Tracing the stemness and malignant transition in a heritable colorectal cancer Lynch Syndrome by single-cell RNA-seq analysis

Abstract

BackgroundLynch Syndrome (LS) is an autosomal dominant disease characterized by germline heterozygous mutations in DNA mismatch repair (MMR) genes. High-risk LS patients may proceed to colorectal cancer (CRC). However, the drivers or biomarkers of LS benign colon tissue approaching malignant CRC are not completely understood. The similarity and difference between LS-related and nonLS-related CRC are also not well interrogated (LS-CRC vs. nonLS-CRC). This study aimed to understand the cellular changes during malignant transition in LS. MethodsSingle-cell RNA sequencing (scRNA-seq) was used to analyze paired biopsy samples from 3 patients with LS (cancer tissues vs. adjacent normal tissues, n=3). Single-nuclear RNA sequencing (snRNA-seq) was used to analyze a frozen biopsy sample from a patient with LS (n=1). scRNA-seq or snRNA-seq datasets from CRC were downloaded from the open source. Integrative computational analysis was performed to conclude the distinct pattern in the single-cell atlas. Immuno-histo-fluorescence staining (IHF) were also performed for three key markers. ResultsIn the single-cell atlas, we observed an increase in primitive cancer stem-cell-like cells with high expression of a general cancer biomarker CarcinoEmbryonic Antigen-related Cell Adhesion Molecule 5 (CEACAM5) in the epithelium of the LS. Infiltration of immune cells and DNA repair biological activity are dramatically increased in LS carcinoma. The burden in LS is fundamentally elevated compared to that in CRC or healthy donors when the mutations are in the coding-sequence-wide range. Furthermore, T cell and macrophage-related tumor immunity is readily mobilized in carcinomas compared to paraCArcinomas. ConclusionsThis study provides single-cell transcriptomic resource using affected tissues from patients with Lynch Syndrome and describes an integrative profile covering the alterations (cancer stem cell markers, mutation burden, and tumor immunity) during the malignant transition from healthy to Lynch Syndrome and to colorectal cancer at the single-cell level. Highlights and Figure/Table Index (Take-Home Messages)O_LIColon tissues from HD (n=4), patients with LS (n=6) and with CRC (n=3) were collected for snRNA-seq or scRNA-seq analysis; Diagnosis of the 6 LS patients are clearly supported by our pedigree documentations (Fig. 1 and Supp. Fig. 1) and briefed in Table 1; C_LIO_LIAmong the 6 patients with LS, 1 for snRNA-seq analysis (a carcinoma tissue only), 3 for scRNA-seq analysis (3 paired para-carcinoma and carcinoma tissues) and 3 for experimental validations; (Fig. 1) C_LIO_LIFollowing studies are focused on the comparisons between HD, LS and CRC; In addition, comparisons between paired carcinoma and para-carcinoma tissues from patients with LS were also performed; (Fig. 2-7) C_LIO_LIWhen HD and LS are compared, a cancer stem cell marker CEACAM5 is readily detected by snRNA-seq analysis; (Fig. 5) C_LIO_LIWhen carcinoma and para-carcinoma from LS patients are compared, increased immune cell infiltration of and enhanced DNA repair activity in the malignant tissues is readily detected by scRNA-seq analysis; in addition, three upregulated markers in LS carcinoma (BACE2, GRPC5A and OLFM4) were identified and validated by immunohistoinfloresnce; (Fig. 3 and Fig. 4) C_LIO_LISComatic algorithm-based mutation calling analysis suggests a comparable mutation burden between carcinoma and para-carcinoma from LS patients, but such mutation burdens are grossly greater than that from HD or that from patients with CRC; (Fig. 6) C_LI O_FIG O_LINKSMALLFIG WIDTH=170 HEIGHT=200 SRC="FIGDIR/small/647495v1_fig1.gif" ALT="Figure 1"> View larger version (50K): org.highwire.dtl.DTLVardef@174af7dorg.highwire.dtl.DTLVardef@1ad3e71org.highwire.dtl.DTLVardef@dea1e4org.highwire.dtl.DTLVardef@568f91_HPS_FORMAT_FIGEXP M_FIG O_FLOATNOFig. 1C_FLOATNO Enrollment of LS patients and overall design of the study. (A) Three family pedigrees of the LS patients (LS-CRC) who donated the samples for single-cell transcriptomic analysis. Cropped pedigree information is provided in the Main text. See the Supp. Fig. 1 for the complete pedigree information. (B) Two different platforms were used in the study: snRNA-seq and scRNA-seq. (C) Various types of comparison were performed to characterize the malignant transition of LS. C_FIG O_TBL View this table: org.highwire.dtl.DTLVardef@eb8e4eorg.highwire.dtl.DTLVardef@cbafaaorg.highwire.dtl.DTLVardef@79ca83org.highwire.dtl.DTLVardef@1de7544org.highwire.dtl.DTLVardef@1e4fcbc_HPS_FORMAT_FIGEXP M_TBL O_FLOATNOTable 1.C_FLOATNO O_TABLECAPTIONClinical information of enrolled patients with LS. An Excel version of the table is provided in the Supp. Material C_TABLECAPTION C_TBL O_FIG O_LINKSMALLFIG WIDTH=160 HEIGHT=200 SRC="FIGDIR/small/647495v1_fig2.gif" ALT="Figure 2"> View larger version (56K): org.highwire.dtl.DTLVardef@8cd71corg.highwire.dtl.DTLVardef@7e75b3org.highwire.dtl.DTLVardef@61757eorg.highwire.dtl.DTLVardef@68ca5f_HPS_FORMAT_FIGEXP M_FIG O_FLOATNOFig. 2C_FLOATNO snRNA-seq analysis of a frozen colon biopsy sample from a LS patient reveals a cluster of cancer stem cells (A) Frozen colon biopsy samples were subjected for snRNA analysis (Healthy donors, N=4; Lynch Syndrome, N=1) and epithelial cells were filtered out for further study. Data were integrated by the Seurat and Harmony standard procedure. Clusters 0-9 were identified in a total of 11,353 cells (the Healthy Controls have 7,373 cells; the LS patient has 3,980 cells). The malignancy of the epithelial cells was inferred by the inferCNV and SCEVAN algorithm (two major compartments: tumor and no-tumor). The portion of tumor or non-tumor cells in each cluster was also illustrated (right panel). Note that the major contributor in Cluster 4, 6, and 7 is the tumor-like cells from the LS patient. (B) Each cluster of the epithelial cells was further annotated according to the combination of Seurat cluster identity and SCEVAN malignant identity. Mix_1 to 5 contain both tumor-like cells and non-tumor-like cells; Other clusters are marked as BEST4+-Enterocytes, Enterocytes and Goblet cells where much fewer tumor-like cells were assigned. Note that the epithelial cells from the LS patient are mainly classified into Mix_2, 4, and 5 tumor-like cells. (C) Three clones were inferred by SCEVAN algorithm. In the phylogenetic tree, Clone_1 and Clone_2 are closer and probably originated from Clone_3. (D) A Sankey plot showing overall belonging relationships in the epithelial snRNA-seq dataset. Note that the Mix_2 pool is dominated by cells from Lynch_01 and that in the Mix_2 pool Clone_3 is the most dominant clone. (E) The Mix_2 cells specifically express numerous colon epithelial stem cells signature markers such as SMOC2, RGBM, OLFM4, AXIN2 and LGR5. Of note, all these intestine stem cell signature genes were reported as marker genes for cancer stem cells, especially LGR5. (F) Pseudotime trajectory analysis of the Lynch Syndrome epithelium cells indicates the Mix_2 cells are the most primitive in development. (G) Top 36 differentially expressed genes (DEGs, both upregulated and downregulated) were highlighted when the non-tumor-like cells were compared with tumor-like cells in the Mix2. (H) The Mix_2, 4, 5 epithelial cells all have enriched signaling pathways involved in DNA repair response and MYC targets (fonts in red) while Mix_4 and 5 manifest Interferon alpha or gamma response (fonts in blue). C_FIG O_FIG O_LINKSMALLFIG WIDTH=158 HEIGHT=200 SRC="FIGDIR/small/647495v1_fig7.gif" ALT="Figure 7"> View larger version (54K): org.highwire.dtl.DTLVardef@1be2c6aorg.highwire.dtl.DTLVardef@55f267org.highwire.dtl.DTLVardef@9d60fborg.highwire.dtl.DTLVardef@4a2d37_HPS_FORMAT_FIGEXP M_FIG O_FLOATNOFig. 7C_FLOATNO Distinct tumor immunity microenvironment profiles in LS carcinoma (A) T cells from LS-CA and LS-paraCA patients were filtered for further annotation. The LS-paraCA group had 1410 T cells, whereas the Lynch CA group had 2125 T cells. Six major T cell subtypes were identified: Naive CD4-T, NKT, Treg, IL7R+CD8_T, GZMA+CD8_T, and GZMK+CD8_T. Note that GZMA+CD8_T cells appear to be more abundant in Lynch CA (arrows). (B) The pseudotime trajectory indicating the T cells developed towards two branches: the Naive CD4-T and Treg branches, and GZMA+CD8_T and GZMK+CD8_T branches. GZMA+CD8_T cells appear to be more mature in LS-CA than in LS-paraCA (arrows). (C-F) Biological function-instructed scoring of the paired single-cell datasets of T cells from the three LS patients, including T cell cytotoxicity (B), T cell exhaustion (C), T cell immune escape (D), and T cell surveillance (E). (G) Circle plots showing interaction profiles across indicated cells in the paired LS samples. The direction of the lines indicates the direction of the cell-cell communication, the width of the lines indicates the interaction strength, and the color of the lines indicates the cell types of the senders. (H) Summary of the Study. In this study, we profiled Lynch Syndrome biopsy samples using single-cell or single-nucleus RNA-seq analysis. Paired carcinoma and adjacent para-carcinoma samples were included to facilitate the analysis, including tracing tumor clones, calculating mutation burden, identifying biomarkers for cancer stem cells, and tumor immune microenvironment changes during the transition from pre-carcinoma to carcinoma in the inheritable cancer Lynch Syndrome. See the text in the Discussion section for further detail. C_FIG O_FIG O_LINKSMALLFIG WIDTH=148 HEIGHT=200 SRC="FIGDIR/small/647495v1_fig5.gif" ALT="Figure 5"> View larger version (44K): org.highwire.dtl.DTLVardef@1de7f42org.highwire.dtl.DTLVardef@74e8ddorg.highwire.dtl.DTLVardef@1642847org.highwire.dtl.DTLVardef@907a23_HPS_FORMAT_FIGEXP M_FIG O_FLOATNOFig. 5C_FLOATNO Distinct pre-cancer signs in epithelial and immune cells of LS para-carcinoma (A) The scRNA-seq datasets of LS-paraCA were integrated with those of the colons from healthy donors (HD) (N=3, respectively; the HD group had 7885 cells in total, whereas the LS-paraCA group had 4766 cells). A dominance of CEACAM5high epithelial cells were identified in LS-paraCA. The sample from the patient with relapsed LS is marked with an arrow in the bar plot (left panel). (B) The portion of CEACAM5high epithelial cells, infiltrated CD8_T cells and macrophages was quantified. The sample from the relapsed LS patient is marked with arrows. (C, E) Top marker genes co-expressed in CEACAM5high epithelial cells and the enriched pathways are illustrated in the gene expression heatmap. Cells from the Enterocytes, Tufts, and Goblet subtypes were used as controls. (D, F) Quantification of the expression of CEACAM5 in epithelial cells of HD and LS-paraCA cells. C_FIG O_FIG O_LINKSMALLFIG WIDTH=178 HEIGHT=200 SRC="FIGDIR/small/647495v1_fig3.gif" ALT="Figure 3"> View larger version (48K): org.highwire.dtl.DTLVardef@142c991org.highwire.dtl.DTLVardef@1bf5d99org.highwire.dtl.DTLVardef@99edd5org.highwire.dtl.DTLVardef@1cd5c15_HPS_FORMAT_FIGEXP M_FIG O_FLOATNOFig. 3C_FLOATNO Distinct infiltration of immune cells in carcinoma and para-carcinoma tissues of LS (A) Fresh paired carcinoma (LS-CA) and adjacent para-carcinoma (LS-paraCA) biopsies of colon tissue was subjected to scRNA-seq analysis (N=3). Colorectal cancer (nonLS-CRC, they are not diagnosed as LS) tumor samples from the public dataset were also included as controls (N=4). Cells were first grossly annotated as stromal, epithelial, or immune cells. Note that the LS-CA group showed increased infiltration of myeloid cells compared with LS-paraCA group, but T cell infiltration was comparable between the two groups. (B) Expression of classic markers for annotating the major 5 cell types. (C) Detailed annotations of 16 cell types in paired LS-CA and LS-paraCA samples. (D) Stacked bar plot showing the contribution of sample sources in each cell type. (E) Quantification of immune cell infiltration in LS-CA compared with LS-paraCA. Although no significant differences were observed, the fold-changes were approximately 2-10. Although the overall T cell infiltration was comparable, infiltration of GZMA+CD8_T and GZMK+CD8_T cells appeared to be higher in LS-CA than in LS-paraCA. Of the three LS patients enrolled in the study, one was a relapsed patient (gray dashed line, surgery treatment performed; whose son donated the frozen sample in Figure 2), while the other two were treatment-naive (gray solid lines). (F) Boxplots show DNA repair score quantification per sample (above) or per epi-cell. (bottom). C_FIG O_FIG O_LINKSMALLFIG WIDTH=152 HEIGHT=200 SRC="FIGDIR/small/647495v1_fig4.gif" ALT="Figure 4"> View larger version (51K): org.highwire.dtl.DTLVardef@1588458org.highwire.dtl.DTLVardef@58db3borg.highwire.dtl.DTLVardef@1a12084org.highwire.dtl.DTLVardef@15de70e_HPS_FORMAT_FIGEXP M_FIG O_FLOATNOFig. 4C_FLOATNO Prioritizing novel epithelial biomarkers for tracing the transition between para-carcinoma and carcinoma in LS (LS-paraCA vs. LS-CA) (A) Epithelial cells were filtered out for further analysis and annotated as clusters 0-9 in the UMAP plot. (B, C) Bar and pie plots show the contribution of sources to epithelial clusters. Cluster 0 contained cells largely from Lynch CA_02, whereas Cluster 9 was composed of Lynch paraCA_02. The patient who donated Lynch CA_02 and paraCA_02 is identified as a relapsed LS. (D) KEGG enrichment analysis showing the enriched bioprocesses in the four epithelial cell clusters. (E) Five genes with upregulated expression in LS carcinoma epithelial cells were prioritized based on the scRNA-seq and snRNA-seq datasets. (F) Expression of BACE2, GPRC5A and OLFM4 in the paired LS-paraCA and LS-CA groups. Epithelial cells are circled by dashed lines. Note that the expression of genes in epithelial cells from nonLS-CRC samples was also very strong. (G) Verification of the low-to-high transition in the expression of BACE2, GPRC5A and OLFM4. Paraffin sections from paired LS-paraCA and LS-CA samples were prepared, and immunohistochemical (IHF) staining was performed. The expression of the indicated proteins is labeled in green, whereas the nuclei are stained with DAPI in blue. In total, biopsy samples from 4 patients with LS were used for IHF verification. C_FIG O_FIG O_LINKSMALLFIG WIDTH=148 HEIGHT=200 SRC="FIGDIR/small/647495v1_fig6.gif" ALT="Figure 6"> View larger version (28K): org.highwire.dtl.DTLVardef@1ec5210org.highwire.dtl.DTLVardef@1ffd6f9org.highwire.dtl.DTLVardef@17a7a8forg.highwire.dtl.DTLVardef@4819e_HPS_FORMAT_FIGEXP M_FIG O_FLOATNOFig. 6C_FLOATNO Increased mutation burden in the colon tissues from Lynch Syndrome (A) The schematic diagram for calculating mutation burden based on the scRNA-seq datasets using the SComatic algorithm. Only single-base substitutions (SBS) were called in this study. Once mutation calls were achieved, the mutation burden value was quantified by normalizing the total mutation calls to the total cell number in the samples or cell types (See Methods section for details). (B) The total SBS mutation burden in the four groups of samples was indicated. (C) The six different SBS mutation burdens in the four groups of samples as indicated. (D) The SBS mutation burden in the cell types of the four groups is indicated. Note that the The mutation burden appears to be higher in mast, epithelial, plasma, myeloid, and Treg cells in the LS-CA group than in that the LS-paraCA group. C_FIG