RNA sequencing: new technologies and applications in cancer research

Genomic data, such as RNA-seq, have become widely available due to the popularity of high-throughput sequencing technology [59]. As an important part of next-generation sequencing, RNA sequencing has made great contributions in various fields, especially cancer research, including studies on differential gene expression analysis and cancer biomarkers, cancer heterogeneity and evolution, cancer drug resistance, the cancer microenvironment and immunotherapy, neoantigens, etc. (Fig. 5).

Differential gene expression analysis is one of the most common applications of RNA sequencing [60]. Samples from different backgrounds (different species, tissues and periods) can be used for RNA sequencing to identify differentially expressed genes, revealing their function and potential molecular mechanisms [61]. More importantly, differential gene expression analysis facilitates the discovery of potential cancer biomarkers [62]. Many studies have shown that gene fusions are closely related to oncogenesis and are appreciated as both ideal cancer biomarkers and therapeutic targets [63]. Gene fusions in clinical samples are primarily detected by RNA-CaptureSeq. Compared to whole transcriptome sequencing, RNA-CaptureSeq has significantly higher sequencing depth [27, 64, 65]. It has been reported that the NUP98-PHF23 fusion gene is likely to be a novel therapeutic target in acute myeloid leukemia (AML) [66]. Recently, a variety of recurrent gene fusions, including ESR1-CCDC170, SEC16A-NOTCH1, SEC22B-NOTCH2 and ESR1-YAP1, have been identified in breast cancer, indicating that recurrent gene fusion is one of the key drivers for cancer [67]. Several novel configurations of  BRAF, NTRK3 and RET gene fusions have been identified in colorectal cancer [68]. These fusions may promote the development of malignancy and provide new targets for personalized treatment [68]. In addition, some special genomic factors have been discovered as biomarkers by RNA sequencing, including miRNA, lncRNA and circRNA, which are widely present in various types of cancer [69‐71]. A recent example is circRNA_0001178 and circRNA_0000826, which are biomarkers of colorectal cancer metastasis to the liver [72]. By applying both RNA sequencing and small RNA sequencing, a study on pancreatic cancer identified differential expression of simple repetitive sequences (SSRs) and demonstrated that the frequency of SSR motifs changed dramatically, which is expected to become a tumor biomarker [73]. In addition to nucleic acid biomarkers, RNA-seq combined with immunohistochemistry and western blot has also identified certain proteins as cancer biomarkers, such as nuclear COX2 (cyclooxygenase2) in combination with HER2 (human epidermal growth factor receptor type 2), which may serve as potential biomarkers for the diagnosis and prognosis of colorectal cancer [74]. Similar examples identified using RNA-seq profiling analysis include ISG15 (Interferon-stimulated gene 15) in nasopharyngeal carcinoma [75] and DMGDH (dimethylglycine dehydrogenase) in hepatocellular carcinoma [76]. Data-mining analysis of RNA sequencing data and other clinical data has identified that the isoforms of peroxiredoxins also can be expected the prognostic biomarkers for predicting overall survival and relapse-free survival in breast cancer [77]. Increasing differentially expressed genes are being identified by RNA sequencing, and new potential cancer biomarkers are being continuously discovered (Table 2). However, sufficient clinical practice is needed to confirm the diagnostic and predictive applications of these biomarkers in cancer.

RNA-seq could detect early mutations as well as high molecular risk mutations, thus can discover novel cancer biomarkers and potential therapeutic targets, monitoring of diseases and guiding targeted therapy during early treatment decisions. Tumor mutation burden (TMB) is considered as a potential biomarker for immune checkpoint therapy and prognosis [78, 79]. RNA-seq can be used to explore the application value of TMB in diffuse glioma [78]. Through the RNA-seq, MET exon 14 mutation and isocitrate dehydrogenase 1 (IDH1) mutation were identified as new potential therapeutic targets in lung adenocarcinoma and chondrosarcoma patients, respectively [80, 81]. Several studies have shown that RNA sequencing can effectively improve the detection rate on the basis of DNA sequencing, provide more comprehensive detection results and achieve a better curative effect for targeted therapy [82]. In addition, it has been proved that IDH mutation is a good prognostic marker for glioma by RNA-seq [83]. Targeted therapy is also considered to enhance or replace cytotoxic chemotherapy regimen in cancer including AML [84‐86].

ScRNA-seq also has some new discoveries in diagnosis. For example, scRNA-seq data can be used to infer copy number variations (CNV) and to distinguish malignant from non-malignant cells. The infer CNV algorithm, which was used in the study of glioblastoma, uses averaging relative expression levels over large genomic regions to infer chromosome copy number variation [87]. Similar examples include head and neck cancer [88] and human oligodendroglioma [89]. It is reported that RNA sequence of tumor-educated blood platelets (TEPs) can also become a blood-based cancer diagnosis method [90]. It should be noted that the lack of detailed functional implications of the identified RNAs in platelets in the field of platelet RNA research is also an urgent problem to be solved [91].

Heterogeneity has always existed during the transformation of normal cells to cancer cells. The continuous accumulation of heterogeneity may reflect the evolution of cancer [109]. Early RNA sequencing detected all RNA transcripts in a given tissue or cell group, ignoring differences in individual cells. Transcriptome profiling of single-cell RNA sequencing solves this problem by providing single-cell resolution of the transcriptome [3]. In melanoma, single-cell RNA-seq was used to analyze 4645 tumor cells from 19 patients, including cancer cells, immune cells, mesenchymal cells and endothelial cells. Transcriptomic data from different single cells revealed that heterogeneity of cells within the same cancer is associated with cell cycle, spatial background and drug resistance [110]. A recent single-cell RNA-seq study of 49 samples of metastatic lung cancer revealed changes in plasticity induced by non-small cell lung cancer treatment, providing new directions for clinical treatment [111]. Single-cell RNA sequencing also integrates a variety of information in a single cancer cell, deciphering the secrets of cancer heterogeneity and evolution [112]. Compared with scRNA-seq, another emerging technology spatial transcriptome sequencing incorporates information on the spatial location of cells. In prostate cancer, using spatial transcriptomics technology, the transcriptome of nearly 6750 tissue regions was analyzed, revealing the whole-tissue gene expression heterogeneity of the entire multifocal prostate cancer and accurately describing the range of cancer foci [113]. In a study of breast cancer tissues, the results of spatial transcriptome sequencing revealed that gene expression among different regions was surprisingly highly heterogeneous [22]. In recent years, single-nucleus RNA sequencing (snRNA-seq) has also received extensive attention due to its solving the problem that single-cell RNA sequencing cannot be applied to frozen specimens and cannot obtain all cell types in a given tissue [114, 115]. The emerging technology of RNA sequencing will contribute to research on cancer heterogeneity and evolution.

Drug resistance is a main reason leading to cancer treatment failure. However, the molecular mechanisms underlying drug resistance are still poorly understood [116]. RNA sequencing became a vital tool for revealing the mechanisms of cancer drug resistance. In breast cancer, single-cell RNA sequencing identified a tumor-infiltrating immunosuppressive immature myeloid cell that leads to drug resistance [117]. Another study identified a new COX7B gene related to platinum resistance and a surrogate marker CD63 in cancer cells by single-cell RNA-seq [118]. RNA sequencing has also demonstrated that cancer cells that wake up from a dormant state produce large amounts of BORIS (brother of the regulator of imprinted sites), which can regulate the expression of survival genes in drug-resistant neuroblastoma cells [119]. Identifying special molecules that mediate these processes could help us understand the occurrence of drug resistance. Single-cell transcriptomics can be used to study different modes of chemoresistance in tumor cells and has shown that pre-existing drug-resistant cells can be selected through higher phenotypic intratumoral heterogeneity, while phenotypic homogeneous cells use other mechanisms to trans-differentiate under drug-selection [120]. In one study of pancreatic ductal adenocarcinoma, human pancreatic cancer (PANC-1) cells and gemcitabine-resistant PANC-1 cell lines were compared by RNA sequencing, and two circRNAs were identified as both novel biomarkers and potential therapeutic targets for gemcitabine resistant patients [121]. RNA-seq has also conducted in-depth research on the drug resistance of hematological malignancies. Through RNA-seq, it has been found that non-coding RNAs and fusion genes play an important role in mediating the drug resistance of hematological malignancies [122]. A good example is to compare the circRNA expression profile of the drug-resistant acute myeloid leukemia cell with its parent cell, and determine the circRNAs involved in drug resistance [123]. Similarly, the novel MEF2D-BCL9 fusion transcript identified by RNA-seq was found to increase HDAC9 (histone deacetylase 9) expression and to enhance the resistance to dexamethasone in acute lymphocytic leukemia (ALL) [124]. Leukemia stem cells (LSCs), a rare cell population assumed to be responsible for relapse, is crucial to improve the prognosis of patients [125, 126]. RNA-seq analysis showed that LSCs have a unique lncRNA signature with functional relevance and therapeutic potential, providing an explanation for chemotherapy resistance and disease recurrence [127].

The immune system plays a critical role in the cancer microenvironment, affecting several stages of cancer development, including tumorigenesis, progression and metastasis, through tumor-infiltrating lymphocytes (TILs) [128]. TILs and their interactions with malignant cells and stromal cells make up the cancer immune microenvironment. Due to the heterogeneity of cancer, it is difficult to define the exact pro- or anti-cancer function of certain immune cells. Cancer heterogeneity also causes the varied clinical efficacy observed in patients treated with immunotherapies due to different responses of different subclones [129]. Transcriptomic profiling by RNA-seq, in particular scRNA-seq, provides comprehensive information on cellular activity and interactions among cells in the tumor microenvironment (TME). ScRNA-seq enables genomic and molecular profiling of high quantity and quality individual immune cells and assessment of cellular heterogeneity to depict the immune system spectrum in the cancer microenvironment [130‐132]. ScRNA-seq data demonstrated that compared to normal tissues, cancer tissues exhibited significantly higher heterogeneity in the immune microenvironment, and a continuity in T cell activation resulting from polyclonal T cells and heterogeneous antigen-presenting cells has been identified [133].

ScRNA-seq of tumor-infiltrating T cells in metastatic melanoma identified transcription factor NFATC1 (nuclear factor of activated T cells 1) as a potential molecular signature of T cell exhaustion programs and revealed the depletion of low-exhaustion T cells in expanded clones of T cells [110]. Combining scRNA-seq with assembled T cell receptor (TCR) sequences, 11 T cell subsets, such as CD8 ⁺ T cells and CD8⁺ FOXP3 ⁺ regulatory-like cells, and their genomic signatures, were identified in hepatocellular carcinoma (HCC), providing valuable insights for understanding the immune landscape of infiltrating T cells in HCC [134]. Cancer infiltrating T cells also play an anti-tumor role through impairment of an autophagy protein, LC3 (microtubule-associated protein 1A/1B-light chain 3, often short for LC3)-associated phagocytosis (LAP), demonstrating the role of autophagy in oncogenesis and suppression revealed by scRNA-seq [135]. In addition to solid tumors, scRNA-seq of acute myeloid leukemia patients detected diverse immunomodulatory genes that suppress T cell function [136]. By CSOmap, a computational tool for scRNA-seq, the CCL4-CCR8 directed interaction between Tregs and Texs, as well as reduced proliferation of Texs, was characterized [137]. Notably, findings also revealed that tumor-infiltrating T cells exhibited more interactions among themselves than with T cells from peripheral blood and different interactions between tumors and T cells, indicating a varied response to immunotherapy and a potential trend for immune escape [137].

In the cancer immune microenvironment, neutrophils, in addition to T cells, are also key components of cancer progression and cancer drug resistance [138‐141]. Through scRNA-seq of murine sarcomas and certain human cancers, neutrophils with CSF3R (colony stimulating factor 3 receptor) expression were found to be a part of type 1 antitumor immunity associated with unconventional CD4⁻CD8⁻αβ T cells (UTCαβ) in anti-cancer immunity, indicating better prognosis [142]. ScRNA-seq of metastatic breast cancer and CD45 cells from primary cancer identified neutrophils as pro- and anti-tumorigenic or metastatic, in which pro-tumorigenic and metastatic neutrophils are induced by IL11 expressing cancer subclones, resulting in polyclonal metastasis [143]. This observation also provides new insight into anti-cancer immunotherapy by targeting neutrophils [143]. With scRNA-seq of CD4 and CD8 T cells, several crucial pathways with anti-cancer function were revealed [144].

The balance between immune reaction and immune tolerance is the basis of immune homeostasis, which is also involved in anti-cancer immunity and oncogenesis. scRNA-seq of monocytes and dendritic cells (DCs) separated from a single lymph node melanoma metastasis revealed a conserved homeostatic module regulated by suppressor-of-cytokine-2 (SOCS2) protein and IFNγ [145]. SOCS2 serves an essential regulatory role in anti-tumor immunity and T cell priming through DCs. This highly conserved homeostatic program establishes a connection between autoimmune prevention and immune surveillance in cancer [145].

Immunotherapies, especially immune checkpoint blockade (ICB), has opened a new chapter for anti-cancer therapy with remarkable responses from targeting programmed death 1 (PD1), programmed death-ligand 1 (PD-L1) and cytotoxic T-lymphocyte-associated protein 4 (CTLA4) [146‐148]. However, only a few patients benefit from ICB, and severe side effects were observed [149, 150]. Obviously, various unknown determinants are correlated with the outcome of immunotherapies in addition to well-known factors such as PD1/PD-L1/CTLA-4 expression and mismatch repair deficiency [151‐155]. Therefore, it is paramount to identify potential effectors for ICB efficacy. By analyzing RNA-seq data from melanoma patients who underwent anti-PD1 and anti-CTLA4 treatment, a potential ICB resistance effector SERPINB9 (a member of the serine protease inhibitor (serpin) family) and the connection between cytotoxic T lymphocytes (CTL) infiltration level and ICB response were characterized [156]. A sub-population cells with immunotherapy persistence have been identified by scRNA-seq and were found to have stem cell-like states with the expression of stem cell antigen-1 (Sca-1) and Snai1 [157].

Another immunotherapy, myeloid-targeted immunotherapy, is based on the complexity of tumor-infiltrating myeloid cells, including DCs and tumor-associated macrophages (TAMs) revealed by scRNA-seq [158]. Through scRNA-seq of immune cells from colorectal cancer patients, C1QC⁺ and SPP1⁺ TAMs, two subsets of TAMs, were identified, and the mechanism of myeloid-targeted immunotherapy, such as anti-CSF1R (colony stimulating factor 1 receptor) and CD40 agonist, was revealed [159]. Intracellular staining and sequencing (INs-seq), a novel technology integrating scRNA-seq and intracellular protein activity measurements, revealed novel Arg1⁺ Trem2 ⁺ regulatory myeloid (Mreg) cells and demonstrated that depletion of Trem2 led to deduction of exhausted CD8 T cells with increased NK and cytotoxic T cells and cancer suppression by reducing accumulation of intratumoral Mreg cells [160].

Neoantigens, human leukocyte antigen (HLA)-bound peptides derived from cancer-specific somatic mutations or gene fusions during tumor growth, are another crucial regulator of the clinical response to immunotherapy [161]. Higher intratumor neoantigen heterogeneity and clonal neoantigen burden increases sensitivity to ICB and contributes to better clinical outcome in patients with melanoma and advanced non-small cell lung cancer [162]. This kind of antigen is an optimal target for anti-cancer immunotherapy, enhancing neoantigen-specific T cell activity, and a vaccine targeting personal neoantigens for melanoma patients has been developed [163, 164]. Given all these promising features of personalized medicine targeting neoantigens in tumors, massive parallel profiling of tumor neoantigen burden is necessary for improving clinical efficacy and a deeper understanding of the neoantigen landscape. An RNA-seq-based transcriptomic approach is an efficient tool for neoantigen profiling in many studies. It was revealed that homology of neoantigen and somatic-mutation induced pathogens are important in response prediction in anti-CTLA4 treated melanoma [165]. In addition to melanoma, other studies found reduced neoantigen load in triple-negative and HER2 breast cancers [166], diverse neoantigen abundance in non-small-cell lung cancer patients with different treatment strategies [167], a decreased ratio of neoantigen expression to predicted neoantigens in recurrent glioma due to immune selection pressure [168], a negative correlation between neoantigen abundance and clinical outcome in selected solid tumors [169] and different neoantigen landscapes in immune filtration and T cell dysfunction based on histology in salivary gland carcinoma (SGC) patients [170].

A neoantigen prediction program, Neopepsee, based on RNA-seq data and somatic mutation, can be utilized to detect potential neoantigens for personal vaccine development with reduced false-positive rate compared to binding affinity prediction [171]. ScanNeo is another prediction computational pipeline based on RNA-seq that aims to identify insertion and deletion derived neoantigens, which was validated in prostate cancer [172], and ASNEO, which identifies personal-specific alternative splicing derived neoantigens [173]. Several neoantigens have been identified to be related to cancer prognosis and might be potential targets of immunotherapies, such as the TP53 neoantigen for HCC patients [174]. For the anti-neoantigen immunotherapy in cancer, a new strategy involving neoantigen-specific TCRs modification has been proposed, and scRNA-seq has been applied to isolate neoantigen-specific TCRs for further clinical application [175].