Page 123 - Read Online
P. 123
Fang et al. Cancer Drug Resist. 2025;8:42 Page 3 of 13
demonstrating efficacy [13-15] . However, the role of somatic ROS1 mutations (ROS1-Mut) in HNC remains
unexplored. Emerging evidence suggests that somatic mutations in driver genes can modulate tumor
immunogenicity; for example, ALK rearrangements, EGFR mutations, and KRAS mutations correlate with
immunosuppressive tumor microenvironments (TMEs) [16,17] . Paradoxically, colorectal cancers with an
ultramutated phenotype exhibit significantly higher objective response rates and more favorable outcomes
following ICI treatment compared to dMMR/MSI-H tumors . These findings raise the intriguing possibility
[18]
that ROS1-Mut may similarly influence immune response in HNC.
Given the critical need to overcome ICI resistance in HNC, we hypothesize that ROS1-Mut might drive
immunosuppressive mechanisms similar to those induced by oncogenic drivers such as ALK or EGFR. To
investigate this, we integrated multi-omics analyses of 139 ICI-treated HNC patients (MSKCC cohort) and
502 treatment-naïve cases (TCGA cohort) with three specific aims: (1) determine whether ROS1-Mut predict
poor ICI response independently of TMB/PD-L1; (2) characterize the immunogenomic landscape of
ROS1-Mut tumors; and (3) elucidate mechanistic links between ROS1-Mut and MYC-driven immune
evasion. This study identifies ROS1-Mut as candidate mediators of ICI resistance and proposes potential
therapeutic strategies.
METHODS
Study cohorts and data acquisition
We analyzed two independent HNC cohorts: 139 advanced HNC patients treated with ICIs
[anti-programmed death 1 (PD-1)/PD-L1 ± anti-cytotoxic T-lymphocyte-associated protein 4 (CTLA-4)]
from the MSKCC cohort, and 502 treatment-naïve HNC cases from The Cancer Genome Atlas (TCGA)
cohort. Clinical and genomic data were retrieved from the following sources: (1) MSKCC cohort from
Samstein et al. ; (2) whole-exome sequencing (WES) and TMB data from Hoadley et al. ; (3) RNA-seq
[19]
[9]
data from the Genomic Data Commons (GDC; https://portal.gdc.cancer.gov/); and (4) survival data from the
UCSC Xena Browser (https://xenabrowser.net).
Genomic profiling
ROS1 mutation was defined as non-synonymous somatic mutations, including missense, nonsense,
splice-site mutations, and in-frame indels in the coding region of the ROS1 gene. TMB was defined as the
total count of non-synonymous mutations per megabase (mut/Mb), with a TMB-high (TMB-H) threshold
set at > 10 mut/Mb. Neoantigen prediction was performed as previously described . Expressed somatic
[20]
variants and patient-specific HLA alleles (predicted using POLYSOLVER) were used as inputs for the
NetMHCpan 4.0 algorithm . Strong-binding peptides (IC50 < 500 nM) were counted as neoantigens to
[21]
calculate tumor neoantigen burden (TNB).
Transcriptomic and immune analyses
Immune cell composition was estimated using CIBERSORT, which quantified 22 immune cell subsets from
TCGA RNA-seq data based on the LM22 signature matrix (1,000 permutations) . Differential gene
[22]
expression was analyzed using the R package DESeq2 , with thresholds of FDR < 0.05 and log fold change
[23]
2
> 0.5. Immune-related genes were obtained from Danaher et al. to compare expression between ROS1-Mut
and ROS1-wild-type (ROS1-WT) HNC cases in TCGA . Gene set enrichment analysis (GSEA) was
[24]
performed using the R Package ClusterProfiler v3.18.1 . Gene sets were considered significantly enriched at
[25]
an adjusted P value < 0.05 (Benjamini-Hochberg correction).
Statistical analysis
All statistical analyses were performed in R version 4.0.3 (http://www.r-project.org). Categorical variables
were compared using Fisher’s exact test, and continuous variables were assessed with the Wilcoxon
116
