As the capabilities of spatial biology evolve, so do the questions we ask ourselves as researchers. The field of spatial biology has seen rapid developments in theoretical frameworks and instrumentation, all of which revealing a clear path towards the innovation of novel perspectives allowing us to think differently about how we study tissue and cancer biology.
Laying the Foundations for Neighborhoods
Dr. Garry Nolan, Rachford and Carlota A. Harris Professor in the Department of Pathology at Stanford University, and co-inventor of the PhenoCycler™ (formerly CODEX) technology, is a well-recognized immunologist. In a September 2020 Cell publication, he and his colleagues developed the seminal concept of “cellular neighborhoods”1. A concept that focuses on the dynamic nature of the immune tumor microenvironment in which the combination of immune cell types, location, and functional orientation leads to a tumor rejecting- or tumor-promoting environment. Characterizing tumor-immune phenotypes in detail has long been done with other techniques, but tumor-immune phenotypes in the context of spatial orientation had not been extensively investigated. So, Dr. Nolan and his team developed a technological and analytical framework to systematically describe the components of complex tissue types and their coordinated behaviors, a concept known as cellular neighborhoods. Using PhenoCycler they investigated cellular organization, functional states, and communication between tissue regions of patients with advanced-stage colorectal cancer. To simplify how we think about cellular neighborhoods we can liken this concept to ourselves and how we act in different environments. When we find ourselves in different environments we tend to behave differently, we are the same human being, but will act differently at work versus at a rock concert or when we are having friends over for dinner. The concept is the same for cells – in that the cells’ phenotype will remain the same, however, its behavior and interactions will change based on its contextual phenotype.
Figure 1:Neighborhood structures in CLR patients are compartmentalized with a separate “tumor compartment” and “immune compartment,” while in DII patients, tumor and immune compartments were mixed. The separate compartments correlate with a higher survival rate, suggesting that tumor cells may interfere with anti-tumor effects of immune cells in the mixed model.1
This framework demonstrates how spatial behavior within the immune tumor microenvironment can be used to stratify patients into a particular group, for example potential responders and non-responders to a given therapy. This spatial biology algorithm can be used to demonstrate how the organization and interactions of these neighborhoods predict patient outcomes, underscoring the importance of spatial context in developing prognostic biomarkers. The cellular neighborhood algorithm works by assigning a cell located at any given position as the central position, then the Euclidean distance from the central cell and the nine nearest neighboring cells is measured. Going back to our analogy of how we behave in different environments, we can consider ourselves the central cell, then those who surround us will represent our cellular neighborhood, for example we are surrounded by nine other fans at a rock show. The ideal method for conducting cellular neighborhood analyses begins with spatial phenotyping – the simultaneous capture of phenotypic compositions, quantities, and spatial complexities of a tissue. Spatial phenotyping is enabled through ultrahigh-plex imaging in which a tissue is labeled with an antibody panel. The tissue is stained with fluorophore conjugated antibodies then imaged through a cyclical process. Once the tissue is imaged at cellular and subcellular resolution, these imaging files are then ready to be analyzed. When multiple neighborhoods are identified, these regions are then clustered based on the composition of their respective microenvironment. The resulting group of cells and neighborhoods can then be interrogated and explored. This information can be used to identify how neighboring cells interact may indicated how a given neighborhood is impacting the central cell’s function.
Since the introduction of the cellular neighborhood concept, we have seen several cases were this algorithmic approach was used to stratify patients with very promising results2,3,4. This concept has been explored in various tissue types and patient populations including cancers of the colon, bladder, intestine, and skin1,2,3,4. However, we could argue that a limiting factor associated with these types of studies is the depth of a given antibody panel. Typically, high-plex antibody panels are maxed out when investigators must stain with foundational markers for cell lineages, checkpoints, and structural detail which may require up to 50 biomarkers. If we are limited by the markers of interest in our panels this means we can only achieve a limited picture of both, the constituents, and the behaviors of cellular neighborhoods. Ultrahigh-plex spatial phenotyping (defined as 100+ plex) present the opportunity to look several layers deeper. With the power of ultrahigh-plex spatial phenotyping, cellular neighborhoods will become more nuanced and may show how cells act in the presence of a tumor differently than we once thought.
Defining Spatial Frameworks and Conceptual Hierarchies
Ultrahigh-plex spatial phenotyping with powerful instrumentation preserves the native context of the tissue environment, offering researchers a tool to simultaneously study multiple facets of tissue biology ranging from metabolomics, proliferation, and stress responses to the immune microenvironment. Coupled with the identification of distinct cellular lineages and structural context, ultrahigh-plex panels build upon our understanding of key cellular players to add further insights into the functional role of cellular neighborhoods and, ultimately, the complex coordination of tissue schematics.
Figure 2 Simultaneous identification of 4 distinct tumors of oropharyngeal squamous cell carcinoma (shown here in blue, purple, green and orange) with varying abundance of immune, proliferating, epithelial, and vascular cell types compared to the normal lymphoid area (red) and the normal squamous epithelium (black) facilitated by an ultrahigh-plex spatial phenotyping assay.
In the budding field of spatial biology, we have seen the power provided by a conceptual framework like cellular neighborhoods, but what if we could take things a step further? Tissue schematics are another conceptual framework developed by Dr. Garry Nolan that builds upon cellular neighborhoods. To think about schematics simply, they define how simple components assemble to create more nuanced interactions. Let us go back to our rock concert analogy. As discussed above, we act differently at home versus at a rock concert (different neighborhoods) but the schematic represents how the different interaction we may have at a rock concert could influence us. At the rock concert we may experience different lighting, perhaps a ballad is playing, or maybe an anthem accompanied by a few too many libations. We find ourselves within the exact same neighborhood but acting completely differently based on what is going on around us. Cellular neighborhoods aim to investigate how different environment influences a cell i, while tissue schematics outline how the environmental assembly within a neighborhood has a nuanced but large effect on cellular interactions. Using high parameter tissue imaging data, the tissue schematic approach begins to become possible These data are then used to assess how complex functions are achieved via the orchestration of many simpler components. The tissue schematic framework leverages our increasing capabilities to look at tissue through a more intricate lens (pun intended), a more detailed picture of how cells organize and interact allows for the construction of hierarchical structures that mediate tumor progression and open the doors to more opportunities for discovering therapeutic targets. We expect that with the advent of new technologies and techniques comes the potential to dive even deeper.
Breaking Through a Bottleneck
When we think about breaking past a particular barrier or bottleneck the natural thought progression is to think about what the next roadblock will be. Experiments continue to become more elegant, improving upon spatial resolution, plex, complexity, all at a reduced cost. These types of innovations are incredible, but they also lead to downstream problems such as massive data bottlenecks. More data means more problems, but the biologists, bioinformaticians, and computer scientists have solved this problem many times before and continue to maintain significant pace. Several promising computational tools such as deep generative modeling and psuedotime analysis may hold keys to help deal with these interesting problems6. Cancer and tissue biology are dynamic, unique, and incredibly complex, but we continue to arm ourselves with the capabilities necessary to conduct truly powerfully research. The tools are available today, so it is time to stop thinking about what if and to start thinking about what is next.
- Schürch, et al. Coordinated Cellular Neighborhoods Orchestrate Antitumoral Immunity at the Colorectal Cancer Invasive Front. 2020; 182, 1341–1359
- Philips, et al. Immune Cell Topography Predicts Response to PD-1 Blockade in Cutaneous T Cell Lymphoma Nat Commun. 2021 Nov 18;12(1):6726
- Gouin, K. H. 3rd et al. An N-Cadherin 2 expressing epithelial cell subpopulation predicts response to surgery, chemotherapy and immunotherapy in bladder cancer. Commun. 12, 4906 (2021).
- Hickey, et al. High Resolution Single Cell Maps Reveals Distinct Cell Organization and Function Across Different Regions of the Human Intestine. [preprint]. 2021, Nov 25.
- Bhate, et al. Tissue schematics map the specialization of immune tissue motifs and their appropriation by tumors, Cell Systems 2021 Oct 14; 13(2), 109-130
- Zeng, Z., Li, Y., Li, Y. et al. Statistical and machine learning methods for spatially resolved transcriptomics data analysis. Genome Biol 23, 83 (2022).
Author: James DeRosa, MPH