C12 - Endothelial cell-centered neighborhood analysis highlights vascular niches of the small intestine and colon

Introduction: : The cardiovascular and lymphatic systems rely on the endothelial cells (ECs) that line the inner surface of vessels to regulate the transport and exchange of nutrients, waste, immune cells, and other factors to surrounding tissues. As all tissues require vasculature to survive, ECs exist in highly varied microenvironments and interact with many different cell types throughout the body. Characterizing the microenvironment of ECs in relation to surrounding tissues, in conjunction with patient metadata, can provide novel insights into the organization of vasculature in different tissue types and its potential clinical relevance.
Recent advancements in spatial omics technologies have enabled molecular profiling at a single-cell or subcellular resolution within the spatial context of the tissue. Previous work has made it possible to investigate cell-cell interactions and cellular organization at different scales in a tissue. However, such global analyses of data from the whole tissue tend to average similar cell type interactions together and mask more nuanced and potentially meaningful relationships. In this work, we demonstrate a cell type-focused approach and investigate ECs in the small intestine and colon. These results indicate the value of cell-centered neighborhood analysis in identifying more specific cell-cell interactions or significant correlations with patient metadata.

Materials and

Methods: : Data: This work uses 64 tissue sections imaged using CODEX, 32 from the small intestine and 32 from the colon, collected from 12 donors and paired with patient metadata. Images were annotated in a previous publication, from which all data was downloaded.
Endothelial cell-centered neighborhood analysis: For each EC, windows, i.e., the 100 nearest neighbors based on spatial coordinates, were computed. Cell type counts within these windows were clustered using k-means clustering, producing neighborhoods. These were labeled based on cell type composition of their centroids (Figure 1a). An initial set of 20 neighborhoods was refined by merging similar clusters, resulting in a final set of 11 distinct EC-centered neighborhoods. Radial neighborhood analysis was also conducted by varying the nearest neighbors (5, 10, 30, 50, or 100) and assigning neighborhoods in the same manner (Figure 1b).
Analysis with donor metadata: Comparing EC neighborhood percentages between the small bowel (SB) and colon (CL), a paired t-test was conducted for each neighborhood using each patient’s neighborhood composition in the SB and CL as data pairs. Grouping the data by tissue type, EC neighborhood percentage was defined as the proportion of cells that belonged to each neighborhood type. To investigate correlations with patient BMI, linear regression was performed using neighborhood percentages across all regions per patient. For assessing the effect of history of hypertension, an unpaired t-test was conducted between the hypertension and no hypertension groups for each neighborhood. EC neighborhood percentage was calculated across regions per patient, with each donor considered a replicate.

Results, Conclusions, and Discussions:: In this work, we adapt neighborhood analysis techniques and focus on ECs to study their specific microenvironments in the intestine. We identified 11 distinct endothelial neighborhoods, some that are consistent with common structures of the intestine, e.g., follicle, and neighborhoods observed in global neighborhood analysis (Figure 2a, b). The identification of additional EC-specific neighborhoods, visualized as unique UMAP clusters, indicates that finer-grained relationships between vasculature and surrounding tissue can be uncovered within the same tissue units (Figure 2c, d).
In radial analysis, changing the number of nearest neighbors considered captures variation due to increasing distance from the center EC (Figure 2e). Notably, in the Cycling TA neighborhood, the relative percentage of ECs decreases dramatically as the number of nearest neighbors increases (Figure 2f). This suggests that tissue surrounding transit-amplifying (TA) cell-enriched areas tend to be more hypoxic, which could be important in the multipotency and differentiation of TA cells.
Comparing cell type percentages in the SB and CL, the SB is enriched with neighborhoods that are more pro-inflammatory due to increased prevalence of cells like CD4+ T and M1 macrophages. The CL has enriched with neighborhoods associated with structural/restorative roles, with higher amounts of smooth muscle, M2 macrophage, stromal, and cycling TA cells (Figure 3a, b). We also observed associations between metadata and neighborhood prevalence. Although BMI was previously found to correlate with M1 macrophage levels, no linear relationship was found between BMI and the percentage of immune EC-centered neighborhoods (Figure 3c). For hypertension, significance was found only in the prevalence of the Immune Enriched neighborhood (Figure 3d). This difference, visualized in the ileum, indicates lower M1 macrophage or CD8+ T cell presence and disruption of normal immune behavior (Figure 3e).
Overall, the cell-centered paradigm presented here enables high-granularity characterization of cellular niches across tissue types and disease states. As spatial datasets increase in accessibility/ complexity, analysis techniques at multiple scales and with different focuses are necessary to maximize the insight derived from such rich data. Thus, this cell-centered neighborhood analysis method is a robust approach for studying specific cell types, their microenvironments, and their relation to clinical phenotypes.

Acknowledgements and/or References (Optional)::