Untargeted metabolomic analysis in cats with naturally occurring inflammatory bowel disease and alimentary small cell lymphoma

Feline chronic enteropathy (CE) is a common gastrointestinal disorder in cats and mainly comprises inflammatory bowel disease (IBD) and small cell lymphoma (SCL). Differentiation between IBD and SCL can be diagnostically challenging. We characterized the fecal metabolome of 14 healthy cats and 22 cats with naturally occurring CE (11 cats with IBD and 11 cats with SCL). Principal component analysis and heat map analysis showed distinct clustering between cats with CE and healthy controls. Random forest classification revealed good group prediction for healthy cats and cats with CE, with an overall out-of-bag error rate of 16.7%. Univariate analysis indicated that levels of 84 compounds in cats with CE differed from those in healthy cats. Polyunsaturated fatty acids held discriminatory power in differentiating IBD from SCL. Metabolomic profiles of cats with CE resembled those in people with CE with significant alterations of metabolites related to tryptophan, arachidonic acid, and glutathione pathways.

www.nature.com/scientificreports/ Type 2)), such as a monomorphic infiltration of the intestinal mucosa with small to medium lymphocytes of predominantly T-cell origin, epitheliotropism, and a predilection for the small intestine 4,18,19 . However, while there are histopathological parallels, the clinical course of feline SCL and EATL Type 2 is very different. While alimentary SCL in cats is slowly progressing with a median survival time of 1.5 to 3.5 years 20,21 , MEITL in people is a clinically aggressive disease and generally associated with a poor prognosis 18 . Recently, other authors suggested that the cat might be a suitable model for indolent digestive T-cell lymphoproliferative disorder 22,23 . This rare disorder in people is characterized by a superficial monoclonal intestinal T-cell infiltrate and the disease is generally indolent or slowly progressive 19,22,[24][25][26][27] .
We hypothesized that cats with CE have metabolomic perturbations and that metabolomic profiling can distinguish cats with IBD from cats with SCL. We further hypothesized that metabolic perturbances observed in cats with CE would be similar to those observed in humans with IBD.

Results
A total of 36 cats were enrolled into this study, 14 healthy cats and 22 cats with chronic enteropathy (11 with IBD and 11 with SCL). A fecal sample was collected from all cats. Demographic characteristics are shown in Table 1.
Age (p = 0.353) and sex distribution (p > 0.999) were comparable between control cats and cats with CE (Table 1). Cats with CE had a significantly lower body weight (p = 0.007) and body condition score 28 (p < 0.001).
A total of 856 named metabolites were detected. Nonparametric univariate analysis using a FDR of 5% revealed a total of 84 metabolites that differed significantly between control cats and cats with CE (Table 2).
Amongst biochemicals found to be increased in feces from cats with CE were various amino acids and their metabolites (e.g., aspartate, cysteine sulfinic acid, phenylalanine, leucine, and valine), fatty acids (e.g., arachidonate and eicosanoids), metabolites within vitamin E metabolism, and simple sphingolipids (e.g., ceramide, sphingomyelins, and sphingosines). Compounds significantly less abundant in cats with CE than in healthy controls included those involved in the tryptophan (i.e., indole-derivates) and vitamin A metabolism (i.e., carotendiols), and sterols.
Upon analysis of the CE subgroups, IBD and SCL, multivariate analysis revealed no visible clustering between cats with IBD and cats with SCL (Fig. 1a,c). Random forest analysis revealed poor group prediction with an OOB of 47.2% and class error rates of 82% and 55% for cats with IBD and cats with SCL, respectively. However, univariate analysis revealed 18 metabolites to differ significantly between healthy cats, cats with IBD, and cats with SCL. Post hoc analysis showed that 3 polyunsaturated fatty acids (i.e., eicosapentaenoate, heneicosapentaenoate, and stearidonate) within the eicosanoid family differed between cats with IBD and cats with SCL (Table 4).

Discussion
Our study revealed global metabolic changes in cats with chronic enteropathy as compared with healthy controls, with many metabolic pathways affected. Both hierarchical cluster analysis and PCA demonstrated clustering among cats diagnosed with chronic enteropathy and healthy control cats. Random forest analysis revealed a good class prediction of 80.6%. Univariate analysis showed a total of 84 metabolites to differ significantly between the two groups after controlling for a 5% FDR. Similar studies in humans with inflammatory bowel disease have also shown significant metabolomic differences between affected patients and healthy controls [10][11][12]30 . Although IBD in humans and CE in cats share only a few characteristics, the use of metabolomics data in humans with chronic enteropathies would support this type of approach in further assessing cats with various forms of CE. www.nature.com/scientificreports/ In addition, metabolic consequences of feline CE and human IBD appear to be very similar. Metabolites or metabolite families commonly found to be affected by IBD in humans are amino acids 8,10-13,30 , bile acids 10 , fatty acids 10,11 , and metabolites of the tryptophan pathway 10,31 . We found multiple amino acids to be increased in the feces from cats with CE, indicating malabsorption, likely as consequence of mucosal inflammation, neoplastic infiltration, or both. Interestingly, 2-hydroxybutyrate/2hydroxyisobutyrate and gamma-glutamylglutamine, two metabolites related to the glutathione-metabolism, were found to be significantly more abundant in feces from cats with CE based on both univariate and multivariate analysis. Gamma-glutamyl amino acids are precursors for the formation of glutathione in the gamma-glutamyl cycle (glutathione salvage pathway) 32 (Fig. 3). Reactive oxygen species have been implicated to contribute to tissue injury in patients with Crohn's disease and ulcerative colits 33,34 . Glutathione is the major intracellular antioxidant and thus a critical part of the defense mechanism against oxidative stress in inflammatory conditions such as CE (Fig. 3) 35,36 . Consequently, glutathione precursors are in high demand during catabolic conditions and corresponding higher loads of oxidative stress 30 . Increased fecal concentrations of gamma-glutamyl www.nature.com/scientificreports/ glutamine in cats with CE might indicate increased loss and correspond to a decreased mucosal availability and glutathione synthesis.
Our study showed arachidonate to be more abundant in feces from cats with CE. Arachidonate is a well described mediator of inflammation, including intestinal inflammation, and a precursor for prostaglandins, which are essential immune signaling molecules 37 . In addition, arachidonate has been found to increase the expression of intercellular adhesion molecule 1 (ICAM-1) 38 . Adhesion molecules such as ICAM-1 and mucosal addressin cell adhesion molecule (MAdCAM) are involved in recruiting leucocytes to the site of inflammation and have been identified as therapeutic targets for human patients with IBD 39 . Increased fecal arachidonate may reflect mucosal upregulation during inflammation and subsequent leakage into the fecal stream. On the other hand, omega-3 polyunsaturated fatty acids, including eicosapentaenoate, have been found to be significantly more abundant in the SCL subgroup of cats with CE compared to healthy controls. , green (IBD, n = 11), and blue (SCL, n = 11). Clusters can be identified between healthy cats and cats with chronic enteropathy (CE) but not between the disease subgroups IBD and SCL. (b) PCA score plots of metabolites in feces from healthy cats (green) and cats with chronic enteropathy (CE, red). (c) PCA score plots of metabolites in feces from healthy cats (red), cats with inflammatory bowel disease (blue), and cats with alimentary small cell lymphoma (green). A cluster can be identified for healthy cats but no specific clusters can be seen for the subgroups of IBD and SCL. Data was mean centered and divided by the standard deviation of each variable (autoscaled). www.nature.com/scientificreports/ Eicosapentanoate and other omega-3 fatty acids have anti-inflammatory properties and inhibit conversion of arachidonate into the pro-inflammatory thromboxane-2 and prostaglandin-2 families 40 . Increased fecal levels of omega-3 fatty acids might indicate an increased loss or malabsorption and thus may correspond to the increased arachidonate concentration found in this study.
We also found several metabolites of the indole family and pathway to be altered in cats with CE. Amongst a wide range of structurally divergent exo-and endogenic chemicals 41 , indole derivates are important ligands for the aryl hydrocarbon receptor (AhR) 42 . AhR signaling participates integrally in intestinal mucosal homeostasis by acting on innate and adaptive immune cells as well as on epithelial renewal and mucosal barrier function 43 . Tryptophan plays a central role in AhR activation because it is transformed into indole and indole derivates by the gut microbiota. Studies in murine and porcine models of colitis have found significantly reduced disease activities after oral supplementation with indole-3-propionate or L-tryptophan by activating anti-inflammatory pathways mediated by IL-10 and IL-22 42,44 . We found indole derivates, such as 2-oxindole-3-acetate and 5-hydroxyindoleacetate, to be significantly less abundant in feces from cats with CE. Although, tryptophan did not significantly differ between CE and control cats, it showed a trend (p = 0.0855, q = 0.2385) towards a higher fecal excretion in cats with CE 42,45,46 (Appendix C, Tables 5.5, 5.6). These findings might indicate disrupted transformation of tryptophan into indole derivates, possibly linked to intestinal dysbiosis in this cohort of cats with CE.
Finally, our study showed increased concentrations of several simple sphingolipids in feces from cats with CE. The role of sphingolipids in IBD is complex and incompletely understood. It appears, however, that a functional equilibrium between simple sphingolipids (e.g., sphingomyelin, sphingosine, ceramide) and complex sphingolipids (e.g., gangliosides GM3 and GD3) is essential in maintaining intestinal homeostasis 47 . Proinflammatory sphingolipid patterns have previously been described, with increased concentrations of sphingomyelin and ceramide in feces of animal models of colitis 48,49 as well as in the ileal mucus of human patients with Crohn's disease 50 indicating that sphingomyelin and ceramide accompany and possibly aggravate chronic intestinal inflammation.
Fecal extracts obtained from cats with IBD and SCL revealed similar global changes in metabolic profiles. However, changes observed in cats with IBD were less pronounced than those observed in cats with SCL. This suggests that metabolic consequences are more severe in SCL than in IBD. In particular, higher concentrations of polyunsaturated fatty acids of the eicosanoid family were found in SCL.
This study had several limitations. A clinical diagnosis of IBD was based on the presence of chronic gastrointestinal signs of at least 3 weeks duration, the absence of known enteropathogens or other causes of signs of gastrointestinal disease, and the histopathologic confirmation of intestinal inflammation 51 . Hence, we did not exclude food-responsive enteropathy in all participating cats. However, in response to dietary interventions human IBD patients can show complete or partial resolution of clinical signs without being reclassified as having food-responsive enteropathy [52][53][54][55] . Fecal cultures for specific pathogens were not performed. However, gastrointestinal disease caused by bacteria usually cause acute diarrhea. In addition, all cats showed either lymphoplasmacytic enteritis or lymphoma on histopathologic examination. Cats with neutrophilic enteritis, a hallmark of www.nature.com/scientificreports/  www.nature.com/scientificreports/ gastrointestinal bacterial infections, were excluded from the study. Differentiating IBD from SCL can impose significant challenges with clinical signs and histopathologic changes commonly overlapping 3 . Moreover, mucosal changes may have an irregularly distribution, inflammatory and neoplastic lesions often coincide and inflammatory lesions might even progress to lymphoma over time 3,4 . Therefore, using currently available measures, the classification into groups of IBD and SCL will always be associated with a degree of ambiguity. However, in all cats intestinal biopsies were available and results of immunohistochemistry and clonality testing were utilized, where indicated, to provide the highest accuracy in category assignment.
Inter-individual variation of metabolomic profiles have been described in humans, dogs, and cats and might be related to a variety of exogenous and endogenous factors such as the environment, diet, gut microbiota, xenobiotics, and the genome 14,17,56 . This study intended to identify clinically relevant perturbations of the fecal metabolome of cats with CE compared to healthy subjects. Demographic characteristics like age, sex, and breed were controlled for in this study. However, we did not control for environmental factors, such as diet and housing, since this likely would have impacted the clinical relevance of our results. This concept is supported by recent studies showed that standardization in is a major source of poor reproducibility preclinical trials 57,58 . Although, we cannot exclude that these factors confounded our results, most are in line with findings across different species with spontaneous or induced IBD and thus likely reflect true changes of the fecal metabolome. In addition, most cats in both groups were housed indoors and fed a variety of different commercial diets. Another limitation is the limited number of animals in this study. However, untargeted "omics" analyses usually provide only relative changes between subjects or groups and are generally conducted using a smaller sample size. Therefore, all untargeted analyses have to be followed by targeted assays to confirm results of those fishing expeditions on a larger number of subjects. However, this was beyond the scope of this study.
In summary, our study revealed global metabolic changes in cats with CE compared to healthy controls. Many metabolic pathways were affected such as amino acids, fatty acids (e.g., arachidonic acid and eicosanoids), sphingolipids, and metabolites of the tryptophan pathway (i.e., indole-derivates). Metabolic profiles resembled patterns found in humans and other animal models with IBD and thus metabolic consequences of feline CE and human IBD might be similar. Future studies of the mucosal and serum metabolome of cats with CE should be conducted to further elucidate the origin of metabolic perturbations and allow further insights into pathogenesis. In addition, targeted analysis of the compounds found to be altered in cats with CE is indicated to confirm results and to investigate their value as non-invasive biomarkers. The health status of cats in the group considered healthy was verified by an owner questionnaire on general and gastrointestinal health 59 . The questionnaire covered the following areas: attitude/activity, appetite, drinking, urination, chronic illnesses, weight loss, vomiting, diarrhea, and treatment with antibiotics, antacids, anti-inflammatory drugs, or steroids 59 . In addition, a physical examination and blood testing was available and performed by a single board-certified internist (SM) in 11 of 14 cats. The body condition score was assessed using a previously established a nine-point condition scoring system 28 . Blood was collected from a peripheral vein or the jugular vein and the following tests were performed: complete blood count, serum chemistry profile, total T4, cobalamin, folate, feline pancreatic lipase immunoreactivity (fPLI), and feline trypsin-like immunoreactivity (fTLI). Cats with gastrointestinal signs (weight loss, hyporexia, vomiting > 2×/ month, diarrhea) within 6 months prior to enrollment were excluded. In addition, cats with systemic diseases, chronic illnesses or clinically significant laboratory abnormalities were excluded from the study 59 . Finally, cats that had received any antibiotics, antacids, anti-inflammatory drugs, or corticosteroids within the past 6 months were excluded 59 .

Materials and methods
Cats with clinical signs of chronic enteropathy (weight loss, hyporexia, vomiting, diarrhea) of at least 3 weeks duration were eligible for enrollment into the group of cats with chronic enteropathy 59 . Extra-gastrointestinal disease as well and, where indicated, infectious intestinal diseases were excluded based on a complete blood count, serum chemistry profile, total T4 and fecal flotation 59 . All cats in this group underwent gastro-duodenoscopy and ileo-colonoscopy for diagnostic purposes. Histopathologic examination of H&E stained endoscopic formalinfixed, paraffin-embedded (FFPE) tissue sections was performed by board-certified pathologists (MA or JSE) blinded to the clinical status of the cats 59 .
Cases with a histopathological diagnosis of SCL or cases where the pathologist was suspicious of an underlying SCL underwent additional diagnostic testing with immunohistochemistry and PCR for antigen receptor rearrangement testing for diagnostic confirmation. A final diagnosis of IBD or SCL was reached upon integration of results from histopathology, immunohistochemistry, and PARR based on the current EuroClonality/BIOMED-2 guidelines for interpretation and reporting of Ig/TCR clonality testing in suspected lymphoproliferations [59][60][61][62] . Cats that had received antibiotics within 4 weeks or corticosteroids within the past 2 weeks prior to fecal sampling were excluded from the study 59  Metabolite extraction. Untargeted fecal metabolomic analysis was performed by Metabolon, Inc. (Durham, NC) as previously described [63][64][65] . Metabolites were extracted from lyophilized and homogenized samples using methanol extraction. Extracts were analyzed by an ACQUITY ultra-performance liquid chromatographer (Waters, Milford, CA) and a ThermoFisher Scientific Q-Exactive high resolution mass spectrometer interfaced with a heated electrospray ionization (HESI-II) source and Orbitrap mass analyzer (ThermoFisher Scientific, Waltham, Massachusetts) operated at 35,000 mass resolution. The scan range covered 70-1000 m/z. Metabolite identification was performed by automated comparison of the ion features in the experimental samples to a reference library. Chemical standard entries included retention time, molecular weight (m/z), preferred adducts, and in-source fragments as well as associated MS spectra, and were curated by visual inspection for quality control using software developed at Metabolon. Peaks were quantified using the area-under-the-curve. Compounds were corrected for inter-day variation by registering the medians to equal 1.00 and normalizing each data point proportionally. Missing values were imputed with the observed minimum for each compound.
Statistical analyses. Patient demographics were compared by the Mann-Whitney or Fisher's exact test as appropriate. Differences in the abundance of fecal metabolites between control cats and cats with CE were evaluated using a Mann-Whitney test. A post hoc analysis to assess differences in the abundance of metabolites between the subgroups (i.e., controls, IBD, SCL) was performed by Dunn's test. Statistical significance was set at p < 0.05. Results were adjusted by False Discovery Rate (FDR) and reported as the q-value where appropriate. Univariate analysis was performed using Prism 7.0b (Graph Pad Software, La Jolla, CA) and JMP Pro 14.1.0 (SAS Institute Inc., Cary, NC). Multivariate analysis was performed using MetaboAnalyst 66 . Data was mean centered and divided by the standard deviation of each variable. PCA and hierarchical clustering was performed and a heatmap was created as a visual aid for the dendrogram. Random forest regression analysis was used to evaluate the classification performance of metabolomics.
Ethical approval. This article does not contain any studies with human participants performed by any of the authors.
Research involving animal rights. All applicable international, national, and/or institutional guidelines for the care and use of animals were followed.

Data availability
The datasets generated and/or analyzed during the current study are available from the corresponding author on reasonable request.