Fetal Exposure to Maternal Inflammation Does Not Affect Postnatal Development of Genetically-Driven Ileitis and Colitis

Background: Chronic inflammatory disorders have been increasing in incidence over the past decades following geographical patterns of industrialization. Fetal exposure to maternal inflammation may alter organ functions and the offspring's disease risk. We studied the development of genetically-driven ileitis and colitis in response to maternal inflammation using mouse models. Methods: Disease susceptible (Tnf ΔARE/+ and IL10−/−) and disease-free (Tnf +/+ and IL10−/+) offspring were raised in inflamed and non-inflamed dams. Ileal, caecal and colonic pathology was evaluated in the offspring at 8 or 12 weeks of age. Ly6G-positive cells in inflamed sections from the distal ileum and distal colon were analysed by immunofluorescence microscopy. Gene expression of pro-inflammatory cytokines was measured in whole tissue specimens by quantitative PCR. Microarray analyses were performed on laser microdissected intestinal epithelium. Caecal bacterial communities were assessed by Illumina sequencing of 16S rRNA amplicons. Results: Disease severity, the number of infiltrated neutrophils as well as Tnf and Il12p40 mRNA expression were independent of maternal inflammation in the offspring of mouse models for ileitis (Tnf ΔARE/+) and colitis (IL10−/−). Although TNF-driven maternal inflammation regulated 2,174 (wild type) and 3,345 (Tnf ΔARE/+) genes in the fetal epithelium, prenatal gene expression patterns were completely overwritten after birth. In addition, co-housing experiments revealed no change in phylogenetic diversity of the offspring's caecal microbiota in response to maternal inflammation. This is independent of the offspring's genotype before and after the onset of tissue pathology. Conclusions: Disease risk and activity in mouse models of chronic ileitis and colitis was independent of the fetal exposure to maternal inflammation. Likewise, maternal inflammation did not alter the diversity and composition of offspring's caecal microbiota, clearly demonstrating that changes of the gene expression program in the fetal gut epithelium were not relevant for the development of chronic inflammatory disorders in the gut.


Introduction
Lifestyle changes in industrialized countries are associated with a sharp increase in the incidence of immune-mediated chronic pathologies such as inflammatory bowel diseases (IBD), multiple sclerosis or type-1 diabetes. Th1-driven inflammatory processes are key mechanisms in the pathogenesis of these chronic pathologies [1]. IBD are spontaneously relapsing, immunologically mediated disorders of the gastrointestinal tract [2]. Although the pathogenesis of these multifactorial diseases is still not fully understood, the combination of genetic predisposition [3] and environmental factors (microbiota, diet and lifestyle) [4] drives disease development. Apart from the major impact of IBD on life quality of the affected patients, epidemiological studies link Crohn's disease (CD) activity during pregnancy to adverse outcomes such as preterm birth, spontaneous abortion and labor complications [5,6].This is not clear in the context of ulcerative colitis.
Despite the fact that alterations of the cytokine milieu during CD-associated pregnancy lead to peri-and postnatal complications [6][7][8], the transmission of maternal inflammation to the offspring with consequences for later disease susceptibility, severity or phenotype is completely unknown. A potential impact of inflammatory processes during pregnancy on fetal organ functions was recently shown for endotoxin-induced chorioamnionitis [9]. In this context, the maturation of the fetal gut barrier was prevented [10]. Furthermore, maternal exposure to high-fat diet was recently reported to induce intestinal inflammation in fetal sheep, suggesting that even low grade maternal inflammation might affect intestinal functions and IBD susceptibility in the offspring [11]. Maternally transmitted compositional changes of the microbiota might be an important factor that contributes to this influence on disease susceptibility in the offspring [12].
In this context, we asked the question whether chronic maternal inflammation is a risk factor for postnatal disease susceptibility in the normal and genetically susceptible host. We applied sophisticated breeding systems with heterozygous Tnf DARE/+ and homozygous IL10 2/2 mice to generate genetically-driven inflammatory disease environments in utero in two well-established models of chronic ileitis and colitis, enabling us to study the role of maternal inflammation on postnatal disease onset. Exemplarily, we took also advantage of the Tnf DARE/+ model with ileitis to assess transcriptional profiling of fetal and postnatal intestinal epithelial cells (IEC) obtained by laser microdissection in order to provide high resolution of cellular specificity at this disease relevant interface. This second objective fits with the hypothesis that the pathogenesis of CD is characterized by a failure of innate immune mechanisms to recognize microbial triggers at the early stage of disease development [4]. Selective overexpression of TNF in the intestinal epithelium seems to be sufficient to trigger CD-like ileitis [13], suggesting an important role of the epithelium in the pathogenesis of chronic intestinal inflammation in Tnf DARE/+ mice [14].
Thus, the aim of the present study was to characterize the role of genetically-driven maternal inflammation in programming the fetal epithelium towards postnatal development of intestinal inflammation.

Ethics Statement
Mouse experiments were performed between 2009 and 2011 in accordance to the German guidelines for animal care (Regierung von Oberbayern, Bavaria, Germany). No animal approval was obtained, because no intervention was performed with living mice. All mice were reported as mice ''sacrificed for research purposes'' at the Regierung von Oberbayern.

Animals and Experimental Design
All mice were conventionally housed in groups of 3-5 mice per cage at a 12 h light/dark cycle at 24-26uC. They received fresh tap water and breeding diet (Ssniff Chow) ad libitum and were sacrificed by neck dislocation. Tnf DARE/+ (ARE) dams were bred with Tnf +/+ wildtype (WT) sires (C57BL/6N genetic background) and vice versa (n = 5-10 breeding pairs), generating offspring from healthy WT dams (WT and ARE) and inflamed ARE dams (iWT and iARE) ( Figure 1A) (n = 5 each). The period of conception was set from 8-12 weeks in order to avoid suffering in genetically susceptible ARE dams. Dams were sacrificed at an average age of 13 weeks +/2 5 d (WT dams) and 11 weeks +/2 1 d (ARE dams) for the prenatal time point and at 17 weeks +/2 1 d (WT dams) and 14 weeks +/2 3 d (ARE dams) for the weaning time point (3 weeks after giving birth). Offspring were sacrificed at the age of 17.5 days post conception (dpc), 3 (weaning) and 8 weeks by neck dislocation. Morphological criteria of fetuses at 17.5 dpc were evaluated according to Theiler stage 25 (TS25) based on the onset of skin wrinkling, whiskers and eyelid closure. Placentas and offspring's gut were embedded in Optimal Cutting Temperature (O.C.T.) matrix (Sakura Finetek, Torrance, USA) and stored at 280uC until laser microdissection of intestinal epithelial cells and subsequent microarray experiments.
For co-housing experiments, ARE and WT dams were kept together in one cage to generate mixed beddings from the age of 4 weeks. From week 8 on, dams were mated with WT males overnight only until pregnancy was observed. During the day, dams were co-housed again. After giving birth, beddings of ARE and WT dams and litters were exchanged to create a synchronized environment independent of the dam's genotype, so as to be able to analyse only the effect of inflammation in utero on shaping gut bacterial colonization. Dams and offspring were sacrificed before weaning, i.e., 3 weeks after birth.

Histopathology
Scoring was performed on 10% formalin-fixed paraffin-embedded or cryo-fixed tissue for the postnatal or prenatal time point, respectively. The histological score was ascertained in a blinded fashion on H&E-stained transversal sections of the terminal ileum (WT, iWT, ARE, iARE) or of the cecum tip, proximal colon and distal colon (IL10 +/2 , IL10 2/2 , iIL10 +/2 , IL10 2/2 ), resulting in a score from 0 (non-inflamed) to 12 (highly inflamed) per section as previously described [15]. The total colitis score per mouse was determined by calculating the mean of all three colonic compartments.

Plasma Measurements of Maternal Tumor Necrosis Factor
Plasma TNF was measured using Mouse TNF alpha ELISA Ready -SET-Go! ELISA, according to the manufacturer's instructions (eBioscience, San Diego, USA). A volume of 100 ml total plasma was incubated on a pre-coated plate for 2 h at RT (room temperature), followed by incubation of anti-TNF detection antibody linked with Avidin/Biotin. HRP-conjugated antibody was incubated for 30 min and substrate conversion was stopped after 15 min with 2N H 2 SO 4 . The product absorbance of standard dilutions and plasma sample was measured at 405 nm to the reference wavelength of 570 nm. Quantification was performed using the linear equation of the standard dilutions.

Immunofluorescence Staining of Ly6G and REG3B in the Intestine
Transversal sections (5 mm thick) were cut from either formalinfixed paraffin embedded (FFPE) or from frozen distal ileum or colon and transferred onto Superfrost Plus slides (Thermo Scientific, Braunschweig, Germany). After deparaffinization of FFPE tissue (Leica ST5020 Multistainer system), antigen demasking was performed by boiling in 16 sodium citrate buffer (pH 6, 900 W, 23 min). After cool down to RT, slides were washed 3 times in dH2O for 5 min, followed by 5 min in PBS. Frozen sections were equilibrated to room temperature (30 min), fixed in ice cold acetone (220uC) for 10 min, followed by 30 min air drying at RT and 3 times 5 min washing in PBS. Sections were blocked with 50 ml blocking buffer raised against the species of the secondary antibody for 60 min at RT in a humidified chamber.
Primary antibody against Ly6G (rat-anti-Ly6G, BD Pharmingen, 1:500) was incubated overnight at 4uC. Primary antibody against REG3B (sheep-anti-REG3B, R&D Systems, 1:100 dilution) was incubated at RT for 1-3 hours in a humidified chamber. Fluorochrome-conjugated secondary antibodies, namely goat anti-rat IgG (H+L) Alexa Fluor 546 (Invitrogen) or donkey antisheep IgG (H+L) DyLight 488 (Jackson ImmunoResearch), were diluted 1:200 and incubated for 1 h at RT. Nuclei were counterstained using DAPI (1:2000) in secondary antibody solution. Sections were visualized using a confocal microscope (Olympus Fluoview 1000 using the FV10-ASW software). The amount of Ly6G positive cells per area was counted using the Volocity 5.51 software (Perkin Elmer) defining the lamina propria as region of interest. For each individual mouse, 3 microscopic fields at a 600-fold magnification were quantified for mean Ly6Gpositive cells per mm 2 . Immunofluorescence intensity of REG3B was quantified with Volocity Demo 5.5 software (Perkin Elmer) defining epithelial cells as region of interest. For each individual mouse 3 different areas were quantified as mean intensity of the fluorescence signal per mm 2 . Offspring developed under non inflamed conditions (white background: white mouse = WT or IL10+/2; grey mouse = ARE or IL102/2 ) and under maternal inflammation (grey background: white mouse = iWTor iIL10+/2; grey mouse = iARE or iIL102/2). (C) Representative H&Estained transversal sections of the distal ileum and individual plots of ileitis score in WT (n = 6) and ARE (n = 8) dams. (D) Total colitis score in IL10+/2 (n = 6) and IL102/2 dams (n = 8) and representative H&E-stained sections of distal colon from non-inflamed IL10+/2 and inflamed IL102/2 dams. Scores [0, not inflamed, to 12, highly inflamed] were determined using tissue sections from dams sacrificed 3 weeks after giving birth. TNF in maternal plasma indicates that the inflammation is also systemically relevant in both the ileitis (E) and colitis (F) models. Individual data and means are shown; t-test, *p,0.05, ***p,0.001. doi:10.1371/journal.pone.0098237.g001 RNA from cryosections of distal ileum (3 sections of 10 mm each) and colonic swiss rolls (2 sections of 10 mm each) was isolated using the RNA isolation kit according to the manufacturer's instructions (Macherey & Nagel). Reverse transcription was performed using SuperScript III Reverse Transcriptase (Invitrogen). RNA of laser microdissected IEC (10 ng) was preamplified with the QuantiTect Whole Transcriptome Kit (Oiagen, Hilden, Germany) according to the manufacturer's instructions. Genespecific nucleotide sequences and accession numbers were obtained from the National Center for Biotechnology Information (NCBI) website (http://www.ncbi.nlm.nih.gov/gene). Primer pairs ( Table 1) were designed using the Universal ProbeLibrary (UPL) design center (Roche Diagnostics, Mannheim, Germany). Quantitative real-time PCR was performed on 10 ng cDNA using the LightCycler 480 System (Roche Diagnostics, Mannheim, Germany). Crossing points (Cp) were determined using the second derivative maximum method by the LightCycler 480 software release 1.5.0. Data were normalized to the Cp mean of reference genes (Gapdh for whole gut tissue and 18s, Rpl13a and Gapdh for pre-amplified cDNA of IEC) and expressed as 2 2DCt values in order to compare expression levels among all groups.

Laser Microdissection (LMD) Microscopy of Intestinal Epithelial Cells
Ileal cryo-sections (Microm, Walldorf, Germany) were generated at 220uC. PET frame slides (MicroDissect, Herborn, Germany) were treated with RNase ZAP (Sigma-Aldrich, Steinheim, Germany) before use and dried at RT. Transversal sections (10 mm) were mounted on slides, air-dried and stored at 280uC for short periods of time (,7 d) until use. Each slide was stained directly before LMD microscopy. Briefly, after equilibration to RT (2 min), slides were fixed for 1 min with 70%(v/v) EtOH, rinsed with Diethylpyrocarbonate (DEPC) water for 30 sec, stained with Harris hematoxylin for 1 min and rinsed with DEPC water for 30 sec. After bluing with 0.1% (v/v) NH 4 OH for 30 sec, slides were counterstained with 2.5% Eosin for 2 min. Finally, sections were dehydrated in ascending EtOH series (96%, 100%) for 30 sec each and air dried at RT for 5 min. Ileal IEC were cut at a magnification of 6306 using the UV lasercutting system LMD 6000 and the Leica Application Suite software (Leica, Wetzlar, Germany). Lysis buffer (100 ml) supplied in the AllPrep DNA/RNA Micro Kit (Qiagen, Hilden, Germany) was added to epithelial pieces directly after dissection. Samples were kept frozen at 280uC until DNA/RNA isolation. In total, a mean area of 1.33060.024610 6 mm 2 IECs was collected per sample ( Figure S1).

RNA Isolation and Quality Control
Total RNA was isolated using the column-based AllPrep DNA/ RNA Micro Kit (Qiagen, Hilden, Germany). RNA concentration was measured using the Quant-iT RiboGreen RNA Assay Kit (Invitrogen, Eugene, USA). RNA integrity was determined using the RNA 6000 Pico Kit and the Bioanalyzer 2100 (Agilent, Waldbronn, Germany). An amount of 50 ng total RNA (3 ml sample volume adjusted by vacuum centrifugation) was used for microarray analysis.

Microarray-based Gene Expression Analysis
Microarray analysis was performed with ileal IEC from 17.5 dpc and 8-week-old mice (n = 20 in total). All samples from one time point were run together. RNA preparation, reverse-transcription, amplification and biotin labeling were performed using the GeneChip 39 IVT Express kit (Affymetrix, Santa Clara, USA). Hybridization, washing and staining were performed using the GeneChip Hybridization, Wash and Stain (Affymetrix) and the GeneChip Fluidics Station 450 (Affymetrix). Labeled RNA samples were hybridized to customized murine genome Nu-GO_Mm1a520177 microarrays containing 23,865 probe sets (covering 15,313 genes, Affymetrix). Gene chips were visually inspected for irregularities and scanned with GeneChip Scanner 3000. Data were analyzed using the Affymetrix GCOS Manager software and the R-and Bioconductor-based MADMAX interface (Management and Analysis Database for Multi-platform micro-Array eXperiments, https://madmax.bioinformatics.nl), including R version 2.11.1, Bioconductor version 2.6 and AnnotationDbi version 1.10.2. Data were normalized using the gc Robust Multichip Average (slow) algorithm. Probe sets were annotated to the transcripts with Custom Chip Definition files version 13.0.0. (nugomm1a520177mmentrezg.cdf). No previous filtering was applied to the datasets. MADMAX calculated Fold Changes (FCs), p-values, FDR-values and q-values to the study groups with the LIMMA procedure. The LIMMA (log2 based) FCs were calculated comparing iWT, ARE and iARE groups to the WT control group (n = 5 each). Genes were considered to be significantly regulated according to the raw p-value of LIMMA with p,0.05 and a threshold FC of 61.5. Heatmaps were generated using the MultiExperiment Viewer (TigrMEV) software. Gene Ontology (GO) terms were computed using the GeneRanker program (Genomatix, München, Germany). Overrepresentation of biological terms were calculated and listed in the output together with respective p-values. Illumina Sequencing of 16S rRNA Gene Amplicons from Caecal Contents Bacterial DNA was obtained after bead beating and ethanol precipitation [16] from directly frozen caecal contents (co-housing experiment) or from ceacal contents embedded in O.C.T. Amplicons of the V4 region of 16S rRNA genes were obtained after 25 PCR cycles as described previously [17]. They were sequenced in paired-end modus (PE200) using the MiSeq system (Illumina Inc., San Diego, USA). Sequences were analyzed using in-house developed pipelines partly based on UPARSE [18], the open source software package QIIME [19] and the Ribosomal Database Project [20]. Sequences were filtered at a base call accuracy of 99%. Sequences containing any ambiguous nucleotide (N character) were discarded. The presence of chimeras was checked after dereplication using UCHIME [21]. Operational taxonomic units (OTUs) were picked at a threshold of 97%. Only those OTUs occurring in at least one sample at abundances .0.5% total sequences were included in the analysis. Sequence proportions of bacterial taxa were analyzed for significant differences using F-Test followed by Benjamini-Hochberg correction for multiple testing in the R programing environment (2008, ISBN 3-900051-07-0).

Maternal Inflammation does not Influence Postnatal Development of Intestinal Inflammation
Conventionally raised Tnf DARE/+ (ARE) and IL10 2/2 mice were used to study the impact of maternal inflammation on the intestine in healthy and genetically susceptible offspring. As expected, all ARE dams were affected by moderate inflammation in the distal ileum (score 5.2 6 0.5) ( Figure 1C). IL10 2/2 dams showed a more heterogeneous inflammation in the large intestine and an inflammation score of 2 was set as minimal value to discriminate inflamed dams, resulting in a mean inflammatory score of 3.3 6 0.8 for the dams that were included in the study ( Figure 1D). The local inflammation in the intestine of dams was reflected systemically by significantly elevated TNF levels in the plasma of both inflamed Tnf DARE/+ (2.9 6 1.9 pg/mL vs. 100.06 31.6 pg/mL) and IL10 2/2 mice (7.6 6 3.3 pg/mL vs. 23.5 6 10.4 pg/mL) ( Figures 1E and F). The abundance of plasma TNF significantly correlated with the inflammation grade in the intestine (data not shown). The intestinal inflammation of healthy and genetically susceptible offspring was found to be unaffected by maternal inflammation in both models at the age of 8 weeks (ileitis) and 12 weeks (colitis) (Figure 2A and B). Both, ARE and iARE mice developed moderate ileitis with comparable histological scores (4.4 6 0.9 and 4.0 6 0.7). Likewise, IL10 2/2 and iIL10 2/2 offspring showed similar histological grades of colitis (3.0 6 1.1 and 2.7 6 1.1). There was no correlation between maternal and offspring's intestinal inflammation (data not shown). Additionally, the number of infiltrated Ly6G-positive cells into inflamed tissue of the distal ileum of Tnf DARE/+ offspring or the distal colon of IL10 2/2 offspring was unaffected by maternal inflammation (ARE vs. iARE: 473 6 262 cells/mm 2 vs. 621 6 434 cells/mm 2 ; IL10 2/2 vs. iIL10 2/2 : 521 6 454 cells/mm 2 vs.497 6 441 cells/ mm 2 ) ( Figures 2C and D). The variation of neutrophil numbers in the distal colon of IL10 2/2 was strongly associated (r = 0.82, p,0.0001) with histological scores in the distal colon, which showed high heterogeneity (total colitis score from 0.6 to 8.8). The fact that histological scores were not affected by maternal inflammation was supported by unchanged mRNA expression of pro-inflammatory cytokines in whole gut sections from the distal ileum (Tnf DARE/+ ) and colonic swiss rolls (IL10 2/2 ) between the different offspring groups (Figures 2E and F). In other words, Tnf and Il12p40 mRNA levels were not altered between ARE and iARE or IL10 2/2 and iIL10 2/2 , but were significantly increased between WT and ARE or IL10 +/2 and IL10 2/2 offspring. All these data consistently show that maternal inflammation does not impact genetically-driven disease phenotypes in both mouse models of intestinal inflammation. In addition, these results raised the question of whether there is any early transcriptional programming effect on the offspring's epithelium that is induced by maternal inflammation.

Transcriptional Fingerprints of the Fetal Epithelium in Response to Maternal Inflammation do not Persist in Grownup Mice
Selective overexpression of TNF in the intestinal epithelium is sufficient to trigger CD-like ileitis [13], suggesting an important role of the epithelium in the pathogenesis of chronic intestinal inflammation in Tnf DARE/+ mice. Furthermore, TNF plays a pivotal role in the development of CD, which has been effectively treated with anti-TNF agents [22]. To further characterize the role of maternal inflammation on the offspring's intestinal homeostasis, we focused on the impact of maternal inflammation on IEC. Since IL10 2/2 dams and offspring showed variable degrees of disease severity along the large intestine, we focused on offspring from Tnf DARE/+ and Tnf +/+ mice to perform this more detailed cellspecific analysis. Therefore, we measured gene expression profiles in IEC at pre-(17.5 dpc) and postnatal (8 weeks) time points using laser dissected ileal epithelium. Randomly selected fetuses of different stages (15-19 dpc) are shown in Figure 3A. At 17.5 dpc, fetuses had comparable sizes, but body weight was decreased in iARE compared to ARE fetuses ( Figure 3B). Equivalent surface areas of laser microdissected IEC were collected in all groups ( Figure S1A). RNA integrity numbers (RIN) were 5.0 6 0.8 and 5.5 6 1.2 for fetal and postnatal RNA, respectively.
Gene expression profiles showed two distinct patterns according to maternal and postnatal inflammation (GEO data base accession number GSE44433). In fetal IEC, gene expression was highly influenced by maternal inflammation with 2,174 and 3,345 significantly regulated genes in iWT and iARE progeny, respectively ( Figure 3C). Heatmap comparisons showed similar patterns in iWT and iARE, indicating that similar genes were regulated under conditions of maternal inflammation (1,614 common genes) ( Figure 3D, Table 2). The 5 most up-and down-regulated genes at pre-and postnatal time points are shown in Table 3. Reg3b (regenerating islet-derived 3 beta) and Fabp6 (fatty acid binding protein 6) were identified among the top-regulated genes under maternal inflammation. The fetal ARE genotype was apparently not relevant at this early gut developmental stage (only 117 regulated genes). In contrast to the fetal stage, gene expression profiles in postnatal IEC were highly influenced by the ARE genotype with 1,154 and 1,197 significantly regulated genes under ARE and iARE conditions, respectively ( Figure 3A). Only 229 genes were significantly regulated in iWT compared to WT mice, indicating that the postnatal environment and the disease susceptible genotype of the offspring almost completely overwrote the gene expression program in the fetal gut. Heatmap analysis confirmed these different patterns and showed that postnatal tissue inflammation (ARE and iARE) most strikingly triggered C3 and S100a8 gene expression. The majority of regulated genes were shared according to postnatal tissue inflammation (830 commonly regulated genes between ARE and iARE mice).
The two distinct gene expression patterns (pre-and postnatal) were reflected as well in the Gene Ontology (GO) terms for 'biological processes' (GO_BP) ( Figure 3F). 'Metabolic' (GO:0008152) and 'cellular processes' (GO:0009987) were predominantly overrepresented according to maternal inflammation in the prenatal state and according to postnatal inflammation in adult mice, respectively. For example, among 'metabolic processes', 989 (iWT) and 1,501 (iARE) genes were significantly regulated in the fetal stage. Fabp6 was highly up-regulated due to maternal inflammation and is part of both 'metabolic' and 'cellular processes'. In the absence of maternal inflammation, fetal IEC from ARE mice were characterized by only 47 significant genes involved in 'metabolic processes' (Figure 3F, red bars). In contrast, the transcriptional profile of postnatal IEC revealed a lower significance of association (log10 p-value) for 'biological processes', highly influenced by postnatal but not maternal inflammation. In ARE mice, 507 and 695 genes were significantly regulated in 'metabolic' and 'cellular processes', respectively, whereas in iWT mice only 105 and 124 genes were responsible for overrepresentation ( Figure 3F, grey bars). At this postnatal stage, C3 encoding complement component 3 had the strongest impact on 'metabolic' and 'cellular processes' in ARE and iARE groups, but not in iWT mice.

Microarray Validation of Tnf and C3 mRNA Expression by qPCR
In order to validate the microarray data, we performed targeted gene expression analysis of two candidate genes via qPCR. Our first target was Tnf mRNA, which is stabilized in the genetically engineered Tnf DARE/+ mouse model and therefore served as control. We observed a very high correlation between microarray fluorescence data and 2 2DCt values ( Figure 3E). In the prenatal stage, Tnf was not expressed and not regulated in both microarrays and qPCR. At 8 weeks of age, when the inflammation in the offspring was fully established, Tnf was highly up-regulated in Tnf DARE/+ mice in both datasets. These data mirror the increased Tnf transcript levels in whole distal ileal tissues from Tnf DARE/+ offspring seen in Figure 2E. We further analyzed complement component C3 (C3), which was identified by microarray analysis as the top up-regulated gene between WT and ARE offspring (FC 39.04). We again observed a high correlation between the microarray fluorescence intensities and qPCR results ( Figure 3E). Reg3b was identified as the strongest induced gene under maternal inflammation at 17.5 dpc, albeit at low gene expression levels (GEO data base accession number GSE44433). In essence, the success of microarray validation is fragile as many factors can influence both methodologies [23]. For instance, low array spot intensities for Reg3b ranged from , 10-300 (prenatal stage) and might have resulted in a discrepancy of results indicated by a low correlation between microarray fluorescence intensities and 2 2DCt values (r = 0.51, p = 0.0007). Therefore, we investigated the intestinal expression of REG3B at the level of proteins.
Maternal Inflammation Influences REG3B Protein Expression in Adult WT Offspring, but has no Influence on the Intestinal Inflammation in Tnf DARE/+ Offspring Immunofluorescence analysis of REG3B expression overtime was performed in ileal gut sections from WT and Tnf DARE/+ offspring at 17.5 dpc, 3 and 8 weeks of age. In all groups, REG3B protein was not detectable in the prenatal gut, but protein expression was clearly detectable in the epithelium of 8-week-old mice ( Figure 4A-C). REG3B protein expression appeared earliest in the ileal epithelium at weaning (3 weeks of age) and further increased at 8 weeks. Quantification of REG3B proteins in IEC of 8-week-old mice revealed significant reduction in iWT mice compared to WT control (p = 0.008), suggesting that maternal inflammation slightly affects postnatal protein expression in the absence of any tissue pathology. But, under intestinal inflammation, both ARE and iARE mice showed almost a complete loss of REG3B protein expression when compared with WT (p,0.0001). The loss of REG3B protein suggests that postnatal tissue inflammation profoundly impacts the expression of REG3B protein ( Figure 4C and D). Tissue pathology of the distal ileum in Tnf DARE/+ offspring was not detectable in the pre-and perinatal period (until 3 weeks of age), but was moderate in 8-week-oldoffspring, the age where an inflammation-driven loss of epithelial REG3B occurs. With the finding that maternal and postnatal inflammation was associated with decreased REG3B expression in the epithelium, we measured REG3B in aqueous and bacterial fractions of caecal contents from 8-week-old mice in order to evaluate an alteration in luminal secreted REG3B. Representative blots revealed the presence of REG3B in bacterial fractions ( Figure 4E). Large amounts of REG3B were detected in bacterial fractions of WT mice, whereas moderate and low amounts were found in WT and inflamed ARE/iARE mice, respectively.

Caecal Bacterial Communities are not Altered in Response to Maternal Inflammation
Down-regulation of REG3B expression in healthy WT offspring suggested that maternal inflammation has a priming effect on the offspring's gut bacterial ecosystem. We were preferentially interested in early development of the microbiota due to maternal inflammation independently of postnatal inflammation and REG3B regulation. Therefore, we analyzed 16S-based diversity and composition in the caecal content from offspring at 3 weeks of age (pre-weaning). At this stage, we circumvented possible effects of postnatal diet on gut bacteria. We followed two different breeding strategies. Our first strategy was to co-house WT and Tnf DARE/+ dams in order to synchronize the dam's microbiota prior to colonization of the pups at birth. We obtained a total of 1,755,593 quality-checked sequences (28,633 to 86,232 per sample) representing a total of 263 OTUs with an average of 122 6 33 molecular species per sample. Members of the phyla Firmicutes, Bacteroidetes and Proteobacteria were dominant, with abundances of 70.9%, 21.3% and 4.6% total sequences, respectively. Lachnospiraceae (59.3%) and Porphyromonadaceae (8%) were the most abundant families. We observed no effect of maternal pictures per mouse were counted. Individual data and means are shown (Two-Way ANOVA, *p,0.05, **p,0.01). Correlation analysis in Tnf DARE/+ and IL10 2/2 offspring indicated strong associations between histopathological scores and infiltration of Ly6G-positive neutrophils. (E+F) Whole tissue specimens were analyzed for Tnf and Il12p40 gene expression in offspring from ileitis and colitis mouse models as described in the method section. RNA was isolated from distal ileal cryosections (3610 mm) of WT, iWT, ARE and iARE offspring and from colonic swiss rolls of IL10 +/2 , iIL10 +/2 , IL10 2/2 and iIL10 2/2 offspring (n = 5 each). Single values and means are indicated as 2 2DCt . Two Way ANOVA, *p,0.05, **p,0.001, ***p,0.0001. doi:10.1371/journal.pone.0098237.g002  inflammation on the phylogenetic make-up of caecal bacterial communities between WT and iWT offspring at the age of 3 weeks, i.e., there was no clustering of samples according to dam genotypes based on the analysis of weighted Unifrac distances ( Figure 4F). Our second strategy was to assess whether there is a general effect of Tnf DARE/+ dams on the offspring's microbiota compared to WT dams. Therefore, we performed a second breeding of WT and Tnf DARE/+ dams that were housed separately. We obtained in this case a total of 1,940,009 quality-checked sequences (16,865 to 61,102 per sample) representing a total of 286 OTUs. Again, we found no influence of the maternal Tnf DARE/+ genotype on the phylogenetic make-up of bacterial communities in caecal contents in both WT and Tnf DARE/+ offspring (n = 4 to 6 mice per group) ( Figure 4G). The number of phylotypes was also not influenced by maternal inflammation and offspring's genotype (Table S1). These data demonstrate that overall bacterial diversity is not affected by maternal inflammation or offspring's genotype before the onset of postnatal inflammation. Interestingly, analysis of the caecal microbiota at 8 weeks of age, when inflammation of the distal ileum is established (n = 4 to 6 mice per group), revealed that REG3B does not affect bacterial diversity and composition of Chow diet-fed offspring ( Figure S2). Taxonomic assignment revealed no statistically significant differences in sequence abundances due to maternal or postnatal inflammation after adjustment for multiple testing (Tables S2A  and S2B). However, feeding offspring a well-defined semisynthetic experimental diet revealed clear differences in bacterial diversity according to the offspring's genotype, i.e., PCoA analysis indicated a change in beta-diversity between WT and Tnf DARE/+ offspring (n = 4-5 mice each). Statistical analysis on mean phylogenetic distances between groups (inter-group distances) clearly indicated significance ( Figure S2). Besides inflammation-associated alterations driven by the offspring genotype, we did not observe any shifts in diversity in response to maternal inflammation (Fig S2D  and S2F). When comparing chow-fed mice at the age of 3 and 8 weeks, we observed age-dependent shifts in beta-diversity independently of dam and offspring genotype ( Figure S3A). Major agedependent effects on bacterial taxa included an increase in members of the Lactobacillaceae, and Prevotellaceae as well as decreased proportions of Lachnospiraceae and Deferribacteriaceae ( Figure S3B and Table S3).

Discussion
In the present work, we provide clear experimental evidence that maternal inflammation has no impact on the offspring's risk to develop genetically-driven ileitis and colitis, despite the fact that TNF-driven maternal ileitis extensively modulates transcriptional responses in the fetal epithelium. This clearly suggests that effects of maternal inflammation are overwritten in genetically-driven models for intestinal inflammation, indicating that the disease end point might be independent of fetal exposure to inflammation [24].
CD-associated pregnancy in humans is characterized by alterations of the cytokine milieu leading to peri-and postnatal complications including preterm labor and low birth weights [6,7]. Most importantly, we showed that 8-week-old ARE and iARE mice developed similar levels of ileal inflammation. This effect was confirmed in a second mouse model for colitis, clearly supporting the hypothesis that maternal inflammation does not affect the onset or severity of disease in the genetically susceptible offspring. Nevertheless, our experiments demonstrate for the first time that TNF-driven maternal inflammation substantially modulates the transcriptional profile in the fetal intestinal epithelium, supporting the hypothesis that not only infection-driven acute but also chronic maternal inflammation impacts on the progeny [25,26]. We identified two clearly distinct clusters of genes that were strongly associated with fetal exposure to maternal inflammation (1,614 genes) or postnatal development of TNF-mediated tissue pathology (830 genes). In fetuses, similar patterns were observed in iWT and iARE with 1,614 common genes in both groups (around 70% overlap). The fetal ARE genotype however seemed not to be relevant at the early developmental stage (117 regulated genes), suggesting that the disease-susceptible genotype does not have a major influence on fetal gut programing. In contrast, gene expression profiles in postnatal IEC were highly influenced by the genetically-driven disease phenotype with 1,154 and 1,197 significantly regulated genes under ARE and iARE conditions, respectively. Heat map analysis confirmed the substantial overlap of similarly regulated genes (around 70%; 830 genes) in these two groups of ARE mice, implying that transcriptional fingerprints in the fetal gut were completely overwritten by signals derived from the postnatal environment independently of the disease-susceptible genotype. These findings may suggest that genetically-driven ileitis in mice is largely induced by the postnatal environment.
Gene ontology analysis identified 'metabolic and cellular' processes as being significantly linked to maternal inflammation to a higher extent in the fetal than in the postnatal stage. Consistent with the over representation of 'metabolic processes', Fabp6 was strongly up-regulated in the fetal gut in response to maternal inflammation. FABP6 facilitates efficient transepithelial transport of both bile and fatty acids [27]. Although other biological processes such as 'localization' and 'death' were not influenced to the same extent as 'metabolic and cellular processes', they included top-regulated genes such as C3, S100a8 and Lyz2 in both ARE and iARE offspring. This reflects a disease-associated inflammatory program [28,29] as well as bacterial defense mechanisms [30,31]. Additionally, the strong down-regulation of Il18 by approximately 25-fold, hints at the relevance of missing immune tolerance in the Tnf DARE/+ model for CD [32].
Both the fetal and postnatal stage revealed a significant changes within the GO category 'response to stimulus', albeit with lower number of observed genes when compared with top-categories like 'metabolic and cellular processes'. Interestingly, this category comprises Reg3b as top-regulated gene under maternal inflammation at 17.5dpc. The Reg3 gene family belongs to a group of C-type lectins, based on their carbohydrate recognition domains [33]. Interestingly, REG3B protein was not detectable in the prenatal gut, but was clearly visible in the epithelium of mice at 8 weeks of age. Immunofluorescence analysis of REG3B expression over time showed appearance earliest at weaning (3 weeks). This observation is consistent with a previously published Reg3g expression analysis from Matsumoto et al. [34]. The closely related lectin REG3G is known to drive host-bacterial segregation [35], preferentially targeting Gram-positive bacteria [36]. Thus, loss of REG3B seems to be associated with inflammatory signals at postnatal stages and may contribute to the development of genetically-driven ileitis through mechanisms that involve the intestinal microbiota of Tnf DARE/+ mice [16,37]. Quantification of REG3B protein levels in IEC of 8-week-old mice revealed significant reduction in iWT, suggesting that maternal inflammation can slightly affect postnatal protein expression. However, ARE and iARE mice almost completely lost REG3B, suggesting that postnatal tissue inflammation profoundly impacts REG3B expression independently of earlier gene transcriptional up-regulation by maternal inflammation. This is supported by the fact that, fetal epithelial programming is completely overwritten in adult offspring, clearly suggesting that postnatal disease-relevant signals initialize genetically-driven ileal pathogenesis.
Interestingly, loss of REG3B in the epithelium was linked to lower REG3B levels preferably bound to bacterial fractions in the caecal content, clearly suggests that the postnatal microbiota harbors disease-relevant signals that show stronger potential to initialize the genetically-driven ileal pathogenesis compared to maternal stimuli. This is supported by the fact that maternallyinduced changes in the fetal epithelium did not cause any shifts in diversity and composition of caecal bacterial communities at 3 weeks of age before weaning. This was shown in two different breeding strategies, with and without co-housing of WT and Tnf DARE/+ dams. However, at 8 weeks of age, we observed an inflammation-driven loss of REG3B in IEC and bacterial fractions of the caecal content that is not additionally affected by maternal inflammation. The fact that the microbial composition and diversity was not affected by maternal or offspring's inflammation indicates that REG3B expression in the epithelium has no influence on the overall phylogenetic make-up of caecal bacterial communities in the Tnf DARE/+ mouse model. We cannot fully explain this absence of any major signs of maternal or inflammation-driven dysbiosis in the offspring's microbiota despite changes in REG3B expression. Diet might be one main confounding factor, since wheat-based Ssniff Chow diets are characterized by highly varying quality due to heterogeneity of raw products used during production. This suggestion was supported by clear inflammation-driven shifts in beta-diversity between WT and ARE mice when fed a well-defined semisynthetic experimental diet based on corn starch.
Taking all these findings together, we summarize that neither maternal nor genetically-driven inflammation impact the overall phylogenetic diversity of caecal bacterial communities in the offspring under conventional conditions. The observation of agedependent shifts in bacterial diversity and composition between 3 and 8 weeks of age goes in line with a previously published work by Garrett et al. [12]. But, unlike their findings that a colitogenic microbiota is transmitted from mothers to offspring under SPF conditions, we could not observe any changes through maternal inflammation in our experiments. Altogether, we conclude from our studies that maternal inflammation impacts the ileal transcriptome in fetuses, but these effects do not persist in grownup mice and are therefore not relevant for the modulation of intestinal inflammation in the genetically susceptible Tnf DARE/+ mouse. Consistent with these findings is the occurrence of age-and diet-but not maternal-dependent shifts in bacterial diversity and composition, indicating that postnatal factors largely overwrite a possible maternal influence on the offspring's microbial ecosystem in both non-disease and disease susceptible offspring. This is also in line with the fact that the host immune response at the site of inflammation is consistently unaffected by maternal inflammation in two different mouse models for genetically-driven ileitis and colitis inflammation. Consequently, maternal inflammation during gestation in mouse models did not alter the genetically-driven risk to develop chronic inflammation in the intestine.  Figure 2A, n = 5-7 mice each), or (B) experimental diet (n = 6-16 mice each). With both diets, there was no difference in inflammatory scores relative to maternal inflammation. (C+E) Analysis of phylogenetic distances indicated no significant change in beta-diversity between offspring fed the Ssniff chow diet. Comparisons of mean phylogenetic distances (weighted UniFrac) between individual mice (WT, iWT, ARE, iARE) within groups (intra-group distances, e.g. all WT) and between mice from different groups (inter-group distances, e.g. WT vs. iWT) revealed no significant differences related to the offspring's genotype or maternal inflammation. (D+F) PCoA analysis indicated an inflammation-driven change in beta-diversity between 8-week-old WT and Tnf DARE/+ offspring fed an experimental diet (n = 4-5 mice each). Statistical comparisons of phylogenetic distances indicated significant separation between WT and ARE or iWT and iARE but not between WT and iWT or ARE and iARE (Two-Way ANOVA, ***p.0.0001). (TIF) Figure S3 Age-dependent shifts in bacterial diversity and composition. (A) PCoA analysis indicated a change in diversity between mice at the age of 3 and 8 weeks (n = [18][19]. (B) Major bacterial taxa that were characterized by significantly different sequence proportions at 3 and 8 weeks of age are shown in box plots (F-test followed by Benjamini-Hochberg adjustment). Individual data for all taxa are given in Table S3. (TIF)

Supporting Information
Table S1 Observed phylotype numbers in caecal contents of WT, iWT, ARE and iARE offspring at 3 and 8 weeks of age. Indicated are means 6SD (n = 4-5 per group). Three Way ANOVA. (XLSX)

Table S2
Sequence proportions of bacterial taxa in ceacal contents from offspring (n = 4-5) at (A) 3 or (B) 8 weeks of age. OTUs occurring in less than 2 mice and ,0.05% total sequences per sample were excluded from the analysis. Sequence proportions were analyzed for significant differences using F-Test followed by Benjamini-Hochberg correction for multiple testing in the R programing environment after adjustment for multiple testing. (XLSX)

Table S3
Sequence proportions of bacterial taxa in ceacal contents from 3-and 8-week old offspring (n = 4-6) show age differences. OTUs occurring in less than 2 mice and ,0.05% total sequences per sample were excluded from the analysis. Sequence proportions were analyzed for significant age-differences using F-Test followed by Benjamini-Hochberg correction for multiple testing in the R programing environment after adjustment for multiple testing. (XLSX)