This site uses cookies, tags, and tracking settings to store information that help give you the very best browsing experience. Dismiss this warning

Untargeted global metabolomic profiling of healthy dogs grouped on the basis of grain inclusivity of their diet and of dogs with subclinical cardiac abnormalities that underwent a diet change

Darcy B. Adin Department of Large Animal Clinical Sciences, College of Veterinary Medicine, University of Florida, Gainesville, FL

Search for other papers by Darcy B. Adin in
Current site
Google Scholar
PubMed Close
 DVM
,
Dana Haimovitz Department of Large Animal Clinical Sciences, College of Veterinary Medicine, University of Florida, Gainesville, FL

Search for other papers by Dana Haimovitz in
Current site
Google Scholar
PubMed Close
 DVM
,
Lisa M. Freeman Department of Clinical Sciences, Cummings School of Veterinary Medicine, Tufts University, North Grafton, MA

Search for other papers by Lisa M. Freeman in
Current site
Google Scholar
PubMed Close
 DVM, PhD
, and
John E. Rush Department of Clinical Sciences, Cummings School of Veterinary Medicine, Tufts University, North Grafton, MA

Search for other papers by John E. Rush in
Current site
Google Scholar
PubMed Close
 DVM

Click on author name to view affiliation information

DOI:
https://doi.org/10.2460/ajvr.22.03.0054
Volume/Issue:
Volume 83: Issue 9
Online Publication Date:
25 Jul 2022
Open access
Sign In Reprints and Permissions

Abstract

OBJECTIVE

To compare metabolomic profiles of dogs eating grain-free (GF) versus grain-inclusive (GI) diets (1) for healthy dogs at baseline and (2) for dogs with subclinical cardiac abnormalities at 12 months after a diet change.

SAMPLE

Serum samples from 23 dogs eating GF diets and 79 dogs eating GI diets, of which 17 (8 eating a GF diet and 9 eating a GI diet) were reevaluated 12 months after a diet change.

PROCEDURES

Metabolomic profiles were developed by means of ultrahigh-performance liquid chromatography–tandem mass spectroscopy of serum samples. Baseline results for the GF group were compared with those for the GI group. Dogs from both groups with subclinical cardiac abnormalities were transitioned to a GI, pulse-free, intervention diet, and samples collected 12 months later were compared between diet groups. Statistical significance for biochemical group differences was defined as P < .05 with a false discovery rate (q) < .10.

RESULTS

Baseline differences in lipid metabolism and amino acid metabolism were found between the GF and GI diet groups. There were 46 metabolites that were higher and 82 metabolites that were lower in the GF group (n = 23), compared with the GI group (79). Comparison of the GF (n = 8) and GI (9) groups 12 months after the diet change showed only 6 metabolites that were higher and 11 metabolites that were lower in the GF group, compared with the GI group.

CLINICAL RELEVANCE

Metabolomic pathway differences between dogs eating GF versus GI diets highlight the important effect of diet in metabolomics analyses. The clinical importance of these differences and how they might relate to cardiac disease in dogs remains undetermined.

Abstract

OBJECTIVE

To compare metabolomic profiles of dogs eating grain-free (GF) versus grain-inclusive (GI) diets (1) for healthy dogs at baseline and (2) for dogs with subclinical cardiac abnormalities at 12 months after a diet change.

SAMPLE

Serum samples from 23 dogs eating GF diets and 79 dogs eating GI diets, of which 17 (8 eating a GF diet and 9 eating a GI diet) were reevaluated 12 months after a diet change.

PROCEDURES

Metabolomic profiles were developed by means of ultrahigh-performance liquid chromatography–tandem mass spectroscopy of serum samples. Baseline results for the GF group were compared with those for the GI group. Dogs from both groups with subclinical cardiac abnormalities were transitioned to a GI, pulse-free, intervention diet, and samples collected 12 months later were compared between diet groups. Statistical significance for biochemical group differences was defined as P < .05 with a false discovery rate (q) < .10.

RESULTS

Baseline differences in lipid metabolism and amino acid metabolism were found between the GF and GI diet groups. There were 46 metabolites that were higher and 82 metabolites that were lower in the GF group (n = 23), compared with the GI group (79). Comparison of the GF (n = 8) and GI (9) groups 12 months after the diet change showed only 6 metabolites that were higher and 11 metabolites that were lower in the GF group, compared with the GI group.

CLINICAL RELEVANCE

Metabolomic pathway differences between dogs eating GF versus GI diets highlight the important effect of diet in metabolomics analyses. The clinical importance of these differences and how they might relate to cardiac disease in dogs remains undetermined.

Investigations into a potential link between consumption of grain-free (GF), high-pulse diets, and the development of a dilated cardiomyopathy (DCM) phenotype in dogs have been ongoing since 2018.1–7 Pulse ingredients are defined as dry-harvested legume crops and include lentils, peas, chickpeas, and a variety of beans.8 Recent data suggest that high-pulse canine diets might be associated with myocardial disease in some dogs.1–6 Grain-free diets, which are usually high in pulse content, have been associated with low-level increases in serum concentrations of high-sensitivity cardiac troponin I, possibly indicating subclinical cardiomyocyte injury.4 Similarly classified nontraditional diets that were also high in pulse content have been shown in 2 studies1,7 of overtly healthy dogs to be associated with cardiac enlargement and reduced systolic function. The exact cause has not yet been identified.

Interest in the use of untargeted metabolomic profiling to understand and improve the health of dogs has grown in recent years.9 Metabolomic profiling offers a unique tool for the evaluation of small molecules in biological samples that could be used to identify biomarkers of disease or diet, possibly differentiating between dogs eating GF versus grain-inclusive (GI) diets.10 Specific markers of nutrient intake in people have been identified through this approach, and several studies11–14 in dogs have identified differences in the metabolome as a result of certain dietary ingredients such as fish and raw meat. Metabolomic assessments of healthy dogs grouped on the basis of dietary properties related to grain and pulse inclusivity have not yet been reported. A recent study15 applied a metabolomics approach to dog food and showed that diets seemingly associated with DCM differed biochemically from diets seemingly not associated with DCM, but similar studies of blood samples from dogs are lacking. Additionally, metabolomic profiling of dogs before and after a diet change could improve our understanding of how disease improvement or reversal occurs after a diet change, which is a unique feature of diet-associated DCM.2,3,5,6

The objectives of the study reported were to investigate whether the biochemical footprint of apparently healthy dogs differed on the basis of diet type being consumed and to determine whether a diet change would alter the metabolomic profile in dogs with subclinical cardiac abnormalities. We hypothesized that dogs eating GF diets would differ biochemically from dogs eating GI diets and that fewer differences would be found after a diet change.

Materials and Methods

The study involved untargeted global metabolomic profiling of serum samples that had been frozen at –80°C as part of a larger study.4,16 Sample collection had been approved by the University of Florida College of Veterinary Medicine’s Institutional Care and Use Committee (No. 201810504 and No. 202110504), and client consent had been obtained.

Briefly, healthy dogs of 3 breeds (Doberman Pinscher, Golden Retriever, and Miniature Schnauzer) that had undergone echocardiography and measurement of serum cardiac biomarker (N-terminal pro-B-type natriuretic peptide and high-sensitivity cardiac troponin I) concentrations were assigned to 2 groups on the basis of grain inclusivity of the diet (GF or GI) the dog had been eating for a minimum of 6 months before study enrollment. Serum samples were obtained at baseline from all dogs and stored at –80°C until untargeted global metabolomic analysis was performed. Owners of dogs found to have echocardiographic abnormalities (normalized left ventricular diameter in diastole > 1.8, normalized left ventricular diameter in systole > 1.2, or fractional shortening < 25%), an N-terminal pro-B-type natriuretic peptide concentration higher than the laboratory-provided reference range (> 735 pmoL/L for Doberman Pinschers and > 900 pmoL/L for Golden Retrievers and Miniature Schnauzers), or a high-sensitivity cardiac troponin I concentration higher than the laboratory-provided reference range (> 0.06 ng/mL) were offered participation in the longitudinal arm of the study if the owner was willing to change the diet to 1 of 6 extruded, GI, pulse-free, intervention diets. Serum was collected from these dogs 12 months after the diet change.16

Serum samples were transported to Metabolon Inc overnight on dry ice and kept at –80°C until untargeted global metabolomic profiling could be performed. Metabolic profiling was conducted by means of standard methods (Supplementary Appendix). Metabolites were quantified by means of ultrahigh-performance liquid chromatography–tandem mass spectroscopy and identified by comparison to a reference library of 4,500 purified standards identified on the basis of retention time, molecular weight, mass–charge ratio, and mass spectroscopy spectral data.

Statistical analysis

To allow for a high-level view of the dataset, principal component analysis was performed by constructing linear mixtures of unrelated metabolites to generate the uncorrelated components that contributed the most to dataset variability at baseline. A χ2 test was used to compare breed distribution between baseline diet groups. Data were log transformed, and metabolites were compared between groups with Welch 2-sample t tests (GF diet group compared with GI diet group at each time point) and paired t tests (baseline samples compared with 12-month samples) with standard software (Jupyter Notebook [The Jupyter Project] and R [The R Project]). Baseline samples from dogs eating GF diets were compared with baseline samples from dogs eating GI diets. Samples obtained 12 months after the transition to an intervention diet were compared between dogs in the GF group that had subclinical cardiac abnormalities and underwent the diet change and dogs in the GI group that had subclinical cardiac abnormalities and underwent the diet change. Separately for the GF and GI groups, baseline samples from dogs with subclinical cardiac abnormalities were compared with samples obtained 12 months after the diet change.

Statistical significance was defined as values of P < .05 and values of q (false discovery rate) < .10, which was considered to provide an acceptable degree of confidence (< 10% chance) that there were no false discoveries. This estimate of the false discovery rate (q value) was calculated to account for multiple comparisons inherent to metabolomic-based studies.17 Random forest analysis was used to define metabolites that contributed the most to group binning when comparing groups (irrespective of statistical significance), and these were displayed as biochemical importance plots.

Results

For the present study, baseline serum samples from 23 dogs eating GF diets (7 Doberman Pinschers, 8 Golden Retrievers, and 8 Miniature Schnauzers) and 79 dogs eating GI diets (30 Doberman Pinschers, 35 Golden Retrievers, and 14 Miniature Schnauzers) were analyzed (Figure 1). In addition, blood samples obtained 12 months after changing to a GI, pulse-free, intervention diet were available for 17 dogs (8 that were originally eating a GF diet and 9 that were originally eating a GI diet). Clinical details of these dogs and the diets have been previously reported.4,16

Figure 1
Figure 1

Flow chart of study enrollment for dogs included in a study designed to compare metabolomic profiles of dogs eating grain-free (GF) versus grain-inclusive (GI) diets for healthy dogs at baseline and for dogs with subclinical cardiac abnormalities at 12 months after a diet change. DP = Doberman Pinscher. GR = Golden Retriever. MS = Miniature Schnauzer.

Citation: American Journal of Veterinary Research 83, 9; 10.2460/ajvr.22.03.0054

Untargeted global metabolomic analysis detected 881 metabolites in these dogs, of which 832 were named (ie, had a known structural identity) and 49 were unnamed (ie, had an unknown structural identity). Principal component analysis did not reveal high-level diet group differences but did show some separation by breed (Supplementary Figure S1). Breed distribution was not significantly (P = .22) different between the GF and GI groups.

Diet group differences at baseline

There were 46 metabolites that were significantly higher and 82 metabolites that were significantly lower in the GF group than the GI group (Table 1). Notable metabolite differences between diet groups were found for some biochemical pathways. Urea cycle intermediates (eg, homoarginine, N-acetylcitrulline, N-delta-acetylornithine, and N2,N5-diacetylornithine) were higher in the GF group than the GI group. Metabolites related to glutathione synthesis and turnover (eg, cysteinylglycine and gamma-glutamylglutamine) were lower in the GF group than the GI group. Succinylcarnitine (C4-DC), which is related to glycolysis and energy production, was lower in the GF group than the GI group. Some metabolites related to phospholipid metabolism (eg, 1-palmitoyl-2-palmitoleoyl-GPC [16:0/16:1], 1-palmitoyl-2-oleoyl-GPC [16:0/18:1], 1-palmitoyl-2-arachidonoyl-GPC [16:0/20:4n6], and 1-palmitoyl-2-oleoyl-GPE [16:0/18:1]) were lower in the GF group than the GI group, whereas 1-linoleoyl-2-linolenoyl-GPC (18:2/18:3) was significantly higher in the GF group than the GI group. Microbiome-associated metabolites related to aromatic amino acids (eg, kynurenate and anthranilate) were significantly lower in the GF group than the GI group. Several vitamin A and B6 metabolites such as retinol (vitamin A), carotene diol (2), and pyridoxal were significantly lower in the GF group than the GI group. Metabolites related to pyrimidine nucleotides (eg, pseudouridine, cytidine, and 2'-deoxycytidine) were significantly lower in the GF group than the GI group. Notably, taurine was not significantly different between diet groups, and only 2 of 23 metabolites related to taurine pathways (cystathionine and cysteine sulfinic acid) were different between groups (ie, significantly higher in the GF group than the GI group).

Table 1

Metabolites that were significantly different between diet groups at baseline, as determined by untargeted global metabolomic profiling, for 23 dogs eating grain-free (GF) diets and 79 dogs eating grain-inclusive (GI) diets.
Metabolite Fold change P value q value
X-25419 6.65 < .0001 < .0001
Trigonelline (N'-methylnicotinate) 2.91 < .0001 < .0001
Homoarginine 3.68 < .0001 < .0001
X-26008 9.60 < .0001 .0001
Tryptophan betaine 14.86 < .0001 .0001
1-Linolenoyl-GPC (18:3) 1.86 < .0001 .0004
Maltol sulfate 3.40 < .0001 .0004
1-Linoleoyl-2-linolenoyl-GPC (18:2/18:3) 1.90 .0001 .0038
1-Palmitoleoyl-2-linolenoyl-GPC (16:1/18:3) 1.60 .0001 .0041
N2,N5-diacetylornithine 5.25 .0001 .0054
Argininate 8.86 .0002 .0070
Cysteine sulfinic acid 1.25 .0002 .0072
N-delta-acetylornithine 5.02 .0003 .0085
1-Lignoceroyl-GPC (24:0) 1.33 .0003 .0085
X-11795 1.35 .0003 .0085
1-Stearoyl-2-oleoyl-GPI (18:0/18:1) 1.27 .0006 .0129
Valylglycine 3.28 .0008 .0157
5-HEPE 1.50 .0018 .0250
Ethyl beta-glucopyranoside 2.18 .0022 .0286
Linolenate [alpha or gamma; (18:3n3 or 6)] 1.93 .0032 .0356
N-acetylasparagine 1.22 .0035 .0388
Quinate 15.47 .0050 .0464
1-Oleoyl-GPI (18:1) 1.41 .0054 .0482
Linolenoylcarnitine (C18:3) 1.35 .0074 .0579
Eicosenoate (20:1) 1.43 .0082 .0611
18-Methylnonadecanoate (i20:0) 1.27 .0101 .0679
1-Stearoyl-2-linoleoyl-GPI (18:0/18:2) 1.18 .0101 .0679
2'-Deoxyuridine 1.16 .0105 .0680
1-Oleoyl-GPC (18:1) 1.15 .0108 .0680
Cis-3,4-methyleneheptanoylcarnitine 2.33 .0111 .0693
1-Eicosapentaenoylglycerol (20:5) 1.27 .0116 .0706
Indoleacrylate 2.47 .0128 .0751
1-Oleoyl-2-docosahexaenoyl-GPC (18:1/22:6) 1.31 .0132 .0751
2-Oxoarginine 1.52 .0153 .0813
Solanidine 6.43 .0158 .0831
Imidazole lactate 1.22 .0167 .0871
Methylnaphthyl sulfate (2) 1.74 .0173 .0874
Arachidate (20:0) 1.33 .0181 .0874
Gamma-tocopherol/beta-tocopherol 3.12 .0184 .0874
N-acetylcitrulline 1.77 .0186 .0874
Glycerophosphoinositol 1.15 .0189 .0875
X-11478 4.98 .0191 .0875
Cystathionine 1.38 .0217 .0955
Glycerol 3-phosphate 1.14 .0231 .0993
1-Linolenoylglycerol (18:3) 1.82 .0234 .0994
Gamma-CEHC sulfate 1.51 .0236 .0996
N-monomethylarginine 0.68 < .0001 < .0001
2-Oxindole-3-acetate 0.37 < .0001 .0001
2'-Deoxycytidine 0.71 < .0001 .0004
Cytidine 0.72 .0001 .0031
Pyridoxal 0.61 .0002 .0054
1-Palmitoyl-GPE (16:0) 0.77 .0002 .0061
X-25371 0.71 .0003 .0085
4-Vinylphenol sulfate 0.41 .0004 .0089
Pyridoxate 0.53 .0005 .0104
Gamma-glutamylvaline 0.64 .0006 .0121
Sphingomyelin (d18:0/18:0, d19:0/17:0) 0.63 .0009 .0169
3,5-Dichloro-2,6-dihydroxybenzoic acid 0.44 .0010 .0182
Kynurenate 0.52 .0012 .0212
X-25247 0.27 .0012 .0213
Gamma-glutamylleucine 0.79 .0014 .0228
N-acetylalanine 0.88 .0015 .0244
Gamma-glutamylglycine 0.81 .0016 .0244
Anthranilate 0.69 .0017 .0244
Gamma-glutamyltyrosine 0.73 .0017 .0244
Retinal 0.70 .0017 .0244
1,5-Anhydroglucitol (1,5-AG) 0.72 .0018 .0250
Succinylcarnitine (C4-DC) 0.74 .0021 .0286
1-Palmitoyl-2-oleoyl-GPC (16:0/18:1) 0.89 .0023 .0289
4-Vinylguaiacol sulfate 0.10 .0025 .0310
Glycine 0.85 .0026 .0318
Betaine 0.84 .0031 .0356
Leucine 0.88 .0031 .0356
Gamma-glutamylglutamine 0.87 .0032 .0356
Palmitoylcarnitine (C16) 0.76 .0040 .0433
Stachydrine 0.49 .0042 .0440
2,6-Dihydroxybenzoic acid 0.56 .0043 .0447
N1-methyl-2-pyridone-5-carboxamide 0.58 .0049 .0464
X-13695 0.51 .0049 .0464
Tyrosine 0.85 .0050 .0464
N-stearoyl-sphinganine (d18:0/18:0) 0.63 .0050 .0464
Retinol (vitamin A) 0.78 .0050 .0464
5,6-Dihydrouridine 0.85 .0053 .0479
1-Palmitoyl-2-arachidonoyl-GPC (16:0/20:4n6) 0.86 .0056 .0489
Cysteinylglycine 0.75 .0058 .0493
Pseudouridine 0.88 .0058 .0493
3-Bromo-5-chloro-2,6-dihydroxybenzoic acid 0.48 .0060 .0499
1-Palmitoyl-2-linoleoyl-GPC (16:0/18:2) 0.88 .0068 .0562
Palmitoyl dihydrosphingomyelin (d18:0/16:0) 0.84 .0070 .0569
1-Methylnicotinamide 0.57 .0072 .0573
N2,N2-dimethylguanosine 0.88 .0075 .0579
Carotene diol (2) 0.49 .0076 .0579
1-Methylhistamine 0.68 .0081 .0611
Proline 0.85 .0084 .0617
1-(1-Enyl-palmitoyl)-GPE (P-16:0) 0.77 .0085 .0619
N-acetylglucosamine/N-acetylgalactosamine 0.82 .0088 .0627
5-Dodecenoylcarnitine (C12:1) 0.68 .0098 .0679
1-(1-Enyl-palmitoyl)-2-arachidonoyl-GPE (P-16:0/20:4) 0.76 .0098 .0679
Alpha-CEHC glucuronide 0.63 .0100 .0679
Sphingosine 1-phosphate 0.85 .0107 .0680
X-25422 0.80 .0107 .0680
3-Decenoylcarnitine 0.55 .0108 .0680
Cys-gly, oxidized 0.73 .0113 .0698
1-(1-Enyl-stearoyl)-2-linoleoyl-GPE (P-18:0/18:2) 0.80 .0122 .0736
2-Aminophenol sulfate 0.55 .0124 .0736
X-11843 0.38 .0131 .0751
1-Palmitoyl-2-linoleoyl-GPE (16:0/18:2) 0.75 .0132 .0751
Dodecanedioate (C12-DC) 0.25 .0133 .0751
Nicotinamide 0.36 .0134 .0751
1,2-Dipalmitoyl-GPC (16:0/16:0) 0.86 .0138 .0761
Glycosyl-N-stearoyl-sphingosine (d18:1/18:0) 0.79 .0143 .0783
5-Methylnorleucine 0.70 .0145 .0788
Ethylmalonate 0.79 .0148 .0792
[N(1) + N(8)]-acetylspermidine 0.72 .0173 .0874
1-Palmitoyl-2-palmitoleoyl-GPC (16:0/16:1) 0.83 .0174 .0874
1-(1-Enyl-palmitoyl)-2-palmitoleoyl-GPC (P-16:0/16:1) 0.81 .0176 .0874
N-palmitoylglycine 0.69 .0177 .0874
Gamma-glutamylphenylalanine 0.87 .0183 .0874
X-12411 0.51 .0185 .0874
N-palmitoyltaurine 0.57 .0186 .0874
1-Stearoyl-GPE (18:0) 0.89 .0190 .0875
1-Palmitoyl-2-linoleoyl-GPA (16:0/18:2) 0.76 .0204 .0925
Phenol glucuronide 0.56 .0207 .0931
1-Palmitoyl-2-oleoyl-GPE (16:0/18:1) 0.80 .0213 .0949
Guanidinoacetate 0.80 .0214 .0949
Octadecanedioylcarnitine (C18-DC) 0.82 .0228 .0992
Alanine 0.87 .0231 .0993
16-Hydroxypalmitate 0.73 .0238 .0996

Metabolites with a fold change > 1.0 were higher in the GF group than in the GI group, and metabolites with a fold change < 1.0 were lower in the GF group than in the GI group. The q value represents the false discovery rate.

Random forest analysis computed the top 30 metabolites that differentiated the GF and GI groups at baseline with 93% predictive accuracy (Figure 2). Most of these metabolites suggested key differences in lipid metabolism (9 metabolites) and amino acid metabolism (7 metabolites). The top differentiating metabolite was unnamed (X-25419), and 3 other unnamed metabolites were part of this biochemical importance plot.

Figure 2
Figure 2

Biochemical importance plot, developed by means of random forest analysis, showing the top 30 metabolites that contributed most to group binning (GF or GI diet group) at baseline for 23 dogs eating GF diets and 79 dogs eating GI diets. Predictive accuracy was 93%. The super pathway for each metabolite is indicated by color.

Citation: American Journal of Veterinary Research 83, 9; 10.2460/ajvr.22.03.0054

Effect of diet change

Comparison of diet groups at 12 months

Comparison of dogs with subclinical abnormalities in the GF group (1 Doberman Pinscher, 4 Golden Retrievers, and 3 Miniature Schnauzers) and GI group (3 Doberman Pinschers and 6 Golden Retrievers) 12 months after the diet change found 6 metabolites that were higher and 11 metabolites that were lower in the GF group than the GI group (Table 2). Only 3 of these 17 metabolites were also significantly different between diet groups at baseline, and all 3 of these were unnamed (X-25419 was higher in the GF group and X-25247 and X-13695 were lower in the GF group at both time points).

Table 2

Metabolites that were significantly different between diet groups 12 months after a diet change, as determined by untargeted global metabolomic profiling, for 8 dogs eating GF diets and 9 dogs eating GI diets before the diet change.
Metabolite Fold change P value q value
X-25419 6.07 .0003 .0642
Myristoleate (14:1n5) 2.19 .0004 .0642
X-25420 8.62 .0004 .0642
Palmitoleoylcarnitine (C16:1) 1.74 .0012 .0901
X-25417 3.04 .0020 .0901
S-methylcysteine sulfoxide 1.39 .0026 .0992
Uridine 5'-monophosphate (UMP) 0.32 .0004 .0642
2R,3R-dihydroxybutyrate 0.64 .0008 .0901
1-(1-Eenyl-stearoyl)-2-arachidonoyl-GPE (P-18:0/20:4) 0.63 .0014 .0901
1,2-Dilinoleoyl-GPC (18:2/18:2) 0.57 .0015 .0901
Gamma-glutamylthreonine 0.64 .0017 .0901
12-HHTrE 0.24 .0017 .0901
X-25247 0.22 .0017 .0901
Threonine 0.63 .0018 .0901
4-Ethylphenylsulfate 0.26 .0019 .0901
Thromboxane B2 0.24 .0023 .0973
X-13695 0.33 .0025 .0992

Comparison of baseline and 12 months after diet change within each diet group

No metabolites were significantly different between baseline and 12 months after diet change for the GF group or for the GI group (Supplementary Tables S1 and S2).

For dogs in the GF group, the biochemical importance plot for the within-group time point comparison showed high predictive accuracy (87%), with the top-ranking metabolites suggesting key differences 12 months after diet change related to lipid metabolism (14 metabolites), cofactors and vitamins (4 metabolites), and xenobiotics (4 metabolites; Figure 3).

Figure 3
Figure 3
Figure 3

Biochemical importance plots, developed by means of random forest analysis, showing the top 30 metabolites that contributed most to group binning (baseline vs 12 months after diet change) for 8 dogs with subclinical cardiac abnormalities fed a GF diet at baseline that were switched to a GI, pulse-free, intervention diet for 12 months (A) and for 9 dogs with subclinical cardiac abnormalities fed a GI diet at baseline that were switched to a GI, pulse-free, intervention diet for 12 months (B). Predictive accuracy was 87% (A) and 60% (B). The super pathway for each metabolite is indicated by color.

Citation: American Journal of Veterinary Research 83, 9; 10.2460/ajvr.22.03.0054

For dogs in the GI group, the biochemical importance plot for the within-group time point comparison suggested key differences after diet change for metabolites related to lipid metabolism (14 metabolites) and amino acid metabolism (5 metabolites), with 60% predictive accuracy (Figure 3).

Discussion

Untargeted global metabolomic analysis of the healthy dogs in this study revealed numerous biochemical differences between dogs eating GF diets and those eating GI diets. Most of the biochemical differences between groups were related to lipid metabolism and amino acid metabolism and were statistically significant at a low false discovery rate. More granular group differences were related to pathways involving urea cycle intermediates, glutathione synthesis and turnover, phospholipid metabolism, vitamins and cofactors, and pyrimidine nucleotides. The results of this study showed some similarities to a publication that applied metabolomic profiling to food samples (foodomics).15 For example, the food-based study also found lower vitamins and cofactors and higher xenobiotics in diets associated with DCM, compared with diets not associated with DCM, which was similar to our findings for these metabolites in the blood of dogs eating GF and GI diets.15 While it is not surprising that biochemical compounds found in food would be detectable in the blood of dogs, the metabolomic profile is not expected to be identical to the foodomic profile because of changes that biochemical compounds undergo associated with absorption and metabolism.

Several unnamed compounds were also found to be biomarkers of GF diet ingestion, but their identity and clinical importance are not known. The top differentiating metabolite (X-25419, which, in the GF group, was 6.65 times that in the GI group at baseline) was reported to be higher by a factor of 7.67 fold in dog foods associated with DCM, compared with dog foods that have not been associated with DCM in the previous foodomics study, suggesting that diet is the source of this metabolite in the blood of dogs in the GF group.15 This unnamed compound was also present in GI diets but to a lesser degree. A study18,19 in people identified X-25419 in blood and urine to be associated with fish and olive oil ingestion, and a study20 of dogs with mitral valve disease found that X-25419 was significantly different between healthy dogs predisposed to heart disease and dogs with mild heart disease. Fish is a common protein source for many types of dog diets, but diet was not reported in the dog study.18 Until structural elucidation is performed, speculation about the clinical importance of X-25419 is not possible.

The present study cannot be used to determine whether any of these biochemical differences (involving named or unnamed metabolites) predispose some dogs that eat GF diets to cardiac abnormalities, but it demonstrated that ingestion of GF diets affects the biochemical footprint of dogs. Therefore, the impact of diet should be considered in the design and analysis of future studies evaluating this clinical problem and of metabolomic studies in companion animals in general. Some of the super pathways that were different between diet groups in this study might be a basis for hypothesis generation in future studies.

We evaluated the effect of transition to an intervention diet in a subset of dogs found to have subclinical cardiac abnormalities by looking for between-group differences at baseline and again 12 months after a diet change and by looking at within-group time point comparisons (ie, baseline vs 12 months after diet change for each group). Fewer diet group biochemical differences were found between the GF and GI groups 12 months after this subgroup of dogs with subclinical cardiac abnormalities transitioned from their original diet to an intervention diet. Most metabolite differences between groups at baseline were no longer found 12 months after diet change, with the exception of 3 unnamed metabolites, which remained significantly different (2 remained lower and 1 remained higher in the GF group). The clinical importance of these unnamed compounds for which group differences persisted cannot be determined until the compounds are structurally elucidated. The loss of most metabolomic group distinctions at 12 months was likely attributable to the dietary unification provided by the intervention diets; however, the clinical importance of these changes is unknown. Some of these changes, such as higher vitamin and cofactor metabolites after diet change, might benefit cardiac health, but this requires further research. The finding of new metabolites that differentiated diet groups at 12 months but had not differentiated diet groups at baseline might be a result of the comparison involving a smaller group of dogs with subclinical cardiac abnormalities at the 12-month time point, compared with larger group of dogs (most of which did not have subclinical cardiac abnormalities) at baseline. Therefore, this comparison might reflect metabolomic differences between these 2 groups that were unrelated to diet or due to other factors that could have varied between groups (eg, environmental, genetic, microbiomic, and residual cardiac abnormalities).

When within-group comparisons of dogs with subclinical cardiac abnormalities were performed for each individual diet group before and 12 months after a diet change, no metabolite differences were statistically significant. Low statistical power could have impacted these results, because a relatively small number of dogs in each group were assessed before and after diet change (8 dogs in the GF group and 9 dogs in the GI group). Random forest analysis, which ranks the top metabolites contributing to group differences regardless of statistical significance, showed a higher degree of predictive accuracy for these metabolites before and after a diet change for the GF group (87%) than for the GI group (60%). Fewer changes for dogs in the GI group after diet change could be explained by the fact that the intervention diets were similar to the GI diets in their grain inclusivity and pulse content. Key differences after the diet change for the GF group were related to lipid metabolism, cofactors and vitamins, and xenobiotics. These changes were probably largely due to the diet change, but their clinical importance is uncertain. The relationship between these metabolomic changes and the potential for disease predisposition or resolution cannot be determined from this study, but these findings can provide some hypotheses to drive future research.

A strength of the present study was the absence of advanced cardiac disease, which can change the metabolome.18 However, it is important to recognize that this methodology is only a biochemical snapshot in time. There are several limitations that should be considered for study interpretation. The diet groups were relatively unbalanced, with more dogs in the GI group, because they reflected the population of dogs presenting for study enrollment. The time point comparisons might have suffered from low power and breed imbalance because of the small number of overtly healthy dogs that were found to have subclinical cardiac abnormalities and underwent a diet change with subsequent reanalysis of metabolomics. The complex interplay of diet and health is not fully understood, and there could be other modifying influences that determine the effect of diet on the metabolome, including the microbiome, environmental influences, and genetic influences. Some of these factors could also lead to changes in gene expression, which could affect the metabolites detected in the blood. The dogs in the present study were overtly healthy, but we did not account for environmental factors. High-level analysis with principal component analysis suggested some metabolomic profile separation by breed for all dogs independent of diet, but this approach was not able to identify the specific metabolites or pathways contributing to these differences. Although breed influence was limited in this study and breed distribution was not significantly different between diet groups at baseline, this finding adds to other reports9 of metabolomic breed differences, supporting breed consideration in future studies. The present study provides unique information, but we could not determine whether the observed biochemical differences were simply markers of dietary ingredients and their metabolites, were disease promoting, or were both. Likewise, the observed metabolite differences between diet groups indicated the biochemical image at the time of sampling and the synthesis or degradation of other compounds that led to these metabolites are unknown and could be important. Therefore, metabolomics can be viewed as a footprint, but the trail leading to it is not always elucidated. Nevertheless, some of the biochemical differences between diet groups were clustered in major areas of lipid metabolism and amino acid metabolism and in specific super pathways, such as urea cycle metabolism, redox homeostasis, glycolysis, phospholipid metabolism, vitamins, and pyrimidine nucleotide metabolism, which is helpful information for planning future studies. Finally, 5% of metabolites could not be structurally identified, several of which were differentiators of GF diet ingestion. Their importance, or lack of importance, cannot be determined without structural elucidation, and without knowledge of the chemical structure, it is not possible to conclude whether these metabolites could affect health or are incidental markers of the diet.

In conclusion, this study found significant differences in the biochemical footprint of dogs eating GF versus GI diets. Most of the differentiating metabolites at baseline were not differentiators of the original diet group 12 months after a diet change for dogs with subclinical cardiac abnormalities, supporting the role of diet in metabolomic modulation. The clinical importance of baseline diet group differences and temporal changes was not determined in this study, but the results highlight the need to consider diet type in metabolomics studies in dogs.

Supplementary Materials

Supplementary materials are posted online at the journal website: avmajournals.avma.org

Acknowledgments

This project was supported by the American Kennel Club Canine Health Foundation (grant No. 02661). The contents of this publication are solely the responsibility of the authors and do not necessarily represent the views of the Foundation. Funding sources did not have any involvement in the study design, data analysis and interpretation, or writing and publication of the manuscript.

Dr. Adin has received support from Nestlé Purina PetCare and is a consultant and sponsored lecturer for Ceva Animal Health and Boehringer Ingelheim. Dr. Freeman has received research or residency funding from, given sponsored lectures for, or provided professional services to Aratana Therapeutics, Elanco, Guiding Stars Licensing Co LLC, Hill’s Pet Nutrition, Nestlé Purina PetCare, P&G Petcare (now Mars), and Royal Canin. Dr. Rush has received funding from, given sponsored lectures for, or provided professional services to Aratana Therapeutics, Elanco, Hill’s Pet Nutrition, Nestlé Purina PetCare, Royal Canin, Idexx, and Boehringer Ingelheim. Dr. Haimovitz was a student representative for Nutramax Laboratories Inc. None of these companies had any role or involvement in data collection, data interpretation, or manuscript preparation.

The authors thank Melissa Pisaroglo de Carvalho, Lana Fagman, Courtney Hanner, and Viviana Heinekin for technical assistance and thank Dr. Michael Aherne for case enrollment.

References

  • 1.↑

    Ontiveros ES, Whelchel BD, Yu J, et al. Development of plasma and whole blood taurine reference ranges and identification of dietary features associated with taurine deficiency and dilated cardiomyopathy in golden retrievers: prospective, observational study. PLoS One. 2020;15(5):e0233206.

    • PubMed
    • Search Google Scholar
    • Export Citation
  • 2.↑

    Kaplan JL, Stern J, Fascetti AJ, et al. Taurine deficiency and canine dilated cardiomyopathy in golden retrievers fed commercial diets. PLoS One. 2018;13(12):e0209112.

    • PubMed
    • Search Google Scholar
    • Export Citation
  • 3.↑

    Adin D, DeFrancesco TC, Keene B, et al. Echocardiographic phenotype of canine dilated cardiomyopathy differs based on diet type. J Vet Cardiol. 2019;21:1–9.

    • PubMed
    • Search Google Scholar
    • Export Citation
  • 4.↑

    Adin D, Freeman L, Kellihan H, et al. Effect of type of diet on blood and plasma taurine concentrations, cardiac biomarkers, and echocardiograms in 4 dog breeds. J Vet Intern Med. 2021;35:771–779.

    • PubMed
    • Search Google Scholar
    • Export Citation
  • 5.↑

    Walker AL, DeFrancesco TC, Bonagura JD, et al. Association of diet with clinical outcomes in dogs with dilated cardiomyopathy and congestive heart failure. J Vet Cardiol. 2022;40:99–109.

    • PubMed
    • Search Google Scholar
    • Export Citation
  • 6.↑

    Freid KJ, Freeman LM, Rush JE, et al. Retrospective study of dilated cardiomyopathy in dogs. J Vet Intern Med. 2021;35(1):58–67.

  • 7.↑

    Owens E, LeBlanc N, Freeman L, Scollan K. Comparison of echocardiographic measurements and cardiac biomarkers in healthy dogs eating non-traditional or traditional diets (abstract). J Vet Intern Med. 2021;35(6):2955–2956. doi:10.1111/jvim.16220

    • Search Google Scholar
    • Export Citation
  • 8.↑

    Mansilla WD, Marinangeli CPF, Ekenstedt KJ, et al. Special topic: the association between pulse ingredients and canine dilated cardiomyopathy: addressing the knowledge gaps before establishing causation. J Anim Sci. 2019;97(3):983–997.

    • PubMed
    • Search Google Scholar
    • Export Citation
  • 9.↑

    Carlos G, dos Santos FP, Fröehlich PE. Canine metabolomics advances. Metabolomics. 2020;16(2):16.

  • 10.↑

    Gibbons H, Brennan L. Metabolomics as a tool in the identification of dietary biomarkers. In: Proceedings of the Nutrition Society. Vol 76. Cambridge University Press; 2017:42–53.

    • PubMed
    • Search Google Scholar
    • Export Citation
  • 11.↑

    Boretti FS, Burla B, Deuel J, et al. Serum lipidome analysis of healthy beagle dogs receiving different diets. Metabolomics. 2020;16(1):1.

    • Search Google Scholar
    • Export Citation
  • 12.

    Rafiq T, Azab SM, Teo KK, et al. Nutritional metabolomics and the classification of dietary biomarker candidates: A critical review. Adv Nutr. 2021;12(6):2333–2357.

    • PubMed
    • Search Google Scholar
    • Export Citation
  • 13.

    Schmidt M, Unterer S, Suchodolski JS, et al. The fecal microbiome and metabolome differs between dogs fed Bones and Raw Food (BARF) diets and dogs fed commercial diets. PLoS One. 2018;13(8):e0201279.

    • PubMed
    • Search Google Scholar
    • Export Citation
  • 14.↑

    Pasławski R, Kurosad A, Zabek A, et al. Effect of 6-month feeding with a diet enriched in epa + dha from fish meat on the blood metabolomic profile of dogs with myxomatous mitral valve disease. Animals. 2021;11(12):3360.

    • PubMed
    • Search Google Scholar
    • Export Citation
  • 15.↑

    Smith CE, Parnell LD, Lai CQ, Rush JE, Freeman LM. Investigation of diets associated with dilated cardiomyopathy in dogs using foodomics analysis. Sci Rep. 2021;11(1):15881.

    • PubMed
    • Search Google Scholar
    • Export Citation
  • 16.↑

    Haimovitz D, Vereb M, Freeman L, et al. Effect of diet change in healthy dogs with subclinical cardiac biomarker or echocardiographic abnormalities. J Vet Intern Med. 2022;36(3):1057–1065.

    • Search Google Scholar
    • Export Citation
  • 17.↑

    Storey JD, Tibshirani R, Green PP. Statistical significance for genomewide studies. PNAS. 2003;100(16):9440–9445.

  • 18.↑

    Wang Y, Hodge RA, Stevens VL, Hartman TJ, McCullough ML. Identification and reproducibility of urinary metabolomic biomarkers of habitual food intake in a cross-sectional analysis of the cancer prevention study-3 diet assessment sub-study. Metabolites. 2021;11(4):248.

    • Search Google Scholar
    • Export Citation
  • 19.↑

    Wang Y, Hodge RA, Stevens VL, Hartman TJ, McCullough ML. Identification and reproducibility of plasma metabolomic biomarkers of habitual food intake in a US diet validation study. Metabolites. 2020;10(10):382.

    • Search Google Scholar
    • Export Citation
  • 20.↑

    Li Q, Larouche-Lebel É, Loughran KA, Huh TP, Suchodolski JS, Oyama MA. Metabolomics analysis reveals deranged energy metabolism and amino acid metabolic reprogramming in dogs with myxomatous mitral valve disease. J Am Heart Assoc. 2021;10(9):e018923.

    • PubMed
    • Search Google Scholar
    • Export Citation
All Time Past Year Past 30 Days
Abstract Views 0 0 0
Full Text Views 2602 891 65
PDF Downloads 1988 573 34
Cover American Journal of Veterinary Research
American Journal of Veterinary Research
Publication Date:
01 Sep 2022
Advertisement