The role of plasma cytokine levels, CRP and Selenoprotein S gene variation in OA.

Abstract

Investigating the association between plasma levels of cytokines and chemokines, Selenoprotein S (SELS) gene variation and osteoarthritis (OA) subtypes.
The genetics of osteoarthritis and progression (GARP) study consists of 191 sibling pairs with symptomatic OA at multiple joint sites. We have measured plasma levels of 17 cytokines and chemokines and genetic variation at the SELS gene.
Nine out of 17 serum markers could be assessed quantitatively, whereas eight markers were assessed qualitatively. Principal component analysis (PCA) on the quantitatively assessed markers and serum high sensitive C-reactive protein (S-HsCRP) revealed that three components underlie 61% of the total plasma variation. Three single nucleotide polymorphisms (SNPs) in the SELS gene revealed four common haplotypes, one of which, GAG (frequency 3.5%) showed significant association to an anti-inflammatory (P=0.019) and acute phase related (P=0.036) component. OA subtype analysis showed that one component (mainly representing chemokine variation) was significantly associated to hand OA and disc degeneration (P=0.029 and P=0.010 respectively) as well as a physical component score (PCS) (P=0.042). The CRP related component also showed a strong association to the PCS (P=0.007). SELS haplotypes showed no association to OA subtypes in the GARP study.
Genetic variation in the SELS gene associates to components representing inflammatory signaling. Another component, representing chemokine variation, showed association to hand OA and disc degeneration in the GARP study indicating chemokines may contribute to OA pathogenesis.

Full text

Page 1

The role of plasma cytokine levels, CRP and Selenoprotein S gene
variation in OA
S. D. Bosy*, M. Kloppenburgzx, E. Suchimany, E. van Beelenk, P. E. Slagboomy and I. Meulenbelty
yDepartment of Molecular Epidemiology, LUMC, Leiden, The Netherlands
zDepartment of Rheumatology, LUMC, Leiden, The Netherlands
xDepartment of Clinical Epidemiology, LUMC, Leiden, The Netherlands
kDepartment of Immunohematologie and Blood Transfusion, LUMC, Leiden, The Netherlands
Summary
Objective: Investigating the association between plasma levels of cytokines and chemokines, Selenoprotein S (SELS) gene variation and
osteoarthritis (OA) subtypes.
Methods: The genetics of osteoarthritis and progression (GARP) study consists of 191 sibling pairs with symptomatic OA at multiple joint sites.
We have measured plasma levels of 17 cytokines and chemokines and genetic variation at the SELS gene.
Results: Nine out of 17 serum markers could be assessed quantitatively, whereas eight markers were assessed qualitatively. Principal com-
ponent analysis (PCA) on the quantitatively assessed markers and serum high sensitive C-reactive protein (S-HsCRP) revealed that three
components underlie 61% of the total plasma variation. Three single nucleotide polymorphisms (SNPs) in the SELS gene revealed four com-
mon haplotypes, one of which, GAG (frequency 3.5%) showed significant association to an anti-inflammatory (P¼0.019) and acute phase
related (P ¼0.036) component. OA subtype analysis showed that one component (mainly representing chemokine variation) was significantly
associated to hand OA and disc degeneration (P ¼0.029 and P ¼0.010 respectively) as well as a physical component score (PCS)
(P ¼0.042). The CRP related component also showed a strong association to the PCS (P ¼0.007). SELS haplotypes showed no association
to OA subtypes in the GARP study.
Conclusion: Genetic variation in the SELS gene associates to components representing inflammatory signaling. Another component, repre-
senting chemokine variation, showed association to hand OA and disc degeneration in the GARP study indicating chemokines may contribute
to OA pathogenesis.
ª 2008 Osteoarthritis Research Society International. Published by Elsevier Ltd. All rights reserved.
Key words: Osteoarthritis, Inflammation, Selenoprotein S, CRP, Genetic.
Introduction
Osteoarthritis (OA) is a common joint disease and an impor-
tant cause of pain and disability in the general population.
Elucidation of common pathways that are involved in the
onset and progression of the disease will assist in the devel-
opment of new drug targets and provide a better manage-
ment of this disabling condition in the future. Several
studies have shown that genetic factors play an important
role in OA etiology1,2. Although OA is not regarded to be
an inflammatory disease, there is increasing evidence for
the involvement of an innate low grade systemic inflamma-
tory component which may partly explain the genetic sus-
ceptibility3e7. More specifically, support for the hypothesis
that local variation in cytokine levels in the joint may
influence OA onset and progression is found in the
catabolic effects of pro-inflammatory cytokines, and the
protective effects of anti-inflammatory cytokines in articular
cartilage8,9. Gene expression studies have shown that in
OA cartilage several cytokines and chemokines are highly
expressed10e12. The recent introduction of multiplexed cy-
tokine assays facilitates sensitive measurements of differ-
ent cytokines using small amounts of valuable sample
material from different origins. This sensitive technique al-
lows investigation of the basal levels of chemokines and cy-
tokines in plasma in relation to the ongoing disease
processes for diseases where, as in OA, no pronounced
and obvious inflammatory component is present.
In a previous study, Curran et al. showed that subjects
can be characterized by a high (proinflammatory) or low
(noninflammatory) plasma cytokine profile depending on
a common promoter single nucleotide polymorphism
(SNP) ?105G >A (rs28665122) in the Selenoprotein S
(SELS) gene. A significant association of this SELS SNP
was observed specifically with higher plasma levels of tu-
mor necrosis factor alpha (TNFa), interleukin beta (IL1b)
and interleukin 6 (IL6)13. SELS is a widely expressed pro-
tein involved in maintaining the functional integrity of the en-
doplasmatic reticulum (ER) by participating in the removal
of misfolded proteins and regulating the cellular redox
balance. SELS inhibition by small interfering RNA (siRNA)
revealed that a functionally impaired ER leads to activation
of numerous pro-inflammatory cytokines mediated by
*Address correspondence and reprint requests to: Steffan D.
Bos,LeidenUniversityMedical
Epidemiology (S-05-P), P.O. Box 9600, 2300 RC Leiden, The
Netherlands. Tel: 31-71-5269736; Fax: 31-71-5268280; E-mail:
s.bos@lumc.nl
Received 28 April 2008; revision accepted 18 October 2008.
Centre, SectionMolecular
Osteoarthritis and Cartilage (2009) 17, 621e626
ª 2008 Osteoarthritis Research Society International. Published by Elsevier Ltd. All rights reserved.
doi:10.1016/j.joca.2008.10.007
International
Cartilage
Repair
Society
621

Page 2

nuclear factor kb (NF-kb) activation13. In a cohort of patients
with intestinal inflammation, there was no association to the
SELS locus, however, the pro-inflammatory allele associ-
ated to a high serum CRP levels in Crohn Disease patients
with active disease14. Finally, SELS polymorphisms associ-
ated with coronary heart disease and ischemic stroke15.
These studies trigger the question whether SELS gene vari-
ation also influences inflammatory responses and the etiol-
ogy of OA. Hence, we set off to map the plasma levels of
17 cytokines and chemokines by use of a multiplexed bead
arraysysteminsubjectswithsymptomaticOAatmultiplejoint
sites of the GARP study. We tested whether SELS gene var-
iation influenced these inflammatory plasma mediators and
OA, and whether these markers associated to OA subtypes
and severity.
Materials and methods
THE GARP STUDY
The GARP study consists of 191 sibling pairs. All participants have symp-
tomatic OA at multiple sites in the hand or OA at two or more joint sites of
four joint sites examined16. Symptomatic OA was determined following the
American College of Rheumatology recommendations17e19whereas radio-
graphic OA (ROA) was scored according to Kellgren/Lawrence20. Details
on the GARP OA phenotype and inclusion criteria can be found in previous
publications16. Physical functioning was assessed with the physical compo-
nent score (PCS), a subscore of the Dutch validated RAND 36-item Health
Survey. This questionnaire covers health related aspects including social
functioning, role limitations, mental health and vitality21,22. A higher score
on the PCS indicates a better physical functioning. For the current study
we used the proportionate ROA score based on the presence of ROA at
each joint location and on the number of joints with ROA identical as de-
scribed previously. In short, scores 0, 1 and 2 represent respectively no,
uni- and bilateral hip and knee OA for these joint sites. The hand ROA score
(0e2) represents subjects with, respectively, 0e2, 3e6, and ?7 hand joints
affected out of 20 scored. For spinal discus degeneration (DD) score (0e2)
represents subjects with DD at respectively 0e2, 3e5 and ?6 levels out of
11 levels scored23. Hand OA following the American College of Rheumatol-
ogy (ACR) criteria17was analyzed in addition to the ROA criteria. Genomic
DNA was extracted from peripheral blood leukocytes using standard
protocols. Written informed consent as approved by the ethical committee
was obtained from all subjects in the GARP study.
SERUM AND PLASMA COLLECTION AND MEASUREMENTS
For each participant of the GARP study a morning serum and EDTA
plasma sample was collected. Samples were processed within 4 h upon
collection and stored at ?80?C until measurement. Serum high sensitive
C-reactive protein (S-HsCRP) was assayed earlier24. A 17-plex bead assay
provided by Bio-RAD was used to measure plasma levels of cytokines and
chemokines on a Luminex platform. Intra assay variation was estimated at
6.2%. The standard protocol was adjusted using twice the indicated amounts
of plasma sample in half the amount of sample buffer to increase signal in the
measurements. On each plate 10e15% of samples were in duplicate, in
which no inconsistencies were observed. For calculation of z-scores the
average value of duplos was used as a single value. For cytokines and
chemokines with less than 60% of the fluorescence levels above background
noise, a qualitative measure of detectable vs not detectable level was used
(qualitative analysis, QL). For cytokines and chemokines with more than
60% fluorescence levels above background noise, z-scores of fluorescence
levels were calculated per plate (quantitative analysis, QT). In the principal
component analysis (PCA) individuals with missing values in five or more
out of nine cytokine/chemokines measurements were excluded (N ¼ 13).
For the analyzed individuals remaining missing values were given the
specific marker mean score. Sample distribution was random per plate on
a total of eight plates.
GENOTYPING
In the current study three SNPs of the SELS were genotypes selected from
the original paper of Curran et al.13. SNP positions relative to translation start
of the SELS gene are ?105 (G > A, rs28665122), þ3705 (G > A, rs4965814),
and þ6218 (A> G, rs9874). SNPs were measured using hME? chemistry on
a matrix assisted laser adsorption/ionization time-of-flight mass spectrometry
Mass Spectrometer (Sequenom Inc., San Diego, CA, USA). Assays were de-
signed using the Sequenom MassARRAY Assay Design software (version
3.1). Assay conditions were standard conditions as described earlier25. In ad-
dition to the genotypes obtained by the Sequenom for rs28665122, a Taqman
assay (Applied Biosystems, Foster City, California, United States) was
performed to deal with low genotyping success rates this SNP using the
Sequenom technique. Primers used for the ABI-genotyping were: forward
primer 50GGGTCGGCCTGCGA and reverse primer 50CTTCCGGTGCGC
TCCTA, probes were 50TGGCCGGGACCAC labeled with VIC and 50TTGGC
CAGGACCAC labeled with FAM. Assays were run on a 7900HT (Applied
Biosystems) according to the manufacturers specifications. Genotypes of
both techniques were used in addition to each other where necessary and
were used as a control where both techniques provided reliable genotypes.
No inconsistencies were observed between techniques for the reliable
genotypes.
STATISTICAL ANALYSIS
PCA was applied to reduce the correlated data of the cytokines and che-
mokines plasma levels. Subjects (N ¼ 341) with available levels of S-HsCRP
and with >4 out of nine quantifiable markers were entered in the PCA anal-
ysis. In PCA analyses random missing data (see Table II) was replaced by
mean values. Both empirical criteria (percentage of variance explained by
factors and Eigen values >1) and interpretability were used to determine
the number of factors. We explored the interpretability of these factors after
applying a Varimax rotation with Kaiser Normalization. The loading score of
each variable onto the individual factors represents the contribution of that
variable to the variance observed in the resulting factor. For analysis only
individual variable contributions of >0.4 qualified for loading a component26.
A factor loading represents the linear relationship (Pearson correlation under
Varimax rotation) between a variable and a factor.
In order to assess the relationships between OA characteristics, genetic
variation at the SELS gene and the clusters of cytokines and chemokines,
a mixed model regression analyses was performed. To investigate the indi-
vidual associations, subject specific regression scores of each extracted
cluster were used as dependent variable and age, sex, BMI (Body Mass In-
dex) and all OA subtypes as co-variables. OA features and subtypes tested
were specific ROA scores as defined previously; knee (0e2), hip (0e2),
hand (0e2) and spinal DD (0e2)24and the PCS derived from the SF36 ques-
tionnaire. Furthermore, for each linear mixed model analysis family identity
numbers (representing family relations) as random variables in order to
model the familial dependencies that might occur for the levels.
Genotype distributions of SELS SNPs were checked by use of the HWE
program available at http://linkage.rockefeller.edu. Thesias V3.1 was used to
assess linkage disequilibrium between the SNPs and to assess the expected
haplotypic contribution to a mean quantitative measure in carriers of a spe-
cific haplotype27. Quantitative measures analyzed by use of Thesias are e.g.,
the subject specific regression scores as determined by the PCA analyses.
In individuals the expected quantitative measure is determined by the contri-
bution of the two carried haplotypes. The Thesias program allows for adjust-
ments of co-variables but not for familial relationship. Chi square analyses
were performed to test for association of individual SNPs or proportionate
ROA scores to the qualitatively analyzed cytokines and chemokines.
P-values are unadjusted to multiple testing. Analyses were done in SPSS
version 14 (SPSS, Chicago, IL, USA) unless mentioned otherwise.
Results
CYTOKINE AND CHEMOKINE MEASUREMENTS
Characteristics of the GARP study are displayed in
Table I. For all participants of the GARP study cytokine
and chemokine levels were analyzed in a plasma sample
by use of a 17-plex bead array system. Table II shows cy-
tokines and chemokines measured and the percentage of
samples with levels above detection limit. In nine instances
(see Table II) we were able to assess a semi quantitative
measure by the use of z-scores reflecting relative plasma
levels. For the remaining eight instances (see Table II),
plasma levels were analyzed in a qualitative matter (detect-
able vs not detectable).
ASSOCIATION ANALYSIS OF CYTOKINE/CHEMOKINE
LEVELS WITH OA FEATURES
Nine cytokines and chemokines and S-HsCRP were an-
alyzed in a quantitative manner, the remaining eight cyto-
kines and chemokines were qualitatively analyzed. In the
622
S. D. Bos et al.: Cytokine, CRP & SELS gene variation in OA

Page 3

quantitatively analyzed cytokine and chemokine z-scores
we observed substantial correlations (Supplementary Table
1). In addition to the association between S-HsCRP and
BMI shown previously24, Supplementary Table 1 also
shows the frequently observed association between BMI
and plasma IL6 levels. GARP subjects in the highest BMI
quartile (BMI > 29.1 kg/m2) had an OR of 1.7 (95% C.I.
1.01e2.82) to reside in the highest quartile of IL6 plasma
levels (P ¼ 0.042) when corrected for age, sex and familial
relationship.
To reduce the redundancy between the markers to more
independent components in which these variables cluster,
a PCA including all inflammatory markers for which a quan-
titative measure was available (see Table II) and S-HsCRP
was performed. Table III shows the three components that
were extracted. The coefficients depicted in Table III explain
how well each individual marker is represented within the
clusters. The marker levels of IL2, IL6, Granulocyte CSF
(GCSF) and IL10 loaded together on the first component,
explaining 34.1% of the total variation in the GARP study.
Component 2 is determined by three chemokines MCP
(Monocyte Chemotactic Protein), IL8 and MIP (Macrophage
Inflammatory Protein), explaining 15.7% of the variation,
whereas the third component is determined by S-HsCRP,
IL5 and IL7 explaining 11.6% of the variation. It should be
noted that S-HsCRP has a negative value in the third com-
ponent (Table III), indicating that on average, within sub-
jects there is an inverse relation between S-HsCRP levels
and IL7 and IL5.
Subsequently the relationship between the three compo-
nents as dependant variables and the presence of OA char-
acteristics (Table I) as co-variables was investigated by
mixed model regression analysis. The upper section of
Table IV shows that component 2, consisting of chemo-
kines IL8, MIP and MCP, has significant negative associa-
tions to hand ROA score (beta ¼?0.14 P ¼ 0.039) and to
disc degeneration ROA score (beta ¼ ?0.22 P ¼ 0.005),
independent of age, sex and BMI. This implies that subjects
with high chemokine levels have lower hand ROA and DD
scores. In addition, component 2 showed a similar associa-
tion to subjects that had hand OA according to the ACR
criteria (beta¼ ?0.26; P ¼0.024, data not shown). When
Table I
Study characteristics of the 382 patients with OA at multiple joint
sites (GARP study)
N (%)Mean (sd)Range
Women
Age (years)
BMI (kgm?2)
S-HsCRP (mgl?1)
Clinical hand OA
PCS SF36
312 (81.7)
382 (100)
379 (99.2)
354 (92.7)
271 (70.7)
375 (98.2)
ee
60.27 (7.54)
27.00 (4.67)
3.63 (5.43)
e
54.02 (21.10)
42.66e79.44
19.10e46.48
0.21e56.80
e
8.75e98.75
ROA score*012
Knee
Hip
DD
Hand
232
275
125
169
90
56
181
110
60
51
76
103
PCS physical component score.
*All subjects were affected by OA at multiple joint sites. The
scores 0, 1 and 2 represent a proportionate OA score, as described
earlier24. In genetic analysis hip and knee replacements (respec-
tively 38 and 8) were considered as OA, numbers indicated are
patients with diseased joints at sampling, or had replacements
within the year prior to sampling.
Table II
Cytokines and chemokines in the 17-plex bead assay with levels
above background noise per cytokine and chemokine
Cytokine/ChemokineN Detectable (%)Analysis
Interleukin 1b
Interleukin 4
Interleukin 12
Interleukin 13
Interleukin 17
Interferon g
TNFa
Granulocyte Monocyte CSF
Interleukin 2
Interleukin 5
Interleukin 6
Interleukin 7
Interleukin 8
Interleukin 10
GCSF*
MCP-1
107
26
90
172
25
179
165
144
234
222
374
274
359
327
248
377
28.3
6.9
23.8
45.5
7.1
47.4
43.4
37.8
61.6
58.4
98.4
72.1
94.5
86.1
65.3
98.9
QL
QL
QL
QL
QL
QL
QL
QL
QT
QT
QT
QT
QT
QT
QT
QT
MIP-1b
375 98.4QT
Table III
Individual factor cytokine scores extracted by PCA of 10 inflamma-
tory markers measured in blood
Component*123
IL2
IL6
GCSF
IL10
IL7
MCP-1
IL8
MIP-1b
S-HsCRP
IL5
0.839
0.815
0.764
0.594
0.561 0.474
0.799
0.799
0.765
?0.747
0.585
Total variation explained 34.1%15.7% 11.6%
*Extraction Method: PCA in which missing values were replaced
by mean levels.
Table IV
Effect sizes (b) of the linear relationships between the extracted
principal components (1e3) reflecting variation at plasma chemo-
kines/cytokine levels and OA characteristics (ROA score, clinical
symptoms) of the subjects of the GARP study sample
OA characteristicsComponentsy
213
Radiographic OA
Hip ROA score
Knee ROA score
Hand ROA score
DD ROA score
0.079
?0.079
0.030
0.002
0.098
0.043
?0.140*
?0.223*
0.035
?0.088
?0.042
?0.060
SF36 outcome
PCS 0.001?0.006*0.007**
* P <0.05 ** P <0.01.
yComponent 1 contains IL2, GCSF, IL6, IL10 and IL7, Compo-
nent 2 contains MCP, IL8, MIP and component 3 contains S-
HsCRP (negative), IL5 and IL7. Data was analyzed using mixed
model regression analyses with the components as dependent vari-
able and as co-variables the joint specific ROA scores (top) or the
PCS (bottom) in addition to age, sex and BMI.
623
Osteoarthritis and Cartilage Vol. 17, No. 5

Page 4

analyzing the relationship of the components to the PCS
derived from the SF36 a significant negative association
of component 2 (P ¼0.035) and a positive association to
component 3 (P ¼0.004) was observed, independent of
sex, age and BMI (Table IV, lower section). This indicates
that subjects with high chemokine levels experience more
functional impairment whereas subjects with high IL7, IL5
and low S-HsCRP levels experience less functional impair-
ment. The association of PCS to component 2, however,
appeared not independent of the hand ROA scores (data
not shown). In each of the mixed models significant associ-
ations with BMI were observed for component 1 (positive,
P ¼ 0.049), for component 2 (negative, P < 1 ?10?5) and
component 3 (negative, P <1 ? 10?5). In the qualitatively
analysis of the cytokines no significant associations were
observed for OA subtypes (data not shown).
SELS GENE VARIATION, INFLAMMATORY
PARAMETERS AND OA
In the GARP study we could not confirm the previously
reported association of rs28665122 with TNFa, IL1-b and
IL6 plasma levels by Curran et al.13.
The influence of the SELS SNPs on the cytokine and/or
chemokine levels, as expressed by the three components,
was investigated by haplotype analysis since high linkage
disequilibrium was observed between the three SELS
SNPs (D0> 0.8). As shown in Table V, four common haplo-
types with frequencies over one percent were observed
similar to a Finnish population15. A significant association
was observed between haplotype GAG (frequency 3.5%)
and component 1 (P ¼ 0.019). Since component 1 reflects
variation in IL2, IL6, GCSF and IL10 levels this association
indicates that carriers of this haplotype have higher levels of
the cytokines in this component. Upon further investigation,
this association appeared to be mainly driven by IL10 vari-
ation (univariate analysis P ¼ 0.001). In addition, the GAG
haplotype shows association to component 3 (P ¼0.036)
containing IL5, IL7 and inversely S-HsCRP. This associa-
tion appeared to be mainly attributable to S-HsCRP levels
in the component (univariate analysis P ¼ 0.002). These as-
sociations were independent of BMI, age and sex. The fact
that both components associate to the GAG haplotype indi-
cates some interrelation between the haplotype and these
components. We were not able to asses association be-
tween SELS SNPs or haplotypes and OA subtypes.
Discussion and conclusion
In inflammation driven diseases high circulating plasma
levels of pro-inflammatory cytokines and S-HsCRP are
present well above detection limits of current methods
and readily used for diagnostic and prognostic purposes28.
Cytokines, however, are known to exert their wide ranged
actions also in very low concentrations. The recent introduc-
tion of multiplexed cytokine assays facilitates simultaneous
measurements of multiple cytokines. However, the de-
scribed absence of a large scale upregulation or strong as-
sociation of any of the measured cytokines or chemokines
in the plasma of subjects with familial OA at multiple joint
sites indicates that, in blood plasma, these markers are
not sufficiently suitable to monitor the ongoing OA process.
Synovial fluid measurements might better reflect the ongo-
ing disease process since it better reflects the cytokine ac-
tivities near the site where the disease is mainly active29.
PCA analyses of cytokine/chemokines measurements re-
vealed 3 components. component 1 reflects variation at
IL2, IL6, GCSF and IL10 levels, component 2 reflects vari-
ation at chemokines MCP, IL8 and MIP and component 3 S-
HsCRP (negative), IL5 and IL7. The components seem to
reflect different ongoing (patho) physiological processes
identified by subjects underlying the components. Compo-
nent 1 may be classified as a marker of ongoing anti-inflam-
matory signaling based on the strong involvement of IL10,
whereas component 2 shows chemokine signaling and
component 3 reflects more acute phase related signaling.
The observed (Supplementary Table 1) and known correla-
tion of IL6 to S-HsCRP is, with the current setting in the
PCA (Eigen values >1), not reflected in the components
since together they do not explain sufficient amount of the
variation.
We found a significant negative association for compo-
nent 2 to hand OA and DD as well as to the SF36 derived
PCS. This indicated that especially the subjects that exhibit
high functional impairment and have low hand ROA scores
have high chemokine levels. It should be noted that, by def-
inition of the GARP selection criteria, subjects of the GARP
study with low hand or disc ROA scores have OA at other
joint sites. We could not, however, attribute the negative as-
sociation of hand and disc ROA to positive effects caused
by these other joint sites. Given these results it appears
the higher levels of chemokines act protective in hand
and disc OA among subjects of the GARP study. The strong
association of the PCS to component 3 including S-HsCRP
Table V
Haplotype frequencies within the GARP study with their mean haplotypic contribution to the component scores as extracted from the data of 9
cytokines/chemokines and CRP
HaplotypeNumber (%) C1*y mean (95% C.I.)
?0.25 (?0.61e0.11)
?0.22 (?0.65e0.20)
0.68 (0.24e1.10)z
?0.68 (?2.57e1.21)
e
C2*y mean (95% C.I.)
?0.27 (?0.58e?0.05)
?0.39 (?0.79e0.00)
?0.23 (?1.06e0.61)
?0.37 (?1.38e0.64)
e
C3*y mean (95% C.I.)
0.85 (0.56e1.15)
0.83 (0.48e1.18)
1.98 (1.35e2.61)*
0.92 (0.29e1.54)
e
GGA
AAG
GAG
GAA
Other
638 (83.5)
81 (10.6)
27 (3.5)
10 (1.3)
8 (1.1)
Total764 (100)
*Scores displayed are the expected haplotypic contribution (independent of the BMI effects) to the mean ‘‘principle component regression
score’’ of subjects calculated by the Thesias program. In individuals the expected level is determined by the contribution of the 2 carried hap-
lotypes. The THESIAS program does not allow correction for familial relationship. Alleles are in the following order of SNPs in the SELS gene
?105 (rs28665122G>A), þ3705 (rs4965814G>A), and þ6218 (rs9874A>G).
yComponent 1 reflects variation of IL2, IL6, GCSF, IL10 and IL7 levels, component 2 reflects variation of chemokines MCP, IL8 and MIP
and component 3 reflects variation of HsCRP (negative), IL5 and IL7.
zBMI adjusted P-value <0.05 for contribution of the haplotype GAG to the component score as compared to the other haplotypes as
determined by THESIAS. C1 component 1, C2 component 2, C3 component 3.
624
S. D. Bos et al.: Cytokine, CRP & SELS gene variation in OA

Page 5

might reflect impairment of physical functioning mediated or
reflected by the individual markers in this component, the
observed effects are independent of BMI and ROA status.
Due to the relative large amount of missing values in the
IL5, IL7 and GCSF data we may have missed specific asso-
ciations with these markers.
Previously, it has been shown that there is a major upre-
gulation of chemokines in human OA affected cartilage12.
Our analysis show that, amongst subjects of the GARP
study, this upregulation might be less pronounced in sub-
jects with hand OA and disc degeneration of the spine as
compared to the other subjects, when correcting for all
involved joint sites. This may reflect a different pathophysi-
ological process underlying hand OA and disc degeneration
as compared to knee and hip OA. This needs to be further
explored in other cohorts of OA patients and especially us-
ing control samples.
Given the earlier found associations of SELS SNPs to in-
flammatory factors measured in blood13, we expected to
find associations of cytokines and chemokines especially
to rs28665122, however, no direct associations of these
levels or presence of these cytokines and chemokines
were observed for variation at the SELS gene. A recent pa-
per of Seiderer et al.14also showed no confirmation of the
association for rs28665122 to cytokine levels in a study in-
cluding patients with intestinal inflammation. As compared
to the study in intestinal inflammation patients, however,
Curran et al. used a more sensitive method of measuring
cytokines. Seiderer et al. did observe an association of
the pro-inflammatory allele to higher serum CRP levels in
a subgroup of Crohn disease patients with higher signs of
disease activity14. In our study, possibly the association to
either cytokines or S-HsCRP for rs28665122 is not ob-
served due to a smaller sample size or upregulation of
these cytokines by the ongoing disease processes is not
sufficient to show the genotype effect as observed by
Seiderer et al.. However, haplotype association analyses
revealed a specific SELS haplotype (GAG, 3.5%) signifi-
cantly associated to components of increased IL10 blood
levels and decreased S-HsCRP levels, confirming that the
SELS gene variation may interfere with or affect the homeo-
stasis of the inflammatory pathways.
In the PCA individuals with missing values in five or more
out of nine cytokine/chemokines measurements were ex-
cluded (N ¼13). For the analyzed individuals remaining
missing values were given the specific marker mean score.
Although we could not readily detect the cause of values
being missing most likely these occurred due to bad sample
qualityorassayerrorsandlesslikelyduetoindividualsbeing
out ofrange.Performing PCAusinga listwise caseselection
(using only cases with all markers available), replacing miss-
ing values by the lowest observed value, imputation of the
missing values by use of regression analysis or by use of
multiple imputation using a winMICE implemented expecta-
tion-maximization (EM) algorithm30did not affect formation
of components or subsequent associations (data not
shown). The components reported should be considered
robust. The effect of the familial dependencies on the com-
ponent formation is considered to be minimal as the total
number of pairs in the dataset cancels out possible intra sib-
ling pair correlations. This is strengthened by the results of
an analysis of the data using only one member per sibling
pair (data not shown) which shows highly similar component
and scores.
Moreover, by excluding individuals with over four out of
nine missing values we may have excluded individuals
with a particular low-inflammatory profile. Finally, missing
values in the qualitatively analyzed cytokines may likewise
have been subject to possible misclassification of individ-
uals due to bad sample quality, assay errors or measure-
ment problems.
Since OA patients are likely to use drugs which alter the
immune system we explored whether the use of NSAIDs
significantly influenced our results. In a split analysis for
use vs no use of NSAIDs the formation of the components
in the PCA showed no major changes, except for the third
component in NSAID users where IL5 and IL7 disappeared.
The described subsequent associations remained present
in both separate datasets (data not shown).
In earlier investigations of the GARP study, we have
found that S-HsCRP levels are independently associated
to a CRP haplotype7,13and by the occurrence of the closely
related factors knee ROA, BMI and high Western Ontario
and McMaster Universities Osteoarthritis Index (WOMAC)
scores24. In the current study we show that a SELS haplo-
type is additionally influencing the S-HsCRP level. When
we fitted a mixed model with S-HsCRP as dependent vari-
able and as co-variables the CRP haplotype H7/8, SELS
haplotype GAG, and the factor representing knee ROA,
BMI and WOMAC24, it was shown that all three consistently
and independently influence the S-HsCRP level. In this
model, the SELS haplotype (GAG) significantly decreased
(beta¼ ?0.42; P ¼ 1.9 ?10?5) S-HsCRP levels in carriers
whereas the S-HsCRP haplotype
P ¼0.08)
P ¼1.6 ? 10?9) increased S-HsCRP level.
The strength of the GARP study lies in the availability of
extended clinical and radiological data of OA features for
four joint sites. Furthermore, demographic data for the par-
ticipants is available, as well as additional familial informa-
tion, a range of biological fluids and DNA. A downside for
this study is, however, that the sample size is relatively
small for genetic studies and no synovial fluid samples
and control samples of healthy individuals are available
for the cytokine measurements in this study.
The analysis of a range of cytokines as measured in
blood has not shown any strong associations of one single
cytokine to OA features; however, a component comprised
of several chemokines did show association to OA in
smaller joints and PCS. Another component which has
strong involvement of S-HsCRP shows a highly significant
association to this PCS, indicating physical impairment
might be reflected or mediated by the markers in this com-
ponent, independent of presence of specific ROA subtypes.
Future measurements of chemokines and related signaling
proteins in synovial fluid may shed more light on the origin
of this association. Long term effects of lower circulating S-
HsCRP, cytokine or chemokine levels might be reflected by
OA progression. Upcoming follow up data in the GARP
study will reveal whether these inflammatory parameters
associate to further active progression of OA.
H7/8(beta ¼0.14;
(beta ¼0.14;and thePCA component
Conflict of interest
The authors declare no conflict of interest.
Acknowledgments
We thank all participants of the GARP study. For the GARP
study, the Dutch Arthritis Association, the Netherlands Orga-
nization for Scientific Research and Pfizer Inc., Groton, CT,
USA,providedgenerous support.In addition,we
625
Osteoarthritis and Cartilage Vol. 17, No. 5

Page 6

acknowledge the support of the cooperating hospitals and
referring rheumatologists, orthopedic surgeons and general
practitioners. The work described in this paper was sup-
ported by the Netherlands Organization of Scientific Re-
search (MW 904-61-095, 911-03-016, 917 66344 and 911-
03-012), Leiden University Medical Centre, the Centre for
Medical Systems Biology (CMSB) in the framework of the
Netherlands Genomics Initiative(NGI). Furthermore, we
thankDennisKremerforworkandsupporttothegenotyping.
Supplementary material
Supplementary data associated with this article can be
found,intheonlineversion,atdoi:10.1016/j.joca.2008.10.007.
References
1. IkegawaS.Newgeneassociationsinosteoarthritis:whatdotheyprovide,
and where are we going? Curr Opin Rheumatol 2007;19(5):429e34.
2. Valdes AM, Loughlin J, Oene MV, Chapman K, Surdulescu GL,
Doherty M, et al. Sex and ethnic differences in the association of
ASPN, CALM1, COL2A1, COMP, and FRZB with genetic susceptibil-
ity to osteoarthritis of the knee. Arthritis Rheum 2007;56(1):137e46.
3. Botha-Scheepers SA, Watt I, Slagboom E, de Craen AJ, Meulenbelt I,
Rosendaal FR, et al. Innate production of Tumor Necrosis Factor-
{alpha} and Interleukin-10 is associated with radiological progression
of knee osteoarthritis. Ann Rheum Dis 2007.
4. Pelletier JP, Martel-Pelletier J, Abramson SB. Osteoarthritis, an inflam-
matory disease: potential implication for the selection of new thera-
peutic targets. Arthritis Rheum 2001;44(6):1237e47.
5. Riyazi N, Slagboom E, de Craen AJ, Meulenbelt I, Houwing-
Duistermaat JJ, Kroon HM, et al. Association of the risk of osteoarthri-
tis with high innate production of interleukin-1beta and low innate pro-
duction of interleukin-10 ex vivo, upon lipopolysaccharide stimulation.
Arthritis Rheum 2005;52(5):1443e50.
6. Spector TD, MacGregor AJ. Risk factors for osteoarthritis: genetics(1).
Osteoarthritis Cartilage 2004;12(Suppl A):39e44.
7. Bos SD, Suchiman HE, Kloppenburg M, Houwing-Duistermaat JJ, Hellio
Le Graverand MP, Seymour AB, et al. Allelic variation at the C-reac-
tive protein gene associates to both hand osteoarthritis severity and
serum high sensitive CRP levels in the GARP study. Ann Rheum
Dis 2007;67(6):877e9.
8. Goldring MB. Update on the biology of the chondrocyte and new
approaches to treating cartilage diseases. Best Pract Res Clin
Rheumatol 2006;20(5):1003e25.
9. Goldring MB, Goldring SR. Osteoarthritis. J Cell Physiol 2007;213(3):
626e34.
10. Appleton CT, Pitelka V, Henry J, Beier F. Global analyses of gene ex-
pression in early experimental osteoarthritis. Arthritis Rheum 2007;
56(6):1854e68.
11. Borzi RM, Mazzetti I, Marcu KB, Facchini A. Chemokines in cartilage
degradation. Clin Orthop Relat Res 2004;(Suppl 427):S53e61.
12. Attur MG, Dave M, Akamatsu M, Katoh M, Amin AR. Osteoarthritis or
osteoarthrosis: the definition of inflammation becomes a semantic
issue in the genomic era of molecular medicine. Osteoarthritis
Cartilage 2002;10(1):1e4.
13. Curran JE, Jowett JB, Elliott KS, Gao Y, Gluschenko K, Wang J, et al.
Genetic variationin selenoprotein
response. Nat Genet 2005;37(11):1234e41.
14. Seiderer J, Dambacher J, Kuhnlein B, Pfennig S, Konrad A, Torok HP,
et al. The role of the selenoprotein S (SELS) gene -105G > A pro-
moter polymorphism in inflammatory bowel disease and regulation
of SELS gene expression in intestinal inflammation. Tissue Antigens
2007;70(3):238e46.
15. Alanne M, Kristiansson K, Auro K, Silander K, Kuulasmaa K, Peltonen L,
et al. Variation in the selenoprotein S gene locus is associated with
coronary heart disease and ischemic stroke in two independent Finn-
ish cohorts. Hum Genet 2007;122(3e4):355e65.
16. Riyazi N, Meulenbelt I, Kroon HM, Ronday KH, Hellio Le Graverand MP,
Rosendaal FR, et al. Evidence for familial aggregation of hand, hip,
and spine but not knee osteoarthritis in siblings with multiple joint in-
volvement: the GARP study. Ann Rheum Dis 2005;64(3):438e43.
17. Altman R, Alarcon G, Appelrouth D, Bloch D, Borenstein D, Brandt K,
et al. The American College of Rheumatology criteria for the classifi-
cation and reporting of osteoarthritis of the hand. Arthritis Rheum
1990;33(11):1601e10.
18. Altman R, Alarcon G, Appelrouth D, Bloch D, Borenstein D, Brandt K,
et al. The American College of Rheumatology criteria for the classifi-
cation and reporting of osteoarthritis of the hip. Arthritis Rheum 1991;
34(5):505e14.
19. Altman RD, Asch E, Bloch D. Development of criteria for the classifica-
tion and reporting of osteoarthritis: Classification of osteoarthritis of
the knee. Arthritis Rheum 1986;29:1039e49.
20. Kellgren JH, Lawrence JS. Radiological assessment of osteo-arthrosis.
Ann Rheum Dis 1957;16(4):494e502.
21. Hays RD, Sherbourne CD, Mazel RM. The RAND 36-Item Health
Survey 1.0. Health Econ 1993;2(3):217e27.
22. Ware JE, Kosinski M. Interpreting SF-36 summary health measures:
a response. Qual Life Res 2001;10(5):405e13.
23. Meulenbelt I, Kloppenburg M, Kroon HM, Houwing-Duistermaat JJ,
Garnero P, Hellio Le Graverand MP, et al. Urinary CTX-II levels are
associated with radiographic subtypes of osteoarthritis in hip, knee,
hand, and facet joints in subject with familial osteoarthritis at multiple
sites: the GARP study. Ann Rheum Dis 2006;65(3):360e5.
24. Meulenbelt I, Kloppenburg M, Kroon HM, Houwing-Duistermaat JJ,
Garnero P, Hellio-Le Graverand MP, et al. Clusters of biochemical
markers are associated with radiographic subtypes of osteoarthritis
(OA) in subject with familial OA at multiple sites. The GARP study.
Osteoarthritis Cartilage 2007;15(4):379e85.
25. Min JL, Meulenbelt I, Riyazi N, Kloppenburg M, Houwing-Duistermaat JJ,
Seymour AB, et al. Association of matrilin-3 polymorphisms with spinal
discdegenerationandosteoarthritis of the firstcarpometacarpal joint of
the hand. Ann Rheum Dis 2006;65(8):1060e6.
26. Stevens J. Applied multivariate statistics for the social sciences. Hill-
sdale (NJ): Lawrence Erlbaum Associates; 1986.
27. Tregouet DA, Escolano S, Tiret L, Mallet A, Golmard JL. A new
algorithm for haplotype-based association analysis: the Stochastic-
EM algorithm. Ann Hum Genet 2004;68(Pt 2):165e77.
28. McInnes IB, Schett G. Cytokines in the pathogenesis of rheumatoid
arthritis. Nat Rev Immunol 2007;7(6):429e42.
29. Hulejova H, Baresova V, Klezl Z, Polanska M, Adam M, Senolt L. In-
creased level of cytokines and matrix metalloproteinases in osteoar-
thritic subchondral bone. Cytokine 2007;38(3):151e6.
30. Van Buuren S, Oudshoorn CGM. Multivariate Imputation by Chained
Equations: MICE V1.0 User’s manual. TNO Report PG/VGZ/00.038.
Leiden: TNO Preventie en Gezondheid. <http://web.inter.nl.net/
users/S.van.Buuren/publications/MICE
2000.pdf>; 2007.
Sinfluences inflammatory
V1.0Manual TNO00038
626
S. D. Bos et al.: Cytokine, CRP & SELS gene variation in OA