Unraveling Renal Transplant Rejection: How Can We Measure Intragraft Cell–Cell Interactions?

Recent advancements in tissue analysis, particularly through the widespread adoption of high-plex and omics techniques for protein, transcript, and metabolite profiling of tissue sections, have greatly enhanced our understanding of the cellular interactions within transplant allografts.1 The distribution and composition of inflammation and injury within allografts are well recognized indicators of different pathological processes, each having important implications for patient management.2 By applying novel spatial profiling and bioinformatic methods to transplant biopsies, we can define the fundamental biological processes occurring within allografts and provide clinically relevant insights that experimental models cannot faithfully replicate. In this commentary, we aim to highlight the methods used by Hira et al3 to mathematically interpret the cell interactome during rejection. The study in this issue of Transplantation3 uses high-plex immunofluorescence analysis of 40 protein targets to delineate the cellular interaction networks present during T cell–mediated rejection (TCMR) and antibody-mediated rejection (AMR) after renal transplantation. Notably, AMR took place early posttransplantation (median 17.5 d) with the presence of preformed anti-HLA donor-specific antibodies in 7 of 12 patients. The study utilized an informatics workflow which included cell segmentation, cell type annotation, and the application of 2 interaction methods. In general, the first step in image analysis is cell segmentation, whereby cell boundaries are defined through methods such as nuclear dilation. Each nucleus is identified, and the cell boundary expands from this central point until it reaches another cell or a maximum radius. Cells are then characterized by their immunofluorescent signal intensity, similar to flow cytometry. Cell phenotypes are defined by applying dimensionality reduction methods (such as Uniform Manifold Approximation and Projection, as used in single-cell sequencing for example) and cell clusters manually annotated based on their expression of lineage markers. This transforms the image into thousands of cells defined by their relative protein expression, allowing for measurement and comparison of cell composition. In the current study, cell neighborhood analysis was then used to understand cell interactions by examining their spatial proximity.4 Neighborhoods were defined by a cell of interest and its 14 nearest neighbors, which are then clustered to identify recurring patterns within the tissue samples. Eight distinct neighborhood types were identified, composed of various parenchymal and immune cell interactions. For instance, a prominent neighborhood in TCMR biopsies (NC0) was found to be enriched in CD4+ T cells, CD20+ B cells, CD8+ T cells, and macrophages. In contrast, a separate immune neighborhood (NC7) was defined by proximity between CD8+ T cells and macrophages, 30% of which exhibited an activated inflammatory phenotype. This interaction was observed in both types of rejection and may contribute nonspecifically to alloresponses. Complementary data from a recent cellular indexing of transcriptomes and epitopes-sequencing study, which annotated physically interacting cells resisting tissue dissociation protocols, identified 3 stable immune cell complexes during rejection.5 These complexes are seen across rejection types and include plasmacytoid-like dendritic cell:B-cell complexes, NK/T-cell:myeloid cell complexes, and NK/T-cell:endothelial cell complexes. The enzymatic choices used in tissue dissociation could bias these findings; however, they provide valuable insights into phenotypic differences between the T cells in NC0 and NC7. Further analysis defined spatial relationships between adjacent cells through pairwise cell interaction analysis. This method determines the frequency of contact between each cell type. In TCMR, a significantly greater proportion of interactions occurred between T lymphocytes, whereas AMR was characterized by a higher frequency of macrophage interactions with podocytes, endothelial cells, and stromal cells. Macrophages in the stromal and endothelial cell neighborhood (NC5) exhibited a more proinflammatory phenotype (enriched in Ki67, GzmB, HLA-DR, and CCR7) than those in the same location during TCMR. Both the NC5 neighborhood and the number of interactions between macrophages and endothelial cells correlated with biopsy scores for peritubular capillaritis and glomerulitis, consistent with the histological classification of AMR by microvascular inflammation. The distribution of monocytes, notably CD14+ monocytes in interstitial and tubular regions and FcγRIII+ monocytes in vascular and glomerular areas during mixed or AMR, highlights the heterogeneity of immune infiltrates. Although Hirai et al have focused largely on activated macrophage phenotypes, other studies have used multiplexed histological profiling to highlight different cell subsets identified in single-cell sequencing datasets. In a study from Lamarthée et al,6 multiplexed immunofluorescence was used to show an association between FcγRIII+ NK cells, neutrophils, FcγRIII+ monocytes, CD14+ monocytes, and histologically defined inflammation severity. Rather than explore cell–cell interactions, they assigned cells to renal compartments such as glomerular, large vascular, and tubular areas. Curiously, the distribution of monocytes was not uniform with CD14+ monocytes associating with interstitial and tubular regions in both AMR and TCMR, whereas FcyRIII+ monocytes were observed chiefly in the vascular and glomerular areas during mixed or AMR. Similar to multiplexed immunofluorescence-based imaging, imaging mass cytometry quantifies 30–40 proteins but does this simultaneously by using heavy metal tagged antibodies, without sequential bleaching and staining. Recently, in Science Advances, Barbetta et al7 analyzed immune cell composition in acute and chronic liver transplant rejection. In addition to pairwise cell interaction and neighborhood analyses as above, they applied a distribution of distance analysis to establish any changes in immune composition relative to the endothelium. This revealed a relationship between rejection scores and endothelial inflammation. Proinflammatory T-cell subsets and classical monocytes resided near endothelial cells, whereas CD16+ M1 and CD16+ M2 macrophages were distributed throughout the tissue. The composition of immune infiltrates during rejection is now well established through techniques such as multiplexed immunofluorescence and single-cell transcriptomics. However, understanding how immunosuppressive control of these tissue-infiltrating cells is compromised and how destructive alloresponses develop will be crucial for effective tissue analysis and for grasping the nuances of borderline, mixed, and ambiguous transplant pathologies. With the advent of emerging single-cell in situ molecular analytics, our comprehension of tissue responses has reached unprecedented levels. While identifying cell phenotypes is undoubtedly important, understanding their interactions in situ is likely to be key to determining where and when to intervene during pathogenesis. We are undoubtedly at a pivotal juncture in the field, with spatial profiling techniques becoming increasingly accessible and digital pathology analysis advancing rapidly. It is highly likely that markers indicative of specific pathological processes will soon be identified, validated, and incorporated into current clinical scoring criteria. Collaborative efforts among centers specializing in these areas are likely to accelerate this progress, providing essential training and validation datasets across institutions.