(G-268) Developing a 3-Dimensional Model of Rodent Knee Joints Using Histological Images

Assessing pathogenic joint tissue changes is crucial for diagnosing and treating joint diseases. Micro-computed tomography (μCT) provides high-resolution 3-dimensional (3-D) images of hard tissue, but is expensive, requires access to specialized equipment, and is unable to identify important cellular tissue changes, especially in soft joint tissues like nerves, synovium, and some cartilage. Alternatively, histology is the gold standard to assess all joint tissues at the cellular level and is widely accessible, versatile, and cheap. However, histology can currently only provide 2-dimensional tissue images, making it difficult to determine 3-D whole joint health. There is a need to develop accessible, cheap, and detailed 3-D assessment of all joint tissues at the cellular level to better understand joint disease pathologies; to do this, we will elevate histological techniques to produces μCT-like models by digitally stitching serial two-dimensional histology images into a 3-D model to produce a low-cost, accessible, and detailed alternative method to visualize whole joints. 

Preclinical rodent models of osteoarthritis allow for the study of disease pathogenesis and the therapeutic effects of treatment modalities in a controlled environment. An open-source and inexpensive method of visualizing knee tissue in three dimensions would be quite beneficial as a tool to more effectively analyze candidate treatments’ effects on tissue morphology.  

This project consists of two parts: 1) developing histology skills to produce joint images (completed) and 2) applying computational skills to produce a 3-D model of mouse knees with joint tissues identified (in progress). 

Part one: Healthy contralateral (n=8) and degenerated ipsilateral (n=4) knees from adult mouse cadavers from another study were used. Mice received a unilateral injection of monoiodoacetate (2 μg/10 μl) then were allowed to develop joint degeneration for 6 weeks prior to euthanasia. Dissected cadaver joints were fixed for 48 hours in 10% neutral buffered formalin, decalcified for 2-5 days using Cal-Ex, embedded in paraffin, sectioned on a microtome (10 μm, approximately 120 sections per joint), and mounted on glass slides. A toluidine blue staining protocol was optimized by testing several acidities and buffer combinations to achieve high contrast metachromasia across joint tissues for ideal tissue tracking and differentiation once imported into a 3-D model. One to ten joint sections for every 100 μm were stained to visualize articular cartilage, bone, synovium, and meniscus tissues then imaged using an EVOS XL microscope. 

Part two: Joint section images were imported and stitched together into a preliminary 3-D model using MATLAB. Optimizing the model and characterizing joint tissue features is ongoing however initial targets for the model to identify are articular cartilage, bone, meniscus, and synovium. Future iterations of this model will enable quantification of tissue changes such as articular cartilage lesions, subchondral bone density, osteophyte formation, and synovial inflammation with comparisons to both histological and μCT assessment methods. 

Standard histological techniques produced high quality images of rodent joint sections using an acidic toluidine blue solution (pH 4.0) in 100mL McIlvaine’s buffer (1.10g Na₂PO₄, 1.19g citric acid). Toluidine blue’s metachromatic properties differentiate tissues of different charges (e.g., articular cartilage and bone) by reflecting as either blue (bone) or purple (articular cartilage). Optimizing this metachromasia by adjusting buffer type, acidity, and tissue dehydration will enable accurate color-based differentiation of tissues in the 3-D model. Toluidine blue was chosen as the simplest, single-step stain accessible to the most researchers and the stain recommended for assessing osteoarthritis (the most common joint disease) by the Osteoarthritis Research Society International histological guidelines which will be used as the ground truth to assess accuracy and precision of the 3-D model. Other stains, such as safranin-O/fast green or hematoxylin/eosin, more dramatically differentiate tissues and will be included in future iterations of this model.  

The joint images were easily imported and stitched together in MATLAB. However, slight distortions caused by the histological processing (likely sectioning, de-/rehydration, or slide mounting) currently prevent smooth stitching across serial sections.  These distortions need to be minimized, possibly by increasing section thickness, to develop an accurate 3-D model of the knee joint while maintaining sufficient spatial resolution to identify joint tissues and pathogenic degeneration. This was corrected by taking serial images of sections at each step in the staining process, leading to the determination that the final dehydration step produced too much mechanical strain on the tissue as described in the initial protocol.  

All in all, this method will provide a cost-effective, 3-D assessment of joint tissue alternative to expensive μCT with added detailed cellular joint tissue analysis. Measuring specific cellular tissue features in 3-D joint architecture will provide deeper insight into whole joint health, with implications across many joint diseases. Further work will involve solidifying a sectioning and staining procedure to facilitate the downstream computational work and avoid distortion-related difficulties.