Supplementary Figures from Oxygen-Enhanced and Dynamic Contrast-Enhanced Optoacoustic Tomography Provide Surrogate Biomarkers of Tumor Vascular Function, Hypoxia, and Necrosis

Abstract

<p>Supplementary Figure S1 Diagram of the optoacoustic scan timeline Supplementary Figure S2. CAIX staining shows co-localisation to pimonidazole hypoxia staining in PC3 tumours. Consecutive sections from a representative PC3 tumour stained with CAIX (left) and pimonidazole (right) show very similar staining patterns, indicating that CAIX staining in PC3 tumours may be indicative of hypoxia. Supplementary Figure S3. Demonstration of responding fraction binarisation for metric extraction. Maps of OE-OT Î"SO2MSOT and DCE-OT Î"ICG (left) were compared with maps of standard deviation of the baseline variation in SO2MSOT and ICG signals (centre) and the pixels showing enhancement that exceeded 2x the standard deviation threshold were classified as responding (right). The ratio of responding tumour pixels to all tumour pixels defines the Responding Fraction (RF). Example images from PC3 tumours. Supplementary Figure S4 Schematic layer architecture for the Convolution Neural Networks based model for necrosis identification in H&E sections. Example images from PC3 tumours. Supplementary Figure S5 Examples of the necrosis identification algorithm performance. Necrotic regions were manually identified in 3 example H&E sections (A), as highlighted by red circles. The Convolution Neural Network based model (B) as used in the study was found to correctly identify all the necrotic and viable regions of the tissue. Example images from PC3 tumours. Supplementary Figure S6. Convolution Neural Network (CNN) based model accurately identifies tissue necrosis. Strong positive correlation was observed between the necrotic area fraction as quantified by CNN based model, and by manual identification of necrotic areas. n=18 PC3 tumours, * p<0.05, ** p<0.01, *** p<0.001 Supplementary Figure S7. Hypoxic tumours show weak OE-OT and DCE-OT response. Representative PC3 tumours with low (left) and high (right) CAIX staining (A) show differences in the amplitude of OE and DCE-OT responses (B), which translate to low OE and DCE RF in the binarised response of highly hypoxic tumours (C). Supplementary Figure S8. Necrotic tumours show weak OE-OT and DCE-OT response. Representative PC3 viable (left) and highly necrotic (right) tumours, as shown in H&E stained sections (A) show differences in the amplitude of OE and DCE-OT response (B), which translate to low OE and DCE RF in the binarised response images of necrotic tumours (C). Supplementary Figure S9. Treatment with a vascular disruptive agent causes vascular shutdown. Vascular Disruptive Agent CA4P led to marked haemorrhage in PC3 tumours , as seen in H&E sections (A). Significant differences in in the observed haemorrhagic area (B, n=12 PC3 tumours) and smooth muscle coverage of blood vessels (C, n=12 PC3 tumours) was observed between vehicle and drug treated tumours. * p<0.05, ** p<0.01, *** p<0.001 by paired two-tailed t-test. Box between 25th and 75th percentile, line at median. Supplementary Figure S10. Vascular disruption causes a dramatic decrease in OE-OT and DCE-OT responses. (A) Amplitudes of OE-OT Î"SO2MSOT (top) and DCE-OT Î"ICG (bottom) decrease dramatically following treatment with the vascular disruptive agent CA4P. Averaged kinetic curves for vehicle (B) and drug treated (C) animals confirm this observation. n=5 PC3 tumours were vehicle treated, n=7 PC3 tumours received CA4P. Shaded envelope in (B) and (C) indicates standard error of the mean. Supplementary Figure S11. Vehicle treated PC3 tumours show no change in spatial distribution of OE and DCE response. OE (top) and DCE (bottom) response maps show little change following vehicle treatment, contrary to the dramatic effect of the CA4P drug.</p>