We performed a comprehensive analysis of the phenotypes present in primary breast cancer and metastatic breast cancer. This information enabled the development of antibody panels and highly multiplexed tissue imaging workflows that now provide an in-depth view of tumor cells, T cells, macrophages, and fibroblasts in tumor ecosystems. For the rapid analysis of tumor tissues, we have developed a method that uses X-ray fluorescence imaging of metals bound to antibodies, called MEZ-XRF imaging. To increase speed and sensitivity, we coupled MEZ-XRF to a signal amplification method for metal-based reporters. Together, these methods enable unprecedented tissue imaging speeds in the kilohertz range. In order to enable efficient analysis of these large highly multiplexed tissue imaging datasets and to support statistically sound analyses, we have furthermore developed computational tools. These include workflows for the phenotypic and spatial analysis of the highly multiplexed tissue imaging data generated in this project, spatial statistical methods for optimized sampling, and methods to increase the number of simultaneously measurable antibodies using compressed sensing. Importantly, all antibody panels, technologies, and computational methods are generally applicable and can be used to study other tumor types.
In order to gain a better understanding of metastatic breast cancer, we used these approaches and analysed matched primary breast tumors and lymph node metastases, as well as primary breast tumors and distant metastases. Comparing single-cell phenotypes of primary breast tumors and matched lymph node metastases, we identified phenotypes and spatial organizations of disseminated tumor cells associated with patient survival. Notably, using the lymph node metastasis tumor cells, we identified high-risk patients within a patient group normally considered low risk. We also performed an analysis of the tumor ecosystems at distant breast cancer metastatic sites. We found that, surprisingly, the same tumor cell phenotypes were typically present in primary tumors and metastatic sites, and that the metastatic site did not influence tumor phenotype composition, except for the brain. Analysis of the immune cell phenotypic composition of metastatic breast cancer highlighted the spatial heterogeneity of the disease within patients and across distant metastatic sites, indicating myeloid cells as the predominant immune modulators that could potentially be targeted at these sites. Finally, we also analyzed a large cohort of triple-negative breast cancer patient samples, the breast cancer subtype most prone to metastasize, and managed to develop a novel prognostic classification system reflecting phenotypic and spatial properties as well as metastatic propensity.
With the knowledge of the primary and distant metastasis tumor ecosystems, we were well equipped to study how these ecosystems can be modulated. We therefore established and validated patient-derived breast cancer organoids and ex vivo tumor slice cultures, and we are using these for drug screening and functional genomics experiments to understand the functional properties of the multicellular assemblies. We have performed drug screens testing thousands of conditions. We found that the different breast cancer phenotypes and multicellular assemblies have unique drug response profiles. We currently perform follow-up experiments and develop computational models that predict how to best drug a given tumor ecosystem given its phenotypes and their spatial organization, supporting the future development of individualized drug treatment approaches.
The results have been exploited scientifically through the development of generally applicable antibody panels, imaging workflows, computational tools, and functional model systems. We established and validated patient-derived breast cancer organoids and ex vivo tumor slice cultures, which are now used for drug screening and functional genomics experiments. This work provides a basis for computational models that predict how a given tumor ecosystem may best be treated, thereby supporting future individualized therapy strategies. The results have been disseminated through standard scientific channels, including peer-reviewed publications, preprints, conference presentations, invited lectures, workshops organized by my lab, and direct exchange with the scientific and clinical community. Ultimately, all methods, antibody panels, workflows, computational tools, data, etc. are accessible to the scientific community. Importantly, the project also contributed to training students, postdocs, clinicians, and external researchers in highly multiplexed tissue imaging and spatial single-cell analysis.
Beyond academic dissemination, the project has supported translational exploitation. A key route has been the spin-out of a company developing multiplexed tissue imaging diagnostics. This company aims to translate highly multiplexed tissue imaging into clinically approved diagnostic assays and to support biomarker development, patient stratification, and clinical trial analysis. Thus, the project has generated not only new biological knowledge, broadly useful technologies, and computational methods, but also a clear path toward clinical and commercial application in precision oncology.
