Transcription networks responsible for early regulation of Salmonella-induced inflammation in the jejunum of pigs

Total RNA isolated from mucosal scrapings of an earlier described SISP experiment was used for micro array and QRT-PCR analysis. In the mid jejunum, intestinal segments were prepared in 4 male piglets (all from the same liter) as described [8]. 10 cm of the control segment (0 hour control) was dissected before segments were perfused for 1 hour without and with Salmonella enterica subspecies enterica serovar Typhimurium DT104 10⁹ CFU/ml according to the scheme depicted in Figure 1A. Subsequently, loops were perfused for 1, 3 or 7 hours without Salmonella and samples were dissected at 2, 4 and 8 hours after the first exposure with Salmonella (at the start of the 1 hour perfusion period with Salmonella). Details of this SISP experiment, dissection of mucosal scrapings, and RNA isolation from these scrapings was published previous [9]. RNA from mucosal scrapings of three of these four pigs (pig 2, 3, and 4) was stored as alcohol precipitate at −20°C (not sufficient material was available anymore of pig 1). After centrifugation the RNA pellet was dissolved in RNAse-free water, and the integrity of this RNA was checked by analyzing 0.5 μg on a 1% (w/v) agarose gel. The SISP experiment described in Niewold et. al. (2007) was approved by the Animal Ethics Commission in Lelystad, the Netherlands, in accordance with the Dutch law on animal experimentation [9].

The commercially printed Pig “Operon” expression micro-array was used for all hybridizations. Array slides contained a total of 13297 70-mer oligonucleotide sequences representing 10655 Sus scrofa sequences with a blastn hit to known human, mouse or pig mRNA sequences and some 3’ expressed sequence tags (Operon Array-Ready Oligo Sets™ for the Pig Genome, Version 1.0, plus the Pig Genome Oligo Extension Set, Version 1.0). All probes were printed in duplicate. Dual labeling of total RNA using the RNA MICROMAX TSA labeling and detection kit (Perkin-Elmer), hybridization and washing of slides was performed as described recently, except that 4 μg of template was used instead of 1 μg [8, 9]. A total of 6 comparative hybridizations were performed according to the scheme depicted in Figure 1B and C. For each comparison a dye-swap (duplicate) was performed. Slides were scanned and images were gridded on a GenePix 4200A 01 Autoloader 116826 (Molecular Devices, Apeldoorn, The Netherlands). Data files were processed in GenePix Pro 6.1.0.4 or 6.0.1.25 (Molecular Devices, Apeldoorn, the Netherlands). Data normalization (blank-specific background correction, LOWESS fit function with a fraction of 0.2) was performed using a customized version of the statistical software package R for simultaneous data analysis of dye-swaps. Significantly differential expressed probes with M value (Log 2 scale) of < −1.58 or >1.58 (a ratio greater than 3-fold) and with a p-value <0.025 were selected. For each probe 4 spots were hybridized, 2 on one slide and 2 on the dye-swap slide. Probes with more than one missing values were removed from gene-lists used for bioinformatics analysis. Results of these micro array comparisons are posted in the NCBI GEO database (accession number GSE41630)

Oligonucleotide sequences of differential expressed probes not annotated yet, or annotated as Unigene, tentative consensus sequences (TC) or mRNA accession number, were compared with the NCBI non-redundant nucleotide databases using blastn and blastx options to assign a gene-name to these probes. Probes that did not produce a significant match with any other eukaryotic mRNA/gene were excluded from gene lists used for functional analysis. Throughout this manuscript official human gene-symbols (HUGO Gene Nomenclature Committee) were used in the text and in all (supplementary) figures and tables. Response genes detected in inter-animal comparisons were assigned to a list of genes based on their similarity in response to cytokines IL8 at 2 hours and IL1B at 4 hours (see results section for a detailed explanation). Because of the important role of IL8 and IL1B in inflammatory processes, these lists were named after these cytokines. Lists of gene-symbols were uploaded separately in web-based bioinformatics programs.

The Database for Annotation, Visualization and Integrated Discovery (DAVID version 6.7) website [12] and the “Set Distiller” module of GeneDecks [13] were used to assign genes to a specific pathway. Because far more human genes are annotated, and more information in databases is available for humans than for pigs, the human background was used for this functional analysis. In DAVID pathways (KEGG and Biocarta) with a p-value of <0.2 (EASE score) were retrieved. In Genedecks pathways (KEGG, MLPR, CST, GeneGlobe Pathway Central, Invitrogen, and Ingenuity) were retrieved called significant with a p-value <0.05 using the Set Distiller algorithm. KEGG pathways retrieved from DAVID were only listed when not called significant by Genedecks, or in case more genes were listed than in Genedecks. In case a pathway was called significant in one group (list) of genes, genes regulated on the same time-point in the “opposite” list, and also part of this pathway, were also retrieved from DAVID and GeneDecks.

From DAVID “Functional Annotation charts” transcription factors or genes involved in regulation of transcription were identified by gene-ontology analysis and uploaded as sub-list in GNCPro (free online software developed and maintained by SABiosciences Inc.) to establish relations between these genes (i.e. to build a network of transcription factors and regulators). Non-interacting genes were omitted from the displayed network.

Functional association between proteins encoded by differential expressed genes, ligands, and enzyme substrates/products linked to these proteins, were established using the (protein)-protein-chemical interaction web tool STITCH2 [14]. Relevant chemicals were added to gene lists and uploaded to in STITCH2 to establish associations. Associations with a confidence score of ≥ 0.4 (medium level) were selected from output files and displayed. In Additional file 1: Table S1 the type and confidence level of each association is listed in a separate sheet STITCH interactions.

Based on bioinformatics analysis (see above) a set of regulated genes was selected from 8, 4 and 2 hours micro array comparisons. Each individual gene/protein was loaded into STITCH2 to find associations with chemical compounds scoring a high confidence level (>0.7). For each gene/protein chemical compounds were selected meeting the following criteria: i) direct interaction with the protein, ii) present in the same (chemical) pathway, iii) capable to inhibit or activate the function of the protein, iii) involved in direct interference with transcription of the gene in question or relevant (groups of) genes transcribed/regulated by this gene/protein. In addition, for each of the selected genes/proteins-chemical combination relevant literature linked in the Comparative Toxicogenomics Database [15] (http://​ctdbase.​org/​) or in PubMed (NCBI) was examined.

The relative concentration of NFKBIA, IL1B, IL8, TIMP1, MMP1, and REG3A (alias PAP) mRNA in all RNA samples extracted from mucosal scrapings and isolated from IPEC-J2 monolayers was determined by real-time PCR. The gene-specific primers and specifications for these quantifications are recently described (REG3A [9], NFKBIA, TIMP1 and [16], IL1B and IL8 [17]). RT reactions were performed with Superscript III (Invitrogen) and random hexamer primers (pdN6) according to the manufacturer’s instructions using 250 ng of RNA template. The quantity of 18S rRNA in each RNA sample was determined using the above described RT reactions by real-time PCR [18] and used to normalize NFKBIA, IL1B, IL8, TIMP1, MMP1, and REG3A data. The quantity of 18S ribosomal RNA showed no essential differences among all individual RNA samples extracted from mucosa or from IPEC-J2 monolayers.

IPEC-J2 cells were seeded in 2 cm² tissue culture wells (M24 plate) and grown for 7 days at 37°C and 5% CO2 using 1:1 DMEM/Ham’s F10 1:1 medium (Gibco-BRL) supplemented with 5% FCS without antibiotics [10]. Confluent monolayer were washed twice with medium without FCS (hereafter denoted as medium) and incubated for 1 hour with this medium. Medium was discarded and a mixture of Salmonella bacteria and the chemical dissolved in medium was added. In a pilot experiment the multiplicity of infection (MOI) that did not induce visible (microscopic) damage to the cells (in the absence of chemical) was determined after an exposure times of 6 or 20 hours. Based on this pilot experiment a MOI of 1.0 and 0.1 was used for 6 and 20 hours of incubation, respectively. In a similar pilot experiment the effect of the highest concentration chemical was evaluated after 6 and 20 hours. For most chemicals the morphology of the IPEC-J2 cells was changed after 20 hours, but not after 6 hours. For these chemicals a maximum incubation time of 6 hours was used. After incubation total RNA from cells was extracted using Trizol (Invitrogen) according to the manufacturer’s instructions. RNA was treated with DNase as described [8] and further purified using the QIAamp MinElute Virus Spin Kit (Qiagen Cat no. 57704). The integrity of RNA was checked by analyzing an aliquot of 0.5 μg on a 1% (w/v) agarose gel before it was used as template in QRT-PCR reactions. The effect of all chemicals was tested in duplicate at 3 different concentrations. In each culture plate duplicate control wells containing no Salmonella (only with chemical), or containing no chemical (only with Salmonella), or without chemical and Salmonella (only medium), were incubated for the same period as was done for wells containing mixtures of chemicals and Salmonella. In case another solvent was needed (e.g. ethanol) to prepare a stock solution of chemicals, control wells without chemical were incubated with medium containing similar concentrations of solvents as were used for wells incubated with chemicals. The percent stimulation or inhibition in Salmonella and in mock challenged wells was calculated from the relative concentration of IL8 or NFKBIA mRNA (mean; n = 2) containing a specific concentration of chemical divided by the mRNA concentration measured in corresponding wells containing no chemical, and multiplied with 100% (e.g. IL8 in percent; {[IL8]+/ [IL8]-}× 100%).

For determination of cortisol concentrations in mucosal scrapings the Beckman Coulter ELISA was used (cat.no DSL-10-67100i). To extract cortisol from scrapings, 0.5 g was homogenized in 2 ml of PBS in a tube with screw-cap using a tissue homogenizer. Suspensions were incubated for 2.5 hours at 70°C under constant agitation. After cooling to room temperature 5 ml of ethyl ether was added and the mixture was shaken vigorously for 2 min, centrifuged for 10 min at 3000xg, frozen, and stored for 2 hours at −20°C. The ethyl ether fraction was transferred to a clean tube and the ethyl ether was evaporated under liquid nitrogen. The residue was dissolved in 0.5 ml PBS and the concentration cortisol in the extract was analyzed in the ELISA. A linear correlation (Regression Coefficient = 0.9824) was found between the amount of scrapping (g) used for extraction and the response in the ELISA, showing that the extraction method was reliable. The concentration cortisol (ng/g) in extracts was determined by extrapolation on a standard curve. For each mucosal scraping duplicate extracts were prepared and analyzed. A two-sided Grubbs’ test (p < 0.01) using the mean and standard deviation calculated over all determined values was performed to identify outliers; i.e. values that differed significantly from the population.

In an earlier study a limited number mRNA’s were found differentially expressed at 2, 4 and 8 hours when pools composed of RNA extracted from identical Salmonella treated segments of 4 SISP pigs were compared to pools prepared from mock treated loops of these 4 pigs (Figure 1B) [9]. In part the limited number of genes detected was due to the low complexity of the previous used home-made cDNA array. However, quantification of REG3A (alias PAP; a C-type lectin with antibacterial properties) mRNA expression in individual segments of all pigs (see Figure 2) indicated that responsiveness to Salmonella differed substantially for individual pigs, and, most likely, also accounted for this. This observed plasticity urged us to analyze IL8 and IL1B mRNA responses by Q-PCR in all segments dissected from SISP pigs 2, 3, and 4 (not sufficient material was available anymore from pig 1). IL8 and IL1B mRNA expression profiles (Figure 2) clearly showed that response-time and the type of response differed for all 3 pigs. In addition, quantification of TIMP1, MMP1, and NFKBIA in all segments confirmed this. The results of these Q-PCR analysis were largely in agreement with the below presented micro array data (see Table 1 and Additional file 1: Table S1).

The commercial Operon array platform was used to analyze RNA extracted from segments according to the scheme depicted in Figure 1B and C. In Additional file 1: Table S1 ratios of all genes found differentially expressed for more than 3-fold up- or down are listed. The genes discussed in this manuscript are tagged with letter d and can be easily sorted from this table. In Table 1 the total number of genes regulated in these comparisons, and genes that responded to Salmonella in the 8 hour segments of pig 3 (isogenic comparisons; Figure 1B) were presented. In contrast to pig 2 and 4, in pig 3 a limited number of genes (37) were found regulated 8 hours after challenge with Salmonella, despite this pig showed a strong IL8 response similar as observed in pig 4. Moreover, compared to pig 4 (at 2 hours), and pig 2 (at 4 hours) this pig showed a relatively faint IL1B response at 2 hours that quickly descended to a normal level within 4 hours. Also, this pig failed to produce an REG3A response at 8 hours. Nevertheless, 21 out of the 37 genes regulated in pig 3 were similar to genes regulated in pig 2 and/or 4; e.g. the highly up-regulated CYR61, a gene that responded to several pathogenic bacteria in epithelial cells [19]. This suggested that exposure to Salmonella had proceeded equally well in all three pigs, and prompted us to investigate the role of the 16 genes exclusively regulated in pig 3 at 8 hours. Using bioinformatics program STITCH2 relevant protein-protein associations were detected for 19 out of the 37 genes, including associations of proteins/genes with corticosteroid metabolism (depicted in Figure 3 left panel and genes listed below TIMP1 in Table 1). Among these 19 genes was the enzyme hydroxysteroid 11-beta dehydrogenase B2 (HSD11B2), which interconverts cortisol and cortisone. Determination of cortisol in mucosal scrapings detected a significant higher level in pig 3 at 4 hours than in scrapings of pig 2 and 4 (Figure 3, right panel), suggesting that down-regulation of HSD11B2 at 8 hours in pig 3 was related to the metabolism of cortisol. In addition, a strong down-regulation of TFF2 and RAC1 mRNA expression was observed in pig 3 at 8 hours. TFF2 plays a role in mucosal protection and repair in the intestine and was found up-regulated at sites of ulceration in various chronic inflammatory diseases. The secreted Salmonella-effector protein SopE interacts with host Rho GTPases like RAC1 in epithelial cells and macrophages and stimulates RAC1-mediated cytokine production and cytoskeletal reorganizations, i.e. forming of membrane ruffles that facilitate invasion of these cells with Salmonella[20, 21]. In pig 3, down-regulation of RAC1 could impair these processes and prevent or reduce the production of inflammatory cytokines like IL1B.

Sets of regulated genes in individual pigs were separately loaded into GeneDecks to assign genes to a specific pathway. Significant immunological pathways were selected from output files and presented in Additional file 2: Table S2 (sheet 8 h). The majority of pathways called for pig 2 and 4 reflected an inflammatory response induced by a bacterial infection, and were not called for pig 3. The genes of the MAPK and glucocorticoid receptor (GCR) signaling pathways regulated in pig 3 (down-regulation of RAC1 and up-regulation of HSPA1A, HSPA1B and NR4A1; see Table 1) showed little resemblance with the genes of these pathways regulated in pig 2 and 4. Together with the regulation of cortisol metabolism, the regulation of this set of unique genes in pig 3 may be part of “one” of the mechanisms by which the intestinal immune system tempers inflammation. However, cortisol-cortisone inter-conversion is a complex process. Therefore, further dedicated research is needed to observe similar responses in more than one pig to prove this hypothesis.

The observed plasticity in response was used to detect “potential” genes involved in early induction/tempering of Salmonella-induced inflammation. Microarray comparisons between identical treated segments of the three different responding pigs (depicted in Figure 1C) detected a broad set of genes differentially expressed between these individual pigs at 2 and 4 hours. The genes showing a ratio of more than 3-fold up or down in inter-animal comparisons are presented in Additional file 1: Table S1 (sheets 2 and 4 h). IL8 and IL1B are important cytokines for the immune response in the intestine. Therefore differentially expressed genes of these inter-animal comparisons were grouped based on their similarity in response to that of IL8 at 2 hours and IL1B at 4 hours. From these mutual comparisons it was determined in which pig the highest level of expression was observed for a gene from our list. In case the highest expression of a gene was observed in pig 3 or 4 at 2 hours (the pigs that showed a strong IL8 response; see Figure 2) the gene was assigned to a group called IL8-high. For all other genes detected at 2 hours the highest expression was observed in pig 2 (the pig that showed no IL8 response; see Figure 2). These genes were assigned to a group IL8-low. A similar grouping was performed at 4 hours, except that assignment of genes was based on the IL1B response of pigs (IL1B-high; highest expression in pig 2 or 4, IL1B-low; highest expression in pig 3). In Additional file 1: Table S1 the assignment of genes to a list may be selected by sorting with the “list headers”.

Sets of regulated genes of the IL8-high and -low lists and the IL1B-high and -low lists at 2 and 4 hours were separately loaded into DAVID and GeneDecks to assign genes to a specific pathway. Significant immunological and chemical pathways involved in immunological processes were selected from output files and presented in Additional file 2: Table S2 (sheet 2 and 4 h). Genes specifically regulated in one list, but also part of a pathway called significant for the opposite list, are presented between brackets. Such genes may be important factors and may provoke an opposite/different effect. In contrast to the IL8-high list, no pathways were called involved in recognition of bacterial components by Toll-like receptors (TLR’s) and extracellular and cytosolic pattern recognition receptors (PRR’s) in the IL8-low list. These pathways are responsible for the first production of cytokines like IL1B (e.g. in inflammasomes [22]) or IL8. The appearance of two T-cell pathways in the IL8-low list, representing pig 2, which showed a delayed IL8 and IL1B response, suggests that T-cell activation of residing (e.g. intraepithelial lymphocytes [IELs] and/or γδ T cells) or infiltrated T-cells preceded the IL8 and/or IL1B response. The most prominent pathways called in the IL8-high list were the ERK signaling and Glucocorticoid Receptor Signaling pathway. Glucocorticoids (GC’s) are potent anti-inflammatory agents. ERK signaling in response to binding of ligands like macrophage migration inhibitory factor (MIF) to its receptor CD74 may abrogate the immunosuppressive action of GCR’s. NRF2-mediated Oxidative Stress Response was called in the IL8-high list. The influx of granulocytes (neutrophils, basophiles, and eosinophils) attracted by IL8 may be related to regulation of genes in this pathway. IL1B mediated inhibition of retinoic X receptor (RXR) function was called for both IL8 lists. However, a complete different set of genes in the high list was mapped to this pathway than for the low list. This pathway regulates the metabolism and transport of cholesterol and fatty acid/lipids, for which it is known that they modulate the immune response in the intestine. The list-specific regulation of genes in the PPAR signaling pathway (IL8-high) and of genes in the peroxisome/fatty acid oxidation pathway (IL8-low) confirms that differences in regulation of lipid metabolism are closely linked to regulation of inflammatory responses in the intestine.

In the IL1B-high list at 4 hours regulation of Glucocorticoid Receptor and PPAR signaling pathway-genes continued. In the IL1B-low list, representing the pig [3] in which the IL1B and IL8 response decreased to normal levels, and in which the inflammatory response to Salmonella was tempered after 8 hours, two lipid associated pathways, i.e. the synthesis and degradation of ketone bodies (by-products of fatty acids metabolism) and the ”Arachidonic acid metabolism”, were called. In the opposite IL1B-high list, several different genes mapped to these pathways were regulated; suggesting lipid metabolism and Arachidonic acid related immune modulators like prostaglandins are involved in regulation of Salmonella-induced inflammation. Calling of the “Agranulocyte and Granulocyte Adhesion and Diapedesis” pathways at 4 hours indicates that activation and/or an influx of lymphocytes and monocytes from the blood to the epithelial layer occurred in pigs of the high-ILB list. Probably also related to this influx/activation was the mapping of genes to the “Phagosome” and “Antigen processing and presentation” pathways. Both pathways are involved in engulfing of bacteria and processing them to exposable antigen fragments. The inflammatory state in the two pigs of the IL1B-high list is probably best illustrated by mapping of genes to the “Acute Phase Response Signaling” and “p53 signaling” pathway. Genes in the latter pathway point to a response to damaged DNA, and to induction of apoptosis.

Completely different genes of the “ErbB signaling” pathway were regulated in the IL1B-low and -high lists. High list genes AREG-NRG1 induces AP1-mediated (JUN) transcription, and low list genes NCK1 and GAB1 are involved in cytoskeleton rearrangements in intestinal epithelial cells in response to bacterial adhesion or invasion. Also completely different ring-finger type E3 ubiquitin ligase complex-genes were regulated in the low and high list, suggesting that forming of different type of ubiquitination complexes results in degradation of different proteins. For instance, genes in the low list are part of the SCF Ubiquitin ligase E3 complex (BTRC-RBX1) that ubiquitinates NFKBIA and PER1. The latter protein affects hypoxia-responsive element mediated transcription.

A graphical network was built of all transcription factors/regulators present in gene-lists of both 2 and 4 hours inter-animal comparisons. Genes which showed an interaction to at least one other gene of the sub-list were displayed and non-interacting genes were omitted from the network (Figure 4). Genes central in these networks were selected as key factors, and a comprehensive data-mining was performed to find associations with chemical substances that have potential to influence the transcription of these genes, and/or the cellular processes regulated by these genes (see Methods). In Table 2 selected chemical substances tested, together with their corresponding “central” key factor genes, are listed. In addition, the genes HSD11B2, F3, NR4A1, and MMP1, which according to our STITCH network (Figure 3) may play a role in the cortisol/cortisone regulated tempering of inflammation in pig 3 at 8 hours, were also selected as key factor genes. Preferably, chemical substances were selected, which according to literature had the potential to influence expression of more than one key factor (maximal 3 key factor genes per chemical were listed in Table 2).

To perform a first evaluation whether the 20 selected chemicals have potential to influence Salmonella-induced gene expression in IPEC-J2 cells, we used IL8 and NFKBIA as reporter genes. The expression level of both these mRNA’s was found up-regulated in IPEC-J2 cells (see Figure 5; μM bars Salmonella versus mock) and in SISP loops (see Table 1) after challenge with Salmonella. All chemical were tested for 4 and 20 hours or for 3 and 6 hours incubation periods at three different concentrations chosen around a concentration that affected expression of the genes/proteins in question in cultured cells (thus, in the absence of Salmonella) in earlier studies (studies linked to data in the Comparative Toxicogenomics Database: see Methods). For all chemicals, the turbidity of the culture medium was increased after 6 or 20 hours, even at the highest concentration of chemical tested, indicating that the chemicals did not seriously affected the growth of Salmonella. In Figure 5 an example of the concentration dependent regulation of IL8 mRNA expression by Quercetin and Genistein is depicted for Salmonella and mock challenged wells. In case IPEC-J2 cells were not challenged with Salmonella, none of the 20 chemicals stimulated or inhibited the expression of IL8 and NFKBIA (results not shown). For Salmonella-challenged wells the stimulation/ inhibition index was calculated from the relative concentrations IL8 and NFKBIA mRNA measured in the presence of different concentration chemicals at the incubation times applied (see Methods). For 10 of the chemicals related to the key transcription factors/regulators (listed between brackets below) no significant stimulation/inhibition of Salmonella-induced IL8 or NFKBIA gene expression was observed (Palm oil [NFIL3], alpha-Tocopherol [PPARG], Tretinoin [PER1, TEF], Zymosan [NFKBIA], Aminolevulinic Acid [MMP1], beta-Carotene [CAT], Butyric Acid [IL8], Lycopene [NFKBIA], and Folic Acid [ETS2]). Concentrations of 10 μM or more Menadion (alias; Resveratrol [GADD45A]) resulted in the loss of IPEC-J2 cells from the monolayer already after 3 hours. The observed effect of Menadion on expression of both reporter genes was, therefore, doubtful (results not shown). For the 9 chemicals that significantly affected Salmonella-induced IL8 and NFKBIA mRNA expression the calculated inhibition/stimulation indexes were summarized in Table 2 and bar plots are presented in Additional file 3: Figure S1. Except for Chenodeoxycholic Acid, for which a consistent stimulation of both NFKBIA and IL8 mRNA was observed, most chemicals inhibited mRNA expression of IL8 and/or NFKBIA. However, stimulation and inhibition was observed for Nordihydroguaiaretic acid and Curcumin, depending on the concentration tested.