Osteoarthritis (OA) is the most common joint disease worldwide, characterized by progressive deterioration of articular cartilage. A connective tissue covering the bony ends of synovial joints, cartilage has poor capacity for self-repair. Indeed, currently no reparative cures exist for OA. This leaves disease prevention as the best option, which necessitates understanding of OA pathology and the multiple factors affecting OA progression. Thus, understanding the combined effects of various OA subtypes, pursued via in vivo small animal models, is similarly essential.

The main aim of this thesis was to characterize the content and structural changes of articular cartilage in rat knee cartilage resulting from different OA models. The large volume fraction of cells in rats was assumed to affect the tissue-level analysis. Therefore, the accurate measurement of cartilage proteoglycan (PG) content warranted methodological improvement of digital densitometry (DD) image analysis. Namely, cell removal from densitometry images was automated by using deep learning-based segmentation tool U-Net. Following training, the accuracy of U-Net was evaluated and the effect of the cell removal to DD measurements was measured. After cell segmentation, OA risk factors of obesity and anterior cruciate ligament transection were studied in combination or separately. PG content was measured with DD-imaging and collagen content with Fourier transform infrared (FTIR) spectroscopy imaging. Polarized light microscopy (PLM) was used to characterize the collagen network organization. Further, the cartilage content/structure data underwent correlation analysis with key biomarkers from synovial fluid and blood serum (e.g., IL-1, leptin, and VEGF) to investigate the biological effects of obesity and joint trauma to cartilage. Finally, to estimate biomechanical risks for OA development, a novel rat knee joint computational model was developed. This was achieved with micromagnetic resonance imaging, previously measured biomechanical properties of cartilage, combined with musculoskeletal and finite element (FE) modeling.

In study I, the trained U-Net removed cells from densitometry images with an approximate accuracy of 67%. We found a small, but significant 2-4% difference to the measured optical density caused by cell removal. Additionally, we detected a unique cell distribution pattern with the highest cell density residing in middle cartilage. In study II, it was discovered that animals with only joint trauma had lower PG content and higher collagen orientation angles, especially in the superficial cartilage. Surprisingly, obesity and joint trauma together increased PG content, and moderately increased the orientation angles. However, collagen parallelism was severely disrupted in animals with combined joint trauma and obesity. Despite this, the collagen content in the cartilage was predominantly unaffected by either joint trauma or obesity. In study III, significant correlations were seen between cartilage PG content, collagen orientation, collagen parallelism and collagen content and biomarkers, namely synovial fluid/serum leptin ratio, VEGF and fractalkine. In study IV, an FE model of rat knee was successfully developed. The model indicated increased contact pressure, maximum principal stress, maximum principal strain, and fluid pressure that decreased towards the end of the stance on the lateral side of the joint.

The inaccuracies of U-Net were likely caused by the complex structures between cells and ECM. The small but significant effect to PG content analysis via DD images caused by presence of cells was likely due to the marked difference between PGs in- and outside of the cells. Thus, the removal of cells is recommended for histochemical image-based analysis of cartilage. The alterations in PG content, collagen orientation and collagen parallelism in obese, lean, and transected animals could be related to alterations in gait, joint loading, and inflammatory profiles in the animals. As the collagen content was predominantly unaffected by different OA models, collagen loss seems to characterize late or advanced OA. Negative correlation was found between synovial fluid/serum leptin ratio, cartilage proteoglycan content and collagen parallelism. Based on previous literature it possible that the consumption of leptin inside the joint could have an interaction with the metabolism of cartilage. Positive correlation between collagen content and VEGF was found similarly. The developed FE model simulated cartilage mechanical behavior that matched well to those reported in previous literature. Thus, once further iterated, the model could be used to reveal biomechanically and biochemically driven mechanisms to trigger development of cartilage degradation.

To conclude, this thesis has provided methodological improvements on histochemical imaging -based cartilage content measurements via automated cell removal. Additionally, we have provided novel FE model of rat knee joint and it’s mechanical environment. Importantly, the results in the thesis indicate collagen organization to be severely disrupted when obesity and joint trauma are combined. However, obesity might also have beneficial effect of increasing the PG content in cartilage. Similarly, the correlative patterns between cartilage and leptin, VEGF and fractalkine highlight a need to elucidate the possible interactions between these mediators and cartilage.