Osteoarthritis and Cartilage 24 (2016) 1470e1478

Comparative lipidomic analysis of synovial fluid in human and canine

osteoarthritis
M.K. Kosinska y, S.C. Mastbergen z, G. Liebisch x, J. Wilhelm k, R.B. Dettmeyer ¶,
B. Ishaque y, M. Rickert y, G. Schmitz x, F.P. Lafeber z, J. Steinmeyer y *
y Laboratory for Experimental Orthopaedics, Department of Orthopaedics, Justus-Liebig-University Giessen, Germany
z Department of Rheumatology and Clinical Immunology, University Medical Centre Utrecht, Utrecht, The Netherlands
x Department of Clinical Chemistry and Laboratory Medicine, University Hospital Regensburg, Germany
k Medical Clinic II/IV, Justus-Liebig-University Giessen, Germany
¶ Institute of Forensic Medicine, Justus-Liebig-University Giessen, Germany

a r t i c l e i n f o s u m m a r y

Article history: Objective: The lipid profile of synovial fluid (SF) is related to the health status of joints. The early stages of
Received 23 September 2015 human osteoarthritis (OA) are poorly understood, which larger animals are expected to be able to model
Accepted 25 March 2016 closely. This study examined whether the canine groove model of OA represents early OA in humans
based on the changes in the lipid species profile in SF. Furthermore, the SF lipidomes of humans and dogs
Keywords: were compared to determine how closely canine lipid species profiles reflect the human lipidome.
Osteoarthritis
Methods: Lipids were extracted from cell- and cellular debris-free knee SF from nine donors with healthy
Synovial fluid
joints, 17 patients with early and 13 patients with late osteoarthritic changes, and nine dogs with knee
Lipidomics
Model
OA and healthy contralateral joints. Lipid species were quantified by electrospray ionization tandem
Dog mass spectrometry (ESI-MS/MS).
Human Results: Compared with control canine SF most lipid species were elevated in canine OA SF. Moreover,
the lipid species profiles in the canine OA model resembled early OA profiles in humans. The SF lip-
idomes between dog and human were generally similar, with differences in certain lipid species in the
phosphatidylcholine (PC), lysophosphatidylcholine (LPC) and sphingomyelin (SM) classes.
Conclusions: Our lipidomic analysis demonstrates that SF in the canine OA model closely mimics the
early osteoarthritic changes that occur in humans. Further, the canine SF lipidome often reflects normal
human lipid metabolism.
© 2016 Osteoarthritis Research Society International. Published by Elsevier Ltd. All rights reserved.

Introduction that results in sclerosis and osteophytes, synovial hyperplasia, joint

Osteoarthritis (OA) is characterized primarily by the progressive
loss of articular cartilage e the remodeling of subchondral bone
* Address correspondence and reprint requests to: J. Steinmeyer, Laboratory for
Experimental Orthopaedics, Department of Orthopaedics, Justus-Liebig- University
Giessen, Paul-Meimberg-Strasse 3, 35392 Giessen, Germany. Tel: 49-(0)641-985-
42920; Fax: 49-(0)641-985-42939.
E-mail addresses: m.kosinska@gmail.com (M.K. Kosinska), s.mastbergen@
umcutrecht.nl (S.C. Mastbergen), gerhard.liebisch@klinik.uni-regensburg.de
(G. Liebisch), jochen.wilhelm@patho.med.uni-giessen.de (J. Wilhelm), reinhard.
dettmeyer@forens.med.uni-giessen.de (R.B. Dettmeyer), bernd.ishaque@ortho.
med.uni-giessen.de (B. Ishaque), markus.rickert@ortho.med.uni-giessen.de
(M. Rickert), gerd.schmitz@klinik.uni-regensburg.de (G. Schmitz), f.lafeber@
umcutrecht.nl (F.P. Lafeber), juergen.steinmeyer@ortho.med.uni-giessen.de
(J. Steinmeyer).
effusion, and secondary synovitis1,2. Although OA is the most
prevalent joint disorder and develops slowly over many years, it is
often only diagnosed in the late stages of disease3,4. Most patients
generally seek medical attention due to increasing discomfort and
pain at an advanced stage of disease, after changes in articular
cartilage have occurred4,5. Because early biomarkers are unavai-
lable6, the diagnosis of OA is frequently based on radiographic
signs, such as narrowing of the joint space during the late stages4,7.
One reason that the early phase of OA is poorly understood is the
lack of samples for research. Specifically, human synovial tissue,
cartilage, and synovial fluid (SF) that have undergone early osteo-
arthritic changes are difficult to acquire. However, the alterations
that occur early during the onset of OA have garnered significant
interest with regard to the development of therapies that limit the
progression of OA8.

http://dx.doi.org/10.1016/j.joca.2016.03.017
1063-4584/© 2016 Osteoarthritis Research Society International. Published by Elsevier Ltd. All rights reserved.
 M.K. Kosinska et al. / Osteoarthritis and Cartilage 24 (2016) 1470e1478
Over 30 OA animal models have been developed to study the
pathophysiology of OA and examine treatment modalities9,10.
Proper selection of an animal model is an essential consideration in
the design of a research study11. Larger animals mimic the human
condition better regarding the anatomy, cartilage composition (e.g.,
cell matrix distribution), and loading patterns of joints. In contrast,
smaller laboratory models, such as mice and rats, are less trans-
lational but incur lower costs. Animal models of OA can be divided
into spontaneously occurring OA12; surgically-induced OA, effected
by destabilization of the medial meniscus, transection of the
anterior cruciate ligament (ACLT), or transection of the cranial
cruciate ligament13,14; and chemically-induced OA, through intra-
articular injection of chemicals, such as mono-iodoacetate15.
However, none of these models reproduces the full spectrum of
changes observed in human OA.
The frequently used groove and ACLT models of OA were
developed in domestic dogs16e20. Canine ACLT leads to permanent
instability in the knee joint, causing degenerative changes in
articular cartilage and inflammation in synovial tissue, similar to
human OA13,20. In the canine groove model, artificial grooves
damage the articular cartilage of the weight-bearing areas of the
femoral condyles in one knee, triggering the development of OA. To
exacerbate this inducer of OA, loading of the affected joint can be
intensified by temporarily fixing the contralateral control limb to
the trunk of the dog16e20. By biochemical and histological evalua-
tion, degenerative changes occur in the joint, are similar to those of
the ACLT model17,18. Moreover, the canine groove model resembles
the human condition regarding the biochemical changes in the
cartilage17. This model also results in mild synovial inflammation
and negligible joint effusion17,18, and is advantageous because the
contralateral joint of each dog can serve as an internal control.
SF lubricates articular cartilage, effecting near frictionless mo-
tion of joints. Hyaluronic acid21, lubricin22,23, and surface-active
phospholipids24,25 are three major SF constituents that provide
boundary lubrication independently and cooperatively21,26,27.
Changes in the glycerophospholipid composition and concentra-
tion of SF can impair lubrication whereas an altered level of
sphingolipids can modulate the inflammatory status of the
joints28,29.
Because lipids are related to the health status of human synovial
joints, we hypothesized that similar alterations to the lipid
composition of SF would be detected in the canine groove model of
OA. In this study, the levels of lipid species were measured in the
canine groove model of OA to determine whether they resemble
the early or late stages of human OA. Also, we compared the SF lipid
profile of knee joints in healthy humans and dogs to evaluate how
closely the canine SF lipidome resembles that of humans.
Materials and methods
Description of the cohorts
Canine SF
Nine skeletally mature mixed-breed domestic dogs [9 females,
mean age 26 (25e28) months, mean weight 17.5 ± 1.3 kg] were kept
in the animal laboratory of Utrecht University, The Netherlands. The
dogs were housed in groups of 2e3 per pen and were allowed to
exercise in groups on a large patio daily for at least 2 h. The Utrecht
University Medical Ethical Committee for Animal Studies approved
this study.
OA was induced in the right knee joint of each dog using the
groove model as described17,18. Under general anesthesia, surgery
was performed through a 2e2.5-cm medial incision near the liga-
mentum patellae. In a state of maximum flexion, approximately 10
longitudinal and diagonal grooves were made on the weight-
1471
bearing sections of the femoral condyles using a 1.5-mm-diam-
eter Kirschner wire, which was bent at a 90 angle 0.5 mm from the
tip. This kink ensured that the depth of the grooves was restricted
to roughly 0.5 mm, preventing a damage to the underlying sub-
chondral bone. The contralateral left knee joints were not treated
and served as internal controls. The dogs were sacrificed 45 weeks
after induction of OA with an intravenous injection of pentobarbital
(Euthesate).
At the time of death, both hind limbs were amputated, and SF
was collected by direct aspiration for processing within 2 h. All
knee joints were evaluated macro- and microscopically, and the
harvested SF (described below) was stored at 86 C. SF samples
were shipped on dry ice to laboratories in Giessen and stored
at 86 C prior to analysis.
Compared with the contralateral control knee joints with
healthy smooth cartilage (canine OARSI macroscopic score 0; scale
0e5), OA knee joints were characterized by increased damage to
the cartilage surface and cartilage fibrillation [OARSI macroscopic
score: tibia: 1.8 (1.0e1.8), condyle: 2.8 (2.5e3.0)]. Moreover, there
was no evidence of effusion in OA knees. Mild inflammation was
evaluated histologically and macroscopically in synovial tissue
[OARSI macroscopic score 2.0 (2.0e2.0)]. By histology, OA cartilage
suffered damage to the cartilage surface and chondrocyte clusters
were observed. Also, compared with the contralateral control knee,
proteoglycan content was lower and the levels of denatured
collagen were elevated in OA cartilage as determined
biochemically30,31.
Human SF
Human SF was obtained from the knee joints of nine postmor-
tem donors with no documented history of joint disease [8 males, 1
female, mean age 22 (21e25) years, mean BMI 24.8 (20.8e25.2)]
and 30 patients with OA as described28. Briefly, OA stage was
defined per the Kellgren/Lawrence (K/L) scale for radiographic
severity (range 0e4) and based on the macroscopic appearance of
six cartilage surfacesdpatella, trochlea, medial and lateral femur,
and medial and lateral tibia-per the Outerbridge (OU) classification
guide32 as already described28. Seventeen patients with early
osteoarthritic changes in the cartilage [defined as OU 2, K/L
0 (0e1), 11 males, 6 females, mean age 36 (25e49) years, mean BMI
24.7 (22.6e26.9)], and 13 patients with late OA [defined as OU> 2,
K/L 3 (3e4), 6 males, 7 females, mean age 70 (67e74) years, mean
BMI 27.7 (25.8e30.1)] were included.
Sampling of SF
SF samples were collected by arthrocentesis of the knee joints
from dogs or during arthroscopy (early OA, human) and total knee
replacement surgery (late OA, human) and processed as
described28. In brief, fresh unfrozen SF samples were immediately
analyzed macroscopically to exclude SF with blood contamination
and high turbidity before they were diluted when necessary and
incubated for 15 min at 37 C. Then, the samples were passed
through a 1.2-mm filter to remove any remaining cells before 10% (v/
v) protease and phospholipase inhibitors were added; the resulting
mixture was centrifuged at 16,100 g for 45 min at room temper-
ature to remove cellular particles, and the supernatant was stored
at 86 C for later analysis28.
Chemicals
Unless otherwise specified, all chemicals were obtained from
Sigma (Taufkirchen, Germany). HPLC-grade methanol and chloro-
form, gentamycin sulfate, and neomycin sulfate were purchased
from Merck (Darmstadt, Germany). The lipid standards were pur-
chased from Avanti Polar Lipids (Alabaster, AL, USA).
1472 M.K. Kosinska et al. / Osteoarthritis and Cartilage 24 (2016) 1470e1478
Lipid extraction and mass spectrometry analysis
Lipids were extracted in the presence of nonnaturally occur-
ring lipid species, which served as internal standards as
described28,33,34. Phospholipid and sphingolipid species were
quantified by electrospray ionization tandem mass spectrometry
(ESI-MS/MS) on a Quattro Ultima™ triple quadrupole mass
spectrometer (Micromass, Manchester, United Kingdom) using
the analytical setup and data analysis algorithm as reported34.
The isotopic overlap of lipid species was corrected and data
analysis was performed with self-programmed Excel Macros for
each lipid class34. Lipid species were annotated per the Lip-
idomicNet proposal for shorthand notation of lipid structures
that are derived from mass spectrometry35. This method, estab-
lished for humans, was applied to dogs without any further
adaptations.
Statistical analysis
Statistically significant differences in lipid concentrations be-
tween cohorts were analyzed by paired t-test of the log-values
(Figs. 1e3, 7). Analysis of variance (ANOVA) and t-tests of the log-
values were performed for the three cohorts (Figs. 4e6). P-values
of less than 0.05 were considered to be statistically significant.
Figures show the medians (or median differences) and interquartile
ranges. Unless otherwise indicated, the values in the text are me-
dians, for which the interquartile ranges are indicated in brackets.
The statistical analysis was performed using the statistical software
program R, version 3.1.0.
Results
SF lipid profile in the canine OA model
OA was induced in the right knee joints of nine dogs, and the
contralateral left knee joints served as untreated normal controls
(Fig. 1). Compared with canine control SF, canine OA SF had
significantly higher levels of total lipids and lipid classes,
including phosphatidylcholine with ether phosphatidylcholine
[PC(O)], lysophosphatidylcholine (LPC), phosphatidylinositol
(PI), and sphingomyelin (SM). The total lipid content was
Fig. 1. The concentrations of lipid classes in normal and OA SF from the knee joints of nine dogs. Lipids were quantified by ESI/MS-MS as described in Materials and methods. Data
are presented as medians with interquartile ranges. P-values less than 0.05 were considered significant. P-values marked with * are less than 0.001. PC(O)-
phosphatidylcholine þ ether phosphatidylcholine, LPC-lysophosphatidylcholine, PI-phosphatidylinositol, SM-sphingomyelin.
elevated 2.6-fold (P  0.001) in OA SF [375.4 nmol/mL
(191.6e416.9 nmol/mL)] compared with controls [147.1 nmol/mL
(103.6e239.1 nmol/mL)].
Further, compared with control SF, the concentrations of lipid
classes in canine OA SF rose 2.6-fold for PC(O) [122.0 nmol/mL
(82.5e200.8 nmol/mL) to 313.9 nmol/mL (155.4e356.5 nmol/mL),
P ¼ 0.021], 1.8-fold for LPC [6.0 nmol/mL (4.5e8.6 nmol/mL) to
11.1 nmol/mL (7.4e18.1 nmol/mL), P ¼ 0.020], 2.2-fold for PI
[5.1 nmol/mL (3.1e7.0 nmol/mL) to 11.2 (8.1e13.4), P ¼ 0.050], and
2.5-fold for SM [14.6 nmol/mL (11.8e25.3 nmol/mL) to 36.9 nmol/
mL (20.7e39.1 nmol/mL), P ¼ 0.020] (Fig. 1).
A total of 389 individual lipid species were analyzed, only 54
different lipid species belonged to these four mentioned lipid
classes and were above the limit of detection in canine control and
OA SF (Figs. 1e3 and Supplementary Table I) e 30 PC(O), 6 LPC, 14
SM, and 4 PI species. Notably, versus canine control SF, 34 lipid
species were elevated in canine OA SF (Figs. 2 and 3 and
Supplementary Table I).
Knee SF lipidome of the canine OA model vs human OA
Knee SF lipid classes and species profiles in canine OA were
measured and compared with early and human late OA27 to
determine whether the canine OA groove model resembles changes
in early or late human OA. The levels of PC(O), LPC, and SM classes
and species profiles were comparable between canine OA and early
human OA SF. However, the amounts were significantly elevated in
late human OA compared with canine OA SF (Fig. 4).
The relative differences [2.4-fold (1.1e3.9-fold)] in PC(O) con-
centrations in canine OA SF were similar to those in early human OA
[2.3-fold (1.5e3.0-fold)] but less than the magnitude in late human
OA [3.8-fold (2.5e7.5-fold), P  0.001]. Versus canine OA SF [2.4-
fold (1.1e3.2-fold)], the median change in LPC class was more
similar to early human OA [3.23-fold (2.8e4.7-fold)], but elevated
in late human OA [4.8-fold (3.6e8.7-fold), P  0.001]. Similar
findings were observed for the SM class (Fig. 4). The median change
in SM class was similar between canine OA SF [2.7-fold (1.0e3.0-
fold)] and early human OA [2.1-fold (1.4e3.5-fold)], but was
higher for late human OA [3.5-fold (2.4e6.9-fold), P  0.001].
Notably, human SF and canine SF had comparable lipid com-
positions. The measurements for lipid species often reflected the
 M.K. Kosinska et al. / Osteoarthritis and Cartilage 24 (2016) 1470e1478
Fig. 2. The concentrations of PC (A and B) and PC O (C) species in normal and OA SF
from the knee joints of nine dogs. Lipid species were quantified by ESI/MS-MS as
described in Materials and methods. Ether species were assigned based on the
assumption that the PCs contain only fatty acids with an even number of carbon atoms.
Data are presented as medians with interquartile ranges. P-values less than 0.05 were
considered significant. P-values marked with * are less than 0.001. PC-
phosphatidylcholine, PC O-ether phosphatidylcholine.
results that were observed for the corresponding classes (Figs. 5
and 6). For instance, compared with controls, the relative differ-
ences were similar between canine OA and early human OA SF, but
were higher in late human OA for most SM species [P  0.001,
Fig. 6(B) and (C)].
1473
Fig. 3. The concentrations of LPC (A), and SM (B and C) species in normal and OA SF
from the knee joints of nine dogs. Lipid species were quantified by ESI/MS-MS as
described in Materials and methods. SM species were assigned based on the
assumption of a sphingoid base with two hydroxyl groups. Data are presented as
medians with interquartile ranges. P-values less than 0.05 were considered significant.
P-values marked with * are less than 0.001. LPC-lysophosphatidylcholine, SM-
sphingomyelin.
Lipid composition of healthy human and canine SF
To measure the differences in the composition of SF lipids be-
tween the human and dog and determine how closely the lipidome
in dogs resembles that of human SF, we compared SF lipids from
nine humans with no documented history of joint disease and nine
dogs (left, healthy knees, Fig. 7). Human SF and canine SF had
1474 M.K. Kosinska et al. / Osteoarthritis and Cartilage 24 (2016) 1470e1478

Fig. 4. The levels of lipid classes in SF from the canine OA model compared with the early and late stages of human OA. Knee SF from early human OA (n ¼ 17), human donors with
healthy joints (n ¼ 9), late human OA (n ¼ 13), and nine dogs with OA and healthy control knees were analyzed by ESI-MS/MS as described in Materials and methods. The data for
lipid classes from human and canine OA SF are presented as relative differences to the average values from healthy human control SF (¼1) respectively to the canine contralateral
control SF values (¼1). Data are expressed as medians with interquartile ranges. P-values less than 0.05 were considered significant: P-values marked with * are less than 0.001.
PC(O)-phosphatidylcholine þ ether phosphatidylcholine, LPC-lysophosphatidylcholine, SM-sphingomyelin.

comparable lipid compositions, with PC, LPC and SM being the
major classes.
Although the concentration of the predominant lipid class,
PC(O), was similar between the human and dog, the concentrations
of the LPC and SM lipid classes were lower in canine vs human SF.
Compared with human SF, the concentration of the LPC class
decreased 0.39-fold (0.31e0.58-fold, P  0.001) in canine SF.
Similarly, the concentration of the SM class fell 0.39-fold
(0.33e0.68-fold, P  0.001) in canine vs human SF.
The levels of certain lipid species from the PC, LPC and SM
classes underwent mammal-specific changes (Fig. 7,
Supplementary Table I). For instance, compared with human SF, PC
species with highly polyunsaturated, long-chain fatty acids, such as
PC 38:4, PC 40:4, PC 38:5, and PC 40:5, were higher in canine SF,
whereas PC species with short-chain fatty acids, such as PC 32:0, PC
32:1 and PC 34:1, declined, Moreover, versus human SF, LPC spe-
cies, such as LPC 16:0, LPC 18:1, LPC 18:2 and LPC 20:4, and poly-
unsaturated SM species, such as SM 34:2, SM 38:2, SM 42:2 and SM
42:3, were markedly lower in canine SF vs human SF (Fig. 7,
Supplementary Table I).
Discussion
Changes in the SF lipid profile are associated with diminished
lubrication and modified inflammatory status in articular
joints28,29. Thus, lipid species can contribute to the development
and progression of OA. In this study, we established the first SF lipid
profile for a canine model of OA compared with human OA. We
have also provided data on the species dependency of the lip-
idomes from humans and dogs with healthy knee joints. The PC,
LPC, PI, and SM classes were determined in both mammalian spe-
cies. However, the levels of the phosphatidylethanolamine (PE), PE-
based plasmalogen (PLPE), and ceramide (Cer) classes, which are
detectable in human SF27,29, were below the limit of detection in
canine SF.
Understanding the differences in lipid profiles between healthy
and early OA can guide us toward the identification of diagnostic
markers for early OA. Greater insight into the pathophysiological
processes during the early stages of this widespread joint disease
can also accelerate the development of novel chondroprotective
treatments for OA. We have shown that SF from the canine groove
model of OA contains elevated lipid levels, consistent with our
findings in humans, wherein the levels of glycerophospholipids and
sphingolipids are higher in osteoarthritic SF28,29. Increased lipid
concentrations might confer protection to damaged articular
cartilage surfaces. However, long-term studies are required to
determine the effects of these changes in lipid concentrations on
cartilage integrity.
Seven of the 14 SM species, which might be involved in in-
flammatory and apoptotic processes, were elevated in the canine
OA model. Moreover, the sphingomyelinase pathway hydrolyzes
SM to release ceramides, which also induce proinflammatory cy-
tokines and apoptosis in cultured human fibroblast-like syn-
oviocytes36e38. Consistent with this evidence, macroscopic signs of
mild inflammation were observed in canine OA joints30.
Compared with other canine OA models, osteoarthritic changes
occur slowly in the groove model of OA17,18,39,40, which resembles
the early osteoarthritic changes that are observed in human knee
joints17. Our data support these observations, because similar
changes in lipid concentrations were observed in human OA and
the canine model of OA. Thus, the groove model resembles the early
stages of human OA, rendering it suitable for studying lipid meta-
bolism in early OA.
Several animal models of OA have been studied extensively9.
Compared with smaller laboratory animals, bigger animals are
advantageous, because larger and better distinguishable tissue
samples can be acquired from a single joint. Consequently, dogs
are frequently used to study OA11. However, it was unknown
whether a canine animal model had a SF lipid profile that
resembled human SF closely. The SF lipidomes of human and dog
primarily comprise PC, PI, LPC, and SM. Our results indicate that
the SF profiles of humans and dogs are in principle similar, but
there are clear differences in the levels of individual lipid
species.
The most prominent differences between humans and dogs
were found for PC. Compared with human SF, highly poly-
unsaturated species with longer fatty acid chains, such as PC 38:4,
PC 40:4, PC 38:5, and PC 40:5, were higher in canine SF. The
 M.K. Kosinska et al. / Osteoarthritis and Cartilage 24 (2016) 1470e1478 1475

Fig. 6. The levels of LPC (A), and SM (B and C) species in SF from canine OA model
Fig. 5. The levels of PC (A and B), and PC O (C) species in SF from canine OA model
compared with the early and late stages of human OA. Knee SF from early human OA
compared with the early and late stages of human OA. Knee SF from early human OA
(n ¼ 17), late human OA (n ¼ 13), human donors with healthy joints (n ¼ 9), and nine
(n ¼ 17), late human OA (n ¼ 13), human donors with healthy joints (n ¼ 9), and nine
dogs with OA and healthy contralateral control knees were analyzed by ESI-MS/MS as
dogs with OA and healthy contralateral control knees were analyzed by ESI-MS/MS as
described in Materials and methods. SM species were assigned based on the
described in Materials and methods. Ether species were assigned based on the
assumption of a sphingoid base with two hydroxyl groups. The data obtained for lipid
assumption that the PCs contain only fatty acids with an even number of carbon atoms.
classes from human and canine OA SF are presented as relative differences to values
The data obtained for lipid classes from human and canine OA SF are presented as
obtained from healthy control SF (¼1) of the same species. Data are expressed as
relative differences to average values from healthy human control SF (¼1) respectively
medians with interquartile ranges. P-values less than 0.05 were considered significant.
to the canine contralateral control SF values (¼1). Data are expressed as medians with
P-values marked with * are less than 0.001. LPC-lysophosphatidylcholine, SM-
the interquartile ranges. P-values less than 0.05 were considered significant. P-values
sphingomyelin.
marked with * are less than 0.001. PC-phosphatidylcholine, PC O-ether
phosphatidylcholine.
1476 M.K. Kosinska et al. / Osteoarthritis and Cartilage 24 (2016) 1470e1478

Fig. 7. The levels of PC (A and B), PC O (C), LPC (D), and SM (E and F) species in normal canine SF compared with SF from healthy human joints. Knee SF from human donors with
healthy joints (n ¼ 9) and from healthy canine joints (n ¼ 9) were analyzed by ESI-MS/MS as described in Materials and methods. Species annotation is based on the assumptions
shown in Figs. 2 and 3. The data obtained for lipid species from canine contralateral normal SF are presented as relative differences to average values obtained from SF of healthy
human joints (¼1). Data are expressed as medians with interquartile ranges. P-values less than 0.05 were considered significant. P-values marked with * are less than 0.001. PC-
phosphatidylcholine, PC O-ether phosphatidylcholine, LPC-lysophosphatidylcholine, SM-sphingomyelin.

lubricating properties of lipids depend on fatty acid chain length
and the number of double bounds41,42. Thus, lipids with longer fatty
acid chains and higher levels of unsaturation might have better
lubricating abilities. However, they have greater proinflammatory
potential, especially those with u-6 fatty acids. We can merely
speculate about the discrepancies in the lipidomes of the two
mammals that we investigated. Humans and especially dogs often
use their joints, exposing them to dynamic loads. Thus, SF with
highly lubricating properties appears to be essential.
Another explanation for the difference in lipid profiles between
the species is disparate dietary habits43. For example, humans are
generally omnivores, whereas dogs received a diet containing some
dry meat. However, the exact influence of dietary habits on lipid
composition remains poorly understood. Moreover, the ESI-MS/MS
 M.K. Kosinska et al. / Osteoarthritis and Cartilage 24 (2016) 1470e1478
method, established for human SF samples, was applied to canine
SF without any further adaptations e for instance, with regard to
fatty acids and sphingoid base composition.
This report is the first study to compare a wide range of phos-
pholipid and sphingolipid species in human and canine SF. Our
analysis reveals that the lipid profiles of dogs often reflect those of
humans. In addition, the canine groove model of OA closely re-
sembles the SF lipid profile of early OA SF in humans, supporting
domestic dogs as an appropriate species to study lipids during the
progression of OA.
Author contribution
All authors were involved in revising the manuscript critically
for important intellectual content, and all authors approved the
final version to be published. Marta K Kosinska had full access to all
of the data in the study and takes responsibility for the integrity of
the data and the accuracy of the data analysis. Study conception and
design: Juergen Steinmeyer. Acquisition of data: Marta K. Kosinska,
Simon C. Mastbergen, Gerhard Liebisch, Reinhard B. Dettmeyer,
Bernd Ishaque, Markus Rickert, Floris P. Lafeber. Analysis and
interpretation of data: Marta K. Kosinska, Simon C. Mastbergen,
Gerhard Liebisch, Jochen Wilhelm, Gerd Schmitz, Juergen Stein-
meyer. Drafting of the manuscript: Marta K. Kosinska, Juergen
Steinmeyer.
Conflict of interest
Neither author has any financial conflict of interest.
Role of the funding source
Funding for this work was provided in part by a grant of the DRB
Foundation and by the “LipidomicNet” project, funded within the
Seventh Framework Programme of the European Union (EU FP7
grant agreement no 202272). The sponsors had no involvement in
the study design and execution, data interpretation, writing of the
manuscript or the decision to submit the manuscript.
Acknowledgments
The authors wish to express their sincere thanks to Christiane
Hild, and Simone Düchtel for their expert technical support, and to
Manuela Do €ller for her assistance with the study organization.
Supplementary data
Supplementary data related to this article can be found at http://
dx.doi.org/10.1016/j.joca.2016.03.017.
References
1. Felson DT, Lawrence RC, Dieppe PA, Hirsch R, Helmick CG,
Jordan JM, et al. Osteoarthritis: new insights. Part 1: the dis-
ease and its risk factors. Ann Intern Med 2000;133:635e46.
2. Loeser RF. Osteoarthritis year in review 2013: biology. Osteo-
arthritis Cartilage 2013;21:1436e42.
3. Murray CJ, Vos T, Lozano R, Naghavi M, Flaxman AD,
Michaud C, et al. Disability-adjusted life years (DALYs) for 291
diseases and injuries in 21 regions, 1990e2010: a systematic
analysis for the Global Burden of Disease Study 2010. Lancet
2012;380:2197e223.
4. Hunter DJ, Eckstein F, Kraus VB, Losina E, Sandell L,
Guermazi A. Imaging biomarker validation and qualification
report: sixth OARSI Workshop on Imaging in Osteoarthritis
combined with third OA Biomarkers Workshop. Osteoarthritis
Cartilage 2013;21:939e42.
1477
5. Kean WF, Kean R, Buchanan WW. Osteoarthritis: symptoms,
signs and source of pain. Inflammopharmacology 2004;12:
3e31.
6. Lotz M, Martel-Pelletier J, Christiansen C, Brandi ML,
Bruyere O, Chapurlat R, et al. Value of biomarkers in osteoar-
thritis: current status and perspectives. Ann Rheum Dis
2013;72:1756e63.
7. Emrani PS, Katz JN, Kessler CL, Reichmann WM, Wright EA,
McAlindon TE, et al. Joint space narrowing and Kell-
greneLawrence progression in knee osteoarthritis: an analytic
literature synthesis. Osteoarthritis Cartilage 2008;16:873e82.
8. Steinmeyer J, Konttinen YT. Oral treatment options for
degenerative joint diseaseepresence and future. Adv Drug
Deliv Rev 2006;58:168e211.
9. Teeple E, Jay GD, Elsaid KA, Fleming BC. Animal models of
osteoarthritis: challenges of model selection and analysis.
AAPS J 2013;15:438e46.
10. Aigner T, Cook JL, Gerwin N, Glasson SS, Laverty S, Little CB,
et al. Histopathology atlas of animal model systems e over-
view of guiding principles. Osteoarthritis Cartilage
2010;18(Suppl 3):S2e6.
11. Longo UG, Loppini M, Fumo C, Rizzello G, Khan WS, Maffulli N,
et al. Osteoarthritis: new insights in animal models. Open
Orthop J 2012;6:558e63.
12. Mistry D, Oue Y, Chambers MG, Kayser MV, Mason RM.
Chondrocyte death during murine osteoarthritis. Osteoar-
thritis Cartilage 2004;12:131e41.
13. Pond MJ, Nuki G. Experimentally-induced osteoarthritis in the
dog. Ann Rheum Dis 1973;32:387e8.
14. Kamekura S, Hoshi K, Shimoaka T, Chung U, Chikuda H,
Yamada T, et al. Osteoarthritis development in novel experi-
mental mouse models induced by knee joint instability.
Osteoarthritis Cartilage 2005;13:632e41.
15. Naik SR, Wala SM. Arthritis, a complex connective and syno-
vial joint destructive autoimmune disease: animal models of
arthritis with varied etiopathology and their significance.
J Postgrad Med 2014;60:309e17.
16. Intema F, DeGroot J, Elshof B, Vianen ME, Yocum S,
Zuurmond A, et al. The canine bilateral groove model of
osteoarthritis. J Orthop Res 2008;26:1471e7.
17. Mastbergen SC, Marijnissen AC, Vianen ME, van
Roermund PM, Bijlsma JW, Lafeber FP. The canine ‘groove’
model of osteoarthritis is more than simply the expression of
surgically applied damage. Osteoarthritis Cartilage 2006;14:
39e46.
18. Marijnissen AC, van Roermund PM, TeKoppele JM, Bijlsma JW,
Lafeber FP. The canine ‘groove’ model, compared with the
ACLT model of osteoarthritis. Osteoarthritis Cartilage 2002;10:
145e55.
19. Marijnissen AC, van Roermund PM, Verzijl N, Tekoppele JM,
Bijlsma JW, Lafeber FP. Steady progression of osteoarthritic
features in the canine groove model. Osteoarthritis Cartilage
2002;10:282e9.
20. Brandt KD, Myers SL, Burr D, Albrecht M. Osteoarthritic
changes in canine articular cartilage, subchondral bone, and
synovium fifty-four months after transection of the anterior
cruciate ligament. Arthritis Rheum 1991;34:1560e70.
21. Schmidt TA, Gastelum NS, Nguyen QT, Schumacher BL, Sah RL.
Boundary lubrication of articular cartilage: role of synovial
fluid constituents. Arthritis Rheum 2007;56:882e91.
22. Jay GD, Torres JR, Warman ML, Laderer MC, Breuer KS. The role
of lubricin in the mechanical behavior of synovial fluid. Proc
Natl Acad Sci USA 2007;104:6194e9.
23. Jay GD, Waller KA. The biology of lubricin: near frictionless
joint motion. Matrix Biol 2014;39C:17e24.
1478 M.K. Kosinska et al. / Osteoarthritis and Cartilage 24 (2016) 1470e1478
24. Hills BA, Crawford RW. Normal and prosthetic synovial joints
are lubricated by surface-active phospholipid: a hypothesis.
J Arthroplasty 2003;18:499e505.
25. Schwarz IM, Hills BA. Surface-active phospholipid as the
lubricating component of lubricin. Br J Rheumatol 1998;37:
21e6.
26. Mirea DA, Trunfio-Sfarghiu AM, Matei CI, Munteanu B,
Piednoir A, Rieu JP, et al. Role of the biomolecular interactions
in the structure and tribological properties of synovial fluid.
Tribol Int 2013;16:302e11.
27. Kosinska MK, Ludwig TE, Liebisch G, Zhang R, Siebert HC,
Wilhelm J, et al. Articular joint lubricants during osteoarthritis
and rheumatoid arthritis display altered levels and molecular
species. PLoS One 2015;10:e0125192.
28. Kosinska MK, Liebisch G, Lochnit G, Wilhelm J, Klein H,
Kaesser U, et al. A lipidomic study of phospholipid classes and
species in human synovial fluid. Arthritis Rheum 2013;65:
2323e33.
29. Kosinska MK, Liebisch G, Lochnit G, Wilhelm J, Klein H,
Kaesser U, et al. Sphingolipids in human synovial fluid-a lip-
idomic study. PLoS One 2014;9:e91769.
30. Wiegant K, Intema F, van Roermund PM, Barten-van
Rijbroek AD, Doornebal A, Hazewinkel HA, et al. Evidence of
cartilage repair by joint distraction in a canine model of
osteoarthritis. Arthritis Rheumatol 2015;67:465e74.
31. Cook JL, Kuroki K, Visco D, Pelletier JP, Schulz L, Lafeber FP. The
OARSI histopathology initiative e recommendations for his-
tological assessments of osteoarthritis in the dog. Osteoar-
thritis Cartilage 2010;18(Suppl 3):S66e79.
32. Outerbridge RE. The etiology of chondromalacia patellae.
J Bone Joint Surg Br 1961;43B:752e7.
33. Liebisch G, Drobnik W, Lieser B, Schmitz G. High-throughput
quantification of lysophosphatidylcholine by electrospray
ionization tandem mass spectrometry. Clin Chem 2002;48:
2217e24.
34. Liebisch G, Lieser B, Rathenberg J, Drobnik W, Schmitz G. High-
throughput quantification of phosphatidylcholine and
sphingomyelin by electrospray ionization tandem mass spec-
trometry coupled with isotope correction algorithm. Biochim
Biophys Acta 2004;1686:108e17.
35. Liebisch G, Vizcaino HA, Ko€ feler H, Tro
€ tzmüller M, Griffiths WJ,
Schmitz G, et al. Shorthand notation for lipid structures
derived from mass spectrometry. J Lipid Res 2013;54:
1523e30.
36. Gerritsen ME, Shen CP, Perry CA. Synovial fibroblasts and the
sphingomyelinase pathway: sphingomyelin turnover and
ceramide generation are not signaling mechanisms for the
actions of tumor necrosis factor-alpha. Am J Pathol 1998;152:
505e12.
37. Ichinose Y, Eguchi K, Migita K, Kawabe Y, Tsukada T, Koji T,
et al. Apoptosis induction in synovial fibroblasts by ceramide:
in vitro and in vivo effects. J Lab Clin Med 1998;131:410e6.
38. Cutler RG, Mattson MP. Sphingomyelin and ceramide as reg-
ulators of development and lifespan. Mech Ageing Dev
2001;122:895e908.
39. Frost-Christensen LN, Mastbergen SC, Vianen ME, Hartog A,
DeGroot J, Voorhout G, et al. Degeneration, inflammation,
regeneration, and pain/disability in dogs following destabili-
zation or articular cartilage grooving of the stifle joint. Oste-
oarthritis Cartilage 2008;16:1327e35.
40. Myers SL, Brandt KD, O'Connor BL, Visco DM, Albrecht ME.
Synovitis and osteoarthritic changes in canine articular carti-
lage after anterior cruciate ligament transection. Effect of
surgical hemostasis. Arthritis Rheum 1990;33:1406e15.
41. Chen Y, Crawford RW, Oloyede A. Unsaturated phosphatidyl-
cholines lining on the surface of cartilage and its possible
physiological roles. J Orthop Surg Res 2007;2:14.
42. Trunfio-Sfarghiu AM, Berthier Y, Meurisse MH, Rieu JP. Role of
nanomechanical properties in the tribological performance of
phospholipid biomimetic surfaces. Langmuir 2008;24:
8765e71.
43. Huang T, Yu X, Shou T, Wahlqvist ML, Li D. Associations of
plasma phospholipid fatty acids with plasma homocysteine in
Chinese vegetarians. Br J Nutr 2013;109:1688e94.