TL;DR
This paper introduces an automated, high-throughput screening workflow using 3D light-sheet microscopy to analyze mitotic phenotypes in 3D cell cultures. This approach overcomes limitations of conventional microscopy, enabling detailed characterization of gene function and drug candidates in more physiologically relevant tissue-like models.
What Problem Does This Paper Address?
Studying dynamic biological processes like the cell cycle in 3D cell cultures is crucial for understanding disease and developing treatments. However, conventional microscopy struggles with the speed, resolution, and light exposure needed for high-throughput screening of these complex 3D structures. This makes it difficult to accurately assess how genetic or epigenetic changes affect cell division in a tissue-like environment. The authors aim to bridge this gap by developing a robust and automated workflow that leverages the advantages of light-sheet microscopy for large-scale screening.
Methods
The core of this work is a fully automated workflow for high-content screening of 3D cell cultures. Key components include:
- Automated Sample Preparation: A liquid handling robot automates the mixing of cells with Matrigel and their precise spotting onto imaging plates, minimizing variability and cost.
- 3D Light-Sheet Microscopy: A dual inverted selective plane illumination (diSPIM) microscope is used for imaging. This technique offers improved signal-to-noise ratio, higher imaging speed, isotropic resolution, and reduced phototoxicity compared to traditional confocal microscopy. Spheroids are imaged for 24 hours at 5-minute intervals.
- Image Processing Pipeline ('hSPIM'): A custom pipeline is developed to handle the large datasets generated by diSPIM. It performs image fusion, deconvolution, and data depth reduction, significantly improving XYZ resolution and signal-to-noise ratio.
- Phenotype Classification: A convolutional neural network (CNN) is employed for accurate cell cycle phase detection (interphase, prophase, metaphase, anaphase) directly from image slices, achieving 96% accuracy. This is combined with random forest-based classification for quantifying phenotypic traits.
Datasets
The paper utilizes the following:
- MCF10A H2B-GFP cells: These cells, expressing a fluorescent protein to label DNA, are used to form 3D spheroids for screening.
- HeLa H2B-GFP cells: Data from a previously established MitoCheck database, generated from HeLa cells undergoing siRNA knock-down, is used for comparison and validation of identified phenotypes.
No new benchmark datasets are explicitly introduced, but the paper demonstrates the application of their workflow to analyze phenotypes in a novel context (3D spheroids) and with epigenetic modifications.
Key Insights and Ideas
- Bridging the gap between 3D culture and high-throughput screening: The workflow successfully adapts light-sheet microscopy, previously limited in throughput, for large-scale screening of 3D cell models.
- The power of automated pipelines: The combination of robotic sample handling and a dedicated image processing pipeline is crucial for achieving reproducibility and efficiency in analyzing complex 3D imaging data.
- CNNs excel at 3D cell cycle phase detection: The deep learning approach significantly outperforms traditional methods, achieving high accuracy in classifying cell cycle phases directly from 3D image data.
- Epigenetic modifications impact mitosis: The study demonstrates that targeting epigenetic regulators can indeed lead to observable mitotic phenotypes in 3D cultures, opening new avenues for research.
- Phenotypic clustering reveals distinct cellular responses: Hierarchical clustering of quantified phenotypes effectively groups genes with similar effects, providing a higher-level understanding of their roles in mitosis.
Key Results
- The developed workflow successfully identified mitotic phenotypes in 3D cell cultures after siRNA knock-down of 28 mitotic target genes and after epigenetic perturbation of the RGMA gene.
- The CNN-based cell cycle phase detection achieved 96% accuracy on manually annotated data.
- The 'hSPIM' processing pipeline reduced data size by approximately 75% while improving resolution and signal-to-noise ratio.
- Hierarchical clustering identified five distinct phenotypic groups, revealing conserved and novel roles for genes in mitosis.
- The identified phenotypes from siRNA knock-downs in MCF10A spheroids largely agreed with those previously reported in 2D HeLa cells (MitoCheck database), validating the workflow's relevance.
- Epigenetic modification of RGMA led to phenotypes such as macronuclei formation, extended mitosis, and reduced spheroid growth.
Limitations and Future Work
- Restriction to intracellular perturbations: The current diSPIM setup, with its dipping lenses, primarily allows for the evaluation of intracellular perturbations like siRNA or CRISPR-Cas9 screens.
- Extending to small molecule screens: To adapt the workflow for screening small molecule libraries, samples would need to be physically separated into distinct wells. The authors suggest novel cell culture plate formats, like those with invertible pillar structures, could facilitate this.
- Combined approaches: The authors envision future work integrating this imaging data with other modalities, such as single-cell genomic or transcriptomic analyses, for a more comprehensive understanding of cellular processes.
- Further applications: The workflow is presented as adaptable for various high-throughput screening applications in tissue-like microenvironments, with potential impact in cancer research and beyond.
Summary updated