Single-cell Long Non-coding RNA Landscape of T Cells in Human Cancer Immunity

Abstract

The development of new biomarkers or therapeutic targets for cancer immunotherapies requires deep understanding of T cells. To date, the complete landscape and systematic characterization of long noncoding RNAs (lncRNAs) in T cells in cancer immunity are lacking. Here, by systematically analyzing full-length single-cell RNA sequencing (scRNA-seq) data of more than 20,000 libraries of T cells across three cancer types, we provided the first comprehensive catalog and the functional repertoires of lncRNAs in human T cells. Specifically, we developed a custom pipeline for de novotranscriptome assembly and obtained a novel lncRNA catalog containing 9433 genes. This increased the number of current human lncRNA catalog by 16% and nearly doubled the number of lncRNAs expressed in T cells. We found that a portion of expressed genes in single T cells were lncRNAs which had been overlooked by the majority of previous studies. Based on metacell maps constructed by the MetaCell algorithm that partitions scRNA-seq datasets into disjointed and homogenous groups of cells (metacells), 154 signature lncRNA genes were identified. They were associated with effector, exhausted, and regulatory T cell states. Moreover, 84 of them were functionally annotated based on the co-expression networks, indicating that lncRNAs might broadly participate in the regulation of T cell functions. Our findings provide a new point of view and resource for investigating the mechanisms of T cell regulation in cancer immunity as well as for novel cancer-immune biomarker development and cancer immunotherapies.

Keywords: Functional annotation; Immune regulation; Long non-coding RNA; Metacell; Transcriptome assembly.

Figure 1

Statistical analysis of assembled transcripts and workflow for novel lncRNA identification process in T cells during cancer immunity A. Violin plots showing the number of assembled transcripts (All) and the number of those matched to the reference (Match) at single-cell level across five HCC patients (P0205, P0322, P0407, P0508, and P1116). B. Number of assembled transcripts matched to reference across five HCC patients based on four different strategies. ***, P < 0.001 (Wilcoxon rank-sum test). C. Scheme of the pipeline used to identify novel lncRNAs expressed in T cells during cancer immunity using three full-length scRNA-seq datasets. D. Correlation of the number of cells and the number of assembled transcripts across different subsets for CRC, HCC, and NSCLC. 95% confidence intervals were added and shown as colored shades. E. The statistics of assembled transcripts that matched to reference protein-coding genes and reference lncRNA genes. CRC, colorectal cancer; HCC, hepatocellular carcinoma; NSCLC, non-small-cell lung cancer; P, peripheral blood; N, adjacent normal tissue; T, tumor tissue; CPC, Coding Potential Calculator; CNCI, Coding-Non-Coding Index.

Figure 2

Validation of novel lncRNAs using qRT-PCR A. Flow cytometric analysis for CD8 T cells. T cells from three NSCLC patients were separated by magnetic beads and stained with antibody CD8-APC. B. Flow cytometric analysis for CD4 T cells. T cells from three NSCLC patients were separated by magnetic beads and stained with antibody CD4-APC. Isotype was used as a negative control. C. An example of novel intergenic lncRNA that was validated by Sanger sequencing. The genomic views are generated from the UCSC Genome Browser. The spliced sequence outputted by Sanger sequencing is shown.

Figure 3

Expression features of reference genes and novel lncRNA genes at the single-cell level A. Number of protein-coding, reference lncRNA, and novel lncRNA genes expressed in T cells across three cancer types. ***, P < 0.001 (Wilcoxon rank-sum test). B. Plots showing the percentage of expressing cells against the mean expression level (log counts) for protein-coding, reference lncRNA, and novel lncRNA genes across three cancer types. The number of genes that are expressed in at least 25% of the cells is provided in the figure.

Figure 4

Characterization of T cell states based on the 2D projection of T cells and the annotation of metacell maps A. 2D projection of CD8 T cells from three cancer types into 43 metacells. B. 2D projection of CD4 T cells from three cancer types into 65 metacells. In panels A and B, metacells are indicated as nodes, which are numbered (1–43) and color-coded for differentiation. Cells are shown as small data points, which are color-coded according to the metacells they belong to. C. CD8 metacells (rows) ordered by groups and organized within each group. D. CD4 metacells (rows) ordered by groups and organized within each group. The left bar plot shows the number of cells of different clusters in each metacell. The middle and right bar plots show the percentage of cells from different cancer types (CRC, HCC, and NSCLC) and tissues (P, N, and T) in each metacell, respectively. Heatmaps show the confusion matrices (the pairwise similarities between metacells) for CD8 or CD4 metacells. The annotations of different metacell groups are shown on the right. E. 2D projections of the composition of CD8 T cells from different clusters. F. 2D projections of the composition of CD4 T cells from different clusters. In panels E and F, cells from different clusters are colored-coded according to the metacells they belong to.

Figure 5

Correlation and expression analyses of signature lncRNAs associated with different T cell states A. Pearson correlation heatmaps for signature lncRNA genes and anchor genes in CD8 metacells. B. Pearson correlation heatmaps for signature lncRNA genes and anchor genes in CD4 metacells. The signature gene modules and two anchor genes (CTLA4 and FOXP3) are labeled on the right. C. Expression (log fold enrichment values; lfp values) of signature lncRNA genes and anchor genes across CD8 metacells. Signature lncRNA genes and anchor genes are marked in black and red on the right, respectively. Metacells are numbered at the bottom. Metacell groups associated with effector and exhausted functions are underlined with yellow and green bars, respectively.

Figure 6

Functional annotation of signature lncRNAs A. Functional enrichment maps of signature lncRNAs for CD8 effector/exhausted T cells. B. Functional enrichment maps of signature lncRNAs for CD4 effector/exhausted T cells. C. Functional enrichment maps of signature lncRNAs for CD4 Treg cells. The enriched gene sets from Gene Ontology based on the predicted functions of signature lncRNA genes are visualized by Cytoscape plugin Enrichment Map. Each node represents a gene set; the size of the node is indicative of the number of genes and the color intensity of the node reflects the level of significance. Effector signature gene sets are shown in red circles, exhausted or Treg ones in green circles, and the common gene sets in orange circles. Maps are magnified differently for easy visualization.

Figure 7

Genomic and functional characterization of example signature lncRNAs A. Genomic view of a known effector signature lncRNA TM4SF19-AS1. The genomic view is generated from the UCSC Genome Browser. B. Co-expressed genes of TM4SF19-AS1. Co-expressed protein-coding genes, reference lncRNA genes, and novel lncRNA genes are colored in pink, light green, and light yellow, respectively. C. Functional annotations of TM4SF19-AS1 based on co-expression network. D. Genomic view of a novel exhausted signature lncRNA XLOC-633950. The genomic view is generated from the UCSC Genome Browser. E. Co-expressed genes of XLOC-633950. Co-expressed protein-coding genes, reference lncRNA genes, and novel lncRNA genes are colored in pink, light green, and light yellow, respectively. F. Functional annotations of XLOC-633950 based on co-expression network.