From 2D to 3D
Transcriptomics documents what RNA is being actively expressed in cells, which indicates the activities the cell is engaged in and reveals the cell types present in a sample. Spatial transcriptomics does this but for individual cells to build a 2D map.
The team got early access to the CosMx instrument from the company NanoString, which does this at extremely high resolution – 1,000 different RNA molecules can be detected at one time, compared to traditional methods that identify just a handful of molecule types at once. The team analyzed 340,000 individual cells from the lung tumor, identifying 18 cell types.
The 3D analysis was powered by a new computational algorithm, STIM, which aligns datasets to reconstruct 3D virtual tissue blocks. “We realized that spatial transcriptomics datasets can be modeled as images,” says Dr. Nikos Karaiskos, a postdoctoral researcher in the Rajewsky lab and co-corresponding author of the second “Cell Systems” paper describing STIM in detail. Leveraging imaging techniques, STIM marries the fields of computer vision and spatial transcriptomics. The team worked closely with Dr. Stephan Preibisch, a former principal investigator at MDC-BIMSB who is now at Howard Hughes Medical Institute’s Janelia Research Campus in the U.S., to bring this collaborative effort to fruition.
They then worked with the Systems Biology Imaging Platform in Mitte to apply a separate imaging technique, called second harmonic generation, to map elastin and collagen in cellular neighborhoods, which in the lung are the main extracellular matrix constituents. Areas with more elastin were healthier, while those with more collagen surrounded the tumor cells, which indicates harmful tissue remodeling.
“So not only do we know what cell types are present, we know how they are grouped with their neighbors, and we could begin to understand how tumor cells rewire non-malignant cells at the tumor surface to support tumor growth,” explains Tancredi Massimo Pentimalli, MD, the first paper author who is pursuing a PhD in the Rajewsky Lab and the Berlin School of Integrative Oncology at Charité – Universitätsmedizin Berlin.
Cells talk
But the analysis did not stop there. The team was able to understand precise phenotypes – for example, if fibroblasts, which form connective tissue, were activated and remodeling the tissue or not. They were also able to listen in on cell-to-cell communication and determine how tumor cells were blocking immune cells from entering the tumor.
“This immune suppression mechanism is well-known and suggests immunotherapy would help,” Pentimalli says. “Immune checkpoint inhibitors would reverse the suppression and then you have this army of immune cells that are already in position ready to attack. It was exciting to see how our approach identified this dynamic and could help direct a personalized immunotherapy plan.”
Notably, these key insights were only possible with data in 3D – in 2D it was impossible to distinguish between the tumor and other immune cells embedded in the tumor surface.
Pathology 2.0
The beauty of this approach is that, while very high-tech, it starts with a routine tissue sample found in any pathology lab. For this study, the group used a tissue sample of a lung tumor that was several years old, preserved with formalin and embedded in paraffin wax – the standard method pathologists use to preserve archival tissues.
“We were able to extract all this wealth of molecular information from one very thin section of a sample that has been sitting around at room temperature for years,” Pentimalli says. “This is pathology 2.0 – not just looking at the cells under a microscope to make a diagnosis, but bringing molecular insight to the clinic.”
Next steps
Now that the proof-of-concept has been established, the team plans to apply the approach to larger datasets. They are currently working on 700 samples from 200 patients and collaborating with Dr. Fabian Coscia, who leads the Spatial Proteomics Lab at Max Delbrück Center, to integrate protein activity into the analysis.
Reference: Preibisch S, Innerberger M, León-Periñán D, Karaiskos N, Rajewsky N. Scalable image-based visualization and alignment of spatial transcriptomics datasets. cels. 2025. doi: 10.1016/j.cels.2025.101264
Pentimalli TM, Schallenberg S, León-Periñán D, et al. Combining spatial transcriptomics and ECM imaging in 3D for mapping cellular interactions in the tumor microenvironment. cels. 2025. doi: 10.1016/j.cels.2025.101261
This article has been republished from the following materials. Note: material may have been edited for length and content. For further information, please contact the cited source. Our press release publishing policy can be accessed here.
