Analysis of genomic heterogeneity and the mutational landscape in cutaneous squamous cell carcinoma through multi-patient-targeted single-cell DNA sequencing

Abstract

Background

Cutaneous squamous cell carcinoma (CSCC) is a prevalent skin cancer with aggressive progression that poses significant challenges, especially in metastatic cases. Single-cell DNA sequencing (scDNA-seq) has become an advanced technology for elucidating tumor heterogeneity and clonal evolution. However, comprehensive scDNA-seq studies and tailored mutation panels for CSCC are lacking.

Methods

We analyzed the genomic landscape of Chinese CSCC patients via a Multi-Patient-Targeted (MPT) scDNA-seq approach. This method combined bulk exome sequencing with Tapestri scDNA-seq. Mutations identified through bulk sequencing were used to design a targeted panel for scDNA-seq. Comparative analysis was conducted to explore the associations between specific gene mutations and clinical characteristics such as tumor stage and patient sex. Clonal evolution analysis was performed to understand the evolutionary trajectories of the tumors.

Results

Bulk sequencing revealed a diverse spectrum of somatic mutations in CSCC tumors, with missense mutations being predominant. The top tumor mutations, such as those in NOTCH1, TP53, NOTCH2, TTN, MUC16, RYR2, PRUNE2, DMD, HRAS, and CDKN2A, presented similar frequencies to those reported in studies in Korean and Caucasian populations. However, the mutation frequencies of HRAS, TTN, MUC16 and MUC4 were significantly different from the Korean and Caucasian populations. Comparative analysis revealed associations between specific gene mutations and clinical characteristics such as tumor stage and patient sex. Clonal evolution analysis via scDNA-seq revealed distinct evolutionary trajectories and their potential correlation with tumor development and patient prognosis. Furthermore, scDNA-seq identified two low-frequency mutation clones, NLRP5 and HMMR, which play important roles in the clonal evolution of CSCC.

Conclusions

Our study introduced a novel MPT sequencing approach for CSCC, providing insights into tumor heterogeneity and clonal evolution. We identified novel mutations and their potential associations with patient survival and tumor characteristics. Overall, our study layed the groundwork for personalized treatment strategies and provides a reference for future applications of the MPT panel in CSCC patients.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12885-025-14585-z.

Introduction

Cutaneous squamous cell carcinoma (CSCC) is a prevalent form of cancer that accounts for approximately 20% of all skin tumors [1]. Originating from keratinocytes, it typically arises on sun-exposed skin and progresses aggressively [2]. While early-stage CSCC can often be managed surgically, metastatic CSCC poses a significant challenge due to limited treatment options and poor long-term survival rates [3]. Therefore, there is an urgent need to identify prognostic markers for early-stage CSCC and elucidate the molecular mechanisms underlying its progression to explore novel therapeutic approaches for patients across all stages of the disease. Major risk factors for CSCC include UV radiation exposure, aging, and fair skin [4]. Additionally, immunosuppression, autoimmune diseases, and chronic lymphocytic leukemia are known to increase the risk [4].

Recent advancements in high-throughput omics technologies have greatly enhanced our understanding of CSCC. For example, genome-wide association studies have identified several genetic variants linked to CSCC, affecting pathways related to skin pigmentation, immunity, cellular growth, and apoptosis [5]. Moreover, RNA sequencing studies have revealed significant transcriptomic differences between CSCC and healthy skin samples, highlighting alterations in genes related to cell cycle regulation, epidermal function, inflammation, and cell death pathways and providing valuable insights into the molecular mechanisms of this disease [1]. Whole-exome sequencing, also known as bulk sequencing, has been pivotal in revealing genetic alterations in CSCC [6]. However, bulk sequencing can detect only average allele frequencies without clear reference to specific cells or zygosity, so it is unable to study clonal evolution and tumor heterogeneity. Furthermore, the variant allele frequency (VAF), which represents the proportion of the mutant allele among all alleles analyzed, often reaches approximately 50%, posing challenges in assessing the normal (nonleukemia) component in a sample.

Single-cell DNA sequencing (scDNA-seq) has emerged as a powerful tool for dissecting intratumoral heterogeneity and understanding clonal evolution in cancer. Unlike traditional bulk sequencing methods, which provide an averaged view of genomic alterations across a tumor population, scDNA-seq allows for the examination of individual cells, enabling the identification of rare subclones and the reconstruction of evolutionary trajectories within tumors [7]. Some studies have investigated the utility of Tapestri scDNA-seq in exploring clonal evolution and the diversity of tumors in patients with myeloproliferative disorders [8, 9]. While scDNA-seq has been widely applied in various cancer types, its utility in cutaneous squamous cell carcinoma (CSCC) has rarely been reported. In addition, scDNA-seq has several inherent limitations. Unlike bulk sequencing methods, single-cell sequencing often suffers from higher levels of technical noise and amplification biases, leading to incomplete coverage and increased false positives [10, 11]. Moreover, the economic feasibility of single-cell sequencing hinges on the necessity of targeting specific genomic regions (gene panels) for sequencing, which might be insufficient for certain research questions that require unbiased study of the cancer exome/genome [12, 13]. As a result, the design of targeted panels becomes paramount in maximizing the cost-effectiveness and accuracy of single-cell sequencing assays, ensuring the comprehensive capture of relevant genetic information while minimizing sequencing costs and technical artifacts [14, 15].

In this study, we conducted a thorough investigation into the genomic landscape of CSCC via multi-patient-targeted (MPT) scDNA-seq, which integrates bulk exome sequencing with Tapestri scDNA-seq. By pooling somatic mutations that were identified in the bulk sequencing analysis, we designed an optimal gene panel for scDNA-seq. Our research explored the clonal relationships, pinpointed crucial driver mutations, and elucidated the evolutionary dynamics of CSCC across multiple patients. This detailed analysis not only enhanced our understanding of the genomic structure of CSCC but also holded promise for advancing personalized and precisely targeted therapeutic strategies.

Materials and methods

Patient cohort

The current investigation included twelve individuals who were diagnosed with CSCC, the fundamental details of which were outlined in Table 1. These participants exhibited normal kidney and liver functionality and did not take any medicine prior to sampling to prevent metabolic interference. The CSCC samples along with their corresponding noncancerous tissue samples (located > 2 cm away from the tumor periphery) were collected from 12 patients, promptly frozen in liquid nitrogen, transferred into Eppendorf tubes, and subsequently preserved at −80 °C until further analysis. The age range of the patients was between 56 and 97 years, with 5 males and 7 females. Approval for the study was obtained from the ethics committee of Jiangxi Cancer Hospital, and the subjects provided appropriate informed consent. Bulk exome sequencing was performed on the tumor and normal tissues from all 12 patients. The single cell sequencing was performed on Patient 2, Patient 8, Patient 9, Patient 12.

Table 1.

Summary of human samples in the present study

Subject	Age	Gender	TNM stage	Tumor size (cm)	Progression-free Survival (months)	Survival status
Patient 1	92	Female	T3N0M0	4*4*2.5	124	Live
Patient 2	87	Female	T2N0M0	4*3	45	Live
Patient 3	65	Female	Unknown	2*2	39	Live
Patient 4	91	Female	Unknown	Unknown	19	Dead
Patient 5	97	Male	T1N0M0	Unknown	38	Dead
Patient 6	81	Male	Unknown	3*3.5	20	Live
Patient 7	80	Male	T1N0M0	4.5*5.5	36	Live
Patient 8	56	Female	T1N0M0	4.5*3.0	29	Live
Patient 9	81	Male	T2N0M0	2*2	30	Live
Patient 10	75	Female	T2N0M0	2*1	Unknown	Unknown
Patient 11	96	Female	T2NxMx	Unknown	1	Live
Patient 12	58	Male	T3N2M1	5*3.5	31	Live

Enrichment of cell Lysis and nuclear suspensions

The frozen tumor tissues were sectioned via a surgical blade and then lysed in NST solution (146 mM NaCl, 10 mM Tris base at pH 7.8, 1 mM CaCl₂, 0.05% BSA, 0.2% Nonidet P-40, and 21 mM MgCl₂). The cell nuclear was stained with DAPI. The acquired cell nuclear were filtered and transferred into a 1.5 ml EP tube. Then the cell nuclear was enriched via a Cytomics FC500 cytometer (Beckman Coulter, FL, USA), with DAPI used as a label.

Bulk exome sequencing

The genomic DNA was then fragmented to an average size of 200–300 bp via a Covaris S220 focused-ultrasonicator (Covaris, USA). Following fragmentation, the DNA was subjected to end-repair, A-tailing, and adapter ligation via the KAPA HyperPrep Kit (Kapa Biosystems, USA). Adapter-ligated fragments were purified with AMPure XP beads (Beckman Coulter, USA), amplified via PCR, and subsequently purified. Exome capture was conducted via the SureSelect Human All Exon V7 Kit (Agilent Technologies, USA), where the hybridization of the library with the exome capture probes was carried out overnight. The captured DNA was then enriched via PCR and purified. The exome libraries were quantified, assessed for fragment size distribution, and pooled in equimolar amounts for sequencing. Sequencing was performed on an Illumina NovaSeq 6000 (Illumina, USA) via paired-end 150 bp reads. The bulk exome sequencing was performed to pool the tumor mutations from all 12 patients and design a targeted panel for scDNA-seq.

Data processing and analysis

Raw sequencing reads from bulk exome sequencing of tumor and matched normal tissues were processed using the GATK Best Practices workflow. Initially, the reads underwent quality trimming and adapter removal via Trimmomatic, followed by alignment to the reference human genome (GRCh38) via BWA-MEM. Post-alignment processing included duplicate removal, base quality score recalibration, and variant calling via the GATK toolkit. The identified variants were annotated with ANNOVAR and filtered to retain high-confidence exonic variants. Further analysis of these variants was conducted to assess their potential functional impact via SnpEff or VEP. Comparative analysis of the variant data against matched normal samples enabled the identification of somatic mutations, with key variants validated through Sanger sequencing or targeted re-sequencing.

Single-cell DNA sequencing and mutation detection via mission bio

The tumor mutations identified by bulk sequencing of tumor tissues from all 12 patients were sorted by allele frequencies ranging from 0.1 to 100%. On the basis of the mutation frequencies and previous literature, we selected 66 genes as a tumor-targeted panel for single-cell DNA sequencing. The 66 genes in the MPT panel were selected using a multi-faceted approach: (1) ​​Mutation frequency​​: Genes with high mutation rates in our bulk sequencing data (e.g., NOTCH1, TP53, TTN) were prioritized. (2) ​​Literature validation​​: Genes with established roles in CSCC pathogenesis (e.g., TP53, HRAS, and CDKN2A) were included based on prior studies in cutaneous and head/neck squamous cell carcinomas [16–18]. (3) Functional impact​​: Mutations predicted as deleterious by computational tools (CADD > 20, PolyPhen2 > 0.8) were prioritized. (4) ​​Novelty​​: Rare mutations (e.g., NLRP5, HMMR) identified in our cohort but not widely reported in CSCC were included to explore their potential roles in clonal evolution. For example, HMMR was selected for its pancancer association with poor prognosis [19–21]. The Mission Bio Tapestri^® platform was employed for targeted single-cell DNA sequencing using the following workflow: (1) ​​Single-cell isolation​​: Nuclei from frozen tumors were enriched via fluorescence-activated cell sorting (FACS) and loaded onto a microfluidic cartridge for encapsulation. (2) ​​Multiplex PCR amplification​​: A custom MPT panel targeting 66 genes amplified mutation sites identified in bulk sequencing, with unique molecular identifiers (UMIs) incorporated to mitigate amplification bias. (3) ​​Library preparation and sequencing​​: Amplified DNA was indexed, purified, and sequenced on an Illumina HiSeq4000 (150 bp paired-end reads). (4) ​​Data analysis​​: Mission Bio’s Tapestri Insights^® software integrated read depth normalization, UMI correction, and hierarchical clustering to resolve clonal substructure. This workflow enabled high-resolution detection of somatic mutations at single-cell resolution, validated against bulk sequencing data.

Estimation of mutation impact

To assess the functional impact of detected mutations through single-cell DNA sequencing, a comprehensive approach combining multiple tools was employed. Initially, the Combined Annotation Dependent Depletion (CADD) scoring system [22] which integrates over 60 genomic features to generate a comprehensive mutation score, was used. To further validate the functional impact of the mutations, POLYPHEN2 [23] and SIFT [24] tools were utilized. Specifically, mutations were initially screened by selecting those with POLYPHEN2 scores greater than 0.8 or (1-SIFT) scores greater than 0.8. The CADD scores were subsequently used to rank the initially screened mutations, and the top 30 most deleterious genes in each sample were selected on the basis of the highest CADD scores, with CADD > 10 indicating a mutation in the top 10%, CADD > 20 indicating the top 1%, and CADD > 30 indicating the top 0.1%.

Construction of a comprehensive phylogenetic tree

To uncover the evolutionary relationships between cell populations from single-cell mutation data, a neighbor-joining (NJ) phylogenetic tree was constructed via the Ape package [25]. Initially, single nucleotide variant (SNV) information was extracted from the sequencing data of scDNA-seq, and the NJ tree was constructed on the basis of these variant data. Next, cell cluster annotations defined by uniform manifold approximation and projection (UMAP) and hierarchical clustering methods were mapped onto the generated NJ tree. This step aimed to clarify the distribution of each cell cluster within the phylogenetic tree. Finally, mutations within each sample’s identified clusters were annotated, and harmful genes showing significant copy number variations (CNVs) in each cluster were further analyzed.

Copy number Estimation from mission bio data

In this study, we conducted single-cell CNV analysis using the Python-based Mosaic toolkit. The workflow included stringent quality control with the sample.cnv.filter_amplicons() function to retain high-quality amplicon data detected in ≥ 50% of cells and with read depth > 0. Subsequently, dual normalization was performed using the sample.cnv.normalize_reads() function, based on total read depth per cell and median read depth of amplicons to correct for biases. Absolute copy numbers were calculated with the sample.cnv.compute_ploidy() function, referencing wild-type diploid cells (WT). Finally, clone-specific CNV variation patterns were visualized using the sample.cnv.heatmap() function.

Results

MPT sequencing

A significant obstacle in targeted single-cell DNA sequencing of tumor mutations is the limited profiling of somatic mutations per patient via a small unbiased panel, resulting in high costs and hindering parallel cell profiling. To address this, we introduced a method called MPT sequencing (Fig. 1). Initially, bulk deep-exome sequencing was conducted on a cohort of patients to identify somatic mutations (Table 1). A custom targeted panel was subsequently developed to cover all identified mutation sites across the patient cohort. This panel was synthesized commercially and utilized for single-cell DNA sequencing of each tumor, with a focus on mutation sites of interest. This approach significantly reduced sequencing costs by avoiding noninformative genome profiling, enabling high-throughput analysis of thousands of single cells per patient at a low cost.

Fig. 1.

Schematic representation of the Multi-Patient-Targeted (MPT) sequencing workflow. A Initial bulk exome sequencing conducted on a group of human tumors. B Collective mutations from all patients combined to create a customized MPT targeted panel and scDNA-seq. C Utilization of the MPT panel for single-cell DNA sequencing on individual patients, enabling parallel profiling of thousands of cells via a microfluidics platform

Clinicopathological characteristics of patients

A diverse cohort of 12 patients diagnosed with CSCC participated in this extensive investigation. The cohort exhibited a spectrum of demographic attributes, including age, sex, and ethnic diversity. A detailed demographic breakdown was presented in Table 1, illustrating the heterogeneous nature of the patient population. The average age of all patients was 80 years. Histopathological examination revealed characteristic features inherent to CSCC. The participants included 3 patients with T1 stage disease, 4 patients with T2 stage disease, 2 patients with T3 stage disease and 3 patients with uncertain T stage disease. Patients had better overall survival, with an average progression-free survival of approximately 37 months.

Bulk exome sequencing of CSCC tumors

We collected frozen tumor samples from 4 untreated CSCC patients for bulk exome sequencing and MPT profiling via a microfluidics system (Tapestri, Mission Bio). To enhance tumor cell populations, we isolated single-nucleus suspensions and sorted them via fluorescence-activated cell sorting (FACS) within a ploidy range of 2.65–3.7 (N). Aneuploid tumor cell fractions were enriched via this method. Subsequently, sequencing libraries were prepared from aneuploid cells, and both whole-genome sequencing (at low depth for copy number estimation) and exome capture (using Roche, Nimblegen V2) were conducted to detect point mutations. Exome sequencing of the aneuploid-sorted nuclei, along with matched normal tissues, revealed a variety of somatic mutations in the tumors (Supplemental file 1). The VAF of each variant in every sequenced sample was provided in Supplemental file 2. The mutation frequencies of the top 44 genes, along with the tumor mutation burden (TMB) for each patient, are shown in Fig. 2A. On the basis of the integrated analysis of all mutations, we found that the mutations were mainly missense mutations, nonsense mutations, and splice site mutations. Among these, missense mutations occurred with the highest frequency (Fig. 2B). Further comparison of the three mutation types, single nucleotide polymorphisms (SNPs), insertions, and deletions, revealed that SNPs were the most frequent mutation type (Fig. 2C), with C > T mutations being the predominant mutation subtype (Fig. 2D). The tumor mutational burden (TMB) values exhibited significant heterogeneity among patients, ranging from 15 to 1106 (Fig. 2E). The top 10 mutated genes included NOTCH1 (50%), TP53 (50%), NOTCH2 (50%), TTN (42%), MUC16 (42%), MUC4 (42%), RYR2 (42%), PRUNE2 (42%), DMD (42%), and CDKN2A (42%) (Fig. 2G). These findings underscored the complex mutational landscape and heterogeneity present within the patient cohort.

Fig. 2.

Mutation landscape and tumor mutational burden (TMB) analysis of CSCC tumors. A Mutation frequencies of top 44 genes and TMB values for each patient. B Distribution of mutation types, with missense mutations being the most prevalent. C Comparison of variant types, including single nucleotide polymorphisms (SNPs), deletion mutations and insert mutations. D Predominant mutation subtype, highlighting C > T mutations. E Heterogeneity in TMB values among patients. G Top 10 mutated genes in CSCC tumors

Next, we conducted a comparative analysis of mutations with respect to tumor clinical stage, identifying 20 genes mutated exclusively in T1-stage patients (n = 3) (Fig. 3A). Further analysis of patient sex revealed that two mutated genes, OBSCN and USH2A, were correlated with female sex, and five mutated genes, OTOGL, ANK3, FAM205A, LOXHD1, and PCDH15, were correlated with male sex (Fig. 3B). The values of TMB and mutant-allele tumor heterogeneity (MATH) across different sexes were shown in Fig. 4, suggesting that there were no significant differences between males and females. These results provide valuable references for exploring certain gene mutations that are associated with CSCC clinical progression and patient sex.

Fig. 3.

Comparative analysis of mutations with tumor clinical staging and gender in CSCC patients. A Identification of genes mutated exclusively in T1 stage patients. B Occurrence frequency of mutations in T1 stage patients. C and D Mutated genes correlated with males and females. OR represented Odds Ratio while the P-value meaned the significant differences of gene expression between females and males

Fig. 4.

Comparison of tumor mutational burden (TMB) and mutant-allele tumor heterogeneity (MATH) across different genders

Mutational substructure of CSCC tumors

After bulk exome sequencing for patients with CSCC, we selected 4 patients to investigate their tumor clonal substructure via scDNA-seq analysis. First, we combined all of the somatic mutations identified across the 4 CSCC patients to design a custom MPT panel (Mission Bio), which included 66 genes (Supplemental file 3). A total of 29,128 cells (ranging from 2,789 to 10,079 cells per tumor) were sequenced via the Tapestri microdroplet scDNA-seq platform on the basis of the above MPT panel. The resulting dataset revealed 330 targeted sites with an average coverage depth of 164x across the five tumors. Somatic mutations were identified in individual cells through a filtering process using matched germline bulk exome sequencing data. This approach provides comprehensive insights into the clonal architecture of CSCC tumors, leveraging the MPT panel and scDNA-seq technology.

For Patient 2 with T2-stage middle skin squamous cell carcinoma, an in-depth analysis of 5,590 single cells spanning 66 genes revealed two predominant mutated genes, namely, NLRP5 and GAS2L2 (Fig. 5A). Copy number assessments derived from read depth analysis revealed chromosomal gains and losses, with PCLO, NOTCH1, and KIAA0100 exhibiting amplifications, and WNK2, CNM1, and STK11 demonstrating copy number losses (Fig. 5B and C). Subclones are defined from scDNA-seq data by clustering cells with shared mutation profiles (single nucleotide variants, SNVs and CNVs). Evolutionary relationships are inferred via phylogenetic trees, distinguishing truncal (shared) and subclonal (branch-specific) mutations. The temporal sequence of single nucleotide variants (SNVs) involved in tumor evolution was elucidated through a constructed clonal evolutionary tree, and WT cells were branched into two clones, NLRP5 p.T891M and GAS2L2 p.R441* (Fig. 5D). Notably, Patient 2 had a PFS of 45 months, which was the longest among all four patients.

Fig. 5.

Clonal substructure and phylogenetic reconstruction of Patient 2 with a T2 middle CSCC. The MPT panel covering 66 genes was utilized for scDNA-seq analysis of CSCC tumors. A Heatmaps of hierarchical clustered mutations in 5490 CSCC cell nuclei. The x-axis represents tumor mutations, and the y-axis represents tumor subclones. B and C Copy number variations (CNV) of tumor mutations in targeted panel. The x-axis represents mutated genes, and the y-axis represents tumor subclones. D Clonal evolution of CSCC tumor

For Patient 8, who had T1-stage low skin squamous cell carcinoma, an extensive analysis of 6,984 single cells across 66 genes revealed predominant mutations in PCLO, NLRP5, and HMMR (Fig. 6A). Copy number estimates revealed chromosomal gains and losses, with DNAH5 and EGFR-AS1/EGFR exhibiting amplification (Fig. 6B and C). A clonal evolutionary tree delineated NLRP5 (p.M424I)/PCLO (p.D4152V)/HMMR (p.L403F and p.L403S) as the major evolutionary mutations (Fig. 6D).

Fig. 6.

Clonal substructure and phylogenetic reconstruction of Patient 8 with a T1 low CSCC. A Heatmaps of hierarchical clustered mutations in 6984 CSCC cell nuclei. The x-axis represents tumor mutations, and the y-axis represents tumor subclones. B and C Copy number variations (CNV) of tumor mutations in targeted panel. The x-axis represents mutated genes, and the y-axis represents tumor subclones. D Clonal evolution of CSCC tumor

For Patient 9, who had T2-stage middle skin squamous cell carcinoma, an analysis of 10,079 single cells across 66 genes revealed predominant mutations in TP53 (Fig. 7A). Copy number estimates revealed chromosomal gains, with ZDHHC11 exhibiting amplification (Fig. 7B and C). The clonal evolution of the tumor followed a linear trajectory, evolving as WT-TP53 (p.P239L)-TP53 (p.C137S) (Fig. 7D).

Fig. 7.

Clonal substructure and phylogenetic reconstruction of Patient 9 with a T2 middle CSCC. A Heatmaps of hierarchical clustered mutations in 10,079 CSCC cell nuclei. The x-axis represents tumor mutations, and the y-axis represents tumor subclones. B and C Copy number variations (CNV) of tumor mutations in targeted panel. The x-axis represents mutated genes, and the y-axis represents tumor subclones. D Clonal evolution of CSCC tumor

For Patient 12, who had T3-stage high-skin squamous cancer, an analysis of 2,789 single cells across 66 genes revealed predominant mutations in TP53, TTN, and HRAS (Fig. 8A). Copy number estimates revealed chromosomal gains, with IGFN, OBSCN, RYR2, ZDHHC11, and HRAS exhibiting amplification (Fig. 8B and C). The clonal evolutionary tree of Patient 12 exhibited the greatest degree of clonal evolution among all patients, evolving from WT-HRAS (p.G13D) to TP53 (p.R303*) and subsequently to TTN (p.V6842I) (Fig. 8D). Furthermore, we also found that Patient 12 presented with the largest tumor size and highest tumor stage, indicating that tumor growth and progression might arise from the evolution of mutation clones within the tumor.

Fig. 8.

Clonal substructure and phylogenetic reconstruction of Patient 12 with a T3 high CSCC. A Heatmaps of hierarchical clustered mutations in 2789 CSCC cell nuclei. The x-axis represents tumor mutations, and the y-axis represents tumor subclones. B and C Copy number variations (CNV) of tumor mutations in targeted panel. The x-axis represents mutated genes, and the y-axis represents tumor subclones. D Clonal evolution of CSCC tumor

Discussion

In this study, we employed a novel MPT sequencing approach to elucidate the mutation landscape and clonal evolution lineage in Chinese CSCC patients. Initially, bulk deep-exome sequencing was employed to delineate the tumor gene landscape in 12 patients, followed by the construction of an MPT panel for single-cell DNA sequencing. Mutation signature analysis revealed a predominance of missense mutations, followed by nonsense and splice site mutations. Among the mutation types, SNPs were the most common, with C > T mutations being the predominant subtype. Comparative analysis with tumor clinical staging revealed that 20 genes were exclusively mutated in T1-stage patients. Gender-based analysis revealed specific gene mutations associated with males (OBSCN and USH2A) and females (OTOG, ANK3, FAM205A, LOXHD1, and PCDH15). Notably, there were no significant differences in TMB or MATH values between sexes. Previous studies have investigated the tumor mutation landscape of CSCC in various populations including Korean, British, American, and other Caucasian cohorts [6, 17, 26–28]. Compared with these studies, our identified top mutations presented similar frequencies, including those of NOTCH1, TP53, NOTCH2, TTN, MUC16, RYR2, PRUNE2, DMD, HRAS, and CDKN2A. For example, the frequency of TP53 mutation in Chinese patients is 50%, which is 79% in Korean patients and 66.3%, 70% and 94.9% in Caucasians [6, 17, 26–28]. TP53 is recognized as a key tumor suppressor gene crucial for maintaining genomic stability by regulating cellular processes such as the cell cycle, DNA repair, and apoptosis. Mutations in TP53 are commonly associated with various cancers, including CSCC [17, 29]. NOTCH1 and NOTCH2, encoding single-pass transmembrane receptors for multiple EGF-like repeats, are considered important drivers in CSCC [30, 31]. Mutations in these two genes usually lead to impaired cell differentiation, increased stem cell activity, and disease progression in CSCC [30, 31]. The mutation frequencies of NOTCH1 and NOTCH2 are comparable across Chinese (50%), Korean (68%, 58%), and Caucasian (ranging from 36.1 to 75%) populations [6, 17, 26–28].

HRAS showed similar mutation frequencies in Chinese and Korean patients (33% and 32%), which was higher than that of Caucasians (ranging from 9.6 to 22.5%). HRAS is a tumor oncogene. Gain-of-function mutation or amplification of this gene promotes CSCC occurrence and development [30]. Interestingly, HRASs are all missense mutations in Chinese patients, whereas in Korean patients, deletion mutations are predominant (82%). The mutation frequencies of TTN and MUC16 in Chinese patients (42%) were significantly lower than those in Korean patients (79%) and Caucasians (ranging from 70 to 90% and 39–87%, respectively) [6, 17, 26–28]. The function of the TTN in the CSCC has been relatively rarely studied. However, TTN/TP53 co-mutation is considered a significant driver in colon adenocarcinoma and lung squamous cell carcinoma [32]. MUC16, a transmembrane member of the mucin family, is involved in epithelial cell barrier formation and pathogen defense [33]. Notably, MUC4 exhibited a high frequency distribution (42%) in our study, in contrast with the lack of significant mutations in Korean, American, and British patients [6, 17, 26–28]. However, in a study of Australian patients, MUC4 exhibited a high frequency of mutation in patients with head and neck squamous cell carcinoma (HNSC) [34]. MUC4 was also found to be highly expressed in the tumor tissues of cutaneous condyloma acuminatum and squamous cell carcinoma and modulated by tissue inflammation and adjacent cutaneous malignancy [35]. These findings suggest a potential pivotal role for MUC4 in the tumorigenesis of CSCC.

Although bulk sequencing provides extensive information about the tumor mutation landscape, particularly regarding driver genes, it has several limitations. For example, bulk sequencing homogenizes DNA content from thousands to millions of cells, limiting its ability to identify potentially rare and genetically heterogeneous neoplastic cell populations, including subclones that may significantly impact tumor evolution and the development of therapeutic resistance [36–38]. Additionally, bulk sequencing often fails to detect rare alleles and cannot definitively determine whether mutations co-occur within the same cell. Single-cell sequencing has the potential to address these key issues. This technique offers unprecedented resolution by enabling the analysis of individual cells within a heterogeneous population, providing insights into intratumoral diversity and evolutionary trajectories [39, 40]. In addition, scDNA-seq has been employed to investigate the mechanisms of therapy resistance, tumor metastasis and relapse [38, 41, 42]. In the present study, by using a microdroplet-based scDNA-seq platform (Tapestri), we conducted an extensive analysis of thousands of individual cells from 4 CSCC patients. By focusing on regions exhibiting somatic mutations, as identified beforehand through bulk exome sequencing, we successfully characterized 66 mutations across thousands of individual cells via MPT. Investigation of the mutational substructure of CSCC tumors via scDNA-seq elucidated the clonal architecture and evolutionary trajectories of individual tumors. For Patient 2 with a T2-stage CSCC, a constructed clonal evolutionary tree was established, in which WT cells were branched into two clones, NLRP5 p.T891M and GAS2L2 p.R441*. For Patient 8 (T1-stage), WT cells were branched into three clones, including NLRP5 (p.M424I), PCLO (p.D4152V) and HMMR (p.L403F and p.L403S). For Patient 9 (T2 stage), the tumor’s clonal evolution followed a linear trajectory, evolving as WT-TP53 (p.P239L)-TP53 (p.C137S). For Patient 12 (T3 stage), the clonal evolutionary tree depicted a linear trajectory, evolving from WT-HRAS (p.G13D) to TP53 (p.R303*) and subsequently to TTN (p.V6842I). scDNA-seq not only identified high-frequency mutations such as GAS2L2, PCLO, HRAS, TTN, and TP53, which were previously detected via bulk sequencing, but also revealed low-frequency tumor mutations, such as those in NLRP5 and HMMR. NLRP5 gene mutations were observed in 2 tumors. NLRP5 belongs to the Nod-like receptor (NLR) family and is implicated in inflammation and innate immunity. Although its role in skin cancer is less understood, emerging research indicates its potential involvement in malignancies [43, 44]. The functional role of the HMMR in tumors is currently unknown. However, in a pancancer study, HMMR was found to be highly expressed in various cancer types, including HNSC, and was negatively correlated with patient prognosis [45]. These findings suggest that mutations in HMMR, particularly gain-of-function mutations, may play a role in the development of CSCC. Furthermore, we observed significant differences in tumor stage, size, and survival status among patients with different clonal mutations and evolutionary structures. Patient 2 exhibited a progression-free survival (PFS) of 45 months, which was significantly greater than that of the other three patients (approximately 30 months), suggesting a potential correlation between shallow-depth simple mutation evolution and better prognosis. In contrast, Patient 12, with a WT-HRAS (p.G13D)-TP53 (p.R303*)-TTN (p.V6842I) profile, presented with the largest tumor size and highest tumor stage, indicating a potential correlation between clonal evolution depth and tumor development.

However, it is important to acknowledge several limitations in our study. First, the sample size was limited to 4 CSCC patients as the primary focus was on developing the MPT approach. Therefore, generalizing our biological conclusions regarding tumor evolution to a larger population of CSCC patients may not be appropriate. Moreover, the custom amplicon panel utilized in our study targeted a restricted number of mutation sites (N = 66). Future advancements in microdroplet technology may enable the expansion of amplicons to thousands of sites, increasing genomic coverage.

Conclusions

In conclusion, our study introduced a MPT sequencing approach to analyze the mutation landscape and clonal evolution of CSCC tumors. Through bulk exome sequencing and scDNA-seq, we identified diverse somatic mutations, highlighting the prevalence of missense mutations in key genes such as NLRP5, TP53, and TTN. Comparative analysis revealed associations between specific gene mutations and disease characteristics. Using single-cell sequencing, we elucidated the mutational substructure and evolutionary trajectories of CSCC tumors, revealing distinct patterns associated with tumor stage and size. Our findings provide insights into tumor heterogeneity and evolution, laying the groundwork for future research aimed at personalized treatment strategies and improved patient outcomes in CSCC and other cancers.

Abbreviations

scDNA-seq

Cutaneous squamous cell carcinoma

scDNA-seq

Single-cell DNA sequencing

MPT

Multi-Patient-Targeted

VAF

Variant allele frequency

CADD

Combined Annotation Dependent Depletion

CNV

Copy number variation

FACS

Fluorescence-activated cell sorting

TMB

Tumor mutation burden

SNPs

Single nucleotide polymorphisms

MATH

Mutant-allele tumor heterogeneity

SNVs

Single nucleotide variants

HNSC

Head and neck squamous cell carcinoma

PFS

Progression-free survival

Authors’ contributions

Wen Chen, Jiawei Xu, and Chengdong Yu contributed to study conceptualization and project administration. Yong Ai, Wenbing Rao and Qingchuan Wang contributed to data curation and formal analysis. Wen Chen, Jiawei Xu, Chengdong Yu and Jun Rao conceived and designed the experiments; Siyi Xu and Lei Tang contributed to validation and visualization; Wen Chen, Jiawei Xu, and Jun Rao contributed to the writing of the original draft. Si Gong, Meng Zhou and Chengdong Yu contributed to the investigation and methodology. All the authors read and approved the final manuscript.

Funding

This work was supported by the “Five-level Progressive” talent cultivation project of Jiangxi Cancer Hospital & Institute (WCDJ2024JQ02), Jiangxi Key Laboratory of Oncology (2024SSY06041) and the Top Program (20224BAB206080) from Natural Science Foundation of Jiangxi Province, Jiangxi Provincial Administration of Traditional Chinese Medicine Projects (2024B0278), the General Science and Technology Program (202210056 and 202211009) from Jiangxi Provincial Health Commission, and the Jiangxi Cancer Hospital “Excellent Young Talent Program” (2021EYS05).

Data availability

The datasets generated during the current study have been deposited in the GenBase [46] in the National Genomics Data Center [47], China National Center for Bioinformation/Beijing Institute of Genomics, Chinese Academy of Sciences (GSA-Human: HRA008813 and HRA008816) that are publicly accessible at https://ngdc.cncb.ac.cn/gsa-human.

Declarations

Ethics approval and consent to participate

The study was performed in accordance with the principles of the Declaration of Helsinki. The study was approved by the Ethics Committee of Jiangxi Cancer Hospital (Grant Number: 2022ky306). All the subjects provided appropriate informed consent.

Consent for publication

The informed consent for publication was signed by each subject.

Competing interests

The authors declare no competing interests.