The main feature of lung cancer revealed by scRNA-seq is intratumoral and intertumoral heterogeneity. The scRNA-seq can make more precise diagnosis and prognostic predictions, and facilitate the development of new anti-lung cancer agents. For example, Wu F et al. (35) utilized scRNA-seq and analyzed 42 samples from patients with advanced NSCLC at various stages. They found that the intertumoral and intratumoral heterogeneity of lung adenocarcinoma (LUAD) was lower than that of lung squamous cell carcinoma (LUSC). In addition, cancer cells from different patients exhibited higher heterogeneity. Next, the researchers used scRNA-seq data to infer copy number changes (CNAs) in the cancer cell population and to reveal heterogeneity between and within patients. Most patients, especially LUAD patients, had dominant clones, while in a few LUSCs, malignant cells were distributed in multiple clusters. To quantify intratumoral heterogeneity, they obtained expression-based and CNA-based intratumoral heterogeneity scores (ITH), which were designated ITH_GEX and ITH_CNA, respectively. The patients were further divided into three categories in terms of lung cancer types and mutations: LUAD patients with driving mutations (LUADm), LUAD patients without driving mutations (LUADn) and LUSC patients without driving mutations (LUSCn). Their results showed that, compared with LUADm patients, LUSCn patients had significantly higher ITH_CNA, but no significant difference was found in ITHGEX. ScRNA-seq demonstrated a transcriptional heterogeneity within the malignant cell population, which was associated with driving mutations. Identification of alterations in more diverse subpopulations may have implications for immunotherapy.

It is also feasible to use scRNA-seq to identify ubiquitous tumor cells with specific transcriptomic status in lung cancer patients, which helps us better understand the tumor type and cell hierarchy of lung cancer, and single out transcriptome signature related to the response and resistance to treatment (23, 58). Mounting evidence shows that EGFR mutation is an important factor affecting the therapeutic efficacy of PD-1 inhibitors in NSCLC patients, and patients with EGFR mutation responded to the treatment less well than their counterparts without EGFR mutation (59–62). He D et al. (23), by employing scRNA-seq, revealed a significant heterogeneity in EGFR mutation in patient with early-stage LUAD, and found that ELF3 was one of the most up-regulated genes in advanced tumor cells. Under the effect of immune infiltrating inflammatory cytokines (such as IL1B), ELF3 in tumor cells was up-regulated, thereby activating PI3K/Akt/NF- κB pathway, and up-regulating the expression of proliferation and anti-apoptosis genes, such as BCL2L1 and CCND1. These findings suggested that there existed an involved interaction among tumor cells, stromal cells and immune infiltration cells in TME. These results may pave the way to immunotherapy targeting EGFR mutant LUAD.

Meanwhile, scRNA-seq analysis has multiple advantages in that it not only can reveal the molecular diversity of different samples and show the impact of clinical treatment on different cell subsets. Ke-Yue Ma, et al. (24) utilized scRNA-seq to examine the heterogeneity of genes associated with response of LUAD to immunotherapy. They compared LC2/ad (Vandetanib sensitive) and LC2/ad-R (Vandetanib tolerant) cell lines, and found that LC2/ad had a higher level of MHC II gene and IFN-γ signal pathway coexpression gene. However, the IFN-γ signaling pathway in LC2/ad-R was down-regulated and the expression of MHCII gene was low. They revealed a possible mechanism of Vandetanib resistance: that is, MHC II and IFN-γ signaling pathways jointly determined the development of immunotherapy resistance.

Tian Y, et al. (25), by using scRNA-seq, examined about 5000 matched normal adjacent tissues (NAT) and primary tumors (PT) cells from 11 SCLC patients (including a patient with both primary tumor (PT) and recurrent tumor (RT)). They found that human SCLC had a significant inter-tumor and intra-tumor heterogeneity, and many tumors contained separate subpopulations, indicating there is a remarkable intra-tumor heterogeneity at the transcriptomic level. In addition, most SCLCs with neuroendocrine (NE) characteristics, such as SCLC-N and SCLC-A, tended to have strong immunological features, while non-NE SCLCs, such as SCLC-P and SCSC-Y, tended to possess weak immune traits. Patients with SCLC having weak immune features were more likely to benefit from immune checkpoint blockade (ICB) than their counterparts with strong immune characteristics. The scRNA-seq data revealed that multiple SCLC subtypes showed different proportions in practically all SCLC patients, highlighting the importance of scRNA-seq and the need for functional research on tumor progression and immunotherapy of ITH. Moreover, Chan JM et al. (26) applied scRNA-seq to analyze the transcriptome of 21 fresh SCLC samples from 19 patients and 155098 cells from 24 LUAD samples and 4 normal lung tissue samples from the area adjacent to cancer. They found that the level of copy number variation (CNV) was higher in SCLC than in LUAD and had significant heterogeneity. They exhibited that PLCG2 overexpression subsets were associated with metastasis, immunosuppression and poor prognosis. Therefore, it is potentially of great significance for the design of novel strategies of the targeted therapy and immunotherapy. These observations collectively demonstrated that ScRNA-seq can help researchers better understand tumor heterogeneity and the intricate interactions between tumor cells and their microenvironment, thereby facilitating the identification of lung cancer cell subpopulations amenable to immunotherapy.

Infiltrating immune cells, cancer-associated fibroblasts and vascular endothelial cells are important components of TME (63). By means of scRNA-seq, we can identify the features of various types of immune, stromal cells, heterogeneous expression profiles, and look into mechanisms involved in immunosuppression, thus better understanding the heterogeneity and diversity of cancer immune responses.

In view of the intricacies of the TME, in recent years, single-cell transcriptome sequencing has been incrementally used in the study of infiltrating immune cells in lung cancer (25, 27–31, 45). It has an important implication for the research of the mechanism underlying the lung cancer tumor immunity, especially for the study of the functional status of T cells in the tumor, which play a pivotal part in killing tumor cells, for the development of efficacious immunotherapy and the identification of sensitive targets and markers. Guo X et al. (27) conducted scRNA-seq on 12346 T cells from the peripheral blood, cancer-adjacent tissues and cancer tissues of 14 NSCLC patients prior to drug treatment, worked out the immune map of lung cancer T cells at the single-cell level, revealed the heterogeneity of lung cancer T cells, and provided a new notion for immunotherapy to specifically target T cell subsets. This study identified 16 major clusters of T cells (7 CD8 and 9 CD4 types). In addition to exhausted CD8 T cells, the infiltrating CD8 T cell population of lung cancer was also found to include two groups of “pre-exhausted” CD8 T cells that may bear a state transition relationship with exhausted CD8 T cells. Moreover, higher ratios of “pre-exhausted” and exhausted CD8 T cells were associated with a more favorable prognosis for lung adenocarcinoma. Apart from that, in terms of the expression of TNFRSF9 (4-1BB), a group of activated Tregs could be distinguished from lung cancer-infiltrating regulatory T cells (Tregs). The expression of inhibition-related genes in this group of Tregs was higher, suggesting that they were the Treg cells that actually serve the inhibition function in tumors. At the same time, the proportion of activated Treg cells was negatively correlated with the prognosis of lung adenocarcinoma.

In NSCLC patients receiving ICI, Zhang Y et al. (28) found that patients with MET gene amplification were refractory to the treatment. After analysis of more than 20000 immune cells with scRNA-seq, the researchers identified a new XTIST/CD96/KLRG1 triple positive NK cell subpopulation in patients with MET amplification. In immunotherapy-resistant patients, the proportion of this subpopulation was elevated and the proportion of NK cells and CD8+T cell subpopulations dropped. Moreover, some researchers used scRNA-seq to study the dynamic change of peripheral blood T cell clones in NSCLC patients receiving PD-1 inhibitors (45), and found that the number of a CD4+T cell clone related to tumor significantly dropped upon tumor progression, and the proportion of PD-1+T cells also decreased significantly. What’s more, an NSCLC patient with negative PD-L1 expression benefited from the treatment of pabolizumab (PD-1 inhibitor) (29). An scRNA-seq analysis of the patient’s peripheral blood revealed that the NKG7+NK cells and NKG7+T (NKT) cells of the patient were significantly lowered, while the CD8+T cells and Naive T cells were prominently increased, suggesting that the change in the percentage of NK cells and T cells might be related to the efficacy of pabolizumab treatment.

Liu S et al. (30) examined whether the patient’s previous response to EGFR-TKI was related to the subsequent immunotherapy results. They found that, in patients receiving TKI (Tyrosine Kinase Inhibitor) treatment, the objective response rate (ORR) of immunotherapy was significantly higher in patients with short progression free survival (PFS) than in those with long PFS. By comparing the TME of the two groups using scRNA-seq, the researchers found that the infiltration rate of INF-γ+CD8+T cells and CD8+T cells in the immune microenvironment was higher in patients with short PFS, and the ratio between M2- and M1-like macrophages was significantly lower in short-PFS patients than in their counterparts with long PFS. Therefore, this study provided a marker reference from the angle of a single cell for the ensuing treatment in patients who had received EGFR-TKI targeted therapy. scRNA-seq analysis by Yang L et al. (31) revealed that the TME of EGFR mutant LUAD and wild type LUAD had different heterogeneity in cell composition, function and their interaction. The loss of proinflammatory cells, enrichment of inhibitory cell types and the low expression of immune checkpoint proteins may lead to an immune silence environment for EGFR-mutated LUAD, i.e., EGFR-mutated tumor cells secrete cytokines to recruit various immunosuppressive cells, while activated immune cells (CXCL9+TAM and CD8+TRM) were seriously insufficient. Therefore, in future, effective immunotherapy can be accomplished in EGFR mutant LUAD patients by improving the inhibitory tumor immune microenvironment (TIME).

Tian Y, et al. (25) employed scRNA-seq to develop a detailed immunity map of SCLC. Compared with normal adjacent tissues, the proportion of lymphocytes in primary SCLC was higher and the proportion of myeloid cells lower, indicating that adaptive immunity in TME played a more important role. Then, the researchers re-classified T cells and myeloid cells and categorized macrophage into four groups: a tumor-associated macrophage groups and three groups of resident alveolar macrophages. T cells from normal adjacent tissues and TME were predominantly CD8+ T cells and highly expressed cytotoxic markers, suggesting that immunological assessment of SCLC is of great significance. Moreover, reclassification of T cells in SCLC patients can better mirror the expression pattern of T cells dysfunction and exhaustion markers (such as HAVCR2, CTLA4, LAYN, PDCD1, LAG3, TIGIT), which may be used as immunotherapeutic targets. scRNA-seq of T cell subsets of TME in SCLC revealed that HAVCR2 had the highest expression level in exhausted T cell subsets, while LAYN was sporadically expressed in exhausted CD8 + T cells, and CTLA4 was preferentially expressed on other T cells. T cell heterogeneity and co-inhibitory receptor expression preference in SCLC patients provide potential immunotherapeutic targets.

Tumor-infiltrating myeloid cells (TIMs), such as monocytes, macrophages, dendritic cells and neutrophils, have been identified as a key regulator of cancer growth (64, 65). Zilionis et al. (32) used scRNA-seq to locate TIMs in NSCLC patients and identified 25 TIMs states, most of which could be found repeatedly in patients. This study provides a new theoretical basis for future elucidation of the role of myeloid cells in cancer, and TIMs may serve as a new target for immunotherapy. Lavin et al. (33) conducted scRNA-seq to analyze the TME of 18 LUAD patients, and identified the characteristic genes of tumor infiltrating macrophages, such as TREM2, CD81, MARCO, APOE, etc. In addition, analysis with scRNA-seq found that vascular endothelial cells from NSCLC patients reduced their antigen presentation and the homing activity of immune cells through remodeling, thereby promoting tumor immune tolerance (34, 66). A subgroup of cancer-associated fibroblasts (CAFs) that highly expressed extracellular matrix protein genes were found to induce immunotherapy resistance by increasing the levels of PD-L1 and CTLA-4 proteins in Treg cells through cell cross talk (67).

Overall, a comprehensive understanding of the composition and state of immune cells is crucial to the elucidation of the responsiveness and resistance to current immunotherapies, and to the design of new immunomodulatory therapies. Use of the single-cell transcriptomic analysis in lung cancer patients receiving immunotherapy also revealed the heterogeneity and diversity of lung cancer immune responses. Clusters based on non-targeted transcriptional assessment of cell state often cannot be completely separated into traditional immune cell subpopulations in terms of cell surface protein expression. For immunotherapy, application of these methods to immune cells and stromal cells in TME can help elucidate the response to immunotherapy (such as ICIs) and the transcriptional state of drug-resistant cells. Importantly, the analysis aimed at mapping the immune picture of multiple lung cancer types has identified multiple new transcriptional states, which are related to the responsiveness of lymphocytes, DC cells, monocytes, macrophages and fibroblast compartment of TME to immunotherapy. Researchers can now look into how lung cancer and the immune system co-evolve during treatment and recurrence at the single-cell level.

Very complex interactions take place between tumor or immune cells and/or stromal cells, which together dictate the tumor progression and response to treatment. This cell-cell interaction can be studied by calculating the expression level of receptor and ligand from scRNA-seq data. At the same time, for the exploration of the complex interaction between tumor cells and TME, new technologies, including single-cell space transcriptome methods, are also developing rapidly, making it possible to look into the cell-cell physical interaction. At present, the extensively used cell interaction analysis algorithms include CellPhoneDB, CellChat, iTALK, NicheNet, among others (68–71), which is out of the scope of the review.

To look into the interaction between different cell types in TME, Wu F et al. (35) examined the cell-cell interaction by using scRNA-seq to get a full picture of TME in NSCLC patients, including angiogenesis, T cell activation, CAF activation, immunosuppressive cell recruitment, and activation of checkpoint routes. Obvious interactions were found between cancer cells and fibroblasts, endothelial cells and macrophages. An analysis of the cross cell interaction molecules revealed a complex network involving multiple carcinogenic and inflammatory signaling pathways. The researchers found that, in patients’ immune environment, macrophages played a key role in the inhibition of T cells via checkpoint pathway. Furthermore, predominant pathways vary with different subgroups of NSCLC. For instance, LUAD driving gene mutation had a high level of TIGIT pathway activation, but a low level of TIM3 (HAVCR2) pathway activation. Except in few LUSC patients, the authors did not detect any significant activation of PD1/PD-L1 axis, which might be ascribed to the low expression of PD1/PD-L1 at the transcriptomic level. Even in the same subtype of lung cancer, the interactions related to immunotherapy were different, highlighting the need for more precise biomarkers to improve the efficacy of immunotherapy. In addition, Tian Y, et al. (25) used CellPhoneDB to study the interaction between ligand-receptor pairs, and confirmed that the non-neuroendocrine small cell lung cancer subtypes (SCLC-non-NE) interacted more with other cells, including immune and stromal cells than other subtypes of SCLC, and may be related to the clinical outcomes of immunotherapy. By correlating the expression of ligand receptor pairs between different cell types, important information on cell-to-cell interactions related to lung cancer immunotherapy can be inferred from scRNA seq data, which may help us better understand the complicated arrangement and interaction between tumor cells and TME, as well as to find new indicators or signatures related to lung cancer immunotherapy.

ScRNA-seq is the most widely used single-cell sequencing technique. With its extensive application in scientific research, scRNA-seq gradually some methodologically inherent problems began to emerge. First, efficiency of mRNA capture in the single-cell transcriptome method is low, standing at somewhere between 5% and 15%, leading to data sparsity, sampling deviation, and loss of low-level transcript gene expressions (72). Second, it is only applicable to fresh tissue samples. For frozen samples, since cells have lost their activity, scRNA-seq cannot be performed. This substantially restricts its application, increases the difficulty of operation and reduces the sample flux. For example, in order to ensure the stability of RNA, many clinical samples need to be frozen. Such archived frozen samples are no longer amenable to scRNA-seq and their value could not be fully tapped by the latest technology (73). Third, the dissociation process induces the expression of stress genes, resulting in “artificial transcriptional stress responses” of cell transcription and eventually to transcription bias. The data so obtained cannot truly reflect the cell transcription status of the sample, and the reliability of the results is greatly impaired. This has been demonstrated by a great many experiments. For example, Brink et al. found that the process of protease dissociation at 37°C would induce the expression of stress genes, thereby introducing human errors, and leading to inaccurate results of cell type identification (74). The latest comparative experiment further confirmed this limitation: Dissociation at 37°C induced an increased expression of multiple stress genes, which yielded seriously distorted results, and the low-temperature dissociation could effectively avoid this phenomenon (75). Fourth, for many solid tissues, such as cerebral, cardiac and renal tissues, protease tends to dissociate the cell types that are subject to dissociation, thus losing the cells that are not easy to dissociate. At the same time, some sensitive cells may be damaged due to excessive dissociation. Therefore, the dissociation process cannot effectively obtain all cell types in the tissue, and the accuracy of the results is substantially affected (76–79). It is believed that, with the continuous improvement and breakthroughs of the technology, single-cell sequencing technology has a good prospect of being widely used in the research and treatment of lung cancer.

Single-cell proteomics allows for analysis of protein expression at the single-cell level, thus revealing fine differences between individual cells. It provides a powerful tool for the analysis of cell and tumor heterogeneity, specific cell types, circulating tumor cells (CTC), immunological research, the genetic study of cell cycle, and the examination of trace/rare samples. Working on different principles and depending on various scenarios, many methods for the quantitative detection of single-cell proteins have emerged in recent years, including microfluidic techniques, microporous methods, optical fiber nano biosensoring, fluorescent probing and mass spectrometry-based single-cell protein detection (80, 81). Especially in recent years, with the rapid development of mass spectrometry technology, the bottleneck of proteomic research based on mass spectrometry has been removed, and the scanning speed and detection sensitivity have been greatly improved, which makes it possible to detect extremely trace protein samples.

CyTOF is a flow cytometry technique based on mass spectrometry principles and is used for multi-parameter detection of individual cells. By integrating mass spectrometry and traditional flow cytometry, CyTOF not only keeps the high-speed of traditional flow cytometry, but also attains the high-resolution of mass spectrometry, overcomes the problem of overlapping light spectra of traditional flow cytometry fluorescence emission groups, and is able to simultaneously detect more than 30 protein markers in tumor cells. This high-dimensional single-cell technique is described as the “single-cell atlas” of the tumor ecosystem. As a single-cell high-dimensional immune analysis, it can better link the tumor immunity map with its clinicopathological characteristics (82, 83).

Lavin Y et al. (33) separated immune cells from tumor tissue, normal lung tissue and peripheral blood from lung cancer patients, detected specific transcripts of cells and more than 30 protein markers on the surface using CyTOF and other technologies, and drew a detailed map of immune cells in the TME of early-stage lung adenocarcinoma, to provide an experimental basis for the design of immunotherapeutic regimen for early lung cancer. The unsupervised cluster analysis of the three tissues divided T cells into 21 subgroups in terms of different surface marker expression patterns, including new subgroups (IX, XX, etc.) that had not been previously identified by traditional methods. In addition, the analysis of the proportion of subpopulations in different samples showed that the content of Treg cells in early-stage tumor patients was significantly increased, and it grew rapidly at the early stage of tumor, and PD-1 expression was significantly elevated in CD4+ and CD8+ T cells in tumor tissue.

Karacosta LG et al. (36) used CyTOF to identify and characterize the epithelial mesenchymal transformation (EMT) state of clinical lung cancer specimens according to the immune state map obtained in lung cancer cell lines. The researchers used HCC827 cell line for study of EMT, and examined 28 protein expression markers to characterize the EMT status and proliferation-, signal transduction- and apoptosis-related status by employing qualitative CyTOF. Their study observed an increased expression of PD-L1 during EMT, confirming that EMT was related to tumor immune escape. Notably, the researchers found that, among the EMT transcription factors, Oct3/4 and Nanog expression was significantly up-regulated during the entire EMT process, indicating that the co-expression of Oct3/4 and Nanog was crucial to the EMT of lung cancer cell lines. Shaul ME et al. (37) clinically assessed the level of circulating high-density neutrophils (HDN) and low-density neutrophils (LDN) in patients with advanced lung cancer by using CyToF, and found that the three main subtypes of LDN/HDN possessed immune characteristics and inherent plasticity. These findings laid foundation to the development of new tumor immunotherapies.

Imaging mass spectrometry (IMC) combines high-resolution imaging technology and CyTOF technology to generate tissue structure images involving multiple factors such as cell markers, transcripts, and transduction signals, so as to achieve single-cell proteome spatial analysis (84). In the field of immune oncology, IMC can classify infiltrating immune cells in a high parameter space while maintaining its spatial coordinates, which may provide useful information about host responses and inform the selection of appropriate immunotherapies (85).

Sorin M et al. (38) used IMC to describe histopathological patterns of pulmonary adenocarcinoma and the immune cell landscape in 416 patients, and analyzed more than 1.6 million cells, and conducted spatial analyses on immune cell lineages and activation status with different clinical relevance (including survival). Their analyses on the category of cellular neighborhood and survival time confirmed the association between specific cell interactions and survival rate, indicating that the tissue relationship of cells in the immune microenvironment is of uniquely prognostic value. They also studied the relationship between cell phenotype and survival in the TME. The results showed that the total number of neutrophils exerted no significant impact on survival, but the increased proportion of HIF1α+ subgroup was significantly correlated with poor overall survival.

Sorin M and their colleagues (39) also performed IMC on 114,524 single cells from 27 NSCLC patients receiving ICI, and achieved spatial resolution of immune spectrum and activation state with different clinical results. These studies proved that the expression of CXCL13 was related to the ICI efficacy, and the recombinant CXCL13 enhanced the response to anti-PD-1 in vivo, which could be ascribed to increased T-cell subpopulations subjected to antigen stimulation and decreased CCR2+ monocytes. These findings highlighted the importance of major immune cell lineages and their functional states in the response to ICIs and help us better understand the role of the tumor immune microenvironment in such response.

These observations highlighted the importance of evaluating immune cell phenotypes at the single-cell level. In fact, both CyTOF and IMC are particularly useful methods for characterizing the specific phenotypes of cells involved in responses to immunotherapy at the single-cell level. With immunotherapy, single-cell proteomic analysis can provide insights into the signaling pathways related to the effectiveness of immunotherapies and drug resistance. At the entrance into the era of single-cell proteomics, we are still faced with great challenges in proteome coverage depth and flux. We believe that these challenges can be addressed by integrating mass spectrometry flow cytometry, measurement strategies, and algorithms.