High-fat diet-induced and genetically inherited obesity differentially alters DNA methylation profile in the germline of adult male rats

Background Paternal obesity has been associated with reduced live birth rates. It could lead to inheritance of metabolic disturbances to the offspring through epigenetic mechanisms. However, obesity is a multifactorial disorder with genetic or environmental causes. Earlier we had demonstrated differential effects of high-fat diet-induced obesity (DIO) and genetically inherited obesity (GIO) on metabolic, hormonal profile, male fertility, and spermatogenesis using two rat models. The present study aimed to understand the effect of DIO and GIO on DNA methylation in male germline, and its subsequent effects on the resorbed (post-implantation embryo loss) and normal embryos. First, we assessed the DNA methylation enzymatic machinery in the testis by Real-Time PCR, followed global DNA methylation levels in spermatozoa and testicular cells by ELISA and flow cytometry, respectively. Further, we performed Methylation Sequencing in spermatozoa for both the groups. Sequencing data in spermatozoa from both the groups were validated using Pyrosequencing. Expression of the differentially methylated genes was assessed in the resorbed and normal embryos sired by the DIO group using Real-Time PCR for functional validation. Results We noted a significant decrease in Dnmt transcript and global DNA methylation levels in the DIO group and an increase in the GIO group. Sequencing analysis showed 16,966 and 9113 differentially methylated regions in the spermatozoa of the DIO and GIO groups, respectively. Upon pathway analysis, we observed genes enriched in pathways involved in embryo growth and development namely Wnt, Hedgehog, TGF-beta, and Notch in spermatozoa for both the groups, the methylation status of which partially correlated with the gene expression pattern in resorbed and normal embryos sired by the DIO group. Conclusion Our study reports the mechanism by which diet-induced and genetically inherited obesity causes differential effects on the DNA methylation in the male germline that could be due to a difference in the white adipose tissue accumulation. These differences could either lead to embryo loss or transmit obesity-related traits to the offspring in adult life.

obese fathers are at higher risk of developing metabolic syndrome in adult life [2,3]. This evidence supports the notion that paternal obesity has a heritability component responsible for the transmission of obesity-related traits to the offspring, which is carried by the spermatozoa of obese fathers.
Epigenetic changes are heritable in nature and are brought about by DNA methylation, histone modifications and non-coding RNAs [4]. Several studies have shown that diet plays a critical role in modulating the epigenome [5]. Male rodents fed with a diet rich in fats show altered methylome profiles in the spermatozoa [6]. In addition, genome-wide studies in the spermatozoa of obese fathers show aberrant methylation and non-coding RNA expression profiles [7]. However, obesity is multifactorial in nature and could have the involvement of the environmental and/or genetic component [8]. The effects of these components individually on the epigenome of the male germline have not been explored earlier.
Animal-based obesity models are predominantly of two types (1) genetic models which are based on mutations or manipulations of one or more genes and (2) genetically healthy animals which are exposed to obesogenic environments, mainly high calorie-rich diets [9]. Our earlier studies have shown the differential effects of highfat diet-induced (DIO) and genetically inherited obesity (GIO) on metabolic, biochemical and hormonal profiles, fertility parameters and spermatogenesis [10]. The study involved the use of WNIN/Ob mutant rats (Wistar origin) as the genetically inherited obese model and Wistar rat strain as the high-fat diet-induced obese model [10]. In WNIN/Ob rats, the mutation is present on chromosome 5 upstream of the leptin receptor. However, the gene harboring the mutation is yet to be characterized [11]. These mutant rats follow an autosomal incomplete dominant pattern as the mode of inheritance [12,13]. The metabolic indices like insulin, cholesterol, triglycerides and leptin levels are high in this model with normal glucose levels [12]. Earlier, we reported differences in the metabolic parameters (body weights, adiposity index), biochemical parameters (serum glucose, cholesterol, triglyceride levels), hormonal profiles (serum testosterone, estradiol, LH, FSH, prolactin, leptin, insulin levels), fertility parameters (implantation losses, litter size, potency) and spermatogenesis (mitosis, meiosis, spermiogenesis) between the two types of obesity could be attributed to the difference in the white adipose tissue accumulation despite both the groups having similar body weights [10,14]. This also explained the discrepancies in the literature on the effects of male obesity on sperm parameters, endocrine profile and fertility. Our fertility data reported that diet-induced obese group showed sub-fertility with increased embryo loss and reduced litter size whereas the genetically inherited obese group showed complete infertility due to reduced testosterone levels [10]. Quantitation of testicular cells based on ploidy and cell type-specific expression markers, to study the effect of obesity on spermatogenesis, demonstrated that both GIO and DIO altered mitosis that was evident from the increase in the spermatogonia and S-phase population. However, differential effects were noted on meiosis and spermiogenesis in the two groups [10]. Other significant and differential alterations noted in the testis of DIO and GIO groups include expression of hormone receptors, cytokines, markers of oxidative stress as well as cell cycle mediators [14]. We hypothesized that in addition to the differences in metabolic, hormonal, fertility and spermatogenic parameters observed earlier, there could be differences in the sperm epigenome in the DIO and GIO rats that could have led to differential effects postfertilization. Thus, our present study aims to understand the effects of high-fat diet-induced and genetically inherited paternal obesity on DNA methylation in the male germline.

Diet-induced and genetically inherited obesity differentially affect DNA methylating enzymatic machinery, global 5-mC density in different cell populations in the testis and global CpG methylation in spermatozoa
The metabolic and phenotypic profiling of the animals were performed in our earlier study [10]. The schematic diagram of study design is depicted in Fig. 1.
To study the effect of high-fat diet-induced and genetically inherited obesity on DNA methylation in the male germline, we first assessed the DNA methyltransferases expression pattern in the testis. We observed a significant decrease in the Dnmt1 transcript levels in the testis of the HFD group compared to the CD group (Fig. 2a). However, Dnmt3a, Dnmt3b, and Dnmt3l were found to be unaffected (Fig. 2a). On the contrary, we observed a significant increase in the Dnmt1, Dnmt3a, and Dnmt3b expression levels in the testis of the WNIN/Ob group compared to the CROb and LEAN groups, respectively (Fig. 2b). Dnmt3l expression, however, was down-regulated in the testis of the WNIN/Ob group when compared to the CROb group and was up-regulated in the CROb group as compared to the LEAN group (Fig. 2b). Dnmt3l transcript levels were unaffected in the testis of the WNIN/Ob group compared to the LEAN group (Fig. 2b).
To study whether Dnmt transcript levels correlate with the global DNA methylation levels in the testis, we further analyzed the global 5-methylcytosine (5-mC) levels in the different cell types of testis for both the groups.
Knowing the fact that the different testicular cell types could be separated based on ploidy, we performed combined DNA methylation and the cell cycle analysis by flow cytometry. We observed a significant decrease in the 5-mC density in different testicular cell populations, mainly haploid population [elongated, elongating and round spermatids (RS)], 2 N population [spermatogonia, Sertoli cells, Leydig cells, peritubular myoid cells and secondary spermatocytes], S phase population [proliferating cells] and 4 N population [primary spermatocytes] in the testis of the HFD group compared to the CD group that correlated with the decrease in Dnmt1 transcript levels in the testis observed in the HFD group as compared to the CD group (Fig. 2c). Interestingly, we did not observe any changes in the 5-mC density in the different testicular cell populations of the WNIN/Ob group compared to the CROb and LEAN groups, respectively (Fig. 2d). To substantiate that the obesity-induced epigenetic changes observed in the testis of DIO and GIO are not due to any anatomical defects, histological examination of testis was done. We did not observe any specific morphological changes in the testis of the two groups (Additional file 5: S1). Further as mentioned above, we had observed changes in the number of different testicular cell populations in the two groups [10]. Besides, data on methylation have been normalized to DNA content whereas data on gene expression to housekeeping genes to account for these differences.
To ascertain whether the DNA methylation changes in the spermatogenic cells translated in the mature spermatozoa of the two groups, we assessed global CpG methylation using ELISA. We noted a significant reduction in global CpG methylation in the spermatozoa of the HFD group compared to the CD group (Fig. 2e). In contrast, we observed a significant increase in the global CpG methylation in the spermatozoa of the WNIN/Ob group compared to the CROb group (Fig. 2f ). However, upon calorie restriction, we found a significant reduction in the global CpG methylation in the spermatozoa of the CROb group compared to the LEAN group (Fig. 2f ). No effect was observed on the global CpG methylation in the spermatozoa of the WNIN/Ob group compared to the LEAN group (Fig. 2f ).

DNA methylome analysis of spermatozoa of diet-induced and genetically inherited obese groups
To ascertain the effect of high-fat diet-induced and genetically inherited obesity on sperm epigenome, we performed targeted methylome sequencing in the spermatozoa of both the groups (DIO and GIO). Single basepair resolution profile of 5-methylcytosine was generated Fig. 2 Effect of diet-induced (DIO) and genetically inherited obesity (GIO) on the transcript levels of DNA methylating enzymatic machinery, cell type-specific 5-methylcytosine (5-mC) density in the testis and global CpG methylation levels in spermatozoa. a, b Represent the transcript levels of DNA methyltransferases in the testis, c, d represent the 5-mC density in different testicular cell populations and e, f represent the percent CpG methylation levels in the spermatozoa of HFD group compared to CD group and WNIN/Ob group compared to CROb and LEAN groups, respectively. Data are expressed as mean ± S.E.M, N = 6 per group. Asterisks indicate significant differences compared with the CD group (*p < 0.05, **p < 0.01, ***p < 0.001); One-way ANOVA, Bonferroni multiple comparisons test: a P < 0.05, b P < 0.01, c P < 0.001 compared with the LEAN group; e P < 0.05, f P < 0.01, g P < 0.001 compared with the CROb group Deshpande et al. Clin Epigenet (2020) (Fig. 3i-l). The significantly hypermethylated CpGs are shown in red circles while hypomethylated CpGs are shown in black circles ( Fig. 3i-l). Green circles represent the CpGs that were unchanged (Fig. 3i-l). It was interesting to note that there were similarities between the volcano plots of CD versus HFD, LEAN versus CROb, CROb versus WNIN/ Ob and LEAN versus WNIN/Ob (Fig. 3i-l). Comparative analysis of CD versus HFD and CROb versus WNIN/Ob groups showed 3499 genes to be common between both the groups (Additional 5: S8). Interestingly, 10,543 genes were found to be unique to the HFD group compared to the CD group and 5927 genes were exclusive to WNIN/ Ob group compared to the CROb group (Additional 5: S8). Comparative analysis of CROb versus WNIN/ Ob, LEAN versus WNIN/Ob, and LEAN versus CROb groups showed 5927, 5830 and 10,218 genes unique to each group, respectively (Additional 5: S8). About 2540 genes were common to the three groups (Additional 5: S8). The majority of the differentially methylated CpG sites were mapped in the intergenic and intronic regions for all the four groups ( Fig. 3m-p).

Diet-induced and genetically inherited obesity differentially reprogram DNA methylation profile of molecular players of developmental pathways in spermatozoa
To ascertain the effect of high-fat diet-induced and genetically inherited obesity on developmental pathways in spermatozoa, few differentially methylated genes from Wnt, Hedgehog, TGF-beta and Notch signaling pathways that were exclusive to CD versus HFD and CROb versus WNIN/Ob groups were selected to validate the methylation changes by pyrosequencing (Tables 1, 2).
We first generated a hierarchically clustered heatmap based on the differences in the methylation value of CpGs for the genes associated with each developmental pathway for CD versus HFD (Figs. 4a, 5a, 6a, 7a) and CROb versus WNIN/Ob groups (Figs. 4b, 5b, 6b, 7b). The heatmap in every pathway in every group revealed differential changes in the methylation pattern (hypermethylation and hypomethylation) of developmental genes in spermatozoa (Figs. 4a, 5a, 6a, 7a) (Figs. 4b, 5b, 6b, 7b). 13 differentially methylated genes were randomly selected from Wnt, Hedgehog, TGF-beta and Notch signaling pathways from both the CD versus HFD and CROb versus WNIN/Ob datasets for pyrosequencing-based validation (Tables 1, 2, Additional file 1: S1.8-S1.11, Additional 2: S2.7-S2.10). Among these, 6 genes (Skp1, Wnt9b, Mapk10, Spop, Ppp2cb, Kat2b) were found to be differentially methylated in the spermatozoa of the HFD group compared to the CD group (Figs. 4c, 5c, 6c, 7c) and 7 genes (Sox17, Wnt1, Spopl, Gli1, Dcn, Bmpr1b, Rbpj) showed differential methylation pattern in the spermatozoa of the WNIN/Ob group compared to the CROb group (Figs. 4d, 5d, 6d, 7d). The differentially methylated CpG sites of all the selected genes, acquired by high throughput sequencing, represented in the form of heatmap, correlated with the pyrosequencing data in the spermatozoa of the HFD group compared to the CD group and of the WNIN/Ob group compared to the CROb and LEAN group, respectively (Figs. 4a, 5a, 6a, 7a) (Figs. 4b, 5b, 6b, 7b). The specific genes enriched in the CD versus HFD group for each pathway were also validated in the CROb versus WNIN/Ob group and vice versa. We did not observe any significant changes in the spermatozoa suggesting these genes were exclusive to their respective groups and thus additionally validated the NGS data (Data not shown).

Methylation defects in spermatozoa are partially translated in the resorbed and normal embryos sired by high-fat diet-induced obese male rats
Our earlier studies had demonstrated that high-fat dietinduced obese male rats showed increased pre-and postimplantation loss and a significant reduction in the litter size, while the genetically inherited obese male rats were infertile [10]. In the present study, we observed methylation defects in the genes of developmental importance in the spermatozoa of high-fat diet-induced obese male rats. Therefore, we examined the gene expression patterns of those differentially methylated genes in the resorbed and normal embryos sired by both the HFD and CD groups, respectively. In the Wnt signaling pathway, the expressions of Skp1 and Ppp3cc were upregulated, whereas the expressions of Mapk10, Rhoa, Btrc, Lrp6, Lrp5, Wnt9b, Tcf7l2 and Tbl1xr1 were downregulated in the resorbed embryos sired by the HFD group compared to the CD group (Fig. 8a). On the contrary, Skp1, Mapk10, Rhoa, Plcb2, Ppp3cc, Tcf7l2 and Tbl1xr1 transcript levels were unaffected whereas Btrc, Lrp6, Lrp5 and Wnt9b were upregulated in the normal embryos sired by the HFD group compared to the CD group (Fig. 8b). In the Hedgehog signaling pathway, Spop transcript levels were upregulated, while that of Kif7 and Gli3 transcript levels were downregulated in the resorbed embryos of the HFD group compared to the CD group (Fig. 8c). However, in the normal embryos, Spop, Kif7 and Gli3 transcript levels were unaltered in the HFD group compared to the CD group (Fig. 8d). In the TGF-beta signaling pathway, we observed significant upregulation of Smurf2 and Ppp2cb transcript levels and downregulation of Smad5 levels in the resorbed embryos of the HFD group compared to the CD group (Fig. 8e). Conversely, Ppp2cb and Smad5 transcript levels were unchanged, whereas Smurf2 was upregulated in the normal embryos of the HFD group compared to the CD group (Fig. 8f ). Similarly, in the Notch signaling pathway, we noted a downregulation of Aph1a, Dll3, and Dtx4 expression levels and the upregulation of Kat2b in the resorbed embryos of the HFD group compared to the CD group (Fig. 8g). On the other hand, Aph1a and Dll3 transcript levels were unaltered, whereas Dtx4 and Kat2b were upregulated in the normal embryos sired by the HFD group compared to the CD group (Fig. 8h). Interestingly, the expression patterns of all the genes in the different pathways in the resorbed embryos correlated with the methylation pattern in the spermatozoa of the HFD group compared to the CD group, which was not the case with the normal embryos sired by the HFD group compared to the CD group (Figs. 4,5,6,7,8).
Interestingly, we also noted the expression of a few genes in the resorbed embryos that did not correlate with the methylation pattern in the spermatozoa. For example, in the Wnt pathway, we observed hypermethylation of Ppard, Mapk8, Prkcb, and Fosl1 in the spermatozoa of the HFD group compared to the CD group (Additional 5: S13). However, expression of these genes was significantly upregulated in the resorbed embryos of the HFD group compared to the CD group, whereas, Fosl1 showed a reduced expression pattern but the change was not significant (Additional 5: S13). Similarly, genes in other pathways also showed a similar trend (Additional 5: S13). Furthermore, gene expression in the normal embryos sired by the HFD group also did not correlate with the methylation changes in the spermatozoa (Additional 5: S13).

Discussion
Here, we provide evidence that high-fat diet-induced and genetically inherited obesity differentially alters the DNA methylome in the spermatozoa and affects the pathways involved in embryo development. DNA methylation is a key epigenetic contributor responsible for gene silencing with the help of the enzymes known as DNA methyltransferases (Dnmts) [15]. Therefore, we first assessed the Dnmt transcript levels in the testis of both the groups. We observed down-regulation of the Dnmt1 transcript levels of the DIO group, up-regulation of Dnmt1, Dnmt3a, and Dnmt3b and down-regulation of Dnmt3l levels in the GIO group in the testis. Dnmt1 is a maintenance methyltransferase that is associated with replication foci and maintains DNA methylation after replication, and localized in all the male germ cells [16,17]. On the other hand, Dnmt3a and Dnmt3b are involved in catalyzing the active de novo methylation process on the DNA [16]. Dnmt3l lacks catalytic activity but is involved in the regulation of DNA methylation through its association with Dnmt3a and Dnmt3b [18]. Several studies have shown the importance of de novo methyltransferases in regulation of spermatogenesis. De novo methylation in the germ cells could be involved in genomic imprinting, germ cell-specific methylation or post-DNA repair methylation [19]. In the rat testes, Dnmt3a and Dnmt3b are both localized in the spermatogonial cells. Dnmt3a is also expressed in spermatocytes whereas Dnmt3l is localized in the nuclei of round spermatids, in the head region of the elongating spermatids, and in the peritubular cells of the adult rodent testis [20,21]. Dnmt levels One-way ANOVA, Bonferroni multiple comparisons test: a P < 0.05, b P < 0.01, c P < 0.001 compared with the LEAN group; e P < 0.05, f P < 0.01, g P < 0.001 compared with the CROb group are mainly correlated with the global 5-mC levels. In the present study, we noted a significant reduction in global 5-mC density in different testicular cell populations that correlated with the Dnmt1 expression levels, which were downregulated in the DIO group. Interestingly, Fullston et al. 2013 also demonstrated that paternal high-fat dietinduced obesity decreases global 5-mC levels in the testes that corroborates with our findings [2]. However, in the GIO group, no change was observed in the 5-mC density in the testis, despite the Dnmts being upregulated. Histological examination of the testis in the two groups also confirmed that the epigenetic changes are not due to the differences in testicular morphology but rather due to the high-fat diet-induced and genetically inherited obesity. Our histological findings corroborate with the findings of previous published studies in which no morphological defects were observed in the testis of DIO and GIO groups [13,22]. Our previous studies have shown a significant increase in the spermatogonial cell number in both the DIO and GIO groups and a significant decrease in the primary spermatocyte number in the GIO group [10]. In the present study, we did not observe any changes One-way ANOVA, Bonferroni multiple comparisons test: a P < 0.05, b P < 0.01, c P < 0.001 compared with the LEAN group; e P < 0.05, f P < 0.01, g P < 0.001 compared with the CROb group in the Dnmt3a and Dnmt3b expression levels in the testis of the DIO group whereas in GIO group we noted an increase in the same. This suggests that the changes in the levels are not due to the alterations in cellular composition or number in the testis but rather due to the difference in the physiological changes in the two obese rat models. Moreover, we also assessed the CpG methylation levels in the spermatozoa of both the groups. In the DIO group, the decrease in CpG methylation levels in spermatozoa correlated with the decrease in Dnmt1 levels in the testis whereas an increase in CpG methylation corroborated with the increase in the Dnmt levels in the testis in the GIO group. Changes in Dnmts and/or global 5-mC levels due to two different types of obesities in the male germline could be attributed to a variety of factors like high oxidative stress, pro-inflammatory cytokines and methyl donor imbalance (for example, S-Adenosyl Methionine (SAM)) as well as endocrine dysfunction. Changes in these factors have been shown to be associated with Fig. 7 Effect of high-fat diet-induced and genetically inherited obesity on Notch signaling pathway in spermatozoa. a, b Hierarchically clustered heatmap of the differentially methylated CpGs of genes associated with Notch pathway. Black rectangle on the heatmap represents the genes validated by pyrosequencing. c, d Pyrosequencing validation of differentially methylated CpGs of genes involved in Notch signaling in spermatozoa of DIO and GIO groups. Data are expressed as mean ± S.E.M, N = 3 samples pooled per group for methylation sequencing and N = 5 per group for pyrosequencing validation. Asterisks indicate significant differences compared with the CD group (*p < 0.05, **p < 0.01, ***p < 0.001). One-way ANOVA, Bonferroni multiple comparisons test: a P < 0.05, b P < 0.01, c P < 0.001 compared with the LEAN group; e P < 0.05, f P < 0.01, g P < 0.001 compared with the CROb group Deshpande et al. Clin Epigenet (2020) 12:179 DNA methylation defects [23][24][25][26]. Another plausible reason for the differential changes in the Dnmt and global DNA methylation levels could be due to the difference in white adipose tissue accumulation in the two groups. As in the GIO group, we have observed that upon calorie restriction, Dnmt levels in the testis were restored in the CROb group and were comparable to that of the LEAN controls. Also, another mechanism could be due to the loss of equilibrium between DNA methylation and demethylation. This balance between the two pathways is critical for the normal functioning of the germ cells [27]. We have also noted distinct changes in ten-eleven translocation (Tets) expression and 5-hydroxymethylcytosine (5-hmC) levels in the testis of both the groups (unpublished observations) suggesting an imbalance in the enzymatic machinery of 5-mC and 5-hmC. Interestingly, upon calorie restriction, we noted a decrease in CpG methylation levels in the spermatozoa of the CROb group compared to the LEAN group. This could probably be a combinatorial effect of genetic factors and calorie restriction. Since we observed differential changes in the global DNA methylation levels in the spermatozoa of DIO and GIO groups, we next performed high-throughput methylation sequencing on the spermatozoa of the two groups to acquire a genome-wide picture. Sequencing data analysis revealed that spermatozoa harbor both CpG and non-CpG methylation. However, the percentage of methylated cytosines in each sequence context, namely, CpG, CHG and CHH in the control sperm is ~ 77%, ~ 0.5-1.4% and ~ 0.4-1.3%, respectively. Our data corroborated with the findings of Tomizawa et al. 2011 [28]. Hence, further genome-wide methylation data analysis was performed to specifically assess CpG methylation. We identified a total of 16,966, 9113, 9249 and 16,471 differentially methylated CpG sites in the spermatozoa of the HFD group compared to the CD group, WNIN/Ob group compared to the CROb group, WNIN/Ob group compared to the LEAN group and CROb group compared to the LEAN group respectively. Interestingly, the global methylation levels of the total CpGs analyzed from the sequencing data did not correlate with the ELISA-based CpG methylation levels in the spermatozoa for each group. This could probably be due to difference in the techniques and usage of pooled genomic DNA samples for sequencing. Further, comparative analysis between the DIO and GIO groups showed the presence of both unique and overlapping differentially methylated regions/genes in the spermatozoa. Our previous studies have shown significant changes in the white adipose tissue accumulation between the two groups despite their body weights being similar. This could have led to the differences in the biochemical, endocrine and fertility profile as well as spermatogenesis observed between the DIO and GIO groups [10,14]. From the literature, we understand that any changes in metabolic and hormonal profile could alter the epigenetic mechanisms [2,6]. Obesity is also known to be associated with increased oxidative stress [29]. Thus, we speculate that the methylome differences in spermatozoa could be a combination of these various altered parameters. Also, presence of unique regions/genes in spermatozoa is an indication of the exclusive effects of high-fat diet-induced obesity and genetically inherited obesity. In addition, the majority of the CpG sites were found to be differentially methylated in the intergenic (~ 45%) and intronic (~ 34%) regions in the spermatozoa. Both intergenic and intronic regions are associated with enhancers [30][31][32]. Interestingly, several studies have shown that methylation at intergenic and intronic regions where the enhancers reside regulate gene expression [32][33][34][35]. Aberrant methylation in these regions has been shown to be associated with aberrant gene expression in breast cancer [30]. Genome-wide methylome study by de Castro Barbosa et al. 2015 reported 109 differentially methylated genes in the spermatozoa of rats fed with high-fat diet [6]. Comparison with our DIO dataset showed 44 differentially methylated genes to be common between the two studies (Additional file 1: S1.15). Further, pathway analysis of the differentially methylated genes of all the four groups showed the involvement of these genes in numerous pathways, namely, metabolic, pathways in cancer, PI3K-Akt, MAPK, thermogenesis, JAK-STAT, autophagy, insulin, etc. Dysregulation of these fundamental pathways has been shown to be associated with obesity [2,3,[36][37][38][39][40][41][42][43]. It is interesting to note that the genes/pathways enriched in the spermatozoa of both the DIO and GIO groups in the present study has similarity to phenotypic or physiological changes, namely, metabolic, adiposity, endocrine, spermatogenesis, oxidative stress, inflammation, apoptosis and cell cycle observed in the two groups in our previous studies [10,14]. These observations establish a causal link between the sperm methylome and phenotypic or physiological changes induced due to DIO and GIO. Since some of these pathways have been explored before with respect to obesity, we identified a few other pathways of developmental importance, namely Wnt, Hedgehog, TGF-beta, and Notch that have never been investigated in the spermatozoa and linked to the embryo loss that had been earlier observed in our obese models [10].
Wnt signaling plays a pivotal role in both embryonic development and adult physiology and is involved in body axis patterning, cell fate specification, cell migration and proliferation [44,45]. Another developmental pathway, Hedgehog is required for the development of intercellular communication, organogenesis, regeneration and homeostasis [6,47]. Transforming growth factor-beta (TFG-beta) signaling regulates various cellular functions namely cell proliferation, differentiation, migration, adhesion and apoptosis [48,49]. Notch proteins are single-pass receptors that are activated by delta and Jagged/Serrate families of membrane-bound ligands [50]. The notch pathway regulates various cellular processes like proliferation, differentiation, cell fate and apoptosis [51]. Several knockout studies have shown that loss of upstream and downstream molecular players of these developmental pathways are associated with embryo lethality, suggesting their critical role during embryo development [52][53][54][55][56][57][58]. It was interesting to note that almost all the differentially methylated genes in these developmental pathways were unique to the DIO and GIO groups as evident from the methylation pattern depicted in the heatmap. Validation of the methylation status of all the selected genes associated with these pathways in the spermatozoa by pyrosequencing confirmed the high-throughput sequencing results for DIO and GIO groups.
It is well evident in the literature that the sperm epigenome is involved in embryogenesis and offspring health [59]. Earlier studies in our laboratory have shown that paternal obesity leads to embryo loss and caused significant reduction in litter size [10]. Therefore, we analyzed the methylation status of developmental genes in spermatozoa and their expression in resorbed and normal embryos sired by the DIO model. The functional validation of gene expression of differentially methylated genes from these pathways was carried out in resorbed and normal embryos sired by the DIO group only. Expression of the genes associated with developmental pathways in the resorbed embryos but not the normal embryos correlated with the methylation pattern in spermatozoa of the DIO group. Overall, these results indicate that DIO and GIO differentially altered the developmental pathways in spermatozoa and methylation defects in these pathways in the spermatozoa could have led to aberrant expression of these genes in the embryos leading to embryo loss in the DIO group. The functional validation of the pathways in the embryos of the GIO group was not possible, as the WNIN/Ob group was infertile and did not produce any litters [10].
On the other hand, in normal embryos, we did not observe any correlation of the gene expression to that of sperm methylome. This could be due to the presence of 'epigenetically fit' spermatozoa in the sperm pool. Our previous studies on tamoxifen treatment in adult male rats had reported that not all spermatozoa in the treated group had abnormal methylation at imprinted gene loci Igf2-H19 ICR [60]. This suggests that the normal embryos could be a result of fertilization of the oocyte with 'epigenetically fit' spermatozoa, whereas resorbed embryos could be a result of fertilization with 'epigenetically unfit' spermatozoa.
In resorbed embryos, expression of few of the differentially methylated genes did not correlate to their methylation pattern in the spermatozoa of the DIO group. This indicates that differential gametic methylation is either partially translated or not persistent in the embryos. This could be due to either post-fertilization genome-wide reprogramming or changes in other regulatory mechanisms like histone modifications (unpublished observations) and non-coding RNAs [61,62]. These epigenetic changes may modify embryonic development during genome-wide reprogramming that occurs after fertilization or could persist to alter the gene expression and accumulate in the key tissues to show its effect at adulthood. In the Methyl Seq data, we also observed genes of several DNA modifying enzymes (Dnmts and Tets) and histone modifying enzymes [histone acetyltransferases (HATs), histone deacetylases (HDACs), histone demethylases (HDMs) and histone methyltransferases (HMTs)] to be differentially methylated in spermatozoa for both the groups. In addition, another reason for the differences in the gene expression patterns could be sex specific [63], as the sex of the embryos was not assessed in the present study.
In embryos, non-CpG methylation is prevalent and is mainly mediated by the de novo methyltransferases. Non-CpG methylation is critical for cellular development and differentiation [64]. It is possible that the changes in expression pattern of the genes associated with developmental pathways in the resorbed and normal embryos could be contributed by non-CpG methylation changes. Since the methylation status of the de novo methyltransferases are altered in spermatozoa of both the DIO and GIO groups, the non-CpG methylation in these embryos could be altered. Taken together, these evidences indicate that paternal obesity could be responsible for changes in the epigenome modifying enzymatic machinery in the spermatozoa that could alter the genome-wide reprogramming after fertilization thus leading to either embryo loss or cause defects at adulthood in the DIO group.

Conclusion
Taken together, our study provides a mechanism through which diet-induced and genetically inherited obesity differentially reprograms the DNA methylome profile in the male germline. We provide a comprehensive picture of the genes and the pathways that are common and uniquely altered in the spermatozoa of diet-induced and genetically inherited obesity. We show that paternal diet-induced obesity could partially transmit the defective epigenetic signatures of developmental importance via sperm to the embryo thereby leading to embryo loss. We also show that calorie restriction of the genetically obese male rats could bring about partial changes in the germline by modulating the epigenome leading to partial restoration of the fertility in the WNIN/Ob group [10]. The differential changes in the methylation profiles in the germline between the DIO and GIO groups could be due to the difference in the white adipose tissue accumulation, which we had reported, in our previous study [10]. Knowledge of the altered genes or pathways in the spermatozoa due to diet-induced and genetically inherited obesity will further our understanding of the role of paternal contribution in the transmission of defective epigenetic signatures, leading to either embryo loss or transgenerational inheritance of deleterious traits to the offspring at adulthood as reported by other studies [2,3,6].

Animals
High-fat diet-induced obese (DIO) Wistar male rat model was developed at the ICMR-National Institute for Research in Reproductive Health (ICMR-NIRRH), Mumbai, India. A study on genetically inherited obese (GIO) male rat model of Wistar origin was performed at ICMR-National Institute of Nutrition (ICMR-NIN), Hyderabad. All animals were maintained under controlled temperature (22 °C ± 1 °C) and humidity conditions with a 12:12 h light: darkness cycle. Prior approval for the use of animals for the experimental study was taken from the Institutional Animal Ethics Committee of both the Institutes.

Group 1. Diet-induced obese (DIO) group consisted of 2 sub-groups (n = 10 animals per group)
At 21 days of age, randomly bred Wistar male rats were maintained either on a high-fat diet (HFD) containing 39.72 kcal% fat, 27.92 kcal% protein and 32.35 kcal% carbohydrate or on control diet (CD) containing 16.50 kcal% fat, 23.83 kcal% protein, 59.65 kcal% carbohydrate for 16 weeks. Both the diets were formulated at National Institute of Nutrition Animal Facility, ICMR-NIN, Hyderabad. Food and water were available ad libitum. Body weights of each animal were monitored once a week for 16 weeks as reported in our earlier study [10]. The detailed composition of the diets is provided in Additional file 1: S1.1.

Group 2. Genetically inherited obese (GIO) group consisted of 3 sub-groups (n = 10 animals per group)
21-day-old Wistar origin WNIN/Ob (homozygous mutant), CROb (calorie-restricted WNIN/Ob-homozygous mutant) and LEAN (homozygous wild type) male rats were used for the study. Male rats age-and body weight-matched to the DIO group were used. Calorie restriction was carried out for WNIN/Ob group so as to rule out the confounding factors of the mutation and report obesity-induced effects. Calorie restriction was achieved by pair feeding, such that the amount of control diet consumed by the LEAN control male rats was calculated and the same amount was fed to 35 days old WNIN/Ob male rats [10,65,66]. The calorie restriction experiments on GIO group were performed at National Institute of Nutrition Animal Facility, ICMR-NIN, Hyderabad. The age (17 weeks old) and body weights of the GIO group were matched to the DIO group for the experimental study. All animals in the GIO group were maintained on control diet (CD) containing 16.50 kcal% fat, 23.83 kcal% protein and 59.65 kcal% carbohydrate. Food and water were available ad libitum.

Mating studies and sample collection
All the male rats (17 weeks old) were cohabited with normal cycling lean female rats (n = 20 female rats per group) at a ratio of one male rat: two female rats. Mating was confirmed by the presence of sperm in the vaginal smear and was considered as day 0.5 of gestation. Our earlier data on fertility studies in the GIO group have shown that WNIN/Ob male rats were infertile and did not produce any litters [10]. Thus, the resorbed and normal embryos were acquired only from the DIO group. Gravid female rats were separated and sacrificed at day 18.5 of gestation and the fertility parameters, namely, potency, litter size, number of implantation sites, corpora lutea, number of live, dead and resorbed embryos were assessed as reported earlier [10]. Resorbed and normal embryos were dissected, snap-frozen in liquid nitrogen and immediately stored at −80 °C for RNA extraction. The collected resorbed and normal embryos were sired by different male rats within the same group (n = 6 per group).
After the mating studies, the male rats from both the DIO and GIO groups were sacrificed and both the testes excised. One testis was snap-frozen in liquid nitrogen and stored at -80 °C for RNA extraction and the other testis was processed for flow cytometry studies. For the collection of caudal spermatozoa, both the cauda epididymides were excised in DMEM (Dulbecco's Modified Eagle Medium, pH-7.4) (Sigma, St Louis, MO, USA) and incubated at 37 °C for 30 min, allowing sperm to diffuse in the medium. The sperms were passed through a 40 µm cell strainer and washed with hypotonic solution containing 0.45% NaCl so as to remove any contaminating cells. The sperms were then washed with PBS (0.01 M phosphate buffer containing 0.154 M NaCl; pH 7.4) and stored at −80 °C for DNA extraction. A schematic representation of the experimental design is shown in Fig. 1.

Genomic DNA extraction from spermatozoa
Genomic DNA was extracted from 10 million caudal spermatozoa (n = 10 animals per group) using GeneAll Exgene Cell SV mini kit (GeneAll Biotechnology, Seoul, South Korea) as per manufacturer's instructions. Briefly, spermatozoa were lysed using H2 buffer containing 20 mM Tris HCl (pH-8.0), 20 mM EDTA, 200 mM NaCl, 4% SDS, 80 mM DTT and proteinase K and incubated at 56 °C for 1 h. The lysate was then subjected to RNase A treatment for 2 min followed by further lysis using BL buffer at 56 °C for 10 min. DNA was precipitated using chilled absolute ethanol, washed and eluted in nucleasefree water. Genomic DNA yield and purity were assessed using a UV spectrophotometer. Genomic DNA with the absorbance ratio 260/280 nm of 1.6-1.8 was further processed for global DNA methylation and methylation sequencing.

Global DNA methylation analysis
Global DNA methylation levels of caudal spermatozoa were estimated using a 5-mC DNA ELISA kit (Zymoresearch, CA, USA) as per the manufacturer's protocol. Briefly, 100 ng of genomic DNA (n = 10 animals per group) from all the five groups were diluted in coating buffer and denatured at 98 °C for 5 min in a thermal cycler followed by incubation on ice for 10 min. The denatured DNA was immobilized on the microtitre plate (in duplicate) and incubated at 37 °C for 1 h. This was followed by blocking with a 5-mC ELISA buffer at 37 °C for 30 min. A standard curve was generated of known 5-mC percentage by preparing the mixtures of negative control and positive control, provided in the kit. Antibody mixture containing anti-5-methylcytosine and secondary antibody in 5-mC ELISA buffer was added in all the wells and incubated at 37 °C for 1 h, followed by washing. The monoclonal antibody binds to 5-mC in single-stranded DNA, with no detectable cross-reactivity to non-methylated or hydoxymethylated cytosines. However, the antibody could detect methylated cytosines in any context and it was not sequence specific. The reaction was detected using an HRP developer followed by color development in 10 min at room temperature (RT). The absorbance was measured at 450 nm using ELISA plate reader. The global 5-mC content was calculated from the standard curve. Percent CpG methylation (%5mC/CpG sites) levels in spermatozoa were calculated by multiplying the observed optical density (OD) from the standard curve by the fold difference between E. coli (control DNA derives from E. coli) and the rat genome according to the manufacturer's instructions.

Combined 5-methylcytosine (5-mC) and propidium iodide (PI) staining using flow cytometry
Testicular cell-type-specific 5-mC content was determined using flow cytometry (protocol adapted from Desjobert et al. 2015 [67]). Testicular single-cell suspension (n = 6 animals per group) was prepared by passing the seminiferous tubules through 100 and 70 µm cell strainers. One million testicular cells were fixed using 4% paraformaldehyde for 15 min at RT. The cells are then washed with PBS, followed by permeabilization with 0.5% Triton X-100 prepared in PBS, for 15 min at RT. Cells were again washed and treated with 2 N HCl for 30 min at 37 °C. This was followed by neutralization with 100 mM Tris-HCl (pH-8.8) for 10 min at RT and extensive washes with 0.05% Tween 20 in PBS. The cells were then blocked for 2 h with 1% BSA and 0.05% Tween 20 in PBS and incubated either with a 5-mC antibody (host: mouse; 1:50; catalog no. A3001-50; Zymoresearch, CA, U.S.A.) or isotypic control antibody for 90 min at RT. This monoclonal antibody was both sensitive and specific for 5-mC in single-stranded DNA and showed no cross-reactivity to non-methylated or hydroxylated cytosines. It could detect methylated cytosines in any context and was not sequence specific. Cells were then washed with 1% BSA and 0.05% Tween 20 in PBS and incubated with secondary antibody conjugated to a fluorophore, for 45 min at RT. The samples were then washed twice with 1% BSA and 0.05% Tween 20 in PBS and stained with PI (5 µg/ ml) for 2 h at 4 °C for flow cytometry analysis. Cell labeling was analyzed using FACS Aria SORP (Becton Dickinson; San Jose, CA, USA) and 10,000 ungated events were recorded. Flow cytometry data were analyzed using FACS Diva 6.1.3 software (BD, San Jose, CA, USA). Briefly, testicular cells were selected according to their forward scatter (FSC) and side scatter (SSC) parameters (P1 region) to exclude cell debris and then gated based on their PI content (P2 region) to exclude cell doublets, aggregates and apoptotic cells (Additional file 5-S3). 5-mC and PI fluorescence intensities of the gated cells (P2 region) were reported as dot plots and histograms, respectively (Additional file 5-S3). Dot plots display DNA methylation according to the DNA content (Additional file 5-S3). In order to compare the 5-mC density of the DNA of each testicular cell type, taking into consideration the difference in their DNA content, the mean fluorescence intensity (MFI) of 5-mC was combined with the MFI of PI (DNA) to calculate 5-mC/DNA index, as shown by Desjobert et al. 2015 [67]. Representative plots for CD versus HFD and LEAN versus CROb versus WNIN/Ob are shown as Additional 5: S4.

Methylation sequencing (Methyl-Seq)
For high-throughput DNA methylation analysis in caudal spermatozoa (n = 3, pooled genomic DNA samples per group), methylation sequencing libraries were prepared with Illumina-compatible Sure-SelectXT Methyl-Seq Target Enrichment System for Illumina Multiplexed Sequencing (Agilent Technologies, Santa Clara, CA, USA) at Genotypic Technologies Pvt. Ltd., Bangalore, India. Briefly, approximately 1 μg of genomic DNA was sheared using Covaris S2 sonicator (Covaris, Woburn, Massachusetts, USA) to generate approximate fragment peak size between 150-200 bp. The fragment size distribution was checked on Agilent Bioanalyzer (Agilent Technologies, Palo Alto, CA, USA). The fragments were end-repaired, adenylated and ligated to Illumina multiplex barcode adaptors as per SureSelectXT Methyl-Seq Target Enrichment protocol. Adapter-ligated DNA was purified with HighPrep PCR clean up system. The ligated product was then used for hybridization as outlined in the SureSelectXT Methyl-Seq Target Enrichment protocol for Illumina Paired-End Sequencing Library manual using SureSe-lectXT Rat Methyl-Seq Reagent Kit. Hybridized library fragments were isolated by magnetic capture using Dynabeads ™ MyOne ™ Streptavidin T1 (Invitrogen, Carlsbad, CA, USA). The captured library was eluted in 20 μl of SureSelect Elution Buffer. The captured library was then subjected to bisulfite conversion using the EZ DNA Methylation Kit. Here, unmethylated cytosine residues in the library are converted to uracil residues and methylated cytosine residues remain unmodified. After desulphonation, the treated DNA was amplified by PCR, converting uracil residues in the sample to thymidine, using 8 PCR cycles. Clean up of PCR product was done using HighPrep PCR clean up system. The modified captured DNA library was indexed by PCR indexing amplification using 6 PCR cycles. The final PCR product (sequencing library) was purified with HighPrep PCR clean up system. Quantification and validation of the Captured Library were carried out using Qubit fluorometer and running an aliquot on Agilent Bioanalyzer. Finally, the sequencing library was accurately quantified by quantitative PCR using the Kapa Library Quantification Kit (Kapa Biosystems, Wilmington, MA, USA). The qPCR quantified libraries were pooled in equimolar amounts to create final equimolar amounts for multiplexed paired end sequencing on Illumina NextSeq 550 sequencer for 75 cycles.

Sequencing data analysis
The Illumina paired end raw reads (~ 35 to ~ 40 million raw reads were generated for each sample) were quality checked using FastQC (https ://www.bioin forma tics.babra ham.ac.uk/proje cts/downl oad.html). The raw reads were processed using a PERL script developed at Genotypic Technologies Pvt. Ltd. for removal of adapters and low-quality bases. The processed data was mapped to the rat reference genome [Jul.2014 (RGSC 6.0/rn6)] using Bowtie2 version 2.2.7 [68] and methylated cytosines were identified using the Bismarkv0.16.3 tool [69]. The reference genome was converted into a bisulphite-converted version (conversion of C to T and G to A) and then indexed using Bowtie2-build program [68]. Methylation levels were extracted from the alignments using Bismarkv0.16.3 methylation extractor [69]. From alignment output, every C's which depends on the context as CpG, CHG and CHH (where H is A, C or T), the methylation call were extracted and reported using Bismarkv0.16.3 tool [69]. The percentage of methylation level was calculated by the number of methylated counts divided by all counts (methylated + non-methylated) summed across CpGs within the region. PCA plots were generated using MethylKit [70]. Methylated positions were annotated using Homer tool. Differentially methylated regions across samples were assessed using DESeq tool [71]. Log of global methylation data was retrieved from the DGE (differential gene expression) report and whisker box plots for global methylation of DMRs were developed using R script. Volcano plots were also created using the R script. Differences in fold coverage of DNA methylation were considered significant when the p value was < 0.05 and that the log2 foldchange > 1 was considered hypermethylated and log2 foldchange < 1 was considered hypomethylated. Gene ontology analysis and pathway analysis was done using PANTHER (PANTHER 14.1) and KEGG databases. A p value and false discovery rate (FDR) cutoff of < 0.05 were applied to obtain the list of significant pathways. Further, normalized methylation values from DGE were retrieved and heatmaps for each pathway were generated using Heatmap2 R package.

Pyrosequencing
Validation of differentially methylated genes in spermatozoa was carried out using Pyrosequencing. Genomic DNA from spermatozoa (n = 5 per group) was subjected to bisulfite modification using MethylCode Bisulfite Conversion Kit (Invitrogen) as per manufacturer's instructions. Briefly, 1.5 µg of genomic DNA was sequentially denaturated, sulphonated, deaminated and desulphonated to convert the unmethylated cytosines to uracil whereas the methylated cytosines remain unmodified. The modified DNA was eluted and used for PCR amplification of the genes of interest. PCR was performed using primers specific for modified sequences of genes of interest (Sigma, St Louis, MO, USA) using the Pyromark PCR amplification kit (Qiagen, Hilden, Germany). The primers were designed based on the coordinates for the point methylation change at the CpG site for the genes of interest obtained after methylation sequencing data analysis (Tables 1, 2). Either the forward or reverse primer was biotinylated. In order to check for completion of bisulfite reaction, the modified DNA was first subjected to PCR with primers specific for genomic DNA sequence (i.e., wild-set primers) of the same locus that is to be analyzed. A completely converted sample would not give any amplification with these primers. Genomic DNA was used as a template for positive control (Additional 5: S6). Primer sequences, coordinates, annealing temperature and amplicon sizes are listed in Additional file 1: S1.12 and Additional file 2-S2.11. Amplification conditions used were as follows: initial denaturation for 15 min at 95 °C, followed by 45 cycles of denaturation at 95 °C for 30 s, primer annealing for 30 s, extension at 72 °C for 30 s, and final extension at 72 °C for 10 min. The PCR products were bound to streptavidin-coated Sepharose beads (GE Healthcare Bio-Sciences AB, Uppsala, Sweden). Bound PCR products were denatured, and the non-biotinylated strand was washed off. The bound biotinylated strand was annealed to the sequencing primer, which was then subjected to pyrosequencing in Pyro-Mark Q96 ID (Qiagen). Pyrosequencing detects the percentage of nucleotides at a given position in a sequence; therefore, it could be used to identify the level of C or T at a given CpG site. It was important to note that the PyroMark Q96 ID Pyrosequencer (Qiagen) used in the present study could detect only CpG methylation and not non-CpG methylation. The sequence of the template and the positions of the CpGs were marked and instructed before the run. At non-variable positions only the expected nucleotide was dispensed while at the variable position, i.e., of CpG, both C and T were dispensed and the amount of incorporated nucleotides revealed the proportion of C or T at the CpG position. The dispensation order was predetermined by the software for nucleotide incorporation. The nucleotides expected to be incorporated and nucleotides which were expected to give no signal were included (called blank dispensation). Each generated signal that was not associated with a CpG site (methylation detection site) is called a reference peak, used as internal control concerning quality issues and to calculate the expected single peak height. The histogram function provided by the software represents the theoretical peak height, which is used to validate the observed peak height. In-built internal controls and conversion controls were provided for DNA methylation analysis. As conversion control, thymines in the forward direction or adenines in the reverse direction were provided by the software to control the bisulfite-mediated reaction. Qiagen pyro Q-CpG software provided an efficient quality control system for each sequencing run. Conversion controls of completely bisulfite converted genomic DNA at a specific locus did not show any intensity signal. Default settings of the software for the threshold to pass quality control (blue assay) was a percentage of 4.5 and a percentage of 7.0 for conversion controls that have to be rechecked manually (yellow colored) [72].Representative pyrograms are shown as Additional 5: S7.

RNA extraction and quantitative real-time PCR
Total RNA was extracted from 100 mg of the testis (n = 6, per group) and whole resorbed and normal embryos (n = 6, per group, day 18.5 of gestation) using TRI pure (Roche Diagnostics) as per the manufacturer's instructions. Total RNA was treated with DNAse I enzyme at 37 °C for 1 h. RNA concentration and purity were assessed by measuring the absorbance at 260 and 280 nm. Total RNA (2 µg) was reverse transcribed using a High-Capacity cDNA reverse transcription kit from Applied Biosystems (Foster City, CA, USA) as per manufacturer's protocol. Quantitative Real-time PCR was performed on Light cycler 96 Real-time PCR system (Roche) using Takyon SYBR green master mix (Eurogentec, Seraing, Belgium). The relative expression levels of genes of interest were estimated in relation to Rn18s. Amplification reactions (20 µl) containing 1.6 µl cDNA, 10 pM of respective primers and SYBR green master mix with the thermal cycling conditions of initial denaturation of 10 min at 95 °C followed by 40 cycles of 95 °C for 10 s, primer annealing temperature for 10 s, and extension at 72 °C for 10 s. Amplification reactions were run in duplicate and a no template control was included. Primer sequences, accession number, product size, and annealing temperature are mentioned in Additional file 1: S1.13. To check the specificity of the primers, melt curve analysis was performed and all the PCR products yielded the predicted melting temperature (Additional file 5-S2). Primer efficiencies were calculated using a standard curve (Additional file 1: S1.13). Pfaffl method was used to quantitate the relative gene expression. All the quantitative Real-time PCR procedures and data analysis followed MIQE guidelines (Additional file 1: S1.14).

Statistical analysis
Data analysis was performed using Graph Pad Prism (version 5; Graph Pad Inc.; San Diego, CA, USA). For comparison between CD and HFD groups, unpaired Student's t-test was used with Welch's correction. For comparison between LEAN, CROb, and WNIN/Ob groups, one way ANOVA was used. The level of significance was considered as p ≤ 0.05. All the data are represented as mean ± S.E.M.