Development and Characterization of a Weaned Pig Model of Shiga Toxin–Producing E. coli-Induced Gastrointestinal Disease

Post-weaning enteropathies in swine caused by pathogenic E. coli, such as post-weaning diarrhea (PWD) or edema disease (ED), remain a significant problem for the swine industry. Reduction in the use of antibiotics over concerns of antibiotic resistance and public health concerns, necessitate the evaluation of effective antibiotic alternatives to prevent significant loss of livestock and/or reductions in swine growth performance. For this purpose, an appropriate piglet model of enterotoxigenic E. coli enteropathy is required. In this study, we attempted to induce clinical signs of post-weaning disease in a piglet model using a one-time acute or lower daily chronic dose of a Shiga toxin–producing and enterotoxigenic E. coli strain. The induced disease state was monitored by determining fecal shedding and colonization of the challenge strain, animal growth performance, cytokine levels, fecal calprotectin, histology, fecal metabolomics, and fecal microbiome shifts. The most informative analyses were colonization and shedding of the pathogen, serum cytokines, metabolomics, and targeted metagenomics to determine dysbiosis. Histopathological changes of the gastrointestinal (GI) tract and tight junction leakage as measured by fecal calprotectin concentrations were not observed. Chronic dosing was similar to the acute regimen suggesting that a high dose of pathogen, as used in many studies, may not be necessary. The piglet disease model presented here can be used to evaluate alternative PWD treatment options. Furthermore, this relatively mild disease model presented here may be informative for modeling human chronic gastrointestinal diseases, such as inflammatory bowel disease, which otherwise require invasive procedures for study. Importance Post-weaning diarrhea remains a significant problem in swine production and appropriate models of pathogenesis are needed to test alternative treatment options. In this study, we present an E. coli induced piglet model for post-weaning diarrhea, and also explore its translational potential as a model for human intestinal inflammation. Our study here presents two firsts to our knowledge. 1) The first simultaneous analysis of the intestinal microbiome and metabolome through fecal sampling of piglets challenged with Shiga toxin–producing E. coli. This is valuable given the limited metabolomics data from swine in various disease states. 2) A comparison of the clinical signs caused by a daily chronic vs one-time dosing regimen of E. coli. This comparison is key as infection by pathogenic E. coli in real-world settings likely occurs from chronic exposure to contaminated food, water, or environment rather than the highly concentrated dose of pathogen that is commonly given in the literature.


73
Post-weaning diarrhea (PWD) and edema disease (ED) following the weaning period in piglets 74 remain significant problems for the swine industry and can result in significant economic losses 75 (1-3). PWD is characterized by diarrhea which can lead to severe dehydration, emaciation, and 76 death. While ED of swine is characterized by submucosa edemas of the stomach and mesocolon 77 resulting in swelling of eyelids, forehead, and in some cases hemorrhagic gastroenteritis leading 78 to eventual death (2). Pathogenic Escherichia coli is the primary cause of these diseases in 79 swine, and the transitionary period of weaning leaves piglets susceptible to infection by 80 pathogenic strains of E. coli (3,4). While PWD and ED are generally caused by enterotoxigenic 81 E. coli (ETEC) and Shiga toxin-producing E. coli (STEC), respectively, they affect similarly 82 aged pigs and there can be considerable crossover between serotypes and associated virulence 83 factors. PWD ETEC are primarily associated with E. coli producing heat-stable and/or heat-84 labile enterotoxin, while ED STEC are associated with Shiga toxin, primarily the 2e subtype 85 (Stx2e), producing strains, which can be expressed with or without other enterotoxins (5). The 86 antibiotic colistin has been the classical treatment for pathogenic E. coli in swine, however given 87 concerns over antibiotic resistance, alternative treatment options should be explored (1). To anatomy, immune response, and ETEC clinical signs closely mimic that of humans (6). Swine 96 inoculated with ETEC experience sloughing of intestinal villi, increased crypt depths, and scours 97 (7). It has been shown that ETEC infections in weanling pigs can be caused by a single dose of 98 approximately 10 9 CFU (8,9). However, this high acute single dose most likely does not 99 accurately represent the real-world scenario of PWD or ED in which piglets are more likely 100 initially infected by chronic exposure to lower doses of E. coli as ETEC/STEC can be found in 101 contaminated feed, water, soil, and elsewhere in the barn environment (2). The objective of the 102 present study was to develop and characterize a Shiga toxin-producing E. coli induced weaned 103 swine model of PWD/ED. Given the E. coli strain used in this work encodes heat-labile 104 enterotoxin IIA (LT-IIA), heat-stable enterotoxin II (STIIB), as well Shiga toxin (Stx2e) it will 105 henceforth be referred to simply as an STEC strain, despite it technically classifying as both an 106 ETEC and STEC. We also sought to evaluate differences in dosing regimens, comparing a one-107 time high acute dose to a lower daily chronic dose of STEC. To our knowledge, this is the first 108 reported comparison on the effects of a one-time high acute dose vs a lower chronic daily dose in 109 an animal model. Furthermore, comparing the single-or repeated-dose models in swine is critical 110 to being able to evaluate the piglet model as potential model for human ETEC or STEC induced 111 enteropathies, particularly for the study of chronic inflammatory gastrointestinal disorders.

113
Growth performance 114 Pigs used as an experimental model for enteric enteropathy were challenged with a spontaneous 115 nalidixic acid-resistant mutant of Escherichia coli strain NCDC 62-57 (ATCC 23545) referred to 116 hereafter simply as E. coli 62-57nal in either a single acute high-titer dose (~10 9 CFU), or in a 117 series of daily lower-dose challenges (~10 7 -10 8 CFU). All pigs were held for two days prior to 118 the start of the trial and were asymptomatic for gastroenteritis. Additionally, pigs were not 119 colonized by organisms capable of forming colonies on MacConkey amended with 50 µg/ml 120 nalidixic acid (MacConkey+nal), and no endogenous phage infecting E. coli 62-57nal were 121 identified. Thirty-six presumptive coliform colonies from pooled fecal samples plated on 122 MacConkey agar (0 µg/ml Nal; three colonies per pen) were also tested by PCR for the presence 123 of Shiga toxin type 1 (Stx1), Shiga toxin type 2 (Stx2), heat-stable enterotoxin I (ST1), heat-124 stable enterotoxin II (ST2) and heat-labile toxin (LTI). All colonies were negative for Stx1,

125
Stx2, ST2 and LTI, but three colonies were positive for ST1. Presence of ST1 gene alone is not a 126 strong predictor of ability to cause disease (8, 10, 11) and pigs were asymptotic, so all animals 127 were retained in the study.

128
In general, pigs administered E. coli 62-57nal via both the acute and chronic dosing regimens 129 presented similar clinical signs with the majority of pens developing scours by day 2 and 130 continuing through day 6. Control pens had visibly soft feces on day 5 and 6 with a single 131 incident of scours on day 6, however the animals in control pens remained visibly healthy 132 throughout the trial period. Additionally, the control pen with the incidence of scours was culture 133 negative for the inoculated E. coli 62-57nal throughout the trial, so scours may have been 134 induced by stress or other native microbiota. There was no evidence of difference for overall 135 average daily gain (ADG), average daily feed intake (ADFI), and gain:feed (G:F) of the different 136 treatment groups (P > 0.184, Table 1). However, there were numerical differences between pigs 137 fed the treatments, suggesting that the modest number of replicates and the inherently high post 138 weaning variability in performance were responsible for the failure to detect significant 139 differences in growth performance. This lack of evidence for significant growth differences is 140 similar to previously reported results (12). Pigs administered the acute and chronic dose of E. 141 coli 62-57nal had a 54.7% and 14.9% reduction in ADG compared to the control pigs, 142 respectively (Table 1). The control group had the lowest ADFI among the three treatments with 143 acute and chronic dosing regimens increasing feed intake by 17.3% and 29.95%. These findings 144 are in agreement with previous work that showed a 24% decrease in control pigs ADFI 145 compared to the pigs inoculated with ETEC O149 on d 3 to d 6 (9). Madec et al. (13) had similar 146 results with a decrease in weight of weaned piglets inoculated with pathogenic E. coli expressing 147 K88 fimbriae from day 0 to day 2 which then recovered by day 9 of the trial. In this study, the 148 acute challenge group had the poorest mean G:F conversion with the control group having the 149 highest mean feed efficiency. Piglets experiencing PWD have been reported to exhibit reduced 150 weight gains (3,14), however statistically significant reductions in weight performance were not 151 observed, perhaps due to the relatively brief duration of the trial or small sample sizes.

152
Bacterial colonization and fecal shedding 153 The ability of E. coli 62-57nal to colonize the gastrointestinal tract of inoculated piglets was 154 determined by measuring colony-forming units recovered from intestinal mucosa, intestinal 155 luminal contents, and in feces. Inoculated strain counts adherent to the mucosal lining were found 156 to be variable, with ~50% of samples, ranging from 0.21 to 1.71 g of intestinal scraping, yielding 157 counts above the detection limit (5000 CFU/ml of tissue homogenate). Of the samples yielding 158 enumerable colonies, bacterial counts ranged from ~10 4 to ~10 7 CFU/g in the duodenum, jejunum, 159 ileum, cecum and colon ( Fig 1A). Bacterial counts in the cecal and colonic luminal contents, 160 ranging from 0.14 to 11.07g of digesta, were more reliably above the detection limit and ranged 161 from ~10 3 to 10 9 CFU/g, suggesting bacterial proliferation in the unattached population.

162
Acute and chronic treatments had higher prevalence of STEC in feces (~10 5 to 10 7 CFU/g) than 163 control pigs on all sampling days ( Fig 1B) Control pens sporadically shed E. coli 62-57-nal in the 164 feces at levels near the lower detection limit (500 CFU/g), likely reflecting low levels of pen cross-165 contamination. Pigs administered the acute STEC dose exhibited significantly higher fecal 166 shedding on day 1 (~10 7 CFU/g, P = 0.001) compared to the chronic dose, however there was no 167 statistically significant difference in fecal STEC counts between the acute and chronic treatments

Markers of inflammation and intestinal leakage 173
Infection-induced inflammation is mediated by increased levels of pro-inflammatory cytokines 174 (17). Interleukins 6 and 8 (IL-6 and IL-8) are useful biomarkers since they have been linked to 175 intestinal inflammation (18-21). On d 6 of the study, pigs challenged with STEC had increased (P 176 < 0.05) concentrations of serum IL-6 compared to control pigs (Fig 2). However, there was no 177 difference in IL-6 concentrations between acute and chronic treatments. Similar elevations of IL-   In the present study, there were no significant treatment effects on fecal calprotectin concentration.

193
This is the first study to our knowledge to test fecal calprotectin in pigs inoculated with E. coli, 194 and this indicates calprotectin may not be an informative biomarker in this model.  (22). Similarly, post-weaning anorexia in piglets has been shown to be 208 associated with reduced villus heights (29). Therefore, given a longer trial period and/or 209 malnourishment, blunting may have been eventually observed in our present model. Histology is 210 also only able to evaluate a tiny fraction of the intestinal tract, so lesions must be broadly 211 distributed throughout the tissue to be detectable by this method. Based on this data, histologic 212 analysis does not appear to be a useful method for evaluating this model.

Effects on the microbiome by 16S qPCR analysis 214
To observe any changes of the gut microbiota caused by our acute or chronic dosing treatments, 215 targeted 16S qPCR was performed for select bacterial groups on fecal samples collected from 216 pens at day -1, day 1, day 3, and day 6. Relative abundances obtained were consistent with 217 previous examinations of the piglet microbiome, showing a microbiome dominated by 218 Bacteroidetes and Firmicutes (30, 31). Overall, the bacterial groups tended to increase relative to 219 control and pre-treatment samples, likely due to natural microbiome succession. A summary of 220 these significant (P < 0.05) or marginally significant (P < 0.10) bacterial group changes at each 221 time point is shown in Table 2. Both acute and chronic STEC doses impacted relative quantities   (Fig 3A), the acute and chronic day 6 samples clearly cluster separately from their pre-treatment 256 samples, while the control samples did not separate. The stability of the control group indicates 257 the natural enzymatic, microbial, and structural maturation of the weaned piglet gut (32) was not 258 responsible for the observed shifts in the acute or chronic treatment groups.

259
As with the 16S qPCR approach, metabolomic comparison of post-treatment samples with their 260 respective pre-treatment samples was more informative when identifying significant changes in 261 individual metabolites. Volcano plots (p-value <0.10 and >2-fold change) were used to identify 262 metabolites that significantly changed following treatment (33) (Fig 4). A full list of metabolites 263 identified by volcano plot is provided in S1 between the chronic and control groups, and ten were common to both the chronic and acute 268 treatment groups (Fig 4). Changes in metabolites in the control group were presumed to be 269 associated with the normal development of the weaned piglet gastrointestinal tract.  Fecal metabolites that were significantly reduced in post-STEC treatment samples were 285 primarily fatty acid metabolites (Fig 4), including stearic acid, myristic acid, and arachidic acid.

286
The levels of lipid-soluble alpha-tocopherol (vitamin E) was also reduced in both treatment  (38, 39). In our current study a 296 reduction in the SCFA metabolites butyrate, alpha-ketoglutarate and fumaric acid were observed.

297
Our metabolomic findings indicate both the chronic and acute STEC treatments caused sufficient 298 dysbiosis to statistically distinguish pre-and post-treatment samples (Fig 3B) in large part due to 299 amino acid malabsorption and reduction in fatty acid metabolites (Fig 4), generating 300 metabolomic profiles resembling those of human inflammatory gastrointestinal diseases.

301
Whole genome sequencing 302 To better understand the gene content of E. coli 62-57nal that may contribute to its 303 pathogenicity, its genome was sequenced. The genome of E. coli 62-57nal was assembled into 304 378 contigs of greater than 200 bp totaling in 5.6 Mbp length and at an average 45-fold coverage.

305
The entire set of 378 contigs was submitted for sequence typing using SerotypeFinder v1.1 (42) 306 and confirmed to be O138 and H14 as reported previously (43). Analysis of the assembled  (Fig 1, 2). Furthermore, both treatments induced similar levels of dysbiosis as measured 355 by targeted 16S qPCR and untargeted metabolomics (Fig 3). These findings imply that high   with 2 barrows per pen that had solid concrete flooring and was equipped with a nipple waterer, 406 rubber mat, and feeder. Pigs were provided ab libitum access to water and feed; the diet was a 407 standard phase 1 nursery pig pelleted diet (S3 Table)

458
The minimum detection for IL-6 was 18.8 pg/ml and 62.5 pg/ml for IL-8. Assays were 459 conducted as outlined by the manufacturer. (65) and family/genus/species specific primers described previously (65,66). qPCR data is 500 reported as log10 of starting 16S copy number per 5 ng of DNA isolated. Specific primer sets 501 used were for Universal, Faecalibacterium, Streptococcus,E. coli,Fusobacterium,Firmicutes,502 Bacteroidetes, Lactobacillus,Ruminoccaceae,and Enterococcus. 503 Statistical analysis 504 Growth performance along with cytokine and intestinal bacterial population data were analyzed 505 using the PROC MIXED procedure in SAS 9.3 (SAS Inst. Inc., Cary, NC). The model fixed 506 effect was treatment with pen set as a random effect for growth performance, cytokine and 507 intestinal bacterial population data. Fecal samples were collected on a pen basis, therefore pen 508 was not included as a random effect. Calprotectin levels and fecal colony counts were analyzed 509 as repeated measures using the PROC GLIMMIX procedure. Treatment, day, and treatment × 510 day served as fixed effects. Day of collection also served as the repeated measure with pen as the 511 subject. Metabolite data was normalized to sum, mean-centered and divided by the standard 512 deviation of each variable, and analyzed for significant or trending metabolites between 513 treatments using MetaboAnalyst version 4.5 (33). Comparison of qPCR LOGSQ values was 514 carried out using JMP Version 13 (SAS Inst. Inc., Cary, NC.). qPCR treatment means were 515 compared pairwise on a per time point basis; many of the datasets did not pass the Shapiro-Wilk 516 test for normality and were of small sample size, therefore treatments were compared using the 517 nonparametric Wilcoxon Exact Test. qPCR data was also considered using multivariate methods 518 on a pre/post treatment basis using principal component analysis (PCA). Results were 519 considered significant at P ≤ 0.05 and marginally significant between P > 0.05 and P ≤ 0.10. Research. The authors would like to thank Amanda Blake and So Young Park for assistance with 526 qPCR. We would also like to thank Justin Leavitt, Jacob Chamblee, Jacob Lancaster, Lauren are greater than or less than relative to control or Pretreatment. Significant** Marginally Significant*