Single Cell Transcriptomic Study Of Lung Cancer Heterogeneity

Objective and background:Different tumor cells within solid tumors are of high heterogeneity in genetics and epigenetics.Tumor heterogeneity leads to different pathological morphologies,cell biological behavior,and biological functions of different tumor cells in same tumor,which are the main factors that enable the tumor cells to expand and adapt to the new microenvironment of the metastatic sites,as well as to develop resistance to chemotherapy.Recent years,the development of single-cell transcriptome studies have made it possible to study the intratumor heterogeneity of tumor cells at the single-cell level,which has greatly promoted our understanding of the process of tumorigenesis and metastasis.However,most of the previous studies on sc RNA-seq of lung cancers mainly focus on cell heterogeneity within the tumor microenvironment and little is known about the heterogeneity within tumor cells.In this study,by single cell RNA sequencing（sc RNA-seq）we generated a single cell transcriptomic map for both human primary invasive lung adenocarcinoma（IADs）and small cell lung cancer（SCLCs）to systematically identify canonical and novel cell types and cell clusters,different cell states and key regulators governing cell transitions,which should allow us to better understand the mechanisms underlying tumor development and progression.Research contents:Sc RNA-seq analysis were perform on 6 primary IAD and 2 primary SCLC samples resected from patients and tumor cell subtypes were identified according to transcriptional characteristics.We then conducted comparative analysis of different cancer cell clusters to identify the key regulators to control the transition of tumor phenotypes,thus,to give us a better understanding of tumor development and progression.Methods:1.We obtained 6 primary IAD and 2 primary SCLC samples resected from patients before target therapy and performed sc RNA-seq analysis on these samples.Major cell types were identified by expression of canonical cell type marker genes andcancer cell fraction was subset for further analysis.2.Functional annotation of different tumor cell subtypes was performed by gene differential expression analysis and gene set enrichment analysis（GSEA）.3.The correlation between tumor subtype specific proteins and tumor prognosis was analyzed using public database.4.In vitro experimental studies including cancer metastasis mouse models,as well as in vitro assays including migration and Matrigel invasion assays,soft agar assay and sphere formation assay were used to verify the phenotypes of different cancer cells and validate the functions of revealed target genes in this study.Results:1.Based on EPCAM expression level,malignant cells were grouped into two main clusters:EPCAM ^loand EPCAM ^hiclusters.The majority of EPCAM ^locells resided in the quiescent state,while EPCAM ^hicells were highly proliferating cells.2.We could identify distinct EPCAM ^locancer cell clusters in all six IAD tumors,indicating that these cells shared similar transcriptional profiles.But EPCAM ^locancer cells from different patients showed different cellular states.Specifically,EPCAM ^lotumor cells in IAD2 expressed VIM,MMP2,ITGA5,FN1,CD44,and ZEB1（a well-known transcription factor that drives EMT）,indicating that these cells were in a mesenchymal state.While in IAD1,IAD3 and IAD5,EPCAM ^lotumor cells expressed HES1,HES6,ID2,SOX2,POU3F1 and other neural development related genes,indicating that these cells initiated a neural development program and proceeded of different stages confirmed by expression of markers for different neural developmental stages.Furthermore,the transcriptional shift from EPCAM ^hito EPCAM ^lowas significantly similar to that from LUAD to SCLC.3.We analyzed the TCGA lung adenocarcinoma RNA-seq dataset and found that high expression of neural signature genes in tumors was significantly correlated with lymphatic metastasis.We selected HES6,a potent transcriptional regulator to promote neural development with specifical expression in EPCAM ^lotumor cells of our sc RNA-seq data,for further study to validate the role of neural development program on cancer progression.HES6 was ectopically expressed in lung cancer,which was negatively correlated with the prognosis of lung cancer patients significantly（p<0.05）.HES6 conferred cancer cells with higher tumor initiating and metastatic capabilities.These data confirmed the role of neural developmental signatures,represented by HES6 expression,in enhancing cancer metastasis.4.The majority of SCLC cells in SCLC1 and SCLC2 had a NE signature with high expression of NE genes.ASCL1 and NEUROD1 are two potent regulators of SCLC differentiation.Most of the NE-SCLC cells showed transcription of these two genes in a mutually exclusive pattern,indicating different transcriptional profiles of cell clusters with expression of these two genes.NEUROD1⁺SCLC cells showed reduced epithelial features but remarkably increased neural migration phenotype than ASCL1⁺SCLC cells.Notably,EPCAM expression was specific in ASCL1⁺SCLC cells but was excluded in NEUROD1⁺SCLC cells,suggesting that EPCAM was a potential cell surface marker to distinguish the ASCL1⁺and NEUROD1⁺SCLC cells.5.Flow cytometry analysis demonstrated that H1155 cells were highly heterogeneous in EPCAM expression.We enriched H1155 cells with high or low expression of EPCAM by FACS.Consistent with the above sc RNA-seq analysis,EPCAM ^hi-H1155 cells showed elevated expression of ASCL1 whereas EPCAM ^lo-H1155 cells showed upregulated NEUROD1.After culturing these enriched cells for days,EPCAM ^hi-H1155（ASCL1⁺SCLC）cells showed some potentials to transform into EPCAM ^lo-H1155（NEUROD1⁺SCLC）cells,confirmed by loss of EPCAM expression in EPCAM ^hi-H1155 cells.Pseudo-time analysis of the sc RNA-seq data further revealed a developmental trajectory from ASCL1⁺SCLC cells to NEUROD1⁺SCLC cells.6.Migration assay and subcutaneous injection mouse model on these enriched cells suggested that NEUROD1⁺SCLC cells showed higher metastatic abilities compared to ASCL1⁺SCLC cells both in vitro and in vivo.Immunostaining on slides of primary and metastatic SCLC tissues confirmed that NEUROD1 was significantly upregulated in human SCLC tissues at metastatic sites and its expression was relatively low in primary human SCLC tissues.Conclusion:By Single-cell transcriptomics analysis we identified a substantial subset of adenocarcinoma cells in the primary tumors,which was specific for the expression of neural stem cell signatures.These cells were demonstrated to show enhanced tumor initiation and metastasis capacities and maight be the precursor cells of SCLC.This is consistent to the clinical phenomenon that transformation into SCLC has been found in 15%of the LUADs.Validation by a set of in vitro and in vivo experiments suggested that LUAD cells might be the drivor of this transition before clinical treatment.SCLC was previously regarded as a“homogenous”disease,but recent studies have indicated that there is considerable heterogeneity among SCLCs.We first generated a single cell transcriptomic map of human primary SCLCs and provided the evidence of heterogeneity in vivo.We found that NEUROD⁺SCLC cells were mostly derived from ASCL1⁺SCLC cells and NEUROD⁺SCLC cells exhibited higher metastatic capability than ASCL1⁺SCLC cells.These findings provided confidential evidence to identify the characteristics of different SCLC subtypes and investigate to new therapeutic strategies for each subtype in clinical.