Esophageal carcinoma ranks as the seventh most prevalent malignancy and the sixth leading cause of cancer-related mortality worldwide, accounting for over 540,000 annual deaths globally [1]. The two major pathological subtypes of esophageal cancer are esophageal squamous cell carcinoma (ESCC), and esophageal adenocarcinoma (EAC). They exhibit marked differences in epidemiology, molecular profiles, and oncogenic pathways [2]. ESCC is the predominant subtype in Asian populations, whereas EAC is more common in Western countries [3, 4]. ESCC is considered to be more analogous to other types of squamous cell carcinoma, whereas EAC exhibits notable similarities to gastric cancer [5]. Despite these differences, both subtypes follow a progressive carcinogenic sequence from precursor lesions to invasive carcinoma. The progression from normal epithelium to precursor lesions, and ultimately to invasive cancer, provides an invaluable model for studying the full trajectory of esophageal tumorigenesis. Uncovering the molecular mechanisms driving the transition from pre-invasive to invasive disease is critical for improving early detection and therapeutic strategies [6]. Over the past decade, significant progress has been made in understanding the mechanisms underlying the progression of Barrett’s esophagus (BE), a metaplastic precursor lesion to EAC. This has resulted in significant advances in surveillance protocols of EAC [7, 8, 9, 10, 11]. However, in contrast to EAC, research on esophageal squamous dysplasia and its progression to ESCC remains limited, leaving significant gaps in our understanding of this pathway.

Although several studies have compared the genomic sequencing data of ESCC and adjacent precancerous lesions, many critical questions remain unanswered, particularly regarding molecular differences between the two histopathological stages. Liu Xi et al. [12] reported that dysplastic lesions adjacent to tumors exhibit similar alterations to ESCC in terms of both mutations and copy number alterations (CNAs). Chen Xixi et al. [13] found that “two-hit” events on TP53 are rare in dysplasia samples from tumor-free patients, in contrast to dysplastic lesions that are adjacent to tumors. Chang Jiang et al. [14] inferred that TP53 inactivation was one of the earliest steps in the malignant transformation of esophageal epithelium.

Most of the above studies focused on morphological precancerous lesions that were adjacent to tumors, and relied primarily on data from whole-exome sequencing (WES) for their analysis. However, it is important to note that due to field cancerization, dysplastic lesions adjacent to tumors are more likely to represent the final stages of cancer evolution. Consequently, it is not known to what extent these lesions accurately reflect earlier time points in the natural history of neoplastic transformation [7]. For instance, genetic alterations observed in precancerous lesions adjacent to tumors may differ from those found in precancerous lesions in patients who have not progressed to cancer [15]. Furthermore, many insights into neoplastic transformation can only be obtained through whole-genome analysis, and would not be seen with targeted exome sequencing [16]. Currently, there are few reports on the state of structural variants (SVs) in precursors to ESCC.

To address these gaps, we conducted whole-genome analysis on precancerous samples from patients without concurrent cancer, as well as on ESCC samples. By analyzing the genomic landscape of esophageal premalignant lesions, we detected molecular similarities and differences between dysplasia and cancer. Apart from finding some well-known CNAs known to act as drivers, our findings also revealed that SVs occur frequently in esophageal squamous dysplasia, indicating they emerge early in the evolution of ESCC. Moreover, our data showed that the level of ploidy alterations could distinguish esophageal squamous dysplasia from ESCC, and were also associated with the local recurrence of dysplasia. Based on these findings, we propose that elevated levels of aneuploidy occur relatively late in the progression of ESCC and exacerbate genomic instability, thereby providing cells with a competitive advantage for malignant transformation.

The current study comprehensively analyzed the whole genomic landscape of esophageal squamous precancerous lesions and ESCC. Field effects, which encompass tumor-host interactions and gene-environment interactions [44], can remodel the neighboring microenvironment of tumor cells. Tumorous dysplasia exhibits a more malignant molecular phenotype compared to non-tumorous dysplasia, with a similar pattern and burden of genomic alteration to that of tumor tissue [12]. Such tumor-adjacent precancerous lesions may already have acquired some transformative potential [13]. Once the cancerous tissue is removed and suppression of the dominant ESCC clone is lifted, the lesion may rapidly progress into carcinoma. However, among precursor cases from patients with no sign of ESCC, 33% progressed or recurred, while others regressed or remained static [45]. To reduce the effects of field cancerization and to conduct risk stratification, we analyzed biopsies of squamous dysplasia obtained from patients without concurrent tumor, as well as performing long-term follow-up of these patients.

Our analysis revealed that precursor lesions without tumors harbor frequent CNAs and SVs, which differs from some previous reports on esophageal carcinogenesis. Our use of whole-genome sequencing generates a more comprehensive genomic profile compared to targeted or exome sequencing approaches, potentially explaining the discrepancy with earlier studies. The inherent heterogeneity of dysplasia may also contribute to the differences observed between studies. Despite the comparable burden of CNAs and SVs between precancerous lesions and ESCC, the overall genomic alteration burden was generally higher in ESCC. This result suggests that chromosomal instability, a hallmark of cancer, progressively accumulates during carcinogenesis. Similar to our finding, a large burden of canonical CNAs were also detected in the hyperplasia/metaplasia stages of the LSCC progression model, with their frequency also increasing during the progression of lesions [46].

Our results showed a high mutation frequency of TP53 in precancerous lesions. Chang Jiang et al. [14] reported that the frequency of TP53 biallelic inactivation increases dramatically in early precancerous lesion stage and TP53 inactivation leads to CNAs in TP53 biallelic loss clones. These results suggested that TP53 may be an early initiator of carcinogenesis. Additionally, we found the MUC5B gene was mutated in 7.7% of precancerous lesions and 6.7% of ESCC. This gene is implicated in cell-to-cell and cell-to-matrix interactions. Interestingly, mutations in MUC5B have been associated with increased TMB in ESCC, HNSCC, LSCC, and pan-cancer datasets, suggesting they may represent a novel early event in carcinogenesis and a biomarker of genomic instability. Moreover, our analysis revealed that samples with MUC5B mutations were associated with immunologically “hot” tumors, indicating a potentially heightened responsiveness to immunotherapy. Functional enrichment analysis of differentially expressed genes indicates that the MUC5B-mutant group is more prone to invasive metastasis and platinum-based chemoresistance and exhibits increased genomic instability owing to attenuated chromatin assembly. Due to the deficiency in homologous recombination function in the MUC5B-mutant group, these patients may derive therapeutic benefit from polyadenosine-diphosphate-ribose polymerase (PARP) inhibitors. Meanwhile, DNA-repair defects in the mismatch repair pathway within the MUC5B-mutant cohort are also biomarkers for guiding the use of immune-checkpoint inhibitors [47, 48]. Concurrently, individuals harboring MUC5B mutations may be more sensitive to novel treatment agents that target genomic instability, including KIF18A inhibitors, p53-reactivating agents and PLK4 inhibitors, all of which are presently being tested in clinical trials. Furthermore, the APOBEC mutagenesis signature, a known contributor to genomic instability, was detected exclusively in ESCC but was absent in precancerous lesions. This observation indicates the APOBEC mutagenesis pattern may represent a relatively late event in the carcinogenic process, potentially conferring invasive capabilities to neoplastic cells.

The recent advances in genomic research have focused increasing attention towards the study of genomic alterations in morphologically normal cells. Li Ruoyan et al. [49] reported that normal esophageal tissues already harbor CNAs. These are predominantly characterized by whole-chromosomal amplifications of chromosomes 3, 5, and 7. The present study also detected amplification of 7q in both dysplasia and ESCC, suggesting that gain of 7q may represent an early genomic event and play a role in the accumulation of genomic instability during the progression to ESCC.

SVs are well-known hallmarks of cancer, affecting a larger fraction of the genome than point mutations by increasing or deleting gene copy numbers. These frequent genomic rearrangements can result in the amplification of oncogenes or the disruption of tumor suppressor genes, thus contributing to tumorigenesis. In our study, CASP8, a recently identified driver gene, exhibited frequent deletions in premalignant lesions, indicating that SVs of this gene may be early driver events in the development of ESCC. In HNSCC, the CASP8 inactivating mutations have been reported to confer cells the ability of resistance to death receptor [50]. However, recent studies have reported that the function of CASP8 in cancer represents a complex double-edged sword, and its specific role in ESCC remains to be further elucidated [51, 52]. While most genome-scale rearrangements are thought to be neutral, some give rise to oncogenic gene fusions [32]. Gene fusions are potential targets for personalized therapeutics, and may also serve as potential prognostic and diagnostic biomarkers [53]. The current study presented the gene fusion landscape of esophageal squamous precursors and carcinoma. However, further studies are needed to determine whether these gene fusions are drivers or passengers.

Our study found that a high level of ploidy alterations can effectively distinguish between the two histologically different cohorts, suggesting that it may be an adaptive mechanism enabling cell survival under selective pressure. Aneuploidy is closely associated with increased genomic instability [54], thus aligning with the observation that ESCC is characterized by high levels of genomic instability [36]. Similarly, the transition from carcinoma in situ to LSCC is associated with a high level of chromosomal instability, which was identified as an early marker in this progression [55, 56]. Previous research comparing the genomic landscapes of early-stage and late-stage ovarian high-grade serous carcinoma (HGSC) also observed a significant increase in overall ploidy in late-stage tumors compared to their early-stage counterparts [57]. Moreover, genomic instability and aneuploidy may facilitate metastatic progression by inducing epithelial-to-mesenchymal transition (EMT), thereby promoting tumor invasion [58]. High levels of aneuploidy and genomic instability also correlate with the expression of immune evasion markers [59] and with poor prognosis [58, 60]. Our results suggest that a high level of ploidy confers cells with the ability to outcompete the original diploid dysplastic cells. This may result in phenotypic changes, such as the ability to infiltrate the basement membrane, which can be recognized as a distinct pathological change by histological examination.

Long-term follow up in our study revealed that patients with precancerous lesions characterized by elevated ploidy alterations displayed a greater risk of recurrence following ESD treatment. This result provides further evidence that intraepithelial neoplasia characterized by a high level of ploidy alterations has an increased risk for malignancy. Furthermore, the identification of individuals with high-risk dysplasia may facilitate more targeted and effective endoscopic surveillance and monitoring strategies. In summary, ploidy alterations appear to be closely associated with malignant transformation and with local recurrence of esophageal dysplasia.

Our comprehensive molecular profiling revealed distinct temporal phases in ESCC tumorigenesis. It also identified critical early-stage alterations that initiate malignant transformation, followed by subsequent transitional genomic events that are associated with tumor progression and phenotypic evolution. However, the findings of our study are subject to several limitations. First, due to the very small tissue volume of dysplasia samples, direct validations using these specimens were not feasible in the present study. And due to the limited samples number, we may not be able to capture a broader spectrum of genomic variations. We therefore plan to accrue a prospective clinical cohort of premalignant samples to replicate our findings. Second, although the current study includes long-term clinical follow-up validation and confirmation through genomic and transcriptomic analyses in independent cohorts, it is limited by the absence of molecular experiments. The functional effects of MUC5B mutations and CASP8 deletions still require mechanistic molecular experiments in further studies. Third, our next-generation sequencing analysis was restricted to the genomic and transcriptomic level, without the incorporation of epigenetic data. To overcome these constraints, subsequent investigations should employ multi-omics strategies that combine whole-genome, transcriptome, epigenome, and proteome analyses across a well-curated sample cohort.

In summary, this study comprehensively characterized the genomic alterations in ESCC and precancerous lesions. The cancer hallmark of genomic instability was prevalent even in the precancerous stages, and increased further during the transition from precursor to invasive cancer (Fig. 6). Mutations in TP53 and MUC5B, as well as the deletion of CASP8, may represent early driver events in carcinogenesis that precede emergence of the APOBEC mutagenesis pattern. In addition, ploidy alterations confer a selective advantage to genomically unstable cells, promoting their progression toward malignant transformation and local recurrence. These findings provide new insights into the genomic evolution of ESCC and highlight potential avenues for early detection and targeted intervention in esophageal carcinogenesis.

Fig. 6.

Preliminary model of the evolution from dysplasia to ESCC. The schematic diagram of esophageal precancerous lesion and cancerous tissue in Fig. 6 was generated by ChatGpt (version 5.0; OpenAI, San Francisco, CA, USA).

The datasets used and analyzed during the current study are available from the corresponding author on reasonable request.

HA performed the advanced bioinformatics analysis and wrote the manuscript. XC helped to verify the analysis results. GW participated in the experimental design and provided the dysplasia specimens used in this research. LX provided ESCC specimens for the research and conducted pathological diagnosis of all the samples. XZ performed basic bioinformatics analysis of data. JL assisted with literature retrieval and proofread the manuscript. SC took the lead in designing the study and overseeing the experimental framework. TX played a critical role in data analysis and drafting the manuscript. All authors contributed to editorial changes in the manuscript. All authors read and approved the final manuscript. All authors have participated sufficiently in the work and agreed to be accountable for all aspects of the work.

This study was performed in accordance with the Declaration of Helsinki and was approved by the Institutional Review Board of Cancer Hospital, Chinese Academy of Medical Sciences (Ethic Approval Number: 12-71/605). A written consent was signed by the patients or their families/legal guardians.

This work was supported by the National Natural Science Foundation of China (No. 82372718), the National Key R&D Program of China (No. 2023YFC3503205), and the Chinese Academy of Medical Sciences Innovation Fund for Medical Sciences (No. 2023-I2M-2-004).

The authors declare no conflict of interest. Xiuli Zhu is an employee of Geneplus-Beijing Institute, the judgments in data interpretation and writing were not influenced by this relationship.

During the preparation of this work we used ChatGpt (version 5.0; OpenAI, San Francisco, CA, USA) to generate a schematic diagram of esophageal precancerous lesion and cancerous tissue in Fig. 6. After using this tool, we reviewed and edited the content as needed and takes full responsibility for the content of the publication.