Topics in Companion An Med 33 (2018) 126135

Research Article

A Comparative Study of Serum Biochemistry, Metabolome and Microbiome

Parameters of Clinically Healthy, Normal Weight, Overweight, and Obese
Companion Dogs
D1X XGenevieve M. ForsterD2Xa,b,yX , D3X XJonathan StockmanD4Xa,b,y X , D5X XNoelle NoyesDc6X X , D7X XAdam L. HeubergerD8XdX ,
D9X XCorey D. BroecklingD10X X , D1X XCollin M. BantleD12X X , D13X XElizabeth P. RyanDb,
e b
14X X *

Keywords: A B S T R A C T
canine
obesity The aim of this study was to compare fecal microbiome, plasma, fecal and urine metabolomes, and serum
metabolome
biochemistry of adult companion dogs according to body condition scores. Blood, serum/plasma, urine, and
microbiome
fecal samples were collected from 66 clinically healthy, adult companion dogs of either normal weight (NW),
overweight (OW), or obese dogs (OB). analyses included fecal microbiome analyses via 16S ribosomal RNA
a
College of Veterinary Medicine and Biomedical gene amplicon; sequencing, nontargeted plasma, fecal, and urine metabolomics using liquid chromatogra-
Sciences, Departments of Clinical Sciences,
phy/gas chromatography-mass; spectrometry, and serum biochemistry for each dog. Few significant differ-
Colorado State University, Fort Collins, CO, USA
ences in serum biochemistry and fecal microbiome Operational Taxonomic Unit (OTU) were found between
b
College of Veterinary Medicine and Biomedical weight groups and there was high OTU variation between individual dogs. NW dogs had higher relative
Sciences, Environmental and Radiological Health
abundance of the genus Eubacterium (log-fold change 4.3, adjusted P value = .003) and lower relative abun-
Sciences Colorado State University, Fort Collins,
dance of the family Bifidobacteriaceae (log-fold change ¡3.6, adjusted P value = .02) compared to OB dogs.
CO, USA
c
The microbiome of NW dogs had higher OTU richness compared with OB dogs. Metabolome analysis showed
College of Veterinary Medicine and Biomedical
185 plasma, 37 fecal, and 45 urine metabolites that significantly differed between NW and OW or OB dogs.
Sciences, Microbiology, Immunology, and
There were notable significant differences in relative abundance of several plasma phospholipid moieties
Pathology, Colorado State University, Fort
Collins, CO, USA and fecal volatile fatty acids between weight phenotypes. The combinations of host and gut microbiota and
d metabolic shifts suggest a pattern that could help detection of early metabolic changes in overweight dogs
College of Agricultural Sciences, Departments of
before the development of obesity related disease. The results of this study support the need for continued
Horticulture and Landscape Architecture,
Colorado State University, Fort Collins, CO, USA investigation into sensitive measures of metabolic aberrancies in overweight dogs.
e © 2018 The Authors. Published by Elsevier Inc. This is an open access article under the CC BY-NC-ND license.
College of Agricultural Sciences, Proteomics and
(http://creativecommons.org/licenses/by-nc-nd/4.0/)
Metabolomics Facility, Colorado State
University, Fort Collins, CO, USA

Introduction High throughput techniques such as metabolomics and micro-

Obesity is a multifactorial nutritional disorder,1 highly preva-
lent in dogs in developed countries.2-4 Overweight (OW) and
obese (OB) dogs suffer from decreased quality of life, shorter life
expectancy, and are at greater risk for multiple diseases including
osteoarthritis, dyslipidemia, cardiovascular disease, and cancer.5,6
Obesity is readily diagnosed on physical examination, where
excess body fat can be estimated with body condition scores
(BCS), or measured by bioimpedance spectroscopy, deuterium
water, or dual energy X-ray absorptiometry scans.7-9
Some animals may be at higher risk to develop a metabolic, obe-
sity-related, secondary disease than others; however, identifying the
overweight animals with increased risk is a challenge. The detection
of abnormal clinical parameters has had limited success in early iden-
tification of developing weight related comorbidities. Elevated white
blood cell counts, blood glucose, blood urea nitrogen, creatinine,
phosphorus, calcium, cholesterol, and alkaline phosphatase have
been reported in overweight dogs, but these appear to be inconsis-
tent findings and have low specificity to obesity-related diseases.10-15
* Corresponding author.
E-mail address: E.P.Ryan@colostate.edu (E.P. Ryan).
yThese authors contributed equally to this report.
biome analysis have shown promise for evaluating altered metabolic
states in people.16-18 The identification of metabolic patterns/profiles
associated with early detection of obesity may become useful to dis-
ease prevention of diabetes, renal disease, liver disease, and even can-
cer. For example, patterns of metabolic shifts, altered fecal volatile
organic compounds and fecal microbiota were reported in human
obesity-related fatty liver disease as means of early and noninvasive
detection.16
The role for the gut microbiome in obesity is suggested to be
related to enhanced capacity to harvest energy from food,19 particu-
larly via fermentation and increased carbohydrate bioavailability. In
people, obesity-related differences exist in abundance of Clostridium
spp., Faecalibacterium prausnitzii, Bifidobacteria, Eubacterium, Bacter-
oides, Lactobacilli, and Prevotellacea.20-24 However, only few studies
evaluated differences in the fecal microflora composition between
normal weight (NW) and OW dogs.25,26
Metabolomics is the study of all detectable small molecules in
a biological sample, and can provide information on subclinical
metabolic alterations that are associated with disease out-
comes.27 Metabolome analysis has revealed previously unknown
alterations in amino acid, lipid, and carbohydrate metabolism
across species with underlying links to obesity, inflammation,
oxidative stress, and perturbations in the gut microflora.28-32 Sev-
eral studies showed potential relationships between stool

http://dx.doi.org/10.1053/j.tcam.2018.08.003
1938-9736/© 2018 The Authors. Published by Elsevier Inc. This is an open access article under the CC BY-NC-ND license. (http://creativecommons.org/licenses/by-nc-nd/4.0/)
 G.M. Forster et al. / Topics in Companion An Med 33 (2018) 126135 127

metabolites and certain bacterial species providing further insight
into microbial functions and the association between gut micro-
flora, health, and disease. 16,33,34 The plasma, fecal, and urine
metabolomes of overweight/obese humans were shown to have
altered volatile organic compounds,16 plasma lipid metabo-
lism, 35-37 and amino acid metabolism36,37 that reflects shifts in
energy metabolism pathways.
The objective of this study was to determine differences in
fecal microbiome composition plasma, fecal, and urine metabo-
lites, and serum biochemical analytes in companion dogs accord-
ing to weight phenotype. We hypothesized that there will be
differences in fecal microbiome composition and diversity and
the plasma, fecal, and urine metabolome would show altered
lipid, amino acid, and purine metabolic pathways when compared
with NW dogs. Also, OW and OB dogs were hypothesized to have
a higher number of serum biochemical values that are outside of
established normal ranges.
Materials and Methods
Study participants were clinically healthy adult dogs recruited
for future participation in 1 of 3 previously reported clinical tri-
als.38-40 All samples collected and analyzed for this report were
acquired from companion dogs at enrollment/baseline and they
had not received any intervention or manipulation prior to sam-
pling. Dogs were enrolled at the Colorado State University Veteri-
nary Teaching Hospital (Fort Collins) and the Wellington
Veterinary Hospital (Wellington). IACUC approved clinical trial
operations, etc, and the informed, written consent obtained from
all owners. A single veterinarian determined that all enrolled
dogs were healthy (aside of obesity status) according to the pro-
vided medical history and physical examination. In addition, a
complete serum biochemistry panel and a complete blood count
(CBC) with differential count were conducted for each participant.
All dogs enrolled were free of clinical evidence for metabolic dis-
ease according to the results of these standard tests. A BCS was
assigned to each dog on a 9-point scale7where a score of 4 or 5 is
considered “ideal” and every unit increase in BCS over 5 corre-
sponds to an approximate 10% increase in bodyweight, such that
a BCS of 8/9 corresponds to approximately 30% excess body-
weight.8 Dogs were divided into weight groups based on BCS:
dogs with a BCS of 4-5 were considered “normal weight” (NW),
dogs with a BCS of 6-7/9 were considered “overweight” (OW),
and dogs with a BCS 8-9/9 were considered “obese” (OB).8 Under-
weight dogs were not included in the study. Signalment data
(breed, age, and sex) for each dog was collected from the owner
and medical records (Table 1).
Serum Biochemistry and Complete Blood Cell Counts
Blood was collected via venipuncture into a serum separation tube
without anticoagulant for 15 minutes and serum was collected fol-
lowing centrifugation for 15 minutes at 1500 £ g. Blood for analysis
was collected into tubes containing EDTA and stored at 4°C for up to
4 hours until analysis. All clinical analyses were performed at the
Clinical Pathology Laboratory at Colorado State University as previ-
ously described.39 Briefly, serum biochemistry was performed under
standard conditions on a clinical chemistry analyzer (Hitachi 917,
Roche Diagnostics; Indianapolis, IN) and the CBC was performed on
an automated analyzer (Advia 120, Bayer; Tarrytown, NY).
Canine Fecal Microbiome Analysis
Owners were instructed to collect fresh fecal samples (samples
accepted were within 5 hours of being voided). Fecal samples were
immediately frozen after collection and stored at ¡20°C, lyophilized
until dry, and stored at ¡80°C until extracted for microbial analysis.
In brief, DNA was extracted from 150 mg of lyophilized stool using an
isolation kit (MO BIO PowerFecal DNA Isolation Kit, MO BIO Laborato-
ries, Carlsbad, CA) per the manufacturer’s instructions. The targeted
16S small-subunit ribosomal genes at region V4 were amplified using
universal 515F (50 -GTGCCAGCMGCCGCGGTAA-30 ) and 806R GGAC-
TACVSGGGTATCTAAT primers41 via PCR using a 12 bp unique Golay
barcoded primer for each sample.42 The sample amplicons were sub-
sequently cleaned using cleaning kits (MO BIO UltraClean-htp 96
Well PCR Clean-Up Kits, MO BIO Laboratories, Carlsbad, CA), quanti-
fied (Quant-iT PicoGreen dsDNA Reagent and Kits, Life Technologies
Corporation, Grand Island, NY), and pooled for sequencing. DNA
extraction, amplification, and cleaning were performed at the Natural
Resource Ecology Laboratory at Colorado State University.
Sequencing of 2 £ 250 cycles was performed on a MiSeq plat-
form (Illumina, San Diego, CA) at the Research Technology Sup-
port Facility Genomics Core at Michigan State University.
Sequencing data were processed using the Quantitative Insights
into Microbial Ecology (QIIME 1.7.0) pipeline43 for determining
operational taxonomic unit (OTU), taxonomy assignment, and
UniFrac analyses.44,45 Two reads were performed on each sample.
Sequence data were demultiplexed and filtered for quality using
the QIIME pipeline default settings and chimera sequences were
removed. 43 Sequence reads were normalized using the rarefying
procedure as implemented in QIIME. The level for rarefying was
taken from the sample with the lowest number of reads across
the entire dataset. Taxonomies were assigned to each sequence
using a default confidence setting of 0.5 (RDP Classifier: http://
rdp.cme.msu.edu/index.jsp). OTUs were picked using the

Table 1
Characteristics of 66 Normal Weight, Overweight, or Obese Clinically Healthy Dogs Analyzed to Determine Metabolic Aberrancies Between Weight Classes*.

Normal weight Overweight Obese

BCS 4-5 BCS 6-7 BCS 8-9 P valuey
y a b
Age , yr 3.0 5.0 5.5 .037
(2.5-4.5) (3.0-6.0) (3.0-7.0)
Weight, kg 22.5a 30.6 37.7b .031
(19.4-30.8) (22.4-37.3) (24.6-39.8)
Sexz .905
Female 9 16 12
Male 8 11 10
Data are reported as median and interquartile range.
* Weight class was determined by Body Condition Score (BCS) using a 9 point scale (Laflamme, 1997). Continuous variables (age and weight) were evaluated for differences
across groups using a KruskalWallis test and categorical variables (sex) were evaluated using a chi-square test. P < .05 was considered significant. Groups with significant differen-
ces are indicated with different letter superscripts.
y
Age as reported by owner.
z
All dogs were neutered with the exception of 1 male in the normal weight group, and 1 female in each of the overweight and obese groups.
128 G.M. Forster et al. / Topics in Companion An Med 33 (2018) 126135
pick_open_reference_otus.py command, with default settings.
The resulting OTU matrix contained OTUs as rows, sample num-
bers as columns, and counts of hits as matrix cells. This matrix
was imported into a computer software for descriptive and infer-
ential analyses (R version 3.2.3, R-Core Team, Vienna, Austria).46
OTUs present in fewer than 5 samples were removed from further
analyses to avoid nonrobust estimates due to sparseness. Remain-
ing OTU counts were normalized using cumulative sum scaling as
implemented in metagenomeSeq; 47 this method accounts for
uneven sequencing depth while avoiding bias that can be intro-
duced with the rarefying procedure. 48 A default value of 0.5 % was
used for the procedure. In order to perform a taxonomy-based
analysis (in addition to an OTU-based analysis), OTUs were
matched to a reference taxonomy and the taxa summaries were
created with standard QIIME workflows with gg_otus97 as the
reference set. Normalized OTU counts were aggregated to various
taxonomic levels (species through phylum) using the “aggTax”
function in metagenomSeq.47 The raw data files were uploaded to
National Institute of Health, Bioproject Database (http://www.
ncbi.nlm.nih.gov/bioproject/417964). The fecal microbiome rich-
ness and diversity and the relative abundance of the fecal micro-
biome composition are presented in Supplementary Tables 1 and
2, respectively.
Nontargeted Metabolomics Using Gas Chromatography and Liquid
ChromatographyMass Spectrometry
Plasma, fecal, and urine samples available from each dog were pre-
pared, extracted and derivatized as previously described.49 Briefly, lyophi-
lized fecal samples were ground to a powder using a mortar and pestle,
and metabolites were extracted by adding 1 mL of ice cold methanol:
water (80:20, v:v) to 100 mg of feces and incubated for
1 hour at ¡80°C. Samples were then centrifuged at 1500 £ g for
5 minutes at 4°C and 800 mL of the extract was transferred into 1 mL 96
well plates. For plasma and urine, 150 mL and 400 mL of each sample was
transferred to micro well plates, respectively. For gas chromatography
(GC)-mass spectrometry (MS) analysis, extracts were dried in a speedvac
(Thermo Fischer Scientific Inc., Waltham, MA) and methoximated by
resuspending in pyridine (50 mL) with 25 mg/mL of methoxyamine
hydrochloride and incubated at 60°C for 45 minutes twice, with
10 minutes sonication in between. After methoximation, samples were
incubated with 50 mL of N-methyl-N-trimethylsilyltrifluoroacetamide
with 1 % trimethylchlorosilane (MSTFA + 1% TMCS, Thermo Fischer Scien-
tific Inc., Waltham, MA) at 60°C for 30 minutes, followed by centrifugation
at 3000 relative centrifugal force for 5 minutes. Samples were then cooled
to room temperature (»22°C), and the supernatant (80 mL) was trans-
ferred to a 150 mL glass insert in a GCMS autosampler vial. For liquid
chromatography (LC) analysis, 100 mL of each sample was centrifuged at
13,000 relative centrifugal force at 4°C. The supernatant was transferred
to an autosampler vial and directly injected. GC and LCMS run condi-
tions were as previously described.49 Individual features, described by
mass, charge, and retention time were generated in a cloud-based metab-
olomic data processing platform (XCMS Online: https://xcmsonline.
scripps.edu/landing_page.php?pgcontent=mainPage).50 The mean abun-
dance of duplicate injections was normalized to total ion current and clus-
tered together into individual metabolites using the RAMClust
deconvolution algorithm.51 The relative abundance of each cluster was
calculated based on the weighted sum of all features within the cluster
and scaled to a median of one. Spectra were annotated based on mass
spectral and retention time/index matching against in-house and external
libraries (NIST v12: www.nist.gov; Massbank: http://www.massbank.jp/?
lang=en; Metlin: https://metlin.scripps.edu/landing_page.php?pgcon
tent=mainPage; Golm: http://gmd.mpimp-golm.mpg.de metabolite data-
bases). For this study, each annotated metabolite was reported with an
experimental library identification number "C" for compound number,
identification confidence,52 and structural and database ID numbers. Sup-
plementary Table 3 shows the relative abundance of metabolites that
were significantly different between weight groups and Supplementary
Table 4 lists the GC and LC mass spectral library used in this study and
each metabolite is referenced by its library/cluster number.
Statistical Analysis
Baseline characteristics of the dogs were evaluated using a Krus-
kalWallis test for the continuous variables of age and weight, fol-
lowed by Dunn’s Multiple Comparison test to determine differences
between specific groups (NW vs. OW, NW vs. OB, and OW vs. OB). A
chi-square test was used to evaluate groups for differences in the cat-
egorical variables of sex and presence of abnormal clinical blood val-
ues (outside the normal ranges established by the Colorado State
University Diagnostic Laboratory for dogs; where all outcomes were
dichotomous and recoded as yes or no for whether the value was
abnormal or not). All parameters evaluated had expected values
greater than 1 and at least 20% of the expected values were greater
than 5 to meet the assumptions for the chi-square test. Parameters
not meeting these assumptions are noted in Supplementary Table 1
and Supplementary Fig 1. Differences in proportional data were con-
sidered significant when P < .05, and trends were considered present
when P < .10. Statistical analyses were performed via computer soft-
ware (GraphPad Prism, Version 5.03, San Diego, CA).
Calculations performed on clinical blood parameters were ana-
lyzed with an ANCOVA where both weight group and age were evalu-
ated. Pair-wise comparisons were performed with TukeyKramer
and parameters without normal distributions were log transformed
for statistical evaluation. The raw values, not the transformed values
are reported for all parameters.
The fecal microbiome data were analyzed using generalized linear
models of normalized counts to test for differences in richness and
Shannon’s diversity between NW, OW, and OB dogs. The influence of
age on richness and diversity was evaluated using a chi-Square likeli-
hood ratio test of models with and without age as a covariate. When
age was found to exert a significant effect on model fit, it was
included in the final multivariable model as a confounder. Compari-
sons of richness and diversity were made at each taxonomic level, as
well as across all OTUs. Zero-inflated Gaussian mixture models were
used to model abundance of zero-inflated and over dispersed micro-
biome counts at each taxonomic level and across all OTUs,47 and pair-
wise comparisons of abundance (measured as log-fold change)
between NW, OW, and OB dogs were assessed using limma,53 with a
critical a of 0.05 for BenjaminiHochberg adjusted P values to
account for multiple comparisons. We were not able to include age as
a covariate in these models due to low degrees of freedom. Overall
microbiome composition was compared between NW, OW, and OB
dogs using nonmetric multidimensional scaling ordination of Euclid-
ean distances between Hellinger-transformed normalized counts,
specifying use of 2 dimensions.54 The nonmetric multidimensional
scaling stress value, as well as the analysis of similarities R-statistic
and corresponding P value were all used to evaluate the significance
of the ordination by body weight classification (NW, OW, or OB).
The relative abundance of each plasma, fecal, and urine metabolite
was scaled to a median value of 1 and is presented as the median and
interquartile range. To determine differences in the relative abun-
dance of each cluster/metabolite between dogs in each group (NW,
OW, OB), a mixed model ANOVA was used. Age was evaluated using
the distribution pattern of P values for age for each metabolite. The
distribution was found to be random (data not shown) indicating no
effect of age; therefore, age was not accounted for in the metabolo-
mics model.
 G.M. Forster et al. / Topics in Companion An Med 33 (2018) 126135
Results
Characteristics of Canine Participants
Baseline samples from 66 adult NW (n = 17), OW (n = 27), and OB
(n = 22) dogs, enrolled in 3 different dietary intervention studies38-40
were used in this study. The median age in years as reported by own-
ers was 3.0 (2.5-4.5) in NW dogs, 5.0 (3.0-6.0) in OW dogs, and 5.5
(3.0-7.0) in the OB group (Table 1). There was a significant difference
in age across the 3 weight groups (P = .04) with the median age of the
NW dogs lower than the OB dogs (P < .05). This was found to affect
the statistical evaluation of serum biochemistry analytes and fecal
microbiome richness and diversity and was included in the statistical
evaluation of these parameters. The median weight of the NW dogs
was 22.5 kg (19.4-30.8 kg), 30.6 kg (22.4-37.3 kg) for the OW dogs,
and 37.7 (24.6-39.8 kg) for the OB dogs (Table 1). Dogs in the OB
group were significantly heavier than dogs in the NW group (Table
1; P < .05). However, when the estimated ideal weights of each dog
according to BCS were calculated to account for size differences, there
was no difference between groups (data not shown). There were no
differences in the distribution of sexes between the groups (Table 1).
The range of breeds represented in this study included: Labrador
Retrievers, n = 8; Labrador cross, n = 6; Australian Cattle Dogs, n = 5;
Terriers (American Pit Bull and Boston), n = 5; Border Collie, n = 1;
Border Collie cross, n = 4; Golden Retriever, n = 4; Australian Shep-
herd, n = 3; Australian Shepherd cross, n = 1; Hounds (Daschund and
Basset), n = 3; Dalmatian, n = 2; Saint Bernard, n = 2; Spaniels, n = 2;
Boxer, n = 1; Corgi, n = 1; Karelian Bear Dog cross, n = 1, Keeshond,
n = 1, Rottweiler, n = 1; Shih Tzu, n = 1; Standard Poodle, n = 1; Wire-
hair Pointer, n = 1; and otherwise unknown mixed breeds, n = 12. All
dogs were fed commercial maintenance diets at the time of the
study.
CBC and biochemical analysis was completed in 66 dogs (NW,
n = 17; OW, n = 27; OB, n = 22). Due to insufficient or missing samples,
not all analyses were completed for every dog. Plasma metabolomics
was performed in 59 dogs (NW, n = 13; OW, n = 25; OB, n = 21), fecal
metabolomics in 64 dogs (NW, n = 17; OW, n = 25; OB, n = 22), and
urine metabolomics for 51 dogs (NW, n = 17; OW, n = 22; and OB,
n = 12). Fecal microbiome was analyzed from 50 dogs (NW, n = 17;
OW, n = 21; and OB, n = 12).
Fecal Microbiomes of Normal Weight, Overweight, and Obese Dogs
The sequence data from the fecal microbiome analysis were clus-
tered into 808 distinct OTUs, 454 of which were present in fewer
than 5 samples and therefore removed from further analyses. The
remaining 354 OTUs were assigned to 6 known phyla, which com-
prised 12 classes, 18 orders, 35 families, and 39 genera of bacteria.
Across all samples, the majority of normalized counts belonged to the
phylum Firmicutes (56.3%), followed by Bacteroidetes (38.1%). Actino-
bacteria, Cyanobacteria, Proteobacteria, and Verrucomicrobia
together comprised 3.6% of all normalized counts, and 2.0% of nor-
malized counts could not be assigned to a phylum. Extensive varia-
tion was present between individual dogs at the phyla level (Fig 1).
Within Firmicutes, the top 3 most abundant genera were Blautia
(51.2%), Ruminococcus (17.6%), and Clostridium (9.9%). Within Bac-
teroidetes, normalized counts were distributed between the genera
Bacteroides (50.9%) and Prevotella (49.0%). Based on nonmetric mul-
tidimensional scaling stress value, as well as the analysis of similari-
ties R-statistic and corresponding P value, the overall microbiome
composition of NW, OW, and OB dogs was not significantly different
at any level (analysis of similarities R-statistic < 0.10, P > .05; Supple-
mentary Table 1). The fecal microbiomes of OB and OW dogs showed
a trend toward lower microbial richness and Shannon’s diversity
when compared to NW dogs, but this comparison did not reach sta-
tistical significance (Supplementary Table 1).
129
There were significant differences in relative abundance of specific
taxa when comparing NW, OW, and OB dogs (Fig 1; Supplementary
Table 2). At the class level, Erysipelotrichi were more abundant in
OW compared to OB dogs (log-fold change 4.3, adjusted P value = .01),
and this was driven by differences in the genus Eubacterium, although
the difference was apparent at all levels of the taxonomy including
Erysipelotrichales and Erysipelotrichaceae (order and family levels,
respectively). Class Actinobacteria was more abundant in OB dogs
compared to NW dogs (log-fold change ¡2.3, adjusted P value = .04).
At the order level, Bifidobacteriales were less abundant in OW com-
pared to OB dogs (log-fold change ¡3.6, adjusted P value = .02) and
Aeromonadales tended to be more abundant in OW compared to OB
dogs (log-fold change 3.6, adjusted P value .08). Compared to OB
dogs, NW dogs also had higher relative abundance of Erysipelotrichi,
Erysipelotrichales, and Erysipelotrichaceae (class, order and family
levels, respectively), and we speculate that this finding was also
driven by difference in the genus Eubacterium (log-fold change 4.6,
adjusted P value = .002). As with OW compared to OB dogs, NW dogs
also had lower levels of the order Bifidobacteriales (log-fold change
¡3.8, adjusted P value = .01). There were no significant differences of
any taxa at any level when comparing normal-weight and over-
weight dogs (Fig 1B).
When analyzed as OTUs, the genus Blautia was significantly more
abundant in NW and OW dogs compared to OB dogs (log-fold change
3.4 and 2.2 and adjusted P value = .002 and .03, respectively), as was
the family Lachnospiraceae (log-fold change 2.6 and 2.7 and adjusted
P value = .02 and .03, respectively) and the species Eubacterium
biforme (log-fold change 4.3 and 4.0 and adjusted P value = .003 and
.004, respectively) (Fig 1C). The family Ruminococcus was more abun-
dant in NW than OB dogs (log-fold change 2.3, adjusted P value = .01).
OTUs within the species Prevotella copri (log-fold change 4.3, adjusted
P value = .03), and the genus Clostridium (log-fold change 3.2,
adjusted P value .03) were more abundant in OW than OB dogs.
Plasma, Fecal, and Urine Nontargeted Metabolome of Normal Weight,
Overweight, and Obese Dogs
Plasma, fecal, and urine samples were analyzed via GC and
LCMS and the relative abundance of compounds were compared
between NW, OW, and OB dogs.
The relative abundance of each metabolite that significantly dif-
fered by weight class is presented in Supplementary Table 3. Fig 2
shows the relative abundance for each metabolite that was signifi-
cantly different between weight groups as a ratio between OW/NW
or OB/NW.
For plasma, 30/554 (GCMS) and 155/348 (LCMS) of the identi-
fied metabolites varied between weight class (Supplementary Table 3).
A total of 1112 compounds were detected in feces by GCMS and 795
compounds were detected by LCMS. No fecal metabolites detected
by GCMS were significantly different between the weight groups,
whereas 37 fecal metabolites detected by LCMS showed differences
according to weight phenotype. In urine, only 1 of 1139 metabolites
detected by GCMS was different by weight group. There were 1318
urine compounds detected with LCMS, 44 of which were different by
weight class. Of the 267 compounds detected in plasma, feces, or urine
by GC or LCMS with significant differences between weight classes,
79 were subsequently identified with database spectral matches.
Metabolites that were increased in OW or OB compared to NW dogs
were primarily lipids (33/42 identified metabolites) and included 18
Glycerophospholipids, 1-stearoyl-phosphatidylethenolamine, 6 Sphin-
golipids, 4 Sterol lipids, 3 Fatty Acyls, and 2 Glycerolipids (triglycerides;
Supplementary Table 3 and Fig 2). Five metabolites associated with
protein metabolism were increased on OW or OB dogs compared to
NW, including 3 amino acid derivatives and 2 peptides. There were 2
carbohydrates elevated in OW or OB dogs, glucose and glycerol; 1
cofactor-dihydrobiopterin, and 1 nucleic acid. A total of 37 metabolites
130 G.M. Forster et al. / Topics in Companion An Med 33 (2018) 126135

Fig 1. Phyla level variation of the fecal microbiome from 51 companion dogs. (A) Phyla level variation in the fecal microflora of normal weight (NW), overweight (OW), and obese
(OB) companion dogs (n = 51). Each bar represents the total relative abundance and is shaded by phyla. Dogs are separated by weight group based on body condition score (BCS).
(B) Fecal microbiome differences between NW, OW, and OB dogs. Log-fold changes in the fecal microbiome between NW and OB dogs and OW vs. OB dogs compared at varying tax-
onomic level. (C) Log-fold changes in the fecal microbiome between NW vs. OB dogs and OW vs. OB dogs compared at operational taxonomic unit level. Data are shown as means
and error bars indicate 95% confidence intervals. Positive log fold changes indicate increased phyla abundance in NW and OW dogs compared with OB dogs. All shown differences
were significant (adjusted P value < .05) except for #, which indicates a trend (adjusted P value < .10).

were decreased in OW, and OB dogs compared with NW dogs. A total
of 28 were lipids, 5 proteins, 1 phytosterol, 1 cofactor, 1 bile pigment,
and 1 nucleic acid (Supplementary Table 3).
Serum Biochemistry and Complete Blood Cell Counts of Normal Weight,
Overweight, and Obese Dogs
Table 2 shows the median and range for 39 clinical serum and
whole blood parameters for all 66 dogs. Aspartate aminotransferase,
gamma-glutamyl transpeptidase, and chloride were significantly dif-
ferent between NW, OW, and OB dogs when adjusted for age. Aspar-
tate aminotransferase was decreased by an average of 4 U/L in OB
dogs compared to NW; chloride was decreased on average by
2 mmol/L; and g-Glutamyl transpeptidase was decreased in both OW
and OB dogs by 3 U/L and 2.5 U/L, respectively.
The most commonly observed parameter outside of normal
ranges was the albumin:globulin ratio which was elevated in 38/66
dogs; however, the distribution of dogs with elevated albumin:globu-
lin ratios was not different by weight group (P = .962). Hemolysis of
blood samples was noted in 25/66 dogs, and the distribution of sam-
ple hemolysis was different across weight group (P = .014) with OW
dogs having the highest proportion of hemolytic samples (15/27), fol-
lowed by OB dogs (8/22), and then NW dogs (2/17). Creatinine kinase
was elevated in 9/66 dogs, all of which were OW or OB, and this
 G.M. Forster et al. / Topics in Companion An Med 33 (2018) 126135 131

Fig 2. Relative abundance ratios for detected and annotated metabolites from plasma, urine, and feces of normal weight, overweight, and obese dogs. White boxes represent the
overweight:normal weight ratios whereas the gray boxes represent the obese:normal weight ratios. The lines represent median abundance. Positive values indicate metabolites
with higher relative abundance in overweight (OW) and obese (OB) dogs compared with normal weight (NW) dogs, and vice versa. (A) Ratios for relative abundances of plasma
metabolites in NW, OW, and OB dogs (n = 59). (B) Ratios for relative abundances of urine metabolites in NW, OW, and OB dogs (n = 51). (C) Ratios for relative abundances of fecal
metabolites in NW, OW, and OB dogs (n = 64).

distribution was also different across weight group (P = .038) with
OW dogs exhibiting the highest proportion (7/27), followed by OB
dogs (2/22). AST was elevated in 5/66 dogs and slightly below 16.3 U/
L in one NW and one OW dog. All 5 dogs with elevated AST levels
were OW (5/27), thus the proportion of abnormal values by weight
group was different (P = .020). Cholesterol was elevated in 5/66 dogs
and reduced in 1 OB dog, and the proportion of dogs with abnormal
values tended to differ by weight group (P = .080): NW (2/17), OW (0/
27), and OB (4/22). The fourth most frequent parameter outside of
the normal range was platelet count, which was decreased in 14/66
dogs, 13 of which had clumped platelets (data not shown). The pro-
portion of dogs with low platelets was equally distributed across
weight classes: 5/17 NW, 5/27 OW, and 4/22 OB (P = .631).
Discussion
The tandem analysis of fecal microbiome, plasma, fecal and
urine metabolomes, and serum biochemistry of NW, OW, and OB
companion dogs was implemented in this investigation to enable
identification of subclinical differences that may result from
excess adiposity. The fecal microbiome was analyzed to
determine distinctions that could contribute to metabolic altera-
tions in OW or OB dogs. Consistent with previous reports, the
phyla Firmicutes were the most abundantly represented, 26,55,56
followed by Bacteroidetes; yet, possibly due to individual varia-
tion, the fecal microbiome composition differences did not reach
statistical significance by weight group. OB dogs had lower abun-
dance of Erysipelotrichi, Erysipelotrichales, Erysipelotrichaceae,
and Eubacterium, (class, order, family, and genus, respectively) at
the taxonomic level, while OTU analysis also revealed lower
abundance of E. biforme compared to NW and OW dogs. High lev-
els of Eubacterium dolichum have been found to be associated
with increased visceral fat mass in people;57 whereas our results
display an inverse relationship between obesity and E. biforme in
dogs. Eubacterium ferment dietary nondigestible carbohydrates
into short-chain fatty acids in the gut, with propionate and buty-
rate having important health promoting and anti-inflammatory
roles on the host colonocytes. 58 One interpretation of the findings
in the present study is that the decrease in Eubacterium spp. in
the gut could reduce production of short-chain fatty acids that
supports a higher, yet subclinical inflammatory stress that reflects
decreased health status in the OB dogs.
132 G.M. Forster et al. / Topics in Companion An Med 33 (2018) 126135

Table 2
Serum Biochemistry and Whole Blood Counts of Normal Weight, Overweight, and Obese Dogs.

Analyte* Normal range Normal weight Overweight Obese P valuey

N = 17 N = 27 N = 22

Nucleated cells 4.5-15.0 £ 109/L 7.9 (5.8-12.1) 9.3 (5.6-14.7) 10.05 (5.8-15.4) .260
Seg. neutrophils 2.6-11.0 £ 109/L 4.9 (2.5-9.3) 6.3 (3.2-10.2) 7.25 (4.3-11.4) .109
Lymphocytes 1.0-4.8 £ 109/L 2.3 (0.9-4.4) 1.8 (1-3.8) 1.6 (0.6-3.0) .641
Monocytes 0.2-1.0 £ 109/L 0.5 (0.1-1.0) 0.5 (0.1-0.9) 0.6 (0.1-1.3) .978
Eosinophils 0.1-1.2 £ 109/L 0.35 (0.2-2.1) 0.4 (0.1-1.8) 0.4 (0.1-1.1) .677
Plasma Protein 0.06-0.075 g/L 0.65 (0.61-0.7) 0.68 (0.63-0.75) 0.7 (0.64-0.78) .206
RBC 5.5-8.5 £ 1012/L 7.3 (6.7-8.5) 7.3 (6.5-9.2) 7.6 (6.7-9.03) .510
HGB 13.0-20.0 g/L 17.2 (15.7-19.6) 18 (16-21) 18.25 (15.8-21) .309
Hematocrit 0.40-0.55 l/l 0.50 (0.46-0.56) 0.51 (0.46-0.60) 0.53 (0.44-0.60) .542
MCV 62.0-73.0 fl 68 (64-73) 70 (65-78) 68.5 (37-73) .331
MCHC 330-360 g/L 340 (340-360) 350 (340-370) 350 (330-360) .221
RDW 12.0-15.0 % 12.5 (12.1-13.3) 12.9 (11.9-14.5) 13.2 (11.4-13.9) .206
Platelets 200.0-500.0 £ 109/L 239 (142-324) 264 (106-453) 290 (123-444) .283
MPV 7.5-14.6 fl 11.1 (9.6-14.2) 10.4 (8.5-15.3) 11.2 (8.2-16.7) .589
Glucose 4.1-7.1 mmol/L 5.3 (3.9-6.6) 5.2 (4.4-10.8) 5.4 (4.4-7.6) .641
BUN 2.5-11.4 mmol/L 6.4 (4.2-11.0) 5.3 (3.9-11.4) 5.7 (3.2-13.9) .700
Creatinine 35.3-132.6 mmol/L 97.2 (70.7-123.8) 97.2 (70.7-132.6) 79.56 (62-114.9) .206
Phosphorus 0.67-1.93 mmol/L 1.22 (0.83-1.61) 1.13 (0.77-1.55) 1.06 (0.77-1.68) .827
Calcium 2.3-2.9 mmol/L 2.6 (2.5-2.9) 2.6 (2.5-2.9) 2.6 (2.5-2.9) .987
Magnesium 0.78-1.1 mmol/L 0.9 (0.7-0.98) 0.84 (0.78-1.1) 0.8 (0.7-0.9) .136
Total Protein 5.3-7.2 g/L 6.1 (5.6-6.8) 6.3 (5.4-7.0) 6.4 (5.8-7.0) .860
Albumin 2.5-4.0 g/L 3.8 (3.5-4.2) 3.9 (3.4-4.7) 4.0 (3.6-4.5) .125
Globulin 2.0-3.8 g/L 2.4 (2.0-3.2) 2.4 (1.6-3.3) 2.3 (1.7-2.8) .641
A/G Ratio 0.8-1.6 1.7 (1.1-2) 1.7 (1-2.4) 1.7 (1.3-2.4) .331
Cholesterol 3.38-7.8 mmol/L 5.95 (4.4-8.8) 6 (4.4-7.61) 6.3 (3.22-8.9) .542
Total bilirubin 0.0-0.51 mmol/L 0.17 (0-0.35) 0.17 (0.17-0.51) 0.17 (0-0.34) .221
z
CK 50.0-275.0 U/L 90 (52-212) 146 (62-2179) 150.5 (73-350) .109
z
ALP 20.0-142.0 U/L 40 (10-138) 45 (16-830) 47 (21-477) .987
z
ALT 10.0-110.0 U/L 31 (27-164) 41 (19-118) 35 (21-109) .743
z
AST 16.0-50.0 U/L 29 (15-47)a 29 (14-93)a 25 (16-39)b .043
GGT 0.0-9.0 U/L 3.0 (0.0-5.0)a 0.0 (0.0-5.0)b 0.5 (0.0-3.0)b .040
Sodium 142.0-152.0 mmol/L 147 (144-151) 146 (140-157) 147 (142-150) .199
Potassium 4.0-5.0 mmol/L 4.4 (3.9-4.7) 4.3 (3.7-4.9) 4.3 (3.5-4.7) .677
Chloride 108.0-120.0 mmol/L 113 (110-118)a 112 (109-119)a 111 (108-115)b .040
Bicarbonate 16.0-25.0 mmol/L 21.5 (18.8-28.4) 20.2 (17.6-23.9) 20.15 (16.8-25.3) .330
Anion gap 13.0-22.0 mmol/L 16 (11-21) 17 (12-22) 19 (14-23) .125
Calc. osmolality 284.0-304.0 mmol/L 293 (287-303) 290 (282-314) 291 (284-306) .237
z
Lipemia 0.0-0.53 mmol/L 0.18 (0-0.50) 0.1 (0-8.7) 0.09 (0.04-1.4) .987
z
Hemolysis 0.0-0.6 g/L 0.35 (0.07-0.82) 0.72 (0.13-4.03) 0.46 (0.03-2.20) .283
* A/G, albumin:globulin; ALP, alkaline phosphatase; ALT, alanine aminotransferase; AST, aspartate aminotransferase; BUN, blood urea nitrogen; CK, creatinine kinase; GGT,
gamma-glutamyl transpeptidase; HGB, hemoglobin; MCHC, mean corpuscular hemoglobin concentration; MCV, mean corpuscular volume; MPV, mean platelet volume; PCV,
packed cell volume; RBC, red blood cell; RDW, red blood cell distribution width.
y
Data are shown as median and range (minmax) and were evaluated for differences across groups using an ANCOVCA with the variables of age and body weight classification
and false discovery rate-corrected. Differences between groups were determined with pairwise comparisons and a TukeyKramer correction. P < .1 was considered significant.
Groups with significantly different medians are indicated with different letters.
z
Parameters not normally distributed were log transformed for statistical analysis however the actual, untransformed values are reported.

OB dogs had higher abundance of order Bifidobacteriales than NW
and OW dogs, and also had higher abundance of class Actinobacteria
compared to NW dogs. Within the order Bifidobacteriales, genus Bifi-
dobacterium is implicated in preventing obesity and altering lipid
metabolism in both human and rodent studies, however strain varia-
tions exist that can promote excess weight gain.59 Previous studies
reported increases in the phylum Actinobacteria (to which the family
Bifidobacteriaceae belongs) in obese companion dogs.26 Increased
Actinobacteria abundance was also demonstrated in obese humans.60
Obese dogs had lower abundance of the genus Blautia than NW and
OW dogs. Blautia are a dietary carbohydrate responsive, butyrate pro-
ducing bacteria.61 Decreases in Blautia spp. in dogs have been associ-
ated with acute diarrhea,62 and Blautia spp. in human fecal
microbiomes have been found to vary by adiposity.63 Altered lipid
and carbohydrate metabolism as well as increased adiposity may be
related to microbiome shifts through several mechanisms, including
interactions with the innate immune system and metabolic inflam-
mation through toll-like receptors, gut absorption and permeability
to lipopolysaccharides, and bacterial bile acid modifications.64,65
These results support the need for follow-up investigations into the
roles for microbiome composition and metabolism changes during
canine weight gain and development of obesity.
Nontargeted metabolomics analysis of the plasma, feces, and urine
revealed 267 compounds that varied among these weight groups,
with the majority of the variation detected in the plasma. Most differ-
ences in the identified metabolites trended similarly for both the OB
and the OW dogs when compared to the NW. As expected, the differ-
ences between OB and NW were larger than differences between OW
and NW dogs (Fig 2). From the list of annotated plasma metabolites,
the majority included lipids such as fatty acids and phospholipids.
Changes in plasma lipids in humans with increased body mass index
have been previously documented,37,66 and further support the
hypothesis that altered energy metabolism and glucose homeostasis
occur in OW and OB dogs. The OW and OB dogs in the present study
showed elevated plasma lysophosphatidylcholine (LysoPC) (20:4), 4
other LysoPC moieties were decreased, and several phosphatidylcho-
line lipids were elevated. While the association between phosphati-
dylcholine lipids, LysoPC, and obesity is unclear, similar findings have
been documented in obese mice, humans, and pigs.29,30,67 In addition,
a previous study that evaluated the effects of rapid weight gain in
 G.M. Forster et al. / Topics in Companion An Med 33 (2018) 126135
beagles on the plasma metabolome showed changes in plasma phos-
pholipids in the chronic stages of weight gain.68 Shifts in plasma
phospholipids may represent changes in lipid metabolism and/or
increased b-oxidation for energy production in individuals with
excess adiposity.35-37 Plasma phospholipids may act as targets for the
research and development of targeted tests to evaluate the risk of
obesity related metabolic alterations in dogs. In light of this finding,
consistency across species is needed since obesity related derange-
ments in lipoprotein metabolism may be a risk for several concurrent
morbidities such as hyperlipidemia and pancreatitis.69
A previous investigation of the fecal metabolome in overweight
and obese humans demonstrated alterations in volatile organic com-
pounds and fatty acids that may be associated with altered micro-
biota.16 The fecal metabolome of the obese and overweight dogs
showed decreased abundance of several fatty acids including linoleic
acid, ferulic acid, 1,28-dicaffeoyloctacosanediol (a coumaric acid
derivative), and colnelenic acid which act as functional antioxidants
as well as modulate inflammation, and reduce risk for cardiovascular
disease.70,71 The relative decrease in these metabolites may reflect
the subclinical chronic inflammation and oxidative stress associated
with obesity in dogs. The observed adiposity-related differences in
gut microbiome and fecal metabolome are likely interlinked, as the
gut microbiota is responsible for the production of many organic
compounds including fatty acids, which in turn influence the micro-
biota’s composition and diversity. Decreased abundance of fecal vita-
min D metabolites in the OW and OB dogs may be explained by shifts
in the gut microbiota and changes in bile acids as a fat-soluble vita-
min.72 In turn, vitamin D was shown to have an important role in
maintaining gut microbiome composition.72,73
The present study showed lower urinary levels of 1-methylhypox-
anthine in OB and OW dogs, while urinary methylxanthine abun-
dance was increased. Similar alterations in purine metabolism have
been described in overweight people where the decrease in xanthine
was also associated with hyperuricemia, suggesting that the enzy-
matic activity of xanthine oxidase is increased in obese individuals.74
Urine of the OB and OW dogs had increased N-acetyltryptophan,
betaine, and carnosine and decreased amino acid metabolism includ-
ing conjugated arginine, saccharopine, and biopterin. These changes
differ from those previously described for overweight Labrador
Retrievers during fasting and as postprandial response.75
Altered amino acid metabolism was documented in humans, rats,
and mice suffering from type-2 diabetes.76 Adiposity-related insulin
resistance, resulting in gluconeogenesis and perturbation in energy
metabolism and eventually increased amino acid oxidation may in
part explain the decreased urinary amino acids and their metabolites.
Increased urinary excretion of carnosine and betaine, which are com-
pounds that act as an antioxidant and as a methyl donor, respectively,
may reflect a subclinical metabolic stress from increased adiposity.
Since carnosine is present in high concentrations in muscle tissues,
the increased urinary abundance of this compound may also reflect
increased muscle catabolism.
Clinical blood parameters, such as g-Glutamyl transpeptidase, and
chloride were reduced in OW and OB dogs (Table 2), yet their values
were within normal ranges. g-Glutamyl transpeptidase increase was
previously described in obese dogs compared with nonobese dogs.15
g-Glutamyl transpeptidase may become elevated as a result of
changes in the biliary system or cholestasis;77 therefore this finding
suggests differences in bile composition or metabolism in overweight
dogs. In this cohort, the level of hemolysis was not associated with
obesity; however more OW and OB dogs had hemolytic blood sam-
ples than NW dogs, and could also represent a true phenomenon in
which dogs with excess adiposity could have friable red blood cell
membranes from the alterations in lipids, and thus a higher probabil-
ity of red cell lysis. Erythrocyte membranes from overweight and
obese humans have been shown to have altered lipid profiles and
decreased membrane fluidity;78 therefore, further investigation into
133
erythrocyte membrane integrity is warranted in OW and OB dogs. An
alternative explanation to the increase in hemolysis may be that veni-
puncture might have been technically more difficult from the subcu-
taneous fat in OW and OB dogs.
A previous retrospective analysis of hematological and biochemi-
cal parameters of OB and OW dogs showed higher leukocyte count;
higher plasma protein and globulin; higher sodium, calcium, and
anion gap; and lower chloride compared with NW dogs.14 A possible
explanation to the reduced chloride in OW and OB dogs would be
metabolic alkalosis, which may result from metabolic disturbances
and increased loss of chloride in the gastrointestinal tract.79
Several study limitations should be considered such as the rela-
tionship between age and weight phenotype, whereby excess body-
weight was positively correlated with age, although no dogs had
evidence of age related morbidities. This limitation may be more pro-
nounced when using client-owned dogs rather than colony dogs
because obesity has higher prevalence in middle-aged dogs.4 In addii-
ton, this study limitation also impacts interpretation due to age
effects on the microbiome and metabolome.80 Similarly, the dogs in
the study were of different breeds, neuter status, activity level, and
diet. These variables are all challenging to balance between the
weight groups, and future studies should stratify these parameters
with a larger number of dogs and weight groups. Another limitation
of metabolomics is the use of existing databases that include primar-
ily humans and rodent metabolite identifications which may differ
from those in canines. Finally, the inherent variability noted in the
fecal microbiome of companion dogs suggests that larger numbers of
dogs are needed to characterize microbiome shifts that occur in over-
weight and obese dogs.
Conclusion
The evaluation of fecal microbiome, plasma, fecal and urine
metabolomes, and serum biochemistry provided an integrative
assessment of the metabolic alterations that occur as a result of
weight gain and obesity. Our preliminary investigation provided
compelling evidence that microbiome analysis combined with
metabolomics merits further study to improve the understanding of
the impact of obesity on metabolism and health. The plasma metabo-
lites were identified from this study to show the greatest differentia-
tion according to obesity status when compared to minor changes
found in hematological parameters, serum biochemistry, fecal micro-
biota, and the fecal and urinary metabolomes. Further investigation
of the plasma phospholipid moieties in OW canine patients is war-
ranted for metabolic shifts to be considered candidate biomarkers of
canine obesity related comorbidities and disease risk.
Acknowledgments
This study was supported by a research grant from The CSU
Research Council of The College of Veterinary Medicine And Biomedi-
cal Sciences, Morris Animal Foundation Predoctoral Training Grant,
and CSU Department of Environmental Health and Radiological Sci-
ences funding. We would like to thank Guy Beresford for processing
the fecal samples for microbiome analysis, Ann Hess for her assis-
tance with the statistical analysis, Kelly Santangelo for her assistance
with interpreting the clinical blood panel, Husen Zhang for his help
with the fecal microbiome analysis, and the Clinical Trials Core and
Cadie Tillotson for her assistance with the collection of dog blood,
urine, and fecal samples.
Supplementary materials
Supplementary data associated with this article can be found in
the online version at http://dx.doi.org/10.1053/j.tcam.2018.08.003.
134 G.M. Forster et al. / Topics in Companion An Med 33 (2018) 126135
References
1. Hamper B. Current topics in canine and feline obesity. Vet Clin N Am 46:785–795, 2016
2. German AJ. The growing problem of obesity in dogs and cats. J Nutr 136:1940S–
1946S, 2006
3. Mao JF, Xia ZF, Chen JN, et al. Prevalence and risk factors for canine obesity sur-
veyed in veterinary practices in Beijing, China. Prev Vet Med 112:438–442, 2013
4. Lund EM, Armstrong PJ, Kirk CA, et al. Prevalence and risk factors for obesity in
adult dogs from private US veterinary practices. Int J Appl Res Vet Med 4:177, 2006
5. Linder DE. Top 5 clincial consequences of obesity. Clin Brief 19–21, 2014. Available
online at: https://www.cliniciansbrief.com/article/top-5-clinical-consequences-
obesity. Accessed September 5, 2018
6. Kealy RD, Lawler DF, Ballam JM, et al. Effects of diet restriction on life span and
age-related changes in dogs. J Am Vet Med Assoc 220:1315–1320, 2002
7. Laflamme D. Development and validation of a body condition score system for
dogs. Canine Practic 22:10–15, 1997
8. German AJ, Holden SL, Bissot T, et al. Use of starting condition score to estimate
changes in body weight and composition during weight loss in obese dogs. Res Vet
Sci 87:249–254, 2009
9. Van Loan MD, Withers P, Matthie J, et al. Use of bioimpedance spectroscopy to deter-
mine extracellular fluid, intracellular fluid, total body water, and fat-free mass.
Human Body Composition. Springer; 1993. p. 67–70
10. Eirmann LA, Freeman LM, Laflamme DP, et al. Comparison of Adipokine Concentra-
tions and Markers of Inflammation in Obese Versus Lean Dogs. Int J Appl Res Vet
Med 7:196–205, 2009
11. Chikamune T, Katamoto H, Ohashi F, et al. Serum lipid and lipoprotein concentra-
tions in obese dogs. J Vet Med Sci 57:595–598, 1995
12. Diez M, Michaux C, Jeusette I, et al. Evolution of blood parameters during weight
loss in experimental obese Beagle dogs. J Anim Physiol Anim Nutr 88:166–171, 2004
13. Pena C, Suarez L, Bautista I, et al. Relationship between analytic values and canine
obesity. J Anim Physiol Anim Nutr 92:324–325, 2008
14. Radakovich LB, Truelove MP, Pannone SC, et al. Clincally healthy overweight and
obese dogs differ from lean controls in select CBC and serum biochemistry values.
Vet Clin Pathol 0:1–6, 2017
15. Tribuddharatana T, Kongpiromchean Y, Sribhen K, et al. Biochemical alterations
and their relationships with the metabolic syndrome components in canine obe-
sity. Kasetsart J 45:622–628, 2011
16. Raman M, Ahmed I, Gillevet PM, et al. Fecal microbiome and volatile organic com-
pound metabolome in obese humans with nonalcoholic fatty liver disease. Clin
Gastroenterol Hepatol 11:868–875, 2013. e863
17. Oberbach A, Blü her M, Wirth H, et al. Combined proteomic and metabolomic profiling
of serum reveals association of the complement system with obesity and identifies
novel markers of body fat mass changes. J Proteome Res 10:4769–4788, 2011
18. Ley RE. Obesity and the human microbiome. Curr Opin Gastroenterol 26:5–11, 2010
19. Turnbaugh PJ, Backhed F, Fulton L, et al. Diet-induced obesity is linked to marked but
reversible alterations in the mouse distal gut microbiome. Cell Host Microbe 3:213–
223, 2008
20. Brignardello J, Morales P, Diaz E, et al. Pilot study: alterations of intestinal micro-
biota in obese humans are not associated with colonic inflammation or disturban-
ces of barrier function. Aliment Pharmacol Ther 32:1307–1314, 2010
21. Kardinaal AF, van Erk MJ, Dutman AE, et al. Quantifying phenotypic flexibility as
the response to a high-fat challenge test in different states of metabolic health. The
FASEB Journal 29:4600–4613, 2015
22. Balamurugan R, George G, Kabeerdoss J, et al. Quantitative differences in intestinal
Faecalibacterium prausnitzii in obese Indian children. Br J Nutr 103:335, 2010
23. Armougom F, Henry M, Vialettes B, et al. Monitoring bacterial community of
human gut microbiota reveals an increase in Lactobacillus in obese patients and
Methanogens in anorexic patients. PloS One 4:e7125, 2009
24. Walters WA, Xu Z, Knight R. Metaanalyses of human gut microbes associated
with obesity and IBD. FEBS Lett 588:4223–4233, 2014
25. Park HJ, Lee SE, Kim HB, et al. Association of obesity with serum leptin, adipo-
nectin, and serotonin and gut microflora in beagle dogs. J Vet Intern Med
29:43–50, 2015
26. Handl S, German AJ, Holden SL, et al. Faecal microbiota in lean and obese dogs.
FEMS Microbiol Ecol 84:332–343, 2013
27. Zhang A, Sun H, Wang X. Power of metabolomics in biomarker discovery and min-
ing mechanisms of obesity. Obes Rev 14:344–349, 2013
28. Ding W, Mak RH. Early markers of obesity-related renal injury in childhood. Pediatr
Nephrol 30:1–4, 2015
29. Kim H-J, Kim JH, Noh S, et al. Metabolomic analysis of livers and serum from high-
fat diet induced obese mice. J Proteome Res 10:722–731, 2010
30. Kim JY, Park JY, Kim OY, et al. Metabolic profiling of plasma in overweight/obese
and lean men using ultra performance liquid chromatography and Q-TOF mass
spectrometry (UPLC¡ Q-TOF MS). J Proteome Res 9:4368–4375, 2010
31. Kim OY, Lee JH, Sweeney G. Metabolomic profiling as a useful tool for diagnosis
and treatment of chronic disease: focus on obesity, diabetes and cardiovascular
diseases. Expert Rev Cardiovasc Ther 11:61–68, 2013
32. Calvani R, Brasili E, Pratico G, et al. Application of NMR-based metabolomics to the
study of gut microbiota in obesity. J Clin Gastroenterol 48:S5–S7, 2014
33. Weir TL, Manter DK, Sheflin AM, et al. Stool microbiome and metabolome dif-
ferences between colorectal cancer patients and healthy adults. PloS One
8:e70803, 2013
34. Minamoto Y, Otoni CC, Steelman SM, et al. Alteration of the fecal microbiota and
serum metabolite profiles in dogs with idiopathic inflammatory bowel disease.
Gut Microbes 6:33–47, 2015
35. Sun H, Wang X. Power of metabolomics in biomarker discovery and mining mech-
anisms of obesity. Obes Rev 14:344–349, 2013
36. Mihalik SJ, Michaliszyn SF, De Las Heras J, et al. Metabolomic profiling of fatty acid
and amino acid metabolism in youth with obesity and type 2 diabetes. Diabetes
Care 35:605–611, 2012
37. Fiehn O, Garvey WT, Newman JW, et al. Plasma metabolomic profiles reflective of
glucose homeostasis in non-diabetic and type 2 diabetic obese African-American
women. PloS One 5:e15234, 2010
38. Forster GM, Hill D, Gregory G, et al. Effects of cooked navy bean powder on appar-
ent total tract nutrient digestibility and safety in healthy adult dogs. J Anim Sci
90:2631–2638, 2012
39. Forster GM, Ollila CA, Burton JH, et al. Nutritional weight loss therapy with cooked
bean powders regulates serum lipids and biochemical analytes in overweight and
obese dogs. J Obes Weight Loss Ther 2:149, 2012
40. Forster GM, Nealon NJ, Hill D, et al. Navy and black bean-based dog foods are
digestible during weight loss in overweight and obese adult companion dogs.
J Appl Anim Nutr E4:4, 2016
41. Caporaso JG, Lauber CL, Walters WA, et al. Global patterns of 16S rRNA diversity at
a depth of millions of sequences per sample. Proc Natl Acad Sci 108:4516–4522,
2011
42. Lauber CL, Strickland MS, Bradford MA, et al. The influence of soil properties on the
structure of bacterial and fungal communities across land-use types. Soil Biol Bio-
chem 40:2407–2415, 2008
43. Caporaso JG, Kuczynski J, Stombaugh J, et al. QIIME allows analysis of high-
throughput community sequencing data. Nat Methods 7:335–336, 2010
44. Lozupone C, Knight R. UniFrac: a new phylogenetic method for comparing micro-
bial communities. Appl Environ Microbiol 71:8228–8235, 2005
45. Lauber CL, Hamady M, Knight R, et al. Pyrosequencing-based assessment of soil pH
as a predictor of soil bacterial community structure at the continental scale. Appl
Environ Microbiol 75:5111–5120, 2009
46. R-Core-Team. R: A Language and Envrionment for Statistical Computing. Vienna,
Austria; 2014
47. Paulson JN, Stine OC, Bravo HC, et al. Differential abundance analysis for microbial
marker-gene surveys. Nat Methods 10:1200–1202, 2013
48. McMurdie PJ, Holmes S. Waste not, want not: why rarefying microbiome data is
inadmissible. PLoS Comput Biol 10:e1003531, 2014
49. Forster GM, Heuberger AL, Broeckling CD, et al. Consumption of cooked navy bean
powders modulate the canine fecal and urine metabolome. Curr Metabolomics
3:90–101, 2015
50. Smith CA, Want EJ, O'Maille G, et al. XCMS: processing mass spectrometry data for
metabolite profiling using nonlinear peak alignment, matching, and identification.
J Analyt Chem 78:779–787, 2006
51. Broeckling CD, Afsar FA, Neumann S, et al. RAMClust: a novel feature clustering
method enables spectral-matching-based annotation for metabolomics data. Anal
Chem 86:6812–6817, 2014
52. Sumner LW, Amberg A, Barrett D, et al. Proposed minimum reporting standards for
chemical analysis Chemical Analysis Working Group (CAWG) Metabolomics
Standards Initiative (MSI). Metabolomics 3:211–221, 2007
53. Smyth GK. Linear models and empirical bayes methods for assessing differential
expression in microarray experiments. Stat Appl Genet Mol Biol 3, 2004. Article3
54. Legendre P, Gallagher ED. Ecologically meaningful transformations for ordination
of species data. Oecologia 129:271–280, 2001
55. Middelbos IS, Boler BM, Qu A, et al. Phylogenetic characterization of fecal microbial
communities of dogs fed diets with or without supplemental dietary fiber using
454 pyrosequencing. PloS One 5:e9768, 2010
56. Kerr KR, Forster G, Dowd SE, et al. Effects of dietary cooked navy bean on the fecal
microbiome of healthy companion dogs. PLoS One 8:e74998, 2013
57. Pallister T, Jackson M, Martin T, et al. Untangling the relationship between diet and
visceral fat mass through blood metabolomics and gut microbiome profiling. Int J
Obes 41:1106, 2017
58. Louis P, Flint HJ. Formation of propionate and butyrate by the human colonic
microbiota. Environ Microbiol 19:29–41, 2017
59. Yin Y-N, Yu Q-F, Fu N, et al. Effects of four Bifidobacteria on obesity in high-fat diet
induced rats. World J Gastroenterol 16:3394–3401, 2010
60. Turnbaugh PJ, Hamady M, Yatsunenko T, et al. A core gut microbiome in obese and
lean twins. Nature 457:480–484, 2009
61. Barcenilla A, Pryde SE, Martin JC, et al. Phylogenetic relationships of butyrate-pro-
ducing bacteria from the human gut. Appl Environ Microbiol 66:1654–1661, 2000
62. Suchodolski JS, Markel ME, Garcia-Mazcorro JF, et al. The fecal microbiome in dogs
with acute diarrhea and idiopathic inflammatory bowel disease. PLoS One
7:e51907, 2012
63. Kasai C, Sugimoto K, Moritani I, et al. Comparison of the gut microbiota composi-
tion between obese and non-obese individuals in a Japanese population, as ana-
lyzed by terminal restriction fragment length polymorphism and next-generation
sequencing. BMC Gastroenterol 15:100, 2015
64. Joyce SA, MacSharry J, Casey PG, et al. Regulation of host weight gain and lipid
metabolism by bacterial bile acid modification in the gut. Proc Natl Acad Sci
111:7421–7426, 2014

65. Caesar R, Fak F, Ba€ckhed F. Effects of gut microbiota on obesity and atherosclerosis
via modulation of inflammation and lipid metabolism. J Int Med 268:320–328,
2010
66. Kumar J, Karlsson R, D Broeckling C, et al. Associations of body mass index and obesity-
related genetic variants with serum metabolites. Curr Metabolomics 2:27–36, 2014
67. Galili O, Versari D, Sattler KJ, et al. Early experimental obesity is associated with
coronary endothelial dysfunction and oxidative stress. Am J Physiol Heart Circ Phys-
iol 292:H904–H911, 2007
 G.M. Forster et al. / Topics in Companion An Med 33 (2018) 126135
68. RC de Godoy ML, Pappan KW, Grant RS, Swanson K. Plasma metabolite profiling
and search for biomarkers of metabolic dysfunction in dogs undergoing rapid
weight gain. Curr Metabolomics 3:102–121, 2015
69. Jeusette IC, Lhoest ET, Istasse LP, et al. Influence of obesity on plasma lipid and lipo-
protein concentrations in dogs. Am J Vet Res 66:81–86, 2005
70. Farvid MS, Ding M, Pan A, et al. Dietary linoleic acid and risk of coronary heart dis-
ease: a systematic review and meta-analysis of prospective cohort studies. Circula-
tion 137:114, 2014
71. Srinivasan M, Sudheer AR, Menon VP. Ferulic acid: therapeutic potential through
its antioxidant property. J Clin Biochem Nutr 40:92–100, 2007
72. Ooi JH, Li Y, Rogers CJ, et al. Vitamin D regulates the gut microbiome and protects
mice from dextran sodium sulfateinduced colitis. J Nutr 143:1679–1686, 2013
73. Hewison M. Antibacterial effects of vitamin D. Nat Rev Endocrinol 7:337–345, 2011
74. Calvani R, Miccheli A, Capuani G, et al. Gut microbiome-derived metabolites char-
acterize a peculiar obese urinary metabotype. Int J Obes 34:1095–1098, 2010
135
€ der J, Hagman R, Dicksved J, et al. The urine metabolome differs between lean
75. So
and overweight Labrador Retriever dogs during a feed-challenge. PloS One 12:
e0180086, 2017
76. Salek RM, Maguire ML, Bentley E, et al. A metabolomic comparison of urinary changes
in type 2 diabetes in mouse, rat, and human. Physiol Genomics 29:99–108, 2007
77. Center S, Slater M, Manwarren T, et al. Diagnostic efficacy of serum alkaline phospha-
tase and gamma-glutamyltransferase in dogs with histologically confirmed hepato-
biliary disease: 270 cases (1980-1990). J Am Vet Med Assoc 201:1258–1264, 1992
78. Cazzola R, Rondanelli M, Russo-Volpe S, et al. Decreased membrane fluidity and
altered susceptibility to peroxidation and lipid composition in overweight and
obese female erythrocytes. J Lipid Res 45:1846–1851, 2004
79. Boag AK, Coe RJ, Martinez TA, et al. Acidbase and electrolyte abnormalities in
dogs with gastrointestinal foreign bodies. J Vet Int Med 19:816–821, 2005
80. Valdes AM, Glass D, Spector TD. Omics technologies and the study of human age-
ing. Nat Rev Genet 14:601, 2013