Evaluation of rodent models of osteoarthritis using lipidomic profiling and behavioural studies

Osteoarthritis (OA) is a complex, multifactorial, and slowly progressive disease where there is currently no effective medical treatment. Research in understanding the mechanisms of OA has been advanced by preclinical studies in rodent models of OA. Recent evidence highlights the role of different classes of lipids in OA pathogenesis. Therefore, the main aim of this thesis was to apply both targeted and untargeted (global) lipidomics mass spectrometry based an alytical methods, in conjunction with univariate and multivariate statistical analysis, in various tissues from three established rodent models of OA; meniscal transection (MNX), monosodium iodoacetate (MIA), and destabilization of the medial meniscus (DMM). The overall goal was to identify statistically differentiated lipids between controls versus OA rodents that may reflect changes in the pathophysiology of OA and associated pain. In addition, a global lipidomics workflow was developed by me, following the latest trends used within the wider metabolomics community, ensuring robustness and reproducibility in the identification of putative metabolite/lipid biomarkers for diseases. Experiments in this thesis using a targeted oxylipin liquid chromatography tandem mass spectrometry (LC-MS/MS) method showed that statistical significant changes in the levels of certain oxylipins were observed. More specifically, 11,12-DHET (mean concentration: 0.26 pmol/g in control, 0.54 pmol/g in MNX; p<0.01), 14,15-DHET (0.46 pmol/g in control, 0.75 pmol/g in MNX; p<0.05) and 8-HETE (5.46 pmol/g in control, 7.40 pmol/g in MNX; p<0.05) were statistically increased in the MNX compared to control (sham) rats in ventral spinal cord in the MNX rat model of OA. These findings are supported by literature since these three lipids exhibit pro-inflammatory properties and thus are expected to increase in the OA group where inflammation is the main feature of OA. Regarding the MIA rat model levels of other oxylipins in synovial fluid were differentially expressed in the MIA compared to saline (control) rats. Arachidonic acid (AA), (272.3 pmol/g in control, 435.3 pmol/g in MIA; p<0.05) was increased in the MIA-treated compared to saline-treated rats, while 9-HODE (4.42 pmol/g in control; 1.21 pmol/g in MIA; p<0.05) was statistically decreased in the MIA compared to saline rats. Since AA has been reported to be released from membrane phospholipids in OA, the observation that AA is statistically increased in synovial fluid in MIA- compared to saline-treated rats bears strong significance. In addition, maps of oxylipins metabolism were generated to visualize the pathways underlying the changes of lipid concentrations in plasma between control and OA rats for both MNX and MIA rat models. Therefore, applying a targeted oxylipin LC-MS/MS method in different tissues of MNX and MIA rat models of OA is a successful approach and informative about changes in pathophysiology of OA, underlying significant alterations in oxylipins concentrations. Although the global lipidomics approach was able to measure different classes of lipids that might account for differences in plasma between MNX/MIA and sham/saline-treated rats, this approach exhibited weak MVA (multivariate analysis) models. In contrast to MNX and MIA rat models, the global LC-MS lipidomics profile in plasma from a DMM mouse model of OA exhibited excellent MVA models with good prediction scores. Twenty-six statistically significant lipids were identified, using the lipidomics workflow that I have developed, and when four of these lipids were used to build Receiver Operative Curves (ROC) the model produced high prediction (84%) power in separating sham from DMM mice. The identity of these four lipids was classified as being fatty acids (FAs), sterols, sphingolipids, and diacylglycerols (DAG). In addition, MS/MS experiments were performed to confirm the identity of significant lipids. Thus, it was shown herein that applying a global lipidomics LC-MS approach in plasma from the mouse DMM model, using only a small number of mice (15 in total), can be informative about significant changes in the “lipidome” in OA and can be used as a robust means of predicting OA in mice based on their global lipidomics profile. Lastly, correlation statistical analysis was applied between levels of lipids in the various tissues, pain behaviour, and histopathology parameters in the three rodent models of OA. Although many oxylipins/lipids levels were found to be statistically correlated with the aforementioned parameters, the most striking finding is that 9-HODE and AA were both found to be positively correlated with Weight Bearing (WB), a parameter of pain behaviour, in plasma and synovial fluid in the MIA rat model of OA. Since plasma reflects systemic inflammation and synovial fluid reflects local (inflammation) 9-HODE (p<0.01 in plasma; p<0.05 in synovial fluid) and AA (p