Investigation of the tumour microenvironment is scarce in canine cutaneous and subcutaneous soft tissue sarcomas (STSs) and previously has not been carried out with a focus on the spatial distribution (central tumour versus infiltrating margin) of inflammatory cells infiltrating the tumour. In canine cutaneous histiocytoma (CCH), investigations of the tumour microenvironment have not yet allowed to fully elucidate the mechanisms leading to regression. This research project investigated the tumour microenvironment of these two tumours, at the same time examining the utility of the image analysis program QuPath to support the investigations. In the first part of this study, the extent and subsets of infiltrating immune cells, including their spatial distribution, of 15 grade I and 15 grade III STSs was investigated by immunohistochemistry (IHC) with antibodies against CD3, CD20, CD204, CD206, Granzyme B and FOXP3. When examining all 30 STSs combined, the expression of CD3, CD20, FOXP3 and granzyme B was statistically significantly (p<0.05) higher in the invasive margin compared to central tumour areas, however there were no statistical differences when comparing the immune cell infiltrates in invasive margin and central tumour between grade I and III tumours. Dogs with a better clinical outcome exhibited higher CD20, CD3, Granzyme B, FOXP3 and CD204 immunolabelling in the invasive margin when compared to dogs with a worse clinical outcome, warranting further research on the immune environment and clinical outcome of STSs, in particular with a focus on the invasive margin of these tumours. In the second part of the study, immune infiltrates and markers of cell death were investigated in 30 CCH cases at different stages of regression by IHC with antibodies against CD3, CD204, HMGB1 and Activated Caspase 3. CD3, CD204 and Activated Caspase 3 increased significantly (p<0.05) with regression whereas HMGB1 labelling was not significantly modified. Therefore, apoptosis appears to be the main cell death pathway involved in CCH regression. However, bearing in mind the robust histiocytic inflammatory response incited in the regression of CCH, further investigations in the exact mechanisms of cell death continue to be warranted which may also help elucidate the mechanisms underlying regression. Overall, automated image analysis, using QuPath, allowed to gather quantitative (and objective) data one whole scanned slides, which in turn was comparable to results from previously published data acquired by manual counting of cells, in turn validating this approach. However, the training of QuPath is labour and time intensive and necessitates supervision of attentive and trained operators. Furthermore, the algorithms generated are anticipated to only be able to be applied on the present datasets since staining parameters are expected to vary in differing IHC protocols and scanning settings. Therefore, whilst automated image analysis can provide highly objective data, possibilities and drawbacks need to be taken into account before proceeding with the element of machine learning/training.
Samuel Kalpers (Thu,) studied this question.