Citation
Kamphorst, J. J. (2010, March 25). Systems biology of osteoarthritis.
Retrieved from https://hdl.handle.net/1887/15125
Version: Corrected Publisher’s Version
License: Licence agreement concerning inclusion of doctoral thesis in the Institutional Repository of the University of Leiden
Downloaded from: https://hdl.handle.net/1887/15125
Note: To cite this publication please use the final published version (if applicable).
107
6
Lipidomics reveals local and systemic changes in lipid homeostasis during osteoarthritis
Abstract
Objective: While the main characteristic of osteoarthritis (OA) is progressive loss of articular cartilage, recent epidemiological studies suggest a biochemical link between obesity and OA. This biochemical link likely involves lipid regulation, but the details remain elusive. We aim to contribute to the understanding of the role of lipid (dys)regulation in OA, by comprehensively analyzing multiple lipids and lipid classes in plasma and synovial fluid (SF).
Methods: the extracted lipid content of 59 plasma samples with varying grades of OA and 29 synovial fluid samples from controls and late stage OA patients were analyzed with liquid chromatography and mass spectrometry. Using a target list of identified lipids, peak areas were quantified, and multivariate analysis was performed to discern differences in lipid content between OA and control samples.
Results: In total 109 lipids in plasma and 72 lipids in SF from 8 lipid classes were quantified. In OA plasma samples, an overall lowering of lipid levels was observed compared to control samples. This effect was most pronounced for lipids with shorter fatty acid chains and correlated positively with disease severity. For SF, the most pronounced difference between control and OA was the relative abundance of the lipid classes. Cross-compartment analysis revealed that OA plasma and SF are more similar in lipid content than control plasma and SF, which may indicate decreased synovial membrane barrier function.
Conclusion: This study shows that both local and systemic changes in lipid homeostasis/metabolism occur in OA. Further studies need to establish the biological relevance of these differences.
Based on: Kamphorst J. J.; DeGroot J.; van der Heijden R.; Reijmers T. H.;
Lafeber F. P. J. G.; Paliukhovich I.; Troost J.; Vreeken R.; van der Greef J.;
Hankemeier T. Lipidomics reveals local and systemic changes in lipid homeostasis during osteoarthritis. Submitted to Osteoarthritis & Cartilage.
108 1. Introduction
Osteoarthritis (OA) is one of the most prevalent rheumatic diseases affecting more than half of the population above sixty years of age (1). The main characteristic of the disease is progressive loss of articular cartilage which is thought to originate from an imbalance between synthesis and degradation of the cartilage matrix (2;3). However, its precise aetiology is still far from understood, resulting in the lack of adequate biomarkers for early diagnosis, prediction of eventual joint damage, assessment of disease activity and monitoring of the efficacy of therapies (4;5). Consequently no generally accepted disease-modifying treatments exist: the available therapies primarily comprise analgesics. Recent studies suggest that OA is a disease resulting from multiple pathophysiological mechanisms in which local and systemic factors, as well as biomechanical triggers are interplaying (1).
One pathological mechanism that is increasingly receiving attention is the deregulation of lipid homeostasis. In epidemiological studies, obesity coincided with increased incidence of non load-bearing types of OA, such as hand OA (6;7), suggesting a systemic, biochemical link. Indeed, various types of OA associated changes in lipid species have been observed. For example, gene expression analysis of cartilage from an OA mouse model revealed a marked down regulation of gene products involved in lipid homeostasis. Additionally, multiple studies have demonstrated increased lipid peroxidation, as evidenced by the elevated concentration of the peroxidation product 4-hydroxynonenal in OA specimens (8;9). In human articular cartilage the total fatty acid content was increased at sites of cartilage lesions (10). Our recent systems biology driven meta-analysis of transcriptomics, proteomics, and metabolomics profiling methods revealed that lipid metabolism is amongst the most affected processes in OA pathology (11).
The precise mechanisms through which lipids are involved in OA pathology are yet to be uncovered and our knowledge remains fragmented and incomplete to date. Enlarging the 'toolbox' of research methods with approaches that simultaneously analyze multiple lipid species within a class, and even multiple classes of lipids, could speed up research into the
109 involvement of lipids in OA. Due to recent innovations in analytical chemistry in general, and in the field of 'lipidomics' in particular (12-14), such new analytical methods slowly become available. These approaches typically involve sample treatment by extraction of the lipids of interest, separation of the lipid classes and individual lipids by liquid chromatography and detection by mass spectrometry.
In one such method plasma samples were extracted using a modified version of Bligh and Dyer and the extracted lipids were subsequently separated using a C8 reversed phase column and analyzed with high accuracy mass spectrometry (15). This method allowed the global analysis of various lipids from the monoacyl-glycerophosphocholine, glycerophosphocholine,
sphingomyeline, monoacyl-glycerophosphoethanolamine, glycerophosphoethanolamine, cholesterol ester, diglyceride, and triglyceride
lipid classes. Collectively, these lipids exert biological functions that include membrane formation and regulation, fuel storage, bioactive signaling and substrate transport (13).
In this paper we present the results from the global lipid analysis of plasma and synovial fluid (SF) samples from OA and control donors, by means of a method based on Bligh and Dyer lipid extraction, C8 reversed phase chromatography, and mass spectrometry analysis. The goal of this study was to contribute to the elucidation of the occurrence and role of lipid (dys)regulation in OA, so as to find potential biomarkers and hints at potential novel mechanisms of disease.
2. Methods and materials 2.1 Chemicals
Acetonitrile (ACN, HPLC grade), dichloromethane (DCM, HPLC grade), isopropanol (IPA, HPLC grade), and methanol (MeOH, HPLC grade) were purchased from Biosolve B.V. (Valkenswaard, The Netherlands). Ultrapure water (5.5 µS/m) was obtained from a Milli-Q gradient A10 system (Millipore, Bedford, Massachusetts). Ammonium formate (AmFm, >99%) was obtained from Sigma (St. Louis, Missouri). Lipid standards 1-
110 heptadecanoyl-2-hydroxy-sn-glycero-3-phosphocholine GPCho (17:0/0:0), 2-diheptadecanoyl-sn-glycero-3-phosphoethanolamine GPEtn (17:0/17:0), and 1,2-diheptadecanoyl-sn-glycero-3-phosphocholine GPCho (17:0/17:0) were acquired from Avanti Polar Lipids (Alabaster, AL), and 2,3- triheptadecanoylglycerol TG (17:0/17:0/17:0) from Sigma (St. Louis, Missouri).
2.2 Lipid nomenclature
For the lipid nomenclature the guidelines of lipidmaps (http://www.lipidmaps.org) were adhered to throughout this article. For example, lyso-glycerophosphocholine with a 17:0 fatty acid chain was named as monoacyl-glycerophosphocholine GPCho (17:0/0:0) or lyso- GPCho (17:0) and 1,2-dinonadecanoyl-sn-glycero-3-phosphocholine was named as GPCho (19:0/19:0). In case the individual fatty acid chains were not determined, the sum of the carbons and double bonds of all fatty acid chains is given, as for example, GPCho (38:0).
2.3 OA plasma and SF samples
In this study, heparinized plasma samples from 59 female subjects were analyzed. These samples were collected at three hospitals in the Netherlands as part of the CHeCK study (16). Twenty-six of the 59 donors had no reported (by radiological assessment) arthritic conditions at the time of sample acquisition, whereas 6 of those 26 showed early signs of OA at a follow-up two years later. For the remaining 32 donors disease severity was classified based on radiologic features of osteoarthritis (OA) in knee and hip joints (Kellgren-Lawrence (K-L) grading) (17). The K-L grade (0-4) was determined for each joint and a summed OA load was calculated for each subject by summing the K-L grade of the individual joints, resulting in a theoretical range from 0 (no OA in knee or hips) to 16 (severe OA in all joints). Since the CHeCK cohort comprised subjects with mild OA, the actual range in the current 59 subjects was 0-8. All donors had similar body mass indexes (BMI) and ages: average BMI control group 28.3 (S.D. 6.2), early OA group 24.7 (S.D. 4.4) and definite OA group 26.57 (S.D. 3.4), and
111 average age control group 52.9 (S.D. 4.1), early OA group 56.8 (S.D. 3.3) and definite OA group 56.2 (S.D. 5.7).
The synovial fluid (SF) samples (from both male and female donors) were obtained from the knees of patients with late osteoarthritis (OA, during surgery) and from individuals without a joint disease (post mortem), and thus were not from the same donors as the plasma samples. In total 29 SF samples were analyzed of which 13 were from control donors and 16 from diagnosed (late) OA patients. Both plasma and SF samples were stored at - 80 °C. All procedures were conducted according to local ethical standards.
2.4 Sample preparation
The plasma samples were pretreated in duplicate following a protocol based on the Bligh and Dyer method (15;18). Briefly, 60 μl of an internal standard solution with 12.5 μg/ml GPCho (17:0/0:0), 25 μg/ml GPEtn (17:0/17:0), 25 μg/ml GPCho (17:0/17:0), and 12.5 μg/ml TG (17:0/17:0/17:0) was added to 30 μl of plasma. This was followed by the addition of 200 μl MeOH, rigorous vortexing for 30 s, addition of 400 μl of DCM, again vortexing for 30 s, addition of 125 μl H2O, and vortexing for 3 s. Then the samples were allowed to rest at room temperature for 10 min and subsequently the samples were centrifuged for 10 minutes at 8000 rpm at 10 °C. The final steps in the sample preparation procedure comprised transferring the lower, organic phase containing the lipid extract into a glass sample vial, and diluting this extract five times in a mixture of ACN, IPA, and H2O (65%:30%:5%, v/v/v).
The samples are then stored at -20 °C until analysis.
2.5 Analysis
For the lipid profiling, liquid chromatography was coupled to mass spectrometry (LC-MS). The plasma and SF samples were analyzed in separate batches. The separation was performed with an Acquity UPLC (Waters, Milford, Massachusetts) pump and detection with a 6530 LC-MS QTOF mass spectrometer (Agilent, Santa Clara, California). The prepared samples were analyzed in random order with every 10th analysis being a quality control (QC) sample. This QC was created by pooling aliquots of
112 plasma from the other samples and was used to check the quality of the measurement series. Additionally, the quality of analysis (linearity) was checked by analyzing a calibration curve, which was obtained by spiking 8 levels of GPCho(19:0/0:0), GPEtn(15:0/15:0), GPCho(19:0/19:0), and TG(15:0/15:0/15:0) to the plasma or SF samples. Every sample was analyzed by injecting 3 μl onto an Ascentis Express column (2.7 μm C8
particles; 2.1 x 150 mm; Sigma, St. Louis, Missouri) operated at a temperature of 55 °C and a flow rate of 0.5 ml/min. A binary gradient was used, with solution A H2O-ACN (40%:60%, v/v) solution containing 10 mM ammonium formate, and solution B an ACN-IPA (10%:90%, v/v) solution containing 10 mM ammonium formate. The gradient ran from 68% A for the first minute, to 50% A at 3 min, 35% A at 9 min, 20% A at 14 min, then 3%
A at 14.1 min maintained until 17 min, and finally 68% A at 17.10 min maintained to 20 min. Mass spectrometry was performed in full scan mode with the following settings: 2 GHz dynamic range mode in low mass region (100-1700 m/z); ion-source, Jet Stream ESI interface; scan mode, positive ion mode; fragmentor, 215; skimmer 1, 70; octopoleRFPeak, 750; Vcap, 3500 V; nozzle voltage, 1000 V; gas temp, 325 °C; gas flow, 10 l/min;
sheath gas, 400 °C; sheath flow gas, 12 l/min; nebulizer, 2.4 bar; stoptime, 20 min; mass range, 450 – 1600 m/z.
2.6 Data processing
The peak areas of the extracted ion chromatograms of 109 individual lipids for plasma and 72 for SF were integrated with the software program Masshunter Quantitative Analysis (QTOF) v B.01.04 (Agilent, Santa Clara, California), following a target list of previously characterized lipids. For this program the settings were as follows: mass tolerance, +/- 20 ppm; retention time tolerance, +/- 1 minute; relative response, area; noise algorithm, RMS;
smoothing function width, 15; smoothing Gaussian width, 5. The obtained peak areas were manually verified and corrected for analytical variation with an internal standard from the same class or with similar retention time:
(GPCho (17:0/0:0) for the monoacyl-glycerophosphocholines (lyso-GPChos), GPEtn (17:0/17:0) for the glycerophosphoethanolamines (GPEtns) and monoacyl-glycerophosphoethanolamines (lyso-GPEtns), GPCho (17:0/17:0) for the glycerophosphocholines (GPChos), sphingomyelines (SMs) and
113 diglycerides (DGs), and TG (17:0/17:0/17:0) for the triglycerides (TGs) and cholesterol esters (ChoEs). Data analysis and visualization were performed with MATLAB R2006a (The MathWorks, Natick, Massachusetts) and Office Excel 2003 (Microsoft, Redmond, Washington).
3. Results
ANALYTICAL PERFORMANCE AND SAMPLE GROUPING
From a target list of previously characterized lipids, containing the masses and retention times of 136 identified lipids from 8 lipid classes, in total 109 lipids in plasma and 72 lipids in SF peak areas could be accurately integrated and compared across all samples (see Table 1 for the number of lipids per class). In general, the concentration of the lipids found in SF was lower compared to plasma. Scatter plots indicated that the analytical duplicates were highly similar (data not shown) and the average of the duplicates was used for further data analysis.
As calculated from the quality control samples, the average relative standard deviations for the lipids were found to be 4%, for the plasma as well as the SF analyses, and were considered to be satisfactory. Additionally, the constructed calibration curves of some of the non-endogenous lipids spiked to SF and plasma showed an adequate linear response of the analytical system for both matrices, i.e. the calibration curves had an R2 of 0.946 or better.
For the data analysis the plasma samples were grouped into three classes of different OA severity. This was done for visualization purposes and the statistical power is then higher than when using a more continuous scale.
The grouping was done as follows: the control class (with samples from individuals with no radiological OA), the early OA class (with samples from patients with a summed K-L grade of 1 and 2), and the definite OA class (with samples from patients with a summed K-L grade of 3 and higher).
114
Table 1. Number of individual lipids per lipid class of which the peak areas could be integrated and compared across all samples.
LIPID PROFILES DIFFER BETWEEN CONTROL AND OA
As a first step in assessing the changes in lipid profiles that occur in OA in both plasma and SF, principal component analysis (PCA) was performed (see Figure 1). PCA is a multivariate, unsupervised statistical technique that enables the interpretation of potential patterns in complex datasets (19). For the plasma samples, the PCA model already showed some grouping of the three disease classes, although there was considerable overlap (see Figure 1A & B, different principal components from the same model). The control and the definite OA classes were separated most, with the early OA group in the middle. This indicated that the analytical variation was smaller than the biological variation and that the detected biological variation between the three classes was subtle and gradual but definite, as they could already be observed in unsupervised PCA. In addition, these changes could correspond to OA severity.
For performing PCA on the SF samples’ lipid levels, additional normalization of the data was a point of care, as high variation was observed in the total lipid levels (sum of all lipids) between the individual SF samples (this was not observed for the plasma samples). For the PCA model we were interested in the OA-related changes in the relative composition of the lipid profiles rather than the absolute quantities. Therefore, lipid levels relative to the total lipid content (sum of all lipid peak areas) were used to construct the PCA model. Except for two samples, the SF PCA model showed a clear
115 separation of the control and OA samples (see Figure 1C), indicating a considerable difference in the SF lipid profiles of the control and OA samples.
Figure 1. PCA scores plots of the analyzed lipid profiles of the plasma samples (A&B) and of the SF samples (C). The plasma samples are categorized as control (no diagnosed OA), early OA (summed K-L grade of 1-2), or definite OA (summed K-L grade > 2). (A&B) The PCA model already shows some grouping of the three disease classes, although there is considerable overlap. (C) The lipid profiles from the synovial fluid samples show almost complete segregation between the control and OA samples, indicating strong differences in the lipid profiles of both groups.
ASSESSING THE DIFFERENCES BETWEEN OA AND CONTROL LIPID PROFILES FOR PLASMA AND SF
Following the observed differences in the PCA models, the next step was to examine any potential underlying patterns causing these differences. In search of such patterns variations in the individual lipid levels between control and cases in both plasma and SF were examined. In addition the relative contribution of the separate lipid classes (by summing-up the peak areas of all lipids of that class in each sample) to the total lipid composition was assessed. ANOVA analysis was performed to test the significance of the observed differences.
In plasma, although not reaching statistical significance in all cases, the most striking pattern was a general lowering that was observed for all lipids in the definite OA (K-L grade ≥ 3, n=19) samples, relative to the early OA (K-L grade 1 & 2, n=14) and control (K-L grade 0, n=26) samples. Actually, the
116 largest variation occurred in the levels of the lipids with the shorter fatty acid chains (Table 2). This was observed for all major lipid classes. Out of the 109 lipids 10 showed statistically significant (p < 0.01) lower levels in the definite OA samples (Figure 2A and Table 3). Of those, 9 lipids showed decreasing levels that correlated with disease severity: i.e. the lowest levels were found for the definite OA class, intermediate levels in the early OA class samples, and the highest levels in the control samples.
Clinical follow-up studies (2 years after initial sample collection) revealed that six individuals from the original control population (n = 26) had developed early OA (K-L grade 1 or 2). To investigate the prospective predictive power of the 10 statistically significant lipids, their levels in these samples were compared to those in the rest of the control samples. For three GPCho lipids (GPCho(34:1), GPCho(36:1), and GPCho(38:2)), near significant lowering (p-values of 0.057, 0.052, and 0.080, respectively) of their levels in the six samples were observed (see Figure 2B).
117
Table 2. Analysis of the (average) lipid profile differences between control and definite OA plasma samples reveals that especially lipids with short fatty acid carbon lengths (from multiple classes, classes with >10 lipids shown) have lower levels in OA samples than in control samples; the fold change of average of control samples compared to average of definite OA samples per lipid is shown.
118
Figure 2. (A) Ten individual lipids (mostly GPCho) showed significant (p <
0.01) lowering in definite OA (compared to controls), and intermediate levels for the early OA samples. (B) The donors of six control samples showed early OA signs (summed K-L grade 1-2, progressor samples) at a two year follow-up, and comparison of these samples with the control samples suggested that three lipids of the previous ten have the ability to predict the presence of OA prospectively.
Table 3. Individual lipids that have significant (p < 0.01) lower levels in definite OA with respect to the control samples.
119
Figure 3. Analysis of changes in the SF lipid profiles of control and OA samples showed marked differences in total lipid levels (as calculated by summing the peak areas for all analyzed lipids) and class compositions: (A) total levels were significantly higher in OA (p = 0.0002), (B) the total TG levels/total lipid levels ratio lower (p = 0.0019), (C) the total SM levels/total lipid levels ratio significantly higher (p = 0.0001), and (D) the total GPCho levels/total lipid levels ratio also significantly higher in OA (p
= 5.73x10-5).
While in plasma for many lipids lower concentrations were observed in the OA samples, the SF samples showed a different pattern. In the SF compartment the total lipid amount (as calculated by the sum of the all lipids’
peak areas) was significantly lower in the control samples than in the OA samples (Figure 3A). Obviously, the total lipid amount is not exactly proportional to the sum of all lipid peaks, as the response factor per lipid class can be different, but the sum of all lipid areas is a good estimation. In contrast to the plasma samples, there was considerable variation in the lipid class composition rather than in the individual lipids. In the OA samples, the ratio of TGs (sum of TGs’ peak areas) to total lipid peak area was significantly lower (p < 0.01) in the OA samples than in the control samples (see Figure 3B). In addition, the ratio of SMs (sum of SMs’ peak areas) to total lipid peak was significantly higher (p < 0.01) in the cases compared to the control samples (see Figure 3C). This was also true for the GPCho lipid class (see Figure 3D). Consequently, all individual, significantly lower GPCho lipids in OA plasma, were significantly elevated (p < 0.01) in SF, insofar detected (all of the 10, except for GPCho(34:4), GPCho(36:3), and GPCho(38:2)).
120 CROSS-COMPARTMENT COMPARISION OF LIPID CLASS COMPOSITION IN PLASMA AND SF
To gain further insights in the differences in lipid class compositions between SF and plasma, a comparison was mad between each lipid class (as calculated by the sum of all lipids of that class) ratio to the total lipid content (as calculated by the sum of all lipids). This was done for both the control and OA samples.
For the control samples, all class ratios except the SMs and PEs were markedly different for plasma and SF. Significant at p < 0.01 were the glycerophosphocholine / total lipid content, triglyceride/ total lipid content, and monoacyl-glycerophosphocholine/ total lipid content ratios (see Figure 4A). Clearly less variation in the lipid class ratios was observed between the OA plasma and SF. Only significantly differing at p < 0.01 was the monoacyl-glycerophosphocholine/ total lipid ratio (see Figure 4B).
121
Figure 4. Cross-compartmental comparison of lipid profiles of plasma and SF samples reveals that the lipid profiles of (A) control samples of both compartments are more dissimilar than the OA samples (B). Shown are the ratios of lipid classes that are significantly differing (p < 0.01).
4. Discussion
To the best of our knowledge no lipidomics approaches have ever been applied to OA plasma or SF samples. As such, this study presents for the first time a view on OA associated changes in the lipid composition of the monoacyl-glycerophosphocholine, glycerophosphocholine, sphingomyeline,
122 monoacyl-glycerophosphoethanolamine, glycerophosphoethanolamine, cholesterol ester, diglyceride, and triglyceride lipid classes
From the presented results it is clear that changes in the lipid homeostasis are associated with OA, both systemically (plasma) and locally (SF). The most striking observation made from the analysis of the plasma samples is a general lowering of lipids that is proportional to OA disease severity (summed K-L grade). This lowering can be explained by either the increased turnover of the lipids for energy production, an increased degradation through lipid peroxidation, or a combination thereof. Evidence for the involvement of both mechanisms has been provided in the scientific literature, albeit mainly in SF and urine. For example, together with lower levels of glucose and higher levels of lactate (indicating anaerobic metabolism), elevated levels of glycerol and ketone bodies were reported for OA samples, indicating increased degradation of lipids for energy release (20-23). Our finding that mainly the short-chain fatty acid lipids were lower in the OA samples could be due to their better accessibility to digestive enzymes, because of their more favourable lipid-water partitioning coefficients (24). Evidence for increased lipid peroxidation was found in the increased levels of the fatty acid degradation product 4-hydroxynonenal (8;9). This specific degradation is thought to be a direct result of impaired oxidative stress defense, indications for which have been found on the mRNA and protein levels as well (25-27).
The significantly lower total lipid levels in the control samples instead of the OA samples in SF at first glance seem to contradict both the results from the plasma samples and earlier reports. However, it could be possible that through OA-mediated degradation the barrier between the plasma and SF compartments is reduced, enabling a more direct interaction between the compartments. This is certainly suggested by our cross-compartment analysis where the lipid class composition of both OA plasma and OA SF are much more similar than control plasma and control SF. Also the differences in entire lipid classes between the SF control and OA samples can be explained in this respect. Higher GPCho ratios in OA seem to indicate membrane damage, and the higher SM ratios may point towards leakage of lipoproteins from plasma into the SF.
123 As the TG levels are lower in OA, this in turn may indicate that especially the small lipoproteins are transported into SF and/or that there may be a combined effect of degradation of triglycerides for fatty acid and energy production or oxidation. However, more research will be necessary to validate these possible explanations of the results.
While evaluating the significance of the results presented here, some considerations regarding the samples and the analytical method should be taken into account. Firstly, the plasma samples used in this experiment are part of a large sample collection study where, for practical reasons, the samples were collected in a non-fasted state. This likely has introduced diet- related variations that may have obscured subtle, but disease related changes between the samples. However, the disease-related changes that we did observe in the study are likely to be robust as they persisted despite the diet- related additional complexity. Secondly, the plasma and SF samples are not from the same donors, and as the SF samples are from late-stage and in part male OA donors, only a rudimentary comparison with the definite OA plasma samples can be made. Also the fact that the control SF samples were acquired post mortem makes this comparison tentative. Finally, it should be mentioned that the fatty acid constituents of the analyzed lipid classes are mass and geometrical isomers of one another. With the analytical method employed in this study it is impossible to distinguish between these isomers (13). Therefore, it may be possible that for some individual lipids the calculated peak area represents a mixture of lipids that have varying lipid compositions, but the same total number of fatty acid carbons and double bonds. This makes it difficult to study the biochemical events in more detail than was done here.
Overall, the results presented in this study support the accumulating evidence for the involvement of lipids in the OA disease process. While it is still debatable whether the changes in lipid composition are the result from other disease processes or are actually the cause, our results from the prospective analysis provides an indication that they are involved very early in the disease development. As both the underlying biochemistry and the analytical methods for the analyzed lipid classes will improve the coming years, it will be possible to study the OA associated lipid changes more in-
124 depth. Of crucial importance will be the study of the involvement of lipid homeostasis in early OA development, as it will allow for the development of new biomarkers and opportunities for treatment.
125 Acknowledgements
This study was financed by the research programme of the centre for medical systems biology (CMSB), which is part of the Netherlands Genomics Initiative/Netherlands Organization for Scientific Research. In addition we thank Jildau Bouwman and Suzan Wopereis for help with the data analysis and biological interpretation. Adrie Dane is acknowledged for his help with the data analysis.
Reference List
(1) Wieland H. A.; Michaelis M.; Kirschbaum B. J.; Rudolphi K. A. Osteoarthritis - an untreatable disease? Nat. Rev. Drug Discov. 2005, 4, 331-344.
(2) Creamer P. Osteoarthritis. Lancet. 1997, 350, 503-509.
(3) Hunter D. J.; Felson D. T. Osteoarthritis. Br. Med. J. 2006, 332, 639-642.
(4) Rousseau J. C.; Delmas P. D. Biological markers in osteoarthritis. Nat. Clin. Pract.
Rheumatol. 2007, 3, 346-356.
(5) Garnero P.; Delmas P. D. Biomarkers in osteoarthritis. Curr. Opin. Rheumatol. 2003, 15, 641-646.
(6) Felson D. T.; Chaisson C. E. Understanding the relationship between body weight and osteoarthritis. Baillieres Clin. Rheumatol. 1997, 11, 671-681.
(7) Dahaghin S.; Bierma-Zeinstra S. M. A.; Koes B. W.; Hazes J. M. W.; Pols H. A. P.
Do metabolic factors add to the effect of overweight on hand osteoarthritis? The Rotterdam study. Ann. Rheum. Dis. 2007, 66, 916-920.
(8) Morquette B.; Shi Q.; Lavigne P.; Ranger P.; Fernandes J. C.; Benderdour M.
Production of lipid peroxidation products in osteoarthritic tissues. Arthritis Rheum.
2006, 54, 271-281.
(9) Vaillancourt F.; Fahmi H.; Shi Q.; Lavigne P.; Ranger P.; Fernandes J. C.;
Benderdour M. 4-Hydroxynonenal induces apoptosis in human osteoarthritic chondrocytes: the protective role of glutathione-S-transferase. Arthritis Res. Ther.
2008, 10, R107-R118.
(10) Lippiello L.; Walsh T.; Fienhold M. The association of lipid abnormalities with tissue pathology in human osteoarthritic articular cartilage. Metabolism. 1991, 40, 571-576.
(11) Kamphorst J. J.; DeGroot J.; van der Heijden R.; Wopereis S.; Reijmers T. H.; van Ommen B.; van der Greef J.; Hankemeier T. Systems Biology in osteoarthritis:
omics-data integration, biological interpretation and information gap identification.
In preparation. 2009,
(12) Roberts L. D.; McCombie G.; Titman C. M.; Griffin J. L. A matter of fat: An introduction to lipidomic profiling methods. J. Chromatogr. B Analyt. Technol.
Biomed. Life Sci. 2008, 871, 174-181.
(13) Morris M.; Watkins S. M. Focused metabolomic profiling in the drug development process: advances from lipid profiling. Curr. Opin. Chem. Biol. 2005, 9, 407-412.
(14) Hu C.; van der Heijden R.; Wang M.; van der Greef J.; Hankemeier T.; Xu G.
Analytical strategies in lipidomics and applications in disease biomarker discovery. J.
Chromatogr. B Analyt. Technol. Biomed. Life Sci. 2009, 877, 2836-2846.
126
(15) Hu C.; van Dommelen J.; van der Heijden R.; Spijksma G.; Reijmers T. H.; Wang M.; Slee E.; Lu X.; Xu G.; van der Greef J.; Hankemeier T. RPLC-Ion-Trap-FTMS method for lipid profiling of plasma: method validation and application to p53 mutant mouse model. J. Proteome Res. 2008, 7, 4982-4991.
(16) Wesseling J.; Dekker J.; van den Berg W. B.; Bierma-Zeinstra S. M. A.; Boers M.;
Cats H. A.; Deckers P.; Gorter K. J.; Heuts P. H.; Hilberdink W. K.; Kloppenburg M.; Nelissen R. G.; Oosterveld F. G.; Oostveen J. C.; Roorda L. D.; Viergever M. A.;
ten Wolde S.; Lafeber F. P. J. G.; Bijlsma J. W. CHECK (Cohort Hip and Cohort Knee): similarities and differences with the osteoarthritis initiative. Ann. Rheum. Dis.
2009, 68, 1413-1419.
(17) Kellgren J. H.; Lawrence J. S. Radiological assessment of osteo-arthrosis. Ann.
Rheum. Dis. 1983, 16, 494-502.
(18) Bligh E. G.; Dyer W. J. A rapid method of total lipid extraction and purification.
Can. J. Biochem. Physiol. 1959, 37, 911-917.
(19) Miller J. N.; Miller J. C. Statistics and chemometrics for analytical chemistry.
Pearson Education Limited: Harlow. 2000,
(20) Lamers R. J. A. N.; van Nesselrooij J. H. J.; Kraus V. B.; Jordan J. M.; Renner J. B.;
Dragomir A. D.; Luta G.; van der Greef J.; DeGroot J. Identification of an urinary metabolite profile associated with osteoarhtritis. Osteoarthr. Cartilage. 2005, 13, 762-768.
(21) Lamers R. J. A. N.; DeGroot J.; Spies-Faber E. J.; Jellema R. H.; Kraus V. B.;
Verzijl N.; TeKoppele J. M.; Spijksma G. K.; Vogels J. T.; van der Greef J.; van Nesselrooij J. H. J. Identification of disease- and nutrient-related metabolic fingerprints in osteoarthritic guinea pigs. J. Nutr. 2003, 133, 1776-1780.
(22) Damyanovich A. Z.; Staples J. R.; Chan A. D. M.; Marshall K. W. Comparative study of normal and osteoarthritic canine synovial fluid using 500 MHz 1H magnetic resonance spectroscopy. J. Orthop. Res. 1999, 17, 223-231.
(23) Damyanovich A. Z.; Staples J. R.; Marshall K. W. 1H NMR investigation of changes in the metabolic profile of synovial fluid in bilateral canine osteoarthritis with unilateral joint denervation. Osteoarthr. Cartilage. 1999, 7, 165-172.
(24) Fernandez-Quintela A.; Churruca I.; Portillo M. P. The role of dietary fat in adipose tissue metabolism. Public Health Nutr. 2007, 10, 1126-1131.
(25) Lambrecht S.; Verbruggen G.; Verdonk P. C. M.; Elewaut D.; Deforce D.
Differential proteome analysis of normal and osteoarthritic chondrocytes reveals distortion of vimentin network in osteoarthritis. Osteoarthr. Cartilage. 2008, 16, 163-173.
(26) Guo D.; Tan W.; Wang F.; Lv Z.; Hu J.; Lv T.; Chen Q.; Gu X.; Wan B.; Zhang Z.
Proteomic analysis of human articular cartilage: identification of differentially expressed proteins in knee osteoarthritis. Joint Bone Spine. 2008, 75, 439-444.
(27) Aigner T.; Fundel K.; Saas J.; Gebhard P. M.; Haag J.; Weiss T.; Zien A.; Obermayr F.; Zimmer R.; Bartnik E. Large-scale gene expression profiling reveals major pathogenic pathways of cartilage degeneration in osteoarthritis. Arthritis Rheum.
2006, 54, 3533-3544.
