Defining and modeling dynamic spatial heterogeneity within tumor microenvironments

Abstract

Many solid tumors exhibit significant genetic, cellular, and biophysical heterogeneity which dynamically evolves during disease progression and after treatment. This constant flux in cell composition, phenotype, spatial relationships, and tissue properties poses significant challenges in accurately diagnosing and treating patients. Much of the complexity lies in unraveling the molecular changes in different tumor compartments, how they influence one another in space and time and where vulnerabilities exist that might be appropriate to target therapeutically. Recent advances in spatial profiling tools and technologies are enabling new insight into the underlying biology of complex tumors, creating a greater understanding of the intricate relationship between cell types, states, and the microenvironment. Here we reflect on some recent discoveries in this area, where the key knowledge and technology gaps lie, and the advancements in spatial measurements and in vitro models for the study of spatial intratumoral heterogeneity.

Figure 1

Spatial heterogeneity within the tumor microenvironment. Schematic illustrating examples of distinct immune and biomechanical processes occurring at the leading edge (LE) and core (TC) of solid tumors. 1. Deposition of collagen I occurs at the LE, linearly arranged and perpendicular to the tumor mass aiding in cancer cell invasion. 2. ECM is deposited by various cells, including cancer cells and cancer-associated fibroblasts (CAFs). Neutrophils also deposit neutrophil extracellular traps (NETs) which interact with ECM proteins, promoting local tumor cell invasion. 3. When compared to the TC, the LE has increased stiffness and epithelial-to-mesenchymal transition (EMT). Whereas the TC has increased hypoxia, interstitial fluid pressure (IFP), and compression. 4. This increased hypoxia and solid stress can lead to tumor cell apoptosis, driving the formation of a necrotic core. 5. Tumor cells within the LE recruit macrophages and CAFs to the tumor site via chemotaxis, where tumor-associated macrophages (TAMs) are polarized towards an M2 phenotype expressing CD68, CD206, and CD163 markers. 6. TAM phenotype is influenced by proximity to cancer cells, with high proximity correlating with CD68+ IRF8+ M1-like TAMs. 7. Collagen IV primes cancer cells to release more CXCL10. CXCL13 and CCL2 cytokines resulting in increased Th1 differentiation, T cell apoptosis, and T cell exhaustion. 8. The dense ECM at the LE acts as a barrier for infiltrating T cells. 9. TAMs also act as a barrier, preventing the infiltration of T cells. 10. EMT at the LE promotes metastasis as mesenchymal-like cancer cells intravasate into blood vessels and extravasate to secondary sites. Immature, leaky vasculature is present throughout the TME and contributes to the increased IFP observed.

Figure 2

Methodologies for the study of spatial heterogeneity (a) Modeling spatial heterogeneity in vitro. Tumor components: tissue sections, organoids, and spheroids are commonly used in vitro to establish 3D tumor compartments. Ex vivo culture of tumor tissue samples retains a heterogenous population of cancer cells in addition to native TME components. Self-organized 3D spherical cell aggregates (spheroids) can be established from single or multiple cell lines. Organoids are established from tissue samples, iPSCs, ESCs, and ASCs through a process of enzymatic digestion and 3D culture with growth factors retaining a heterogenous population of cancer cells. Organoids and spheroids have gradient of nutrients, oxygen, and other soluble factors and often have a necrotic core. Other TME components that can be incorporated into 3D in vitro culture include immune cells, CAFs, ECM, vasculature, and secretory components. ECM components: ECM components can be sourced from decellularized ECM, artificial hydrogels, and isolated ECM-derived proteins (such as collagen, elastin, and hyaluronic acid). Stiffness can be regulated through crosslinking, increasing concentration of ECM components, or incubation of collagen with riboflavin phosphate (RFP) followed by blue light exposure. Bioprinting: layer-by-layer deposition of bioinks with predetermined and tunable ECM and cellular components for the creation of 3D scaffolds and TMEs. The example shows cancer cells seeded at the core with an ECM component and then surrounded by endothelial cells to establish components including TC, LE, and surrounding tissue. Over time an oxygen gradient forms, leading to hypoxia at the core. Microfluidics: integration of bioprinting and microfluidics allows for the introduction of flow-derived stress to recapitulate the dynamic TME. Microfluidic devices can also incorporate 3D cancer components such as organoids embedding them within an ECM and integrating with a vascular network to further mimic the TME and the LE-TC dynamics. (b) 3D spatial ‘omics examples. I) 3D Imaging mass cytometry (3D-IMC) in which serial tissue sections are stained with metal labeled antibodies. The surface of the sample is ionized turning molecules into ions allowing for the spatial detection and quantification of different ions. A 2D image is produced for each slide, images are then aligned and stacked to generate a 3D reconstruction [61]. II) CODA has been used to create 3D renderings of hematoxylin & eosin-stained tissue slides, serial images are aligned and stacked to create a 3D rendering, and machine learning is applied for the identification of several pathologically relevant TME compartments including cancer cells, epithelium, smooth muscle and nerves, fat, collagen, islets of Langerhans and acini [63]. III) Spatially resolved transcript amplicon readout mapping (STARmap) is a technique to create 3D renderings of thick tissue sections. Specific RNA molecules within the tissue are labeled with signal amplification by exchange reaction (SNAIL) probes. SNAIL probes facilitate amplification of RNA, during signal amplification unique fluorescent barcodes are introduced, specific to each target RNA. A hydrogel precursor solution is infused into the tissue and polymerized in situ to create a cross-linked hydrogel network that embeds the tissue. During polymerization functional groups within the hydrogel bind to DNA amplicons fixing them in place, ensuring they remain in their original spatial position. Tissue is cleared and imaged layer by layer to detect spatial location of fluorescent signals. 2D images are then aligned and compiled into a 3D map [62]. IV) 3D imaging of solvent-cleared organs (DISCO) by mass spectrometry (DISCO-MS), whole tissue is fixed and cleared using the DISCO method and imaged. Regions of interest (ROI) are identified via deep learning algorithms. Ultra-high sensitivity mass spectrometry is used to perform proteomics at these ROIs, resulting in a spatially resolved proteome analysis of tissue within whole organs [68].