Host mechanisms involved in cattle Escherichia coli O157 shedding: a fundamental understanding for reducing foodborne pathogen in food animal production

The host mechanisms involved in Escherichia coli O157 super-shedding in cattle is largely unknown. In this study, the comparison of transcriptomes of intestinal tissues between super-shedders (SS) and cattle negative for E. coli O157 (NS) was performed, aiming to identify genes that are potentially associated with super-shedding. In total, 16,846 ± 639 (cecum) to 18,137 ± 696 (distal jejunum) were expressed throughout the intestine, with the expressed genes associated with immune functions more pronounced in the small intestine. In total, 351 differentially expressed (DE) genes were identified throughout the intestine between SS and NS, with 101 being up-regulated and 250 down-regulated in SS. Functional analysis revealed DE genes were involved in increased T-cell responses and cholesterol absorption in the distal jejunum and descending colon, and decreased B-cell maturation in the distal jejunum of SS. RNA-Seq based SNP discovery revealed that the mutations in seven DE genes involved in leukocyte activation and cholesterol transportation were associated with E. coli O157 shedding. Our findings suggest that T-cell responses and cholesterol metabolism in the intestinal tract may be associated with super-shedding phenomenon, and the SNPs in the DE genes are possibly associated with the observed gene expression difference between SS and NS.

Identification of transcriptomes and differentially expressed genes. The RNA-Seq data analysis was performed using the pipeline previously reported 11 . Briefly, adapter sequence removal and quality filtering of sequencing reads were performed by fastq-mcf 24 , and the remaining reads were mapped against the reference bovine genome UMD3.1 assembly 25 using the splice junction mapper, Tophat2 26 . HtSeq-count 27 was used to quantify the mapped reads to each known bovine gene. The number of mapped reads was then normalized into counts per million (cpm) to eliminate variation introduced by sequencing depth, using the formula: cpm = (Number of reads mapped to a gene) ÷ (total number of reads mapped to all annotated genes) × 10 6 . Any gene with a cpm higher than 1 was considered as being expressed 28 , and the genes expressed in more than 2 of animals in each group were further subjected to differential expression and functional analysis. Bioconductor package edgeR was used for differentially expressed (DE) gene identification between SS and NS in each intestinal tissue 29 , using negative binomially distributed RNA-Seq counts. A false discovery rate (FDR) of 0.05 was used as the cut-off to define DE genes. Log2 fold change of each DE gene was calculated using the equation: log2 fold change = log2(average cpm of SS/average cpm of NS). Log2 fold change threshold for DE genes were set to ≤−1 and ≥1 with negative values indicating down-regulated genes in SS and positive values indicating up-regulated in SS.
Functional analysis of transcriptome and differentially expressed genes. Functional analysis of RNA-Seq data was performed using several bioinformatics tools. The PANTHER (Protein Analysis Through Evolutionary Relationships) classification system 30 was used for gene ontology (GO) terms annotation and enrichment, with p-value 0.05 used as cut-off. Ingenuity Pathway Analysis ® (IPA, QIAGEN, Redwood City, CA, United States www.qiagen.com/ingenuity) was used for functional term enrichment and canonical pathways analysis for core transcriptomes and DE genes for each tissue, and a downstream effects analysis was performed to identify the biological influence of the DE genes. As the maximal number of genes that IPA can analyze is 8,000, the 8,000 genes in the core transcriptome with the highest cpm values were used for functional analysis for core transcriptome of each tissue. Based on the fold change of DE genes, the IPA also predicted if a certain enriched biological function was increased or decreased, and a z-score was used to indicate whether a predicted function was increased (z-score >2.0) or decreased (z-score <2.0). A similar z-score algorithm was used by IPA to predict if a canonical pathway was inhibited (z-score <−2.0) or activated (z-score >2.0).

RNA-Seq based SNP discovery.
To identify potential causal effects of DE genes, single nucleotide polymorphism (SNP) analysis was performed using RNA-Seq reads for these DE genes. To increase read coverage for SNP analysis, RNA-Seq reads of all tissues were combined for each animal and SNP calling was performed by VarScan2 31 , which is a mutation caller that uses a heuristic algorithm for sequence variant detection. For SNP calling, the minimum base quality of reads was 15, minimum read depth at a position to call a SNP was 8, minimal read number that supports an allele was 2, and the minimum variant allele frequency threshold was 0.1. The SNPs were kept for association analysis by the Fisher's exact test if the genotypes for all 10 animals were detected by VarScan2. Shedding of E. coli O157 was considered as a phenotype of SS, and the Fisher's exact test was used to analyze for associations between alleles and super-shedding with significance at 0.05.

Cholesterol quantitation.
To quantify the cholesterol concentration in distal jejunum and descending colon, a Cholesterol Quantitation Kit (Catalog Number: MAK043, Sigma-Aldrich, St. Luis, MO, USA) was used for lipid extraction and quantitation by fluorometric detection following manufacturer's instruction. Prior to cholesterol extraction, adipose tissues were removed manually using a scalpel, as the study of lipid storage was not the objective of this study, and the rest of the tissues were ground and mixed. Approximately 100 mg of tissue was used for cholesterol extraction and measurement according to the kit instructions, and the fluorescence intensity was detected at: λ ex = 535 and λ em = 587 nm. Two tailed sample t.test (significance level: 0.05) was performed to determine whether cholesterol content of the distal jejunum and descending colon differed between SS and NS.
Availability of sequencing data. The RNA-Seq data are available at NCBI Gene Expression Omnibus (GEO) database under accession number GSE85277.

Results
Transcriptome profiling of bovine intestinal tissues. The number of paired-end sequence reads for each tissue sample ranged from 185.0 million in the descending colon to 203.7 million in the cecum ( Supplementary Fig. S1A), with the number of the expressed genes detected in each tissue ranging from 16,846 ± 639 (mean ± standard deviation) in the cecum to 18,137 ± 696 (mean ± standard deviation) in the distal jejunum ( Supplementary Fig. S1B). For each tissue, transcripts with counts per million higher than 1 (cpm ≥ 1) in all 10 animals were defined as the core transcriptome for that tissue. The core transcriptomes consisted of 12,813 expressed genes in the duodenum, 12,552 expressed genes in the proximal jejunum, 12,905 expressed genes in the distal jejunum, 12,627 expressed genes in the cecum, 12,404 expressed genes in the spiral colon, and 12,216 expressed genes in the descending colon ( Supplementary Fig. S1C). The hierarchical cluster analysis of core transcriptomes revealed no clear separation between SS and NS in most of the gut regions, except for descending colon where SS (with the exception of SS 310) and NS were clustered into separate groups (Fig. 1). Principal component analysis (PCA) plots based on the whole transcriptome revealed that four of the NS were clustered closely for duodenum, cecum and spiral colon, three of NS were clustered closely for distal jejunum, and NS and SS were separated on the direction of first principal component (except for SS 310) for descending colon (Fig. 2). For SS animals, they were not closely grouped, with certain SS more closely related to NS in several tissues. For example, the duodenum transcriptome of SS 274 was similar to that of NS, while the spiral colon transcriptome of SS 287 and descending colon transcriptome of SS 310 were similar to those of NS.
Functional analysis of core transcriptomes of bovine gastrointestinal tissues. Functional analysis of the core transcriptome using PANTHER for biological process GO terms (Gene Ontology) enrichment showed consistent enriched GO terms for all intestinal tissues, including metabolic processes, cellular processes, biological regulation, localization and development processes (based on p-values) ( Supplementary Fig. S2). Further functional analysis using IPA revealed 16 enriched biological process categories (Fig. 3A). Although similar functional categories were identified for all intestinal tissues, the likelihood of association (indicated by Fisher's exact test p-value calculated by IPA) between functional categories and the transcriptome of a tissue varied, rendering the tissues of small and large intestines clustered separately (Fig. 3A). Immune functions such as cell mediated immune responses, haematological system development and function, humoral immune responses and lymphoid tissue structure and development were more associated with the core transcriptome of small intestinal tissues than with those of large intestinal tissues, particularly the distal jejunum (Fig. 3A). Functional analysis by the Panther Classification System also showed that more transcripts (t.test, p-value < 0.05) in the core transcriptome of small intestinal tissues were associated with immune system processes: 562 transcripts in the duodenum, 574 transcripts in the proximal jejunum, 579 transcripts in the distal jejunum, 545 transcripts in the cecum, 497 genes in the spiral colon, and 514 transcripts in the descending colon.
Identification of tissue dependent differentially expressed genes between super-shedders and non-shedders. The number of differentially expressed (DE) genes ranged from 1 (in the proximal jejunum) to 248 in the distal jejunum (Fig. 4A) between SS and NS, and the Log2 (fold change) of DE genes ranged from 8.9 for APOB in the distal colon to −7.9 for CPS1 in the spiral colon (Supplementary Table S1) of SS. In total, 101 genes were up-regulated and 250 were down-regulated when the transcriptomes of gut tissues of SS to NS were compared ( Fig. 4A and B). Functional analysis of differentially expressed genes. Functional analysis using PANTHER revealed that the GO terms were enriched for distal jejunum and descending colon DE genes, but not for the duodenum, proximal jejunum, cecum or spiral colon DE genes (p-value > 0.05). Similarly, the IPA functional terms were enriched for DE genes identified from distal jejunum, cecum and descending colon, but not for those from duodenum, proximal jejunum, or the spiral colon. For 67 up-regulated DE genes in distal jejunum, 33 enriched GO terms were identified within 11 functional classes based on the hierarchical structure of GO term (Supplementary Table S2). The most specific GO term in each enriched class includes "regulation of interferon-gamma production", "positive regulation of inflammatory response" and "immune response-activating cell surface receptor signalling pathway" (Table 1). Among these 11 GO terms, 10 of them were associated with immune functions (Table 1), which involved 22 DE genes including five genes encoding cytokines (CCL2, CXCL10, CXCL9, IL1A and TNFSF10) and three genes encoding cytokine receptors (IL18R1, IL1RL1 and CCR9) ( Table 2). According to IPA functional analysis of DE genes of distal jejunum, five up-regulated genes (F3, CCL2, CCR9, GPR132, S1PR2) were associated with increased leukocyte migration (z-score ≥ 2.0) ( Fig. 3B and Table 3), while five down-regulated DE genes (BCL6, BLNK, CD79A, CD79B, EBF1) were involved in B-cell signalling pathway (z-score ≤ −2.0) ( Table 3, Supplementary Fig. S3). For DE genes in descending colon, IPA analysis showed four up-regulated genes (APOA1, CD36, GP2 and GPAM) were associated with increased proliferation of T-cell function (z-score ≥ 2.0) in SS ( Fig. 3B and Table 3).
In addition to immune functions, IPA pathway analysis of DE genes in distal jejunum showed inhibited LXR/ RXR activation pathway (z-score ≤ −2.0) which is associated with increased cholesterol absorption in SS. Six genes were associated with this pathway, and two genes (APOE, ARG2) were down-regulated, while four genes (CCL2, IL1A, IL1RL1, PTGS2) were up-regulated (Table 3). Similarly, for the DE genes in the descending colon, five enriched GO terms were also associated with cholesterol transportation, including "cholesterol import", "sterol import" and "cholesterol transport" (Table 1). Four DE genes (APOB, APOA1, CD36 and AGPAT9) associated with these GO terms were up-regulated (Table 1).  Table S3) were then subjected to SNP identification analysis. In total, 1,575 SNPs were identified based on RNA-Seq dataset of these DE genes, and 1,477 of them have been previously reported in the dbSNP database 32 . Among those SNPs, 67 were three prime UTR variants, 9 were five prime UTR variants, 1,283 were intronic variants, 41 were missense variants, 9 were splice region variants (3 missenses variants), (C) down-regulated differentially expressed genes in super-shedders. Scaled log2(counts per million) was indicated by red and blue colors, and red indicates higher expression level, while blue colour indicates lower expression level. The column labels indicate super-shedders and non-shedders: NS, non-shedders; SS, supershedders. The row labels indicate the tissue where differentially expressed genes were identified. DU, duodenum; PJ, proximal jejunum; DJ, distal jejunum; CE, cecum; SC, spiral colon; DC, descending colon. and 71 were synonymous variants. The association analysis (Fisher's exact test) showed that 33 SNPs (Table 4) in seven genes (p-value < 0.05) including BATF2, THEMIS, ITK, BLNK, IL18R1, EBF1, and APOA1 were associated with super-shedding (Fig. 5A). As indicated by GO term enrichment and IPA pathway analysis, BATF2, THEMIS, ITK, BLNK, IL18R1 and EBF1 were related to leukocyte activation and APOA1 was associated with cholesterol transportation. Of these 33 SNPs, rs42084078 in BLNK and rs109755291 in IL18R1 were missense variants and rs384985356 in APOA1 was a synonymous variant, while the others were intronic variants. Furthermore, rs42084078 (A/G) causes an amino acid difference (A/V) at position 101 of BLNK (Accession number: NP_001039519.1), and rs109755291 (C/T) also causes an A/V difference at position 531 of IL18R1 (Accession number: XP_005212515.1).

Cholesterol quantitation in distal jejunum and descending colon.
To further verify the identified potential cholesterol metabolism difference in distal jejunum and in descending colon between SS and NS, the cholesterol concentrations in these two tissues were evaluated. The cholesterol content ranged from 0.57 to 1.15 mg/g in the distal jejunum, and 0.26 to 0.68 mg/g in the descending colon. The cholesterol content was higher in small intestinal tissue than in large intestinal tissue (t-test, p-value < 0.05); however, there was no statistical difference between SS and NS in both tissues (p-values were 0.89 and 0.48 for descending colon and distal jejunum, respectively) (Fig. 5B).

Discussion
To our knowledge, this study is the first to investigate gene expression (transcriptome) throughout the whole gastrointestinal tract of beef cattle with different E. coli O157 shedding phenotypes. Although the rectal-anal junction has been suggested to be the primary E. coli O157 colonisation site, this pathogen has also been isolated from other regions of the digestive tract 33 , suggesting that the environment throughout the gastrointestinal tract could impact on the colonisation of this pathogen which can attribute to the fecal shedding capacity of the animals. Therefore, the information obtained from the genome wide transcriptome can help us understand gut physiology and biology of bovine intestinal tract at a molecular level and how it may differ between SS and NS, which could partially explain the super-shedding phenomenon. To better assess such information, the whole intestinal tissue (epithelium and tissues underneath) for each region was used for transcriptome profiling as the tissues underlying the epithelium (such as the lymphoid structures) may also play a role in super-shedding phenomena 16 . Similar numbers of expressed genes were detected in all gut tissues, which was not surprising due to their physiological nature as the digestive tract. When the whole transcriptomes were compared, most of tissues (especially small intestinal tissues) showed no segregation between NS and SS, suggesting that the variation in E. coli O157 shedding status was not the key factor to influence the overall host gut transcriptomes. Indeed, these animals were healthy and no significant difference was observed in terms of their growth. The association of immune functions with small intestine regions, especially for distal jejunum, was in agreement with a previous report that jejunum also plays a role in mucosal immunity because Peyer's patches (PP) are present in this region 34 , and ileal PP extend from the distal jejunum to ileocecal valve 34 . Higher level of activity of immune functions in the small   intestine over the large intestine in beef steers suggests potential lower chance for E. coli O157 to colonize in small intestine. Thus, the colonisation of E. coli O157 in large intestine of cattle, especially in the RAJ 16 , may be a result from lower host immune activity of these regions. Further validation is needed to measure whether lymphoid structures in small intestine can contribute to the tropism of E. coli O157. The separation of transcriptome in cecum and descending colon between NS and SS further suggests that difference in hindgut physiological environment regulated by gene expression profiles may contribute to the difference in E. coli O157 shedding. Such findings were also in agreement with previous report that the colonisation of E. coli O157 was prevalent in lower digestive tracts of challenged yearling cattle, including cecum and colon 35 . It is possible that the environment of hindgut of SS is more favourable for E. coli O157 to survive, proliferate and colonise. Future studies to quantify the E. coli O157 population and measure the physiological parameters as well as immunological features in each region of the gut is necessary to identify alternation in which host functions through the gut can regulate E. coli O157 colonisation and the high level faecal shedding of E. coli O157.
Although there were no significant differences in core transcriptomes among the small intestinal regions (indicated by PCA analysis), DE genes were detected for each gut region between SS and NS. In the small intestinal region, lower number of DE genes were identified in duodenum and proximal jejunum compared with distal jejunum, suggesting that the gut environment in anterior part of small intestine is similar between SS and NS. The highest number of DE genes (248, including 181 down-regulated and 67 up-regulated) was detected in distal jejunum with the up-regulated DE genes F3, GPR132, CCR9, CXCL9 and CXCL10 involved in several GO terms including "cytokine-mediated signaling pathway", "positive regulation of intracellular signal transduction", "defense response". Previous findings have revealed that these genes are directly involved in T-cell related functions. The F3 (also called TF, log2 fold change: 1.3) product was reported to play a regulatory role in recruiting leukocytes in intestine of mice 36 , whereas the GPR132 (also called G2A, log2 fold change: 1.3) protein was  Table 4. SNPs which showed association with super-shedding phenomena. *Allele frequency in non-shedders, total frequency of two alleles are 100%. **Allele frequency in super-shedders, total frequency of two alleles are 100%.
Scientific RepoRts | 7: 7630 | DOI:10.1038/s41598-017-06737-4 reported to enhance chemotaxis of T-cells in vitro 37 . Similarly, the requirement of CCR9 (log2 fold change: 1.2) for migration of T helper 17 cells (Th17) to the small intestine was demonstrated using mice 38 , and CXCL9 (log2 fold change: 3.2) and CXCL10 (log2 fold change: 2.5) can induce chemotaxis of T-helper cells 39 . In addition, the cytokine IL1A and two cell-membrane receptor genes, IL18R1 and IL1RL1 were upregulated and were involved in GO terms including "cytokine-mediated signaling pathway" and "leukocyte activation" in distal jejunum SS. IL1A (log2 fold change: 1.6) protein plays a key role in the differentiation of Th17 cells 40 ; IL18R1 (log2 fold change: 1.3) was suggested to promote cell-mediated responses 41 ; and IL1RL1 (log2 fold change: 1.4) is involved in activation of T-helper cells 41,42 . Up-regulation of these genes suggests a potential increase in T-cell migration, differentiation and proliferation in the distal jejunum of SS (Fig. 6A). For descending colon, the functional analysis of the up-regulated DE genes also suggests a trend of increased proliferation of T-cell (z-score ≥ 2.0) in SS. The up-regulation of CD36 (log2 fold change: 4.1) was reported to increase T-cell proliferation in vitro 43 ; GPAM, also known as GPAT-1, (log2 fold change: 2.4) was reported to increase T-cell proliferation in mice 44 . These findings are in agreement with previous findings that increased expression of genes involved in the proliferation of T-cells in the rectum of E. coli O157 challenged cattle 45 . However, different from our previous transcriptomic analysis of rectal tissues which suggested a potential decrease in both humoral and cell-mediated immune functions in SS 11 , our current study indicates potentially enhanced T-cell migration and proliferation at the distal jejunum and descending colon of SS. In the gut associated lymphoid system, T-cells need to migrate from thymus to the gut epithelium to become intraepithelial T-cells (IETs), which are mature and activated cells serve as front line to protect the intestinal epithelium 46 . Compared with NS, the gut transcriptomes of SS suggest a higher level of T-cell migration in locations anterior to RAJ, while lower T-cell migration and quantity at RAJ 11 , suggesting a potential dysregulation of migration of T-cells in SS cattle which may result in reduced IETs accumulation at RAJ, leaving it more vulnerable to E. coli O157 colonisation.
In addition, the functional analysis of down-regulated DE genes in distal jejunum suggest association between genes including CD79A, CD79B, BLNK, BCL6 and EBF1 and decreased B-cell signalling. CD79A (log2 fold change: −2.2) and CD79B (log2 fold change: −4.3) are components of B-cell receptor (BCR) that initiates the B-cell signalling pathway 47 . Upon BCR interaction with antigens, BLNK (log2 fold change: −2.5) can be phosphorylated, leading to downstream signal transduction 48 , such as the mediation of transcription by BCL6 and EBF1, two transcription factors essential to B-cell maturation 49,50 . E. coli O157 mediated activities, such as Shiga-toxin production, were reported to suppress lymphocyte responses in cattle 51 , which may partially explain why the transcriptome analysis suggested inhibited B-cell signalling pathway in distal jejunum and the trend of reduction in lymphocyte activation and migration in cecum of SS.
Besides the potential involvement of immune functions, the pathway analysis and GO term enrichment of DE revealed inhibited LXR/RXR pathway and altered cholesterol transportation in distal jejunum and descending colon of SS. It has been reported that activation of LXR/RXR pathway in intestinal tract leads to decreased cholesterol absorption/synthesis and increased excretion of cholesterol in faeces 52 . Thus, inhibition of LXR/RXR pathway could lead to increased cholesterol absorption in the distal jejunum of SS. In addition, the up-regulated DE genes in the descending colon of SS, including APOB (log2 fold change: 8.9), APOA1 (log2 fold change: 6.0), CD36 (log2 fold change: 4.1) and AGPAT9 (log2 fold change: 3.2) also suggest altered cholesterol transportation in the SS. APOB encodes a key component of chylomicrons and low density lipoproteins which are responsible for cholesterol transportation between luminal content and intestinal tissues 53 . APOA1 was reported to regulate cholesterol transportation between cell membrane 54 and CD36 can facilitate cholesterol absorption in the intestinal tract 55 . Moreover, AGPAT9 was reported to be involved in cholesterol metabolism, because the deficiency of AGPAT9 (in mice) led to dysregulation in cholesterol metabolism with higher free cholesterol and cholesteryl esters in plasma and liver 56 . Our observations that differences in the expression of genes involved in cholesterol transportation and immune functions in the gut between SS and NS are not likely due to coincidence, as lipid metabolism plays a key role in immune system 57 . The LXR pathway was reported to interact with both the innate and adaptive immune systems, especially with macrophage functions 58,59 . In addition, Bensinger et al. reported that increased T-cell activation in mice was accompanied by inhibition of LXR pathway and promotion of cholesterol synthesis 58 . A possible mechanism is that cholesterol is required for cell membrane construction and cell proliferation, and thus increased sterol absorption/synthesis may act as a factor to lymphocytes proliferation 58 . Similar in the report by Bensinger et al. 58 , current findings also indicated potentially increased T-cell migration and proliferation accompanied by a trend of increased lipid absorption in the distal jejunum and descending colon of SS, suggesting that both alteration of host immune functions and lipid metabolism could influence E. coli O157 shedding. However, the increased absorption of cholesterol in the SS may have compensated the potentially increased utilization of cholesterol for T-cell proliferation in epithelial tissues of the distal jejunum and descending colon, which may explain why no differences were observed for cholesterol concentrations in distal jejunum and descending colon. Also, cholesterol may have been transported to the liver via mesenteric lymph upon absorption, as intestine is not an organ of cholesterol storage. In addition, post-transcriptional regulation may also influence the expression of proteins which are the major players in physiological functions. Therefore, although the gene expression suggested the increased cholesterol absorption, the measurable cholesterol level in the distal jejunum and descending colon have remained unchanged, due to immediate utilization, transportation or/and post-transcriptional regulations. Further validation is required to investigate the transcriptome of the liver, and the cholesterol levels in the liver and blood, as well as the post-transcriptional regulations in the intestine of NS and SS. It is noticeable that highly upregulated expression of FOLH1 (log2 change fold: 4.5 in RNA-Seq results; for qPCR measurement, the ΔCq was −1.04 in NS and −1.97 in SS, ΔCq was calculated by Cq FOLH1 -Cq β-actin , lower ΔCq indicated higher expression) occurred in descending colon of SS, suggesting that the expression level of FOLH1 was up-regulated in SS. Up-regulation of FOLH1 was observed in intestine of rat in a folate-deficient situation 60 . Because the environmental and dietary factors are the same in the NS and SS, it is possible that the potential folate-deficiency in descending colon was due to lower density of folic acid synthesizing commensal bacteria, including beneficial microorganisms such as Bifidobacterium species 61 . To identify how host-commensal interactions can influence E. coli O157 shedding requires further research, such as 16S rDNA amplicon sequencing for mucosa attached microbes and metagenomics studies.
Further SNPs discovery analysis identified 33 SNPs in seven DE genes associated with immune functions and cholesterol transportation. Among those SNPs, rs42084078 (in BLNK) and rs109755291 (in IL18R1) are non-synonymous, and allele A of rs42084078 and allele C of rs109755291 were only seen in SS. Thirty of the SNPs showing association with super-shedding were in the intron of genes BATF2, BLNK, IL18R1, ITK, THEMIS and EBF1. Although SNPs in the intron do not alter protein sequences, they are still important since introns are known to regulate gene expression, and intronic variation was reported to be associated with phenotypes 62 . For example, the DE gene EBF1 is a transcription factor essential for B-cell development from innate lymphoid cells 39 , and intronic SNPs in EBF1 was reported to be associated with autoimmune disease in human 63 . Therefore, we speculate that the detected genetic variations may influence their expression and lead to difference in B-cell signalling, T-cell responses and cholesterol absorption in the gastrointestinal tract between SS and NS. Future validation study is needed for the identified SNPs, and current results can be used as a reference for selection of candidate gene/SNPs for testing using larger populations as well as with more defined shedding status through a long-term monitoring approach.
It is known that shedding levels vary greatly in SS, ranging from 10 4 to 10 9 CFU/g of faeces 5 , with three types of SS being observed: non-persistent shedders (shedding <14 days), moderately persistent shedders (shedding ~30 days), and persistent shedders (shedding last several months) 4 . As described by Munns (2015) 22 , the SS used in this study had different faecal E. coli O157 counts before and upon their slaughter. For example, SS 310 was slaughtered on the 5th day of shedding and the faecal E. coli O157 in this animal upon slaughter was still greater than 10 4 CFU/g of faeces. Similarly, SS 274/SS 287 and SS 294/SS 299 were slaughtered on the 8th day and 11th day, respectively, and their shedding level was below 10 4 CFU/g of faeces at the slaughter. Different shedding patterns among these five SS may be the reason underlying the variation observed in gene expression through the gut of SS. However, since they were monitored for only 11 days at most, it is difficult to define the which types of shedder they belonged to. Therefore, future studies using the samples collected from long term monitored SS are needed to determine how the gut transcriptomes differ among different types of SS.

Conclusion
Our results suggest that (1) the bovine small intestine may have a more active immune system than the large intestine, making the distal colon towards rectum more prone to colonisation with E. coli O157; (2) the distal jejunum and descending colon of SS showed potentially higher levels of T-cell migration and proliferation, and cholesterol absorption, as well as inhibited B-cell maturation (Fig. 6A); (3) host genetic variation may be one of the mechanisms to affect the expression of genes involved in immune function and cholesterol absorption that may play a role in E. coli O157 shedding in SS, but genetic association analysis with larger number of animals is required to validate such speculation. Based on the transcriptome data, we speculate that host mechanisms, including genetic variation, alternation in immune functions and cholesterol metabolism, may be critical to super-shedding (Fig. 6B). Although this study is stronger at revealing the alteration of gene expression during high level shedding of E. coli O157 in cattle, potential host functions that could influence SS shedding were also identified. Further research is needed focusing on the associations among host mechanisms (such as genetics, immunity and cholesterol metabolism, as well as host-microbial interaction) and the E. coli O157 immunomodulatory effects, with larger populations of cattle and/or using a longitude approach. In addition, long-term monitoring faecal E. coli O157 for animals used in future studies to define NS as negative control and the type of SS (temporary, intermediate, and persistent) is critical to investigate more specific host effects that are associated with super-shedding. Importance. Escherichia coli O157:H7 (E. coli O157) is a foodborne pathogen which causes human illness as well as significant economic loss for food industry. Cattle are the primary reservoir for E. coli O157, and some animals can shed greater than 10 4 CFU/g faeces, who are considered as super shedders. To date, the host mechanisms involved in Escherichia coli O157 super-shedding in cattle is largely unknown. This study performed genome wide transcriptome analysis of the whole gastrointestinal tract of super-shedders in comparison with non-shedders and have revealed genes involved in immunity and cholesterol metabolism significantly differed between the two groups of animals and identified SNPs in seven of these DE genes. The identified SNPs could serve as potential genetic markers to select and breed cattle with the "suitable" gut environment allowing no or less E. coli O157 shedding in cattle. These findings provide valuable information in understanding of host-microbial interactions in super-shedding phenomena.