Three regions – 6, 8, and 11 – had epithelial composition and programs that suggested advanced malignant-like characteristics, each highlighting a potentially different mechanism for tumor progression (Figure 5G and H). These three ‘malignant-like regions’ were enriched (vs. all other regions) in stem cell, Wnt signaling, and angiogenesis programs (16, #4, and #14; FDR = 9.6 10–21, 8.1 10–10, 3.7 10–10, two-sided Welch’s t-test on CLR transformed compositions) and depleted of normal epithelial programs (#5, #8, and #10; FDR = 5.5 10–12, 1.6 10–12, 4.2 10–10, Figure 5I). Furthermore, the malignant-like regions were enriched in immune cells, including monocytes-macrophages (FDR ≤ 1.5 10–5; excluding Mac02 [Lyve1+]), T cell subsets TNK02 (Th1/Th17), TNK05 (GdT/Il17+), TNK06 (Treg), TNK08 (proliferating T) (FDR = 2.4 10–3, 2.2 10–4, 1.7 10–3, 9.6 10–4; two-sided Welch’s t-test on CLR transformed compositions), infiltrating granulocytes (FDR ≤ 9.7 10–3), and mast cells (FDR = 1.4 10–2), suggesting an ongoing immune response (Figure 5J). However, each one of the three regions had a different epithelial program composition, suggesting that in each type of region, there is a different dominant pathway/feature that may drive tumor progression (Figure 5C, Supplementary file 2).
Region 6 was characterized by an inflammatory and angiogenic multicellular community, with epithelial and immune cells expressing inflammatory programs, endothelial cells and monocytes connected in a pro-angiogenic circuit, and pro-invasive genes expressed by both endothelial and immune cells (Figure 5B and C). Specifically, region 6 was distinctly enriched for proliferation (programs 3 and 11; FDR = 2.2 10–14, 1.2 10–15, two-sided Welch’s t-test on CLR transformed compositions) and inflammatory epithelial programs (programs 6 and 7; FDR = 9.0 10–7, 2.0 10–11), and its non-epithelial compartment was correspondingly enriched for genes from inflammatory pathways, including the response to TNF, IL-1, and IFNγ (FDR = 3.1 10–4, 2.8 10–3, 4.9 10–6, two-sided Fisher’s exact test in GO term enrichment), and chemotaxis of monocytes, neutrophils, and lymphocytes (FDR = 1.5 10–3, 3.2 10–10, 1.0 10–2, Supplementary file 2), suggesting recruitment of inflammatory cells from the circulation or other parts of the tissue. Region 6 was also enriched for collagen binding genes and collagen-containing extracellular matrix (ECM) genes (FDR = 1.4 10–2, 7.6 10–5, two-sided Fisher’s exact test in GO term enrichment, Supplementary file 2), which are important for migration and invasiveness (Winkler et al., 2020). These include Sparc, expressed mainly by endothelial cells and fibroblasts in our data, known to promote CRC invasion (Drev et al., 2019); and Ctss, a peptidase expressed by T cells and monocytes-macrophages that promotes CRC neovascularization and tumor growth (Burden et al., 2009). Finally, gene expression patterns in endothelial cells and monocytes in region 6 suggested active angiogenesis through a multi-cellular feedback loop, with enriched numbers of vascular and lymphatic endothelial cells expressing immune-attracting chemokines (Cxcl9) and adhesion molecules (e.g. Chd5, Mcam), monocytes expressing proangiogenic factors that induce proliferation of endothelial cells (e.g. Mmp12), and monocytes and macrophages expressing Ctsd, which increases tumorigenesis in CRC models (Basu et al., 2019; Figure 5—figure supplement 1E).
Region 8 was enriched for deep crypt cells (program 13; FDR = 1.7 10–14, two-sided Welch’s t-test on CLR transformed compositions), reminiscent of the normal stem cell niche in normal colon, an epithelial innate immune program (program 1; FDR = 3.4 10–100) expressed by secretory cells in AV and AKPV lesions, and plasma and B cell activity. Unlike the canonical (normal) deep crypt region (region 5), which is enriched for MHCII expression (program 18; FDR = 8.6 10–28), this region was depleted for the program’s expression (FDR = 2.1 10–26), which may indicate an earlier stem cell-like state (Biton et al., 2018), or a decoupling of the cell cycle and MHCII programs (which are coupled in normal ISC differentiation, and allow a cross talk with T cells to modulate T cell differentiation) (Figure 5B and C). The region’s non-epithelial compartment was enriched for B cell activation and BCR signaling genes (FDR = 4.0 10–4, 1.9 10–3, two-sided Fisher’s exact test in GO term enrichment, Supplementary file 2). This may be related to B cell function in protection from lumen antigens (Spencer and Sollid, 2016) or to tertiary lymphoid structures (TLS), which are correlated with clinical benefits in cancer patients (Sautès-Fridman et al., 2019). Notably, Epi05 (dysplastic secretory-like) enriched in Region 8 (Figure 5B) expressed higher levels of inflammatory genes and immune chemokines (e.g. Ccl9, Ifitm3) compared to normal counterparts, Epi04 (secretory; Figure 3C, Figure 3—figure supplement 1C and E), and may thus promote the formation of this region. We validated the presence of TLS-like structures in association with deep crypt secretory cells in AV lesions using multiplex RNA analysis (Cartana; Figure 5—figure supplement 2), showing that the dominant population of B cells is accompanied by monocyte-macrophages and T cells characteristic of TLSs, as well as the expression of Reg4 and Muc2, deep crypt goblet/secretory cell markers.
Region 11 was populated by cells expressing the Wnt signaling pathway program (4, FDR = 9.3 10–13, two-sided Welch’s t-test on CLR transformed compositions), with several lines of evidence supporting an active epithelial to mesenchymal transition (EMT) in this region. Epithelial cells in region 11 were enriched for the expression of mesenchymal genes, including Vimentin (Mendez et al., 2010) (Vim, FDR = 7.4 10–245, one-sided Fisher’s exact test), Prox1 (Lu et al., 2012) (FDR = 3.7 10–153), and Sox11 (Oliemuller et al., 2020) (FDR = 7.7 10–224) (Figure 5—figure supplement 1F), as well as for EMT signatures from a mouse model of lung adenocarcinoma (Marjanovic et al., 2020) (FDR = 1.5 10–142, two-sided Mann-Whitney U test) and from human head and neck squamous cell carcinoma tumors (Puram et al., 2017) (FDR = 6.9 10–55, two-sided Mann-Whitney U test). This is consistent with the role of Wnt signaling in promoting EMT and a mesenchymal phenotype in CRC, breast cancer, and other epithelial tumors (Schwab et al., 2018; DiMeo et al., 2009). Region 11 non-epithelial cells also expressed genes encoding MHC-I binding proteins (FDR = 4.6 10–2, two-sided Fisher’s exact test in GO term enrichment, Supplementary file 2) and actin cytoskeleton, filament, and binding proteins (FDR = 3.1 10–6, 7.7 10–3, 2.7 10–10). Organization of the cytoskeleton affects migration, adherence, and interaction of lymphocytes with antigen-presenting cells (Penninger and Crabtree, 1999). Notably, region 11 also concentrated at a more distal part of the tissue at ~900 μm from the muscularis, suggesting an outgrowth of the tissue towards the lumen (Figure 5—figure supplement 1C).
Non-epithelial cells formed two cellular hubs in the malignant-like regions (6, 8, and 11; Figure 5—figure supplement 1G): An endothelial-fibroblast hub, detected in all three regions, and an immune hub with B cells, TNK cells, monocytes, and macrophages, which was prominent in inflammatory region 6, weaker (less spatially correlated) in region 8 (but validated in situ), and not correlated in region 11. Thus, activation of an immune response is reflected by close proximity between immune cells. We further characterized the organization of the vascular niche using our multiplex in situ RNA data, finding that while neighbors of the Pdgfrb-expressing pericytes are mainly other Pdgfrb-expressing pericytes and endothelial cells, Pecam1-expressing endothelial cells appear self-enriched next to themselves at cellular scale distances and close to Pdgfrb-expressing pericytes for larger distances (Figure 5—figure supplement 1H).
Overall, three multicellular community regions were enriched in AV lesions: (1) inflammatory epithelial regions with endothelial cells and monocytes expressing angiogenesis, inflammation, and invasion programs; (2) epithelial stem-like regions, associated with plasma and B cell activity; and (3) regions with epithelial to mesenchymal transition (EMT) and Wnt signaling dysplastic cells. Each region highlights different processes that modulate tumorigenesis or invasion, and the three regions co-exist in the same tumor at different spatial locations.