Skip to main content
2024 BMES Annual Meeting Main banner
Icon Legend
This session is not in your schedule.
This session is in your schedule. Click again to remove it.
Back

Biomedical Imaging and Instrumentation

Biomedical Imaging and Instrumentation - Poster Session E

Poster F2 - Streamlining Subcellular Object Detection Via Advanced Image Analysis

Saturday, October 26, 2024
10:00 AM – 11:00 AM EST
Location: Exhibit Hall E, F & G

Introduction:  Glioblastoma (GBM) is one of the most common and deadliest forms of brain cancer, with a 3-5% survival rate past three years for patients [1]. One contributing factor to the severity of glioblastoma is its ability to quickly become resistant to apoptosis-inducing therapies [2]. To combat this, we are studying methuosis, a non-apoptotic form of cell death characterized by excessive macropinocytosis, leading to catastrophic cell vacuolization and plasma membrane collapse [3]. The events leading up to methuotic cell death (e.g., accumulation of vacuoles) are most easily captured using phase-contrast microscopy without the use of dyes that are challenging to remove for subsequent rounds of imaging. However, conventional image analysis platforms including FIJI, Ilastik, and CellProfiler struggle to quantify phase contrast images reproducibly due to variance in pixel intensities at vacuole edges and related issues. Here, we developed a novel pipeline that builds on the strengths of those platforms and additionally leverages the CellProfiler Analyst (CPA) application to build a pipeline with enhanced accuracy. The resulting pipeline represents a substantial improvement due to the machine learning algorithms embedded in CPA.

Materials and

Methods:  U251 human glioblastoma cells were plated in 96-well plates and treated with 4 µM MOMIPP, a potent small-molecule inducer of methuosis, or 0.2% DMSO (control). After 4 hours of treatment, the cells were stained with the plasma membrane dye CellMask and live-cell DNA stain NucBlue and then washed 3 times with 1× PBS before bright-field imaging on a Cytation5 imaging plate reader at 20×. Cells were then fixed in 4% paraformaldehyde, 0.05% glutaraldehyde, and 0.05% TritonX-100 for 10 minutes before being washed again with 1× PBS. We then performed immunofluorescence staining for PIP3 along with the DNA stain Hoechst and again imaged on at 20×. After collecting live- and fixed-cell images, FIJI was used to increase the contrast on the bright-field images to improve the pixel segmentation processing time for subsequent analysis in Ilastik. Ilastik is a user-trained, machine learning platform that segments images based on pixel classification. The output from Ilastik was used as input for a CellProfiler pipeline designed to quantify vacuole numbers, shapes, and sizes. CellProfiler was additionally used to align live cell images, vacuole objects from Ilastik, and fixed immunofluorescence images. CPA then used measurements produced by CellProfiler to create a reusable model file that learns to recognize what features of size, shape, and intensity fit the definition of vacuolization in cells.

Results, Conclusions, and Discussions: The pipeline was able to identify subcellular vacuoles in U251 cells formed in response to MOMIPP and correlate fluorescence images to the appearance of vacuolization (Fig.1A). PIP3 is highly involved in macropinocytosis, and our image analysis pipeline shows a positive correlation on a cell-by-cell basis between PIP3 signal and vacuole accumulation [4] (Fig.1B). Providing evidence towards the pipelines ability to recognize and correlate new subcellular features towards vacuolization rather than relying on one subcellular feature. The pipeline we have developed produces a confusion matrix with a 90% accuracy (Fig. 1C) using the Random Forest classifier training format for classifying phenotypes between vacuolized and non-vacuolized cells. Over time, the algorithm learns to create connections between the existing information we gave, and the new information provided, allowing the pipeline to analyze, learn and adapt. These results will help advance our understanding of vacuolization accumulation within cells during methuosis as we expand the number of features that can nominate potential novel therapeutic targets to selectively drive the accumulation in GBM.

Acknowledgements (Optional):  This work was supported by funding from the National Cancer Institute (U54 CA274499) and the National Science Foundation (REU award 1950374).

References:
[1] Mohammed, et al. (2022). Rep Pract Oncol Radiother, [2] Kishan Yalamarty, et al. (2023). Cancers (Basel) [3] Maltese & Overmeyer (2014). Am J Pathol, [4] Kay, et al. (2018). Biochem J