Potential immunosuppressive effects of Escherichia coli O157:H7 experimental infection on the bovine host

Background Enterohaemorrhagic Escherichia coli (EHEC), like E. coli O157:H7 are frequently detected in bovine faecal samples at slaughter. Cattle do not show clinical symptoms upon infection, but for humans the consequences after consuming contaminated beef can be severe. The immune response against EHEC in cattle cannot always clear the infection as persistent colonization and shedding in infected animals over a period of months often occurs. In previous infection trials, we observed a primary immune response after infection which was unable to protect cattle from re-infection. These results may reflect a suppression of certain immune pathways, making cattle more prone to persistent colonization after re-infection. To test this, RNA-Seq was used for transcriptome analysis of recto-anal junction tissue and ileal Peyer’s patches in nine Holstein-Friesian calves in response to a primary and secondary Escherichia coli O157:H7 infection with the Shiga toxin (Stx) negative NCTC12900 strain. Non-infected calves served as controls. Results In tissue of the recto-anal junction, only 15 genes were found to be significantly affected by a first infection compared to 1159 genes in the ileal Peyer’s patches. Whereas, re-infection significantly changed the expression of 10 and 17 genes in the recto-anal junction tissue and the Peyer’s patches, respectively. A significant downregulation of 69 immunostimulatory genes and a significant upregulation of seven immune suppressing genes was observed. Conclusions Although the recto-anal junction is a major site of colonization, this area does not seem to be modulated upon infection to the same extent as ileal Peyer’s patches as the changes in gene expression were remarkably higher in the ileal Peyer’s patches than in the recto-anal junction during a primary but not a secondary infection. We can conclude that the main effect on the transcriptome was immunosuppression by E. coli O157:H7 (Stx−) due to an upregulation of immune suppressive effects (7/12 genes) or a downregulation of immunostimulatory effects (69/94 genes) in the ileal Peyer’s patches. These data might indicate that a primary infection promotes a re-infection with EHEC by suppressing the immune function.

to have a chronic excretion over a period of at least 6-12 weeks was observed. Of the intermittent shedders, one animal was positive at the beginning of the study and was also excreting when sampled 11 months later. Furthermore, the shedding patterns showed that positive animals can shed different strains at different sample points [6]. Other studies showed that positive animals became culture negative within 2-3 months after the first testing [7]. The immune response of the animals against the EHEC strains, which has been investigated in few studies, might explain these excretion patterns. One study demonstrated that antibodies against the O157 lipopolysaccharide (LPS) and Shiga toxin-1 and -2 (Stx1, Stx2) frequently occur in bovine sera and colostrum upon experimental infection [8], but this response could not clear the infection, as infected animals secreted bacteria over a period of months [4]. Another study on two farms demonstrated that faecal excretion was not always correlated with E. coli secreted protein A (EspA), intimin and, translocated intimin receptor (Tir) specific serum antibody responses. In contrast, 87.5% of the animals showed a serum antibody response against Escherichia coli secreted protein B (EspB) at the same time that their faecal sample was positive for EHEC O157, O26 or, O111 or 6 weeks after a positive faecal sample. These antibodies persisted, even when shedding had ceased, until the animals were slaughtered, which was 2-8 months later, whereas EspA-specific antibodies disappeared within 2 months [9]. These results indicate that farm animals, which develop an immune response after infection, can become reinfected by different EHEC strains as evidenced by intermittent excretion which may reflect suppression of certain pathways of the immune system by the primary infection, making cattle more prone to persistent colonization by a subsequent infection.
In 2003, Naylor et al. [10] described the preference of E. coli O157:H7 for the terminal rectum up to the recto-anal junction (RAJ) as primary site for colonization. This site is characterized by a high density of lymphoid follicles. Predilection for epithelium above mucosa-associated tissue could be important for modulating the immune system. Indeed, EHEC O157 is capable of suppressing cellmediated immune responses in cattle by targeting lymphocytes via their Shiga toxins [8,11], but enterocytes do not have receptors for these toxins, suggesting that close contact with the immune system might be necessary. Here, the ileal Peyer's patches might play a role as they are of major importance for the mucosal immune responses in cattle [12]. In this study, a Stx negative strain was used for biosafety reasons. Nevertheless immunomodulating effects of other virulence factors of E. coli O157:H7 have been described. The H7 flagellin, bacterial LPS and type IV pilus have been shown to induce proinflammatory responses upon EHEC infection. On the contrary, it has been observed that EHEC as wells as EPEC strains could suppress NF-κB and MAPK activation as well as IκB degradation, [13] and could inhibit the production of proinflammatory cytokines IL-8 and IL-6, early in the infection by different LEE-and non-LEE encoded effectors (Tir, NleB, NleC, NleD, NleE, NleH1 and NleH2) [14]. Clearance of EHEC O157 is associated with an upregulation of Th-1 associated transcripts within the rectal mucosa, the principle site of colonization [10,15], suggesting that a cellular component of the adaptive immune response may be important in EHEC O157 control.
In this study we wanted to obtain insights in genes involved in an immunosuppressive effect of an E. coli O157:H7 Stx negative strain. In our experimental infections prolonged excretion was observed after a second infection with this strain. Transcriptome analysis of the ileal Peyer's patches and the RAJ from calves was performed using RNA-seq technology. Samples were taken from animals infected either once or twice which have never been in contact with E. coli O157:H7.

Bacterial strain
The E. coli O157:H7 strain NCTC 12900, a wellcharacterized Shiga-toxin negative E. coli O157:H7 strain of human origin with naladixic acid resistance [16] was used for experimental infections in calves. We used this Stx-negative strain for biosafety reasons. Bacteria were grown overnight in Luria Bertani broth (LB) with aeration (200 rpm) at 37°C, harvested by centrifugation (11 337 × g, 5 min), re-suspended in sterile phosphate-buffered saline (PBS) to a concentration of 10 10 CFU/10 ml and subsequently used for experimental infections.

Experimental infection of calves and sample collection
Nine 5-week-old Holstein-Friesian calves were randomly assigned to three groups (primary infection, re-infection and uninfected control; n =3), each reared in separate boxes in isolation units (Fig. 1). These animals were screened to be seronegative for intimin, EspA and EspB, as well negative for faecal shedding of E. coli O157:H7 and non-O157:H7. The animals were milk-fed from their arrival until the end of the experiment and allowed free access to hay, water and grain-based pellets. The milkuptake gradually decreased as the animals were able to digest more pellets, to allow a normal development of the gastro-intestinal tract.
Six animals were inoculated at the age of 6 weeks with 10 10 CFU during two consecutive days as previously described by Kieckens et al. [17]. Two weeks after infection, three animals were euthanized (Infection Group, IG). At the same time, the remaining three infected animals were re-infected and euthanized 2 weeks later at 10 weeks of age (Re-infection Group, RG). At that time, the three uninfected control animals (Control Group, C) of the same breed and age, were also euthanized ( Fig. 1). Euthanasia was carried out by captive bolt and samples from the ileal Peyers' patches and the RAJ were collected for RNA isolation to analyze the transcriptome profiles by RNA-Seq. Hereto, tissue samples of 1 cm 2 were cut and rinsed in sterile cold PBS (4°C) and immediately frozen in liquid nitrogen and stored at −80°C until RNA purification. All experimental and animal management procedures were undertaken in accordance to the requirements of the animal care and ethics committee of the Faculty of Veterinary Medicine, Ghent University, Belgium (EC2011/082).

RNA extraction and sequencing using RNA-seq
Total RNA was extracted from the tissues using the Qiagen RNeasy mini kit as described by the manufacturer. Briefly, the samples were ground to a fine powder under liquid nitrogen using a mortar and pestle and homogenized. Then, the lysate was further processed as instructed by Qiagen. The RNA purity was verified using NanoDrop Technology (Thermo Fisher Scientific, USA) and the RNA concentration was measured. High-quality RNA (260/280 nm ratio~2.0; RNA Rin# > 8.0) was processed using an Illumina TruSeq RNA sample prep kit following manufacturer's instruction (Illumina, San Diego, CA, USA). Individual RNA-Seq libraries were pooled based on their respective sample-specific 6-bp adaptors and sequenced at 50 bp/sequence read using an Illumina HiSeq2000 sequencer.

Data analysis and bioinformatics
The mean number of raw reads generated per sample in the study was 18,910,480.56 ± 5,610,677.59 (mean ± SD; n =18). SolexaQA was used for trimming and filtering using default parameters. The resultant reads with < 40 bp in length were discarded. After performing trimming and filtering, the final number of reads for the genome alignment was 16,042,336.72 ± 5,046,779.84 (mean ± SD).
The resultant quality reads were aligned to the bovine reference genome (UMD 3.1) using TopHat2 (v2.0.6) [19] using the following parameters: mismatches allowed: 2 bp; and max insertions: 3 bp and max deletions: 3 bp (using bowtie2 v2.0.2 with INDEL allowed). The SAM output files from the TopHat alignment, along with the GTF file from ENSEMBL bovine genebuild v67.0, were used in the Cuffdiff program in the Cufflink package (v2.0.2) to test for differential gene expression. Mapped reads were normalized based on the upper-quartile normalization method. Cuffdiff models the variance in fragment counts across replicates using the negative binomial distribution [20].
Results were considered significant for p <0.05. Differentially expressed genes identified in the transcriptome were further analyzed using GeneOntology (GO) analysis (https://github.com/tanghaibao/goatools) after FDR correction for FDR <0.1. A Fisher 's exact test was used for enrichment of certain GO terms. A multiple correction control (permutation to control false discovery rate, FDR) was implemented to set up the threshold to obtain the list of significantly over-represented GO-terms as previously described [21]. IPA (Ingenuity Pathway Analysis) and Path designer was used to visualize connections between differentially expressed genes.
The fold change was reported in the result section for every gene in brackets.

Results
Excretion of E. coli O157:H7 (Stx − ) Average excretion patterns of the infected animals are shown in (Fig. 2).Animals were negative before infection and the highest peak of bacteria in faeces was detected at day 4 post infection. Within 14 days after the primary infection, all animals of the primary infection group became negative. The second infection gave rise to a lower level of bacterial shedding but these animals were still shedding bacteria at euthanasia.

Transcriptome analysis of gut tissues
A total of 21,046 genes were detected at least once in one of the nine RAJ samples or in one of the nine ileal Peyer's patches samples. The number of genes expressed per sample was 18,753.22 ± 276.94 (Mean ± SD). The number of genes with mean hits ≥ 5.0 was 16,788 whereas the core transcriptome of the RAJ and the ileal Peyer's patches consisted of 16,948 and 17,197 genes, respectively. This core transcriptome incorporates genes represented by at least one sequence hit in each of the samples tested.
We used the transcript abundance classification as described by Li et al. [21]. Assuming 300,000 mRNA molecules per cell, approximately 70.68% of genes transcribed in the RAJ and 70.55% of genes transcribed in the ileal Peyer's patches can be classified into "very rare" with a relative abundance of ≤15 molecules per cell (Table 1), followed by "rare" (16-99 molecules per cell) at 16.18 and 17.29%, respectively. "Not expressed" genes (0 molecules per cell) were calculated as 11.17 and 10.18%, whereas "moderately abundant" genes (100-500 molecules per cell) accounted for 1.74 and 1.78%. "Abundant" genes were only for 0.23 and 0.21% part of the transcriptome.

Genes significantly influenced by E. coli O157:H7 experimental infections
In the RAJ, the primary site of E. coli O157:H7 colonization in cattle, fifteen genes were found to be significantly affected by a primary infection with E. coli O157:H7 whereas ten genes were affected after re-infection compared to the uninfected control group (false discovery rate FDR < 0.1). Only one gene (FABP2) appeared to be significantly impacted by both primary infection and re-infection with E. coli O157:H7 (Fig. 3). Three out of fifteen genes that were significantly affected during the primary infection could be linked to an immune function (KLRJ1, MARCO, CCL20); one was upregulated and two downregulated ( Table 2). In the ileal Peyer's patches 1159 genes were significantly influenced by a primary infection compared to the control group and only seventeen genes were significantly affected by the reinfection compared to the same control, indicating a larger effect on the transcriptome during primary infection compared to a re-infection. Seven genes were significantly different regulated after a primary as well as after a re-infection (Fig. 4). The function of 103 out of 1159 genes that were differently regulated

Regulatory gene networks
The IPA software was used to further examine the RNA-Seq dataset. Uploading the dataset for the RAJ after a primary infection, the database could assign two relationships between the significantly regulated genes, whereas for the re-infection four relationships could be detected in both cases linked to antimicrobial responses. When the data for ileal Peyer's patches was uploaded, IPA could identify three different networks that might play a role in EHEC infection, related to antimicrobial response (28 relationships) (Fig. 5), inflammatory response (69 relationships) (Fig. 6) and infectious disease (166 relationships) (Fig. 7).
Possible effect of E. coli O157:H7 experimental infection on the function of immune cells In order to gain insight in the impact of the differential mRNA expression, the data were arranged according to their effect on different immune cells in the specific tissues: lymphocytes (3.5.1), natural killer cells (3.5.2), monocytes and macrophages (3.5.3), dendritic cells (3.5.4), granulocytes (3.5.5) Only genes with the highest fold changes per cell type are reported. The gene expression results for RAJ and ileal Peyer's patches after primary infection are represented in Tables 3 and 4, respectively.

Lymphocytes
In the RAJ, a significant downregulation of Chemokine C-C motif ligand 20 (CCL20) (fold change = 0.07) was observed, which is strongly chemotactic for immature dendritic cells, and B-and T-lymphocytes [22].

Monocytes and macrophages
In the RAJ, macrophage receptor with collagenous structure (MARCO) was significantly downregulated (fold change = 9.20). This is a receptor which is part of the innate antimicrobial immune system binding both Gram-negative and Gram-positive bacteria via an extracellular, C-terminal, scavenger receptor cysteine-rich (SRCR) domain [32]. At the ileal Peyer's patches significant upregulation was seen of ectonucleotide pyrophosphatase/phosphodiesterase 7 (ENPP7), PDIA2, carboxypeptidase M (CPM) important for monocyte to macrophage differentiation [33], the inflammatory purinergic receptor ligand-gated ion channel, 7 (P2RX7) and egf-like module containing, mucin-like, hormone receptor-like 4 pseudogene (EMR4P), a member of the EGF-TM7 receptor gene family which is thought to be important for adhesion and migration of macrophages [34] with fold changes of 974.24, 21.28, 3.33, 3.24 and 2.64, respectively. Al these genes result in stimulation of the immune system on level of monocytes and macrophages. On the other hand, macrophage migration inhibitory factor (MIF) which is important for the acute immune response [35] was significantly downregulated as well as chemokine C-C motif receptor 1 (CCR1) playing a role in recruitment of leukocytes to the effector site [36] (fold change = 0.45 and 0.36, respectively). Furthermore, sialic acid binding Ig-like lectin 10 (SIGLEC10) which is a negative regulator of immune signaling by functioning as an inhibitory receptor [37] and B cell linker (BLNK), an adaptor molecule linked to the pathway activated by Bcell antigen receptor signals [38], were significantly downregulated, with fold changes of 0.01 and 0.26, respectively.

Dendritic cells (DC's)
No differential mRNA expression of genes indicating an effect on DC's was seen in the RAJ. In samples from the ileal Peyer's patches, a significant upregulation of PDIA2, toll-like receptor adaptor molecule 1 (TICAM1) (fold change = 2.97) and EMR4P was observed, whereas SIGLEC10 and MIF were significantly downregulated.

Discussion
It is well known that some pathogens have developed mechanisms to prolong their persistence in a host and can modulate the host immune response in different ways in order to increase their survival in the host [43]. This might be done by passive evasion of the immune surveillance for instance by altering expressed antigens, as Streptococcus pneumonia does. Another way to survive is actively modulating and interfering with regulatore networks that are part of the immune defence. This is done when EHEC and EPEC inject modulatory proteins into the host cell using T3SS [43][44][45]. Furthermore, the suppression of immune responses via TLR4 by uropathogenic E. coli (UPEC) leads to a decrease of IL-6 and IL-8 release [46]. This is congruent with our observations in several infection experiments: a primary infection could elicit an antibody response, but this primary immune response was unable to protect against an experimental re-infection with the same strain. On the contrary, whereas the primary infection led to faecal shedding during less than 2 weeks (<14 days; n = 32), a re-infection resulted in excretion for more than 4 weeks (>28 days; n = 21) ( Table 5; unpublished results).
To our knowledge, this is the first RNA-Seq study highlighting the effect of a primary infection and a reinfection with E. coli O157:H7 in cattle, leading to a better understanding of the transient and sometimes recurrent pattern of EHEC infections. In 2007, Li and Hovde described the differential expression of 49 genes in the RAJ after a primary infection with E. coli O157:H7 using cDNA microarray [47]. Here, we used the RNA-Seq technique which allowed us to determine effects on both RAJ and ileal Peyer's patches. We could identify an important difference in regulation of the transcriptome after an initial contact with the bacteria: the ileal Peyer's patches were more influenced by the infection compared to the RAJ, which is part of the primary site for colonization in cattle [10]. In this study we   PDIA2 + Helps to load antigenic peptides into MHC I molecules and is therefore important in antigen recognition and clearing. KLRJ1 + (see Table 3  were able to show suppression of immunity on different levels of the innate and adaptive immune response, indicating that E. coli O157:H7 can modulate this response in a wide variety of ways. The excretion patterns confirm our previous findings, as a first infection mounts a higher level of bacterial shedding during a shorter period of time compared to a re-infection. It might be mooted that there is a correlation between the higher level of infection and a more pronounced effect on the immune system. However, we hypothesize that the priming of the immune system by the first exposure is more important than the level of bacteria since a second infection with E. coli O157:H7 mostly results in a prolongued colonization, whereas for other pathogenic E. coli such as enterotoxigenic E. coli in pigs a complete protection occurs after a first infection. Our data suggests that an infection with E. coli O157:H7 (Stx − ) is capable of modulating the immune response causing dramatic decreases in CCL20. A similar trend was observed in a bacterial infection with Mycoplasma gallisepticum in chicken, where a down-regulation in mRNA expression of CCL20 as well as of IL-1, IL-8 and IL-12p40 genes was seen. These results indicate the importance of lymphocyte and monocyte chemotactic factors in development of disease, but also the fast occurrence of modulations of the host immune responses by bacteria [22].
Granulysin delivers granzymes into bacteria to kill diverse bacterial strains. In Escherichia coli, granzymes cleave electron transport chain complex I and oxidative stress defense proteins, generating reactive oxygen species (ROS) that rapidly kill bacteria [48]. An upregulation of granulysin is part of the adaptive immune response against bacterial infections.
A study of St John et al. [49] demonstrated downregulation of CXCL13 and CCL21 during an infection of draining lymph nodes by Salmonella typhimurium. The pathogen disrupts the lymph node architecture and cellular trafficking, which enhances its virulence and could serve as a mechanism of immune suppression used by pathogens that primarily target lymphoid tissue. In our study, we could not observe a significant downregulation of CXCL13 nor CCL21 altough the bacterium is in close contact to lymphoid dense tissue in the RAJ and the ileum.
We observed a downregulation of IL17REL, a gene which is important for a fast inflammatory response. The cells of the innate immune system are the first line of defense against pathogen and their cytokines govern the differentiation of T-helper cells. Their pattern-recognition receptors, which are not specific for any particular  epitope, allow them to respond to a wide variety of microbial invaders by producing cytokines that activate T-cells of the adaptive immune system. T-helper 17 cells produce IL17 and this is particularly important for immunity at epithelial and mucosal surfaces, as indicated by the pattern of expression of their chemokine receptors and effector cytokines. Several pathogens, like gram-positive Propionibacterium acnes and gram-negative Citrobacter rodentium, induce mainly Th17 responses [50]. Furthermore, Luo et al. [51] have shown that F4 fimbriae of ETEC can elicit an IL17 response in piglets, suggesting a role in protection of the host against ETEC infection. A downregulation of IL21R might play a role in the persistence of colonization during the primary E. coli O157:H7-infection and re-infection of the host, since the IL-21-IL-21R pathway is important in the development of immune responses, as abnormal signaling through the IL-21R/γc/JAK3/STAT3 pathway leads to defective humoral immune responses to both T-dependent and Tindependent antigens and impairs the establishment of long-lasting B-cell memory [52]. A bacterial infection can elicit IgM memory B-cells which requires T celldependent and IL-21R signaling. The study of Yates et al. [53] demonstrates that T cell-dependent IgM memory B cells can be elicited at high frequency and can play an important role in maintaining long-term immunity during bacterial infection.
In the RAJ, we could observe the downregulation of MARCO, a receptor that can bind Gram-negative bacteria and is only found on macrophages of the marginal zone of the spleen and lymph nodes [32]. Pinheiro da Silva et al. [54] have shown that E. coli are capable of hijacking inhibitory ITAMs leading to a decrease in MARCO-mediated phagocytosis. This observation is indicating a decrease in antibacterial protection of the host at the primary site of EHEC infection.
Due to biosafety reasons and limitations on housing facilities we performed this experiment using an E. coli O157:H7 (Stx − ) strain. Although this is not the exact same strain as the strains found during natural EHECinfections, this study may be seen as a first step to investigate the immunomodulating capacities of E. coli O157:H7 strains in cattle using RNA-Seq. We believe that this study is a valuable contribution to the current knowledge on immune suppression in the bovine host and this study is congruent with previous findings showing that also other bacterial factors apart from Stx can play a role in immune suppression.
A limitation of our study is that most information from gene function databases is derived from studies that are not performed in cattle but in human and mice, therefore we have to take into consideration speciesspecific differences which we unfortunately can not control. This study is important since it is the first study using the highly-accurate and sensitive RNA-Seq technique to study the effect of EHEC on the cattle immune responses. These insights are crucial to better understand EHEC colonization and shedding within herds and our data could contribute to effective measures to control EHEC colonization in ruminants, thereby reducing zoonotic food-borne infections in humans.

Conclusions
We can conclude that the main effect on the transcriptome was immune suppression by E. coli O157:H7 due to an upregulation of immune suppressive effects (7/12 genes) or a downregulation of immunostimulatory effects (69/94 genes). Furthermore, the changes in gene expression were remarkably higher in the ileal Peyer's patches (1159 genes) than in the RAJ (15 genes) during a primary infection. This effect was less obvious after the reinfection (17 and 10 genes, respectively). The data might indicate that a primary infection promotes a re-infection with EHEC by suppressing the immune function.