Introduction

Gastric cancer (GC) represents the fifth most common tumor and the fourth leading cause of cancer-related deaths worldwide [1]. According to World Health Organization statistics, the global incidence of GC is increasing continuously, from 1.09 million in 2020 to 1.77 million by 2040 [2]. As early GC is restricted to the mucosa and submucosa, the 5-year survival rate is over 90%. However, the prognosis is poor for advanced GC, with an average survival of only 12 months [3].

The diagnosis of GC is often made at an advanced stage due to the absence of early distinguishable symptoms and the need for a practical mass screening approach for the general population. Although serological tests, including pepsinogen I, pepsinogen II, pepsinogen ratio, gastrin-17, helicobacter pylori antibody, and carbohydrate antigen72-4 (CA72-4) [4], are less invasive, their sensitivity and specificity are limited. The Japanese GC Association has concluded that serum biomarkers are not helpful for early GC diagnosis but can be used to detect recurrence and distant metastases and to predict patient survival and postoperative recurrence [5]. Currently, the mainstay to confirm GC is endoscopy and tissue biopsy, both of which are invasive operations and dependent on the operator's skill. Thus, they are impractical for a mass screening program [6, 7]. Therefore, there is an urgent need for a less invasive, more sensitive, specific, and highly cost-effective test to improve the clinical utility for diagnosis, prognostic assessment, monitoring changes, and guiding treatment options.

During the past decade, liquid biopsy has become a valuable tool in cancer detection by analyzing tumor-derived entities circulating in body fluids, determining the tissue of origin, monitoring prognosis, and assessing response and resistance to the treatment [8, 9]. These biomarkers include cell-free DNA (cfDNA), cell-free RNA, proteins, autoantibodies, circulating tumor cells, circulating tumor DNA (ctDNA), and cancer-derived extracellular vesicles [10]. Among them, ctDNA is the cornerstone of liquid biopsy in cancer applications due to its intimate relationship with tumors and has become a popular research topic in recent years [11, 12]. In this review, following a brief overview of the biology and detection technologies, we summarized the clinical applications of ctDNA, focusing on its potential in the diagnosis, prognosis, and therapy of GC (Fig. 1).

Fig. 1
figure 1

Clinical applications of liquid biopsy in gastric cancer. Liquid biopsy, including circulating tumor DNA (ctDNA), circulating tumor RNA (ctRNA), extracellular vesicle, and circulating tumor cell (CTC), has gained popularity as a valuable tool in clinical applications of gastric cancer

Full size image

Circulating tumor DNA

The biological basis of ctDNA

cfDNA, identified by Mandel and Métais in 1948 [13], refers to extracellular DNA found in blood or body fluids, which can be either single-stranded or double-stranded [14]. In healthy individuals, cfDNA is primarily derived from apoptotic or necrotic cells or secreted from lymphocytes and other nucleated cells, which form small homogeneous DNA fragments less than 180 bp in length and 3.6–5.0 ng/mL in concentration. cfDNA has an estimated half-life between 16 min and 2.5 h, depending on factors such as the type and stage of the tumor [15].

In 1977, Leon et al. reported increased cfDNA derived from tumors [16]. After that, Stroun et al. demonstrated that cfDNA contained tumor-related mutations [17]. Therefore, cfDNA derived from tumors is described as ctDNA produced by lysed tumor cells or micrometastatic sites [18]. As a matter of principle, ctDNA contains the same genetic features as the tumor cells, such as single nucleotide mutations and methylation changes [19]. This distinguishes ctDNA from cfDNA and guides the development of cancer detection technologies. Since then, many studies have investigated the potential clinical utility of ctDNA analysis for various cancers [11, 12, 20, 21]. Researchers have gradually realized that the development of ctDNA research holds promise for advances in oncology diagnosis and prognosis prediction.

Advantages and disadvantages of ctDNA testing

Tissue biopsy is currently considered the gold standard for diagnosing and treating cancers. It enables tumor classification, aggressiveness and progression assessment, and genetic composition and mutational phenotype analysis, thereby facilitating personalized treatment strategies [22]. However, ctDNA detection has several advantages over tissue biopsy. Firstly, tissue biopsy is invasive, expensive, and risks complications such as bleeding, local infection, and damage to adjacent tissues [23]. Sometimes, tissue biopsy is not feasible due to anatomic location or underlying coagulation dysfunction. There may also be an increased chance of false negative results due to the limited retrieval of the tumor tissue [24]. In contrast, ctDNA testing requires only a minimum of invasiveness to acquire cancer-related information, regardless of the location of the tumor. Secondly, tissue biopsy only provides information at a specific site and time point. At the same time, blood can be conveniently drawn for ctDNA testing at any time throughout the disease, thus allowing for real-time monitoring of tumor changes without the need for multiple invasive tissue biopsies or imaging surveillance. The short half-life of ctDNA makes it convincing for dynamic monitoring of disease progression [25]. Finally, analysis of ctDNA provides a comprehensive molecular profile of a patient's malignancy, thereby overcoming the challenges posed by intra-tumor heterogeneity and providing additional supportive information in the diagnosis and treatment selection [26].

Despite its potential, ctDNA has several drawbacks that impede its use. Firstly, ctDNA is generally present in low abundance in early-stage cancer and represents only a tiny fraction of total cfDNA (ranging from less than 0.1% to more than 10%), which is further diluted by DNA from non-tumor sources. Currently, detecting tumor-specific mutations on cfDNA is the only way of identifying the ctDNA [19]. Secondly, the proportion of ctDNA in total cfDNA depends on tumor load, cancer stage, cell renewal, and therapy response. It is estimated that patients with a tumor load of 100 g (about 3 × 1010 tumor cells) release 3.3% of their tumor DNA into circulation each day [27]. Hence, ctDNA is frequently undetectable in patients with a low tumor burden or at early stages. Thirdly, ctDNA fragments have a half-life of less than 2 h, requiring rapid processing and stringent pre-analytical procedures such as blood collection, transport, processing, and storage temperatures [28]. Fourthly, there is no consensus on standard experimental procedures for ctDNA assays, including sampling, storage conditions, cfDNA isolation and concentration, data analysis, and interpretation [29], leading to a lack of comparability between studies [30]. Finally, most current clinical studies are retrospective and small in sample size, highlighting an urgent need for multicenter, long-term prospective clinical trials to validate the feasibility of ctDNA in cancer detection, monitoring, and treatment [31].

ctDNA detection methods

Changes in ctDNA in plasma can be detected by quantitative and qualitative (Fig. 2). The former refers to total ctDNA concentration, while the latter refers to DNA aberrations such as single nucleotide mutations and methylation changes [32].

Fig. 2
figure 2

Detection methods of ctDNA in gastric cancer. Quantitative and qualitative changes of ctDNA in plasma provide valuable information for cancer. Quantitative change refers to the total ctDNA level, while qualitative changes include ctDNA mutations and methylation changes

Full size image

The qualitative analysis of ctDNA can be categorized into two types: targeted and non-targeted [33]. The former is restricted to the detection of single or several biomarkers, focusing on known genetic alterations in primary tumors, such as KRAS (Kirsten rat sarcoma viral oncogene), BRAF (v-Raf murine sarcoma viral oncogene homolog B1), and EGFR (epidermal growth factor receptor) [26]. On the other hand, the non-targeted analysis aims to screen the genome and identifies novel genomic abnormalities, usually through whole genome sequencing (WGS) testing. However, sensitive testing of large target regions is costly, so achieving an appropriate balance between target region size and test sensitivity is essential.

Initially, detecting specific mutations in ctDNA relied on standard quantitative reverse transcription polymerase chain reaction (qPCR). However, due to its limited sensitivity, qPCR was performed mainly in advanced patients with high ctDNA levels [34]. In cases of lower tumor load, where the percentage of ctDNA is significantly lower than 0.1%, digital PCR (dPCR) and droplet-based digital PCR (ddPCR) methods overcome these limitations. For example, Pearson et al. developed a screening tool based on recombinant fibroblast growth factor receptor 2 (FGFR2) ctDNA using ddPCR [35]. In addition, further high-resolution PCR-based methods that have been successfully applied to ctDNA analysis include the BEAMing (beads, emulsion, amplification, and magnetics) [36], ARMS-PCR (Amplification Refractory Mutation System PCR) [37], and COLD-PCR (co-amplification at lower denaturation temperature- PCR) at lower denaturation temperatures [38]. PCR-based technology is faster, less expensive, and highly sensitive, allowing for the detection of tumor-associated mutations at frequencies as low as 0.01% [26]. However, its main drawback is that a single test can detect only one or a few mutations, limiting its ability to study significant numbers and different kinds of genomic alterations [39]. In 2018, Cohen et al. developed a PCR-based test, CancerSEEK, and investigated its utility for the early detection of eight common cancers. The results showed that it could be used to assess cancer-specific characteristics in the early stages (I-III) of more than 82% of cancers [40].

Compared to PCR-based methods, next-generation sequencing (NGS)-based technology is characterized by high throughput, high sensitivity, and extensive coverage. It can identify somatic and germline mutations, copy number alterations, and other chromosomal rearrangements, including translocation, conversion, and inversion. Unlike targeted analysis, NGS does not require prior knowledge of the exact genetic changes in tumors, making it a non-targeted approach. Currently, targeted deep sequencing methods include TAM-Seq (tagged-amplicon deep sequencing) [41], Safe-SeqS (Safe-Sequencing) [42], and CAPP-Seq (Computer Aided Process Planning sequencing) [43]. These technologies allow NGS to provide personalized cancer genetic profiles and facilitate personalized medicine [19]. Based on this, Kato et al. demonstrated the feasibility of NGS for ctDNA evaluation in patients with gastroesophageal adenocarcinoma [44].

Whole exome sequencing (WES) and WGS can detect tumor mutations in all patients, making them ideal for genome-wide copy number analysis and detection of significant structural variants. However, their high cost renders them unsuitable for sensitively detecting single nucleotide variants [26]. Despite lower analytical sensitivity for ctDNA analysis throughout the disease course, WES and WGS can track clonal genomic evolution associated with tumor progression [45]. Li et al. developed fingerprinting profiles based on WES for ctDNA in individual patients. This study demonstrated that ctDNA fingerprinting improves the specificity of several tumor types for monitoring treatment response and sensitivity [46].

Although tumor-associated gene mutations have been the focus of biomarker research for a long time, their wide diversity has always been a challenge for develo** validated biomarkers. To achieve sufficient sensitivity, a significant proportion of genomes must be examined [47]. In contrast, epigenetic alterations appear more stable and homogeneous in cancer, making them a promising alternative for biomarker development [48]. DNA methylation is the most widely studied epigenetic modification [49, 50]. There are two main types of methods for detecting ctDNA methylation, namely bisulfite-based conversion methods and non-bisulfite–based conversion methods. The latter includes restriction enzyme-based methods such as methylation-sensitive restriction enzymes (MSREs) [51], enrichment/immune-precipitation-based techniques such as methylated DNA immunoprecipitation sequencing (MeDIP-seq) [52], and 5-hydroxymethylation profiling. Many methylation detection methods based on bisulfite conversion have been developed, such as whole genome bisulfite sequencing (WGBS), reduced-representation bisulfite sequencing (RRBS), methylated CpG tandems amplification and sequencing (MCTA-seq), and methylation arrays [53].

Clinical applications

Diagnosis and screening

In the last few years, we have witnessed a growing body of clinical evidence supporting the detection of cfDNA for screening and monitoring patients with GC (Table 1). This test would be detected four years earlier than the current “gold standard” [54]. Plasma cfDNA levels in cancer patients, including GC patients, are two to three times higher than in healthy individuals [55]. However, plasma cfDNA levels may also increase in response to infection, inflammation, and other stressful conditions [56]. Therefore, quantifying plasma cfDNA would not be a sufficient biomarker to detect cancer due to its lack of specificity.

Table 1 Ongoing studies using ctDNA for detecting gastric cancer
Full size table

Information on tumor-associated genetic variants can be detected in ctDNA, ranging from simple point mutations to complex structural variants and even chromosomal copy number variants [57]. Therefore, detecting tumor-associated mutations in ctDNA can provide more identification of GC and guide its detection. Bettegowda et al. [58] first caught ctDNA containing tumor-specific single nucleotide variants in the plasma of 15 GC patients. Following this, Fang et al. [59] analyzed eight genetic alterations and found that Tumor Protein 53 (TP53), AT-Rich Interaction Domain 1A (ARID1A), and phosphatidylinositol-3-kinase catalytic subunit α (PI3KCA) were the most frequently mutated in ctDNA of patients with advanced GC. The detection rate of ctDNA was found to correlate with the tumor stage. Tumor-specific TP53 mutations were detected in patients with stage III-IV GC but not in patients with stage II GC [60]. It has been demonstrated that the copy number of Human epidermal growth factor receptor-2 (HER2) in the plasma of GC patients is significantly higher than that of healthy controls [61]. However, Kinugasa et al. [62] found low concordance between HER2 levels in tumor tissue and plasma DNA. This discordance may be caused by intra-tumor heterogeneity or sampling error due to low ctDNA levels [26]. To investigate whether ctDNA can cover tumor heterogeneity, Gao et al. [63] performed paired sequencing of tumor tissue biopsies and plasma samples from five patients. The biopsies confirmed the presence of tumor heterogeneity, but ctDNA only partially covered this heterogeneity. These analyses suggest that ctDNA research may be superior to tissue biopsy when examining GC with extensive intra-tumor heterogeneity.

In addition to single nucleotide variants, many studies have evaluated ctDNA methylation as a potential biomarker for cancer detection. It has been suggested that epigenetic alterations often precede somatic mutations and are more common than previously thought [64]. Circulating cfDNA methylation is highly predictive for GC, compared to methylation biomarkers in tissues [65]. Hypermethylation of p16 and E-calmodulin gene promoter regions has been detected in serum DNA samples from GC patients but not in healthy volunteers [66]. However, the reported ratio of ctDNA p16 promoter methylation in GC varies significantly across different studies [67], indicating the need for further validation. Ras association domain family 1, form A (RASSF1A), and protocadherin 10 (PCDH10) are tumor suppressor genes. Hypermethylation of RASS1A and PCDH10 was detectable in plasma samples from GC patients [68]. The study by Bernal et al. [69] confirmed the high frequency of methylation of seven genes in GC plasma, including Adenomatous Polyposis Coli (APC), SH2 domain-containing protein tyrosine phosphatase 1 (SHP1), E-calmodulin, Estrogen receptor (ER), Reprimo, Semaphorin-3B (SEMA3B) and 3-O-sulfotransferase-2 (3OST2). Additionally, methylation of tissue factor pathway inhibitor 2 (TFPI2) [70], XIAP associated factor 1 (XAF1) [71], Reprimo-like (RPRML) [72], multiple tumor suppressor 1 (MTS) and Cadherin 1 (CDH1) promoter region [10] dedicator of cytokinesis 10 (DOCK10), calcineurin binding protein 1 (CABIN1) and KQT-like subfamily, member 5 (KCNQ5) [73] can all be used as potential non-invasive diagnostic indicators in GC. In a meta-analysis of 16 studies, Gao et al. [26]. Postoperative tumor-informed ctDNA detection in EGC is feasible and allows for enhanced patient risk stratification and prognostication during curative-intent therapy [94]. GC patients with high ctDNA mutation abundance exhibited shorter overall survival (OS) than those with low mutation abundance [95]. Reduced ctDNA mutation frequency after treatment was associated with improved PFS and OS [96]. Patients with peritoneal metastases have more ctDNA mutated genes than non-peritoneal metastases. Mutations in cell division cycle 27 (CDC27) are associated with a higher risk of peritoneal metastases and a lower survival rate [97]. Patients with Mesenchymal-epithelial transition (MET) amplification in ctDNA have shorter OS than those without MET amplification, which indicates that ctDNA can predict disease progression in patients with advanced GC [98]. In some patients with Epstein-Barr virus (EBV)-associated GC, circulating EBV DNA is reduced after surgery and increases before clinically detectable recurrence. This could help monitor tumor load in patients with EBV-associated GC and predict recurrence [99]. HER2 alterations in ctDNA were significantly associated with poor OS [44]. Patients who tested positive for HER2 ctDNA before treatment had significantly shorter survival than those with negative. Still, no difference in survival was found when comparing the survival of patients regardless of tissue HER2 status [62]. This may be due to tumor heterogeneity, but ctDNA testing may provide a more accurate assessment. Based on a special NGS panel, the number of ctDNA mutations before the start of first-line chemotherapy has prognostic value. Moreover, residual ctDNA after three cycles of systemic treatment is associated with an inferior survival [100]. Changes in genomic features of ctDNA could be biomarkers for predicting the response of platinum-based first-line chemotherapy in patients with advanced GC [28].

Analysis of ctDNA has been shown to provide information on mutations that are not found in tissue biopsies due to intra-tumor heterogeneity, which can help stratify patients for testing targeted drugs and may also help identify new therapeutic targets. Moreover, most current clinical studies on ctDNA are retrospective, with small sample sizes. All these must be explored in more multicenter and long-term prospective clinical trials [26].

Conclusions

GC remains one of the most common malignancies worldwide with a poor prognosis, primarily due to the lack of population-appropriate screening, early detection methods, and suitable treatment options. The application of ctDNA as a biomarker is an exciting and emerging area for disease screening and monitoring in GC. Moreover, combining ctDNA with other biomarkers is expected to enhance cancer management for GC patients in the near future.