male LCR from the contemporaneous inbreeding colony. This backcross procedure was repeated 17 additional times to generate LCR.HCR^mt/Tol strain. Similarly, a female LCR progenitor was bred with a male HCR progenitor. The F1 female offspring were backcrossed with male HCR from contemporaneous inbreeding colony. This backcross procedure was repeated 17 additional times to generate HCR.LCR^mt/Tol strain and maintained in subsequent generations by brother-sister mating.

Aerobic exercise capacity test.

Aerobic running capacity test was conducted in 10 wk old rats by using a standard procedure (19). In brief, for the first week, rats run on a treadmill with 15° slope for increasing duration, until they are able to run at least 5 min. For the second week, rats underwent run-to-exhaustion tests on 5 consecutive days. The initial velocity was 10 m/min and increased by 1 m/min every 2 min. The test was stopped at the third attempt when a rat could not keep pace on the treadmill. The best running time and distance out of the 5 days was taken as one estimate of intrinsic aerobic capacity.

Blood pressure measurement by radiotelemetry.

Blood pressure at the age of 34 wk was recorded by radiotelemetry method as described previously (26). In brief, each rat was anesthetized under the influence of isoflurane, a radiotelemetry transmitter (HDS10; Data Sciences International, St Paul, MN) was surgically implanted into the rat’s femoral artery, and the rat was housed individually. Rats were allowed to recover from surgery for 3 days before we recorded their blood pressure. Blood pressure was collected for 2 consecutive days, and a moving average of 2 h were taken as final blood pressure values.

Echocardiography.

Left ventricular function was evaluated at the age of 16 wk by echocardiography (Sequoia C512 System, Siemens Medical) as described previously (31). In brief, left ventricular dimensions were measured in the M-mode, including end-diastolic (LVEDD) and end-systolic (LVESD) diameter. Fractional shortening (FS) was calculated with the formula ([LVEDD-LVESD]/LVEDD) × 100%, where the relative wall thickness (RWT) was calculated with the formula [posterior wall thickness in diastole + septum wall thickness in diastole]/LVEDD. Ejection time and contraction and relaxation times were measured from Doppler pulsed-wave traces of mitral and aortic flow obtained at the level of the left ventricular outflow tract from the apical four-chamber view. The mean velocity of circumferential shortening (mVCF) was calculated as FS/ejection time. Cardiac indices was calculated as cardiac output/body weight.

Blood cell count.

Whole blood was draw from retro-orbital plexus for each animal under anesthesia in EDTA containing tubes. Hematology of peripheral blood was immediately determined with the VetScan HM5 hematology analyzer (Abaxis, Union City, CA).

Behavior tests.

Locomotor activity was assessed with an open field apparatus (72 × 72 × 28 cm). Lines drawn on the floor with a marker divided the floor into 36 12 × 12 cm squares with a central square (24 × 24 cm). Each rat was placed in the center of the open field and allowed to freely explore the apparatus for 5 min. The frequency of line crossing was counted as an indicator for total activity. Time spent in the center square and spent grooming or rearing was recorded manually.

Anxiety-like behavior was assessed with the elevated plus maze as described (44). Rat behavior was recorded during a 5 min test period. The time spent on open and closed arms were recorded. Depressive-like behavior was assessed by the forced swim test (39). For the forced swim test, rats were placed in a clear tank (25.4 cm in diameter and 40.6 cm in height) containing 25°C water to a level of 30 cm that prevented the rat from touching the bottom of the tank. Testing occurred over 2 consecutive days with a 15 min swim on day 1 (pretest) and a 5 min swim on day 2 (test). Time spent immobile (defined as the rat floating in the water without struggling and only making movements necessary to keep its head above water) was scored. Swimming and climbing were defined as horizontal movements throughout the swim cylinder and upward-directed movement respectively.

The Barnes maze was used to test spatial learning and memory, as previously described in detail (37). The Barnes maze is a circular platform with 20 equally spaced holes (10.5 cm in diameter; 5.5 cm between holes) and elevated 70 cm above the floor with spatial cues placed around the maze. An escape hole was equipped with a small chamber attached underneath the platform. In brief, rats were given two training trials a day for 4 days, with a 15 min interval. To start a trial, animal was placed into a start box located in the center of the maze for 10 s. Once released, bright light and loud noise were presented to stimulate animal to find the escape hole in 2 min. The light and noise were turned off once the rats find the escape hole. The time required by each animal to reach the escape hole was recorded as the escape latency. We also recorded the number of errors before the animal reached the goal. On the fifth and 12th days, short-term and long-term memory were assessed via probe tests respectively. At this time, the small chamber underneath the escape hole was removed, and the rat was allowed to explore the maze for 2 min. Escape latency, escape error, and time spent around the escape hole and other holes were recorded.

mtDNA sequencing and analysis.

Sequencing of mtDNA was based on the PCR amplification followed by direct sequencing of the product through commercial services (Eurofins MWG Operon Sequencing Services, Eurofins Genomics LLC, KY) as described before (25). Forward and reverse sequence data generated in the laboratory were aligned with the software Sequencher 4.10.1 (Gene Codes Corporation, Ann Arbor, MI). LCR/Tol and HCR/Tol mtDNA sequences were further comparatively analyzed based on the Brown Norway reference sequence.

Metagenomics sequencing with 16S rRNA from feces.

Bacterial DNA was extracted from fecal samples by using the MoBio Powerfecal DNA Isolation kit following the manufacturer's instructions (MoBio, Carlsbad, CA), followed by DNA quantification with a Qubit 2.0 (Invitrogen) fluorometer according to the protocol of the dsDNA High Sensitivity option. Isolated DNA samples were then subjected to PCR amplification of the V4 region of the bacterial 16s rDNA using the 515F/806R primer set. PCR was carried out on a MJ Research PTC-200 thermocycler (Bio-Rad, Hercules, CA) using the following cycling conditions: 98°C for 3 min; 35 cycles of 98°C for 1 min, 55°C for 40 s, and 72°C for 1 min; 72°C for 10 min. Pooled PCR products were gel purified with the Qiagen Gel Purification Kit (Qiagen, Frederick, MD), and clean PCR products were quantified with the Qubit 2.0 (Invitrogen) fluorometer. Prior to submission for sequencing, libraries were quality-checked with the 2100 Bioanalyzer DNA 1000 chip (Agilent Technologies, Santa Clara, CA). Metagenomics sequencing using 16S RNA was conducted on an Illumina MiSeq 2X250 bp platform at the California State University (North Ridge, CA). Paired-end sequences were trimmed at a length of 250 bp and quality-filtered at an expected error of <0.5% with USEARCH v7 (6). After quality-filtering, reads were analyzed with the QIIME 1.9.0 software package (3). A total of 1,360,896 sequences were obtained after quality-filtering and chimera-checking. Open reference operational taxonomic units (OTUs) were picked by the USEARCH61 algorithm (5), and taxonomy assignment was performed with the Greengenes 16S rRNA gene database (13-5 release, 97%) (http://greengenes.lbl.gov). Assigned taxonomy was organized into a BIOM-formatted OTU (operational taxonomic unit) table, which was summarized within QIIME 1.9.0.

The QIIME 1.9.0 analysis package was used to conduct alpha and beta diversity analysis and to find enriched taxa between groups. For alpha diversity, multiple rarefactions were conducted on sequences across all samples to a maximum depth of 9,500 sequences. Alpha diversities were then collated and plotted using observed species richness. Principal coordinates analyses plot was generated to show beta diversity, and an ANOSIM test was conducted to indicate the significance between groups. Enriched taxa between groups were identified by LefSe analysis within QIIME 1.9.0 analysis package. A LefSe plot was generated to display significantly enriched taxa [linear discriminant analysis (LDA) score > 2, P < 0.05]. Kruskal-Wallis test with Bonferroni correction was performed to compare the relative abundance of species between inbred and conplastic strains. Correlations between the relative abundance of microbiota species and phenotypes were assessed by Pearson’s correlation tests. The P values obtained were adjusted for multiple comparisons by false discovery rate (FDR) method with a corresponding q-value threshold of 0.05.

Statistical analyses.

All data, unless otherwise indicated, are shown as means ± SE. Data sets with more than two groups were analyzed by one-way ANOVA with post hoc Tukey test. Comparison of two independent groups was performed with independent sample t test. All statistical analyses were performed with SPSS software (version 23; IBM, Armonk, NY). Figures were generated with GraphPad Prism software (version 5; GraphPad Software Inc., La Jolla, CA).

RESULTS

Host genome exerts a selection pressure to influence the composition of gut microbiome.

Compared with selectively bred models, inbred strains have advantages of fixed genetic backgrounds and phenotypes allowing investigators to readily and reproducibly compare genomes and obtain insights into the genetic basis of phenotypic divergence. We developed, through over 20 generations of brother-sister mating, inbred models of LCR/Tol and HCR/Tol (Toledo strains) and through successive backcrosses constructed conplastic models by “switching” mitochondrial genomes between LCR and HCR rats but keeping their nuclear genomes unchanged and named them LCR.HCR^mt/Tol (LCR rat with mitochondrial genome of the HCR rat) and HCR.LCR^mt/Tol (HCR rat with mitochondrial genome of the LCR rat) (Fig. 1 shows a schematic diagram for development of inbred LCR and HCR rats and their reciprocal conplastic strains). We sequenced and compared mtDNA sequence with common inbred strains mtDNA sequences (obtained from GenBank nucleotide database) and found that LCR/Tol and HCR.LCR^mt/Tol mtDNA were identical to the mtDNA reported from Wistar Kyoto (WKY) inbred strain, while HCR/Tol and LCR.HCR^mt/Tol mtDNA were identical to the mtDNA reported from Fischer 344 × Brown Norway F1-hybrid strain (Supplemental Table S1; Supplemental Material available at: https://doi.org/10.6084/m9.figshare.9978893.v1).

To test whether vertical selection of host genome by inbreeding results in a coselection of GM, we sequenced DNA obtained from fecal samples of HCR/Tol, LCR/Tol, HCR.LCR^mt/Tol, and LCR.HCR^mt/Tol. Samples contained a total of 1,360,896 sequences after quality control, with a range from 13,012 to 64,523 high-quality sequences per sample. Alpha diversity was measured by observed OTUs to indicate the number of different species in a sample (“richness”). However, there was no significant difference in alpha diversity between strains (P = 0.06 for nonparametric test). ANOSIM tests were conducted to measure beta diversity, which indicated the diversity in a GM community between different samples. Interestingly, we found that GM composition was significantly different between four groups (ANOSIM test, P = 0.001) (Fig. 2A). Analysis at the phylum level revealed that the bacterial population of HCR/Tol was dominated by Bacteroidetes (40%), Firmicutes (45.1%), Proteobacteria (7.8%), and Cyanobacteria (0.47%) (Fig. 2, B and C). The proportion of sequences assigned to Bacteroidetes (57.7%) and Actinobacteria (0.69%) was significantly increased in LCR/Tol animals at the expense of Firmicutes (36.5%) and Proteobacteria (3.0%). Mitochondrial effects were noted. Replacing LCR mitochondria on the HCR increased Bacteroidetes more robustly compared with the replacement of HCR mitochondria on the LCR (Fig. 2B). Significant alterations were noted in both directions of reshaping of Proteobacteria with corresponding replacement of mitochondrial genome (Fig. 2C).

Fig. 2.

Gut microbiota analysis in inbred and conplastic strains. A: unweighted principal coordinates analysis (PCoA) plots reveals significant reshaping of gut microbiota (GM) between 4 holobionts. B, C: taxonomic compositions of gut microbiota at the phylum level. D: LefSe plots were generated to display biomarker taxa driving shifts in microbial community structure between inbred and conplastic strains. The figure displayed 43 taxa found to be significantly enriched within each group [P < 0.05, linear discriminant analysis (LDA) score > 2.0]. E: the heat map represents results of group comparisons to reveal effects of nuclear genome or mitochondrial genome on the relative abundance of enriched taxa. Blue denotes a higher level of relative abundance in HCR/Tol (columns 1, 2, 4), and in HCR.LCR^mt/Tol (column 3) compared with another strain; red denotes a higher level of relative abundance in LCR/Tol (columns 1, 3, and 5), and in LCR.HCR^mt/Tol (column 2) compared with another strain; in column 6, pink denotes a higher level of relative abundance in LCR.HCR^mt/Tol compared with HCR.LCR^mt/Tol, and light blue denotes a higher level of relative abundance in HCR.LCR^mt/Tol compared with LCR.HCR^mt/Tol. Fecal samples from 13 wk old male rats were used for 16S rRNA sequencing (HCR/Tol, n = 8; HCR.LCR^mt/Tol, n = 7; LCR.HCR^mt/Tol, n = 8; LCR/Tol, n = 8).

We then applied LDA effect size analysis to confirm our findings and identified 43 differentially abundant species between the LCR/HCR inbred and conplastic strains. (Fig. 2D). To further dissect the effect of nuclear genome and mitochondrial genome on GM, we compared the relative abundance of enriched taxa between groups (Fig. 2E). We found 18 taxa were significantly different between HCR/Tol and LCR/Tol (column 1). The nuclear genome had the most significant effect on GM composition in inbred strains. Repeated backcrossing LCR (with nDNA) to HCR (with mtDNA) significantly increased the relative abundance of Actinobacteria phylum/class, Bifidobacterium genus, Coriobacteriaceae family, and unclassified Coriobacteriaceae genus in LCR.HCR^mt/Tol, compared with HCR strain (column 2). Likewise, repeated backcrossing HCR (with nDNA) to LCR (with mtDNA) significantly increased relative abundance of Coprococcus genus and unclassified Desulfovibrionaceae genus and decreased relative abundance of Turicibacter genus, Actinobacteria phylum/class, unclassified Coriobacteriaceae, Christensenellaceae, and Peptostreptococcaceae genus in HCR.LCR^mt/Tol, compared with LCR strain (column 3). Mitochondrial effects were also noted. Replacing LCR mitochondria on the HCR significantly decreased relative abundance of Lachnospiraceae family, unclassified Lachnospiraceae genus, and Ruminococcus genus in HCR.LCR^mt/Tol, compared with HCR (column 4). Likewise, replacing HCR mitochondria on the LCR significantly decreased relative abundance of Alphaproteobacteria in LCR.HCR^mt/Tol, compared with LCR (column 5). We also observed 13 taxa as being significantly different between HCR.LCR^mt/Tol and LCR.HCR^mt/Tol (column 6) but saw no difference between inbred strains, which suggests potential effects of mismatched nuclear and mitochondrial genomes on GM composition.

Nuclear genome plays a dominant role in regulating aerobic exercise capacity, body weight, and adiposity.

As predicted from the selected traits, HCR/Tol strains had significantly higher aerobic exercise capacity than LCR/Tol strains as recorded by their best running time and distance to exhaustion (Fig. 3A). Aerobic running exercise capacity of neither the HCR/Tol nor the LCR/Tol was altered by switching their mitochondrial DNA in HCR.LCR^mt/Tol and LCR.HCR^mt/Tol, respectively. These data suggested that the nuclear genome, but not the mitochondrial or microbiome, is a dominant factor regulating aerobic exercise capacity. Second, as a previously reported correlated response to selection, LCR strains weighed more than HCR strains (Fig. 3B). There was a significant increase in body weight of the two mismatched mitochondrial strains compared with their inbred lines. In both HCR and LCR strains, the introduction of mismatched mitochondria by backcross breeding produced rats that exhibited an increased body weight at the early growth period (40–60 days) (data not shown). Compared with LCR, strains with HCR nuclear genomes exhibited significantly higher heart weight and decreased epididymal white adipose tissue (eWAT) accumulation (Fig. 3C). In line with higher body weight, LCR.HCR^mt/Tol strain exhibited elevated eWAT compared with the LCR/Tol (Fig. 3C).

Fig. 3.

Dominant effect of nuclear genome on aerobic running capacity and body weight. Aerobic exercise capacity as measured by best running time and best running distance (A) and body weight (B) in 10–11 wk old male rats (HCR/Tol, n = 37; HCR.LCR^mt/Tol, n = 45; LCR.HCR^mt/Tol, n = 24; LCR/Tol, n = 26). C: weight of heart, liver, and epididymal white adipose tissue (eWAT) in 44 wk old male rats (HCR/Tol, n = 8; HCR.LCR^mt/Tol, n = 10; LCR.HCR^mt/Tol, n = 8; LCR/Tol, n = 9). Data presented as means ± SE. Levels of statistical significance were analyzed by one-way ANOVA with post hoc Tukey test; differences in letters between bars (e.g., a, b, c) indicate statistically significant differences between strains (P < 0.05).

Cardiovascular risk factors were segregated by aerobic exercise capacity and influenced by cross talk between nuclear-mitochondrial genomes.

Aerobic exercise capacity highly depends on the efficient delivery of oxygen, which relies on cardiovascular function and red blood cells. To validate the segregation effect of intrinsic aerobic exercise capacity on cardiovascular disease risks, and test the influence of mitochondria, we evaluated blood pressure in a subset of chronically instrumented rats with radiotelemetry. Compared with the HCR/Tol, LCR/Tol showed significantly increased mean arterial pressures (MAP), systolic blood pressure (SBP), and diastolic blood pressure (DBP) (Fig. 4, A–C). However, we didn’t find any significant difference in either heart rate or activity (Fig. 4, D and E). Interestingly, by introducing HCR mitochondria to the LCR genetic background, a significant reduction in MAP and DBP, as well as marginally lower SBP relative to the LCR/Tol was noted. In contrast, there was no difference in BP parameters between HCR.LCR^mt/Tol and HCR/Tol.

Fig. 4.

Blood pressure segregation by aerobic exercise capacity and influences of nuclear-mitochondrial genomic interaction. Radiotelemetry measurement of blood pressure (A–C), heart rate (D), and activity (E) in 34 wk old male rats (HCR/Tol, n = 5; HCR.LCR^mt/Tol, n = 4; LCR.HCR^mt/Tol, n = 5; LCR/Tol, n = 5). Data are presented as the 2 h moving average of recordings obtained every 5 min continuously for 24 h or as the time-averaged blood pressures. Data are means ± SE. Levels of statistical significance were analyzed by one-way ANOVA with post hoc Tukey test; Differences in letters between bars (e.g., a, b, c) indicate statistically significant differences between strains (P < 0.05). BP, blood pressure; MAP, mean arterial pressure.

To test whether higher SBP and DBP was accompanied with cardiac remodeling, we assessed heart function by echocardiography. We found that the left ventricular FS and mVCF were significantly higher in LCR strains compared with HCR strains (Fig. 5, A and B). Left ventricular end dimension in both systolic (LVESD) and diastolic (LVEDD) phases was significantly lower in LCR/Tol rats (Fig. 5, C and D). Finally, LCR/Tol showed increased RWT, indicating myocardial adaption to higher blood pressure and chronic cardiac stress (Fig. 5E). However, HCR strains had comparable basal cardiac indices as LCR strains (Fig. 5F) but increased oxygen delivery ability as indicated by increased red blood cell (RBC) mean corpuscular volume and mean corpuscular hemoglobin (Supplemental Table S2). With an introduction of HCR mitochondria, LCR.HCR^mt/Tol exhibited a significant increase of LVEDD (Fig. 5D). Additionally, there was a decrease in RWT in LCR.HCR^mt/Tol compared with LCR/Tol, although it didn’t attain statistical significance (Fig. 5E). In contrast, there was no difference in cardiac indices between HCR/Tol and HCR.LCR^mt/Tol. Among hematological indices other than that of RBCs, two other cell types were prominently increased in the HCR compared with LCR. These were the neutrophils and platelets (Supplemental Table S2). The numbers of both neutrophils and platelets were not significantly affected by replacing mitochondrial genomes from controlling strains.

Fig. 5.

HCR mitochondria have beneficial effect on cardiac remodeling. A–F: fractional shortening (FS), velocity of circumferential shortening (mVCF), left ventricular end-systolic diameter (LVESD), left ventricular end-diastolic diameter (LVEDD), relative wall thickness, and cardiac indices were determined by echocardiography in 34 wk old male rats (HCR/Tol, n = 8; HCR.LCR^mt/Tol, n = 8; LCR.HCR^mt/Tol, n = 7; LCR/Tol, n = 6). Data presented as means ± SE. Levels of statistical significance were analyzed by one-way ANOVA with post hoc Tukey test; Differences in letters between bars (e.g., a, b, c) indicate statistically significant differences between strains (P < 0.05).

Distinct behavior and cognitive functions in the four holobionts.

We tested rats for locomotor activity and mood-related behaviors. In the open field, total distance as indicated by total number of line crossing did not differ between strains (Fig. 6A). Compared with the LCR/Tol, HCR/Tol showed decreased time spent in the central area and increased grooming behavior (Fig. 6, B and C). However, there were no differences between the four strains in the time spent on the open arm of elevated plus maze (Fig. 6D). Depression-like behavior was evaluated by the forced swim test (Fig. 6E). Compared with the LCR strains, HCR strains exhibited significantly less immobility time and more swimming and climbing time. All these behaviors showed little effect from mismatched mtDNA.

Fig. 6.

Altered behavior and cognition in inbred and conplastic animals. Locomotion, anxiety-like behavior, and depression-like behavior were assessed by open field test (HCR/Tol, n = 12; HCR.LCR^mt/Tol, n = 11; LCR.HCR^mt/Tol, n = 6; LCR/Tol, n = 10) (A–C), elevated plus maze test (HCR/Tol, n = 12; HCR.LCR^mt/Tol, n = 11; LCR.HCR^mt/Tol, n = 6; LCR/Tol, n = 10) (D), and forced swim test (HCR/Tol, n = 19; HCR.LCR^mt/Tol, n = 18; LCR.HCR^mt/Tol, n = 18; LCR/Tol, n = 20) (E), respectively, in 16–19 wk old male rats. F: learning ability and spatial reference memory as indicated by the number of errors (top) and escape latency to the first encounter of the escape hole (bottom) were assessed by Barnes maze test in 21 wk old male inbred and conplastic rats (HCR/Tol, n = 18; HCR.LCR^mt/Tol, n = 18; LCR.HCR^mt/Tol, n = 11; LCR/Tol, n = 16). Long-term spatial reference memory was assessed in the second probe test conducted 7 days after the last training as indicated by the number of errors and escape latency to the first encounter of the escape hole (G), time spent around each hole (H), and time spent around the target quadrant (I). Data are means ± SE. Levels of statistical significance for two groups comparison were analyzed by independent sample t test; and more than two groups comparison were analyzed by one-way ANOVA with post hoc Tukey test; differences in letters between bars (e.g., a, b, c) indicate statistically significant differences between strains (P < 0.05) (only significant statistically significant differences were shown in H).

We also assessed spatial learning and memory with the Barnes maze test. We found no difference between the four strains in learning (acquisition phase) and short-term memory (probe test on day 5) (Fig. 6F). However, compared with LCR/Tol, HCR/Tol maintained significantly strong spatial reference long-term memory (probe test on day 12). In addition, HCR/Tol exhibited decreased number of errors and escaping latency before reaching the target hole (Fig. 6G, top) and increased time spent around target hole (Fig. 6H, top), despite a comparable time spent in the target quadrant (Fig. 6I, top). Remarkably, compared with the HCR/Tol, HCR.LCR^mt/Tol showed deteriorated long-term memory (probe test on day 12) (Fig. 6F). In addition, HCR.LCR^mt/Tol exhibited increased number of errors and escaping latency before reaching the target hole (Fig. 6G, bottom) and increased time spent in the target quadrant (Fig. 6I, bottom). In contrast, there was no difference in spatial learning and memory between LCR.HCR^mt/Tol and LCR/Tol.

Correlation analyses of relative abundance of bacterial communities and multiple traits.

To test whether the resultant microbial composition is a pivotal determinant of the divide between health and disease of the host, we performed a correlation analysis between relative abundances of certain species and multiple phenotypes, those most notable within the selectively bred lines of LCR and HCR rats, including body weight, eWAT/body weight ratio (~44 wk of age), cardiac function (FS, LVESD, LVEDD, and RWT), and immobility time. (Fig. 7 and Supplemental Table S3). Body weight significantly associated with relative abundance of Coprococcus, Paraprevotellaceae, and ML615J-28. In addition, eWAT/body weight ratio positively correlated with relative abundance of Turicibacter, Clostridiaceae, and Actinobacteria. The relative abundance of Clostridiaceae and Coriobacteriaceae families was further positively correlated with FS. Cardiac systolic and diastolic function seems to be related with distinct bacteria. We found that LVESD negatively correlated with Clostridiaceae abundance, whereas LVEDD negatively correlated with Alphaproteobacteria and Christensenellaceae abundance. Interestingly, we also found depression level as indicated by immobility time of forced swim test positively associated with Actinobacteria and Coriobacteriaceae abundance.

Fig. 7.

Significant correlation between the relative abundance of gut microbiota and multiple traits. Pearson’s correlation tests were used to assess correlations between the relative abundance of gut microbiota (GM) and body weight (total n = 31), epididymal white adipose tissue (eWAT)/body weight ratio (total n = 28), cardiac function (total n = 28), and immobility time (total n = 30) in male rats from all four strains. Levels of statistical significance were analyzed by Pearson’s r correlation with false discovery rate correction.

DISCUSSION

While genetic and environmental factors are well recognized as determinants of health and disease, there is a rising recognition that microbial communities living within a holobiont are also significant as a collective factor causing or contributing to health and disease. Despite this recognition, we lack a complete understanding of what shapes the composition of these communities. Host genetic factors are presumed to contribute to the microbiotal make-up of individuals; however, these factors have remained relatively unexplored given the requirement for large population-based cohorts in which both genotyping and microbiome characterization have been performed. In the current study, rat models bred under the same conditions for horizontal acquisition of microbiota during their lifetimes over many generations were evaluated to understand the impact of heritable variations in the host genome to shape the composition of GM communities. Four different strains underwent inbreeding for homozygotic fixation of their genomic composition. Despite identical breeding conditions, host genotypic selection occurring in each strain resulted in a unique gut microbiotal community to reside within each strain. Furthermore, disease progression in these models was assessed as influenced by their host genomic plus microbiotal compositions. Significant differences were noted in a variety of disease processes between the four strains. These data provide convincing evidence to indicate that the host genomic composition has direct effects on GM composition impacting health and disease.

The progenitor models chosen for this study, the HCR and LCR rats, are selectively bred for divergence in their genetic composition acquired through phenotypic selection for their divergence in exercise capacity. Because these models are selectively bred and not inbred, the heterogenous nature of their genomic composition would only test for associations; therefore, they are not ideal for testing the hypothesis of our study. Using a panel of already inbred strains would also not be ideal because they are not permissive for allelic variation to be introduced during the vertical selection of GM composition. Our design mimics the conditions of human generations of, for example, various racial groups, living in the same geographical region, consuming the same food. If these racial groups were to inbreed, would their heritable (vertical) genetic identities dictate their GM composition or would they acquire GM composition horizontally based on their food and environment? Our study, using an inbreeding scheme, demonstrates that GM composition between the strains is not identical as would be expected by the fact that they are all exposed to the same food and environmental conditions. Instead, each strain undergoing genetic selection had its own GM composition, which is therefore linked to the single variable in the study, the animal’s genetic makeup. Therefore, we concluded that host genetic factors are important shapers of GM composition. Indeed, in line with human studies, the majority of the enriched taxa observed in our studies have showed high heritability and/or associated with genetic variations (2, 12, 13, 43), including Lachnospiraceae, Clostridiaceae, Turicibacter, Christensenellaceae, and Peptostreptococcaceae of phylum Firmicutes, Bifidobacterium and Coriobacteriaceae of phylum Actinobacteria, YS2 of phylum Cyanobacteria, and Deltaproteobacteria of phylum Proteobacteria. However, we should note that GM variability in inbred rodent models may arise between laboratories and suppliers (8) and cause complexity in model phenotypes and affect reproducibility of results. Therefore, further studies are needed to validate findings in different laboratories or to optimize reproducibility of our animal models by constructing germ-free models (8).

Besides determining the above link, our study also confirmed that after over 20 generations of inbreeding, compared with LCR/Tol, HCR/Tol rats had significant higher exercise capacity and were resistant to the development of disease risks such as obesity, visceral fat deposition, and adaptive cardiac remodeling. These results were consistent with previous findings in selectively bred LCR and HCR rats (9), albeit with the expected dip in separation of the phenotypes observed due to loss of hybrid vigor by inbreeding. We also revealed nuclear-mitochondrial genomic interaction in determining adiposity, cardiac function, and long-term memory. We therefore propose that the animal models generated through our study could serve as valuable resources to study host (nuclear and mitochondrial genomic variants)-microbial interaction in common complex diseases as a consequence of low aerobic exercise capacity.

Mitochondrial function is closely associated with exercise capacity (33). However, we did not observe any mtDNA effects on aerobic exercise capacity. Our previous study found that transferring mtDNA from spontaneously hypertensive rats (SHR) to the salt-sensitive hypertensive model (SS) nuclear genomic background increased exercise capacity, although no difference was noted in heart mitochondrial OXPHOS enzyme activity (24). Of note, mtDNA of SHR and WKY (LCR) have several same nonsynonymous substitutions in genes ND2, ND4, CYTB, COX2, and ATP6 (Table 1). Therefore, one should expect that transferring LCR mtDNA to an HCR nuclear genomic background would increase exercise capacity. Similarly, transferring HCR mtDNA to an LCR nuclear background should decrease their exercise capacity. Given the fact that the exercise capacity is extremely high in HCR/Tol and low in LCR/Tol rats, the mitochondrial effects could be smaller in magnitude than might be masked by the nuclear genomic effect. Moreover, this result is not surprising given the fact that mitochondria are under dual control by the nuclear and mitochondrial genomes. The majority of mitochondrial proteins are encoded by nuclear DNA, whereas only 13 are known to be encoded by the mtDNA. These proteins also control and help coordinate mitochondrial dynamics (fusion and fission) in response to external stimuli (1). In any case, given the data, we conclude that genomic variations in the nuclear genome as a result of the original strong selection pressure, but not the mtDNA per se, are the major contributor to aerobic exercise capacity.

Table 1.

Comparison of base substitution or insertion/deletion resulting in AA change between HCR and LCR mtDNA with other strains

AA Substitution or Insertion/Deletion
Complex	I								IV	V	III
Gene	ND2					ND4			COII	ATP6	CYTB
AA No.	18	150	265	304	318	23	356	401	165	35	214
AA Change
LCR/Tol (WKY)	Ala	Ser	Thr	Met	His	Thr	Thr	Ile	Val	Lys	Asn
SHR	Ala	Ser	Thr	Met	His	Thr	Thr	Ile	Val	Lys	Asn
HCR/Tol (F344BN)	Val	Asn	Ala	Thr	:	Ile	Ala	Val	Ile	Glu	Asp
FHH	Val	Asn	Ala	Thr	:	Ile	Ala	Val	Ile	Glu	Asp
SS	Val	Asn	Ala	Thr	:	Ile	Ala	Val	Ile	Glu	Asp
F344	Val	Asn	Ala	Thr	:	Ile	Ala	Val	Ile	Glu	Asp
LEW	Val	Ser*	Thr*	Met*	His*	Thr*	Ala	Val	Val*	Lys*	Asn*

AA, amino acid. HCR, high-capacity running; LCR, low-capacity running.

^*No AA change between LEW and WKY.

Studies suggest altered mitochondrial abnormalities contribute to the development of obesity and diabetes (22). Interestingly, transfer of mtDNA from HCR to LCR nuclear backgrounds resulted in increased weight of adipose depot through life, despite an increase of body weight only observed at the early growth period. Studies have shown that changes in mtDNA copy number and changes in electron transport chain activity in the mitochondria are directly correlated with the lipogenesis in white adipose tissue (46). In the liver of obese insulin resistant humans, upregulated mitochondrial oxidative capacity leading to increased ROS production is thought to contribute to obesity and diabetes development (23). Houstek et al. compared liver oxidative phosphorylation (OXPHOS) enzyme activity in SHR versus SHR.LEW^mt conplastic rats (15) and SHR versus SHR.F344^mt conplastic rats (16). They found that liver mitochondria from SHR.LEW^mt and SHR.F344^mt conplastic rats have significantly higher complex II enzyme activity than SHR rats. In fact, LEW, F344, and F344BN (HCR) have several same amino acid substitutional mutations in their mtDNA in mitochondrial genes ND2, ND4, CYTB, COX2, and ATP6, (Table 1), which could help explain the observed accumulation of adipose tissue in LCR.HCR^mt/Tol rats.

There are reports of heart function and cardiac remodeling studies in conplastic models (16, 38). Sethumadhavan et al. (38) analyzed T2DN.FHH^mt versus T2DN.WKY^mt and found that left ventricular internal diameter at 12 mo was significantly increased in T2DN.FHH^mt rats. They also revealed that T2DN.FHH^mt rats had a lower cardiac ATP level and synthetic capacity than T2DN.WKY^mt, which was associated with reduced electron transport chain complex I and complex IV activity. Similarly, left ventricular end-systolic diameter was found to be increased in SHR.F344^mt rats with reduced heart OXPHOS complex I and II enzyme activity (16). Consistent with previous findings, and given the higher similarity in mitochondrial genomes of FHH, F344, and F344BN, we found that transfer of HCR mtDNA to LCR nuclear background significantly increased LVEDD. For the first time, we also demonstrated a lower blood pressure effect of HCR mtDNA, although previous studies using conplastic models all failed to detect any mtDNA effect in regulating blood pressure. These findings suggest that inherited alteration in mtDNA, without variation in the nuclear genome, alters cardiovascular function. Given the fact that the primary function of the heart is to supply oxygen and nutrients to the entire body, lower left ventricular function (in adaption to lower blood pressure) resulting in decreased cardiac consumption of oxygen (reflexed by decreased ATP production) might be a protective mechanism for cardiomyocytes from ROS damage (41). However, one should note that given the breeding design for this study, it is possible/likely that there are also nuclear genomic differences between the HCR/Tol and HCR.LCR^mt/Tol strains and similarly between the LCR/Tol and LCR.HCR^mt/Tol strains.

We also observed associations between microbiota and multiple phenotypes that have strong analogs in human diseases. For example, the relative abundance of phylum Actinobacteria, enriched in LCR/Tol rats, was positively correlated to white adipose depot. In humans, Turnbaugh et al. (42) characterized the gut microbiota of 154 monozygotic or dizygotic twins and their mothers and found that the proportion of Actinobacteria was elevated in obese individuals. Similarly, a study of 84 youths also reported that the relative abundance of Actinobacteria associated with body mass index and visceral fat (10). Increasing evidence links gut microbiota and cardiovascular diseases (29). We found that the relative abundance of Clostridiaceae family positively and negatively correlated with FS and LVESD, respectively. In addition, the relative abundance of Coriobacteriaceae family was positively associated with FS, while an unclassified genus from the Christensenellaceae family was negatively correlated with LVEDD. In a study of GM of 205 overweight or obese women at 16 wk of gestation, the proportion of Clostridiaceae and Christensenellaceae was negatively correlated to DBP and SBP, respectively (11). In addition, Kelly et al. (18) report depletion of one genus from family Christensenellaceae (R-7) among participants with high lifetime risk for cardiovascular disease. Coriobacteriaceae is also significantly reduced in heart failure patients with reduced ejection fraction and cardiomyopathy (27). These translational observations suggested that GM and its metabolites might be important factors for preventing cardiovascular diseases. In our study, we also found that phylum Actinobacteria positively associated with immobility time in the forced swim test, suggesting a potential role of these bacteria in the onset of depression. In fact, a recent study compared the gut microbiome between major depressive disorder patients and controls and found Actinobacteria significantly overrepresented in the patient group (47). Given the link between adiposity, diet, and depression (17), and our finding that phylum Actinobacteria associated with both white adipose depot and depression, it would be interesting to exam GM-based mechanisms that potentially mediate the adiposity-mental health functional association in future studies.

In summary, we successfully generated and systematically characterized the genotypes and phenotypes of two novel inbred LCR/Tol, HCR/Tol and two novel conplastic LCR.HCR^mt/Tol, HCR.LCR^mt/Tol rat models. Our inbred strains retained key features of selectively bred models. Some of these features are influenced by cross talk between nuclear-mitochondrial genomes as indicated by comparisons between inbred and mismatched mitochondrial strains. We also identified a close link between host genome and the diversity of microbiome as a determinant of health and diseases. We conclude that interactions between nuclear, mitochondrial genome, and microbiome modify the risks for complex diseases emerging with low exercise capacity.

GRANTS

Research support was provided by the University of Toledo.

DISCLOSURES

No conflicts of interest, financial or otherwise, are declared by the authors.

AUTHOR CONTRIBUTIONS

L.G.K. and B.J. conceived and designed research; Y.Z., S.K., B.M., X.C., and E.E.M. performed experiments; Y.Z., S.K., X.C., and E.E.M. analyzed data; Y.Z., S.K., M.V.-K., L.G.K., and B.J. interpreted results of experiments; Y.Z. and E.E.M. prepared figures; Y.Z. drafted manuscript; S.K., B.M., X.C., E.E.M., S.L.B., M.V.-K., L.G.K., and B.J. edited and revised manuscript; Y.Z., S.K., B.M., X.C., E.E.M., S.L.B., M.V.-K., L.G.K., and B.J. approved final version of manuscript.