Dissecting the Immunological Profiles in NSD3-Amplified LUSC through Integrative Multi-Scale Analyses

Abstract

The histone H3 lysine 36 (H3K36) methyltransferase NSD3, a neighboring gene of FGFR1, has been identified as a critical genetic driver of lung squamous cell carcinoma (LUSC). However, the molecular characteristics, especially the immunological roles of NSD3 in driving carcinogenesis, are poorly understood. In this study, we systematically integrated multi-omics data (e.g., genome, transcriptome, proteome, and TMA array) to dissect the immunological profiles in NSD3-amplified LUSC. Next, pharmaco-transcriptomic correlation analysis was implemented to identify the molecular underpinnings and therapeutic vulnerabilities in LUSC. We revealed that NSD3-amplified LUSC presents a non-inflamed tumor immune microenvironment (TIME) state in multiple independent LUSC patient cohorts. Predictably, elevated NSD3 expression was correlated with a worse immunotherapy outcome. Further molecular characterizations revealed that the high activity of unfolded protein response (UPR) signaling might be a pivotal mediator for the non-immunogenic phenotype of NSD3-amplified LUSC. Concordantly, we showed that NSD3-amplified LUSCs exhibited a more sensitive phenotype to compounds targeting UPR branches than the wild-type group. In brief, our multi-level analyses point to a previously unappreciated immunological role for NSD3 and provide therapeutic rationales for NSD3-amplified squamous lung cancer.

Keywords: NSD3; immunotherapy; lung squamous cell carcinoma (LUSC); tumor immune microenvironment (TIME); unfolded protein response (UPR).

Figure 1

NSD3 is a key mutational driver of LUSC carcinogenesis. (A,B) Correlation analysis between the amplification status and mRNA expression of NSD3 in the TCGA LUSC patient cohort (n = 491). The Wilcoxon rank-sum test was used for comparison, and p < 0.05 was considered significant. (C) Receiver operating characteristic (ROC) curves manifest that for NSD3, the corresponding mRNA level is a sensitive marker to predict its amplification state. (D,E) The top 20 differentially expressed genes (DEGs) between the non- and amplification groups of NSD3 shown in the heatmap were ranked by the adjusted p-value (D). Data were downloaded from the TCGA LUSC cohort (n = 491). Gene set enrichment analysis (GSEA) was performed based on the transcriptomic data from the TCGA-LUSC cohort stratified by NSD3 genetic status. (F) Correlation analysis between the copy number of NSD3 and CRISPR genetic knockout score across LUSC (n = 22) and LUAD cell lines (n = 52). Data were downloaded from The Cancer Dependency Map Project (DepMap; 22Q2 Chronos). (G) The difference in the CRISPR genetic knockout score between the non- and amplification groups of NSD3 in LUSC cell lines (n = 22). Data were downloaded from The Cancer Dependency Map Project (DepMap; 22Q2 Chronos). A two-tailed unpaired t-test was used for comparison, and p < 0.05 was considered significant.

Figure 2

NSD3 shapes a non-immunogenic phenotype in the TCGA-LUSC. The Wilcoxon rank-sum test was used for all comparison, and p < 0.05 was considered significant. (A) Distribution plots of tumor purity, ESTIMATE score, immune score, and stromal score were calculated using the ESTIMATE algorithm in the non- and amplification groups of NSD3 across the TCGA LUSC cohort (n = 491). (B,C) The expression pattern of immunomodulators (chemokines, immunostimulators, MHC, and receptors) in the non- and amplification groups of NSD3 across the TCGA LUSC cohort (n = 491). Data with statistical significance were shown in (C). (D) The expression levels of tumor-infiltrating immune cells were calculated using five various algorithms (TIMER, EPIC, MCP-counter, quanTIseq, and TISIDB) in the non- and amplification groups of NSD3 across TCGA LUSC patients’ cohort (n = 491). Data with statistical significance were shown. (E) The activities of various steps involved in the cancer immunity cycle were calculated by the ssGSEA algorithm in the non- and amplification groups of NSD3 across the TCGA LUSC cohort (n = 491). ns: none significant; * p < 0.05; ** p < 0.01. (F) The difference in the T-cell inflamed gene score between the non- and amplification groups of NSD3.

Figure 3

NSD3 defines a non-immunogenic phenotype in the CPTAC-LUSC. The Wilcoxon rank-sum test was used for comparison, and p < 0.05 was considered significant. (A) Correlation analysis between the mRNA and protein expression of NSD3 in the CPTAC LUSC cohort (n = 108). (B) Distribution plots of tumor purity, ESTIMATE score, immune score, and stromal score were calculated using the ESTIMATE algorithm in the NSD3-low and NSD3-high groups according to the corresponding protein levels at the cut-off of zero across the CPTAC LUSC cohort (n = 108). (C) The difference in immunomodulators expression (chemokines, immunostimulators, MHC, and receptors) between the NSD3-low and NSD3-high groups according to the corresponding protein levels at the cut-off of zero across the CPTAC LUSC patients’ cohort (n = 108). (D) The expression pattern of immune cell-related gene markers in the NSD3-low and NSD3-high groups according to the corresponding protein levels at the cut-off of zero across the CPTAC LUSC cohort (n = 108). (E) The activities of various steps involved in the cancer immunity cycle were calculated by the ssGSEA algorithm in the NSD3-low and NSD3-high groups according to the corresponding protein levels at the cut-off of zero across the CPTAC LUSC cohort (n = 108). ns: none significant; * p < 0.05; ** p < 0.01. (F) The difference in the T-cell inflamed gene score between the NSD3-low and -high groups according to the corresponding protein levels at the cut-off of zero in the CPTAC LUSC cohort (n = 108).

Figure 4

The association between NSD3 expression and immune infiltrates in the LUSC tissue microarray. In all, p < 0.05 was considered significant. (A,B) The difference in the NSD3 IHC score (whole slide region) between the para-tumor and tumor groups in a LUSC tissue microarray (n = 90; (B). Representative IHC images are shown in (A). Original overall magnification, ×200. Scale bar: 50 μM. A two-tailed unpaired t-test was used for comparison. (C) The expression levels of the NSD3 IHC score (whole slide region) in different pathological stages of LUSC patients (n = 90). One-way analysis of variance (ANOVA) followed by Dunnett’s multiple comparison test was used for comparison. (D) Kaplan–Meier analysis of LUSC patients based on the protein level of NSD3 in the TMA array (n = 90) with high- (in red) or low- (in blue) were stratified by the optimal cut-off value of the individual protein expression across all patients using the surv_cutpoint function in the R “maxstat” package. Log-rank test was used for comparison. (E) IHC staining for indicated proteins in the LUSC TMA array (n = 90). Representative IHC images are shown. Original overall magnification, ×200. Scale bar: 50 μM. (F–H) Infiltration levels of CD8+ T-cells and the expression levels of NSD3 (G) and PD-L1 (H) in three different immune subtypes of LUSC samples (n = 90). One-way analysis of variance (ANOVA) followed by Dunnett’s multiple comparison test was used for comparison. (I) Correlation analysis between the infiltration of CD8+ T-cells and the PD-L1 IRS score in the LUSC TMA array (n = 90). (J) The difference in the exhausted- and cytotoxic T-cell signature between the non- and amplification groups of NSD3 in the TCGA LUSC cohort (n = 491). Two-way analysis of variance (ANOVA) followed by Sidak’s multiple comparison test was used for comparison. (K) Heatmap showing the Spearman correlation coefficients between the mRNA expression of NSD3 and cytotoxic T-cell-related gene markers in the TCGA LUSC cohort (n = 491). (L) The difference in the protein levels of cytotoxic T-cell-related gene markers (GZMA, GZMB, and GZMK) and exhausted T-cell gene marker (HAVCR2) between the NSD3-low and -high groups in the CPTAC LUSC cohort (n = 108). Two-way analysis of variance (ANOVA) followed by Sidak’s multiple comparison test was used for comparison. ns: none significant; * p < 0.05; ** p < 0.01.

Figure 5

Clinical predictive value of NSD3 for immunotherapy outcome in LUSC. (A) Heatmap showing the Spearman correlation coefficients between the mRNA expression of NSD3 and a panel of inhibitory immune checkpoints in the TCGA LUSC cohort (n = 491). (B) Immunophenoscore (IPS) distribution plot in the non- and amplification groups of NSD3 across the TCGA LUSC cohort (n = 491). The Wilcoxon rank-sum test was used for comparison, and p < 0.05 was considered significant. (C) mRNA expression of NSD3 in the different clinical response groups (DCB: durable clinical benefit; non-durable clinical benefit) based on the transcriptomic dataset from the NSCLC immunotherapy cohort (n = 27; GSE135222). A two-tailed unpaired t-test was used for comparison, and p < 0.05 was considered significant. ns: none significant. (D) Kaplan–Meier analysis based on the mRNA level of NSD3 in the NSCLC immunotherapy cohort (n = 27; GSE135222) with high- (in red) or low- (in blue) expression was stratified by the optimal cut-off value of the individual expression across all patients using the surv_cutpoint function in the R “maxstat” package. The log-rank test was used for comparison, and p < 0.05 was considered significant.

Figure 6

The internal links between UPR and non-immunogenic features of NSD3-amplified LUSC. (A) The difference in the unfolded protein response (UPR) gene signature between the non- and amplification groups of NSD3 in the TCGA LUSC cohort (n = 491). The UPR gene signature was scored as the sum of the hallmark UPR gene sets. The Wilcoxon rank-sum test was used for comparison, and p < 0.05 was considered significant. (B) Kaplan–Meier analysis based on the UPR gene score in NSD3-amplified LUSC patients (upper panel; n = 207) or non-amplified group (lower panel; n = 236), with high (in red) or low (in blue) expression, was stratified by the optimal cut-off value of the individual expression across all patients using the surv_cutpoint function in the R “maxstat” package. The log-rank test was used for comparison, and p < 0.05 was considered significant. (C) Correlation analysis between the UPR gene score and several gene signatures was calculated using the “GSVA” R package in the TCGA LUSC cohort (n = 491). (D,E) Correlation analysis between the UPR gene score and a panel of immune cell-related gene markers (D) as well as inhibitory immune checkpoints (E) in the TCGA LUSC cohort (n = 491). (F,G) The expression levels of UPR gene signature in the indicated groups. Transcriptomic data were obtained from transgenic mice carrying a lung-specific active mutant (T1232A) or genetic knockout of Nsd3 (GSE149212 & GSE149482). A two-tailed unpaired t-test was used for comparison, and p < 0.05 was considered significant. * p < 0.05; **** p < 0.001. (H) The difference in the indicated immune cell types and the type II-IFN response was calculated using the “GSVA” R package between the wild-type and Nsd3 active mutant (T1232A) groups. A two-tailed unpaired t-test was used for comparison, and p < 0.05 was considered significant.

Figure 7

Prioritization of cancer drug targets for NSD3-amplified LUSC. (A) Heatmap showing the estimated half maximal inhibitory concentration (IC₅₀) values of candidate therapeutic drugs calculated using the “pRRophetic” R package across the TCGA LUSC cohort (n = 491). The top 10 compounds ranked by the adjusted p-value are shown. The Wilcoxon rank-sum test was used for comparison, and p < 0.05 was considered significant. (B) Correlation analysis between the estimated IC₅₀ values of candidate therapeutic drugs and pathway enrichment scores calculated using the “GSVA” R package. (C) Heatmap showing the Spearman correlation coefficients between the mRNA levels of NSD3 and a panel of UPR genes, which displayed an elevated expression pattern in NSD3-amplified LUSC. (D) Correlation analysis between the copy number of NSD3 and CRISPR genetic knockout score of EIF4EBP1 and LSM1 across LUSC cell lines (n = 22). Data were downloaded from The Cancer Dependency Map Project (DepMap; 22Q2 Chronos).