Poster R12 - Novel 4 Component Fluorescence Lifetime Imaging Microscopy (FLIM) Analysis

Introduction: Management of breast cancer involves effective diagnostic and treatment
approaches because molecular analysis is directly coupled to therapy, such as for estrogen receptor (ER), progesterone receptor (PR), and HER2. However, triple negative breast cancer (TNBC) lacks targeted therapies with worse prognosis. A major challenge to identifying more drugable molecular drivers is the vast variety of cellular subtypes and phenotypes that comprise the tumor microenvironment, and this heterogeneity is a key factor underlying solid tumor progression, metastasis, and the development of drug resistance. Substantial diversity can exist amongst cancer clones, as well as numerous host cell subtypes that can also promote tumorigenesis. Hence, there is a critical need to identify and molecularly characterize all cells in the tumor to provide a comprehensive diagnosis and guide disease management. Current clinical diagnostic methods lack the needed detection capacity (i.e. histology/IHC) and/ or cellular resolution (i.e. sequencing/microarrays) to phenotype all cell subtypes.

Materials and

Methods: A fourth target (ECadherin, ECad) and lifetime dye (KU530) was added to the panel and again labeled to different breast cancer cell lines. As shown in Fig. part i, complex and unique phasor fingerprints were obtained for each cell line. Total intensity images (part ii) were then deconstructed into the contributions for each probe species using the phasor analysis algorithm. The number of probes were mapped that were found in each pixel. Finally, mean intensity was determined and compared to results obtained using single probes, showing close correlation (part iii).

Results, Conclusions, and Discussions: Phasors tend to attribute some signal to all probe candidates, which can lead to significant errors for very low expression cases. This can be seen for HER2 in Fig. Ciii. However, phasors provide a fingerprint that can readily be recognized visually, making it possible to remove non-contributing probes from analysis and prevent false positives. To do this in an unbiased manner, machine learning classifier was developed by training a deep neural network using images that were stained with varying numbers of probes and different cell lines. This eliminated false positives through order reduction. There is an outstanding need for high content molecular analysis of breast tumor specimens to provide information about key cancer drivers, predict progression, and personalize therapeutic decisions for patients in an integrative and connected manner. By considering tumor heterogeneity, different host and neoplastic cell subtypes, and detailed phenotyping, the next step has been taken towards guiding disease management in a comprehensive and personalized manner. Molecular targeting has been successful for breast cancer in some cases (HER2, ER/PR) in part because the drivers can be detected and drugged, but triple negative breast cancers require a more comprehensive approach. The same will be the case for many other types of cancer. This research can lead to new drug targets, or more likely cocktails. Additionally, one can seek rare clones, possibly cancer stem-like cells (CSCs), that may be therapy-resistant and ultimately cause relapse. Using this technology, on can characterize tumor-associated immune cells to inform immunotherapies. This substantially increases the amount of information that can be obtained per measurement, would solve the TNBC challenges and enhance overall detection capacity.

Acknowledgements (Optional):