Paternal chronic colitis causes epigenetic inheritance of susceptibility to colitis

Inflammatory bowel disease (IBD) arises by unknown environmental triggers in genetically susceptible individuals. Epigenetic regulation of gene expression may integrate internal and external influences and may thereby modulate disease susceptibility. Epigenetic modification may also affect the germ-line and in certain contexts can be inherited to offspring. This study investigates epigenetic alterations consequent to experimental murine colitis induced by dextran sodium sulphate (DSS), and their paternal transmission to offspring. Genome-wide methylome- and transcriptome-profiling of intestinal epithelial cells (IECs) and sperm cells of males of the F0 generation, which received either DSS and consequently developed colitis (F0DSS), or non-supplemented tap water (F0Ctrl) and hence remained healthy, and of their F1 offspring was performed using reduced representation bisulfite sequencing (RRBS) and RNA-sequencing (RNA-Seq), respectively. Offspring of F0DSS males exhibited aberrant methylation and expression patterns of multiple genes, including Igf1r and Nr4a2, which are involved in energy metabolism. Importantly, DSS colitis in F0DSS mice was associated with decreased body weight at baseline of their F1 offspring, and these F1 mice exhibited increased susceptibility to DSS-induced colitis compared to offspring from F0Ctrl males. This study hence demonstrates epigenetic transmissibility of metabolic and inflammatory traits resulting from experimental colitis.


Offspring of DSS colitic males display a lower body weight and increased susceptibility to colitis.
To decipher epigenetically altered marks that might contribute to the etiopathogenesis of colitis, we exposed wildtype C57Bl/6(N) male littermate mice at 5 weeks of age ( Supplementary Fig. 1a) to three cycles of dextran sodium sulphate (DSS), which resulted in body weight loss and the induction of a chronic colitis in F 0 DSS mice compared to their F 0 Ctrl littermates that were meanwhile kept on tap water ( Supplementary Fig. 1b). After recovering from the last DSS cycle, males of both groups were mated overnight with healthy female littermates ( Supplementary  Fig. 1a), before the colon was harvested for histology and length assessment, which is a measure for the extent of inflammation ( Supplementary Fig. 1c-e). Next, we compared the offspring of DSS-treated males ('F 1 DSS ') with the offspring of healthy males not exposed to DSS (F 1 Ctrl ) ( Supplementary Fig. 1f). The litter sizes between offspring of F 0 Ctrl and F 0 DSS mice were comparable (data not shown). However, the body weight of male and female offspring of F 0 DSS mice ('F 1 DSS ') was significantly lower compared to offspring of F 0 Ctrl mice ('F 1 Ctrl '; Fig. 1a). F 1 DSS mice at the age of 14 weeks exhibited a morphologically healthy colon (Fig. 1b). We next studied whether chronic intestinal inflammation in the paternal generation affects their offspring's susceptibility to acute DSS colitis. For this purpose we applied 3.5% DSS to the drinking water of 7 week-old F 1 DSS and F 1 Ctrl mice for 6 consecutive days. Notably, F 1 DSS mice exhibited higher colitis scores in the proximal colon (Fig. 1c,d) and lower hematocrit levels ( Fig. 1e) compared to offspring of healthy F 0 Ctrl mice. Interestingly, female offspring exhibited more pronounced effects compared to male offspring ( Supplementary Fig. 2a-d). We noted no differences in the severity of inflammation in the distal colon of F 1 DSS compared to F 1 Ctrl mice, which, however, was severely inflamed which may have obfuscated our ability to discern modest differences ( Supplementary Fig. 2e,f). Weight loss during DSS colitis and colonic length were comparable between F 1 DSS and F 1 Ctrl mice ( Supplementary Fig. 2g,h). In summary, chronic DSS colitis in the F 0 generation was associated with lower body weight and modestly increased susceptibility to acute DSS colitis in their F 1 offspring, when compared to offspring from non-colitic mice.
Chronic inflammation induced by DSS leads to a differential DNA methylation pattern in sperm cells of the F 0 and F 1 generation. The reduced body weight and increased DSS-susceptibility of F 1 DSS mice suggested the involvement of epigenetic inheritance from F 0 DSS males to their offspring. In order to test this, we analyzed sperm cells together with fluorescence-activated cell sorted (FACS) EpCAM + CD45 − colonic intestinal epithelial cells (IECs), which are primarily affected in DSS-induced colitis 54 , from mice of the F 1 and F 0 generation ( Supplementary Fig. 1f and Supplementary Fig. 3a) and subjected them to RNA-sequencing (RNA-Seq) and reduced representation bisulfite sequencing (RRBS) to assess their global mRNA expression and methylation profile, respectively. The normalized gene expression counts ( Supplementary Fig. 3b) of epithelial (Actb, Cdh1/ EpCAM) and sperm markers (Odf1, Smcp) confirmed the purity of samples used in this study ( Supplementary  Fig. 3c-h). Moreover, multi-dimensional scaling (MDS) analysis of the filtered and preprocessed methylome data showed clear differences between epithelial and sperm cells ( Supplementary Fig. 4a).  Importantly, MDS analysis of the quality-controlled and filtered CpG sites in F 0 and F 1 sperm samples (F 0 : 820,122 CpG sites vs. F 1 : 727,273 CpG sites identified) pointed to stronger methylation differences in the F 1 generation compared to the F 0 generation (Fig. 2a,b). Comparing sperm samples of F 0 Ctrl and F 0 DSS mice resulted in 823 differentially methylated CpG sites with a methylation difference > 0.20 (477 hypo-methylated and 346 hyper-methylated). Notably, 410 of these CpG sites could be annotated to either transcripts and/or promoters (Supplementary Data 1). Surprisingly, comparing the DNA methylation pattern of sperm cells of F 1 Ctrl mice with sperm cells of F 1 DSS mice resulted in 4,617 differentially methylated sites (2,429 hypo-methylated and 2,188 hyper-methylated) of which 1,680 CpG sites could be assigned to a specific gene (Supplementary Data 2). The quality controlled filtered CpGs in both generations were annotated to mainly regions outside of CpG islands (CGIs), inter-genic regions and regulatory regions of CTCF-binding sites (transcriptional repressor 11-zinc finger protein or CCCTC-binding factor) ( Supplementary Fig. 4b,c).
We identified an overlap of 66 significantly differentially methylated genes (Supplementary Data 3) between sperm samples of mice of the F 0 and F 1 generation (37 hypo-methylated and 29 hyper-methylated genes). Importantly, this number is significantly larger than expected by chance in all 10,000 permutations (mean number after 10,000 permutations = 15.518 [3,30]; Permutation based p-value = 0) ( Supplementary Fig. 5a). The 50 most differentially methylated genes overlapping between sperm samples of both generations are shown as heatmaps in Supplementary Fig. 5b,c. Analyzing all 66 overlapping genes, ten CpG sites showed a differential methylation within the promoter or the transcription factor binding region of their annotated genes (Gm128, Pnpla1, Plekhg4, Hdac5, Tjp3, Ttc28, Pnpla1, Mir6991, Hmha1, Gm6484) and two CpGs were annotated to regulatory enhancers near Igf1r and Mcf2l. Moreover, seven differentially methylated sites were located at the same genomic location in both generations, two of which, were present within the intron or exon of the genes Mta1 and Zfp865, respectively. The remaining five sites (chr10: 12 Offspring of DSS-treated males display differential gene expression in colonic IECs at baseline conditions pointing to a dysregulation of the immune response. To explain if the differences in the sperm methylome contribute to a potential gene dysregulation in the F 1 DSS epithelium, which could possibly contribute to reduced body weight and increased susceptibility to colitis (Fig. 1a,c-e), we assessed the global mRNA expression profile of EpCAM + CD45 − colonic IECs of mice of the F 0 and F 1 generation with the latter not being exposed to DSS. MDS analysis of the mRNA expression profile not only showed a significant separation of F 0 DSS and F 0 Ctrl IECs (Fig. 2c) as expected, but notably also of F 1 DSS and F 1 Ctrl IECs (Fig. 2d). Comparative gene expression analysis of IECs of F 0 DSS and F 0 Ctrl IECs revealed 230 statistically significantly differentially expressed genes (P adjusted < 0.05; 13,198 tested genes with mean coverage of 30.83x) (Supplementary Data 4, Fig. 2e), whereas 2,358 differentially expressed genes were detected between F 1 DSS and F 1 Ctrl IECs (P adjusted < 0.05; Supplementary Data 5, Fig. 2f). The 50 most differentially expressed genes (based on their fold changes) of IECs of the F 0 generation and F 1 generation are shown as heatmaps in Fig. 3a,b, respectively. Gene set enrichment analysis for the differentially expressed genes in the F 1 generation points to a strong deregulation of the immune response in F 1 DSS epithelium involving key genes of the type I IFN response (Tbkbp1, Gbp3, Gbp9, Ifit1, Ifit2, Ifit3, Oas2, Oas3, Isg15, Oasl2, Ddx60, Mx2). This analysis also identified key genes involved in immune pathways and pro-inflammatory processes (Cxcl3, Cxcl5, Cd8a, C4b, Igj, Nlrc5, Lbp), energy metabolism (Itln1) and vascular development (Ang4) (Fig. 3c). Taken together, offspring of male mice that experienced chronic colitis exhibited an altered intestinal epithelial transcriptome, which highlights a dysregulation of immunological and inflammatory pathways. Differential DNA methylation in colonic epithelial cells. After finding significantly differentially methylated CpG sites in F 0 sperm (which may shape future F 1 somatic methylomes) affected by DSS treatment and dysregulated gene expression in their offspring, we analyzed the methylomes of EpCAM + CD45 − colonic IECs using RRBS. Despite observing a clear separation of the transcriptomes of IECs between F 1 Ctrl and F 1 DSS mice, MDS analysis of their methylomes (1,037,337 CpG sites identified) did not show a clear separation into two distinct clusters (Fig. 4a). However, comparing the methylomes of F 1 Ctrl and F 1 DSS IECs resulted in 219 differentially methylated CpG sites of which 115 CpGs could be annotated to transcripts and/or promoters (Supplementary Data 6). Consistently, we observed moderate global DNA methylation changes (20% cut-off) in IECs of F 1 DSS mice compared to F 1 Ctrl mice (Fig. 4b) indicating that DSS in F 0 mice might have a small but significant effect on cytosine methylation in the next generation. For control purposes we also analyzed the methylomes of IECs of the F 0 generation (904,007 CpG sites identified), which resulted in 2,107 CpG sites that are differentially methylated between F 0 Ctrl and F 0 DSS IECs (353 hypo-methylated and 1,754 hyper-methylated) (Fig. 4c). 866 of these sites could be annotated to transcripts and/or promoters (Supplementary Data 7). The 50 most differentially methylated sites (based on methylation difference) found in IECs of F 0 and F 1 mice are shown as heatmaps in Fig. 4d,e, respectively. The proportions of genomic region categories for quality-controlled CpGs in IECs of both generations were similar to what was previously observed in the sperm data ( Supplementary Fig. 6a,b).
An overlap of differentially methylated genes in F 0 and F 1 IECs resulted in 10 genes (two hypo-methylated and eight hyper-methylated in both generations) (Supplementary Data 8). Four of the ten sites showed a differential DNA methylation within the promoter or the transcription factor-binding region of their annotated genes in F 1 IECs (associated with genes: Ick, Adcy6, Rbfox3 and Tspear). Only one differentially methylated site was located at the same inter-genic genomic location and was hypo-methylated in both generations (chr14: 32,700,214).
The DNA methylation pattern of a subset of paternally imprinted genes from the GeneImprint database 55 was also checked in the epithelial and sperm samples to show that the bisulfite conversion for the RRBS protocol worked. As expected, we observed either no methylation (Igf2 and Dio3 genes), complete methylation (Gpr1, Peg10 and Magi2 genes) or partial methylation (Mest and Plagl1 genes) for these 'control' loci in IECs Scientific RepoRts | 6:31640 | DOI: 10.1038/srep31640 (Supplementary Fig. 7a-d), however, minor shifts were observed in the imprinting patterns of sperm cells ( Supplementary Fig. 7e-h).
Finally, the overlap analysis of differential expression and differential DNA methylation in F 1 IECs resulted in 13 genes (Supplementary Data 9, Supplementary Fig. 8a,b). 8,662 out of 10,000 permutations resulted in an overlap greater or equal to 13 (mean number after 10,000 permutations = 16.984 [3,37]; Permutation based p-value = 0.86) (data not shown). In summary, these results suggest that DSS-induced colitis in F 0 males affects the DNA methylation pattern of their germ-line and that intergenerational inheritance of these epigenetic modifications are associated with alterations of the IEC transcriptome and methylome in their offspring.
Patterns of epigenetic inheritance. To identify potential candidate genes with DNA methylation patterns that are passed on to the F 1 generation, we screened for positional overlaps between differentially methylated sites in sperm samples of the F 0 and F 1 generation as well as of IECs of the F 1 generation, which resulted in three inter-generationally inherited genes (Rbfox3, Msi2 and Ttc7) (Supplementary Data 3 and Supplementary Data 7, Fig. 5a). Interestingly, this number is significantly larger than expected by chance in 9,700 out of 10,000 permutations (mean number after 10,000 permutations = 0.677 [0, 6]; Permutation based p-value = 0.03) ( Supplementary  Fig. 9a). Furthermore, the overlap analysis, comparing differential DNA methylation in sperm samples of the F 0 and F 1 generation with the differential expression in F 1 IECs, suggested 14 different candidate genes that may be epigenetically inherited (Supplementary Data 10, Fig. 5a). Nine genes were negatively regulated of which five were hypo-methylated and up-regulated (Sema6a, Tjp2, Hdac5, Arhgef3, Dock6) and four hyper-methylated and down-regulated (Ankrd13b, Blvrb, Nr4a2, Eya2). Five were positively regulated with four hypo-methylated and down-regulated (Edar, Bcl9l, Igf1r, Mta1) and one hyper-methylated and up-regulated (Nedd4l). The regional plots showing the differentially methylated sites for Igf1r, Nr4a2, Hdac5 and Mta1 are shown in Fig. 5b,c and Supplementary Fig. 9b,c. Next, we compared our F 1 sperm DNA methylation data with the 17 validated hypo-methylated regions and their associated genes published in the study from Radford et al. 37 , who report that in utero undernourished pups displayed a lower body weight and suffered from metabolic disorders in their later lives. Interestingly, reporting a similar phenotype of lower body weight herein, we also identified three consistently hypo-methylated genes, i.e. Asap2 (also significantly up-regulated in F 1 IECs), Wif1 and Rnf149. However, the associated CpG sites of these hypo-methylated genes were covered at different locations (Supplementary Data 11).
Finally, we studied the herein-identified candidate genes and their interactions by a systematic functional network analysis (Fig. 5d). Their enriched KEGG (Kyoto Encyclopedia of Genes and Genomes) pathways and Gene ontology (GO) biological process categories are shown in Supplementary Data 12, which highlight pathways involved in notch signaling, melanoma, prostate cancer and transcriptional misregulations affecting several, diverse biological processes like brain development, skeletal muscle development, B cell homeostasis, fear/behavioral defense response, abnormalities with cardiac function, and cell type specific apoptotic process.

Discussion
To the best of our knowledge, this is the first study investigating environmentally-induced epigenetic changes in the course of DSS-induced colitis in mice. We report that offspring of males that had developed and recovered from chronic DSS colitis (F 1 DSS ) displayed a lower body weight at baseline conditions compared to offspring of ). F 1 DSS mice were also more susceptible to experimental colitis but did not exhibit spontaneous intestinal inflammation.
We performed RRBS and RNA-Seq to identify differentially methylated genes and its impact on gene expression. Global DNA methylome analysis of sperm cells revealed that the sperm epigenome was affected by prior DSS-induced colitis. Similarly, the methylome and transcriptome of IECs of F 1 DSS mice at baseline showed differentially methylated and expressed genes compared to F 1 Ctrl mice. It had already been reported that unlike any other cell type, massive epigenetic changes take place during spermatogenesis 26 . Environmental triggers may also disturb imprinting of genes that can alter the sperm epigenome, resulting in compromised gene expression in the offspring 56 . Epigenetic modifications may also shape the response and phenotype of offspring to variable environmental conditions 57,58 . While several of the identified epigenetic alterations may not affect the expression of the associated genes, they may still be transmitted by a 'silent carrier' to the next generations without affecting the gene expression and resulting in an asymptomatic phenotype 59,60 .
Notably, we observed moderate global methylation changes (20% cut-off) between IECs of F 1 Ctrl and F 1 DSS mice indicating that DSS might have small but significant effects on cytosine methylation in the next generation. In agreement with our results, Carone et al. 30 reported similar modest effects on cytosine methylations in offspring of males fed a low-protein diet. Strikingly, they found significant alterations in offspring's gene expression of many hepatic genes involved in lipid and cholesterol biosynthesis 30 .
DSS is a large molecular entity (MW 36-50 kDa) that upon oral ingestion disrupts the intestinal epithelial surface and thereby induces colitis 61 . Since DSS is not present at appreciable levels in the circulation while ingested orally, cell membranes are impermeable to intact molecules, and dextran molecules are not supposed to directly or indirectly introduce methyl groups to DNA, direct DSS effects to the genome appear unlikely, although we cannot formally exclude this possibility. Altogether it is fair to conclude that the biological phenotypes in the F 1 Figure 5. Epigenetic alterations induced by chronic inflammation in the F 0 generation were stably transmitted to offspring. (a) Overlap of differentially methylated genes between sperm samples of F 0 and F 1 mice (same direction of effect) with differentially expressed genes in F 1 IECs as well as with differentially methylated genes in F 1 IECs. The genes and their associated methylation or expression values are listed in Supplementary Data 9, Supplementary Data 3 and Supplementary Data 7, respectively. (b,c) Regional browserview plots for Igf1r (b) and Nr4a2 (c) genes that are differentially methylated (DM) in F 0 and F 1 sperm samples and differentially expressed in F 1 IECs. The plots depict the EnsEMBL gene annotations, the log transformed coverage for each quality-controlled site and the median methylation for F 1 Ctrl and F 1 DSS sperm samples. Red arrows indicate significant DM sites in Igf1r (b) and Nr4a2 gene (c). (d) Gene enrichment analysis of the query candidate genes (colored in blue) that remain connected with a relationship confidence of 0.75. Other interacting genes expressed in epithelial tissue are shown as black dots. The displayed network shows predicted functional relationships between the most functionally related genes and the query input genes, which are also analyzed for gene ontology enrichment (Supplementary Data 12). The relationship confidence is supported by several functional genomic and expression datasets. The edges between genes are colored (light red to dark red) by the confidence of the predicted relationship (relationship confidence). generation and the changes in the F 0 sperm methylome, which are -at least partially -also present in F 1 sperm, may indeed be a consequence of intestinal inflammation.
Notably, almost half of the CpG sites in our data were annotated to regulatory regions of CTCF binding sites supporting the role of chromatin modifications in regulating disease susceptibility and the low weight phenotype. Specifically, CTCF is involved in regulating chromatin structure and plays an important role as a transcriptional repressor of Igf2 (insulin-like growth factor 2) [62][63][64][65] . Small size at birth and food deprivation due to prenatal exposure to famine has been also linked to hyper-methylation of sensitive loci including the imprinted locus Igf2 66 . Similarly, we identified Igf1r (insulin-like growth factor 1 receptor) and Nr4a2/Nurr1 (nuclear receptor subfamily 4, group A, member 2) to be dysregulated in the intestinal epithelium of F 1 DSS mice. Interestingly, Igf1r down-regulation has previously been associated with intrauterine and postnatal growth retardation 67 , growth failure, mental retardation and fetal abnormalities 68 . In the herein reported study, Igf1r was found to be hypo-methylated and down-regulated in F 0 DSS sperm cells as well as in F 1 DSS IECs and F 1 DSS sperm cells, suggesting a possible link between this gene's function with the herein identified metabolic phenotype in the F 1 generation. The other possible epigenetically inherited candidate gene identified in this study, Nr4a2, was previously shown to be strongly up-regulated in adipose tissue in human obesity and is also linked to the regulation of body weight 69,70 . Furthermore, functional in vitro analyses revealed that Nr4a2 is involved in stress response as well as immune development and function 69,70 . Specifically, Nr4a2 was hyper-methylated and down-regulated in F 1 DSS IECs and F 1 DSS sperm cells, which may possibly represent a link between our metabolic phenotype, differential DNA methylation and consequently modulation of adipose tissue expansion.
Furthermore, our gene set enrichment analysis of differentially expressed genes in F 1 DSS mice collectively suggests a dysregulation of the immune response in these mice, which is in agreement with the increased susceptibility to inflammation detected in F 1 DSS mice. Altogether, we report that epigenetic modifications acquired during DSS-induced colitis were transmitted to offspring through the paternal germ-line and led to alterations in the methylation pattern of CpG sites situated in genes involved in energy metabolism, which in turn may be associated with a lower body weight and increased susceptibility to colitis in offspring of diseased mice. Moreover, this study provides a list of candidate genes that may contribute to a better understanding of epigenetically modulated pathways during or after chronic intestinal inflammation. As such, our study adds to a growing body of literature 29,[38][39][40][42][43][44][45][46][47][48][49][50][51][52][53] , which suggests that paternal lifestyle and pathological conditions (e.g. chronic inflammation) can affect spermatogenesis and can induce intergenerational transmission of epigenetic marks modulating offspring's metabolism, but also contributing to risk for disease.

Material and Methods
Mice. Newborn male and female C57Bl/6(N) littermate mice were purchased from Charles River Laboratories.
For all mice used in this study it was requested that n ≥ 6 mice of the same gender came from the same litter. Mice were maintained under specific pathogen free (SPF) conditions and were kept with a strict 12-hour light/dark cycle at the ZVTA (Medical University Innsbruck). All mice received the same diet and were fed according to a strict schedule. At 13 weeks of age male littermates were mated overnight with female littermates. After mating, males were single housed and were allowed to refill their sperm reservoir for 4 days, before they were sacrificed at 13.5 weeks of age. The Austrian Ministry of Science and Research has approved all mouse protocols and all experiments were performed in accordance with institutional guidelines.

Induction of chronic colitis by dextran sodium sulphate (DSS). Chronic experimental colitis was
induced in 5-week-old male C57Bl/6(N) littermates (F 0 DSS ) by administration of three cycles of 2% dextran sodium sulphate (DSS Reagent grade, MW 36-50 kDa, MP Biomedicals, LLC) in their drinking water for 5 consecutive days, followed by a 14-day tap water period each. In the meantime normal drinking water was administered to the other half of the litter (F 0 Ctrl ). During DSS treatment the disease activity index was determined daily by combining scores of weight loss, consistency of stool and rectal bleeding 71 . 9 days after the last DSS administration (day 55 of colitis model), F 0 DSS and F 0 Ctrl male littermates were mated overnight with littermate females.

Inflammation susceptibility test by induction of an acute DSS colitis. Acute colitis was induced in
7-week-old offspring of F 0 Ctrl and F 0 DSS mice by administration of 3.5% DSS in their drinking water for 6 consecutive days. During DSS treatment the disease activity index was determined daily by combining scores of weight loss, consistency of stool and rectal bleeding. After sacrifice, histological evaluations and length measurements were performed on extracted colons.

Isolation of intestinal epithelial cells (IECs).
Mice were euthanized at 13 weeks of age and the middle part of the colon was used for the isolation of intestinal epithelial cells (IECs). The colon parts were washed with ice-cold PBS after cut open longitudinally. Mucus was removed by shaking the intestine in 1x HBSS containing 10% FCS, 1 mM DTT, 2 mM EDTA for 10 min at 37 °C. Then the DTT concentration was increased to 2 mM and shaken another 10 min at 37 °C. After moderate vortexing for 3 × 4 min in 1x HBSS containing 10% FCS and Scientific RepoRts | 6:31640 | DOI: 10.1038/srep31640 1 mM EDTA, intestinal epithelial cells (IECs) were poured through 70 μ m cell strainer (BD Bioscience), centrifuged at 1,500 rpm for 10 min and resuspended in PBS for subsequent FACS analysis.

Sperm cell isolation. Sperm cells from sacrificed mice were isolated from the caudal epididymis and
Vas deferens as described previously 30 . Briefly, sperm was allowed to release from punctured and incised caudal epididymis and Vas deferens by incubation for 30 min at 37 °C in M2 medium (Sigma-Aldrich). Released sperm was collected and determined by microscopy for motility and purity. After RNA isolation qRT-PCR of epididymis-specific (Myh11, Actb) and sperm-specific genes (Smcp, Odf1) was performed to assess purity of sperm preparations.
Hematocrit measurement. Mice were anesthetized using a ketamine/xylazine mixture and blood samples were taken from the heart using a heparinized needle and syringe. Hematocrit levels in the whole blood were measured with a veterinary animal blood counter (Scil vet abc, HORIBA Medical).