Abstract

Simple Summary

   Exocrine pancreatic insufficiency (EPI) is a digestive disorder in dogs
   resulting from insufficient secretion of digestive enzymes from the
   exocrine pancreas. EPI is treated with oral pancreatic enzyme
   replacement therapy (PERT), but the persistence of clinical signs,
   especially diarrhea, is common after treatment. We sought to develop
   new insights into EPI using untargeted metabolomics analysis, a method
   that can measure hundreds of biochemicals (metabolites) in a sample. We
   analyzed 759 serum metabolites and identified 114 that varied
   significantly between dogs with EPI and healthy controls. Differences
   in fatty acid and amino acid metabolites were consistent with a state
   of malnourishment, and decreased vitamin B and C metabolites were
   suggestive of micronutrient deficiencies in dogs with EPI. Disturbances
   in gut microbial metabolites indicated the altered composition of the
   intestinal microbiome in dogs with EPI. Increased kynurenine, a
   tryptophan metabolite, in dogs with EPI may be associated with
   intestinal inflammation. As an exploratory study, causation cannot be
   determined from these results, but our findings have generated new data
   that can be used to inform future investigations of gastrointestinal
   and metabolic disturbances underlying the persistence of clinical signs
   in dogs with EPI treated with PERT.

Abstract

   Exocrine pancreatic insufficiency (EPI) is a malabsorptive syndrome
   resulting from insufficient secretion of pancreatic digestive enzymes.
   EPI is treated with pancreatic enzyme replacement therapy (PERT), but
   the persistence of clinical signs, especially diarrhea, is common after
   treatment. We used untargeted metabolomics of serum to identify
   metabolic disturbances associated with EPI and generate novel
   hypotheses related to its pathophysiology. Fasted serum samples were
   collected from dogs with EPI (n = 20) and healthy controls (n = 10),
   all receiving PERT. Serum metabolomes were generated using UPLC-MS/MS,
   and differences in relative metabolite abundances were compared between
   the groups. Of the 759 serum metabolites detected, 114 varied
   significantly (p < 0.05, q < 0.2) between dogs with EPI and healthy
   controls. Differences in amino acids (arginate, homoarginine,
   2-oxoarginine, N-acetyl-cadaverine, and α-ketoglutaramate) and lipids
   (free fatty acids and docosahexaenoylcarnitine) were consistent with
   increased proteolysis and lipolysis, indicating a persistent catabolic
   state in dogs with EPI. Relative abundances of gut microbial
   metabolites (phenyllactate, 4-hydroxyphenylacetate, phenylacetyl-amino
   acids, catechol sulfates, and o-cresol-sulfate) were altered in dogs
   with EPI, consistent with disruptions in gut microbial communities.
   Increased kynurenine is consistent with the presence of intestinal
   inflammation in dogs with EPI. Whether these metabolic disturbances
   participate in the pathophysiology of EPI or contribute to the
   persistence of clinical signs after treatment is unknown, but they are
   targets for future investigations.

   Keywords: EPI, pancreas, dogs, canine, metabolomics

1. Introduction

   Canine exocrine pancreatic insufficiency (EPI) is a malabsorptive
   syndrome caused by insufficient secretion of pancreatic digestive
   enzymes [[34]1,[35]2]. Characteristic clinical signs of EPI include
   diarrhea, weight loss, and polyphagia. Clinical signs result from
   failure to digest dietary macromolecules (i.e., fats, carbohydrates,
   and proteins) resulting in malabsorption, negative energy balance, and
   altered composition of the gut microbiome (small intestinal bacterial
   overgrowth) [[36]2]. Due to malabsorption, EPI is also associated with
   deficiencies in specific micronutrients, such as cobalamin and
   lipid-soluble vitamins [[37]3,[38]4,[39]5]. The most common cause of
   canine EPI is end-stage pancreatic acinar atrophy (PAA), and clinical
   signs of EPI develop following an approximate 90% reduction in
   pancreatic acinar mass [[40]6,[41]7,[42]8,[43]9]. Serum concentrations
   of canine trypsin-like immunoreactivity (cTLI) less than 2.5 µg/L are
   diagnostic for EPI [[44]10,[45]11]. Dogs with EPI require life-long
   pancreatic enzyme replacement therapy (PERT), typically via the
   addition of enzyme extracts to each meal, and most require
   supplementation with cobalamin. The long-term prognosis for EPI is
   good, but responses to enzyme replacement therapy are variable, and a
   substantial proportion (~40%) of dogs with EPI have persistent clinical
   signs after the initiation of PERT, most commonly diarrhea and weight
   loss [[46]12,[47]13,[48]14]. Thus, many dogs require adjunctive
   therapies, including therapeutic diets, antibiotics, probiotics, or
   some combination thereof, to control their clinical signs.

   Little is known about the pathogenesis of PAA or its progression to
   EPI. Immunohistochemical and serologic studies have identified
   lymphocytic pancreatitis and circulating antibodies against pancreatic
   acini in dogs with subclinical PAA; however, in end-stage PAA
   associated with EPI, there is the replacement of acini with adipose
   tissue with minimal inflammation or fibrosis [[49]9,[50]15,[51]16].
   Over-representation of certain breeds, especially German Shepherds,
   among dogs with EPI has prompted investigations to identify a genetic
   basis for the disorder. The heritability of EPI is complex and does not
   follow a Mendelian pattern [[52]17,[53]18]. Recent genomics studies
   have identified a common risk allele in the dog leukocyte antigen
   complex in German Shepherd dogs and Pembroke Welsh Corgis with EPI
   [[54]19,[55]20]. However, the risk allele was not present in all dogs
   with EPI and was identified in some healthy dogs. Collectively, these
   findings suggest that there is an immune-mediated component to the
   pathogenesis of PAA and that one or more genetic or environmental risk
   factors may contribute to its pathogenesis.

   Given the incomplete understanding of its etiopathogenesis, there is a
   need to generate new insights into the pathogenesis of EPI and its
   pathophysiologic consequences. There is also a need to identify
   pathophysiologic disturbances present in dogs with EPI after the
   initiation of PERT to develop novel therapeutic approaches for dogs
   with persistent signs of gastrointestinal dysfunction. Untargeted
   metabolomics analysis is an investigative tool that can provide new
   insights into poorly understood disease processes, especially those
   associated with nutritional and metabolic disturbances, such as EPI.
   The metabolome is the sum of all small molecules (<1500 kDa) present in
   an organism or biologic sample that are precursors, intermediates, and
   end-products of metabolic processes. Metabolites of endogenous and
   exogenous (diet, environment, and microbial) origin contribute to the
   serum metabolome of mammals [[56]21]. The quantification of metabolites
   in a biologic sample can yield valuable insights into the metabolic
   phenotypes (i.e., metabotypes) of disease states by elucidating
   pathophysiologic mechanisms of disease and facilitating the discovery
   of novel biomarkers. The objectives of this investigation were to (1)
   detect differences in serum metabolite profiles between dogs with EPI
   and healthy controls and (2) to identify novel metabolic signatures
   relevant to the pathophysiology of EPI. While this was an untargeted,
   hypothesis-generating investigation, we hypothesized that metabolomics
   profiling would identify novel serum biomarkers plausibly associated
   with persistent gastrointestinal dysfunction in dogs with EPI following
   treatment with PERT.

2. Materials and Methods

2.1. Patient Population and Sample Collection

   Adult (≥1 year of age) dogs with previously diagnosed EPI were
   recruited from an online patient registry (epi4dogs.com/epi-registry;
   registry data accessed 5 February 2016). Dogs diagnosed with EPI were
   identified from the registry data, and their owners and primary care
   veterinarians were contacted to collect patient histories and medical
   records. Dogs were eligible for inclusion if a review of their medical
   records identified subnormal serum cTLI concentrations (cTLI < 2.5
   µg/L; reference interval: 5.7–45.2 µg/L) from two or more consecutive
   measurements. Additionally, dogs with EPI had to have been receiving
   oral PERT with a commercially available enzyme extract of porcine
   origin for at least 30 days prior to enrollment. The diets of
   participating dogs were not standardized, and they were fed their
   current diet under the supervision of their owners and primary
   veterinarians. Information regarding the diets and PERT of dogs with
   EPI included in this study are presented in [57]Supplementary File S1.
   Dogs with EPI were excluded if they had been diagnosed with any other
   gastrointestinal, metabolic, endocrine, systemic, or neoplastic
   diseases by their primary care veterinarian. After an overnight fast,
   samples of whole blood were collected by the dogs’ primary care
   veterinarians via jugular or lateral saphenous venipuncture using
   sterile 20 g hypodermic needles and syringes. Whole blood samples were
   collected into empty, sterile tubes, allowed to clot at room
   temperature for 30 min, and centrifuged on site at 2500–3500 rpm for 10
   min. The resulting serum was aliquoted into empty, sterile
   polypropylene tubes and immediately frozen on site in clinical freezers
   (approximately −20 °C). To standardize sample collection, all required
   materials (hypodermic needles, syringes, transfer pipets, sample tubes)
   and detailed instructions were provided to participating veterinarians
   by the investigators. All samples were mailed overnight on dry ice to
   the investigators, where they were stored at −80 °C following
   acquisition. Samples were collected and received by the investigators
   between September, 2016 and February, 2017. Aliquots of serum were
   submitted for measurement of serum cTLI, cobalamin, and folate using
   solid-phase chemiluminescent assays run on the IMMULITE 2000 PXi
   immunoassay system (Siemens Inc., Munich, Germany) at the Texas A&M
   Gastrointestinal Laboratory (College Station, TX, USA). Dogs with serum
   cTLI concentrations <2.5 µg/L were assigned to the EPI group, and those
   with serum cTLI concentrations ≥2.5 µg/L were excluded.

   Healthy, adult (≥1 year of age), client-owned dogs were recruited as a
   control group. Owners completed a survey to gather information about
   their dogs’ medical history, while their medical records were reviewed
   by the investigators. To screen for confounding diseases, each healthy
   dog received a physical exam, and the following diagnostic screening
   tests were performed: complete blood count, serum biochemistry panel,
   urinalysis, serum total T4, serum cTLI, serum cobalamin, serum folate,
   and fecal flotation. Complete blood count, serum chemistries,
   urinalysis, total T4, and fecal flotation assays were performed at the
   University of Illinois Veterinary Diagnostic Laboratory (Urbana, IL,
   USA). Dogs were excluded from this group if they had any historical
   clinical signs of gastrointestinal disease (vomiting, diarrhea, weight
   loss, anorexia, polyphagia) within 6 months of recruitment; serum cTLI
   concentrations <5.7 µg/L; clinical signs or laboratory abnormalities
   consistent with any systemic, metabolic, gastrointestinal, or
   pancreatic diseases; or received antibiotics, corticosteroids, or
   immunomodulatory drugs within 6 months of recruitment. To account for
   the potential impact of pancreatic enzyme replacement therapy on serum
   metabolite profiles, a powdered pancreatic enzyme extract (Pancreatin
   6X; American Laboratories Inc., Omaha, NE, USA) was administered orally
   (1 tsp/cup of food) to the healthy dogs for 14 days prior to sample
   collection. Following the supplementation period, serum samples were
   collected after an overnight fast as described above and immediately
   frozen at −80 °C.

2.2. Untargeted Serum Metabolomics

   Serum metabolite profiles were generated via ultrahigh performance
   liquid chromatography-tandem mass spectroscopy (UPLC-MS/MS) by a
   commercial laboratory (Metabolon Inc., Morrisville, NC, USA) between
   June and September, 2017. Serum samples (100 µL) were deproteinated
   using methanol (500 µL) precipitation with vigorous shaking for 2 min
   followed by centrifugation. The methanol contained four recovery
   standards (DL-2-fluorophenylglycine, tridecanoic acid, d6-cholesterol,
   and 4-chlorophenylalanine) used to assess extraction efficiency. The
   resulting extracts were dried and reconstituted in solvents compatible
   to each of the four UPLC-MS/MS methods. Two aliquots were reconstituted
   in 50 μL of 6.5 mM ammonium bicarbonate in water (pH 8) for negative
   ion analysis methods, and two were reconstituted using 50 μL 0.1%
   formic acid in water (pH 3.5) for the positive ion analysis methods.
   Reconstitution solvents contained internal standards (d7-glucose,
   d3-leucine, d8-phenylalanine, d5-tryptophan, d5-hippuric acid,
   Br-phenylalanine, d5-indole acetic acid, amitriptyline,
   d9-progesterone) to assess instrument performance and to aid in
   chromatographic alignment. Analytic controls consisting of pooled
   samples, generated from small volumes of each experimental sample, were
   analyzed simultaneously with experimental samples to serve as a
   technical replicate throughout the dataset. Methanol-extracted water
   samples served as process blanks. Experimental samples were randomized
   across the platform run with quality control samples spaced evenly
   among the injections.

   For UPLC-MS/MS, all methods utilized a Waters ACQUITY ultraperformance
   liquid chromatographer and a Thermo Scientific Q Exactive high
   resolution/accurate mass spectrometer interfaced with a heated
   electrospray ionization source and Orbitrap mass analyzer operated at
   35,000 mass resolution. One aliquot was analyzed using acidic positive
   ion conditions optimized for hydrophilic compounds. In this method, the
   extract was gradient eluted from a C18 column (Waters UPLC BEH C18-2.1
   × 100 mm, 1.7 μm) using water and methanol containing 0.05%
   perfluoropentanoic acid (PFPA) and 0.1% formic acid (FA). Another
   aliquot was analyzed using acidic positive ion conditions
   chromatographically optimized for hydrophobic compounds. In this
   method, the extract was gradient eluted from the same C18 column using
   methanol, acetonitrile, water, 0.05% PFPA, and 0.01% FA. Another
   aliquot was analyzed using basic negative ion optimized conditions
   using a separate dedicated C18 column (Waters UPLC BEH C18-2.1 × 100
   mm, 1.7 μm). The basic extracts were gradient-eluted from the column
   using methanol and water with 6.5 mM ammonium bicarbonate at pH 8. The
   fourth aliquot was analyzed via negative ionization following elution
   from an HILIC column (Waters UPLC BEH Amide 2.1 × 150 mm, 1.7 μm) using
   a gradient consisting of water and acetonitrile with 10 mM ammonium
   formate, pH 10.8. The MS analysis alternated between MS and
   data-dependent MSn scans using dynamic exclusion. The scan covered
   70–1000 m/z. Data extraction, compound identification, and data
   processing were performed using a proprietary software platform by
   Metabolon Inc. Compounds were identified by comparison to library
   entries of purified and authenticated standards. Metabolites were
   quantified by measuring the area-under-the-curve of the chromatographic
   peak.

2.3. Statistical Analysis

   Categorical variables from the demographics data were compared using
   the Fisher’s exact or chi-squared test. The normality of numerical
   demographic and clinical variables were assessed using Shapiro–Wilk
   tests and by examining histograms. For data that were normally
   distributed, t-tests were used to compare means, and for those with
   non-normal distributions, nonparametric Wilcoxon rank-sum tests were
   used. Results were considered statistically significant when p < 0.05.
   Relative abundances of serum metabolites were normalized using
   median-scaling, missing values were imputed with the sample set
   minimum, and the data were log-transformed. Unsupervised analysis
   consisted of principal component analysis (PCA) and hierarchical
   clustering of Euclidian distances. Welch’s two-sample t-test was used
   to detect serum metabolites that varied between the EPI and healthy
   control groups. Metabolite set enrichment analysis (MSEA) was used to
   identify metabolic subpathways enriched in metabolites that varied
   between groups. Metabolites were ranked by the product of the
   −log(p-value) from the t-tests and the log2-fold-change in the
   metabolites between groups. This resulted in a ranked vector where the
   top of the vector contained significantly variable metabolites with
   relative abundances that were increased, and the bottom of the vector
   contained significantly variable metabolites that were decreased in the
   EPI group compared with healthy controls. Next, MSEA was implemented by
   using the “fgsea” R package (version 1.22.0) to calculate normalized
   enrichment scores (NES) for each metabolic subpathway [[58]22,[59]23].
   Subpathways with a NES > 0 were significantly upregulated, and those
   with a NES < 0 were significantly downregulated in the EPI group
   compared with healthy controls.

   The Benjamini–Hochberg false discovery rate (FDR) procedure was used to
   calculate FDR-corrected p-values (q-values) for multiple comparisons in
   the metabolomics data [[60]24,[61]25]. We sought to maximize the
   discovery of features that vary in association with EPI while
   controlling for false discovery within reasonable bounds. Metabolomics
   results were considered statistically significant when p < 0.05 and q <
   0.2. All statistical analyses were performed in the R language for
   statistical computing (version 4.2.1) using the RStudio integrated
   development environment running on macOS (Monterey 12.5).

3. Results

3.1. Demographic and Clinical Data

   Serum samples from 30 dogs were analyzed for this study, including 20
   dogs with EPI and 10 healthy controls. There were no statistically
   significant differences in age, breed, or reproductive status between
   the EPI and healthy control groups ([62]Table 1). The persistence of
   clinical signs associated with EPI were common in dogs with EPI
   following treatment with PERT. Diarrhea was reported in 40% (n = 8) and
   persistent weight loss in 15% (n = 3) of dogs in the EPI group. Other
   clinical signs that are considered atypical for EPI were also reported
   in the EPI group, including decreased appetite in 10% (n = 2) and
   vomiting in 30% (n = 6). Dogs with EPI had received a variety of
   adjunct therapies in addition to PERT. Antibiotics were administered in
   35% (n = 7) of dogs with EPI, including metronidazole in 5% (n = 1) and
   tylosin in 30% (n = 6). Cobalamin and folate supplementation was
   administered to 70% (n = 14) and 40% (n = 8) of dogs with EPI,
   respectively. Other adjunctive medications administered to dogs with
   EPI included gastric-acid-suppressing drugs in 30% (n = 6) and
   probiotics in 70% (n = 14). As the presence of the active clinical
   signs of gastrointestinal disease and administration of
   gastrointestinal medication were an exclusion criterion for healthy
   dogs, no statistical comparisons were performed between the EPI and
   healthy control groups. Except for one dog in the EPI group that was
   fed a home-prepared diet, all dogs included in this study were fed
   diets that were formulated to meet the nutritional levels established
   by the AAFCO Dog Food Nutrient Profiles ([63]https://www.aafco.org;
   accessed 7 June 2023) for maintenance or animal feeding tests using
   AAFCO procedures that substantiated that the diet provides complete and
   balanced nutrition for the maintenance of adult dogs. Four dogs in the
   EPI group were fed mixtures of different commercial diets that were
   formulated to meet the AAFCO profiles. With the exception of the dog
   fed a home-prepared diet and those fed mixtures of commercial diets,
   all dogs were fed dry kibble diets. Mean (± SD) dietary fat (as-fed
   basis) was significantly (p = 0.012) higher in dogs in the EPI group
   (15.4% ± 1.7) compared with the healthy controls (13.3% ± 2.1), but
   there were no other significant differences in dietary protein (p =
   0.08) or crude fiber (p = 0.95) between the groups.

Table 1.

   Patient population statistics. Categorical variables compared using
   Fisher’s exact or chi-squared tests. Variables with normal
   distributions are presented as means and standard deviations (SD) and
   are otherwise presented as medians and interquartile ranges (IQRs).
   Comparisons of serum cobalamin and folate omit dogs receiving
   supplementation. Laboratory reference intervals (RIs) are listed for
   cTLI, cobalamin, and folate.
                                          Healthy           EPI      p-Value
   Age (years)                                                        0.36
   Mean (SD)                            5.0 (± 1.74)    4.3 (± 2.0)
   Breed                                                              0.62
   Akita                                   0 (0%)         1 (5%)
   Australian Shepherd                     0 (0%)         2 (10%)
   Border Collie                           0 (0%)         1 (5%)
   Cavalier King Charles Spaniels          0 (0%)         1 (5%)
   German Shepherd                        4 (40%)         9 (45%)
   Labrador Retriever                     1 (10%)         1 (5%)
   Mixed Breed                            4 (40%)         4 (20%)
   Pit Bull Terrier                       1 (10%)         0 (0%)
   West Highland White Terrier             0 (0%)         1 (5%)
   Sex                                                                0.16
   Spayed Female                          3 (30%)        11 (55%)
   Intact Male                             0 (0%)         2 (10%)
   Neutered Male                          7 (70%)         7 (35%)
   Serum cTLI (RI: 5.7–45.2 µg/L)                                    <0.001
   Median (IQR)                       18.8 (15.7–29.1) 1.0 (1.0–1.1)
   Serum cobalamin (RI: 251–908 ng/L)                                 0.01
   Median (IQR)                        411 (297–593)   231 (207–313)
   Serum folate (RI: 7.7–24.4 ng/L)                                   0.39
   Mean (SD)                            12.1 (±5.2)     13.9 (±4.5)
   [64]Open in a new tab

   Serum biomarkers for exocrine pancreatic and gastrointestinal function
   were measured in the sera of dogs included in this study: cTLI,
   cobalamin, and folate ([65]Table 1). Serum concentrations of cTLI
   (median, IQR) were significantly lower in dogs with EPI (1.0 µg/L,
   1.0–1.1) compared with the healthy controls (18.8 µg/L, 15.7–29.1; p <
   0.001). After excluding dogs that had received supplementation with
   cobalamin or folate, serum cobalamin (median, IQR) concentrations were
   significantly lower in dogs with EPI (231 ng/L, 207–313) compared with
   the healthy controls (411 ng/L, 297–593; p = 0.013), but there were no
   statistically significant differences in serum folate concentrations
   between the groups (p = 0.39).

3.2. Untargeted Metabolomics Analysis

   The untargeted serum metabolome dataset contained 759 named
   biochemicals. Among metabolic superpathways, lipids were the most
   numerous (n = 355), followed by amino acids (n = 187), xenobiotics (n =
   79), nucleotides (n = 47), peptides (n = 31), cofactors and vitamins (n
   = 29), carbohydrates (n = 21), and energy (n = 10) metabolites. PCA
   separated samples from dogs with EPI from healthy controls along the
   first principal component ([66]Figure 1A), consistent with
   metabolome-wide differences in the abundance of serum biochemicals
   among groups. Similarly, hierarchical clustering of metabolite relative
   abundances based on Euclidian distances revealed the clustering of most
   samples from the healthy controls away from those with EPI ([67]Figure
   1B).

Figure 1.

   [68]Figure 1
   [69]Open in a new tab

   Unsupervised analysis of serum metabolite profiles. Samples are
   annotated by color according to their group: EPI (orange), healthy
   (blue). (A) Principal component analysis of relative metabolite
   abundances. (B) Statistical heatmap of relative metabolite abundances
   using Euclidian distances and hierarchical clustering.

   We sought to detect individual metabolites that varied in association
   with EPI by comparing relative abundances of serum metabolites between
   EPI and the healthy controls. After controlling for false discovery (q
   < 0.2), the relative abundances of 114 serum metabolites varied
   significantly between the groups, with 76 being increased and 38 being
   decreased in dogs with EPI compared with the healthy controls. A
   selection of significantly variable serum metabolites with a large
   effect size is presented in [70]Table 2. The complete output is
   presented in [71]Supplementary File S2, and the plots of all
   significantly variable metabolites are presented in [72]Supplementary
   File S3.

Table 2.

   Significantly variable serum metabolites. Significantly variable serum
   metabolites with large effect size (|log2-fold-change|> 0.6;
   corresponding > 1.5 fold difference in means). p-values derived from
   Welch’s t-test. q-value, false discovery rate-adjusted p-values;
   log2FC, log2 of the fold difference in means between dogs with EPI and
   healthy controls.
   Biochemical Subpathway p-Value q-Value log2FC
   Amino Acids
   Alpha-ketoglutaramate Glutamate Metabolism 0.001 0.019 0.99
   Cysteinylglycine Disulfide Glutathione Metabolism 0.001 0.023 −1.31
   Cysteine-glutathione Disulfide 0.003 0.043 −0.93
   4-guanidinobutanoate Guanidino and Acetamido Metabolism 0.002 0.034
   1.65
   Homocarnosine Histidine Metabolism 0.001 0.017 1.24
   5-aminovalerate Lysine Metabolism 0.002 0.034 0.99
   N-acetyl-cadaverine 0.001 0.024 1.19
   Cystine Methionine, Cysteine, s-Adenosylmethionine, and Taurine
   Metabolism 0.000 <0.001 −2.29
   Phenyllactate Phenylalanine Metabolism <0.001 0.010 0.93
   Phenylpyruvate <0.001 0.011 1.31
   4-hydroxyphenylacetate <0.001 0.000 2.09
   Phenol Sulfate Tyrosine Metabolism 0.003 0.042 −1.93
   4-hydroxyphenylpyruvate 0.001 0.023 0.83
   4-hydroxyphenylacetatoylcarnitine <0.001 0.001 3.61
   Pro-hydroxy-pro Urea cycle; Arginine and Proline Metabolism 0.001 0.023
   −1.57
   Argininate 0.001 0.014 0.66
   2-oxoarginine <0.001 0.001 2.44
   Carbohydrates
   1,5-anhydroglucitol Glycolysis, Gluconeogenesis, and Pyruvate
   Metabolism <0.001 0.006 −0.73
   Glycerate <0.001 0.001 1.03
   Ribose Pentose Metabolism <0.001 <0.001 1.97
   Cofactors and Vitamins
   Threonate Ascorbate and Aldarate Metabolism <0.001 0.002 −0.98
   Alpha-CEHC Sulfate Tocopherol Metabolism <0.001 0.010 −2.03
   Pyridoxal Vitamin B6 Metabolism 0.002 0.028 −0.79
   Lipids
   2-hydroxydecanoate Fatty Acid, Monohydroxy 0.001 0.023 1.48
   1-arachidonoyl-GPA (20:4) Lysophospholipid <0.001 0.008 1.02
   1-oleoyl-GPA (18:1) <0.001 <0.001 2.27
   1-palmitoyl-GPA (16:0) <0.001 <0.001 2.38
   Heptanoate (7:0) Medium Chain Fatty Acid 0.003 0.046 0.80
   Choline Phospholipid Metabolism <0.001 0.001 0.62
   1-(1-enyl-palmitoyl)-2-oleoyl-GPE (P-16:0/18:1) Plasmalogen <0.001
   0.006 0.81
   1-(1-enyl-stearoyl)-2-oleoyl-GPE (P-18:0/18:1) <0.001 0.011 0.92
   Sphingomyelin (d18:1/25:0, d19:0/24:1, d20:1/23:0, d19:1/24:0)
   Sphingolipid Metabolism 0.001 0.020 −1.02
   Sphingomyelin (d18:1/14:0, d16:1/16:0) 0.001 0.025 0.65
   Nucleotides
   Xanthine Purine Metabolism, (Hypo)Xanthine/Inosine-containing <0.001
   0.001 1.53
   Adenosine 5′-monophosphate Purine Metabolism, Adenine-containing <0.001
   0.001 −2.89
   Adenosine <0.001 0.004 −2.45
   Dihydroorotate Pyrimidine Metabolism, Orotate-containing <0.001 0.001
   −1.17
   Uracil Pyrimidine Metabolism, Uracil-containing <0.001 0.008 1.33
   Peptides
   4-hydroxyphenylacetylglutamine Acetylated Peptides <0.001 0.001 1.93
   Phenylacetylthreonine 0.002 0.028 1.96
   4-hydroxyphenylacetylglycine 0.000 0.001 2.10
   Leucylglutamine Dipeptide 0.001 0.024 1.90
   Xenobiotics
   Tartronate (hydroxymalonate) Bacterial/Fungal <0.001 0.008 0.94
   Perfluorooctanesulfonic Acid Chemical 0.002 0.034 −2.11
   S-(3-hydroxypropyl)mercapturic Acid 0.001 0.026 0.76
   Hydroquinone Sulfate Drug–Topical Agents 0.002 0.028 −1.86
   Quinate Food Component/Plant <0.001 0.006 3.16
   [73]Open in a new tab

   The amino acid and lipid superpathways contained the greatest number of
   metabolites that varied significantly between the groups ([74]Figure
   2A). Metabolite set enrichment analysis was performed to identify
   metabolic subpathways that were enriched among significantly variable
   metabolites (the complete results are in [75]Supplementary File S4).
   After controlling for multiple comparisons, the “Phenylalanine
   Metabolism” subpathway (NES = 1.42, p = 0.005, q = 0.17) was found to
   be increased, and the “Purine Metabolism, Adenine containing”
   subpathway (NES = −1.57, p = 0.002, q = 0.12) was decreased in dogs
   with EPI ([76]Figure 2B).

Figure 2.

   [77]Figure 2
   [78]Open in a new tab

   Metabolic pathway enrichment among serum metabolites. (A) Counts of
   significantly variable metabolites from each metabolic superpathway.
   Each bar represents the number of significantly variable metabolites in
   each superpathway. The red bars show the number of metabolites that
   were increased in dogs with EPI compared with healthy controls. The
   blue bars show the number of metabolites that were decreased in dogs
   with EPI compared with healthy controls. (B) Metabolite set enrichment
   analysis of metabolic subpathways (MSEA). MSEA was used to detect
   metabolic subpathways enriched among metabolites with high (positive
   NES) and low (negative NES) stability. The 20 subpathways with the
   highest absolute value of NES are shown. Bars are annotated using the
   FDR-adjusted p-value (q-value), such that grey bars represent pathways
   with q-value ≥ 0.2 and orange bars represent subpathways with q-value <
   0.2.

4. Discussion

   The purpose of this investigation was to identify differences in serum
   metabolomes between healthy dogs and dogs with EPI. Unsupervised
   analysis with PCA and hierarchical clustering of Euclidian distances
   revealed the separation of samples from dogs with EPI and the healthy
   controls, consistent with metabolome-wide differences in serum
   metabolite profiles between the groups. The relative abundances of 143
   serum metabolites varied in association with EPI (p < 0.05), and 114 of
   these were considered statistically significant after controlling for
   false discovery (q < 0.2). Pathway enrichment analysis identified two
   metabolic subpathways that were differentially enriched among
   significantly variable metabolites. The “Purine Metabolism, Adenine
   Containing” subpathway was downregulated in dogs with EPI. Within this
   subpathway, adenosine and adenosine 5′-monophosphate (AMP) were
   significantly lower in the serum of dogs with EPI. Previous studies
   have identified adenosine receptors in pancreatic acinar and ductal
   cells, and purines are implicated in the regulation of exocrine
   pancreatic secretions in rodents, humans, and dogs
   [[79]26,[80]27,[81]28,[82]29,[83]30]. Additionally, adenosine has been
   found to exert an anti-inflammatory effect on intestinal epithelial
   cells, and it is plausible that the decreased availability of adenosine
   could contribute to clinical signs of gastrointestinal dysfunction that
   persist after PERT due to the dysregulation of intestinal mucosal
   inflammatory responses [[84]31].

   Several gut microbial metabolites were found to vary significantly in
   the sera of dogs with EPI compared with the healthy controls. The
   “Phenylalanine Metabolism” subpathway was significantly upregulated in
   association with EPI. This subpathway contains several gut microbial
   metabolites, including phenylpyruvate, 4-hydroxyphenylacetate, and
   phenyllactate, which were all significantly increased in the serum of
   dogs with EPI compared with the healthy controls. These biochemicals
   are microbial catabolites of phenylalanine that are generated by
   bacteria in the genera Lactobacillus and Bifidobacterium
   [[85]32,[86]33,[87]34,[88]35,[89]36,[90]37]. Previous studies of fecal
   microbiomes identified enrichment in genes associated with
   phenylalanine metabolism and increased abundances of Lactobacillus and
   Bifidobacterium in dogs with EPI [[91]38]. The acetylated peptides
   phenylacetylglutamine and 4-hydroxyphenylacetylglutamine were
   significantly increased in dogs with EPI. A recent study integrating
   the deep metagenomic sequencing of gut microbiota with untargeted serum
   metabolomics in humans revealed strong, inverse correlations between
   serum concentrations of these acetylated peptides and the abundance of
   Faecalibacterium prausnitzii, an important commensal bacterium that
   maintains host health [[92]39]. The sulfated benzoate derivates,
   3-methoxycatechol sulfate and 4-hydroxycatechol sulfate, were
   decreased, and o-cresol-sulfate was increased in the sera of dogs with
   EPI. In a previous metagenomic study of human feces, the genes involved
   in benzoate catabolism via catechol were affiliated with diverse clades
   of bacteria including Bacteroidetes, Proteobacteria, Firmicutes, and
   Chloroflexi [[93]40]. Notably, o-cresol sulfate is a well-characterized
   uremic toxin highlighting the interconnected nature of gut microbial
   and host metabolic systems in dogs. Consistent with our finding of the
   variation in numerous gut microbial metabolites in association with
   EPI, previous studies have identified microbiota dysbiosis in dogs with
   EPI characterized by small intestinal bacterial overgrowth, higher
   fecal abundance of E. coli, Lactobacillus, and Bifidobacterium and
   lower fecal abundance of Fusobacterium, C. hiranonis, and
   Faecalibacterium [[94]38,[95]41,[96]42,[97]43,[98]44]. Interestingly,
   differences in fecal microbiomes in some dogs with EPI persist
   following treatment, and fecal microbiomes of dogs with EPI following
   the initiation of PERT are significantly different from healthy dogs
   and similar to those with EPI [[99]38,[100]44]. Consequently, it seems
   likely that these differences in the serum relative abundances of
   microbial metabolites in dogs with EPI are due to changes in gut
   microbiome composition, metabolic function, or both. As all dogs
   enrolled in this study were treated with PERT, our findings suggest
   that pancreatic enzyme supplementation may not be sufficient to restore
   eubiosis in dogs with EPI. These results also suggest that
   phenylalanine derivatives and acetylated peptides may be promising
   markers for enteric microbiota dysbiosis in dogs with EPI.

   The presence of enteric microbiota dysbiosis in dogs with EPI is likely
   secondary to the malabsorption of dietary macromolecules due to the
   insufficient secretion of pancreatic digestive enzymes. However, it is
   possible that enteric microbiota participate in the pathogenesis or
   progression of PAA. In experiments in the 1970s, PAA-like lesions were
   induced in rats fed a liquid elemental diet [[101]45,[102]46]. The
   PAA-like lesions were rare in germ-free rats but invariably present in
   conventional rats. Germ-free rats purposefully contaminated with cecal
   contents from conventional mice also developed PAA-like lesions, but
   those contaminated with sterilized cecal contents did not. These
   findings suggest that PAA was induced by nutritional/metabolic
   disturbances, the impacts of which were at least partially conditioned
   by the presence of commensal enteric microbiota. Whether enteric
   microbiota participate in the pathogenesis of PAA or its progression to
   EPI in dogs is unknown.

   Some of the metabolic disturbances we identified in dogs with EPI are
   similar to those observed during fasting or starvation, including
   greater relative abundances of urea cycle metabolites, purines,
   pyrimidines, and fatty acids [[103]47]. Urea cycle amino acid
   derivatives (arginate, homoarginine, and 2-oxoarginine) were
   significantly lower, and N-acetyl-cadaverine (a lysine derivative) was
   greater in dogs with EPI, consistent with the increased rate of
   deamination reactions due to proteolysis. Greater relative abundances
   of free fatty acids, docosahexaenoylcarnitine (an acylcarnitine), and
   α-ketoglutaramate (a ketogenic amino acid) are consistent with the
   mobilization (lipolysis) of peripheral adipose stores and increased
   mitochondrial beta-oxidation reactions. Interestingly, some of the
   changes in microbial metabolites discussed above also indicate a state
   of malnourishment. In particular, increased plasma phenyllactate and
   urinary p-cresol have been identified in undernourished mice [[104]35].
   Collectively, these findings are consistent with the presence of a
   catabolic state in dogs with EPI, characterized by the mobilization of
   bodily stores of protein and adipose tissue, to maintain energy
   homeostasis.

   Kynurenine, a derivative of tryptophan, was significantly greater in
   the serum of dogs with EPI ([105]Supplementary Files S2 and S3). This
   metabolite is noteworthy owing to its association with intestinal
   mucosal inflammation in humans and rodent models of inflammatory bowel
   disease (IBD). Kynurenine is a mammalian catabolite of tryptophan that
   is generated in hepatic and peripheral tissues, primarily intestinal
   epithelial cells and leukocytes. In peripheral tissues, kynurenine is
   synthesized by indoleamine 2,3-dioxygenase-1 (IDO-1), the rate-limiting
   step in the peripheral kynurenine pathway [[106]48]. The increased
   expression of IDO-1 and increased circulating kynurenine concentrations
   are associated with endoscopic inflammation and disease activity and
   predictive of outcomes in humans with IBD [[107]49,[108]50,[109]51].
   Thus, greater relative abundances of kynurenine are suggestive of the
   presence of intestinal mucosal inflammation in dogs with EPI, a factor
   that could contribute to the pathophysiology of gastrointestinal
   clinical signs, especially those atypical of EPI (e.g., vomiting,
   anorexia). Additional studies of kynurenine and other tryptophan
   catabolites in EPI are currently underway to confirm and further
   elucidate the importance of this finding.

   We compared pancreatic and gastrointestinal biomarkers and identified
   significantly lower serum cobalamin in dogs with EPI compared with
   healthy controls. Cobalamin deficiency has been documented in up to 67%
   of dogs with EPI and is a negative prognostic indicator for therapeutic
   response [[110]4,[111]5]. The persistence of altered serum cobalamin in
   this population of dogs with EPI treated with PERT emphasizes the
   necessity to monitor serum cobalamin concentrations during treatment.
   Persistent cobalamin deficiency following PERT suggests that oral
   pancreatic enzyme supplementation alone is insufficient to restore
   cobalamin homeostasis. A previous study identified lower concentrations
   of lipid-soluble vitamins in this same population of dogs with EPI,
   consistent with persistent deficiency of micronutrient absorption in
   dogs with EPI treated with PERT [[112]3]. We identified evidence of
   other micronutrient disturbances involving vitamin B6 and vitamin C, as
   evidenced by decreased pyridoxal and threonate, respectively. The
   causes for the altered metabolism of these water-soluble vitamins is
   unknown and should be investigated in future studies.

   Our findings should be interpreted in light of several important
   limitations. First, this was an untargeted study intended to reveal
   novel metabolic markers and associations that are plausibly involved in
   the pathophysiology of EPI. Direct causal associations cannot be
   determined from any of these results. Many hundreds of metabolites were
   measured in a relatively small number of dogs, increasing the risk of
   discovering falsely significant features. We attempted to control for
   false discovery, but it is likely that some of the results we report as
   significant are false positives. Controlling the FDR at q < 0.2 means
   that 22/114 significant metabolites were likely to be false
   discoveries. We considered this to be an acceptable rate of false
   discovery given the objectives of the study. Future, hypothesis-driven
   investigations should confirm our findings using targeted analytic
   methods. As the serum samples from the dogs with EPI were collected in
   primary care veterinary clinics, preanalytical conditions could have
   impacted our results. We provided detailed instructions for primary
   care veterinarians to collect, process, store, and ship the samples.
   However, the samples were collected and processed by different
   personnel and in different laboratories. We cannot exclude the
   possibility that differences in the personnel, methods, and equipment
   (e.g., centrifuges) used to process the samples or a deviation from the
   sample collection protocol could have influenced our results. The
   sample sizes of the EPI and healthy control groups were different. We
   cannot exclude the possibility that the difference in group sizes could
   have influenced our results. Future investigations should assess serum
   metabolite concentrations using larger numbers of animals with similar
   sample sizes, and our results could inform such future investigations.

   Regarding the patient population, we were unable to control
   environmental factors, including the dogs’ geographic locations or
   diets. This was a field study of animals with spontaneously occurring
   EPI, and some patients were receiving therapeutic diets as a component
   of therapy. Owing to this, we were not able to control the brands,
   nutrient compositions, quantities, or feeding frequencies of the diets
   fed to dogs enrolled in this study. Previous investigations have
   revealed the significant effects of diet on the serum metabolomes of
   dogs and other animals [[113]52,[114]53,[115]54]. Though we did not
   standardize the diets of dogs in this study, we compared the
   macronutrient composition (protein, fat, and crude fiber) of their
   diets to understand whether there were significant differences between
   the EPI and healthy control groups. We identified a small, but
   statistically significant, difference in the dietary proportions of fat
   between the groups, with the diets of dogs in the EPI group containing
   a higher proportion of fat compared with the healthy controls. There
   were no significant differences in the dietary proportions of protein
   or crude fiber between the groups. However, because energy-basis
   nutrient profiles were not available for all diets, macronutrient data
   on an as-fed basis was analyzed which does not account for differences
   in moisture content or energy density between the diets. Data related
   to the quantities and frequencies of feedings were not collected, so
   their impacts cannot be assessed. Thus, we cannot exclude the
   possibility that differences in diet could have influenced our
   findings. Future investigations are needed to understand the influence
   of different diets on the serum metabolomes of dogs with EPI; these
   future studies should aim to include assessments of diets on an energy
   basis. There were small differences in the mean ages of the EPI and
   healthy controls groups, though the differences were not statistically
   significant. Serum metabolomes are known to vary with age in dogs and
   other mammals [[116]55,[117]56,[118]57]. To understand whether
   differences in age were likely to have influenced our results, we
   attempted to detect correlations among relative abundances of serum
   metabolites and age. Of 114 serum metabolites that varied significantly
   between the EPI and healthy control groups, only 9 were significantly
   correlated with the age ([119]Supplementary File S5). It is unlikely
   that the small differences in age between groups had a significant
   impact on our findings.

   Treatments administered to dogs in the EPI group could have influenced
   our findings. The brands, strengths, and dosages of oral pancreatic
   enzyme extracts administered to dogs in the EPI group were not
   standardized. A study in piglets in which EPI was experimentally
   induced by pancreatic duct ligation, suggests that PERT can result in
   the normalization of enteric microbiota dysbiosis [[120]58]. Thus, it
   is possible that some of our findings could have been influenced by
   inadequate PERT, resulting in some dogs with EPI having persistent
   digestive dysfunction and associated dysbiosis. While we attempted to
   control for the potential impact of PERT on serum metabolomes by
   administering pancreatic enzyme extracts to the healthy controls, it is
   likely that dogs with EPI respond differently to PERT compared with
   healthy dogs. It is also plausible that differences between the dogs
   with EPI and healthy controls could have been influenced by the
   duration of treatment. Dogs with EPI had been treated for months–years
   with PERT, whereas the healthy controls received enzyme extracts for
   only 14 days. We compared serum metabolite profiles in the healthy dogs
   before and after oral pancreatic enzyme administration, but there were
   no significant differences in metabolite abundances between the two
   groups ([121]Supplementary File S6). Though we detected persistent
   clinical signs of gastrointestinal dysfunction in dogs with EPI, the
   clinical assessment of dogs with EPI was not among the objectives of
   this investigation. For this reason, we did not attempt to compare
   their serum metabolomes to dogs with a resolution of clinical signs.
   Lacking clinical assessments (e.g., weight, body condition, fecal
   quality scoring) before and after the initiation of PERT, we were not
   able to make accurate assessments of the clinical responses to PERT in
   dogs with EPI. The determination as to whether dogs had persistent
   clinical signs was based on subjective assessments made by the dogs’
   owners and provided to the investigators during historical interviews.
   Owing to these factors, we concluded that our clinical data were
   insufficient to compare serum metabolomes between dogs with EPI that
   had resolved vs. persistent clinical signs. Future investigations
   should assess changes in the metabolomes of dogs following the
   initiation of PERT using more rigorous clinical assessments collected
   before and after the initiation of PERT. Based on historical interviews
   and reviews of the medical records, antibiotics, probiotics, and
   gastric-acid-suppressing drugs were administered to some dogs in the
   EPI group within the 30 days prior to sample collection. The exact
   brands, dosages, and administration frequencies of these treatments
   were not known for all dogs. These medications are known to impact
   enteric microbial communities in ways that could have affected our
   findings, especially in relation to the metabolites derived from
   enteric microbiota [[122]59,[123]60,[124]61]. Targeted follow-up
   studies in dogs with EPI not exposed to antibiotics and probiotics are
   needed to confirm our findings.

5. Conclusions

   This investigation identified relative abundances of 114 serum
   metabolites that varied significantly between dogs with EPI and healthy
   controls after pancreatic enzyme replacement therapy. The differences
   in amino acids and lipids were consistent with increased proteolysis
   and lipolysis, indicating a persistent catabolic state in dogs with
   EPI. The relative abundances of gut microbial metabolites were altered
   in dogs with EPI, consistent with disruptions in gut microbial
   communities. Increased kynurenine suggests the presence of intestinal
   inflammation in dogs with EPI. Our findings have revealed evidence of
   persistent gastrointestinal dysfunction, enteric microbiota dysbiosis,
   and altered lipid, amino acid, nucleotide, and peptide metabolism in
   dogs with EPI receiving PERT. The extent to which these metabolic
   alterations participate in the pathogenesis or pathophysiology of EPI
   is unknown. These findings should be interpreted with caution and
   should be subject to future, hypothesis-driven research to confirm and
   further elucidate their pathophysiologic significance. Future studies
   should investigate the metabolic disturbances we have identified in
   larger populations of dogs with EPI while attempting to standardize
   diets and therapeutic strategies. Additionally, efforts to understand
   whether the persistence of gastrointestinal clinical signs and
   dysbiosis in dogs with EPI following the initiation of PERT is related
   to inadequate treatment, a concurrent chronic enteropathy, or an
   unidentified facet of its pathophysiology are needed.

Acknowledgments