A combination of genetics and microbiota influences the severity of the obesity phenotype in diet-induced obesity

Obesity has emerged as a major global health problem and is associated with various diseases, such as metabolic syndrome, type 2 diabetes mellitus, and cardiovascular diseases. The inbred C57BL/6 mouse strain is often used for various experimental investigations, such as metabolic research. However, over time, genetically distinguishable C57BL/6 substrains have evolved. The manifestation of genetic alterations has resulted in behavioral and metabolic differences. In this study, a comparison of diet-induced obesity in C57BL/6JHanZtm, C57BL/6NCrl and C57BL/6 J mice revealed several metabolic and immunological differences such as blood glucose level and cytokine expression, respectively, among these C57BL/6 substrains. For example, C57BL/6NCrl mice developed the most pronounced adiposity, whereas C57BL/6 J mice showed the highest impairment in glucose tolerance. Moreover, our results indicated that the immunological phenotype depends on the intestinal microbiota, as the cell subset composition of the colon was similar in obese ex-GF B6NRjB6JHanZtm and obese B6JHanZtm mice. Phenotypic differences between C57BL/6 substrains are caused by a complex combination of genetic and microbial alterations. Therefore, in performing metabolic research, considering substrain-specific characteristics, which can influence the course of study, is important. Moreover, for unbiased comparison of data, the entire strain name should be shared with the scientific community.

The increased intake of dietary lipids combined with physical inactivity in recent decades has resulted in an enormous global health problem. Overweight and obesity are associated with various diseases, such as metabolic syndrome, type 2 diabetes mellitus, and cardiovascular and gastrointestinal diseases 1 . Overweight and obesity are defined as abnormal or excessive adipose tissue accumulation that may impair health. According to the WHO, overweight (BMI > 25 kg/m 2 ) and obesity (BMI > 30 kg/m 2 ) affect 1.9 billion people worldwide 2 . In addition to environmental factors, genetic susceptibility and altered microbial diversity were identified to increase the risk of obesity 3,4 .
Several animal models have been established to analyze these factors separately and in a standardized manner 5 . Diet-induced obesity (DIO) models are often utilized during metabolic research, as they resemble human obesity 6 . Diets rich in fat primarily induce adiposity as well as insulin resistance, impaired glucose tolerance and hyperlipidemia 5 . The inbred C57BL/6 (B6) mouse strain is often used for various experimental investigations, including for metabolic research. Over the years, several genetically distinguishable B6 substrains have evolved, including the substrains C57BL/6 J (B6J) and C57BL/6 N (B6N) 7 . The phenotypic manifestation of genetic alterations has resulted in behavioral and metabolic differences among these substrains [8][9][10] . Moreover, differences in the DIO response were detected. One of the most commonly described genetic differences between B6 substrains is a mutation within the nicotinamide nucleotide transhydrogenase (Nnt) gene 11 . This mutation has been linked to impaired glucose metabolism and insulin secretion. Several other genetic differences between the B6 substrains were identified using single nucleotide polymorphism (SNP) genotyping.

Results
DIO results in strain-dependent metabolic differences. Our analysis revealed that all mice continuously gained body weight during the study. However, high fat diet (HFD) feeding induced obesity, as HFD-fed mice showed a 1.2-to 1.4-fold higher body weight increase than low-fat diet (LFD)-fed mice (Fig. 1A). The most rapid body weight gain was detected in obese B6NCrl mice, whereas obese B6J and B6JHanZtm mice showed a less pronounced body weight gain. Body weight gain is associated with glucose intolerance. Blood glucose levels during the oral glucose tolerance test (GTT) were similar in lean mice but increased in obese mice (Fig. 1B). Moreover, the oral GTT revealed increased glucose intolerance in obese B6J mice compared to obese B6JHanZtm and B6NCrl mice.

Figure 1.
Obesity-induced differences in body weight and glucose tolerance in B6 substrains. Body weight and glucose tolerance were analyzed after 10 weeks of feeding a LFD or HFD. Two independent experiments were performed. (A) Body weight in % was measured twice per week and calculated at the initiation of LFD or HFD feeding (n = 9-11; mean ± 95%Cl). Body weight at day 70 (n = 9-11; mean ± 95%Cl, two-way ANOVA with Sidak's multiple comparisons test). (B) During the GTT, blood glucose levels were determined before and at 15, 30, 60 and 120 min after administration of a glucose solution. For analysis, concentration-time curves were created (n = 9-11; mean ± 95%Cl). The AUC was calculated from the concentration-time curves (n = 9-11; mean ± 95%Cl, two-way ANOVA with Sidak's multiple comparisons test).
DIO results in strain-dependent immune activation. Obesity is associated with low-grade chronic inflammation. In our study, immunological differences were detected in the cell subset composition of the MAT and colon among the obese mice of the B6 substrains (Fig. 4).
In the MAT, the numbers of MHCII + CD11c + and IgA + cells in B6J mice were higher than those in B6JHanZtm mice. IgA + cells were also increased in B6NCrl mice compared to those in B6JHanZtm mice (Fig. 4A). Ifnγ expression levels were higher in B6JHanZtm and B6NCrl mice, whereas Il6 levels were increased in B6J mice (Fig. 4B). Additionally, HMOX1 concentrations tended to be increased in the MAT of B6J mice (Fig. 4B). Several differences were observed in the cell subset composition of the colon among the B6 substrains (Fig. 4C). CD8 + T cells were increased in the B6NCrl substrain compared to the other B6 substrains. Furthermore, higher numbers of NK1.1 + T cells, B220 + cells, IgA + cells and CD11c + cells were detected in B6JHanZtm mice (Fig. 4C).
Strain-and diet-dependent differences are associated with intestinal microbiota composition. Microbiome analysis was performed on cecal contents from lean and obese mice of each substrain to detect differences in the microbial community. Nonmetric multidimensional scaling (NMDS) analysis revealed strain-and diet-dependent clusters of microbiota (Fig. 5A). In line with this finding, differences in the community composition based on both factors, strain and diet, were observed (permutational ANOVA; adonis (formula = TogNCTH ~ Diet + Strain, permutations = 10000), diet: p < 0.01, strain: p < 0.01). Firmicutes was the most frequent phylum, accounting for approximately 60-85% of the total bacterial sequences in all B6 substrains (Fig. 5B). Other common phyla were Bacteroidetes (10-30%), Proteobacteria (up to 5%) and Actinobacteria (<10%) in lean B6JHanZtm mice only (Fig. 5B). Additionally, a reduced abundance of Bacteroidetes and an increased frequency of Firmicutes were observed in obese B6J and B6NCrl mice. In B6JHanZtm mice, Firmicutes abundance was similar between lean and obese mice, whereas those of Bacteroidetes were increased in obese mice.

Microbiome transfer.
Microbiota transfer experiments were performed to investigate wether transplantation leads to a transfer of phenotype as well. HFD-fed ex-GF B6NRj B6JHanZtm mice showed a less pronounced body weight increase (Fig. 6A) compared to obese B6JHanZtm mice (Fig. 1A). Furthermore, HFD-fed ex-GF B6NRj B6JHanZtm mice developed glucose intolerance and hyperlipidemia comparable to those observed in obese B6JHanZtm mice (Figs. 1 and 6B,C). Serum levels of lipase, CK and GOT showed no differences between lean and obese ex-GF B6NRj B6JHanZtm mice (Fig. 6C). Analysis of the cell subsets in the colon revealed a similar composition in obese B6JHanZtm and ex-GF B6NRj B6JHanZtm mice (Fig. 6D). ; mean ± 95%Cl two-way ANOVA with Sidak's multiple comparisons test). Concentrations of leptin and FGF21 were measured in the serum using a magnetic bead-based multiplex assay (n = 5-11; mean ± 95%Cl, two-way ANOVA with Sidak's multiple comparisons test). Two independent experiments were performed.
The intestinal microbiota of lean and obese ex-GF B6NRj B6JHanZtm mice were analyzed and compared with those of B6JHanZtm mice (Fig. 7), and NMDS revealed diet-and strain-dependent clusters of microbial communities (Fig. 7A). Again, Firmicutes was the most abundant phylum in ex-GF B6NRj B6JHanZtm mice (75-95%). Similar to lean B6JHanZtm mice, a high abundance of Actinobacteria was observed only in lean ex-GF B6NRj B6JHanZtm mice. However, the frequency of Firmicutes was increased, whereas the frequency of Actinobacteria was reduced in lean ex-GF B6NRj B6JHanZtm mice compared to those in lean B6JHanZtm mice (Fig. 7B). Thus, differences in the microbiota of these mice might be mediated by host genetics.

Discussion
Obesity has emerged as a major global health problem and is associated with various diseases such as metabolic syndrome, type 2 diabetes mellitus, and cardiovascular and gastrointestinal diseases 1 . In addition to environmental factors, such as an increased intake of dietary lipids, genetic susceptibility and microbial diversity have been identified as increasing the risk of obesity 3,4,24 .
The B6 inbred mouse strain is often used for various experimental investigations, such as metabolic research. However, over the years, genetic drift has led to B6 substrains 7,25 . The phenotypic manifestation of genetic alterations has resulted in behavioral and metabolic differences 8 www.nature.com/scientificreports www.nature.com/scientificreports/ and B6J mice revealed several metabolic, genetic, microbiological and immunological differences among these B6 substrains. Because standard thresholds (such as BMI) are not available in animals, a 10-25% higher body weight than control-fed mice was defined as moderate obesity and a 40% higher body weight as severe obesity 5 . In this study, HFD-fed B6J and B6NCrl mice showed a more than 40% increase in body weight compared to lean mice and were characterized as severely obese. In contrast, B6JHanZtm mice were characterized as moderately obese.
Various differences in behavior, phenotype and genetics were described previously between B6N and B6J mice 8,21,26 . A well-known genetic difference between B6N and B6J mice is the Nnt mutation in B6J mice 11 . The Nnt gene encodes a mitochondrial enzyme that is involved in NADP + reduction to NADPH 27 . The mutation leads to reduced NNT production and has been linked to glucose intolerance and beta cell function 28,29 . Differences in glucose tolerance between B6N and B6J mice have been described previously 11 , but similarities in glucose tolerance were also observed 30,31 . In our study, the Nnt mutation was detected only in B6J mice, whereas the wild-type Nnt allele was determined in B6NCrl and B6JHanZtm mice. This spontaneous mutation within the Nnt gene arose between 1976 and 1984 32 . As B6N and B6JHanZtm mice diverged from B6J mice before 1976, they inherited the wild-type Nnt allele 7,26 . Accordingly, obese B6J mice have shown the highest impairment in glucose tolerance. Thus, our results strengthened the association between the Nnt mutation and glucose intolerance.
In addition to the Nnt mutation, further genetic differences between B6N and B6J mice have been published 7,33 . Backcrossing experiments between B6N and B6J mice revealed 4 SNPs (rs13481014, rs13480122, rs13478783 and rs4165065) associated with increased body weight 7,21 . In our genetic analysis, these four SNPs were also found to differ between B6NCrl and B6J/B6JHanZtm mice. Therefore, these SNPs might be involved in the more pronounced body weight gain of B6NCrl mice compared to that of B6J and B6JHanZtm mice. The genetic profile of B6JHanZtm mice shared similarities with that of both B6NCrl and B6J mice, but obesity progression was less severe than in B6J and B6NCrl mice. Thus, our results indicated that differences in body weight gain depend on other genetic differences. Genes associated with human obesity are FTO and IRX3 34,35 . Experimental depletion of both genes induced a reduction in body weight compared to wild-type mice 36,37 . FTO and IRX3 were detected in various organs, such as brain and adipose tissue 37,38 . Analysis of Irx3 expression revealed similar expression in the MAT of obese mice. Irx3 expression was substantially increased in the mLN of obese B6JHanZtm mice compared to that in obese mice of both other strains. Thus, Irx3 does not seem to be the cause of the differences in body weight gain, but the role of Irx3 expression in mLN needs to be further analyzed.
Various studies illustrated a strong influence of microbiota on obesity 4,39-41 . Germfree mice are largely protected against obesity 42,43 . However, the transplantation of uncultured feces from obese donors into germfree mice leads to obesity, whereas mice transplanted with feces of lean donors do not develop obesity 4,44 . Obesity in both humans and mice has been associated with reduced diversity and characteristic changes in the gut microbiota 14,15 . A common change in the microbiota during obesity is an altered Bacteroidetes/Firmicutes ratio. In our study clusters of microbial communities that were strain-and diet-dependent were detected during the NMDS analysis and confirmed by the statistical analysis. The analysis of the microbiota revealed reduced abundance of Bacteroidetes and increased frequency of Firmicutes in obese B6NCrl and B6J mice. In contrast, Bacteroidetes were additionally increased in obese B6JHanZtm mice and Firmicutes abundance was similar between lean and obese B6JHanZtm mice in this experimental setup. However, the frequency of Firmicutes increased in HFD donor mice during the GF co-housing experiment. One reason could be an altered microbiota in B6JHanZtm mice on the species level depending on the age of the mice which seemed to have an impact on the whole bacteria composition. This is in line with others who presented age-dependent microbiota changes [45][46][47] . In addition, high levels of Actinobacteria (up to 45%) were found only in lean B6JHanZtm mice. The microbial alterations among the B6 substrains possibly contributed to the differences in obesity.
During our analysis of the influence of strain-specific microbiota, HFD-fed ex-GF B6NRj B6JHanZtm mice showed a less pronounced body weight increase. However, HFD-fed ex-GF B6NRj B6JHanZtm mice developed glucose intolerance and hyperlipidemia similar to those observed in obese B6JHanZtm mice. We detected differences in the microbial composition between B6JHanZtm and ex-GF B6NRj B6JHanZtm mice. Diet-and strain-dependent clusters mean ± 95%Cl, one-way ANOVA with Tukey's multiple comparisons test) and MAT (n = 10; mean ± 95%Cl) was measured by qPCR and normalized to a reference sample set to 1. (2020) 10:6118 | https://doi.org/10.1038/s41598-020-63340-w www.nature.com/scientificreports www.nature.com/scientificreports/ of microbial communities were observed during NMDS. Analysis of the frequencies revealed similar microbial compositions in obese B6JHanZtm and obese ex-GF B6NRj B6JHanZtm . An increased frequency of Firmicutes and reduced frequency of Actinobacteria were observed in lean ex-GF B6NRj B6JHanZtm mice compared to those in lean B6JHanZtm mice. Such differences in the microbiota in these mice might be mediated by host genetics. Previous studies have reported that host genetics have a partial influence on the microbiota 13,48,49 . Thus, the severity of obesity-and obesity-related alterations in metabolism is likely defined by a specific combination of microbiota, host genetics and environment. Furthermore, host genetics and the microbiota influence the immunological phenotype. Immunological differences among mouse strains are well documented, but differences were also observed among substrains 50 . Several studies reported an increased proinflammatory response in B6N mice compared to that in B6J mice 10,20 . Macrophages from B6N mice displayed a more activated M1 phenotype and produced more NO than those from B6J mice in a tumor model 19 . Obesity is often described as a chronic state of low-grade inflammation 16,17,48 . In our study, immunological differences between the substrains were also detected during obesity. However, most differences were observed between B6JHanZtm mice and both B6NCrl and B6J mice. For example, MHCII + CD11c + and IgA + cells were reduced in the MAT of B6JHanZtm mice, whereas NK1.1 + T cells, B220 + cells, IgA + cells and CD11c + cells were increased in the colon of these mice compared to those of both other B6 substrains. Comparison of B6NCrl and B6J mice revealed a strikingly similar immunological phenotype  , Kruskal-Wallis test with Dunn's multiple comparisons test) from obese mice was performed and analyzed by flow cytometry. CD3 + cells, B220 + cells, IgA + cells and MHCII + cells were gated from the leukocyte gate of the MAT. NK1.1 + cells and CD4 + and CD8 + cells were gated from CD3 + cells. CD11c + cells and CD11b + cells were gated from MHCII + cells. Amounts are presented on a logarithmic scale. (B) Relative gene expression of cytokines and HMOX1 levels in the MAT of obese mice were measured by qPCR and normalized to a reference sample set to 1 (n = 4-8; IQR , Kruskal-Wallis test with Dunn's multiple comparisons test) or ELISA (n = 5-6; median ± IQR[25-75]), respectively. (C) Flow cytometry staining of the total cell population from the colon (n = 5-6; mean ± 95%Cl, one-way ANOVA with Tukey's multiple comparisons test) of obese mice was performed and analyzed as described above. Amounts are presented on a logarithmic scale.
Finally, our results corroborated previously published data demonstrating several genetic and phenotypic differences between B6 substrains during obesity. These differences are caused by a combination of genetic and microbial alterations. Therefore, in performing metabolic research, considering substrain-specific characteristics, which can influence the course of study, is important. Moreover, for unbiased comparisons of data, the entire strain name should be shared with the scientific community. To investigate the influence of strain-specific microbiota, four-weeks-old GF B6NRj mice were cohoused with B6JHanZtm mice over a period of 4 weeks in gnotocages 51 for microbiota transfer (GF B6NRj mice will henceforth be referred to as ex-GF B6NRj B6JHanZtm ). Subsequently, ex-GF B6NRj B6JHanZtm mice were fed a LFD or HFD for 10 weeks.

Methods
During the feeding period, body weight was determined twice per week, and an oral GTT was performed at the end of the study. For the GTT, mice were fasted for 6 h and then administered a glucose solution (2 g/kg) by oral gavage. Blood glucose levels were determined at different time points (0, 15, 30, 60 and 120 min) using a glucose meter (Contour XT, Bayer, Leverkusen, Germany). Serum analysis. After 70 days of feeding, mice were sacrificed by CO2 inhalation followed by exsanguina-  www.nature.com/scientificreports www.nature.com/scientificreports/ Figure 6. Microbiota and genetics are involved in obesity-induced parameters. GF C57BL/6NRj mice were cohoused with B6JHanZtm mice over a period of 4 weeks (GF C57BL/6NRj mice referred to as ex-GF B6NRj B6JHanZtm ). Subsequently, ex-GF B6NRj B6JHanZtm mice were fed a LFD or HFD for 10 weeks. Two independent experiments were performed. (A) Body weight in % was measured twice per week and calculated at the initiation of LFD or HFD feeding and at day 70 (n = 3-4; mean ± SD). (B) During the GTT, blood glucose levels were determined before and at 15, 30, 60 and 120 min after administration of a glucose solution. For analysis, concentration-time curves were created, and the AUC was calculated (n = 3-4; mean ± SD). C: Concentrations of cholesterol, HDL, LDL, lipase, CK and GOT were measured in the serum of LFD-and HFD-fed ex-GF B6NRj B6JHanZtm mice (n = 3-4; mean ± SD, unpaired t test). D: Surface staining of the total cell population of the colon from HFD-fed ex-GF B6NRj B6JHanZtm mice and their microbiota donor mice (obese B6JHanZtm) was performed and analyzed by flow cytometry. CD3 + cells, B220 + cells, IgA + cells and MHCII + cells were gated from the leukocyte gate of the colon. NK1.1 + cells and CD4 + and CD8 + cells were gated from CD3 + cells. CD11c + cells and CD11b + cells were gated from MHCII + cells. Amounts are presented on a logarithmic scale (n = 2-4; mean ± SD).

Scientific RepoRtS |
(2020) 10:6118 | https://doi.org/10.1038/s41598-020-63340-w www.nature.com/scientificreports www.nature.com/scientificreports/ Allele specific PCR for the Nnt gene. Mutations in the Nnt gene were assessed using a three primer, two allele PCR assay as previously described 11  Irx3 isoform analysis. RNA was isolated from mesenteric lymph nodes (mLNs) and cDNA was synthe- Quantitative real-time PCR (RT-qPCR). Before RNA isolation, mesenteric adipose tissue (MAT) was lysed at 37 °C for 5 min in RLT (RNeasy Mini Kit, Qiagen, Hilden, Germany), homogenized with a sonicator and then centrifuged for 5 min at 2500 × g. The aqueous phase was used for further analysis. The following RNA isolation steps were performed as described previously 52   Flow cytometry. Cell suspensions were prepared from MAT and colon tissue. MAT was digested at 37 °C for 20 min with 0.75 mg/mL collagenase (from Clostridium histolyticum, Type VIII, Sigma Aldrich, Steinheim, Germany) in Hanks' Salt Solution (HSS; Biochrom, Berlin, Germany). Colon tissue, devoid of attached mesentery and adipose tissues, was rinsed carefully with PBS. Thereafter, the colon tissue was cut longitudinally and incubated in buffer I (3.5% FCS, 100 mM DTT in HSS) at 37 °C for 20 min. Subsequently, the colon tissue was removed and treated twice with 5 ml buffer II (3.5% FCS, 0.5 M EDTA in HSS) at 37 °C for 15 min each time. Both suspensions of buffer II were combined and stored on ice. Colon samples were transferred to buffer III (10% FCS, 5 mg/ml DNAse, 125 U/mg Collagenase D in RPMI) and incubated for 60 min under the same conditions. After incubation, the suspension was vigorously shaken, and the remaining tissue was discarded. All cell suspensions of the colon were mixed and washed once with MACS buffer, and the cell subset composition was analyzed by flow cytometry using the following antibodies: CD3-APC-Cy7, CD8-PE-Cy7, CD11b-AF488, and MHCII-BV510 (all acquired from Biolegend, San Diego, USA); CD4-VioGreen ™ , B220-VioBlue ® , and NK1.1-PerCP-Vio700 (all acquired from Miltenyi, Bergisch Gladbach, Germany); CD11c-APC (BD Biosciences, Heidelberg, Germany); and IgA-PE (Bio-Rad Laboratories, München, Germany). Flow cytometric analysis was performed using a flow cytometer (Gallios ™ , Beckmann Coulter, Brea, USA) and Kaluza Analysis 1.3 software (Beckmann Coulter, Brea, USA).
Enzyme-linked immunosorbent assay (ELISA). For protein isolation, MAT was homogenized in 1 mL extraction buffer (Takara Bio, Kusatsu, Japan) by a tissue homogenizer (Ultra Turrax, IKA ® -Werke, Staufen, Germany) for 30 seconds. Afterwards, samples were incubated for 10 min on ice and subsequently centrifuged at 400 rpm for 10 min. A Bradford protein assay was performed to determine the protein concentrations in the obtained supernatants. A quantitative sandwich ELISA (Mouse Heme Oxygenase-1 EIA Kit, Takara Bio, Kusatsu, Japan) was used to determine the concentration of heme oxygenase 1 (HMOX1) in the obtained protein solutions from MAT. Initially, samples were diluted 1:1000, and the following procedure was performed according to the manufacturer's instructions. Samples and standards were prepared in duplicate and measured at 450 nm with a plate reader (VICTOR ™ X3, PerkinElmer, Waltham, MA, USA).
Microbiome analysis. The intestinal cecum content was removed under sterile conditions. Then, DNA was extracted and the V1-V2 region of the 16 S rRNA gene was amplified and sequenced on the Illumina MiSeq platform as previously described 53 . Merging of paired-end raw reads were implemented according to Cole et al. and resulted in 32369 ± 26255 sequences per sample 54 . These sequences were subsequently assigned a taxonomic affiliation using RDP's naive Bayesian classifier and rarefied to an equal depth (7398 sequences) 54 . Subsequent analyses were performed at the genus level. Calculations on diversity and non-metric multidimensional scaling analysis (metaMDS, auto transform = TRUE) were performed in R.
Statistical analysis. All statistical analyses were performed using GraphPad Prism ® 6 software. Data were tested for normality with the D' Agostino-Pearson (n ≥ 8) normality test. For smaller sample sizes, the Shapiro-Wilk normality test (n ≥ 7) or Kolmogorov-Smirnov test (n ≥ 5) were used.
Quantitative two group parametric data were analyzed with a t test, whereas data from at least three groups were analyzed by one-way analysis of variance (ANOVA) with Tukey's test for multiple comparisons. Nonparametric data for more than two groups were analyzed by the Kruskal-Wallis test with Dunn's multiple comparisons test. Comparison of data with two factors was analyzed by using two-way ANOVA with Sidak's multiple comparisons test. The significance level was set at 5%.

Data availability
The datasets generated and analyzed in the current study are available from the corresponding author upon reasonable request.