Spatial resolved portrait of UC ecosystem
In order to dissect the spatial heterogeneity of the UC microenvironment while preserving the architecture of the UC ecosystem, we established a 39-marker IMC panel for UC ecosystem (Figure 1A,B), based on published protocols for intestinal tissue and cancer 19, 20, 21, 22. Our 39-marker IMC panel covered markers for epithelial, endothelial, stromal cells, and various immune cells; the cytokines IL-1β, TNF-α, and IL-6; the proliferation marker Ki-67; and the apoptotic marker cleaved caspase-3 (Supplementary Table S1), which allowed highest level of multiplexing of IMC so far.
We first analyzed colon tissues of healthy donors, which revealed typical structural markers, such as Collagen I and CD31, immunological markers, such as CD3 and CD20 (Figure 1B), and cytokines, such as TNF-α and IL-1β after staining for the 39-marker panel (Figure 1B).
Distinct histological features, including intestinal crypts revealed by Pan-keratin staining (cyan) of epithelial cells and smooth muscle layer revealed by αSMA staining (yellow) of smooth muscle cells, were consistent with those determined by hematoxylin and eosin (H&E) staining but were resolved in greater detail (Figure 1C).
We detected major cell types, including CD31+ endothelial cells, αSMA+ smooth muscle cells, and Collagen I deposition in the extracellular matrix (Figure 1C). Immune cell types detected included B cells (CD20+), T cells (CD3+). Notably, T cells and B cells were highly enriched in the intraepithelial space and gland surrounding area (Figure 1D). Resident macrophages (CD163+ CD68+ CD16+ CD11blow), and inflammatory macrophages (CD163- CD68+ CD16- CD11b+) 22 could be detected mainly underneath the epithelial cells (Figure 1E).
Furthermore, to characterize the cellular heterogeneity of the UC microenvironment while preserving gross structural features, formalin-fixed, paraffin-embedded (FFPE) human tissue samples from 9 healthy donors and 23 UC patients with different Mayo grading scores (Supplementary Table S2) were processed and stained with 39 metal-conjugated antibodies, before scanning by IMC. The scanned images were subjected to cell segmentation using the established bioinformatic tool, histoCAT, developed by Schapiro et al. 23 (Figure 1A). The expression of markers in each cell was quantified, and 26 cell clusters were identified by PhenoGraph (Figure 1F), which were then annotated based on the expression of key markers.
CD45- Pan-keratin+ cells indicated epithelial cells (Supplementary Figure S1A), endothelial cells were defined by CD31 and Vimentin staining (Supplementary Figure S1B), smooth muscle cells were identified by αSMA+ cells (Supplementary Figure S1C), and CD68+CD163+CD11blow and CD68+CD163-CD11b+ cells were defined as resident and inflammatory macrophages, respectively (Supplementary Figure S1D), which was consistent with scRNA data (Supplementary Figure S2). T cells and B cells were concentrated in the intraepithelial space (Figure 1D), thus, T cell, B cell, and epithelial markers overlapped due to the resolution limit of IMC (Supplementary Figure S1E). However, T cells and B cells can be separated from epithelial cells based on CD45 expression (Supplementary Figure S1E). Overall, the T cell, B cell, and NK cell markers showed co-staining in clusters 6, 7, 9, 18, and 22, which were defined as lymphocytes (Figure 1F). FoxP3+ CD25+ cells were defined as regulatory T cells (Supplementary Figure S1F). In total, the 26 cell clusters included 7 from epithelial cells, 6 from endothelial cells, 1 from smooth muscle cells, 11 from immune cells, and 1 from unstained cells. Their frequencies in different patients are shown in Figure 1G.
In summary, tissue architecture, major stromal cell types, diverse immune cell populations, and cytokine production in the human colon microenvironment could be observed with our 39-marker panel by IMC and our IMC results revealed a detailed landscape of UC ecosystem, including epithelial cells, stromal cells, and various immune cells.
MDR occurs only for resident macrophage during UC progression
Next, we wanted to characterize cellular changes during UC progression. Firstly, 9 healthy donors and 23 UC patients’ samples were separated into 3 major classes based on their cellular frequencies (Figure 2A,B). All samples from the 9 healthy donors were grouped together (green) whereas, samples from UC patients were separated into mild (blue, Mayo grade 1) and severe groups (red, Mayo grade 2-4), Mayo grade was determined by the patients’ colonoscopy (Figure 2C).
Further analysis showed that epithelial cells decreased from healthy donors to severe UC patients (Figure 2D; clusters 4 and 23), perhaps due to increased tissue damage, which was also confirmed by flow cytometry analysis of colon biospecimen (Figure 2E). While lymphocytes increased (Figure 2F; clusters 6, 18, and 22), indicating increased lymphocyte infiltration due to inflammation. Since the resolution of IMC for lymphocytes in UC region was poor, we also performed flow cytometry analysis for UC samples, and we noticed heavy infiltration of T cells and B cells, especially for B cell and CD4+ T cells (Figure 2G).
Furthermore, we also noticed that the macrophage disappearance reaction occurred only for resident macrophage during UC progression (Figure 2H; cluster 11, 14, and 24). While inflammatory macrophages (Figure 2J; cluster 25) increased when UC developed, suggesting a population switch occurred in the macrophage system. Macrophage disappearance reaction (MDR) was confirmed in UC patients’ samples by FACS. CD68+ CD11b- resident macrophages almost disappeared in UC patients while CD68+CD11b+CCR2+HLA-DR+ inflammatory macrophages increased significantly (Figure 2K).
Both IMC and FACS data revealed a change in cellular components with UC progression, especially for the macrophage system, where inflammatory macrophages replaced resident macrophages.
Macrophage replacement kinetics during UC
To analyze the macrophage replacement kinetics during UC, we used the canonical mouse DSS model (Figure 3A) 4, 17, 24. We found that CD11b+ F4/80hi Fraction I cells represented colon tissue-resident macrophages and CD11b+ F4/80int Fraction II cells consisted of Ly6c+ MHC II- monocytes (P1), Ly6c+ MHC II+ inflammatory macrophages (P2), Ly6c- MHC II+ infiltrating macrophages (P3), and Ly6c- MHC II- eosinophils (P4), as described previously 3 (Figure 3B).
Additionally, we performed a time point analysis for resident and inflammatory macrophages in an acute DSS model. 1.5% and 3% DSS treatments represented mild and severe UC (Figure 3C). Similar macrophage replacement kinetics were observed for both the mild and severe DSS models, where resident macrophages reduced to a minimum level around day 9 before returning to a normal level around day 28 (Figures 3 C, D, and E). Simultaneously, Ly6c+ MHC II+ inflammatory macrophages reached a maximum around day 9 and returned to a basal level around day 14, while Ly6c+ MHC II- monocytes reached the basal level slightly later at around day 21 (Figure 3C, D, and E). Ly6c- MHC II+ infiltrating macrophages and CD11b+ F4/80- neutrophils showed different recruitment kinetics during mild and severe UC. Ly6c- MHC II+ infiltrating macrophages decreased to the minimum level around day 9, similar to CD11b+ F4/80hi resident macrophages, and replenishment of Ly6c- MHC II+ infiltrating macrophages was completed around day 35 for mild UC but took longer for severe UC (Figures 3C, D, and E). Neutrophil recruitment reached a maximum between days 7 to 14, which was only resolved around day 28 for mild UC and much later for severe UC (Figures 3C, D, and E). It is well known that monocyte recruitment is dependent on the chemokine CCR2, thus we wanted to try if we could disrupt the inflammation network formation through CCR2 knockout. However, the Ly6c+ MHC II+ macrophage infiltration into the inflammatory region was not affected (Figure 3F), which was probably due to complementary roles of other chemokines like CCR5.
Overall, our dynamic analysis of macrophage replacement using the mouse DSS model showed that inflammatory macrophages could replace resident macrophages during DSS treatment and return to a basal level when inflammation was resolved. In addition, CCR2 knockout was not enough to block the mouse inflammatory macrophage infiltration.
scRNA-seq analysis confirms MDR of the UC ecosystem
To confirm the IMC pathological landscape of the UC microenvironment and illustrate the mechanism of MDR, scRNA-seq analysis was performed on four healthy donors’ colon, four UC patients’ self-control and corresponding lesion samples (Figure 4A). 52716 single cells were captured with high sequencing quality and we found 15 major cell clusters (Figure. 4B), which was annotated based on their specific markers (Figure 4C), such as EpCAM+ epithelial cells (Supplementary Figure. S2), DCN+ smooth muscle cells 25, CD31+ endothelial cells and rare Gnat3+ Tuft cell (Supplementary Figure S2A) 26. In addition, various immune cell subsets were identified, including CD4+, CD8+ T cells, FoxP3+ regulatory T cells (Treg) and IL7R+ memory T cells (Figure. 4B and Supplementary Figure S2B). B cells could also be separated into IgG producing plasma cells, CD83+ activated B cells27 and Lymphotoxin β+ B cells (Supplementary Figure S2B) 28. Furthermore, macrophages were classified into CD68+ CD163+ CD11blow resident macrophage and CD68+ CD11b+ IL-1β+ inflammatory macrophages. CD11b+ inflammatory macrophage expressed higher level of CCR2, which was central chemokine for macrophage recruitment (Supplementary Figure S2B).
We also observed decreased epithelial cells with heavy infiltration of T cells and B cells, including T cells, CD83+ activated B cells and plasma cells, in the UC region compared to the control region (Figure 4D). And Treg cells were also elevated in UC region (Figure 4D), indicating a negative feedback for the inflammation process in the UC region.
In addition, consistent with the results showed in Figure 2J, the scRNA-seq analysis showed that resident macrophage decreased, while inflammatory macrophages increased in UC region (Figure 4D), indicating the MDR occurred only in the resident macrophage subset 11. Overall, our scRNA-seq landscape covers the major ICC TME cellular components and depicts the dynamic changes of the intestinal ecosystem, such as the MDR occurred in the UC region, which was specific to resident macrophage, confirming the IMC results.
Uneven expression of ROS scavenging enzyme SOD1/2 resulted ROS induced resident macrophage apoptosis
Next, we tried to illustrate the mechanism for resident macrophage disappearance during UC progression. First, we tested if MDR observed in UC region also involved the Factor V dependent coagulation process, as previously reported12. However, Factor V was not expressed by resident macrophages or other cells (Figure 5A). In order to explore the mechanism by which resulted in MDR, we analyzed the differentially expressed gene (DEG) between resident macrophage and infiltrating macrophage due to the opposite cellular dynamics between these two macrophage populations observed during UC. Among the DEG,we found differentially expressed of SOD2 which participate in the process of oxidative stress (Figure 5B). Furthermore, we found ROS scavenging enzyme SOD1/2 expression was uneven. SOD1 was generally expressed by many cells except resident macrophage and SOD2 was highly expressed specifically only in inflammatory macrophages (Figure 5C). As a result, ROS level was quite high in resident macrophage compared with inflammatory macrophage, detected by FACS compatible ROS probes (Figure 5D).
In UC region, epithelium barrier is usually disrupted and commensal bacterial invades the intestinal space, which results bacterial infection defense system including ROS generation 29. To mimic this scenario, THP-1 cells were first induced to M0 cell with PMA 30 and then cocultured with bacterial component LPS. LPS induced ROS generation (Figure 5E), as previously reported 31. Enhanced activation of capsase-3 was observed, upon LPS exposure (Figure 5F), which resulted macrophage apoptosis (Figure 5G), consistent with an earlier report on macrophage 32.
In general, we found general elevated oxidative phosphorylation and ROS level in the UC region compared with healthy control and self-control. Meanwhile, ROS scavenging enzyme SOD1/2 were highly expressed only in inflammatory macrophage but not resident macrophage, which resulted ROS induced resident macrophage apoptosis as previously reported33.
Cellular neighborhood changes during UC
It is widely accepted that the changes of macrophages can further influence the nearby immune system in situ34. Hence, it is critical to identify the local changes of the inflammation network following the switch of macrophages. Therefore, we performed the regional cellular neighborhood analysis. Cell neighborhood was identified for each cell, primary neighbors within 4 μm of the target cells and the secondary neighbors of the primary neighbors were defined as interaction partners of the target population (Figure 6A) 19, 22. In order to figure out different cellular neighborhood function units, cellular neighborhoods were annotated according to the major cellular clusters (Figure 6B). Then based on the CN composition, Voronoi plot was used for each IMC image to visualize whether the defined CN function units were associated with the inflammation caused by UC (Fig 6C) as previously reported 35. As shown in figure 6C, the topology map of IMC represented by Voronoi plot with various CN indicated by different colour was aligned nicely with IMC image. We found lymphocyte-enriched CN and infiltrating macrophage centered CN were highest in severe UC patients. Besides, we found epithelial-enriched CN and resident MF-enriched CN were lowest in severe UC patients (Figure 6D).
Thus, CN changes during UC was consistent with the cellular cluster changes.
Infiltrating macrophages recruitment forms the inflammatory cytokine network within the cellular neighborhood
Next, we intend to explore to the mechanism and results of the CN changes. We noticed profound cytokine production in UC patients based on both IMC and scRNA-seq results. IMC showed that TNF-α was produced mainly by lymphocytes (Figure 7A), and scRNA-seq helped to identify the T cells, CD83+ activated B cells and inflammatory macrophages as major TNF-α producers (Figure 7A). In addition, our IMC and scRNA-seq data showed that IL-1β was produced mainly by infiltrating macrophages (Figure 7B).
Furthermore, cell-cell interaction analysis using scRNA-seq data was implemented based on the published tool, NicheNET, by Browaeys et al 36, which models intercellular communication by linking ligands to target genes. Inflammatory macrophages were shown to induce T cells and B cells to produce TNF-α through IL-1β, based on NicheNET analysis (Figure 7C).
Consistent with the scRNA-seq analysis, neighbors of resident macrophage expressed less TNF-α, while neighbors of inflammatory macrophages expressed more TNF-α, As shown in figure 7D, to prove the crucial role of the infiltrating macrophage in formation of inflammation network, we sorted the Ly6c+MHC II+ macrophage from DSS treated mouse and co-cultured with the lymphocytes from the draining mesenteric lymph nodes (Figure 7E). And we found co-culture with Ly6c+MHC II+ macrophage could induce both T cells and B cells to secrete TNF-α (Figure 7E).
In summary, by combining IMC and scRNA-seq data, we suggested that during UC, bacteria invasion caused ROS production led to resident macrophage specific MDR and infiltration of inflammatory macrophage, which formed the inflammatory CN with lymphocytes and orchestrated the inflammatory cytokine network (Figure 7F).