Article Text
PDF
Abstract
Background Lung squamous cell carcinoma (LUSC) remains a leading cause of cancer-related deaths with few therapeutic strategies. Immune checkpoint inhibitors (ICIs) have demonstrated promising efficacy in patients with LUSC. However, ICIs could also lead to a unique spectrum of immune-related adverse events (irAEs), which dampen the clinical outcome. In-depth characterization of the immune hallmarks of antitumor responses and irAEs remains an unmet need to maximize ICI-treatment benefits of patients.
Methods We performed single-cell RNA sequencing (scRNA-seq) on pre-ICI and on-ICI treatment tumor biopsies. We used bulk RNA-seq data of matched pretreatment/on-treatment tumors and irAE affected organs to validate observations from scRNA-seq analysis. Two independent patient cohorts were collected to determine circulating tumor necrosis factor (TNF) protein expression levels.
Results We found that increased proportions of a macrophage subcluster with highly expressed secreted phosphoprotein 1 (SPP1) and two tumor cell subclusters in irAE patients, whereas proportions of two cytotoxic CD8+ T cell subclusters were higher in patients with partial response (PR). TNF signaling pathway was conversely associated with treatment efficacy and irAE development in most macrophage and tumor cell subclusters. Cell–cell communications for TNF ligand-receptor pairs between macrophage/T cells and tumor cells were also bidirectionally remodeled in responders versus non-responders and irAE versus non-irAE patients. Bulk RNA-seq analysis on matched pretreatment/on-treatment tumors and irAE affected organs revealed remarkably enhanced macrophage abundance and TNF signaling pathway in on-treatment tumors and organs developed irAEs. Furthermore, we observed significantly increased circulating TNF protein in plasma or serum of irAE patients but not ICI responders, based on analysis of two independent LUSC patient cohorts and one published ICI patient cohort.
Conclusions Our data depicts specific reprogramming of macrophage, T cells and tumor cells associated with ICI response and irAEs, elucidates divergent roles of TNF signaling in antitumor immunity and irAEs, and highlights the significance of TNF expression in irAE development in the LUSC setting.
- Immune Checkpoint Inhibitors
- Tumor Microenvironment
- Lung Neoplasms
Data availability statement
Data are available in a public, open access repository. All the raw sequencing reads were deposited in Sequence Read Archive (http://www.ncbi.nlm.nih.gov/sra/) under the accession number SRP439674.
This is an open access article distributed in accordance with the Creative Commons Attribution Non Commercial (CC BY-NC 4.0) license, which permits others to distribute, remix, adapt, build upon this work non-commercially, and license their derivative works on different terms, provided the original work is properly cited, appropriate credit is given, any changes made indicated, and the use is non-commercial. See http://creativecommons.org/licenses/by-nc/4.0/.
Statistics from Altmetric.com
WHAT IS ALREADY KNOWN ON THIS TOPIC
Previously studies have identified T cell clones recognizing shared antigens between tumors and immune-related adverse event (irAE)-affected tissues, thereby indicating that immune responses in tumor microenvironment (TME) may involve in irAE development and antitumor immunity. But most investigations on irAEs are based on blood or stool samples. Hence, the TME-related hallmarks of irAEs and antitumor immune responses, which could be used to maximize immune checkpoint inhibitor (ICI)-treatment benefits of patients with lung squamous cell carcinoma (LUSC), remains poorly studied.
WHAT THIS STUDY ADDS
Our study used multi-source data from different centers, including single-cell RNA sequencing, bulk RNA sequencing and proteomics data, to identify and validate the significant association of tumor necrosis factor (TNF) in irAE development, but not antitumor responses.
HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY
This unique role of TNF played in irAE development suggests that anti-TNF drugs may be employed as potential irAE therapies, without reducing ICI-treatment efficacies, in patients with LUSC.
Background
Lung squamous cell carcinoma (LUSC) is the second most common type of lung cancer, and represents a leading cause of cancer-associated mortality.1 2 In contrast to lung adenocarcinoma (LUAD), very limited clinical benefits are derived from targeted therapies in patients with LUSC, due to the lack of established actionable targets.3 4 Recently, immune checkpoint inhibitors (ICIs) have been proved to be a successful therapeutic strategy for patients with LUSC.5 6 However, blocking immune checkpoint pathways may cause excessively activated immune responses across various normal organs or tissues and these adverse events are collectively referred to as immune-related adverse events (irAEs),7 8 which considerably limit the clinical benefit obtained from ICIs. For example, in CheckMate-063, nivolumab showed a comparable rate of high-grade treatment-related adverse events (14.5%) and objective response (17%) in patients with advanced LUSC.9 In KEYNOTE-407, while combination of pembrolizumab with chemotherapy achieved greater response rate (57.9%), grade 3 or higher treatment-related adverse event rate was also increased (69.8%) in tandem.5 Therefore, an in-depth understanding of the molecular and immune hallmarks of antitumor responses and irAEs is fundamental for maximizing treatment benefits of patients with LUSC.
Several studies have revealed identical T-cell clones recognizing shared antigens from tumor lesions and irAE-affected tissues,10–12 thereby indicating that immune activation in tumor microenvironment (TME) may be a mechanistic basis for the association between irAEs and antitumor immune response. At present, most investigations on irAE mechanisms or biomarkers are based on blood or stool samples,8 and as a result, direct findings from TME are limited. Single-cell RNA sequencing (scRNA-seq) has emerged as a powerful tool for defining cellular compositions and functions across a spectrum of cancer types.13 Previous reports have leveraged scRNA-seq technology to identify ICI response-related T-cell signatures in lung cancer, but most of the studied patients are diagnosed with LUAD.14 15 In addition, scRNA-seq analysis of ICI-treated tumors is scarce, primarily due to the challenge of obtaining high-quality specimens after ICI treatment. In this study, we conducted scRNA-seq assessment of patients with LUSC treated with anti-programmed death 1 (anti-PD-1) to characterize the TME features and dynamics related to response and irAEs, and to discover potential irAE therapeutic targets without compromising ICI efficacy.
Methods
Human specimens
Written informed consent was acquired from all patients in this study. All patients were diagnosed with pathologically and clinically confirmed advanced lung squamous cell carcinoma. We obtained tumor biopsies from nine patients pretreatment or post-treatment with ICIs. Details of patient information were in online supplemental table 1. Clinical responses were defined by CT images following the guidelines of response evaluation criteria in solid tumors (RECIST) V.1.1. Multiple rounds of clinical responses were estimated during ICI treatment, and the best response for each patient was used in our study.
Supplemental material
Single-cell dissociation from fresh tumors
Fresh tumors were cut into small pieces of approximately 1 mm3 in the RPMI-1640 medium (Gibco) with 10% fetal bovine serum (FBS; Gibco), and enzymatically digested with Liberase TL (0.1 mg/mL, Roche) and DNase I (0.02 mg/mL, Roche) for 30 mins on a rotor at 37°C. After filtered by 70 µm Cell-Strainer (BD) in the RPMI-1640 medium (Gibco), the suspended cells were centrifuged at 400 g for 5 min at 4°C. After removing the supernatant, the pelleted cells were suspended in red blood cell lysis buffer (Invitrogen) and incubated on ice for 2 min to lyse the red blood cells. The cell pellets were re-suspended in phosphate-buffered saline (PBS) supplemented with 0.02% FBS after washing twice with PBS.
Single-cell capture, cDNA library preparation and sequencing
The concentration of single-cell suspensions was adjusted to 300 cells/µl. Cells were loaded between 7000 and 10 000 cells/chip position using the Chromium Single cell 3′ Library, Gel Bead & Multiplex Kit and Chip Kit (10× Genomics, V2 barcoding chemistry) according to the manufacturer’s instructions. All the subsequent steps were performed following the standard manufacturer’s protocols. Purified libraries were analyzed by an Illumina NovaSeq 6000 sequencer with 150 bp paired-end reads.
Raw scRNA-seq data processing, filtering and batch effect correction
Cell Ranger (2.1.0, 10× Genomics) analysis pipeline with default parameters was used to process raw scRNA-seq reads, including aligning reads against GRCh38 reference genome, barcode processing, unique molecular identifier (UMI) counting, cell identification and generation of a gene-barcode matrix for each sample. Cells with fewer than 200 genes detected or more than 20% of transcripts attributed to the mitochondrial genome were removed. Genes detected in more than three cells across all samples were retained. Potential doublets were detected and removed by using DoubletFinder16 V.2.0.3 with expected doublet rate set to 0.075. Then we applied Seurat (V.4.0.5) R toolkit in the following analysis. Count data were normalized and log-transformed. The effects of cell cycle, counts of genes, percentage of mitochondrial and percentage of ribosome genes were regressed out. The CCA (canonical correlation analysis) method implemented in Seurat was chosen for individual data set integration. Integrated expression matrix was used for dimension reduction and clustering. The log-transformed expression matrix was used for differential expression and signature expression analyses.
Clustering, visualization and identification of cell types
The first 50 principal components generated from principal components analysis were used for clustering of cells, and the resolution was set to 1.0. Uniform Manifold Approximation and Projection was performed for visualization in two dimensions. To determine the broad cell types of each cluster, differentially expressed genes (DEGs) were calculated using the FindAllMarkers function in Seurat and the top 50 most significant DEGs were manually reviewed based on published canonical cell markers (eg, COL1A1/COL1A2 for fibroblasts; EPCAM for epithelial cells; PECAM1 for endothelial; PTPRC for immune cells; KIT for Mast cell; LYZ/C1QC for macrophage; NCAM1 for NK cell, CD3E/G for T cells; FOXP3 for Tregs; XBP1/CD38 for plasma cells; CD19/MS4A1/CD79A for B cells). We determined epithelial cells as tumor cells based on the copy-number alteration generated by the R inferCNV software package (https://github.com/broadinstitute/inferCNV). GSVA R package17 was applied to the signature and pathway score calculation for each cell using ssGSVA function. Gene Set Enrichment Analysis (GSEA) was completed by fgsea R package.18
Cell–cell interaction analysis
We used the iTALK R software package19 20 to analyze and visualize differential cell interaction signals between tumor cells, T cells and macrophages based on built-in receptor-ligand pair databases. The ligand-receptor pairs with p values of <0.05 were considered as significant differential interactions.
Estimation of cell type abundance of bulk RNA-seq
We used the CIBERSORTx software,21 a digital cytometry tool, to detect the relative cell type abundance in the bulk RNA-seq data. Gene lists for the pathways used in this study were obtained from MSigDB.22 The GSVA package was used to estimate pathway scores in the bulk RNA-seq data.
Determination of protein expression level of cytokines
We collected cohort 1 of 15 patients with LUSC from the Department of Medical Oncology Hunan Cancer Hospital during 2016–2019. Details of patient information were in online supplemental table 2. Plasma samples of these patients were collected 1 day prior ICI therapy and after the second cycle of ICI therapy. BD Cytometric Bead Array was performed to detect the protein level of tumor necrosis factor (TNF), interleukin (IL)-2, IL-6 and interferon (IFN). We collected cohort 2 of 115 patients with LUSC from the Peking Union Medical College Hospital during 2019–2023. Details of patient information were in online supplemental table 3. Plasma samples collected 1 day prior to ICI therapy and after the second cycle of ICI therapy were used to detect the protein level of TNF via IMMULITE/IMMULITE 1000 TNFα kit (Siemens Healthineers, China). Written informed consents were obtained from all the patients. Clinical information, including age at diagnosis, sex, cancer types, response, clinical interventions, and ICI drug regimens (dosage and agents) was obtained from the medical records. Information regarding irAEs was collected, including types, symptoms and grade of irAEs. Clinical responses of patients were defined by CT images following the guidelines of RECIST V.1.1. Multiple rounds of clinical responses were estimated during ICI treatment, and the best response for each patient was used in our study.
Results
Single-cell characterization of LUSC before and after PD-1 blockade
To chart a comprehensive single-cell atlas of LUSC TME during ICI treatment, we performed scRNA-seq analyses on fresh tumor biopsies from nine patients with LUSC (online supplemental table 1). Among these patients, eight were men and one was woman. The median age at biopsy collection for all patients was 71 years (IQR 57–75 years). At the time of biopsy, five patients had been exposed to ICI and four patients were treatment-naïve (figure 1A). We closely monitored patients for irAE development, which occurred in two of five treated patients (figure 1A). After stringent quality control and filtering steps, a total of 66,847 cells were analyzed with respect to their transcriptomes. To diminish possible batch effects and allow for combined analysis of cancer cells with patient-specific expression features, we employed Seurat’s CCA-based algorithm to integrate cells across samples and jointly clustered these cells in an unsupervised manner. Thirteen major cell types were identified comprizing various immune and non-immune cell types (figure 1B), including T cells, B cells, myeloid cells, epithelial cells, endothelial cells and fibroblasts based on their marker genes (figure 1D,E). Each cell cluster consisted of cells from all patients, indicating the absence of sample-specific batch effects (figure 1C and E). In line with previous studies, the abundance of CD4+ T cells, CD8+ T cells and natural killer (NK) cells was significantly increased in tumor from patients with partial response (PR) compared with those with stable disease (SD, figure 1F–H & online supplemental figure S1A). In contrast, we did not observe significant alterations of these major cell clusters between irAE and non-irAE patients’ tumors (online supplemental figure S1B), suggesting that different molecular underpinnings might be responsible for ICI response versus irAEs.
Single-cell characterization of the LUSC tumors pre or post ICI therapy. (A) Summary of treatment histories and clinical features across profiled LUSC tumor specimens. (B) t-distributed stochastic neighbor embedding (t-SNE) projection of all captured cells across all tumor lesions, colored by broad cell type. (C) t-SNE projection of all captured cells colored by patients. (D) Heatmap of scaled normalized expression for representative major cell type marker genes as determined by a two-sided Wilcoxon rank-sum test with p.adjust < 0.05. (E) Stacked bar plots show cell proportions by patient. (F) Proportion of each broad cell types in tumor lesions obtained from ICI-naïve (no ICI), PR and SD patients. The boxes indicate the median±1 quartile, with whiskers extending from the hinge to the smallest or largest value within 1.5 IQR from the box boundaries. Each dot represents individual patients. P values were determined by Kruskal-Wallis test. DC, dendritic cell; F, female; ICB, immune checkpoint blockade; ICI, immune checkpoint inhibitor; irAE, immune-related adverse event; LUSC, lung squamous cell carcinoma; M, male; NK, natural killer; PR, partial response; Pt, patient; SD, stable disease; Treg, regulatory T cell.
Response and irAE-related changes in T-cell subcluster abundance and functionality
To further test the above hypothesis, we next investigated the distribution and function of the T-cell subclusters among different response and irAE subgroups. Seven CD8+ T cell and four CD4+ T-cell subclusters were well-defined and annotated based on published signatures and cell markers23 24 (figure 2A,B). To determine whether the characterized T-cell subclusters in LUSC resembled the functional phenotypes of T cells, we scored individual cells for cytotoxicity and exhaustion based on established gene signatures.23 24 Consistent with previous studies, CD8+ killer cell immunoglobin-like receptor-positive tyrosine kinase-positive natural killer-like (KIR+TXK+ NK-like) cells and CD8+ terminal exhausted T (terminal Tex) cells exhibited the highest cytotoxicity and exhaustion scores, respectively (figure 2C). CD4+ interferon gamma-positive T follicular helper/T helper 1 (IFNG+Tfh/Th1) cells had both higher cytotoxicity and exhaustion scores, and CD4+ tumor necrosis factor receptor superfamily member 9-positive T regulatory (TNFRSF9+Treg) cells had highest exhaustion scores (online supplemental figure S2A). Among CD8+ T cells, we found that PR tumors showed significantly increased proportion of granzyme K-positive early effector memory (GZMK+Tem) cells and KIR+TXK+NK-like cells, whereas these cells were less prevalent in post-treatment SD tumors (figure 2D,online supplemental figure S2B). CD8+ terminal Tex cells and CD8+ interleukin 7 receptor-positive memory T (IL7R+Tm) cells were enriched in pre-ICI tumors (figure 2D). When comparing CD4+ T-cell subpopulations, we found that PR tumors exhibited a higher abundance of amphiregulin-positive tissue inhibitor of metalloproteinase 1-positive memory T (AREG+TIMP1+Tm) cells and nucleoside diphosphate kinase 1-positive and C-C motif chemokine receptor 4-positive T (NME1+CCR4+T) cells, and lower abundance of CD4+ IFNG+ Tfh/Th1 cells (online supplemental figure S2C). But we did not observe a significant difference in the abundance of T-cell subclusters between irAE and non-irAE patients (online supplemental figure S2D, E). Considering relative lower abundance of CD4+ T cells, we next examined response and irAE-related functional phenotype shifts in CD8+ T-cell subclusters. Cytotoxicity scores were upregulated and exhaustion scores were downregulated in most of CD8+ T-cell subclusters from ICI-treated tumors (figure 2E). In melanoma and renal cell carcinoma models, progenitor exhausted CD8+ T cells, which persist longer term, differentiate to terminally exhausted phenotype, and CD8+ T cells displaying the latter phenotype kill target tumor cells more efficiently during response to anti-PD-1.25 26 Within CD8+ GZMK+ early Tem cells from ICI-treated samples, we observed significant enrichment of the terminally exhausted gene signature in cells from PR tumors and enrichment of progenitor exhausted gene signature in cells from SD tumors by GSEA analysis (figure 2F). Interestingly, for comparison in the irAE group, CD8+ GZMK+ early Tem cells from irAE patients’ tumor showed significant enrichment of progenitor exhausted gene signature and non-irAE patients’ tumor showed significant enrichment of terminally-exhausted gene signature. (figure 2G). Collectively, these findings suggested that several subclusters of CD8+ T cells and CD4+ T cells were more prevalent in PR tumors and divergent T-cell exhaustion states of CD8+ GZMK+ Tem were observed in response and irAE subgroups.
CD8+ T-cell exhaustion states were differentially modulated in irAE groups and response groups. (A) t-SNE projection of T cells captured across all tumor lesions, colored and labeled by T-cell subtypes. (B) Heatmap of cell-type-defining marker genes for T-cell clusters. (C) Violin plot of cytotoxicity and exhaustion signature score of each CD8+ T-cell subcluster. P values were determined by Kruskal-Wallis test. (D) Proportion of CD8+T-cell subclusters in no ICI, PR and SD patients. The boxes indicate the median±1 quartile, with whiskers extending from the hinge to the smallest or largest value within 1.5 IQR from the box boundaries. Each dot represents individual patients. P values were determined by Kruskal-Wallis test. (E) Comparison of cytotoxicity and exhaustion signature score in ICI and no ICI groups across multiple patients. P values were determined by a two-sided Wilcoxon rank-sum test. Red indicates upregulation in ICI group. Blue indicates upregulation in no ICI group. The shade of the square indicates the P value, and the bold outlier indicates statistically significant results. (F) Gene Set Enrichment Analysis (GSEA) of terminally exhausted (left panel) and progenitor exhausted signatures (right panel) in CD8.GZMK+ early Tem from PR patients compared with SD patients. (G) GSEA of terminally exhausted (left panel) and progenitor exhausted signatures (right panel) in CD8.GZMK+ early Tem from irAE present patients compared with irAE absent patients. AREG+TIMP1+Tm, amphiregulin-positive tissue inhibitor of metalloproteinase 1-positive memory T; DEG, differentially expressed gene; FDR: false discovery rate; GZMK+Tem, granzyme K-positive early effector memory; ICI, immune checkpoint inhibitor; IFNG+Tfh/Th1, interferon gamma-positive T follicular helper/T helper 1; IL7R+Tm, interleukin 7 receptor-positive memory T; irAE, immune-related adverse event; KIR+TXK+ NK-like, killer cell immunoglobin-like receptor-positive tyrosine kinase-positive natural kill; NME1+, nucleoside diphosphate kinase 1-positive; NME1+CCR4+ T, nucleoside diphosphate kinase 1-positive and C-C motif chemokine receptor 4-positive T; PR, partial response; Pt, patient; SD, stable disease; Tc17, type 17 T cells; terminal Tex, terminal exhausted T; Tn, naïve T cells; TNFRSF9+Treg, tumor necrosis factor receptor superfamily member 9-positive T regulatory; tSNE, t-distributed stochastic neighbor embedding.
Distinct reprogramming of macrophages in anti-tumor response versus irAEs
Tumor-associated macrophages (TAMs) are reported to play dual roles in the TME, including inflammation, angiogenesis and tumor cell killing,27 28 but the association between TAMs and irAEs remain largely unknown. Therefore, we recovered six macrophage subclusters from the full data set and characterized their associations with response and irAEs (figure 3A). We annotated macrophage subclusters based on highly expressed genes and previous published literature29 30 (figure 3B). To better understand the functional phenotypes of the subclusters, we analyzed gene expression of M1, M2, angiogenesis signatures29 and inflammatory response to antigenic stimulus. We did not observe a clean binarization of expression of M1 and M2 signature genes among these subclusters (online supplemental figure S3A). Secreted phosphoprotein 1-positive macrophage (macro SPP1+) showed preferential expression of angiogenesis gene signature and inflammatory response to antigenic stimulus (figure 3C). Two signatures were also highly expressed in SPP1 and Complement C1q C Chain (C1QC) co-expressed macrophage subcluster, macro C1QC+SPP1+ (figure 3C). Importantly, the frequency of macro SPP1+ was higher in tumors from irAE patients (figure 3D, online supplemental figure S3C), which suggested a potential linkage between macro SPP1+ mediated inflammation in TME and irAEs. In PR tumors, significantly decreased frequency of macro C1QC+, infiltration of which in tumors indicates clear cell renal cell carcinoma recurrence,31 was observed (figure 3D, online supplemental figure S3B). In parallel, GSEA analysis showed diverse enrichment of pathways associated with aspects of macrophage activity, including angiogenesis, inflammatory response and antigen processing and presentation, in response and irAE subgroups (figure 3E,F). In addition, TNF-α signaling and tumor growth factor (TGF)-β pathway was enriched in almost all macrophage subclusters from SD and irAE patients (figure 3E,F). Given that TNF-α is secreted by macrophages and a master regulator of inflammation in cancer,32 33 our data suggested that TNF might be a key regulator in TME that promoted irAEs and decreased ICI treatment efficacy.
Opposite regulation of macrophage-related inflammatory responses for irAE and antitumor response. (A) t-SNE projection of macrophages captured across all tumor lesions, colored and labeled by macrophage cell subtypes. (B) Heatmap of cell-type-defining marker genes for macrophage subclusters. (C) Violin plot of angiogenesis and inflammatory response to antigenic stimulus signature score of each macrophage subcluster. (D) Proportion of macrophage subclusters in no ICI, PR and SD patients (left panel) and in irAE present and absent patients (right panel). The boxes indicate the median±1 quartile, with whiskers extending from the hinge to the smallest or largest value within 1.5 IQR from the box boundaries. Each dot represents individual patients. P values were determined by Kruskal-Wallis test for test in the response group. P values were determined by a one-sided Wilcoxon rank-sum test in irAE group. (E) GSEA of macrophage related and inflammatory pathways in macrophage subclusters from PR patients compared with SD patients. (F) GSEA analysis of macrophage related and inflammatory pathways in macrophage subclusters from irAE present patients compared with irAE absent patients. The shade of the square indicates normalized enrichment score. Red indicates positive enrichment and blue indicates negative enrichment. C1QC, complement C1q C chain; CX3CR1, C-X3-C Motif Chemokine Receptor 1; GSEA, Gene Set Enrichment Analysis; ICI, immune checkpoint inhibitor; irAE, immune-related adverse event; ISG15, interferon-stimulated protein 15 KD; NES: normalized enrichment score; PR, partial response; Pt, patient; SD, stable disease; SPP1, secreted phosphoprotein 1; tSNE, t-distributed stochastic neighbor embedding; VSIR, V-Set Immunoregulatory Receptor.
Opposite alteration of inflammation pathways for tumor cells associated with response and irAEs
To better characterize active cellular programs in cancer cells that may drive interactions with the immune cell in TME, we next sought to identify features of tumor cells across patients. We identified epithelial cells as malignant via gene expression and inferred copy number alterations (online supplemental figure S4). Consistent with previous studies,2 4 34 LUSC tumor cells mostly harbored 3q insertions and 5q deletions (online supplemental figure S4), which demonstrated the robustness of our analysis. Unsupervised clustering uncovered eight unique subclusters, denoted T1 to T8 (figure 4A), with DEGs specifically distinguishing each cell clusters (figure 4B). None of these subclusters were significantly changed in different response groups (online supplemental figure S5A). In tumor from irAE patients, we observed significantly increased abundance of T2 and T3 (figure 4C, online supplemental figure S5B), which highly expressed tubulin related genes and interferon-stimulated protein 15 KD (ISG15), respectively. Next, we scored tumor cells for activity of various pathways to identify the cellular programs associated with response and irAEs. Almost all tumor cells from irAE patients in different subclusters scored higher for inflammatory programs, with increased in TGF-β, TNF-α, IL-6, IL-2 signaling, whereas PR tumor cells scored lower for these programs (figure 4D) and IFN responses were higher in PR tumors and lower in tumors from irAE patients (figure 4D). The bidirectional alteration of inflammatory pathways and specific factors/cytokines among irAE and response groups indicated potential opportunities to discover specific irAE therapies without decreasing ICI efficacy.
Divergent tumor cell inflammatory characteristics and cell–cell communications in irAEs and response. (A) t-SNE projection of tumor cells captured across all tumor lesions, colored and labeled by tumor cell subtypes. (B) Heatmap of cell-type-defining marker genes for tumor cell subclusters. (C) Proportion of tumor cell subclusters in irAE present and absent patients. The boxes indicate the median±1 quartile, with whiskers extending from the hinge to the smallest or largest value within 1.5 IQR from the box boundaries. Each dot represents individual patients. P values were determined by a one-sided Wilcoxon rank-sum test in irAE group. (D) Comparison of inflammatory and cytokine signaling pathway scores in PR versus SD patients (left panel) and irAE present versus irAE absent patients (right panel). P values were determined by a two-sided Wilcoxon rank-sum test. Red indicates upregulation in PR and irAE present patients. Blue indicates upregulation in SD and irAE absent patients. The shade of the square indicates the P value, and the bold outlier indicates statistically significant results. (E) Circos plots showing details of top 20 TNF ligand-receptor pairs compared between irAE present and absent (left panel) or PR and SD (right panel). C1QC, complement C1q C chain; CX3CR1, C-X3-C Motif Chemokine Receptor 1; ICI, immune checkpoint inhibitor; irAE, immune-related adverse event; PR, partial response; SD, stable disease; TNF, tumor necrosis factor; tSNE, t-distributed stochastic neighbor embedding; VSIR, V-Set Immunoregulatory Receptor.
Given these observations of different T-cell exhaustion status in irAE versus response, upregulation of TNF-α, TGF-β, IFN and ILs in macrophage and tumor cells of irAE patients and downregulation in tumor cells from PR patients, we hypothesized that the crosstalk between cell populations might participate in ICI induced antitumor response and irAEs. Therefore, we used iTALK to identify putative signals between different cell populations based on iTALK’s built-in receptor-ligand pairs database.19 20 We identified significantly increased TNF-related cellular interactions in irAEs versus no irAEs, which were decreased when comparing PR with SD (figure 4E). Among these top differentially modulated ligand-receptor pairs, most of the TNF-related communications were between macrophage and tumor cells in irAE group (figure 4E). And those interactions in the response group were between T cells and tumor cells (figure 4E). However, inversely differentially altered cell–cell communications for TGF, IFN and ILs related ligand-receptor pairs among response and irAE subgroups were not observed (online supplemental figure S6). Collectively, our observations of distinct TNF pathway alterations and tumor-immune cell TNF-related communications indicated the pivotal role of TNF in promoting irAE and decreasing antitumor response.
Increased TNF expression in irAE patients in independent cohorts
To link our observations in the tumor to the immunologic events in irAE affected organs/tissues, we conducted analysis of bulk RNA-seq on pretreatment/on-treatment tumor samples, irAE affected and unaffected tissue biopsies from ICI-myocarditis patient11 and ICI-encephalitis affected and unaffected brain tissue from brain irAE developed patient.35 We deconvoluted these bulk transcriptome data into fractions of immune cells types via CIBERSORTx21 and found that irAE affected samples and on-treatment tumor contained high fractions of macrophages, with more striking increment than CD8+ T cells (figure 5A). Furthermore, we observed increased score of inflammatory response and TNFA signaling via NFKB in both on-treatment tumor and irAE affected samples (figure 5B), however, TGF, IL-6, IFN and IL-2 pathways were not elevated as much as these two pathways (online supplemental figure S7A, B).
Identification of immune cell fractions and TNF pathway upregulation in irAE affected organs. (A) Heatmap of immune cell fractions across multiple tissues from patients with ICI-induce myocarditis (left panel) and encephalitis (right panel). Immune cell fractions were inferred by using CIBERSORTx. (B) Barplots of TNFA signaling via NFKB and inflammatory response signature scores in multiple tissues from patients with ICI-induce myocarditis (left panel) and encephalitis (right panel). CAF, cancer associated fibroblast; ICI, immune checkpoint inhibitor; irAE, immune-related adverse event; NFKB, NF-κB; NK cell, natural killer cell; TNF, tumor necrosis factor.
Next, we proposed that the trafficking of immune cells into irAE affected tissues or tumor tissue induced by ICI might be reflected by changes of cytokines in the blood. We collected an independent LUSC cohort 1 at Hunan Cancer Hospital. The median (IQR) age of patients was 62 (58.5–65) years, and 14 (93.3%) were men. We compared the alteration of TNF, IL-2, IL-6 and TNF in the plasma of LUSC (total=15; 8 with irAEs and 7 without irAEs; 12 responders and 3 non-responders) patients after second cycle of ICI therapy (online supplemental table 2). Consistently, the alteration of TNF in irAE patients after ICI therapy was significantly higher than non-irAE patients, and the comparison between responder and non-responder was not significant (figure 6A, B,). Considering that the sample size of cohort 1 was small, we collected published data from a prospective multicenter study and another LUSC patient cohort,36 cohort 2, with larger sample size (n=115) to validate the robustness of these observations. We analyzed proteomics data, generated by Olink platform, of serum samples from 40 patients before and during treatment with ICIs. Similarly, alterations of TNF expression were significantly elevated in irAE developed patients, not in responders (online supplemental figure S8). Cohort 2 were collected at Peking Union Medical College Hospital (online supplemental table 3) and TNF expression levels in plasma of patients were determined by chemiluminescent technology (Siemens Healthineers). By comparing differences between TNF expression before and during ICI treatment, we observed that TNF differences increased in patients with irAEs (figure 6C). We did not observe significant changes when comparing responders and non-responders (figure 6C).
Comparison of TNF expression in plasma samples from patients with LUSC. (A) Comparison of alterations of TNF, IL-2, IL-6 and IFN protein levels after ICI treatment between patients with and without irAEs in cohort 1. Each dot represents one patient. (B) Comparison of alterations of TNF, IL-2, IL-6 and IFN protein levels after ICI therapy between responder and non-responders in cohort 1. Each dot represents one patient. (C) Comparison of alterations of TNF protein levels after ICI treatment between patients with and without irAEs in cohort 2. (D) Comparison of alterations of TNF protein levels after ICI therapy between responders and non-responders in cohort 2. P values were determined by a two-sided Wilcoxon rank-sum test. Boxplots indicate the median±1 quartile, with whiskers extending from the hinge to the smallest and largest values within 1.5 IQR from the box boundaries. CR, complete response; ICI, immune checkpoint inhibitor; IFN, interferon; IL, interleukin; irAE, immune-related adverse event; LUSC, lung squamous cell carcinoma; PD, progression disease; PR, partial response; SD, stable disease; TNF, tumor necrosis factor. PR and CR patients were classified as responders, PD and SD patients were classified as non-responders.
Taken together, these data obtained from different centers and platforms confirmed the importance of TNF in development of irAEs and indicated that TNF could be harnessed to uncouple irAEs from antitumor response in patients with LUSC.
Discussion
To our knowledge, the present study is the first to show potential discriminative mechanisms of antitumor response and irAEs in patients with LUSC treated by ICIs. Our comprehensive scRNA-seq of LUSC tumor samples before or on ICI treatment identified dynamics in immune and tumor cell compositions and functional programs underlying antitumor response and irAEs induced by ICI, which pinpointed TNF as a pivotal regulator. Cytokine examination in patients with LUSC’s plasma demonstrated a connection between TNF increment and development of irAEs, but not antitumor response. Analysis of bulk RNA-seq data of matched pretreatment/on-treatment tumor and irAE affected tissues identified identical increased macrophage fraction and expression of TNF pathways in on-tumor and irAE affected tissue. Therefore, we hypothesize that elevation of macro SPP1+ proportion and macrophage-tumor cell communication activate inflammatory responses in tumors and stimulate macrophages to migrate to the blood, and subsequently localize in irAE affected organs. Future studies are needed to demonstrate this hypothesis and elucidate the systemic landscape of dynamic trajectories for immune cells and tumor cells that could affect antitumor response and irAEs.
Although previous studies revealed that TNF blockade improved colitis and hepatitis without affecting the antitumor activity in melanoma and colon cancer bearing mice treated by combined anti-PD-1 and anti-cytotoxic T lymphocyte antigen 4 antibodies37 and specific TNF-related cell interactions between CXCR3+ CD8+ T cell and myeloid cell populations from blood were enriched in responsive liver tumor and non-irAE,38 none have explored the links between tumor/immune cell characteristics in tumors for response versus irAEs. Our study extended the scope of previous TNF findings and involvement of tumor cells via interaction with macrophage for irAE development. In particular, we focused on patients with LUSC which were not included in previous studies.
Th17, the IL17A expressing compartment of CD8+ Tc17 in CD4+ T cells, is known to play a major role in several autoimmune diseases. It was reported to be significantly increased in immune-related enterocolitis and arthritis.27 39 However, we did not identify Th17 in CD4+T cells in LUSC tumors and abundance of Tc17 did not significantly change in the irAE group. Further investigations on fluorescence-activated cell sorting sorted CD3+ T cells will help to determine Th17 or Tc17 function in irAE development in the LUSC setting.
This study has several limitations. First, due to difficulties in accurately determining irAE occurrence and obtaining high-quality single-cell suspensions from on-treatment tumors,40 we only collected a small cohort in scRNA-seq analysis. Second, we were not able to perform subgroup analysis based on anatomic sites of irAEs. Therefore, whether our findings are generalizable to other irAE types will need to be investigated since tissue specific factors may impact irAEs.8 Future studies with a larger sample cohort that includes paired tumor, blood and irAE affected tissues could address these limitations.
In conclusion, we identified the distinct role of TNF in antitumor response and irAE and provided valuable insights into the possibility to dissociate irAEs from antitumor responses, which could be harnessed to enhance the benefit of patients with LUSC treated by ICI.
Data availability statement
Data are available in a public, open access repository. All the raw sequencing reads were deposited in Sequence Read Archive (http://www.ncbi.nlm.nih.gov/sra/) under the accession number SRP439674.
Ethics statements
Patient consent for publication
Ethics approval
The study was conducted in accordance with ethical guidelines of US Common Rule, and was approved by the Ethics Committee of Peking Union Medical College Hospital, Peking Union Medical College, Chinese Academy of Medical Sciences (Beijing, China; approval number: I-22PJ1040) and approved by the Ethics Committee of Hunan Cancer Hospital (Approval number: SBQLL-2019-065). Participants gave informed consent to participate in the study before taking part.
Acknowledgments
We thank all laboratory members in Zhuang laboratory and Jing laboratory.
References
Supplementary materials
Supplementary Data
This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.
Footnotes
Twitter @yingjing06
MC, PM, YZ and DW contributed equally.
Contributors Conception and design: YJ, GZ, and MW. Development of methodology: PM, DW, and ZY. Acquisition of data: MC, PM, YZ, and GZ. Analysis and interpretation of data: YJ, GZ, PM, YF, and XZ. Writing, review and/or revision of the manuscript: YJ and GZ. Technical or material support: MC, PM, YZ, DW, ZY, YF, and XZ. Study supervision: GZ, MW, and YJ. Guarantor: YJ.
Funding This work was supported by the National Natural Science Foundation of China (82373351 to GZ), Shanghai Pujiang Program (22PJ1409000 to YJ), Shanghai Municipal Education Commission-Gaofeng Clinical Medicine Grant Support (20161313 to GZ), Collaborative Innovation Center for Clinical and Translational Science by Ministry of Education & Shanghai (CCTS-2022203 to GZ), innovative research team of high-level local universities in Shanghai (SHSMU-ZLCX20210200 to GZ), and 111project (no. B21024 to GZ), the National High Level Hospital Clinical Research Funding (2022-PUMCH-B-106 to MW).
Competing interests None declared.
Provenance and peer review Not commissioned; externally peer reviewed.
Supplemental material This content has been supplied by the author(s). It has not been vetted by BMJ Publishing Group Limited (BMJ) and may not have been peer-reviewed. Any opinions or recommendations discussed are solely those of the author(s) and are not endorsed by BMJ. BMJ disclaims all liability and responsibility arising from any reliance placed on the content. Where the content includes any translated material, BMJ does not warrant the accuracy and reliability of the translations (including but not limited to local regulations, clinical guidelines, terminology, drug names and drug dosages), and is not responsible for any error and/or omissions arising from translation and adaptation or otherwise.