Multigenerational impacts of bile exposure are mediated by TGR5 signaling pathways

Besides their well-known roles in digestion and fat solubilization, bile acids (BAs) have been described as signaling molecules activating the nuclear receptor Farnesoid-X-receptor (FXRα) or the G-protein-coupled bile acid receptor-1 (GPBAR-1 or TGR5). In previous reports, we showed that BAs decrease male fertility due to abnormalities of the germ cell lineage dependent on Tgr5 signaling pathways. In the presentstudy, we tested whether BA exposure could impact germ cell DNA integrity leading to potential implications for progeny. For that purpose, adult F0 male mice were fed a diet supplemented with cholic acid (CA) or the corresponding control diet during 3.5 months prior mating. F1 progeny from CA exposed founders showed higher perinatal lethality, impaired BA homeostasis and reduced postnatal growth, as well as altered glucose metabolism in later life. The majority of these phenotypic traits were maintained up to the F2 generation. In F0 sperm cells, differential DNA methylation associated with CA exposure may contribute to the initial programming of developmental and metabolic defects observed in F1 and F2 offspring. Tgr5 knock-out mice combined with in vitro strategies defined the critical role of paternal Tgr5 dependent pathways in the multigenerational impacts of ancestral CA exposure.

The last decade has presented evidences that offspring development, behavior and physiology can be determined by parental experiences 1 . In particular, rodent models of paternal exposures to malnutrition, stress or environmental contaminants have been associated with higher susceptibility to metabolic diseases across several generations 2,3 . Deeper explorations identified altered epigenetic patterns in mature sperm as strong candidates for multigenerational programming of diseases [3][4][5][6] .
Indeed, all along their differentiation male germ cells undergo dynamic epigenetic changes particularly sensitive to environmental stressors. Germ cell genome is progressively methylated to reach high level of DNA methylation in mature sperm 7 . This process relies on the activity of the de novo DNA methyltransferase enzymes DNMT3a and DNMT3b that are highly expressed within the germline. After fertilization, most alterations of DNA methylation levels that may occur during spermatogenesis are erased to allow de novo programming of the embryo in both somatic and germ cell lineages. However, recent findings suggest that certain alterations can escape a functional erasure and be disseminated over subsequent generations contributing to epigenetically inherited traits 8 .
Beside DNA methylation, incorrect histone patterns have been shown to mediate the effects of parental exposures on offspring health 9,10 . If most histones are replaced by protamines during spermiogenesis, 1% to 4% of the mature sperm genome remains associated with nucleosomes. Maintained on the paternal genome after fertilization, the nature of histone modifications and their position on the genome will initiate the proper transcriptional program required for early development of the embryo. The histones inherited from the spermatozoa will also guide the processes of demethylation-methylation of the future zygote genome in both somatic and primordial germ cells. Independently or in interaction with DNA methylation, retained histones contained in the mature spermatozoa are potential messengers of the effects of parental exposures on offspring physiology.
It has been established that bile acids (BAs) act as signaling molecules and regulate many physiological functions, such as lipid, glucose and energy metabolisms via the modulation of their two main receptors: the G-protein-coupled bile acid receptor-1 (GPBAR-1, TGR5) and the nuclear receptor Farnesoid-X-receptor alpha (FXRα, NR1H4) [11][12][13] . In a recent report, we showed that chronic exposure to a diet supplemented with 0.5% Scientific REpoRTS | (2018) 8:16875 | DOI: 10.1038/s41598-018-34863-0 cholic acid (CA-diet) decreases sperm count and reduces male fertility. Testicular defects associated with the phenotype are dependent upon the activation of the membrane BA receptor Tgr5 within the germ cell lineage 14 .
Male germ cells contain fundamental epigenetic and genetic informations, which constitute the molecular basis of the paternal contribution to the next generations. Considering our previous finding, the question arises of whether BAs could impact germ cell integrity and contribute to the programming of phenotypic traits across multiple generations.
The present study shows evidence of altered DNA methylation levels in sperm cells from wild-type males exposed to CA-diet (F0 CA;+/+ ). These differential methylation patterns are associated with high perinatal mortality in the first generation of pups originating from F0 CA;+/+ founders (F1 CA;+/+ ). In addition, surviving F1 CA;+/+ offspring showed metabolic abnormalities, such as altered BA homeostasis and glucose intolerance at adult age. The F1 CA;+/+ males are in turn able to transmit most of these phenotypic traits to the subsequent F2 CA;+/+ generation. Interestingly, none of these developmental or metabolic abnormalities was observed in progeny born from CA exposed founders deficient for the gene encoding Tgr5 (F0 CA;−/− ). All together these data show that Tgr5 signaling pathways are critical to initiate multigenerational programming of phenotypic traits after paternal exposure to CA.

Results
Paternal exposure to cholic acid affects developmental and metabolic physiology across 2 generations of offspring. We previously demonstrate that pathological level of bile acids alter male fertility via Tgr5 in response to adult exposure to a diet supplemented with 0.5% of cholic acid (CA-diet) 14 . Remaining questions are to define if CA-exposure can impair germ cell integrity and if it could contribute to multigenerational programing of diseases in progenies.

Figure 1.
Experimental approach to analyze the multi-generational impacts of the paternal exposure to CAdiet. Adult wild type (F0 +/+ ) and Tgr5 knock-out (F0 −/− ) male founders were exposed to a control diet (CT) or the corresponding diet supplemented with 0.5%-CA (CA). After 3.5 months of exposure, males of each group were bred with unexposed C57BL6J females generating the experimental F1 offspring. These F1 males were then bred with unexposed C57BL6J females generating the experimental F2 offspring. Further analysis at PND15 showed higher hepatic BA levels in F1 CA;+/+ group compared to F1 CT; +/+ (Fig. 2c). According to previous studies 15 , altered BA homeostasis might contribute to postnatal lethality of F1 CA;+/+ . Again, no change in BA concentrations was observed in offspring originating from Tgr5 knock-out (F0 CA;−/− ) mice exposed to CA diet compared to control F0 CT;−/− animals (Fig. 2c).
To determine whether early BA increase in F1 CA;+/+ offspring was maintained and potentially associated with other metabolic defects at adulthood, we extended our analysis and performed glucose and insulin tolerance tests (GTT, ITT) ( Fig. 2d,g). No change in hepatic concentrations of BAs was detected in adult F1 CA;+/+ males compared to F1 CT;+/+ males (Fig. 2c). However, adult F1 CA;+/+ males showed difficulties to restore glucose levels during GTT compared to their respective controls (Fig. 2d,e). Consistently with an impaired glucose tolerance, the F1 CA;+/+ males showed lower fasting insulin concentrations before and along the GTT protocol (Supplemental 1d) and lower abilities to normalize glycemia when challenged with exogenous insulin (ITT) compared to F1 CT;+/+ males (Fig. 2f,g). Of note, no significant alterations of glucose metabolism due to paternal CA exposure were observed in F1 CA;−/− offspring ( Fig. 2d-g).
To gain functional insight into such altered metabolic signature, F1 CA;+/+ and F1 CT;+/+ male mice aged of 5 weeks were challenged with a high-fat diet. After 35 days of treatment, F1 CA;+/+ males showed a higher body weight gain compared to F1 CT;+/+ (Supplemental Fig. 1e,f).
All together, these results demonstrate that paternal CA exposure increases susceptibility to developmental defects and metabolic disorders.
Paternal CA-exposure of F0 males results in postnatal abnormalities up to the F2 generation.
Several evidences in the literature demonstrated that parental exposures could determine offspring physiology across several generations. After establishing perinatal lethality and impaired metabolism in F1 CA;+/+ males, we raised the question of whether F2 offspring reproduced part of these phenotypic traits in the absence of further exposure. For that purpose, F1 +/+ males were mated to unexposed C57BL/6 females generating F2 CA;+/+ or F2 CT;+/+ offspring (Fig. 1). CA exposure of F0 founders led to postnatal mortality in the corresponding F2 (F2 CA;+/+ ) generation similarly to what was observed in the F1 CA;+/+ offspring ( Fig. 3a). At PND15, the surviving F2 CA;+/+ showed lower body weight (Fig. 3b), associated with a clear but not significant increase in hepatic BA levels compared to the F2 CT;+/+ males ( Fig. 3c). At adulthood, if no difference in glucose tolerance was detected (Fig. 3d,e), F2 CA;+/+ males showed enhanced sensitivity to exogenous insulin during ITT compared to F2 CT;+/+ animals ( Fig. 3f,g). These results show that paternal CA-exposure resulted in phenotypic abnormalities up to the F2 CA;+/+ generation.
Breeding of F0 −/− males with C57BL/6J females leads to genetic heterogeneity in the following generations. The analysis of these offspring is thus sub-optimal to establish if Tgr5 signaling is involved in the initiation of the multigenerational phenotypes resulting from paternal CA exposure. To encompass this limitation, we used a diet supplemented with oleanolic acid (0.07%-OA), known as a TGR5 agonist 16 . The OA specificity was confirmed using F0 −/− males. In F0 OA;+/+ males, activation of Tgr5 pathways after exposure to OA-diet led to lower sperm count and decreased fertility compared to control males (Supplemental Fig. 2a,b). Similarly to CA-diet, paternal exposure to OA-diet induced perinatal mortality, altered body weight at PND15, and reduced glucose tolerance in adult F1 OA;+/+ males compared to F1 CT;+/+ males (Supplemental Fig. 2c-e). Note that OA exposure has no effect on offspring originating from F0 −/− male founders (Supplemental Fig. 2). Interestingly, the F2 OA;+/+ generation showed increased neonatal mortality and lower body weight at PND15 compared to F2 CT;+/+ pups (Supplemental Fig. 2f,g). Altogether, these last results provide evidences of multi-generational impacts of CA initiated in exposed F0 males by Tgr5 dependent pathways.

CA-exposure alters DNA methylation in F0 +/+ sperm cells.
To understand the molecular mechanisms underlying the initial steps of this new model of multigenerational impacts of ancestral exposure, we next analyzed the effects of CA-diet on germ cell quality with a particular focus on epigenetic integrity.
During the last step of spermatogenesis, namely spermiogenesis, germ cells undergo a global chromatin remodeling characterized by the replacement of most histones by protamines. In elongated spermatids, histone-protamine transition results in a tight compaction of the genome within the nucleus. Interestingly, electronic microscopy on F0 testes sections (Supplemental 3a) clearly showed altered DNA condensation in spermatids from F0 CA;+/+ males suggesting defects in chromatin composition following CA treatment. However, no alteration in mRNA accumulation of either transition proteins (Tnp1 & Tpn2) or protamines (Prm1 & Prm2) was observed (Supplemental 3b). In addition, no obvious change in H3 and H4 histones accumulation in mature sperm was detected in F0 CA;+/+ or F0 CA;−/− founders compared respectively to F0 CT;+/+ or F0 CT;−/− male mice (Supplemental 3c). These results suggested that histone-protamine signaling might not be critical to mediate the offspring phenotypes induced by paternal CA exposure. Besides histone patterns, sperm DNA methylation is known to play a major role in developmental programing of diseases. Interestingly, mature sperm cells from CA exposed founders (F0 CA;+/+ ) showed global DNA hypo-methylation compared to those from F0 CT;+/+ founders (Fig. 4a). In contrast, no global influence of CA exposure on DNA methylation levels in F0 CA;−/− sperm cells was observed (Fig. 4a). The interplay between TGR5 signaling pathways and sperm cell genome methylation was further supported by the increased abundance of 5-methyl-Cytosine (5meC) in F0 CT;−/− males compared to F0 CT;+/+ male founders fed a control diet (Fig. 4a).
Conclusions from global DNA methylation analysis has been confirmed and implemented by a reduced representation bisulfite sequencing (RRBS) approach realized on similar sperm DNA extracts (Fig. 4b,c). DMRs between CA-exposed and control founders were defined as regions containing CpG dinucleotides within a 1000 bp window showing a concordant ≥25% methylation change with a q-value cutoff of 0.01.
Functional analysis of RRBS reveals DMRs associated genes relevant with the multigenerational phenotypes induced by paternal CA exposure. To understand how CA-associated methylation changes might have functional significance in the future embryo, we identified all annotated genes that directly overlapped at least one CpG showing 25% of differential methylation (Supplemental Tables 1-8). Focusing on gene boundaries (intron, exon) and their corresponding promoters; differentially methylated CpGs from F0 CA;+/+ spermatozoa are contained within a total of 2165 annotated genes. In F0 CA;−/− spermatozoa, 1928 differentially methylated genes were identified (Fig. 4d). Interestingly, the majority of differentially methylated genes determined in F0 CA;+/+ sperm cells (69.74%) did not overlap with the ones observed in F0 CA;−/− sperm cells (Fig. 4d). This set of 1510 annotated genes, specific to exposure of F0 CA;+/+ males, might be of particular interest to explain the molecular basis of the multigenerational phenotypes induced by paternal exposure to CA-diet.
To gain functional insight, these 1510 associated genes, specifically observed on F0 CA;+/+ sperm cells were analyzed for Biological process over-representation using PANTHER. Among the 7 categories statistically over-represented after applying the Bonferroni correction for multiple testing, 5 were associated with developmental processes. Of note, "embryonic development" presented the highest enrichment fold (+2.45) (Supplemental Fig. 4).
Additional comparisons for common functional annotations pathways were tested using Enrichr. According to the most updated KEGG database, genes associated with differential CpG methylation showed significant overlap with critical metabolic signaling including "insulin secretion" (P-value = 0.0006) and "glycolysis/gluconeogenesis" (P-value = 0.0280) (Supplemental Fig. 4).
Besides enrichment analysis, genes associated with differential CpG methylation only in F0 CA;+/+ spermatozoa, were considered for their relevance with abnormalities described in offspring from CA-diet exposed males. This a priori read out of the RRBS dataset revealed altered methylation patterns associated with Fgfr4 and Klotho, genes encoding two critical regulators of BA homeostasis 17 .
Our next interest focused on the specific family of imprinted genes. Imprinted genes are defined by allele specific expression that relies on asymmetric 5meC patterns established on imprinted control region (ICR) during gametogenesis. This set of genes is thus of interest in the context of germ cell DNA methylation changes in response to CA exposure. Out of a list of 84 known imprinted genes, 6 were overlapping differentially methylated CpGs in sperm cells from F0 CA;+/+ males. Those genes include Peg10, Phlda2, Plagl1, or Kcnq1. Interestingly, improper dosage of Peg10 and Phlda2 are associated with defects in fetal and placental growth as well as sudden perinatal death in mouse. Of similar interest, Plagl1 or Kcnq1 dysregulation are responsible for impaired glucose tolerance associated with insulin secretion defects 18,19 .
Altogether, these functional interpretations are consistent with developmental and metabolic abnormalities described in offspring from CA exposed males. This suggests that sperm cells epigenome and more specifically defects in DNA methylation might contribute to the initiation of the phenotypic impacts of CA-diet across multiple generations.
CA associated DMRs in sperm cells are associated with transcriptional deregulation of functionally related genes in the liver of F1 and F2 offspring. We next asked whether differential methylation in F0 sperm cells might prime the embryo to persistent transcriptomic deregulations underlying phenotypic traits shared across F1 and F2 progenies.
As a representative read out of metabolic defects, total RNA from the liver of both PND15 and/or adult offspring were tested for relative abundance using quantitative PCR. We first tested for transcript deregulation of genes associated with differential CpG methylation in sperm cells of CA treated F0 male founders. As similar deregulations between F0 sperm cells and offspring somatic tissue cannot be strictly expected, we also expended our analysis to a set of genes functionally related to those submitted to DNA methylation change in the CA exposed F0 germline.
Always at PND15, Kcnq1 an imprinted gene associated with DMRs in F0 sperm cells, showed transcript deregulations in the liver of F1 CA;+/+ and F2 CA;+/+ offspring when compared to their respective controls (Fig. 5c,d). In the F1 generation, the decrease in Kcnq1 was dependent on paternal Tgr5 signaling as F1 CA;−/− male transcript levels were not affected compared to F1 CT;−/− (Fig. 5c).
In line with defects in glucose homeostasis, adult F1 CA;+/+ males originated from CA exposed F0 CA;+/+ founders showed consistent deregulation of G6Pase and Pepck mRNA accumulations compared with controls (Fig. 5e). No alteration of G6Pase and Pepck expression was observed in F1 CA;−/− males compared to F1 CT;−/− males (Fig. 5e). If insulin response was altered in F2 CA;+/+ males, those individuals did not show any difference in hepatic mRNA accumulations of G6Pase and Pepck when compared to their respective control group (Fig. 5f). These data suggest that molecular mechanisms specific to each generation underlie common metabolic traits resulting from paternal CA-diet exposure.
Tgr5 dependent regulation of Dnmt3b within the germ cell lineage is a strong candidate to prime the phenotypic impacts of paternal CA exposure across multiple generations. To understand how Tgr5 signaling pathways affect germ cell DNA methylation, we analyzed testicular abundance of transcripts encoding for DNA-methyltransferases (Dnmts) (Fig. 6a and Supplemental 6). No impact of CA-diet was observed for Dnmt1, Dnmt3a and Dnmt3L (Supplemental 6), whereas a decrease of Dnmt3b mRNA accumulation was detected in testis of F0 CA;+/+ compared to F0 CT;+/+ founders (Fig. 6a). Of note, Dnmt3b mRNA accumulation remained unaffected in testis of F0 CA;−/− males compared to their respective controls (Fig. 6b). Interestingly, F0 CT;−/− mice showed an increased Dnmt3b mRNA accumulation compared to F0 CT;+/+ founders (Fig. 6b). These results are consistent with previous quantification of global sperm DNA methylation, as well as the RRBS dataset obtained from F0 sperm. Dnmt3b is expressed in different cell types within the testis such as Leydig, Sertoli and germ cells 7,20,21 . To define if the observed deregulation of testicular Dnmt3b mRNA accumulation could be correlated with the impacts of CA exposure on DNA methylation levels within germ cell lineage, we performed immunohistochemistry for co-staining of Smad6, a marker of post-meiotic germ cells, and 5-methylcytosine (5meC). Consistently with the observed sperm cell DNA hypo-methylation, F0 CA;+/+ males showed a lower proportion of seminiferous tubules containing post-meiotic germ cells co-stained for Smad6 and 5meC compared to F0 CT;+/+ males (Fig. 6c,d). This effect was not observed in the testis of F0 CA;−/− when compared to F0 CT;−/− group (Fig. 6d).
The use of the spermatogonial cell line GC1-spg allowed demonstrating that Tgr5 signaling pathways control Dnmt3b expression within germ cell lineage. Indeed, GC1spg cells transfected with a siRNA directed against Tgr5 showed higher mRNA accumulation of Dnmt3b compared to cells transfected with a control siRNA (Fig. 6e). Moreover, treatment of GC1-spg cells with deoxycholic acid (DCA) induced a significant decrease of Dnmt3b while transcript accumulation remained stable in cells transfected with a siRNA directed against Tgr5 (Fig. 6f). Those significant transcriptomic changes have been confirmed at the protein level (Fig. 6g). Altogether these results suggest that Tgr5 dependent regulation of Dnmt3b within germ cells can impact sperm DNA methylation and prime the phenotypic impacts of paternal BA exposure across multiple generations.

Discussion
A number of experimental animal models have shown that paternal exposure to metabolic stress can have a significant influence on the susceptibility of future generations to diseases. In the present study, we demonstrate that pathological concentrations of BA are responsible for multigenerational phenotypes (Fig. 7). Two generations of progenies from males exposed to CA-diet show developmental and metabolic abnormalities. These defects are associated with differential DNA methylation in the mature sperm of CA-exposed males. The main contribution of paternal Tgr5 signaling is supported by the fact that progeny originated from F0 CA;−/− founders are protected from the adverse effects of paternal exposure to CA-diet. Further explorations suggest that Tgr5 dependent regulation of Dnmt3b expression within germ cells may constitute the primary molecular event linking paternal BA exposure to the epigenetic programming of diseases across multiple generations.
Our previous research identifies BAs as reprotoxic molecules affecting sperm number 14 . Testicular defects associated with the phenotype included a loss of cell junctions within the seminiferous epithelium as well as an increased cell death affecting advance stages of spermatogenesis. Those observations were dependent upon the activation of the membrane BA receptor Tgr5 within the germ cell lineage. Indeed, mice presenting a global deletion of Tgr5 −/− were protected against fertility damages associated with BA exposure. In this present study, we show that the deletion of Tgr5 also protect progenies to negative phenotypic outcomes induced by paternal CA exposure.
These results were unexpected, as it has been demonstrated that upon CA-supplemented diet, the absence of Tgr5 worsens liver injury creating systemic conditions that are themselves responsible for reproductive abnormalities. Interestingly, hepatic parameters known to impact male fertility such as the Insulin-growth-factor-1 (Igf1), and enzymes responsible for the catabolism of testosterone in hepatocytes, were similarly affected in F0 CA;+/+ and F0 CA;−/− males in response to CA-diet exposure 14 . This is the demonstration that systemic changes associated with CA treatment in both genetic backgrounds can't explain the reproductive defects, which remained unique to wild-type mice. We then raised the hypothesis that BA, without systemic intermediates, may act directly within the testis via Tgr5 dependent signaling pathways. In this study, the use of the spermatogonial cell line GC1-spg confirmed the activity of Tgr5 for the regulation of Dnmt3b expression within the germ cell lineage.
In conclusion and despite the limitations of systemic changes associated with non-tissue-specific gene deletion, Tgr5 knock-out mice appear to be a reliable model to define the impacts of CA-diet on testis physiology. However, genetic heterogeneity of progenies originating from the mating of F0 −/− males with wild-type females raised the limitations of our model to explore generational phenotype resulting from paternal BA exposure (in the following generations). We then used alternative strategy allowing genetic consistency while focusing on Tgr5 specificity. Feeding F0 +/+ and F0 −/− males with oleanolic acid, a specific Tgr5 agonist, confirms the critical role of Tgr5 signaling to program phenotypic outcomes up to the second generation.
Adult exposure to CA-diet is known to reduce sperm count and decrease fertility rates. Beside the quantitative impact of BAs on reproductive function, the present study demonstrate that pathological concentrations of BAs are responsible for germ cell epigenetic abnormalities and disease programming on two generations of offspring. The last few years have seen increasing interest in investigating the epigenetic dimension of male fertility defects. Recently, human studies along with experimental model in rodents, describe a strong association between reproductive disorders and sperm epimutations. A good illustration come from recent work showing that sperm collected for the purpose of Assisted Reproductive Technologies (ART) present abnormal chromatin composition as well as altered DNA methylation patterns 22 . In mouse, exposure to endocrine disruptors such as vinclozoline, a fungicide known for its repro-toxic properties, increases spermatogenic cell apoptosis rate and alters sperm DNA methylation 23 . Interestingly, ART or paternal exposures to vinclozolin have both been associated with offspring phenotypes across multiple generations 22 . In this context, the impacts of supra-physiological BA concentrations on sperm epigenome define an original model of paternal inheritance where metabolic compounds can affect offspring.
Three major classes of epigenetic information remain the likeliest candidates to carry paternal information to offspring. Those include cytosine methylation, chromatin structure, and non-coding RNA.
In the present study, electronic microscopy realized on testes sections clearly showed altered DNA condensation of spermatids in response to CA-diet exposure. Several reports link this feature to an alteration of the histone-protamine transition during spermiogenesis that would ultimately lead to abnormal retention of nucleosomes in mature sperm 23,24 . Even if our model of CA exposure is clearly associated to the observation of uncondensed spermatid nuclei, we were not able to detect substantial chromatin change in mature spermatozoa. The elimination of the most affected spermatids prior their release into the epididymis would explain this interesting discrepancy. Supporting this hypothesis, we previously demonstrate that CA exposure is responsible to spermatid cell death. Further investigations using sensitive approaches must be considered to highlight potential subtle changes in chromatin geography or composition in spermatozoa that survived germ cell selection. The major epigenetic change detected in sperm in response to CA exposure concern DNA methylation levels. A first approach allowed us to identify a 30% decrease in global 5meC on sperm from F0 CA;+/+ compared to controls. Interestingly several studies report that global hypo-methylation are correlated with poor pregnancy outcomes in IVF patients 25 . Consistently, reduced sperm DNA methylation is associated with decreased fertility in our model of BA exposure.
Changes in sperm DNA methylation are also known to affect the proper initiation of the transcriptional program of the zygote. As a consequence, paternal DMRs have the potential to interfere with the early processes of cell differentiation and ultimately with the functional determination of embryonic tissues. Sperm DMRs are thus strong candidates to initiate a cascade of molecular alterations increasing susceptibility to diseases in later life.
In our model, changes in sperm DNA methylation induced by CA exposure are associated with high rates of neonatal lethality in the corresponding F1 and F2 progenies. Macroscopic observations of dead newborns reveal a broad range of phenotypic abnormalities. Further experimentations would be useful to characterize the nature of these developmental disorders to understand their origin. Later in life, surviving offspring develop metabolic syndromes associated with increased BA concentrations as well as defect in glucose/insulin homeostasis. Altogether these data demonstrate that differential methylations in the sperm cells from CA exposed males are associated with phenotypic outcomes in two generation of progenies.
Understanding the mechanistic relationship between sperm DMRs and multigenerational phenotypes constitute a real challenge. Many examples in the field demonstrate that differential methylations in sperm might not be strictly maintained in F1 tissues due to the dramatic epigenetic programming required for cell fate determination during development. However, sperm DMRs are expected to influence the expression of gene networks functionally related to those submitted to sperm DNA methylation changes.
Results obtained in the context of CA exposure are consistent with this model. We identified 2160 genes associated with differential CpG methylations in the sperm of F0 CA;+/+ . Among them, 1510 were specific to F0 CA;+/+ males. Even if the majority of them did not show differences in expression in the liver of the corresponding F1 and F2, enrichment analysis show that sperm differentially methylated genes are functionally relevant with transcriptional changes associated with offspring phenotypes.
In this manuscript we demonstrate that paternal CA exposure can reproduce similar phenotypic traits up to the second generation through the male germline. This new example of transgenerational inheritance prompts us to understand how F1 spermatozoa could in turn carry the memory of ancestral CA exposure. In this study, we focused our interest on sperm DNA methylation. Global 5meC levels were clearly but not statistically decreased in sperm from F1 CA;+/+ compared to F1 CT;+/+ males (data not shown). Further analysis would be required to decipher if F1 sperm present similar epigenetic changes than the one observed in F0 germ cells directly exposed to CA diet. Few examples in the literature show that differential methylation can be maintained in the male germline across several generations by escaping the primordial germ cells reprogramming. Another hypothesis would be that the phenotypic traits observed in the first generation of offspring are able to reproduce similar epigenetic alterations on their own germ cells.
The use of Tgr5 knock out mice as well as spermatogonial germ cell in vitro, shown that the activation of Tgr5 signaling by BA reduces Dnmt3b expression. Previous reports have shown similar deregulations of Dnmt3b expression in response to various reprotoxic molecules. In male rat, neonatal exposure to xeno-estrogens leads to decrease Dnmt3b levels in the germ cells leading ultimately to reduce DNA methylation 23 . In our model, Tgr5 dependent regulation of Dnmt3b may constitute the primary molecular event linking paternal CA exposure, sperm hypo-methylation and developmental programming of diseases across multiple generations.
Our study provides fundamental findings that may contribute to a better understanding of how paternal exposure to stressors can shape offspring physiology through several generations.
Altered BA homeostasis is the direct consequence of prevalent metabolic diseases affecting our society worldwide. Along with liver failure, intestinal inflammation and obesity are all known to impact BA levels and biochemical composition.
Beside their repro-toxic property, this study identified BAs as critical mediators of transgenerational metabolic diseases impregnation. Our conclusions will highlight important opportunities for prevention and therapeutic intervention in order to slow the growing epidemic of decreased sperm quality and its consequences on metabolic disease programing across generations.

Methods
Animals. Nine-week-old Tgr5 +/+ (F0 +/+ ) and Tgr5 −/− (F0 −/− ) mice used have been described elsewhere 14 . The mice used in this study were maintained as a C57BL/6J background and housed in temperature-controlled rooms with 12 h light/dark cycles. Mice had ad libitum access to food and water. Nine-week-old mice F0 founders were exposed to the D04 diet (Control) or the D04 diet, supplemented with 0.5% cholic acid (CA-diet) (SAFE, Augy, France) for 4 months. The experimental procedure to study the multi-generational transmission from F0 to F2 is reported in Fig. 1.
Electron microscopy. Electron microscopy was performed as previously described 26 . Briefly, samples were set in 2% glutaraldehyde and 0.5% paraformaldehyde in cacodylate buffer at 4 °C for 24 h. Set testes were subsequently postfixed for 1.5 h in buffered osmium tetraoxide at 4 °C and embedded in Epon Araldite (Delta