Serum proteome of dogs with chronic enteropathy

Abstract Background Chronic enteropathy (CE) is common in dogs and can occur with multiple etiologies including food‐responsive enteropathy (FRE) and idiopathic inflammatory bowel disease (IBD). Hypothesis/Objective To study the protein profile and pathway differences among dogs with FRE, IBD, and healthy controls using serum proteome analysis. Animals Nine CE dogs with signs of gastrointestinal disease and histologically confirmed chronic inflammatory enteropathy and 16 healthy controls. Methods A cross‐sectional study with cases recruited from 2 veterinary hospitals between May 2019 and November 2020 was performed. Serum samples were analyzed using mass spectrometry‐based proteomic techniques. Results Proteomic profiles showed marked variation in relative protein abundances. Forty‐five proteins were significantly (P ≤ .01) differentially expressed among the dogs with CE and controls with ≥2‐fold change in abundance. The fold change of dogs with IBD normalized to controls was more pronounced for the majority of proteins than that seen in the dogs with FRE normalized to control dogs. Proteins involving reactive oxygen species, cytokine activation, acute phase response signaling, and lipid metabolism were altered in dogs with CE. Conclusions and Clinical Importance Cytokine alterations, acute phase response signaling, and lipid metabolism are likely involved in pathogenesis of CE. Although there are insufficient current data to justify the use of proteomic biomarkers for assessment of CE in dogs, our study identifies potential candidates.

enteropathy (ARE), idiopathic inflammatory bowel disease (IBD), and non-responsive enteropathy (NRE). [1][2][3] Diagnosis of the underlying cause of CE can be challenging. Initial assessment revolves around exclusion of extra-gastrointestinal diseases and other primary gastrointestinal diseases such as infectious causes, parasitism, and neoplasia. Definitive diagnosis of the process leading to CE relies on response to dietary modification, and in dogs failing diet trials, histopathological evaluation of the gastrointestinal tract via surgical or endoscopic biopsies. 1,2 Yet, in some cases, differences in histological interpretations and interobserver variability among pathologists can be problematic. 1,4 Furthermore, histopathological findings often do not correlate with treatment response. 3 An extended amount of time and expense might be spent on treatment trials, and management, leading to owner fatigue and financial exhaustion. In recent years, minimally invasive diagnostic approaches using biomarkers have been explored in diagnosing, monitoring and assessing disease severity in CE. 1,3,5 Serological biomarkers investigated in dogs with CE include cobalamin, folate, C-reactive protein, perinuclear antineutrophilic cytoplasmic antibodies, citrulline, soluble receptor for advanced glycation end products and metabolite profiles. 1,3 Only a few serum biomarkers are commercially available for clinical use. In human gastroenterology, serum protein biomarkers are used to diagnose, monitor treatment response, and to differentiate between different forms of CEs. [6][7][8][9][10][11] The proteomic profile in dogs for diseases including infectious diseases, [12][13][14][15][16] neoplasia, 17 myxomatous mitral valve disease, 18 heartworm disease, 19 immune mediated polyarthritis, 20 chronic hepatitis, 21 chronic bronchitis, idiopathic pulmonary fibrosis, 22 meningoencephalitis, 23 cervical spondylomyelopathy, 24 chronic kidney disease, 25 and obesity is reported. 26 Studies investigating proteomics in gastrointestinal disease remain sparse in the veterinary literature. Given the extensive research and use of proteomics in human gastroenterology, serum proteomics might provide an alternative minimally invasive method for diagnosing, assessing disease severity, monitoring treatment response, and act as a prognostic indicator in dogs with CE. The aim of our study was to identify differences in serum proteins in dogs with CE compared to healthy control dogs, to compare differences in serum proteins between dogs with FRE and IBD, and to further understand the pathophysiology of CE in dogs. weight loss, or a combination of these clinical signs for at least 3 weeks were recruited. The signalment, history, current diet, physical examination findings, diagnostic test results, treatment, and comorbidities of all dogs were retrospectively reviewed. Complete blood count, serum biochemical analyses, and urinalysis were performed in all dogs.

| Sample collection
Cobalamin concentrations and C-reactive protein concentrations were also performed in some dogs. Dogs with CE were classified as FRE, IBD, and NRE based on their treatment response. Dogs that responded to dietary therapy and did not require additional therapeutic trials were defined as FRE. Dogs that did not respond to dietary therapy but responded to immunosuppressant therapy were defined as IBD. Dogs that did not respond to dietary modification, antibiotic therapy, and immunosuppressant treatment were defined as NRE.
Apparently healthy control dogs were recruited from among staff-or client-owned dogs presented for annual health checks and vaccination. They were identified as "healthy" by history and physical examination. An attempt was made to match the cases by age, sex, and body condition score as closely as possible. Blood samples were collected via jugular venipuncture and serum was separated by centrifugation. The serum was then collected into a serum tube and stored at À20 C before analysis.

| Protein sample preparation
Total protein concentration was assessed using the 2D Quant kit (Cytiva, Massachusetts, USA) as per manufacturer instructions. Preparation of samples was as described. 28 Briefly, samples of 100 μg total protein were mixed in 50 μL AMBIC buffer (50 mM Ammoniumbicarbonate, 10 mM DTT, 2 M urea at pH 8) and digested with trypsin at 25 C for 16 hours in a 1:100 enzyme-to-protein ratio based on the calculated serum protein concentration. Digestion was halted by acidification. Each sample was then dried and reconstituted in 50 μL of 0.1% formic acid and desalted using C18 stage tips (Thermo Scientific, Illinois, USA) according to the manufacturer's recommendations except that the elution buffer consisted of 80% CH 3 CN, 0.1% formic acid.

| Mass spectrometry
Digested peptides were reconstituted in 10 μL 0.1% formic acid and separated by nano-LC using an Ultimate 3000 HPLC and autosampler (Dionex, Amsterdam, Netherlands) and followed methods similar to those described previously. 28 Briefly, the sample, 1.7 μL from 10 μL, was loaded onto a micro C18 pre-column (300 μm Â 5 mm, Dionex) with

| Protein characterization
Protein dataset-peak lists were generated from raw files using Mascot comparisons. An adjusted false discovery rate (FDR), with significance set to P < .05 and at least 2 peptides per protein identifications were accepted as valid.
For bioinformatics analysis, Protein ANalysis THrough Evolutionary Relationships (PANTHER) was used to investigate molecular function, biological processes, and protein classes. 29 Only differentially expressed proteins between dogs with CE and controls with fold change ≥2 were included in this analysis.
Causal networks were also analyzed with Ingenuity Pathway Analysis (Qiagen) to show the interactions between proteins and physical pathways. Graphical networks for these proteins were constructed based on their connectivity algorithms.

| Statistical analysis
Descriptive statistics of dog information including age, weight, CCECAI, cobalamin concentrations, and C-reactive protein levels were generated using R software.
The top protein candidates are identified using filters based on group means, statistically significant at P ≤ .01, significance >2-fold change, FDR, and 95% confidence intervals. Significant changes in expression of protein fold change were analyzed using Fisher's exact test. Fold change ratios were calculated for dogs with CE, FRE and IRE, and normalized to control abundances. The dog was treated with prednisolone 1 day after endoscopic biopsies were obtained because of concerns for rapid deterioration and inappetence. A dietary trial with a novel protein diet was commenced 3 weeks later when the dog's appetite returned to normal. This dog was classified as having IBD. The other dog with PLE had failed 2 dietary trials including a prescription hydrolyzed protein diet, metronidazole therapy, and had poor response to prednisolone treatment at referral. After endoscopy, the dog was then treated with chlorambucil while prednisolone was continued and the dog had a good response.
Seven dogs with CE had been treated with metronidazole at a dosage of 10-23 mg/kg q12h with minimal responses, thus none of the studied dogs were categorized as ARE.
Cyclosporine was administered at an initial dosage of 6-8.8 mg/kg once daily or in divided doses and chlorambucil was administered at an initial dosage of 0.1 mg/kg q24h. Immunosuppressants were not used or tapered at consistent doses or frequencies. Four dogs, including 1 dog with PLE, were being treated with prednisolone at the time of sample collection. The CCECAI score ranged from 4 to 16, with a median of 7.
Complete blood count, serum biochemical analyses, urinalysis, and abdominal ultrasound were performed in all dogs with CE. The median C-reactive protein, serum cobalamin, resting or post ACTH cortisol and pre-prandial bile acids concentrations are shown in Table 1. Two dogs with CE had abnormally high C-reactive protein concentrations (RI < 10.7 mg/L). Serum cobalamin concentrations were determined in all but 1 of the dogs with CE before cobalamin supplementation. One dog did not have initial cobalamin concentration determined because cobalamin supplementation had been commenced by the referring veterinarian before referral. Two dogs had cobalamin concentrations below this reference interval. Resting cortisol or ACTH stimulation test was available in all but 1 dog with CE, this dog was receiving prednisolone at the time of referral (Table 1). All dogs with CE were up to date with anthelmintics.  Table 1. There were no significant differences in age (P = .98), sex (P = .05) and body condition score (P = .12) between dogs with CE and clinically healthy dogs.
The median protein concentrations of samples in CE and controls measured using the 2D Quant kit were 109 mg/mL (range 53.9-138 mg/mL) and 95.8 mg/mL (range 51.4-145 mg/mL) respectively. There were no significant differences in median protein concentrations between both groups (P = .52).
T A B L E 1 Characteristics of 9 dogs with chronic enteropathy (CE) and 16 healthy controls by sex, age, body condition score and canine chronic enteropathy clinical activity index (CCECAI), cobalamin, C-reactive protein, resting cortisol concentrations or adrenocorticotropic (ACTH) stimulation test results, and pre-prandial bile acids concentrations.  Table 2. The fold change ratios of dogs with FRE and IBD compared to controls are also shown in Table 2. The fold change in dogs with IBD normalized to controls was more pronounced for the majority of proteins of interest than that seen in the FRE normalized to controls. Potential candidate proteins with differential abundance were identified based on SD, mean differences, and power calculation (power > 95%) and are illustrated in Table 2. There was no significant difference in protein abundance between dogs that were receiving prednisolone (n = 3) and dogs that were not treated with prednisolone (n = 6) during sample collection (P > .01).

| Gene ontology analysis
Proteins that were differentially abundant between dogs with CE and controls with fold change ≥2 were used for gene ontology analysis and were categorized according to molecular function, biological processes, and protein classes ( Figure 1A The remaining proteins were involved in immunity, metabolite interconversion enzymes or cell adhesion ( Figure 1C).

Differentially abundant proteins between dogs with CE and controls
with ≥2-fold change were also mapped into enriched pathways with T A B L E 2 Significantly modulated proteins related to enteropathy in dogs with greatest fold change ≥ j5j in dogs with chronic enteropathy (CE) normalized to controls (P < .01), fold change in dogs with food-responsive enteropathy (FRE) and dogs with idiopathic inflammatory bowel disease (IBD) normalized to controls.   Figure 2B.

| DISCUSSION
Our study utilized an untargeted and hypothesis-free mass spectrometrybased proteomic technique. This investigates differences in serum proteomic profiles of dogs with CEs using mass spectrometry-based proteomic techniques. The technique allows the large-scale study of proteins simultaneously and rapid identification of potential protein biomarkers. 30 Our proteomic study identified significant differences in protein profiles between dogs with CE and healthy controls. Forty-five proteins were differentially abundant between dogs with CE and controls with fold change ≥ 2. Comparing the fold change ratio of FRE normalized to controls and IBD normalized to controls, the fold changes were greater in the IBD group. This suggests that the changes in protein profile were likely more functionally important in IBD and might indicate that IBD is a more severe disease when compared to FRE. Differentially abundant proteins were mostly involved in transfer or carrier functions (58.3%). Proteins responsible for immunity, ROS, acute phase response, lipid metabolism, and molecular biochemistry were also found to be enriched in our analysis. Few studies have explored protein profiles in dogs with CE. [31][32][33][34] Serum tryptophan concentrations are significantly lower in dogs with PLE compared to healthy control dogs. 31 Methionine, proline, serine, and tryptophan concentrations are lower in IBD dogs compared to healthy controls. 32 Fecal proteomic analysis using a liquid chromatography mass T A B L E 3 Upstream regulators with molecule type, Z-activation scores, P value, and proteins involved are shown.  phosphorylation of amino acids in proteins. 35 It forms part of a signaling cascade to induce inflammation, producing proinflammatory cytokines. 35 It has been suggested that MAPK3 plays a role in mediating IBD and inflammation in people. 35 Similarly, in our study, the upregulated MAPK3, together with its involvement with fibrinogen and thrombospondin-1 are key indicators of the inflammatory process in dogs with CE. Thrombospondin-1 is an extensively studied protein biomarker used in diagnosis and treatment monitoring in IBD in people, and is a potential protein biomarker in cats with CE. 36  50 and is involved in iron metabolism and facilitates iron transfer by binding to iron loaded transferrin. 50 It maintains intestinal iron homeostasis. 50 Iron metabolism also takes part in immune defense and inflammation. 51 It is possible that the increased fold change in transferrin receptor protein 1 in dogs with CE in our study is related to disruption of intestinal epithelial iron homeostasis and inflammation.
Gelsolin is the second most significant differentiating protein between dogs with CE and controls in our study with fold change 14.8 (P < .001). This protein is proposed to be a potential biomarker for ulcerative colitis and inflammation in people to predict disease severity, treatment efficacy, and clinical outcomes. [52][53][54] Gelsolin is a multifunctional protein and has an important role in cytoskeletal remodeling including actin filament severing, capping and nucleating activity. [52][53][54] It correlates with clinical and endoscopic activities and has a higher sensitivity and specificity than C-reactive protein for diagnosing ulcerative colitis in humans. 52 In contrast to our finding where gelsolin was higher in diseased dogs, gelsolin has an inverse relationship with clinical and endoscopic activity in patients with ulcerative colitis, its level was significantly lower in ulcerative colitis patients compared to healthy controls. 52 Higher gelsolin concentrations is found in intestinal smooth muscle cells in people with Crohn's disease, suggesting an ongoing inflammatory process. 55 In dogs, only limited studies have investigated the utility of gelsolin. 24,56,57 The finding of increased gelsolin levels in dogs in our study could be related to disruption of intestinal mucosa and inflammation in CE. Gelsolin could be a potential biomarker for inflammatory CE in dogs.
There are several limitations to our study. Cats and dogs are less well studied compared to other mammalian species used in research.
This results in a reliance on cross-species identification using a generalized mammalian database. Although over 60 samples were initially collected, the non-standardized methods of data collection across different practices, and between clinicians, created challenges. We used only biopsy confirmed inflammatory CE to increase the specificity and reliability in this discovery phase study. Despite the small sample size with 3 dogs with FRE and 5 dogs with IBD, our study is comparable to the sample sizes used in proteomic discovery phase studies in people and gives basic insight into differences between these 2 population groups. In human medicine, the discovery of proteomic biomarkers is a multi-phase process where the initial phase involves screening for potential protein biomarkers with a small number of samples using mass spectrometry techniques, followed by verification of the markers using alternative techniques and increased number of samples. A validation phase then tracks the successful (single) biomarker candidate(s) within a large sample cohort. 58,59 The higher numbers of IBD cases in our cohort of dogs might reflect referral of cases that were more challenging to manage and required biopsies. Each case was managed at the discretion of the attending veterinarian and the case management was not standardized. Not all diagnostic tests were performed in each dog, thus liver function tests, cPLI and cTLI were only available for some dogs. One dog with CE did not have a dietary trial before immunosuppressant therapy because of concerns for potential rapid deterioration and inappetence, it is possible that this dog could have been misclassified as IBD.
With the diversity of test groups, our data cannot be extrapolated to other dogs with CE. Therefore, the utility of proteomic markers to diagnose, monitor treatment or prognosticate CE in dogs cannot be justified based on our study at this time. However, the application of proteomics allows screening and discovery of potential biomarkers.
The data reported here provide preliminary information about serum proteomics in dogs with CE and discovers potential candidates for