An investigation of photo-activation of psoralen (AMT) during radiation therapy in a novel tissue model.

Purpose: RECA (Radiotherapy Enhanced by Cherenkov photo-Activation) is a novel treatment with potential to add an anti-cancer immunogenic component through Cherenkov activation of a photo-chemotherapeutic agent (psoralen). This work investigates RECA in a novel tissue-representative in-vitro model consisting of 4T1 murine cancer cells grown on thin slices of viable rat-brain tissue.Methods: Accurate estimation of viable tumor burden is of foundational importance to this work. A CellProfiler pipeline was created and optimized and validated on realistic simulated data/images where the ground truth of number of colonies and integrated intensity was known. Simulated data sets mimicked key features of real experimental data including colony spatial and size distributions, contaminant and stray light signals, colony overlap, and noise. The optimized CellProfiler pipeline was then applied to the original 4T1 tumor cell images to determine colony growth over five days. Several experiments were conducted prior to the RECA experiment to determine the best protocol. The first tested the optimal concentration of psoralen, loading technique, and type of psoralen by co-incubating 4T1 cells with psoralen for differing times and concentrations and subsequently exposing them to 365nm radiation at variable energy fluences. The plates were tested for 4T1 cell viability using Celltiter-glo and luciferase assay 48-72 hours later depending on confluence of the control plate. Another experiment tested the output of the CellProfiler image analysis for relative growth over time measuring 4T1 mCherry cells plated on rat brain slices at 10k,20k,30k,40k,50k cells per hemisphere. For the RECA experiment, six 12-well plates, each containing 1cm of agarose supporting a 400 µm thick coronal slice of viable rat brain tissue were created. Each plate represented one arm of an experiment incorporating the psoralen derivative 4’-aminomethyl trioxsalen (AMT): MV control, MV+AMT, kV control, kV+AMT, no irradiation control, and AMT alone control. 20,000 4T1 cells expressing both mCherry-flourescent and firefly luciferase-luminescent reporter proteins plated on each rat brain slice hemisphere. For the AMT arms, the cells were co-incubated with 1 µM AMT for 1 hour prior to plating. The MV arms received 4 Gy from a 15 MV linear accelerator beam, and the kV arms received 4 Gy from 160 keV photons. Images were taken of the plates each day for 5 days with a Zeiss Lumar microscope with rhodamine filter for the mCherry protein signal. Results: The CellProfiler pipeline measured integrated intensity of the 10 simulated images that best approximated the images from the experiment with an accuracy of 99.23% ± 0.75%. Further analysis on images with increasing colonies, background, and noise showed the pipeline was accurate on images with variable features. These results gave confidence that the same pipeline could be used on images from this experiment. AMT was found to be a more effective psoralen (0.43 ± 0.22% cell survival after 48hr) relative to 8-MOP (31.3% ± 3.7% cell survival after 72hr). The psoralen cell loading was found to be optimal at 1µM for 1 hr prior to plating. The analysis of the cell titration images showed a significant increase in signal for each increase in cells plated on day one and for all subsequent days except for the 20k cell arm. Additionally, the growth in signal for the plates was consistent between the arms except for the 20k arm due to extra signal on the periphery of the slice likely from displaced cells. Integrated intensity analysis of the 4T1 mCherry cells revealed a significant decrease in tumor proliferation by day 5 between the MV control (5.65±0.78-fold growth) and MV AMT (3.49±-0.52-fold growth) arms. This result is consistent with the hypothesis that psoralen is being activated, causing the decreased proliferation seen in MV AMT arm. The kV control and kV AMT arms had a smaller decrease in proliferation when compared to their MV counterparts (6.73±1.24 and 5.26±0.59-fold growth respectively). The growth observed in the Dark control arm was consistent with the 13.6 ± 1.5 hour doubling time for 4T1 cells. In the MV AMT arm, there were punctuated regions of increased signal in 7/12 wells not corresponding to colonies, making segmentation for this arm challenging. The viability of the brain slice was assessed each day and found to be stable over the 5 days. Conclusions: The technique of testing image analytic software on simulated images proved to be an effective tool to verify the software’s accuracy. A similar technique can be applied to images with new and challenging features. The rat brain slice model gives the opportunity to both generate Cherenkov in real tissue while providing a 3D matrix for the colonies to grow, which is an improvement to the 2D well plate culture for this experiment. This new model adds challenges of proper image analysis with cell autofluorescence as well as cell clumping. The preliminary results are consistent with psoralen activated in RECA treated cells causing decreased proliferation for the MV arm. Further work is needed to confirm and quantify the effect.