Methods

This study was approved by the Cummings School of Veterinary Medicine at Tufts University Clinical Studies Review Committee (#002.18 and #001-2022) and owners signed an informed consent form. Three groups of client-owned dogs with myxomatous mitral valve disease (MMVD) were enrolled in the study: (1) Dogs with CHF and cardiac cachexia (CHF-cachexia group), (2) dogs with CHF and no cachexia (CHF-no cachexia group), and (3) dogs with MMVD but no CHF and no cachexia (American College of Veterinary Internal Medicine [ACVIM] Stage B2; B2 group).¹³ Dogs were patients of the hospital's Cardiology Service. The diagnosis of MMVD was based on signalment, a left apical systolic murmur, typical changes to the mitral valve leaflets on echocardiography, the presence of mitral regurgitation on color-flow Doppler, and left atrial enlargement. Echocardiograms were performed by board-certified veterinary cardiologists or cardiology residents in training under the direct supervision of a board-certified veterinary cardiologist. The diagnosis of CHF was made by the attending veterinary cardiologist based on the presence of clinical signs (eg, cough, dyspnea), with evidence of fluid accumulation (eg, radiographic or ultrasonographic evidence of pulmonary edema, pleural effusion, or ascites), and a positive clinical response to diuretic administration (eg, resolution of fluid accumulation in combination with improvement in clinical signs). Cachexia was defined as any muscle loss (mild, moderate, or severe muscle loss on the muscle condition score [MCS]),^1,14 but for the purposes of the study, only dogs with moderate or severe muscle loss were enrolled in the CHF-cachexia group. To be eligible, dogs' medications had to be stable with no medication changes within 7 days. Enrollment of dogs with active CHF or other clinical signs requiring medication adjustments was delayed until clinical signs were controlled with a consistent regimen of medications for ≥7 days. Dogs with CHF due to other heart diseases and dogs with major concurrent diseases (eg, neoplasia, diabetes) were excluded.

Dogs were fasted before study visits. In all dogs, an echocardiogram was performed and a serum biochemistry profile was analyzed to rule out important concurrent medical conditions. Blood was collected in a serum tube. After allowing time for a clot to form, tubes were centrifuged and serum was separated and stored at −80 °C until metabolomics analysis as a single batch. Nutritional assessment was performed in all dogs, including body weight, body condition score (BCS, 1–9 scale),¹⁵ MCS,¹⁴ and diet history. The same investigator (L.M.F.) assigned the BCS and MCS for all dogs. Diet history information was collected using a diet history form and a 24-hour food recall. Each dog's main diet (ie, the diet that provided the majority of the dog's calories at the time of enrollment) was used for statistical analysis. Diets were categorized into traditional commercial, nontraditional commercial, and homemade (traditional and nontraditional diets were separated because of previous results showing metabolic differences in dogs eating these 2 types of diets¹⁶). For the purposes of this study, nontraditional diets were defined as commercial diets that were grain-free or contained pulses, potatoes, or sweet potatoes in the first 10 ingredients on the ingredient list.¹⁷ Traditional diets were defined as commercial diets that included grains and did not contain pulses, potatoes, or sweet potatoes in the first 10 ingredients. Each dog's appetite was assessed by the owner using a questionnaire asking 5 questions about willingness to eat, hunger behaviors, anticipation of mealtime, enthusiasm for eating, and amount of food eaten, with answers given using a Likert scale from 1–5. The scores on individual questions were summed to determine the overall appetite score, which ranged from 5–25, with lower scores indicating worse appetite. The owners' perception of their dogs' quality of life was assessed using the Functional EvaluaTion of Cardiac Health (FETCH) questionnaire; scores can range from 0–85, with higher scores indicating worse quality of life.¹⁸

Serum samples were transported overnight on dry ice to a commercial laboratory (Metabolon, Inc) where they were stored at –80 °C until untargeted global metabolomic profiling was performed. Metabolites were quantified using ultra-high-performance liquid chromatography-tandem mass spectroscopy and identified by comparison to a reference library of purified standards containing retention time, mass–charge ratio, and mass spectroscopy spectral data.

Results

Metabolomics

Overall results and principal components analysis—A total of 1129 compounds that met quality control standards were included in the analysis (970 metabolites of known identity and 159 compounds of unknown structural identity). To provide a high-level view of patterns of the metabolomic datasets, principal components analysis was performed that showed separation between the CHF-cachexia and B2 groups along component 1 (Figure 1). There was a greater degree of intragroup variability in the B2 and CHF-cachexia groups than in the CHF-no cachexia group, which clustered more closely together between the other 2 groups.

View Full Size

Principal component analysis of 3 groups of dogs with myxomatous mitral valve disease: Dogs with congestive heart failure (CHF) and cachexia (CHF-cachexia group, green), dogs with CHF and no cachexia (CHF-no cachexia group, magenta), and dogs with no CHF and no cachexia (American College of Veterinary Internal Medicine Stage B2 [B2 group], blue), showing separation of the CHF-cachexia and B2 groups along component 1, as well as a greater degree of intragroup variability in the CHF-cachexia and B2 groups compared to the CHF-no cachexia group.

Citation: American Journal of Veterinary Research 84, 12; 10.2460/ajvr.23.07.0161

• Download Figure
• Download figure as PowerPoint slide

Random Forest analysis—Random Forest analysis (Figure 2) determined the top 30 metabolites that differentiated pairwise comparisons of the 3 groups. For the CHF-cachexia vs CHF-no cachexia comparison, metabolites suggested key differences in lipids (12 metabolites), followed by amino acids (7) and xenobiotics (7), with a predictive accuracy of 80% (AUC = .86; 95% CI [.69, 1.00]). For the comparison of CHF-cachexia vs B2 groups, amino acids (11), xenobiotics (5), and nucleotides (4) were the most common differentiating metabolites, with a predictive accuracy of 90% (AUC = .97; 95% CI [.91, 1.00]). For the comparison of the CHF-no cachexia vs B2 groups, amino acids (8) predominated as key differentiating metabolites, followed by cofactors/vitamins (5), and nucleotides (5), with a predictive accuracy of 80% (AUC = .84; 95% CI [.66, 1.00]).

View Full Size

Random Forest analysis was used to define the top 30 metabolites that contributed most to group binning between 3 groups of dogs with myxomatous mitral valve disease: (A) CHF-cachexia group vs CHF-no cachexia group, predictive accuracy = 80%; (B) CHF-cachexia vs B2 group, predictive accuracy = 90%; and (C) CHF-no cachexia vs B2 group, predictive accuracy = 80%. These metabolites are shown in descending importance order in the biochemical importance plots. The super pathway for each metabolite is indicated by color as defined by the legend in the lower right corner of each plot. See Figure 1 legend for additional details.

Citation: American Journal of Veterinary Research 84, 12; 10.2460/ajvr.23.07.0161

• Download Figure
• Download figure as PowerPoint slide

Significantly different metabolites—When comparing individual metabolites between the CHF-cachexia and CHF-no cachexia groups (Table 2; Supplementary Table S1), 2 metabolites (3-hydroxyadipate and caprylate, both in the lipid superfamily) were significantly higher (P-value < .05 and q-value < .10) in the CHF-cachexia group compared to the CHF-no cachexia group; none were significantly lower. An additional 122 metabolites were nominally significantly higher (P-value < .05 but q-value > .10) and 16 metabolites were nominally significantly lower in the CHF-cachexia compared to the CHF-no cachexia groups. Similar to the Random Forest analysis results, many of the different metabolites were medium- and long-chain dicarboxylates and dicarboxylate acylcarnitines (up to 18 carbons). The 1 dog in this group that was eating a diet that contained a source of medium-chain triglycerides (coconut oil) did not have the highest levels of dicarboxylates and dicarboxylate acylcarnitines, while dogs with the highest individual dicarboxylates and dicarboxylate acylcarnitines levels had no major sources of medium-chain triglycerides in their diets or supplements.

Table 2

Pairwise comparison of biochemical compounds analyzed by metabolomics between 3 groups of dogs with myxomatous mitral valve disease (see Table 1 for additional details).

	Groups compared
Significance level	CHF-cachexia vs CHF-no cachexia	CHF-cachexia vs B2	CHF-no cachexia vs B2
Number of biochemicals that were significantly different between groups (P < .05 and q-value < .10)	2	180	19
	2 higher	163 higher	19 higher
	—	17 lower	—
Number of biochemicals that were nominally significantly different (P < .05 but q-value > .10)	138	71	136
	122 higher	60 higher	90 higher
	16 lower	11 lower	46 lower

Higher or lower refers to the first group of each pair (eg, for CHF-cachexia vs CHF-no cachexia comparison, “higher” indicates the number of biochemicals that were higher in the CHF-cachexia group). All significant and nominally significant metabolites are shown in Supplementary Table S1. CHF = Congestive heart failure.

For the comparison of the CHF-cachexia vs B2 groups (Table 2, Supplementary Table S1), 163 metabolites were significantly higher, and 17 metabolites were significantly lower in the CHF-cachexia group compared to the B2 group. Many of these significant metabolites were also significant (n = 1) or nominally significant (n = 65) for the CHF-cachexia vs CHF-no cachexia comparison. Another 60 metabolites were nominally significantly higher and 11 metabolites were nominally significantly lower in the CHF-cachexia vs B2 group comparison. Amino acid compounds represented the largest number of differences, but differences in dicarboxylates also were seen.

For the comparison of the CHF-no cachexia vs B2 groups (Table 2, Supplementary Table S1), 19 metabolites were significantly higher in the CHF-no cachexia group compared to the B2 group. Another 90 metabolites were nominally significantly higher and 46 were nominally significantly lower in the CHF-no cachexia group compared to the B2 group. Amino acid compounds represented the largest number of differences between these 2 groups.

Pathway enrichment analysis—Pathway enrichment analysis identified 2 metabolic subpathways that were significantly enriched (overrepresented) when comparing the CHF-cachexia vs CHF-no cachexia groups (Supplementary Table S2): Fatty Acid Metabolism (Acyl Carnitine, Dicarboxylate), with an enrichment value of 9.379 (P = .001, q = 0.06) and Fatty Acid, Dicarboxylate, with an enrichment value of 3.520 (P < .001, q = 0.02). The comparison of the CHF-cachexia vs B2 groups identified 2 different subpathways that were significantly different: Ascorbate and Aldarate Metabolism, with an enrichment value of 4.668 (P < .001, q = 0.002) and Histidine Metabolism, with an enrichment value of 2.697 (P < .001, q = 0.02). The comparison of the CHF-no cachexia vs B2 groups identified 2 subpathways that were significantly different: Ascorbate and Aldarate Metabolism, with an enrichment value of 6.151 (P < .001, q = 0.02) and Dipeptide, with an enrichment value of 6.151 (P < .001, q = 0.02).

Discussion

The results of this study showed numerous significant or nominally significant differences in metabolites in dogs with cardiac cachexia. The objective of the study was to identify metabolic differences due to cardiac cachexia but, because CHF itself is associated with metabolic changes, it was important to include 2 control groups to identify changes due to cachexia as opposed to changes due to the presence of CHF. If a metabolic change was due to cachexia, there should have been consistent findings for metabolites in the CHF-cachexia vs CHF-no cachexia comparison and in the CHF-cachexia vs B2 comparison, but not in the CHF-no cachexia vs B2 comparison. Many of the significant and nominally significant metabolites fit this pattern.

Comparison of the CHF-cachexia vs CHF-no cachexia groups is particularly interesting because both groups had CHF and the main difference was the presence or absence of cachexia. For this comparison, key differences in lipids and amino acids predominated in the Random Forest analysis (Figure 2), with many being medium- and long-chain (up to 18 carbon) dicarboxylates and dicarboxylate acylcarnitines (Supplementary Table S1). The enrichment analysis results for the CHF-cachexia vs CHF-no cachexia comparison were consistent, with both the dicarboxylate and dicarboxylate acylcarnitines subpathways being significantly enriched (Supplementary Table S2).

While mitochondrial beta-oxidation predominates as the source of energy production in the healthy heart, there is a shift in CHF towards reliance on glucose for energy.²⁴ In addition, an alternate route to mitochondrial beta-oxidation for energy production is omega-oxidation which can be induced when beta-oxidation capacity is blocked or overwhelmed.²⁵ Omega-oxidation produces dicarboxylates that can then be metabolized by beta-oxidation in the mitochondria or peroxisomes to produce energy, as well as to limit the toxic effects of fatty acid accumulation.²⁵ Dicarboxylate acidemia/aciduria can occur in humans with mitochondrial fatty acid oxidation defects (eg, medium-chain acyl-coenzyme A dehydrogenase deficiency) or peroxisomal disorders (eg, Zellweger syndrome), although peroxisomal disorders typically produce dicarboxylates of ≥ 22 carbons from very long-chain fatty acids,^25-28 so the current results support mitochondrial dysfunction. While higher levels of even-chain dicarboxylates (eg, maleate [4-carbon], octadecanedioate [18-carbon]) can result from other conditions, including nonenzymatic oxidation of fatty acids²⁹ or fasting,³⁰ the presence of significant higher odd-chain dicarboxylates (eg, heptenedioate [7-carbon], azelate [9-carbon], undecanedioate [11-carbon]) supports omega-oxidation due to mitochondrial dysfunction.^25,28

Higher dicarboxylates and dicarboxylate acylcarnitines as a reflection of mitochondrial fatty acid oxidation dysfunction have been reported in some human studies. One study comparing older men with and without sarcopenia (defined as a loss of muscle and muscle strength or function) showed that sarcopenia was associated with higher 6-, 8-, and 10-carbon dicarboxylate acylcarnitines.³¹ In people at risk for heart disease, short-, medium-, and long-chain dicarboxylate acylcarnitines (but not dicarboxylates) were positively associated with morbidity and mortality.³² Also, in a study of human patients in the intensive care unit after aortic surgery, many acylcarnitines, dicarboxylate acylcarnitines, and dicarboxylates increased over time during hospitalization and were negatively associated with muscle mass and strength.³³ Many of that study's acylcarnitine and dicarboxylate acylcarnitine findings were consistent with those in the current study and support the presence of mitochondrial dysfunction.

Other possible causes for higher dicarboxylates and dicarboxylate acylcarnitines in the CHF-cachexia group include higher dietary precursors (eg, medium-chain triglycerides as a source for medium-chain dicarboxylates and dicarboxylate acylcarnitines). While the dogs' diets were not analyzed, only 1 dog in the CHF-cachexia group ate a diet that included a major source of medium-chain triglycerides, but that dog did not have the highest levels of dicarboxylates and dicarboxylate acylcarnitines. These findings also could be the result of worse appetite in the CHF-cachexia group and resulting in negative nitrogen balance or nutrient deficiencies. A related possibility is that dogs in the CHF-cachexia group were less able to tolerate the fasting that was part of the study protocol. In people with liver disease, an overnight fast produces metabolic changes similar to 2–3 days of fasting in a healthy individual.³⁴ The higher dicarboxylates in the CHF-cachexia group also could be related to the fact that more dogs in this group were receiving sildenafil. One study in humans found that sildenafil treatment in patients with heart failure with preserved ejection fraction was associated with increases in several dicarboxylates, but there was minimal overlap with specific dicarboxylates found to be higher in the current study.³⁵

In addition to the dicarboxylate differences, 1-methylhistidine and its acetylated derivative, N-acetyl-1-methylhistidine, were both significantly higher in the CHF-cachexia vs CHF-no cachexia and CHF-cachexia vs B2 comparisons. The 3-methylhistidine and N-acetyl-3-methylhistidine were only significantly higher in the CHF-cachexia vs B2 comparisons. In humans, 1- or 3-methylhistidine has been reported to be elevated in some, but not all studies of cancer cachexia or CHF.^8,11,12,36 In a metabolomics study of dogs with MMVD, 3-methylhistidine and N-acetyl-3-methylhistidine were significantly higher in dogs with Stage C/D disease compared to those with Stage B2, although diet and muscle condition scores were not reported.³⁷ While 3-methylhistidine is a well-established biomarker for skeletal muscle breakdown in humans, it also is a marker for meat consumption (especially red meat).³⁸ In human studies, subjects are required to avoid meat for 24–72 hours before collecting urine for 3-methylhistidine analysis. The 1-methylhistidine is in the muscle of most mammals but not humans, so if it is found in humans, it is more likely a marker of meat intake.³⁸ Dogs in the current study did not eat a meat-free diet before blood collection so the results for 1- and 3-methylhistidine would not be considered ideal for assessing muscle breakdown. Additional research on 1- and 3-methylhistidine in dogs with cachexia is warranted although would require dietary modification before sample collection.

We were unable to identify any published human or veterinary articles on metabolomics in cardiac cachexia for comparison. There are, however, some studies of metabolic changes in human patients with cancer cachexia that can be used for comparison. Some findings in human patients with cancer cachexia (compared to patients with cancer but without cachexia) include lower branched-chain amino acids and higher phenylacetate, lysine, and myo-inositol, although results are not consistent across all studies.^36,39–42 For example, some studies found lower isoleucine and carnosine in cancer cachexia, while others found higher levels.^36,39,42 Results of the current study did not identify significant changes in metabolites that were consistent with those found in studies with human cancer cachexia studies. It is difficult, however, to compare across studies due to different species, methodology/platforms, disease (ie, CHF vs cancer), type of cancer (eg, gastrointestinal, lung, and mixed cancer types), study design, sex, and matrix analyzed (eg, serum, red blood cells, urine, and muscle).

In addition to being able to investigate changes in the current study that were due to cachexia, the comparison of the CHF-no cachexia group to the B2 group also provides information on metabolic changes associated with CHF. Human studies have shown numerous metabolic changes that occur in heart disease or CHF, such as higher citric acid cycle intermediates and lipid beta-oxidation products, as well as reduced levels of some amino acids (eg, glutamine, threonine, and histidine).^8,9 Other studies of DCM have shown an association between disease severity and metabolites including 2-hydroxybutyrate, glycine, methylmalonate, myo-inositol, certain acylcarnitines, sialic acid, 3-methylhistidine, and glutamic acid.^10,11 One study of humans with CHF showed that high 3-methylhistidine, low alanine, and low valine were associated with adverse events.¹² None of these metabolites from human studies were consistent with findings from the current study. However, comparison with those studies is challenging because none stated whether cachexia was present so it is impossible to tell whether the changes were due to CHF, cachexia, or a combination of the 2 (or to differences in species, study populations, and other methodological differences). One study of humans with heart failure with reduced ejection fraction (mostly ischemic etiology) analyzed lipidomics and found higher levels of some long-chain acylcarnitines and dicarboxylate acylcarnitines, supporting dysregulated mitochondrial and peroxisomal fatty acid metabolism, some of which were similar to the current study.²⁶

There also are at least 3 studies in dogs that evaluated metabolomics in dogs with MMVD.^37,43,44 One compared the serum metabolomics of dogs with MMVD to those of controls.⁴³ While that study used the same analysis platform as the current study, it was unclear whether the dogs with MMVD had CHF or cachexia. The authors found that 41 known metabolites were significantly different between groups, suggesting alterations in energy metabolism and oxidative stress. The only overlap with the current study results was for higher erythronate and aconitate, and lower serine and valine. A more recent study, also using the same analysis platform as the current study, evaluated dogs with MMVD at various stages of disease, including comparison of dogs with ACVIM Stages C/D to dogs with ACVIM Stage B2.³⁷ That study had some consistent findings to the current study's comparison of the CHF-no cachexia vs B2 groups, including higher levels 2R,3R-dihydroxybutyrate, 3,4-dihydroxybutyrate, 5,6-dihydrouridine, aconitate, citrate, ethylmalonate, hippurate, maleate, quinolinate, tiglylcarnitine, and trimethylamine N-oxide. These findings suggest changes in energy metabolism and alterations in the gut microbial by-products associated with CHF.

A recent study compared serum lipidomics in dogs that progressed from ACVIM Stage B1 to Stage C, although the presence of cardiac cachexia was not reported.⁴⁴ Dogs in Stage C had higher levels of nervonoylcarnitine (C24:1) compared to the same dogs in Stage B, which is consistent with the current study results for comparisons of both CHF groups to the B2 group. There were also some similar findings of higher levels of other acylcarnitines in the Wilshaw et al study⁴⁴ and the current study, but not consistently for both CHF groups. In addition to diet information, it will be valuable for future studies of metabolomics and heart disease to include information on the presence of cachexia. This will help to differentiate results associated with cachexia, as opposed to changes due to CHF, diet, or other factors.

The current study has many limitations. While dogs were reasonably well matched for many characteristics, there were some differences either based on the study design or the characteristics of the individual dogs. For example, part of the study design was that dogs in 1 group would have no CHF (B2 group) so none of the dogs in this group were receiving furosemide, while all dogs in both CHF groups were receiving furosemide. Sildenafil was administered to 60% of dogs in the CHF-cachexia group and no dogs in the other 2 groups. Therefore, some metabolic differences in the CHF-cachexia group could be due to this medication, as previously noted.³⁵ Only dogs' main diets were used for statistical comparisons although some dogs ate more than 1 diet. Given how different diets can affect metabolomic profiles,¹⁶ effects of combination diets may have influenced results. Similarly, we only statistically compared diet at the time of enrollment and did not account for treats. Diets also were not analyzed for fatty acids, amino acids, and other nutrients to determine the role of specific nutrients or profiles on results. Therefore, differences in the CHF-cachexia group may not be solely the result of cachexia, but due to other differences besides the main diet. The CHF-cachexia group's metabolic differences also may have been influenced by having a more severe disease since some human studies have seen metabolic differences based on disease severity (based on clinical signs or left ventricular ejection fraction).^10,11 Dogs in the 2 control groups, particularly the CHF-no cachexia group, may have had muscle loss that was too subtle to be detected by the MCS, so the CHF-cachexia group may not have been as phenotypically different from the 2 control groups as presumed. Therefore, while the processes contributing to cachexia could also be present at low levels in the control groups, dogs with overt muscle loss, as dogs in the CHF-cachexia group had, would be likely to have more exaggerated metabolic alterations than those with no clinically detectable muscle loss. Possible breed differences also would be useful to investigate in future metabolomics studies. It would have been ideal for healthy dogs without heart disease and dogs with ACVIM Stage B2 MMVD that had cachexia as additional control groups. The cost of performing metabolomics limited the total numbers for the study so the Stage B2 without cachexia and CHF-no cachexia groups were selected as the most important controls to include for comparison to the CHF-cachexia group. In addition, cachexia typically does not occur until dogs have CHF or in late Stage B2 so dogs with Stage B2 with cachexia might have muscle loss due to other causes (eg, CKD, cancer, and aging [sarcopenia]) that may have different mechanisms. Finally, the sample size of this study was small, with only 10 dogs in each of the 3 groups so additional studies with larger numbers and more consistent diets are needed.

Despite the many limitations, these results suggest there might be some metabolic differences between dogs that have CHF with and without cardiac cachexia, and that these might be separate from the changes associated with CHF. These results, however, are only a first step into studying metabolic changes associated with cardiac cachexia and additional research is warranted to potentially identify biomarkers for earlier diagnosis of cardiac cachexia (or even dogs with early heart disease that are predisposed to developing cardiac cachexia), identify possible treatments, and better understand the complex mechanism of this common syndrome.

Supplementary Materials

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

Disclosures

In the last 3 years, 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; Nestlé Purina PetCare; P&G Petcare (now Mars); and Royal Canin. In the last 3 years, Dr. Rush has received funding from, given sponsored lectures for, or provided professional services to Aratana Therapeutics, Elanco, Hill's Pet Nutrition, Increvet, Nestlé Purina PetCare, Royal Canin, IDEXX, and Boehringer Ingelheim.

No AI-assisted technologies were used in the generation of this manuscript.

Funding

This study was supported by Aratana (now Elanco), the Cummings School of Veterinary Medicine Seed Grant, the Barkley Fund, and generous support from Ray and Ann Buono.