Multi-omics profiling reveals CDKN2A mutant early-onset lung squamous cell carcinoma with a “hot” tumor immune microenvironment

Abstract

Background

Early-onset lung squamous cell carcinoma (LUSC) is a rare and poorly characterized entity. Its distinct clinicopathological, genomic, and tumor immune microenvironment (TIME) profiles remain to be elucidated. This study sought to explore the molecular landscape and immunophenotype of early-onset LUSC, with a focus on uncovering potential genomic associations with its clinical features.

Methods

In this retrospective, single-center cohort study, we conducted a comprehensive multi-omics analysis on patients with surgically resected, treatment-naïve early-onset LUSC (subjects aged ≤40 years, n=21) between 2015 and 2024 using whole-exome sequencing (WES) and digital spatial profiling (DSP). To contextualize our findings, we conducted a comparative analysis using publicly available LUSC data from The Cancer Genome Atlas (TCGA) via the Xena platform.

Results

WES revealed a high mutation frequency of CDKN2A (47.6%), which consistently co-occurred with TP53 mutation. Stratification by CDKN2A status identified two subgroups: CDKN2A-mutant (CDKN2A^mut, n=10) and CDKN2A-wild-type (CDKN2A^wild, n=11). The CDKN2A^mut group exhibited more aggressive clinicopathological features, including higher Ki67 index and frequent vascular invasion, alongside a distinct genomic profile. Despite its aggressive phenotype, the CDKN2A^mut group displayed a robust “hot” TIME, characterized by enriched cytotoxic and exhausted CD8⁺ T cells, a higher intra-tumoral CD8⁺/Foxp3⁺ ratio, and significantly elevated programmed death-ligand 1 (PD-L1). Comparative analysis with public LUSC databases indicated that this specific immunogenic association was unique to the early-onset context and not observed in general LUSC populations. Patients in CDKN2A^mut group showed a trend toward shorter overall survival (OS).

Conclusions

Our study identifies CDKN2A mutation as a hallmark of a unique, aggressive yet immunogenic subtype of early-onset LUSC. This association appears to be specific to young patients, highlighting age of onset as a key biological variable. These findings provide a rationale for further investigation of CDKN2A status as a potential biomarker to guide immunotherapy strategies in early-onset LUSC.

Highlight box.

Key findings

• This multi-omics study identifies an aggressive subtype of early-onset lung squamous cell carcinoma (LUSC) defined by CDKN2A mutation, which co-occurs with TP53 mutation. Despite its aggressiveness, this CDKN2A^mut subtype harbors a robust “hot” tumor immune microenvironment, characterized by enriched cytotoxic and exhausted CD8⁺ T cells, elevated programmed death-ligand 1 expression, and a favorable intra-tumoral CD8⁺/Foxp3⁺ ratio. This specific association between CDKN2A mutation and an immunogenic phenotype is uniquely prominent in the early-onset LUSC population, highlighting patient age as a key biological variable.

What is known and what is new?

• Early-onset LUSC is rare and poorly characterized. While CDKN2A and TP53 mutations are common in general LUSC, their clinical and immunological significance in young patients remains unclear.

• This is the first comprehensive multi-omics profiling of early-onset LUSC. It newly demonstrates that CDKN2A mutation defines an aggressive subtype paradoxically associated with a therapeutically relevant “hot” immune microenvironment. Furthermore, it reveals that this genotype-immunophenotype link is specific to the early-onset context, redefining the biological understanding of this patient group.

What is the implication, and what should change now?

• The findings suggest that in young LUSC patients, CDKN2A mutation status could serve as a practical biomarker to identify a subgroup with a pre-existing but suppressed anti-tumor immune response, who may thus be prioritized for immunotherapy.

• This study provides a potential rationale for incorporating CDKN2A mutation testing into the molecular workup of early-onset LUSC. It calls for future prospective trials to validate immunotherapy efficacy in this biomarker-defined subgroup.

Introduction

Lung cancer has an extremely high incidence rate and is one of the most common malignant tumors (1,2). The majority of patients are elderly males, and it is generally accepted the occurrence of lung squamous cell carcinoma (LUSC) is positively correlated with the duration of smoking. However, it is worth noting that approximately 4% of LUSCs occur in young patients (3-5), and the incidence rate of the early-onset LUSC has been increasing (6). These early-onset cases impose a substantial social burden, and extant literature suggests that they possess a number of distinctive characteristics and challenges, which can be summarized as follows:

Unclear age threshold: there is no consensus on the age limit for early-onset LUSC, with proposed thresholds ranging from 35 to 60 years of age (7-11). This variation reflects differences in the consideration of reproductive potential within the definition (12,13).

Uncertain etiology: the etiology of the condition, particularly the role of known risk factors such as smoking, remains incompletely understood (4,14), suggesting the potential for distinct pathogenic mechanisms (3,15).

Incompletely characterized clinicopathological profiles: features including clinical presentation, pathology, staging, prognosis, and optimal therapeutic strategies require further elucidation in this subgroup (16,17).

Undetermined molecular features: the molecular landscape of early-onset LUSC, including the prevalence of common LUSC-associated genetic alterations, remains largely unexplored (18).

Unclear immune microenvironment characteristics: although the majority of patients diagnosed with LUSC have been shown to benefit from immunotherapy, the immune microenvironment characteristics of early-onset individuals remain to be elucidated (19).

Given these knowledge gaps, this study aims to comprehensively explore the multi-omics characteristics of the early-onset LUSC. Through integrated analysis, we intend to elucidate its distinctive features and provide insights for clinical management and treatment strategies. We present this article in accordance with the REMARK reporting checklist (available at https://tlcr.amegroups.com/article/view/10.21037/tlcr-2025-1130/rc).

Methods

Clinical information and cohort definition

Cohort selection and sample size

Due to the extreme rarity of early-onset LUSC, a formal prospective sample size calculation was not feasible. This comprehensive study aimed to include all consecutively available and eligible cases (treatment-naïve, surgically resected, age ≤40 years) from the specimen bank of the Department of Pathology, Cancer Hospital Chinese Academy of Medical Sciences between 2015 and 2024 to form a complete cohort for this initial, in-depth multi-omics characterization. This study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments. The study was approved by Ethics Committee of National Cancer Center/Cancer Hospital, Chinese Academy of Medical Sciences and Peking Union Medical College (approval No. 22/453-3655) and individual consent for this retrospective analysis was waived. A detailed flowchart of patient selection is provided in Figure 1. Finally, a cohort of 21 LUSC specimens was retrospectively obtained. The initial screening pool consisted of all lung cancer cases within the specified age range during the study period, from which LUSC cases meeting the inclusion criteria were selected.

Figure 1.

Flowchart of patient selection for the early-onset LUSC cohort. The initial screening of the institutional pathology database (January 2015 to June 2024) identified 116 samples from 95 patients with a potential diagnosis of LUSC at age ≤40 years. After rigorous pathological review and application of exclusion criteria, 95 samples were excluded. The final study cohort consisted of 21 formalin-fixed paraffin-embedded tumor samples from treatment-naïve patients with surgically resected early-onset LUSC. LUAD, lung adenocarcinoma; LUSC, lung squamous cell carcinoma; SCLC, small cell lung cancer.

Definition of early-onset LUSC

In this study, “early-onset” LUSC was defined as disease diagnosed at or before the age of 40 years. This threshold was established based on two pillars of evidence: (I) a review of epidemiological data from the Global Burden of Disease (GBD) study, which indicates a marked increase in lung cancer incidence after the age of 40 in the general population (Figure 2A); and (II) integration of findings from previous literature, which supports 40 years as a clinically relevant and widely utilized cut-off for defining early-onset lung cancer (8,17,20,21).

Figure 2.

Study rationale and genomic landscape of early-onset LUSC. (A) Age-stratified global incidence rates of lung cancer (data from the Global Burden of Disease, 2015-2021), showing the sharp rise after age 40, which supports the early-onset definition (≤ 40 years). Each line represents a different year. (B) The top 10 mutation spectrum in early-onset LUSC from present data (early-onset LUSC, n=21). LUSC, lung squamous cell carcinoma; TMB, tumor mutational burden.

Data collection and survival endpoint

Basic and clinical data was obtained from patient medical records. Overall survival (OS) was identified as the time from the date of surgery to the date of last follow-up or death. The survival curves were plotted using the GraphPad Prism software (version 10.1.2).

Histopathology evaluation and immunohistochemistry (IHC)

Histological features such as degree of differentiation, lymph node metastasis, and nerve and vascular invasion were recorded. Following the manufacturer’s protocols, a fully automated IHC staining instrument (Roche BenchMark ULTRA) was used to perform two-step IHC staining on 4-µm-thick continuous sections via the Envision program. A panel research IHC marker was applied to all cohort samples to characterize the proliferation activity and immune microenvironment.

• ❖ Proliferation marker: Ki-67;

• ❖ Immune checkpoint protein: programmed death-ligand 1 (PD-L1);

• ❖ T lymphocyte lineage and functional markers: CD4, CD8 Foxp3;

• ❖ Cytotoxic effector molecules: Granzyme B (GZMB), Perforin (PRF1);

• ❖ Immune exhaustion markers: TIM-3, TIGIT, LAG-3.

The complete details of antibodies are provided in Table S1. All hematoxylin and eosin (H&E) and IHC slides were evaluated by two senior pathologists who were blinded to the patients’ pathological information and clinical outcomes to minimize assessment bias.

Quantitative analysis was conducted using the open-source software QuPath (version 5.1). GraphPad Prism software (version 10.1.2) was employed for graphic drawing and statistical analysis using the independent samples t-test or Mann-Whitney test.

Database

Lung cancer incidence data were obtained from the Global Burden of Disease (GBD) database (http://ghdx.healthdata.org/gbd-results-tool) for the years 2015 to 2021. Specifically, we extracted the annual age- and sex-specific incident counts and corresponding population denominators from the GBD 2021 Results Tool released by the Global Burden of Disease Collaborative Network [Seattle, United States: Institute for Health Metrics and Evaluation (IHME), 2022]. Records with missing incidence values or unknown sex were excluded. Age was categorized into <20, 20–24, 25–29, 30–34, 35–39, 40–44, 45–49, 50–54, 55–59, 60–64, 65–69, 70–74, 75–79, and 80+ years. All analyses were conducted in R version 4.5 using the tidyverse (v2.0) and openxlsx (v4.2.8) packages. The extraction and analysis scripts are available from the corresponding author upon reasonable request.

Somatic mutation data of LUSC were retrieved from the UCSC Xena platform (GDC TCGA LUSC cohort, version 08-05-2024, hg38/Illumina, n=496; http://xena.ucsc.edu). Samples were restricted to primary tumors (TCGA sample barcode ending in “-01A”) and only patients with an unequivocal diagnosis age were retained. After matching with clinical information, we obtained somatic mutation data [whole-exome sequencing (WES)] for 471 tumors and corresponding RNA-seq data for 473 tumors for subsequent analyses. Notably, only two early-onset samples (age ≤40 years) were present in this dataset. This dataset was used as a representative cohort of conventional LUSC for comparative genomic and transcriptomic analyses against our early-onset cohort.

WES

Genomic DNA was extracted from surgically removed formalin-fixed paraffin-embedded (FFPE) tumor tissues and their corresponding adjacent normal tissues using the GeneRead DNA FFPE Tissue Kit (QIAGEN, Germany), following the manufacturer’s instructions. DNA quantification was performed using the dsDNA HS Assay Kit (ThermoFisher Scientific, USA). Subsequently, sequencing libraries were constructed using the SureSelect XT Human All Exon V6 Kit (Agilent Technologies, USA), as previously described (22). The libraries were subjected to sequencing on the DNBSEQ T7 platform, generating paired-end reads of 100 bp. The average effective sequencing depths were ≥300× and 150× for tumors and adjacent normal tissues, respectively. The insert size of DNA fragments was ≥120 bp. To ensure data quality, the average Q30 score was ≥85%, and the duplication rate was ≤20%.

Data processing and somatic variant calling

Data processing pipeline

Raw sequencing data were processed following an enhanced pipeline. Adapters and low-quality reads were trimmed using Fastp (v0.23.2). Clean reads were aligned to the human reference genome hg19 using the Burrows-Wheeler Aligner (BWA-mem, v0.7.12) (23). SAMtools (v1.3) was used for format conversion, sorting, and indexing. PCR duplicates were marked using SAMBLASTER (v0.1.22). To ensure high-quality variant calling and address potential sequencing artifacts, we implemented key steps from the Genome Analysis Toolkit (GATK) best practices. Base quality scores were recalibrated using GATK\’s BaseRecalibrator and ApplyBQSR tools (GATK v4.2.0.0). Tumor purity was computationally estimated using two distinct approaches: All-FIT and FACETs (22,24).

Somatic variant calling and filtering

Somatic single nucleotide variants (SNVs) and insertions/deletions (indels) were identified using VarDict (v1.7.0) in tumor-normal paired mode. Called variants were stringently filtered. SNVs and indels with a variant allelic fraction (VAF) <0.02 or a total read coverage <10 reads in either the tumor or the matched normal sample were excluded. Putative germline variants present in the matched normal tissue were subtracted. The remaining high-confidence somatic variants were annotated using the Ensembl Variant Effect Predictor (VEP, v97).

Tumor mutational burden (TMB) was defined as the total number of all non-synonymous somatic mutations and indels detected, divided by the total exonic megabases (Mb) covered by the sequencing panel (34 Mb), reported as mutations per Mb.

Digital spatial profiling (DSP) and data analysis

DSP was performed on all samples using the NanoString GeoMx^® platform. Tissue microarrays (TMAs) were constructed by punching two 2-mm-diameter cores from each FFPE tissue block. For morphological visualization and region selection, sequential sections were stained with a multiplex immunofluorescence panel containing antibodies against Pan-cytokeratin (PanCK, epithelial marker; NBP2-33200, Novus Biologicals, green), CD45 (leukocyte common antigen; NBP2-34528, Novus Biologicals, red), and the nuclear dye Syto13 (S7575, Invitrogen, blue).

Region selection and profiling

Regions of interest (ROIs) encompassing tumor epithelium (PanCK⁺) were manually selected based on the immunofluorescence images, with one ROI selected per tissue core. Within each ROI, the instrument digitally quantified the abundance of a pre-defined panel of RNA targets (the GeoMx Whole Transcriptome Atlas - Human RNA for NGS, NanoString) from the illuminated area of interest (AOI).

Quantitative analysis of each AOI was performed using the nCounter instrument (NanoString Technologies, USA) (25). To meet quality control standards, an AOI was required to have a nucleus count >20 or a surface area >16,000 µm². The top 25% of expression levels in the AOI/ROI distribution were used for standardization.

Differential expression and functional analysis

Differentially expressed genes (DEGs) between the CDKN2A^mut and CDKN2A^wild groups were identified using a false discovery rate (FDR) <0.05 and an absolute log₂(fold change) >1. Functional enrichment analysis of these genes was performed using the MSigDB Hallmark gene sets. To characterize the immune microenvironment, we utilized ImmuCellAI for deconvolution and quantification of immune cell abundances. Statistical comparisons between groups for all DSP-derived metrics were performed using the Wilcoxon rank-sum test, with a two-sided P<0.05 considered significant.

Statistical analysis

The candidate variables initially examined in this study included: patient demographics (age, sex), clinical characteristics (smoking history, tumor differentiation, lymph node metastasis, vascular invasion, neural invasion), key pathological indices (Ki67 proliferation index), genetic alterations (TP53 and CDKN2A mutation status, TMB), and immune microenvironment features (PD-L1 expression, densities of CD8⁺ T cells, Foxp3⁺ T cells, cytotoxic T cells, and exhausted T cells). All analyses were pre-specified as comparisons between the CDKN2A^mut and CDKN2A^wild groups.

Comparative tests

Categorical variables were compared between groups using Fisher’s exact test in SPSS version 26 (SPSS Inc., Chicago, IL, USA). Continuous variables were assessed for normality using the Shapiro-Wilk test. Based on the distribution, group comparisons for continuous variables (e.g., Ki67 index, TMB, immune cell densities) were performed using the independent two-sample Student’s t-test (for normally distributed data) or the Mann-Whitney U test (for non-normally distributed data). Differences in immune cell composition from DSP were assessed using the Wilcoxon rank-sum test.

Survival probabilities were estimated using the Kaplan-Meier method, and differences between groups were compared with the log-rank test. The number of patients at risk at specified time points is indicated in the survival plots.

Multiple testing and data completeness

Given the exploratory nature of the immune cell analyses involving multiple cell types, P values are presented without adjustment for multiple comparisons; however, findings are interpreted with caution, and emphasis is placed on effect sizes and consistency with orthogonal IHC data. No missing data were encountered for the primary variables (CDKN2A status, TP53 status, OS, and key immunohistochemical markers). The dataset was complete for all analyzed variables.

Pathway and visualization analysis

Functional enrichment analysis of DEGs identified from DSP was performed using the clusterProfiler R package (version 4.16.0) with reference to the Kyoto Encyclopedia of Genes and Genomes (KEGG) database. Enrichment results with an adjusted P value (FDR) <0.05 were considered significant. Differential expression results were visualized using volcano plots generated with the ggplot2 package.

Software and visualization

All statistical tests were two-sided, and statistical significance was set at P<0.05. Statistical analyses were performed using a combination of software: R software (version 4.5.0) with the ‘survival’, ‘survminer’, and ‘ggplot2’ packages was used for survival probability estimation (Kaplan-Meier), log-rank testing, and the generation of corresponding survival curves. Categorical data were compared using Fisher’s exact test in SPSS version 26 (SPSS Inc., Chicago, IL, USA). GraphPad Prism software (version 10.1.2) was employed for the generation of other graphical representations (e.g., box plots, bar charts) and for performing complementary statistical tests (e.g., t-tests, Mann-Whitney U tests) where applicable, ensuring consistency with the primary R-based analyses.

Results

Patient cohort and baseline feature

A cohort of 21 patients with treatment-naïve, surgically resected early-onset LUSC (age ≤40 years) was established for analysis, as detailed in the selection flowchart (Figure 1). The clinicopathological characteristics of the cohort are summarized in Table 1.

Table 1. The clinical information of the cohort of early-onset LUSCs.

Variable	Present data (n=21)	CDKN2Amut (n=10)	CDKN2Awild (n=11)	P value
Gender				0.09
Male	17 (81.0)	10 (100.0)	7 (63.6)
Female	4 (19.0)	0 (0.0)	4 (36.4)
Smoking status				0.20
Never smokers	10 (47.6)	3 (30.0)	7 (63.6)
Former/current smokers	11 (52.4)	7 (70.0)	4 (36.4)
Smoking years, average	8.2	11.6	5.1	0.09§
Cigarettes/day, average	8.7	13.0	4.8	0.11§
Morphologic features
Differentiation grade				0.36
Well	2 (9.5)	0 (0.0)	2 (18.2)
Moderately	11 (52.4)	6 (60.0)	5 (45.4)
Poorly	8 (38.1)	4 (40.0)	4 (36.4)
Invasion†				0.09
With	10 (47.6)	7 (70.0)	3 (27.3)
Without	11 (52.4)	3 (30.0)	8 (73.7)
AJCC 8th				>0.99
I–II	14 (66.7)	7 (70.0)	7 (63.6)
III–IV	7 (33.3)	3 (30.0)	4 (36.4)
N stage‡				>0.99
N0	9 (42.8)	4 (40.0)	5 (45.5)
N1–2	11 (52.4)	6 (60.0)	5 (45.5)
NA	1 (4.8)	–	1 (9.0)

Data are presented as n (%) unless otherwise specified. P value: comparison between groups of CDKN2A^mut and CDKN2A^wild. ^†, vascular tumor thrombus and/or nerve invasion; ^‡, statistical calculation without missing values; ^§, Mann-Whitney test. AJCC, American Joint Committee on Cancer; LUSCs, lung squamous cell carcinomas; N, node; NA, not available.

The cohort was predominantly male (81.0%, 17/21). Smoking status was evenly distributed, with 52.4% (11/21) never-smokers and 47.6% (10/21) current or former smokers. The majority of tumors were moderately (52.4%, 11/21) or poorly (38.1%, 8/21) differentiated. Over half of the patients (66.7%, 14/21) presented with stage I or II disease.

Genomic landscape of early-onset LUSC and CDKN2A subgrouping

WES identified somatic alterations in 6,578 genes. An overview of the major genetic alterations is shown in Figure 2B. The most frequently mutated genes were TP53 (76.2%, 16/21), MUC16 (66.7%, 14/21), TTN (66.7%, 14/21), MUC19 (61.9%, 13/21), ZFHX4 (57.1%, 12/21), CDKN2A (47.6%, 10/21). Notably, among the most frequently mutated genes, we focused on CDKN2A and TP53—key tumor suppressor genes of LUSC (26-28) altered in 47.6% and 76.2% separately—as the candidate classifier. Interestingly, CDKN2A mutations were consistently accompanied by TP53 mutations. This genetic stratification established two subgroups: one with co-mutated CDKN2A and TP53 (CDKN2A^mut, n=10), which is highly prevalent in conventional LUSC (28-30), and the other with wild-type CDKN2A (CDKN2A^wild, n=11), which we subsequently compared across multiple dimensions.

Comparative clinicopathological and molecular profiling of the genetic subgroups

Comparative analysis uncovered significant divergences in both clinicopathological and molecular features between CDKN2A^mut and CDKN2A^wild subgroups. Despite the limited sample size, several distinguishing trends and features emerged (Table 1).

Clinically, the CDKN2A^mut group consisted entirely of males (100.0% vs. 63.6%), the vast majority of whom were current or former heavy smokers (defined as ≥20 cigarette-days) (70.0% vs. 36.4%) with a mean smoking history exceeding ten years. Well-differentiated tumors were observed in some CDKN2A^wild cases (18.2%, 2/11) but not in any CDKN2A^mut cases, and vascular/neural invasion appeared more frequent in the CDKN2A^mut group (70.0% vs. 27.3%). Consistent with this aggressive phenotype, the CDKN2A^mut subgroup exhibited a significant higher Ki67 proliferation index (median =60.0% vs. median =40.0%, P=0.03, Figure 3A). All patients underwent radical surgery. Postoperative adjuvant treatments were administered to a higher proportion of patients in the CDKN2A^mut group (60.0%) compared to the CDKN2A^wild group (36.4%) (Table 2).

Figure 3.

Comparative clinicopathological and molecular features of CDKN2A^mut and CDKN2A^wild groups. (A) Ki67 proliferation index is significantly higher in the CDKN2A^mut group (median 60.0% vs. 40.0%; P=0.03). (B) Waterfall plot of somatic mutations in frequently altered genes, showing distinct patterns between groups. (C) TMB is comparable between groups (median: 11.79 vs. 10.06; P=0.31). (D) Base substitution types, and Ti/Tv fractions in the CDKN2A^mut group. (E) Base substitution types, and Ti/Tv fractions in the CDKN2A^wild group. (F) PD-L1 expression (tumor proportion score) is significantly elevated in the CDKN2A^mut group (median 20.0% vs. 1.0%; P=0.001). *, P<0.05; **, P<0.01; ns, not significant. PD-L1, programmed death-ligand 1; Ti/Tv, transition/transversion; TMB, tumor mutational burden.

Table 2. The postoperative treatment and prognosis of early-onset LUSCs.

Variable	TP53mut (N=16)	CDKN2Awild (N=11)	TP53wild/CDKN2Awild (TP53wild, N=5)	TP53mut/CDKN2Amut (CDKN2Amut, N=10)	TP53mut/CDKN2Awild (N=6)	TP53wild/CDKN2Amut (NA†)
Adjuvant therapy						–
Chemotherapy	7 (43.8)	2 (18.2)	1 (20.0)	6 (60.0)	1 (16.7)
Radiotherapy	1 (6.2)	1 (9.1)	0 (0.0)	0 (0.0)	1 (16.7)
Combination	0 (0.0)	1 (9.1)	1 (20.0)	0 (0.0)	0 (0.0)
Without	1 (6.2)	2 (18.2)	2 (40.0)	1 (10.0)	0 (0.0)
NA‡	7 (43.8)	5 (45.4)	1 (20.0)	3 (30.0)	4 (66.6)
Prognosis						–
Dead	4 (25.0)	1 (9.1)	0 (0.0)	3 (30.0)	1 (16.7)
Alive	9 (56.3)	6 (54.5)	4 (80.0)	7 (70.0)	2 (33.3)
Loss	3 (18.7)	4 (36.4)	1 (20.0)	0 (0.0)	3 (50.0)

Data are presented as n (%). NA^†, no matching cases found in the database; NA^‡, lost to follow-up; no postoperative information. LUSCs, lung squamous cell carcinomas.

At the molecular level, the most common divergent mutated genes in the CDKN2A^mut group were CDKN2A (100.0%, 10/10), TP53 (100.0%, 10/10), CSMD3 (70.0%, 7/10), and FAM135B (70.0%, 7/10), which were rarely observed in CDKN2A^wild (Figure 3B, Figure S1). Although the overall burden of somatic mutations was comparable between groups (median TMB: 11.79 vs. 10.06, P=0.31, NS; Figure 3C), their mutational signatures differed markedly. Specifically, the CDKN2A^mut group was characterized by a predominance of smoking-associated C>A transversions (31,32) (47.6%), whereas C>T transitions were most common in the CDKN2A^mut subgroup (38.0%) (Figure 3D,3E). Subsequent comparison of germline mutations between the two patient cohorts revealed no significant differences in TMB or Transition/Transversion ratios (Figure S2).

Having established that the CDKN2A^mut group is associated with aggressive clinicopathological and proliferative features alongside a smoking-related mutational signature, we then investigated whether this extended to the tumor immune microenvironment (TIME).

Immune microenvironment of CDKN2A^mut group

Building upon the genomic stratification, we investigated the transcriptional and immune landscape associated with the CDKN2A^mut group. DSP of the tumor epithelium identified a set of DEGs in the CDKN2A^mut group (Figure S3, Table S2 for the complete DEG list and exploratory KEGG pathway analysis).

The CDKN2A^mut group exhibited markedly upregulated PD-L1 expression compared to the CDKN2A^wild group (median =20.0% vs. 1.0%, P=0.001, Figure 3F), signifying an active anti-tumor immune response. This group was further defined by a profoundly inflamed TIME (Figure 4A). Despite a comparable overall CD8⁺ T cell density, subset analysis uncovered a significant enrichment of both cytotoxic T cells and exhausted T cells within the CDKN2A^mut tumor epithelium, which was validated by IHC for key functional markers (Figure 4B, Table S1).

Figure 4.

Immune microenvironment characterization between the CDKN2A^mut and CDKN2A^wild group. (A) Heatmap depicting the scaled mean abundance of 24 immune cell types within the tumor epithelium and surrounding stroma, stratified by CDKN2A mutation status. Rows represent four tissue-context groups: CDKN2A^mut-Tumor (CDKN2A^mut, tumor epithelium), CDKN2A^mut-Stroma (CDKN2A^mut, stroma), CDKN2A^wild-Tumor (CDKN2A^wild, tumor epithelium), and CDKN2A^wild-Stroma (CDKN2A^wild, stroma). Columns represent immune cell types. The color scale (right) indicates Z-score normalized abundance per cell type across groups, ranging from low (blue) through medium (white) to high (red). (B) Quantitative assessment of key immune features by IHC. Representative images (original magnification, ×200) and corresponding box plots with overlaid individual data points show quantitative IHC results. Exhausted CD8⁺ T cells, identified by co‑expression of CD8 with exhaustion markers (TIM‑3, LAG‑3, and/or TIGIT); Cytotoxic (effector) CD8⁺ T cells, identified by co‑expression of CD8 with cytotoxic effector molecules GZMB and/or PRF1; Tregs, identified by co‑expression of CD4 and Foxp3. Data are presented as mean ± SEM. Statistical significance was determined by two-tailed Student’s t-test or Mann‑Whitney U test, as appropriate. *, P<0.05; **, P<0.01; ns, not significant. GZMB, Granzyme B; IHC, immunohistochemistry; PRF1, Perforin‑1; SEM, standard error of the mean; Tregs, regulatory T cells.

Spatial architecture analysis further highlighted a distinctive pattern: the CD8⁺/Foxp3⁺ T cell ratio was significantly higher inside the tumor nests of CDKN2A^mut cases compared to their stromal compartments. This indicates that CD8⁺ T cells have robustly infiltrated the tumor core, even though regulatory T cells are enriched in the surrounding stroma.

In summary, the CDKN2A^mut group exhibits a “hot” immune phenotype—characterized by enriched effector/exhausted CD8⁺ T cells, elevated PD-L1, and a favorable intra-tumoral CD8⁺/Foxp3⁺ ratio—collectively portraying an inflamed yet immunosuppressed tumor microenvironment.

Prognosis according to CDKN2A mutation status

We then assessed the prognostic impact of CDKN2A mutation with a median follow-up of 70.1 months. Kaplan-Meier analysis showed a trend toward shorter OS in the CDKN2A^mut group compared to the CDKN2A^wild group, though this did not reach statistical significance (P=0.12, Figure 5A). Given the complete co-occurrence of TP53 mutations in all CDKN2A^mut cases, we also evaluated survival by TP53 status alone and by the combined CDKN2A/TP53 status. While no single-gene analysis was significant (for TP53, P=0.20, Figure 5B), patients harboring the CDKN2A^mut/TP53^mut co-mutation exhibited the least favorable survival trend among all subgroups (Figure 5C). The limited sample size precludes definitive conclusions, and the prognostic contribution of each gene within this co-mutation context remains to be disentangled in larger studies.

Figure 5.

Survival analysis based on mutation status. (A) Kaplan-Meier curves for OS stratified by CDKN2A mutation status (CDKN2A^mut vs. CDKN2A^wild; P=0.12). (B) OS stratified by TP53 mutation status alone (P=0.20). (C) OS stratified by combined CDKN2A/TP53 status. Patients with the CDKN2A^mut/TP53^mut co-mutation show the least favorable survival trend (median follow-up: 70.1 months). ns, not significant; OS, overall survival.

Comparison with public databases highlights the distinctiveness of early-onset LUSC

To contextualize our findings within the broader spectrum of LUSC, we performed a comparative analysis using data from The Cancer Genome Atlas (TCGA) LUSC cohort (n=471), which was accessed via the Xena browser, served as a reference for conventional LUSC.

At the genomic level, several key distinctions emerged. First, our early-onset cohort exhibited a significantly higher TMB compared to all older age-decade subgroups (41–50, 51–60, 61–70, 71–79, and ≥80 years) within the dataset (P<0.010 for all comparisons, Figure S4A). Second, the frequencies of neither CDKN2A mutation nor CDKN2A/TP53 co-mutation were markedly elevated in our early-onset cohort relative to every age subgroup in TCGA (Figure S4B,S4C).

Most strikingly, the relationship between CDKN2A alteration and the TIME was fundamentally different In the general LUSC cohort, stratification by either CDKN2A mutation or CDKN2A/TP53 co-mutation status revealed that the wild-type groups were significantly enriched for effector and exhausted CD8⁺ T cell signatures (Figure S4D,S4E)—a pattern diametrically opposed to our findings in the early-onset cohort, where these same immune features were robustly enriched in the CDKN2A^mut group.

This stark reversal confirms that the immunogenic “hot tumor” phenotype linked to CDKN2A mutation is not a general property of LUSC but appears to be specifically associated with the early-onset disease context, highlighting patient age as a critical biological variable modulating genotype-immunophenotype relationships.

Discussion

Early-onset LUSC is a rare and aggressive form of lung cancer whose genetic landscape remains poorly understood. While some reports suggest a favorable prognosis for early-onset NSCLC (5,9,20,33-35), the molecular drivers and immune contexture of early-onset LUSC—particularly the role of key tumor suppressors like TP53 and CDKN2A—remain unclear, with studies even reporting contradictory prognostic associations for TP53 mutations (30,36-38). In this study, we performed a comprehensive multi-omics characterization and identified a distinct clinic-molecular subtype defined by CDKN2A mutation, which exhibited both aggressive tumor features and a pronounced “hot” TIME. Our comparative analysis with public databases further revealed that this specific CDKN2A-associated immunophenotype appears to be uniquely enriched in early-onset LUSC and was not observed in the general LUSC population. This observation highlights the potential importance of age of onset as a biological variable that may modify genotype–immunophenotype relationships.

Consistent with prior studies in conventional LUSC (29,30,39,40), we found a high frequency of CDKN2A alterations, which were invariably co-mutated with TP53 in our early-onset cohort. Although CDKN2A mutation itself is recognized as an adverse prognostic factor in LUSC (30,41), could inducing uncontrolled cell proliferation (29,42-44), its immunomodulatory role is not well defined. In the general LUSC population (e.g., TCGA), CDKN2A mutation status did not correlate with a specific immune landscape. In contrast, within our early-onset cohort, the CDKN2A/TP53 co-mutant subgroup was associated with a dominant smoking-associated mutational signature, a higher proliferative index (Ki67), and—most notably—a more inflamed TIME. This disparity is in line with emerging evidence that host age can shape the tumor immune landscape (45-47) and leads us to hypothesize that the young host milieu—potentially involving differences in immune competence, carcinogen exposure, or tumor evolution—may unmask or modulate the immunomodulatory effects of CDKN2A (and/or its cooperation with TP53) alterations.

The immunophenotype observed in the CDKN2A-mutant early-onset subtype is of particular translational interest. It was characterized by abundant cytotoxic and exhausted CD8⁺ T cells, elevated PD-L1 expression, and a higher intra-tumoral CD8⁺/Foxp3⁺ ratio. This profile aligns with features often described in “hot” tumors, where evidence of pre-existing anti-tumor immunity coexists with immune suppression (48). Such a microenvironment has been correlated with response to immune checkpoint inhibitors in other contexts, as both PD-L1 expression and T-cell exhaustion are among the considered biomarkers (49-51). Furthermore, the elevated TMB observed in this subgroup implies a higher neoantigen load, which could enhance immune recognition (52-54). The spatial distribution of immune cells, with CD8⁺ T cells penetrating tumor nests, could imply a potential for immune reactivation (55,56). Therefore, our findings provide a rationale for investigating the role of immunotherapy in CDKN2A^mut early-onset LUSC, a strategy that warrants evaluation in future clinical studies.

Our findings also highlight FAM135B and CSMD3 as frequent co-mutated genes in this subtype. FAM135B was first identified as an oncogene in esophageal squamous cell carcinoma (40), and CSMD3 mutation has been associated with higher TMB (57). Both have been linked to aggressive disease (40,58-62), and could correlate with immunotherapy response (63-66). Their potential role in shaping the unique microenvironment of early-onset LUSC merits further investigation. It is noteworthy that beyond its canonical role in cell cycle arrest, the CDKN2A and its encoded protein p16 has been implicated in regulating immune responses and aging-related processes, which could influence the tumor microenvironment (67-71). This provides a plausible biological basis for our observed association.

This study has several important limitations. First, the sample size is small, reflecting the rarity of early-onset LUSC; this limits the statistical power of our analyses and necessitates caution in interpretation. Second, the complete overlap between CDKN2A and TP53 mutations in our cohort precludes disentangling their individual contributions to the observed phenotype. The absence of a CDKN2A-mutant/TP53-wild-type subgroup is a notable gap. Third, all patients underwent surgical resection, which may not represent the full spectrum of early-onset LUSC.

Conclusions

In conclusion, our integrated analysis suggests that CDKN2A mutation may define a unique subtype of early-onset LUSC characterized by an aggressive yet immunogenic phenotype. Crucially, this association appears to be more prominent in the early-onset context compared to general LUSC. This work proposes patient age as a relevant factor in LUSC subtyping and puts forward CDKN2A mutation status as a candidate biomarker worthy of future exploration to identify young LUSC patients who might derive benefit from immunotherapy. Validation in larger, prospective studies is essential.

Supplementary

The article’s supplementary files as

Ethical Statement: The authors are accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved. This study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments. The study was approved by Ethics Committee of National Cancer Center/Cancer Hospital, Chinese Academy of Medical Sciences and Peking Union Medical College (approval No. 22/453-3655) and individual consent for this retrospective analysis was waived.

Data Sharing Statement

Available at https://tlcr.amegroups.com/article/view/10.21037/tlcr-2025-1130/dss

tlcr-15-02-37-dss.pdf ^{(109.8KB, pdf)}

DOI: 10.21037/tlcr-2025-1130