Single-nucleus RNA sequencing and deep tissue proteomics reveal distinct tumour microenvironment in stage-I and II cervical cancer

Identification of the tumour cell composition of stage-I and II cervical cancer

We performed snRNA-seq experiments using all nuclei isolated from the tumour tissues of stage-I (n = 4) and II (n = 3) stage donors, respectively (Fig. 1A). After quality control, a total of 72,128 nuclei (42,928 stage-I and 29,200 stage-II) were used for downstream analysis (Fig. 1B and Fig. S1). Unsupervised clustering analysis revealed 22 cell types, which were present in both the stage-I (CCI) and II (CCII) groups, indicating that the cell-type identity was not strongly confounded by the ageing and stage effects (Fig. 1B; Figs. S2A and 2B). However, these cell types were not detected in all individual patients (Fig. S2C). The proportions of different cell types at the two stages were compared (Fig. 1C), and the population comparison among different samples were displayed in Figure S2D. Overall, the percentage populations of the cluster 3, 4, 9, 14, 15, 16, 17, 18, 19 were largely increased in the CCII group, whereas reduced proportions were observed for the cluster 2, 5, 7 and 20 (Fig. 1C and Table S1). Many of cluster 0 and 3 cells did not show the differential expression of the marker genes for definitive cell-cycle (non-cycling) (Fig. 1D). Cluster 1, 2 and 5 possessed high populations of cells at S or G1 cell cycle, while S and M cell cycles were more represented in cluster 7 and 8, respectively (Figs. S3A and 3B). CCI-4, CCII-2 and CCII-3 showed high proportion of non-cycling cells, while the other three CCI samples possessed more cells at G1 and S cell cycles (Figs. S3C and 3D). 

Fig. 1figure 1

The snRNA-seq analysis of the tumour tissues of patients with stage-I and II cervical cancer. (A) Experimental design: the tumour tissues were collected from the patients at The First Affiliated Hospital/School of Clinical Medicine of Guangdong Pharmaceutical University. (B) t-Stochastic neighbour embedding (t-SNE) representation of aligned gene expression data in single nuclei extracted from the TME of CCI and CCII patients shows partition into 22 distinct clusters. (C) The proportions of the 22 cell clusters in the CCI and CCII groups. (D) The distributions of cell cycle phases in the t-SNE space. (E) Selected enriched genes used for biological identification of each cluster and the top 5 DEGs of each cluster (scale: log-transformed gene expression). MΦ represents macrophage; Treg cell, regulatory T cell; NK cell, natural killer cell; NPC, neural progenitor cell; GMPC, granulocyte-monocyte progenitor cell; pDC, plasmacytoid dendritic cell (see Table S2 for the full list of all marker genes detected)

The differentially expressed genes (DEGs) were analysed to determine cell type-specific marker genes (Fig. 1E and Table S2). The clusters were annotated with predicted cell-type identities based on known marker genes derived from the expert annotation in literature [32]. Correlation analysis of the 22 clusters showed that cluster 17 was least correlated with the other clusters, followed by cluster 13, implying distinct phenotypes present in these two clusters (Fig. S4). The cell types directly associated with immune response included T cells (cluster 4; marker genes: NELL2, ITK, and IL7R), macrophage (MΦ, cluster 6; SIGLEC1, FPR3, and MSR1) [33], γδT cells (cluster 8; TOP2A, ASPM, and CENPF), naïve B cells (cluster 9; DCC, CD38, POU2AF1) [34], regulatory T cells (Treg, cluster 10; CTLA4, IL2RA, and F5) [35], NK cells (cluster 11; KLRC1, GNLY, and NCR1), mature B cells (cluster 14; BLK, FCRL1, MS4A1) [36], CD141+CLEC9A+ DC (cluster 16; SLC24A4, FLT3, and ZNF366) [37], and plasmacytoid dendritic cells (pDCs, cluster 21; CLEC4C, IL3RA, and IRF8) [36, 38]. Several clusters had the features of stem cells, including cancer stem cells (CSCs) (cluster0; CLDN10-AS1, TOX3, and PROM1), mesenchymal stem cells (MSCs) (cluster 3; ADAMTS2, LAMA2, and ABI3BP), vascular stem cells (VSCs) (cluster 7; RRM2, EXO1, and SKA3), and endothelial cell/submandibular gland stem cells (cluster 12; FLT1, PCDH17, and VWF) [39]. Two clusters were annotated as progenitor cells, i.e., neural progenitor cells (NPCs, cluster 19; NRXN1, ADAMTSL1, and PPP2R2B) [40, 41] and granulocyte-monocyte progenitor cells (GMPCs, cluster 20; CPA3, MS4A2, and TPSAB1) [42]. Cluster 1 was characterised as basal cells, with significantly high expression of KRT15, KRT17, and KRT5. Epithelial cells and lymphatic endothelial cells were mainly detected in cluster 2 (IL1RN, GPRC5A, SPRR1B) and 18 (FLT4, PROX1, and CD34) [43], respectively. In addition, adipose-derived stromal cells (cluster 5; NTRK2, PTPRZ1, and CHL1), myofibroblast (cluster 13; SPARC, COL1A1, and COL6A2) [44], pericytes (cluster 15; RGS5, ABCC9, and ADGRF5), and astrocytes (cluster 17; FOSB, ATF3, and ITGB4) [45] were also present.

Immunosuppressive and tumour-growth-promoting phenotypes dominated the macrophages in the TME of CCII patients

Relatively high populations of MΦs were identified in CCI (5.30%) and CCII (5.65%) patients, respectively. The gene expression analysis revealed 122 upregulated and 102 downregulated DEGs in the CCII groups relative to the CCI, with the log-transformed expression of the top 50 DEGs hierarchically compared in Fig. 2A (see Table S3 for the full list of DEGs and annotations). The expression of MΦ marker genes was comparatively downregulated in the CCII group, for instance, C1QB, C1QC, STAT1 and IFI44L. In addition, many of chemokines, cytokines and interleukins exhibited significantly higher expression in the CCI group, including the signatures of M1-like MΦs, such as IL12RB1, IL2RA and IL20RB (Fig. 2B). The genes (CCL19 and MMP11) that correlated with immune suppressive MΦs were elevated in the CCII group. The canonical pathways were analysed based on the respective transcriptome profiles using GSEA. It was evident that the epithelial-mesenchymal transition (EMT) was the most enriched pathway in the CCII group, supported by the upregulation of collagen family members, such as COL1A1, COL1A2, COL6A2 and COL3A1. Furthermore, these collagens might function collaboratively with extracellular matrix synthesis regulating genes (e.g., SFRP4, LUM, SPARC, and DCN), to enhance tissue growth (Fig. 2C and 2D; Table S3). Besides, the enrichment of ‘P53 pathway’, ‘TNFα signalling via NFκB’, and ‘apoptosis’ was detected in the CCII group, and so were several other pathways closely related to the development of cell cytoskeleton. The elevation of genes encoding interferon-induced proteins in the CCI group strongly supported the activation of IFN-α and IFN-β response pathways, including IFI44L, IFIT3, IFI44, IFI35, and IFIH1 (Fig. 2E).

Fig. 2figure 2

The MΦs of the CCI and CCII groups shows distinct levels of immune response. (A) The hierarchy diagram compares the expression and correlation of the top 50 DEGs between the CCI and CCII groups. (B) The 2D t-SNE graph compares the distribution of MΦs expressing (the normalised expression) the genes (FC > 1.2, P-value < 0.05) associated with the signalling of chemokines, cytokines and interleukins in the CCI and CCII groups. (C) The GSEA analysis of the Hallmark pathways enriched respectively by the DEGs of the CCI and CCII groups. The ranking of the genes significantly associated with the epithelial mesenchymal transition (D) and the IFN-α response pathway (E)

The MΦ subtypes were analysed to unveil the change in cell heterogeneity between the two stages. There were five subtypes (i.e., cluster-0 to 4) identified, with the expression of the top five marker genes compared in Fig. 3A. Cluster-0 had the phenotype of resident-like macrophage, characterised by MS4A6A, CD163 and CD163L, thus it was referred to as C0-Res. Several marker genes of cluster-1 were associated with tumorigenesis, such as PARD3[46], EGFR[47], and SMAD3[48], was thus labelled as C1-TAM (Table S4). The third subtype showed the significant upregulation of the signatures of M2-like MΦ, including SLC16A10[49], SLC11A1[50] and CTSL[51], and thus we named it C2-M2. The fourth subtype showed the marker genes of both dendritic cells (ADAM19, HDAC9, and MCOLN2)[52,53,54] and macrophages (SLC8A1, RUNX3, and LCP1)[55,56,57], which was referred to as C3-DC. The fifth subtype had several marker genes (SEMA3A, ESRRG, and IL18)[58,59,60] representing M1-like MΦ, which was labelled as C4-M1. The enrichment of biological processes (BPs) in each subtype was analysed (Fig. S5). More immune response associated BPs were observed in C2-M2, C3-DC and C4-M1, with several inflammation relevant processes only present in the C4-M1, such as ‘interleukin-21-mediated signalling pathway’, ‘inflammatory response’ and ‘regulation of cytokine production’. C1-TAM was highly enriched with cell development and tissue growth processes, and occupied a higher proportion in the CCII group, while other subtypes were more abundant in the CCI group (Fig. S6A).

Fig. 3figure 3

The analysis of macrophage heterogeneity of the CCI and CCII groups. (A) The subtype analysis of the macrophages. Five subpopulations were identified, including resident-like (C0-Res), tumour-associated (C1-TAM), M2-like (C2-M2), MΦ/DC (C3-DC) and M1-like (C4-M1). The expression of top 5 DEGs was compared across different subtypes. (B) Comparison of the average expression of selected DEGs associated with macrophage function. (C) Pseudotime trajectory analysis of macrophage subtypes

The DEGs relevant to immune response appeared mostly upregulated in all the subtypes of the CCI group, especially in C0-Res, C1-TAM and C2-M2 (Fig. S6B). The average expression of selected genes associated with macrophage functions was compared across the subtypes, showing that STAT1, HLA-DRB1, TMSB4X, C1QC and FGL2 were upregulated in all subtypes of the CCI group with a higher percentage of expression (Fig. 3B). Although S100A8, DUSP1, MT2A, and JUNB had higher average expression in the subtypes of the CCII group, their percentage of expression was comparatively low. The expression of the genes related to cell proliferation and extracellular matrix development, such as CCN2, AEBP1, MGP, and several members of the collagen family, was highly elevated in all subtypes of the CCII group. Notably, the upregulation of several genes encoding mitochondrial proteins associated with oxidative phosphorylation, for instance, MT-CYO2, MT-CYO3, MT-ND4 and MT-ND1, were observed in the C0-Res, C1-TAM and C2-M2 of the CCII group (Table S4), which might be associated with active cellular metabolism during tumour growth. We then projected MΦs onto the two-dimensional state-space defined by Monocle3 for pseudotime analysis, to infer lineage trajectory for MΦ development (Fig. 3C). The trajectory began with the C0-Res and C2-M2, and then developed to separate directions, with one direction leading to C3-DC and C4-M1, while the other direction eventually linked to C1-TAM with a branch composed of C1-TAM, C3-DC and C4-M1. It appeared that the C3-DC and C4-M1 had similar developmental lineage along the pseudotime.

A total of nine states were thus derived from the trajectory, with large proportions of C0-Res and C2-M2 cells detected at State-1, State-3, and State-5, while C1-TAM cells gradually became the largest population at late pseudotime (Fig. S6C). There were high proportion of C3-DC in State-2, 4 and 9. It was evident C0-Res, C1-TAM and C2-M2 cells aligned more with the cells derived from the CCI group relative to the CCII group, except for State-9. The DEGs of each state were compared between the CCI and CCII groups (Table S4), and the macrophage marker genes showed higher percentage and/or average expression in most states of the CCI group, except for State-3 showing the upregulation of CXCR4 and DOCK8 in the CCII group (Figure S6D). The expression of mitochondria-associated marker genes and several collagens were elevated along the pseudotime, particularly for the State-5, 6, 7, 8 and 9 (Fig. S6E). The composition of each state within each subtype was displayed; C4-M1 and C2-M2 MΦs were respectively dominated by one cell state, whose population was mostly derived from the CCI group. State-4 and 9 cells occupied nearly 90% of C3-DC (Figure S6F).

There were four branches present along the pseudotime. Analysing the genes that were significantly dependent on Branch-1, we found upregulated expression of immune response relevant genes in State-1 and 2 cells, such as CXCL9, CXCL10, IL15, IL18, DOCK8, DOCK10, and STAT1 (Fig. S7A). The expression of DOCK8 and STAT1 was significantly upregulated in State-1 cells of the CCI group, and so were IFI44L, IFI44 and IFNGR2, which resulted in the activation of IFN-γ response. The genes largely downregulated in the other states post-branching were mainly expressed by M2-like MΦs, playing roles in various metabolisms, whereas resident-like MΦs appeared more active in cell proliferation. In terms of Brach-2, many DEGs in State-1, 3 and 4 were derived from the CCI group and associated with innate immune system post-branching (Fig. S7B). State-6 cells had comparatively lower level of antigen processing and presentation post branching, as suggested by the downregulation of MHC class I proteins, such as HLA-A, B and C, as well as HLA-DQA1, HLA-DRB1 and HLA-DRA (Fig. S7C). Separated by Branch-4, the genes associated with cell adhesion and development were upregulated in State-8, which was more representative in the CCII group; State-8 cells of the CCI group exhibited functions in ion import and response to cytokine (Table S4). The genes relevant to chemokine production and neutrophil activation showed higher expression post-branching, such as DENND1B, NLRP3, CD53, FGR, PTPRC, MNDA, and CD58 (Fig. S7D). This largely reflected the activation of cytokine-mediated signalling in the State-9 cells of the CCI group; in contrast, ECM-receptor interaction and PI3K-Akt signalling were more pronounced in CCII State-9 cells.

The function of T cells was suppressed in the TME of CCII patients

Three populations of T cells were identified, including CD8+ T, γδT and Treg cells. The proportion of CD8+ T cells was much higher in the CCII group (12.85%) than that in the CCI group (8.58%) (Table S1). Considering the entire T cell population, the populations of CD8+ T and γδT cells were largely modulated at different stages, with the former increased from 57% (CCI) to 80% (CCII) while the latter reduced by nearly 70% (from 28% in CCI to 9% in CCII) (Fig. 4A). The patients of the CCI group had a higher number of Treg cells than the CCII group (Fig. 4B). The distribution of three T cell populations between the two groups was displayed in 2D-tSNE space (Fig. 4C). It was evident that CD8+ T cells separated into five major subtypes (a, b, c, d, and e), with subtype-a possessing a significantly higher number of cells derived from the CCII group. The CCI group had more cells derived from subtype-b, c, d, and e. The expression of selected maker genes characterising Treg cells was compared between the CCI and CCII groups (Fig. 4C). CD8A and PRF1 were highly expressed by the subcluster-d cells of CD8+ T, while the cells expressing elevated levels of CD4 and CTLA4 were mainly detected in Treg cells of the CCI group. The expression of IFNGR1 and IFNGR2 was more expressed in γδT cells, whereas IFIT3 was upregulated in the Treg cells of the CCI group.

Fig. 4figure 4

T cells were less activated in the CCII group patients. (A) The comparison of the proportions of CD8+, γδT and Treg cells of entire T cell populations of the CCI and CCII groups. (B) The number of three T cell populations detected in the TME of individual patients. (C) The 2D-tSNE graphs comparatively display the distributions of different T cells, and the expression (in normalised value) of selected marker genes (incl. CD8A, CD4, PRF1, IFIT3, IFNGR1, IFNGR1, and CTLA4) in the CCI and CCII groups. (D) The hierarchical clustering of the top 100 DEGs between the three T cell types of the two groups (FC > 1.5 and P-value < 0.05)

The hierarchical clustering of the top 100 DEGs across the three T cell populations of the two groups was displayed in Fig. 4D. The expression of DEGs associated with the activation of T cell immune response was upregulated in CD8+ T of the CCI group significantly, such as CD7, RUNX3, CD3D, CD3G, GZMB, CD8B, and CCL5. This was also partially observed for the Treg cells. Several genes closely related to interferon response were upregulated in the CCI γδT cells, for example, IFIH1, ISG15, GBP1, GBP4, GBP5, OAS1, OAS2 and OAS3. Notably, the expression of two genes (CD38 and CD48) related to antigen-presenting were elevated in the CD8+ T and Treg cells, but downregulated in the γδT cells of the CCI group. The genes regulating complement and coagulation cascades, and inflammation were clustered and showed higher expression in CD8+ T and Treg cells of the CCII group, for instance, S100A8, S100A9, CYBA, CD55, C3, and PLAU. These observations meant that, compared to the CCI group, the T cells of the CCII group were more immunosuppressive, although the population of CD8+ T was significantly higher.

B cells and DCs in the TME of CCII showed suppressed MHC class I antigen process

The CCI group had a higher population of pDCs, while the population of B cells was more abundant in the CCII group (Fig. 5A). For both CD141+CLECL9A+ DCs and pDCs, the DEGs associated with antigen-presenting processes, such as HLA-DRB5, HLA-DQB1, and HLA-DPA1, were upregulated in the CCI group, whereas the CCII group had a higher average expression of HLA-B, HLA-A, HLA-C, and HLA-DMB (Fig. 5B). In addition, several members of the proteasome subunit gene family appeared more pronounced in terms of both average expression and percentage of expression in pDCs of the CCI group, for instance, PSMB2, PSMD14, PMSC6, and PSM8. The biological process analysis suggested that tissue development signalling was highly activated in the DCs of the CCII group, such as ‘angiogenesis’, ‘blood vessel morphogenesis’, ‘circulatory system development’, and ‘cell–cell adhesion’ (Fig. 5C). In contrast, many immune response processes were highly activated in the DCs of the CCI group, including ‘oxidation reduction process’, ‘immune effector process’, ‘response to virus’, ‘defence response to virus’, and ‘antigen processing and presentation’. It appeared the multicellular organism process was enriched in the CD141+CLECL9A+ DCs of the CCII group, but was suppressed significantly in the pDCs of the CCI group. The expression of selected DEGs associated with MHC class I antigen-processing was compared, and their upregulation in the B cells of the CCI group can be seen (Fig. 5D). The elevation of COL3A1, COLA1 and COLA2 was present in the native B cells of the CCII group (Fig. 5E). Two clusters of genes associated with the signalling of type I and II interferons were upregulated in B cells of the CCI group, such as STAT1, OAS1, OAS2, ISG15, HLA-DPA1, IL12RB2, and JAK2. The signalling of IL-18 and chemokine activity appeared more enhanced in the native B cells of the CCII group, supported by the higher expression of CD81, MMP2, LTB, ITK, and CX3CL1. The thymic stromal lymphopoietin (TSLP) and B cell receptor signalling were overrepresented by the upregulation of TEC, FYN, IL7R, and STAT1 in the B cells of the CCII group.

Fig. 5figure 5

The B cells and DCs in the TME of the CCI group exhibiting enhanced MHC class-I pathway. (A) The proportions of B cells and DC cells detected in the CCI and CCII groups. (B) Comparison of the average expression of the genes associated with antigen presentation in the CD141+CLEC9A+ and pDC populations, with respect to their percentage of expression. (C) The top 25 biological processes enriched in CD141+CLEC9A+ and pDC populations. (D) Comparison of the normalised expression of selected genes associated with antigen presenting processes in B cells and DCs in the two groups. Student’s t-test was used to evaluate the significance (*: P < 0.05, **: P < 0.01, and ****: P < 0.0001). (E) The hierarchical clustering of the top 60 DEGs identified in the B cell populations between the two groups

NK cells were more activated in the CCI group

The populations of NK cells occupied 2.52% and 0.85% of CCI and CCII cells, respectively (Table S1). The top 50 DEGs relevant to immune system and/or response KEGG pathways were compared, which clearly showed that most NK cells of the CCI group expressed higher levels of the marker genes positively associated with activated NK cell function, such as KLRD1, KLRC1, GZMA, GZMB and NKG7 (Fig. 6A). The genes regulating IL17 and TNF signalling were highly expressed by the NK cells of the CCII group, including TRAF4, FOSB, CXCL1, and FOS; in addition, HSPB1, APOE and IL7R, which were related to the apoptosis due to altered Notch3, were more abundant and closely hierarchically clustered. The distribution of cells expressing selected marker genes of activated NK cells was displayed in a 2D-tSNE space, showing that GZMA, KLRC1, GZMB and NKG7 were expressed only in a few NK cells of the CCII group (Fig. 6B). The DEGs associated with chemokine and cytokine signalling mostly had elevated average expression and percentage of expression in the CCI group, such as CD7, IL2RB, CCL5, and CCL4 (Fig. 6C). The GSEA detected ‘natural killer cell mediated cytotoxicity’ as the only KEGG pathway enriched in the CCI group, supported by the upregulation of GZMB, PRF1, KLRC1/2 and CD244 (Fig. 6D). On the other hand, the enrichment of ‘ECM receptor interaction’ and ‘focal adhesion’ was detected in the CCII group (Table S5). In terms of Hallmark pathways, ‘INF-γ response’ and ‘allograft rejection’ were the only two pathways highly activated in the CCI group, whereas EMT and ‘coagulation’ were significantly enriched in the CCII group (Fig. S8A).

Fig. 6figure 6

The function of NK cells was suppressed in the TME of CCII patients. (A) The hierarchical clustering of the top 50 DEGs of NK cells between the CCI and CCII groups. (B) The distribution of the NK cells expressing GZMA, GZMB, NKG7, and KLRC1 in a 2D-tSNE space. (C) The comparison of the average expression of the genes associated with chemokine and cytokine signalling with respect to their percentage of expression in the two groups. (D) The ranking of the DEGs supporting the enrichment of natural killer cell mediated cytotoxicity pathway in the CCI group, with respect to the CCII group. (E) The 2D-tSNE graphs show the distribution of NK cells of the two groups and the subtypes, as well as the proportions of each subtype in individual patient. (F) Pseudotime trajectory analysis of NK cells. (G) The seven states identified along the pseudotime and their composition of the subtypes of NK cells. (H) Comparison of the normalised expression of the marker genes in each state of the two group, including GZMA, GZMB, NKG7, KLRC1, CCL5 and CCL4I

We then analysed the subtypes of the NK cells, which showed that the CCI group possessed higher populations of mature (characterised by the marker genes, PRF1, GZMA, GZMB and CFL1), CD56bright (KLRC1, NCAM1 and GNLY) and terminal (WDR74, HIST1H1D, and HIST1H1E) NK cells in general (Fig. 6E). The population of CD56dim (IL23R, IL7R, and TCF7) NK cells was more present in the CCII group, so were the transitional (FOS, FOSB, and JUNB) NK cells. The development of NK cells in the two groups was inferred based on the trajectory analysis, where many cells present at early pseudotime belonged to in the CCI group, as well on the two branches along the time (Fig. 6F and Table S5). A total of seven states was thus identified, with higher proportion of mature NK cells detected in the CCI group (Fig. 6G). The proportion of terminal NK cells was the second largest in state-1, and then decreased in state-2 and 7 along the time, whereas the proportion of transitional NK cells increased. State-6 was mostly composed of CD56bright and mature NK cells. The expression of the marker genes associated with the activated NK cells, including GZMA, GZMB, NKG7, KLRC1, CCL5 and CCL4, was significantly upregulated in all states of the CCI group except for the state-5 (containing only mature NK), compared with the CCII group (Fig. 6H). The CCII group had more state-2 NK cells, while significantly less NK cells of the other states. For branch point 2, the CD56bright cells of state-1 and 2 expressed higher level of DEGs associated with ‘natural killer cell mediated cytotoxicity’, such as CD247, KLRC1, and NCAM1, with respect to the other states post-branch (Fig. S8B). While the mature NK cells exhibited activated KEGG pathways related to tissue and cell development, such as ‘focal adhesion’, ‘regulation of actin cytoskeleton’, and ‘PI3K-Akt signalling pathway’, due to the upregulation of ITGAT, ITGA1 and BCL2 for state-1 and 2. Thus, in terms of immune response, CCI NK cells were more activated relative to those of the CCII group.

Cancer cells of CCI and CCII patients showed distinct heterogeneity

Several other cell types associated with tumour progression, including stem cells and epithelial cells, were identified with the highest populations (Fig. 1; Table S1 and 2). The proportion of CSCs increased largely in the CCII group to 24.75%, compared to only 15.63% in the CCI group. Similarly, MSC population was higher (13.88%) in the CCII group. In contrast, the ADSCs showed lower proportion (3.13%) in the CCII group. The relative expression density of selected marker genes was displayed in a 2D-tSNE space (Fig. 7A). The top 3 marker genes of CSCs, including CLDN10-AS1, AC024230.1 and FYB2, also showed certain expression in epithelial cells. A cluster of DEGs, such as CFTR, PROM1, CD55, and RHEX, were significantly upregulated in the CSCs and more pronounced in the CCII group, implying a high level of cell growth (Fig. 7B). The genes associated with inflammatory response, such as STAT1, ITGB4, SAT1, and ANTXR1, as well as several collagens, had higher expression in MSCs with respect to other stem cells. The GSEA identified that the top 6 pathways enriched in the CSCs of the CCI group were relevant to immune response, such as ‘cellular immune response to IFN-γ’, ‘IFN-γ mediated signalling’, and ‘B cell mediated immunity’ (Fig. 7C). In contrast, the pathways associated with cell growth were highly activated in the CCII CSCs. The signalling of TGFβ and extracellular matrix organisation appeared more enriched in ADSCs of the CCII group, whereas the activation of ion membrane transport was detected in those of the CCI group (Fig. 7D).

Fig. 7figure 7

The tumour associated stem cells (TASCs) and epithelial cells were more immune response active in the TME of CCI patients. (A) The 2D-tSNE graphs show the expression density of selected marker genes of cancer stem cells (CSCs), mesenchymal stem cells (MSCs), adipose-derived stromal cells (ADSCs) and vascular stem cells (VSCs). (B) The heatmap hierarchically compares the top 10 DEGs of each stem cell-like population in the two groups. The GSEA of the top 6 biological processes enriched (P-value < 0.05) in the CSCs (C) and ADSCs (D) of CCI and CCII groups, respectively. The KEGG pathways enriched in the basal cells (E) and epithelial cells (F) of the CCI group. (G) The comparison of the average expre

留言 (0)

沒有登入
gif