Transcriptomic analysis of ROS1+ non-small cell lung cancer reveals an upregulation of nucleotide synthesis and cell adhesion pathways

Abstract

Introduction: The transcriptomic characteristics of ROS1+ non-small cell lung cancer (NSCLC) represent a crucial aspect of its tumor biology. These features provide valuable insights into key dysregulated pathways, potentially leading to the discovery of novel targetable alterations or biomarkers.

Methods: From The Cancer Genome Atlas (TCGA) and the Gene Expression Omnibus (GEO) databases, all available ROS1+ (n = 10), ALK+ (n = 5) and RET+ (n = 5) NSCLC tumor and ROS1+ cell line (n = 7) RNA-sequencing files were collected. In addition, 10 healthy lung RNA-seq samples were included. Differential gene expression with DESeq2 (R package) and gene co-expression (WGCNA, R package) analyses were performed. Functional annotation was performed through Gene Set Enrichment Analysis (GSEA) using Webgestalt and RNAseqChef, Over-Representation Analysis (ORA) through Enrichr. iRegulon was used to identify enriched transcription factors that regulate a gene co-expression module.

Results: ROS1+ NSCLC samples were significantly enriched for the nucleotide synthesis and cell adhesion KEGG pathways compared to ALK+ and RET+ samples. Moreover, NOTCH1 was significantly downregulated in ROS1+ NSCLC and PD-L1 was weakly expressed. When comparing ROS1+ tumor versus cell line transcriptomes, an upregulation of MYC and MET was found in cell lines together with a significantly decreased expression of HER3, HER4 and BRAF. Within ROS1-tumors, GJB2 was overexpressed in the CD74- and CLTC-ROS1+ subgroups. The differential expression of IL20RB and GJB2 in cell lines was confirmed through RT-qPCR. Finally, the gene co-expression analysis unveils a gene cluster involving cell cycle-related genes which significantly correlates with the disease stage of patients. In addition, we propose TFDP1 and ISL1 as key ROS1-specific transcription factors.

Conclusion: This study highlights cell adhesion and nucleotide synthesis as crucial signatures in ROS1+ NSCLC. The upregulation of GJB2 may serve as a prognostic biomarker, along with IL20RB, a known mediator of bone metastases. Furthermore, TDFP1 and ISL1 were identified as relevant transcription factors that could potentially regulate the biological processes in ROS1-rearranged NSCLC.

Keywords: RNA-sequencing; ROS1+ NSCLC; cell adhesion; gene co-expression; nucleotide synthesis; prognostic biomarker.

Figure 1

ROS1+ NSCLC signature. (A) Gene expression levels of the oncogenic kinases in ALK+, RET+, ROS1+ tumors and normal lung tissue. (B) Venn diagram reflecting the significant differentially expressed genes (DEGs) between two comparisons: (1) ROS1+ tumor specimens and cell lines versus normal lung tissue and (2) ROS1+ tumor specimens and cell lines versus ALK+ and RET+ tumor samples. (C) Circos plot reflecting the top 10 enriched biological processes in ROS1+ NSCLC. (D) IL20RB mRNA levels in HCC-78, CUTO-28 and CUTO-37 cell lines expressed as calibrated and normalized relative mRNA quantity (CNRQ) ± SEM. A 1-way ANOVA test was performed considering a Bonferroni correction and a chosen α = 0.05. (E) Over-representation analysis (ORA) depicting the key enriched pathways in ROS1+ tumor samples and cell lines versus normal lung tissue. (F) GSEA summarizing the dysregulated KEGG pathways in ROS1+ NSCLC compared to normal lung tissue and ALK+/RET+ tumor specimens.

Figure 2

Differences between ROS1+ and ALK+/RET+ tumors. (A) Clustergram reflecting the differences between ROS1+ and ALK+ tumor specimens. (B) GSEA resulting from the significantly DEGs between ROS1+ and ALK+ tumors. (C) Clustergram reflecting the differences between ROS1+ and RET+ tumor specimens. (D) GSEA resulting from the significantly DEGs between ROS1+ and RET+ tumors. (E) Gene expression levels of the NOTCH family and (F) canonical tumor-suppressor genes and oncogenes.

Figure 3

Transcriptomic differences between ROS1+ tumors and patient-derived cell lines. (A) Clustergram highlighting the differences between ROS1+ tumor specimens and patient-derived cell lines. (B) Volcano plot containing more than 6,000 DEGs. (C) Differential expression of relevant oncogenes and tumor-suppressor genes across ROS1+ NSCLC tumor subtypes and patient-derived cell lines. (D) Distribution across chromosomes of the top 500 significant up- and downregulated genes between tumor specimens and cell lines. (E) GSEA elucidating the KEGG pathways up- or down-regulated in patient-derived cell lines versus tumor specimens.

Figure 4

Transcriptomic traits of ROS1+ patient-derived NSCLC cell lines. (A) Phenotypes of some of the cell lines characterized in the study. Phase contrast microscopy images taken at 10X magnification. (B) Principal component plot. (C) Dendrogram and (D) clustergram reflecting the differences across cell lines. (E) GSEA indicating the significantly enriched hallmarks in each cell line. (F) Enriched hallmarks in CUTO-28 and (G) CUTO-37 cell lines.

Figure 5

Impact of the ROS1 fusion partner in the tumor transcriptome. (A) Principal component analysis (PCA) of CD74-, SLC34A2- and EZR-ROS1 tumor samples. (B) Clustergram of the cell lines based on their transcriptomic traits (C) GSEA depicting the enriched hallmarks in CD74-ROS1 and (D) EZR-ROS1 NSCLC samples.

Figure 6

The role of GJB2 expression in ROS1+ NSCLC. (A) GJB2 expression levels across oncogene-driven NSCLC and normal lung tissue. (B) GJB2 expression across different ROS1 rearrangements in tumor specimens. (C) GJB2 mRNA levels in HCC-78, CUTO-28 and CUTO-37 cell lines expressed as calibrated and normalized relative mRNA quantity (CNRQ) ± SEM. A 1-way ANOVA test was performed considering a Bonferroni correction and a chosen α = 0.05. (D) Kaplan-Meier survival plot of The Cancer Genome Atlas (TCGA) lung adenocarcinoma (LUAD) cohort divided in GJB2-high or low expression levels. (E) GJB2 single-nucleotide variants identified in LUAD patients.

Figure 7

Gene co-expression analysis. (A) Eigengene expression patterns of the green-yellow module across samples. (B) ORA comprising the green-yellow gene module and (C) red module. (D) Workflow performed to identify key transcription factors (TFs) that regulate the expression of genes in a given co-expression module. (E) ORA performed with the TFDP1 regulon.