Tumor-reactive CD8+ tumor-infiltrating lymphocytes (TILs) represent a subtype of T cells that can recognize and destroy tumor specifically. Understanding the regulatory mechanism of tumor-reactive CD8+ T cells has important therapeutic implications. Yet the DNA methylation status of this T cell subtype has not been elucidated. Results
In this study, we segregate tumor-reactive and bystander CD8+ TILs, as well as naïve and effector memory CD8+ T cell subtypes as controls from colorectal cancer patients, to compare their transcriptome and methylome characteristics. Transcriptome profiling confirms previous conclusions that tumor-reactive TILs have an exhausted tissue-resident memory signature. Whole-genome methylation profiling identifies a distinct methylome pattern of tumor-reactive CD8+ T cells, with tumor-reactive markers CD39 and CD103 being specifically demethylated. In addition, dynamic changes are observed during the transition of naïve T cells into tumor-reactive CD8+ T cells. Transcription factor binding motif enrichment analysis identifies several immune-related transcription factors, including three exhaustion-related genes ( NR4A1 , BATF , and EGR2 ) and VDR , which potentially play an important regulatory role in tumor-reactive CD8+ T cells. Conclusion
Our study supports the involvement of DNA methylation in shaping tumor-reactive and bystander CD8+ TILs, and provides a valuable resource for the development of novel DNA methylation markers and future therapeutics. Background
Colorectal cancer (CRC) is one of the most common cancers globally. CRC incidence has traditionally been the highest in affluent western countries, but it is now increasing rapidly in other countries with economic development. CRC treatment usually involves surgical removal of the tumor followed by adjuvant chemotherapy. In recent years, various kinds of immunotherapies, such as checkpoint blockade immunotherapy, have been used to enhance the antitumor potential. However, the responses to these treatments vary among patients. Recent literatures supported the notion that not all tumor-infiltrating lymphocytes (TILs) are tumor reactive [ 1 , 2 , 3 , 4 ]. Rather, bystander cells exist, which recognize a wide range of epitopes unrelated to cancer [ 1 , 2 , 3 , 4 ]. For tumor immunotherapy, it is valuable to target those cells of which the T cell receptor (TCR) repertoire is intrinsically tumor reactive. Co-expression of CD39 (ENTPD1) and CD103 (ITGAE) identifies such a unique T cell population [ 1 , 3 ]. These cells have a tissue-resident memory (RM) signature with high expression of exhaustion markers, such as PDCD1 and HAVCR2 (also known as Tim-3 ). Interestingly, these TILs also exhibited low expression of CCR7 , CD127 , and CD28 , indicative of an effector memory (EM) phenotype [ 3 , 5 ]. Understanding the molecular basis of memory CD8+ T cells is key to developing effective therapies against cancers. Further investigation is needed to better distinguish the molecular natures of T EM and this tumor-reactive T cell subtype.
Gene expression patterns, a key determinant for a cellular feature, are believed to be controlled by epigenetic changes [ 6 ]. Decoding the epigenome specific to tumor-reactive T cells is a pivotal step toward understanding the activation and expansion of this T cell population in cancer. However, how they are regulated epigenetically has not been addressed thus far. DNA methylation, a covalent modification of the DNA molecule, is a stable and heritable form of epigenetic modifications which participates in establishing and maintaining chromatin structures and regulates gene transcription [ 7 ]. In general, DNA methylation is critical for establishing stable gene-silencing programs, by affecting the interactions of DNA with chromatin proteins and transcription factors [ 8 , 9 ]. Many studies have highlighted the importance of DNA methylation in regulating complex gene expression programs underlying immune responses [ 10 , 11 , 12 ]. It is thus important to define how the identities of tumor-reactive CD8+ T cells and bystanders are shaped at methylation level, including particular genes and networks.
In this study, we sorted tumor-reactive and bystander CD8+ TILs from treatment-naïve primary CRC patients based on the expression of CD39 and CD103, and naïve and T EM CD8+ T cells from peripheral blood based on the expression of CD45RO, CD45RA, and CCR7. Adapted smart-seq2 and whole-genome bisulfite sequencing (bisulfite-seq) were performed to characterize the transcriptomic features, DNA methylome programming, methylation dynamics, and key transcription factors (TFs) in each T cell subtype. Our study can help understand the underlying mechanisms leading to the specific expression patterns of tumor-reactive CD8+ T cells, thereby facilitating the development of new therapeutic strategies targeting these cells. Results Transcriptomic characteristics of five CD8+ T cell subtypes
Within CD8+ TILs, CD103+CD39+ T cells have been recently demonstrated to be tumor-reactive, while CD103−CD39− T cells and CD103+CD39− T cells are bystanders [ 1 , 3 ]. To further characterize the transcriptional profiles of these cell populations, we isolated naïve, T EM , CD103+CD39+, CD103+CD39−, and CD103−CD39− T cell subtypes from eight CRC patients for gene expression profiling using adapted Smart-seq II (Fig. 1 a–c; S1A, B; As shown in the heat map displaying differentially expressed genes (DEGs) among five CD8+ T cell subtypes, the naïve subtype exhibited high expression of known naïve markers LEF1 and SELL (also known as CD62L ) (Fig. 1 d; S1C). T EM subtype showed enhanced expression of classically defined T EM molecules, such as TBX21 [ 13 ] and CX3CR1 [ 14 ] (Fig. 1 d). Notably, CD103+CD39+ TILs displayed hallmarks of an “exhausted” phenotype, with high expression of CTLA4 , HAVCR2 , and LAYN (Fig. 1 d; S1C, D). Recent literatures reported that the thymocyte selection-associated high mobility group box (TOX) protein is required for the development and maintenance of exhausted T cell populations in chronic infection [ 15 , 16 , 17 , 18 ]. Removal of its DNA binding domain reduced the expression of PD-1 and resulted in a more polyfunctional T cell phenotype [ 16 ]. Here, we observed that TOX expression is also upregulated (Fig. 1 d; S2A). Intriguingly, our previous single-cell RNA-sequencing (scRNA-seq) data identified the specific expression of TOX in exhausted CD8+ TILs [ 19 , 20 , 21 ] S2B-D). These data together supported the important role of TOX in intratumoral T cell exhaustion. Fig. 1
Comparative transcriptional analysis reveals tumor-reactive CD8+ T cells to have a T RM signature with high expression of exhaustion markers. a Experimental design for the isolation of different CD8+ T cell populations from CRC patients. b , c Representative plots of FACS-isolated T cell populations. d Gene expression heat map of five CD8+ T cell populations. Rows represent signature genes, and columns represent cell types. Selective specifically expressed genes are marked in red. e GSVA was performed to identify enriched significant biological pathways in five CD8+ T cell subtypes. Five gene sets for each T cell population are depicted in a heat map. f PCA analysis of transcriptome expression of five CD8+ T cell populations. Each symbol represents one patient. g Volcano plot showing differential gene expression of CD103+CD39+ T cells vs. CD103−CD39− T cells (log2-transformed). Each red dot denotes an individual gene with a false-discovery rate (FDR) 99% (or almost 99%) bisulfite conversion rates were retained for the DNA methylation analysis. Most post-alignment analysis was performed by functions from MethPipe [ 53 ] Package.
Methylation levels for each symmetric CpG site were calculated by the methcounts and symmetric-cpgs commands in MethPipe. Average CpG methylation levels of about 5 kb tils for all chromosomes in human genome were calculated with MethPipe roimethstat command for each T cell population. Only high-confidence genomic regions with at least 40 CpG observations from reads in the genomic tile were used for the PCA analysis (R prcomp function) to compare the overall methylation level in genome wide of the five T cell populations.
Specific hypomethylated regions (HypoMRs) for each T cell population were calculated in a “one vs. rest” fashion by using radmeth regression , radmeth adjust , and radmeth merge commands in MethPipe. Then, we overlapped these HypoMRs with promoter regions (defined as − 2.5 kb and 1 kb from transcription start site (RefSeq gene model downloaded from UCSC Table browser [ 54 ])) to identify genes affected by HypoMRs in each T cell population. The average methylation level of all HypoMRs in each T cell population was calculated by roimethstat command. To visualize the methylation pattern of given genes in different populations, lollipop plot from CGmapTools [ 55 ] was used. The Wilcoxon test was applied to compare the methylation levels of HypoMR between each group, with p value
Also in Industry News
How to decide whether or not to start treatment for prostate cancer?
Analysis of the SARS-CoV-2 proteome via visual tools
$65m investment increases British Patient Capital’s exposure to life sciences and health technology