Utility of Cellular Imaging Modality in Subcellular Spatial Transcriptomic Profiling of Tumor Tissues

2024-07-30

Discover ROICellTrack, a deep learning framework integrating cellular imaging with spatial transcriptomics to enhance tumor tissue analysis and insights into heterogeneity.


Spatial transcriptomics (ST) technologies, such as the GeoMx Digital Spatial Profiler (DSP) by NanoString and the Visium platform by 10x Genomics, have revolutionized cancer genomics by elucidating spatial heterogeneity and complex interactions within the tumor-immune microenvironment. However, these technologies often underutilize the spatial information from immunofluorescence imaging. To address this, the research team developed ROICellTrack, a deep learning-based framework to integrate cellular imaging with spatial transcriptomic profiling, aiming to provide a comprehensive spatial omics analysis.

Methods

ROICellTrack was applied to 56 regions of interest (ROIs) from urothelial carcinoma of the bladder (UCB) and upper tract urothelial carcinoma (UTUC). Multiplex immunofluorescence (mIF) staining was performed, and the GeoMx DSP platform captured RNA reads from the ROIs. The deep learning model Cellpose was used for cell segmentation, followed by quantification of cell numbers and average color intensities for each cell.

Results

ROICellTrack demonstrated high accuracy in identifying cancer-immune mixtures and associated cellular morphological features. The integration of image-based analysis with spatial transcriptomics revealed different spatial clustering patterns and receptor-ligand interactions. These findings emphasize the importance of combining imaging and transcriptomics for detailed within-sample heterogeneity analysis.

  1. Cell Identification: ROICellTrack accurately identified cancer cells and estimated cancer cell burden within samples, providing a reliable measure of tumor purity.
  2. Spatial Clustering: The degree of tumor-stroma mixture was assessed, with results showing distinct transcriptomic features based on the spatial clustering patterns.
  3. Receptor-Ligand Interactions: The analysis identified more receptor-ligand interactions in regions with mixed tumor-stroma patterns compared to separative patterns, highlighting differences in cell-cell communication.

Conclusion

Combining cellular imaging with spatial transcriptomics enhances the analysis of tumor tissues, providing new insights into tumor heterogeneity and potential implications for targeted therapies and personalized medicine. ROICellTrack represents a significant advancement in integrating spatial and transcriptomic data for comprehensive cancer research.

SUBSCRIBE TO OUR NEWSLETTER