Liquid Biopsy: An Emerging Diagnostic, Prognostic, and Predictive Tool in Gastric Cancer

Abstract

Liquid biopsy, a minimally invasive procedure that causes minimal pain and complication risks to patients, has been extensively studied for cancer diagnosis and treatment. Moreover, it facilitates comprehensive quantification and serial assessment of the whole-body tumor burden. Several biosources obtained through liquid biopsy have been studied as important biomarkers for establishing early diagnosis, monitoring minimal residual disease, and predicting the prognosis and response to treatment in patients with cancer. Although the clinical application of liquid biopsy in gastric cancer is not as robust as that in other cancers, biomarker studies using liquid biopsy are being actively conducted in patients with gastric cancer. Herein, we aimed to review the role of various biosources that can be obtained from patients with gastric cancer through liquid biopsies, such as blood, saliva, gastric juice, urine, stool, peritoneal lavage fluid, and ascites, by dividing them into cellular and acellular components. In addition, we reviewed previous studies on the diagnostic, prognostic, and predictive biomarkers for gastric cancer using liquid biopsy and discussed the limitations of liquid biopsy and the challenges to overcome these limitations in patients with gastric cancer.

INTRODUCTION

Gastric cancer (GC) is the fifth most common cancer and the fourth leading cause of cancer-related death worldwide [1]. Currently, endoscopic or surgical resection and surgical resection and/or systemic chemotherapy are the primary therapies used for patients with early gastric cancer (EGC) and advanced gastric cancer (AGC), respectively. Molecular targeted therapy and immunotherapy are gradually being used in some patients; however, the overall survival (OS) of patients, particularly those with AGC, remains poor [2]. Therefore, a reliable tool for early diagnosis, monitoring recurrence, and predicting prognosis or treatment responses for patients with GC is required.

Imaging techniques and histopathological examinations are widely used to detect tumors in clinical settings. However, clinically detectable tumors have a burden of 1–5×10⁹ tumor cells and imaging techniques can detect only a certain tumor volume. Histopathological examinations yield tumor cells or tissues through biopsy or surgery; however, invasive procedures can potentially cause pain and complications in patients. More importantly, these examinations involve evaluating tumors at specific time points and spatial locations and cannot represent the overall tumor status, reflecting tumor heterogeneity and dynamic tumor progression.

Recently, liquid biopsy, a novel tool used to detect tumors, has been extensively studied for cancer diagnosis and treatment and precision medicine for patients with cancer [3]. Liquid biopsy refers to the sampling and analysis of non-solid biological samples, such as blood and other non-blood body fluids, including saliva, breast milk, gastric juice, urine, stool, cerebrospinal fluid, pleural fluid, and ascitic fluid. Compared to traditional tissue biopsy, liquid biopsy is a minimally invasive and easily obtainable procedure. Furthermore, it may comprehensively quantify the whole-body tumor burden and facilitate the longitudinal monitoring of tumor burden and molecular changes through serial assessments (Fig. 1). The development of liquid biopsy for GC is relatively slower than that for other cancers, such as lung, breast, colorectal, and prostate cancers. However, liquid biopsy is possible using not only blood, but also saliva, gastric juice, urine, stool, peritoneal lavage fluid, and ascites from patients with GC, and cellular or acellular soluble biosources can be analyzed through these various liquid biopsy samples. Substantial evidence has shown that liquid biopsy has a broad application potential in several fields, including early diagnosis, minimal residual disease (MRD) detection, as well as recurrence, prognosis, and treatment guidance prediction in patients with GC (Fig. 2) [4,5,6,7].

Fig. 1. Liquid biopsy in a patient with cancer.

Fig. 2. Biosources and clinical application of liquid biopsy in gastric cancer.

cfDNA = cell-free DNA; CTC = circulating tumor cell; ctDNA = circulating tumor DNA; PBMC = peripheral blood mononuclear cell.

Several comprehensive reviews have recently been published on liquid biopsy for GC, but most have specifically focused on blood-based biomarkers or circulating tumor DNA (ctDNA) [8,9,10]. Here, we aim to review previous studies on liquid biopsy using various biosources including blood- and non-blood-based liquid biopsy samples in patients with GC. Finally, we discuss the limitations of liquid biopsy and the challenges in overcoming them for the successful clinical application in patients with GC.

BIOSOURCES OF LIQUID BIOPSY

Liquid biopsies contain numerous cellular and acellular components that can be used as novel targets for early diagnosis, recurrence monitoring, prognostic prediction, and treatment monitoring. The detection techniques and biomarker measurements depend on the biosource of the liquid biopsy.

Cellular components

Circulating tumor cells (CTCs)

CTCs are shed into peripheral circulation from primary or metastatic tumors. However, most CTCs entering the peripheral blood are eliminated by the immune system or die owing to shear forces in the bloodstream. Only a small fraction of CTCs with stem cell-like or epithelial-mesenchymal transition (EMT) properties can survive and migrate to other sites. Additionally, the half-life of CTCs is short, approximately 1–2.5 hours [11]. Therefore, detecting and counting CTCs is highly challenging because the peripheral blood has very few CTCs.

CTC detection methods consist of enrichment, detection, and analyses. CTCs are enriched using include physical property-based techniques, such as size, density, and charge and biological techniques, including positive selection using antibodies specific to tumor antigens such as epithelial cell adhesion molecules, cytokeratin, mucin 1, and epidermal growth factor receptor 1 or 2 and negative selection using antibodies against the common leukocyte antigen cluster of differentiation (CD) 45. The CellSearch system is the most used technology and has been approved by the Food and Drug Administration for CTC detection. CTC detection techniques include traditional polymerase chain reaction (PCR) and cellular protein detection methods such as immunofluorescence, immunohistochemistry, and fluorescence-assisted in situ hybridization.

Peripheral blood mononuclear cells (PBMCs)

PBMCs are heterogeneous population of lymphocytes (T cells, B cells, and natural killer cells), monocytes, and dendritic cells and ideal liquid biopsy approach to evaluate systemic immunity in patients with cancer. These cells are isolated from whole blood using Ficoll^®, a hydrophilic polysaccharide that separates blood layers, and gradient centrifugation. PBMCs are analyzed using flow cytometry to determine the expression of cell-surface and intracellular molecules and characterize distinct single-cell types or fluorescence-activated cell sorting (FACS). FACS further adds to the degree of functionality using highly specific fluorescent-conjugate labeled antibodies. Multicolor FACS provides a multiparametric analysis of various cellular subpopulations, cell-surface antigens, and transcription factors. Recently, as tumor immunity and immunotherapy have become important for patients with cancer, biomarker studies using PBMCs have been actively conducted to analyze the tumor immune microenvironment.

Acellular components

Cell-free DNA (cfDNA) and ctDNA

cfDNA is an extracellular DNA fragment released by cells. In healthy individuals, cfDNA may be of hematopoietic origin and is typically found at low levels. Although cfDNA substantially accumulates in malignant tumors, it is rapidly cleared from the liver (half-life: approximately ≤2.5 hours) [12]. Compared to short-fragment (<150 bp) cfDNAs, long-fragment (>150 bp) cfDNAs originate from necrotic tumor cells. Thus, the presence of long-fragment cfDNAs indicates the presence of tumor cells [13]. cfDNA methylation has long been investigated as a cancer biomarker [14]. Therefore, long fragments or aberrant methylation of cfDNA have emerged as front-runners in the development of cfDNA biomarkers.

ctDNA is a tumor cell-derived fragmented cfDNA with genetic and epigenetic signatures released by different tumor subclones that accounts for <1% cfDNA, and therefore a key component for developing liquid biopsy-based biomarkers in patients with cancer (half-life: <2 hours) [8,9,10,15]. A major challenge in liquid biopsy approaches is the development of methods to detect and characterize small amounts of ctDNA in large populations of cfDNA. Although serum ctDNA concentrations are higher than plasma ctDNA concentrations, serum also contains DNA released from white blood cells [16]. Therefore, plasma-based assays are more sensitive for ctDNA isolation and analysis. PCR-based techniques, including the standard quantitative PCR, digital PCR (dPCR), droplet dPCR (ddPCR), and quantitative methylation-specific PCR, are the most used method for detecting cfDNA and ctDNA. Although PCR-based techniques inexpensive and have a high sensitivity, the number of detectable mutations is limited. Thus, next-generation sequencing (NGS)-based technologies are increasingly being used for cfDNA and ctDNA detection.

Non-coding RNAs (ncRNAs)

Recently, several studies have demonstrated that RNA biomarkers have unique advantages in the field of liquid biopsy owing to their high sensitivity, tissue specificity, and low detection cost. Circulating tumor RNAs include messenger RNA (mRNA) and ncRNA fragments, such as microRNA (miRNA), long ncRNA (lncRNA), circular RNA (circRNA), and PIWI-interacting RNA (piRNA), which originate from different cells and tissues. Although ncRNAs are not translated into a protein, they act as “regulators” of several genes and proteins. ncRNAs are usually secreted into the blood or other body fluids due to cell necrosis, apoptosis, or active secretion by tumor cells [17]. MiRNAs are small, single-stranded ncRNAs comprising 21–23 nucleotides that repress gene expression by promoting the degradation of their mRNA targets or inhibiting translation. lncRNAs are transcripts consisting of >200 nucleotides and can assume several complex secondary and tertiary structures owing to their nucleotide chain length. circRNAs form covalently closed continuous loops and are more stable than linear lncRNAs. ncRNAs can be encapsulated within extracellular vesicles or exosomes to maintain their stability in the extracellular environment and protect them from nuclease activity.

Since ncRNAs and cfDNAs are nucleic acid products, they are detected similarly. Quantitative reverse transcription PCR (qRT-PCR)-, dPCR-, ddPCR-, and NGS-based RNA sequencing are the most used methods for detecting ncRNAs. qRT-PCR is the gold standard for the quantitative detection of ncRNAs with high sensitivity, reproducibility, and accuracy.

Exosomes

Exosomes are small (30–150 nm) lipid bilayer extracellular vesicles found in almost all body fluids, secreted by large multivesicular bodies, and released into the extracellular environment through fusion in response to various pathological processes [18]. They secrete various mRNAs, ncRNAs, transmembrane or encapsulated cytoplasmic proteins, and lipids from tumor or non-tumor cells into the body fluids. Owing to their lipid bilayer structure, exosomes are relatively abundant and stable in circulation, making them potential biomarkers for early diagnosis, prognosis, treatment efficacy, and drug resistance in patients with cancer.

Exosomes are difficult to isolate with high purity owing to their unique formation and transport processes. Additionally, tumor cell-extracted exosomes account for only a small portion of the total exosomes in body fluids; thus, highly sensitive and specific detection methods are required. Exosome enrichment and detection methods are based on exosome properties such as density, size, surface composition, and precipitation. Currently, ultracentrifugation, ultrafiltration, precipitation, immunoaffinity enrichment, lipid-based separation, and microbead, microfluidic chip, or thermal enrichment are used for exosome enrichment. Western blot analysis and enzyme-linked immunosorbent assay (ELISA) are traditionally used to detect exosomes. Electron microscopy, nanoparticle tracking analysis, colorimetry, electrochemical analysis, fluorescence, clustered regularly interspaced short palindromic repeats (CRISPR)/CRISPR-associated system-assisted, and single exosome detection methods have been developed as new techniques for exosome detection.

Proteins

Proteins found in the blood have a long history of use as cancer biomarkers. For example, proteins such as cancer antigen (CA) 125 and prostate-specific antigen have been used clinically as diagnostic and prognostic biomarkers of ovarian and prostate cancer, respectively. Several serum protein biomarkers, including carcinoembryonic antigen (CEA), CA19-9, CA125, and CA72-4, have been widely used to predict prognosis, recurrence, and treatment response in patients with GC. However, they have low sensitivity and specificity, and lack GC-specific properties. Recently, several protein biomarkers, including inflammatory cytokines, chemokines, and enzymes, have been reported as diagnostic or prognostic biomarkers for patients with GC in clinical samples, such as blood, saliva, gastric juice, and stool [4,5,6,7,8,9,10]. ELISA is an economical and relatively easy method for detecting protein biomarkers.

CLINICAL SIGNIFICANCE OF LIQUID BIOPSY IN GC

Several studies on liquid biopsies of various tumor types, including GC, have shown major trend changes, as noted below. The application of liquid biopsy has shifted from early diagnostic and prognostic prediction to recurrence monitoring and treatment guidance for systemic anticancer therapy. Furthermore, the detection and analysis techniques for liquid biopsies have shifted from quantification to genomic, epigenomic, and multiomic analyses.

Blood-based liquid biopsy in GC

Blood is the most important and common source for liquid biopsies. In tumor tissue specimens, tumor heterogeneity may represent a significant problem, and blood-based specimens can be used to identify a comprehensive dynamic molecular profile for each patient with GC.

Early diagnosis

Although endoscopic surveillance is considered the standard method for GC screening in several Asian countries, this approach is inadequate for screening average-risk populations because of its invasiveness, high cost, and low compliance. Therefore, non-invasive approaches, including liquid biopsy-based molecular biomarker assays, are required. Patients with EGC or resectable AGC, who are candidates for screening for the early diagnosis, have significantly low circulating tumor-related biomarker levels in the peripheral blood. Both CTCs and ctDNAs are fragile, heterogeneous, and present in extremely small amounts. Therefore, most previous studies on early detection of circulating tumor-related diagnostic biomarkers have used highly sensitive biosources, such as cfDNA, ncRNA, or exosomal ncRNAs (Table 1) [19,20,21,22,23,24,25,26,27,28,29,30,31,32,33,34,35,36,37,38,39,40,41,42,43,44,45,46,47,48,49,50,51,52,53,54,55,56,57,58,59,60,61,62,63,64,65,66,67,68,69,70,71,72,73,74,75,76,77,78].

Table 1. Blood-based diagnostic biomarkers for the early diagnosis of GC.

Biosources		Biomarkers	No. of participants (GC/control)	Sample	Methods	Sensitivity (%)	Specificity (%)	AUC	Reference
Cellular components
	CTCs	≥2/7.5 mL	116/31	Whole blood	“FAST disc” centrifugal microfluidic system (size-selective isolation)	85.3	90.3	0.928	Kang et al. (2017) [19]
		EpCAM+CD44+	26/12	Whole blood	Flow cytometry	92.3	100	0.974	Watanabe et al. (2017) [20]
Acellular components
	cfDNA	cfDNA level	73/40	Plasma	qPCR targeting the Alu repeats	NP	NP	0.744	Pu et al. (2016) [21]
		cfDNA level	124/92	Serum	Branched DNA-based Alu assay	79.0	91.8	0.940	Qian et al. (2017) [22]
		cfDNA level	428/95	Serum	TaqMan qPCR assay	NP	NP	0.931	Lan et al. (2017) [23]
		CancerSEEK	1,817 (34 GC)/812	Plasma	Multiplex PCR	70.0	99.0	0.910	Cohen et al. (2018) [24]
		EpiPanGI Dx	37	Plasma	Targeted bisulfite sequencing reads	NP	NP	0.900	Kandimalla et al. (2021) [25]
		Methylated RUNX3	50/61	Serum	Combined restriction dPCR	50.0	80.3	0.695	Hideura et al. (2020) [26]
		Methylated RPRML	25/25	Plasma	5-Azacytidine assay and direct bisulfite sequencing	56.0	88.0	0.726	Alarcón et al. (2020) [27]
		Methylated SFRP2	92/50	Plasma	Bisulfite conversion kit	60.9	86.0	0.784	Miao et al. (2020) [28]
		Methylated H19 and MALAT1	150/100	Leukocytes	Bisulfite conversion kit	69.4 (female), 65.8 (male)	92.6 (female), 75.3 (male)	0.821 (female), 0.741 (male)	Hu et al. (2021) [29]
	miRNAs	miR-16, miR-25, miR-92a, miR-451 and miR-486-5p	160/160	Plasma	qRT-PCR	72.9	89.2	0.812	Zhu et al. (2014) [30]
		miR-21	50/50	Serum, PBMCs	qRT-PCR	88.4	79.6	0.912	Wu et al. (2015) [31]
		miR-627, miR-629, miR-652	108/96	Plasma	qRT-PCR	86.7	85.5	0.941	Shin et al. (2015) [32]
		miR-23b	138/50	Plasma	qRT-PCR	71.0	74.0	0.800	Zhuang et al. (2016) [33]
		miR-376c	65/108	Plasma	qRT-PCR	71.0	78.0	0.770	Hung et al. (2017) [34]
		miR-21, miR-93, miR-106a, miR-106b	111/63	Plasma	ddPCR	84.8 (78.3*)	79.2 (70.7%*)	0.887 (0.809*)	Zhao et al. (2018) [35]
		miR-25	184/56+78	Serum	qRT-PCR	67.3–69.4	80.4–81.0	NP	Kong et al. (2019) [36]
		miR-214	168/74	Plasma	qRT-PCR	73.2	91.9	0.880	Ji et al. (2019) [37]
		miR-200c	202/250	Serum	qRT-PCR	74.0	66.0	0.750	Huang et al. (2019) [38]
		miR-425-5p, miR-24-3p, miR-1180-3p, miR-122-5p	30/90	Plasma	qRT-PCR	NP	NP	0.829*	Zhu et al. (2019) [39]
		miR-212	110/100	Serum	qRT-PCR	96.0 (94.6*)	78.7 (60.8*)	0.960 (0.951*)	Shao et al. (2020) [40]
		hsa-miR-320a, hsa-miR-1260b, hsa-miR-6515-5p	235/3,210	Serum	qRT-PCR	>93.0	>85.0	>0.950	Yao et al. (2021) [41]
		miR-4257, miR-6785-5p, miR-187-5p, miR-5739	1,417/1,417	Serum	DNA chip (3D-Gene®)	99.6	95.3	0.998	Abe et al. (2021) [42]
		miR-18a, miR-181b, miR-335	176/173	Serum	qRT-PCR	71.6 (72.0*)	87.9 (88.0*)	0.860 (0.850*)	Izumi et al. (2021) [43]
		miR-1306-3p, miR-936 and miR-6083, miR-659-3p, miR-6792-3p, miR-3185	52/30	Plasma	qRT-PCR	NP	NP	0.730	Yu et al. (2021) [44]
								0.825
		12 miRNA panel	445/4,803	Serum	qRT-PCR	87.0	68.4	0.848	So et al. (2021) [45]
	lncRNAs	CUDR, LSINCT-5, PTENP1	110/124	Serum	qRT-PCR	81.8 (77.8*)	85.2 (97.0*)	0.830 (0.830*)	Dong et al. (2015) [46]
		TINCR, CCAT2, AOC4P, BANCR, LINC00857	162/110	Plasma	qRT-PCR	81.0	86.0	0.900	Zhang et al. (2017) [47]
		ZNFX1-AS1	50/50	Plasma	qRT-PCR	84.0	68.0	0.850	Xian et al. (2018) [48]
		LINC00978	38/31	Serum	qRT-PCR	80.0	70.0	0.831	Fu et al. (2018) [49]
		CTC-501O10.1, AC100830.4, RP11-210K20.5	100/100	Plasma	qRT-PCR	NP	NP	0.724, 0.730, 0.737	Liu et al. (2018) [50]
		PANDAR, FOXD2-AS1, SMARCC2	109/106	Plasma	qRT-PCR	NP	NP	0.839	Yang et al. (2019) [51]
		CTC-497E21.4	100/84	Serum	qRT-PCR	81.8	75.0	0.848	Zong et al. (2019) [52]
		SNHG17	67/67	Plasma	qRT-PCR	NP	NP	0.748	Zhang et al. (2019) [53]
		ARHGAP27P1	53/53	Plasma	qRT-PCR	75.5	60.4	0.732	Zhang et al. (2019) [54]
		LINC00086	168/74	Plasma	qRT-PCR	72.6	83.8	0.860	Ji et al. (2019) [37]
		B3GALT5-AS1	107/87	Serum	qRT-PCR	64.5	87.4	0.816	Feng et al. (2020) [55]
		C5orf66-AS1	200/278	Serum	qRT-PCR	77.5 (81.2*)	53.36 (62.9*)	0.688 (0.789*)	Zhou et al. (2020) [56]
		HCP5	98/82	Serum	qRT-PCR	80.0	70.0	0.750	Qin et al. (2021) [57]
		lnc-MB21D1-3:5, lnc-PSCA-4:2, lnc-ABCC5-2:1	52/30	Plasma	qRT-PCR	NP	NP	0.904	Yu et al. (2021) [44]
	circRNAs	hsa_circ_0000745	60/60	Plasma	qRT-PCR	85.5	45.0	0.683	Huang et al. (2017) [58]
		hsa_circ_0000520	45/17	Plasma	qRT-PCR	82.4	84.4	0.900	Sun et al. (2018) [59]
		circPTPN22	120/104	Plasma	qRT-PCR	78.0	84.0	0.857	Ma et al. (2021) [60]
		8-circRNAs	194/94	Serum	qRT-PCR	89.0 (90.0*)	62.0 (60.0*)	0.830 (0.820*)	Roy et al. (2022) [61]
	Exosomal miRNAs	miR-19b-3p, miR-106a-5p	90/90	Serum	mirVana PARIS Kit, TEM/western blot, qRT-PCR	95.0	90.0	0.814	Wang et al. (2017) [62]
		miR-92b-3p, let-7g-5p, miR-146b-5p, miR-9-5p	86/62	Serum	ExoQuick, qRT-PCR	60.0*	84.0*	0.773*	Tang et al. (2020) [63]
		miR-1307-3p	70/60	Serum	exoRNeasy Serum/Plasma Kits, NTA/TEM/western blot, qRT-PCR	NP	NP	0.845, 0.820, 0.754, 0.732	Ge et al. (2020) [64]
		miR-590-5p	168/50	Serum	Ultracentrifugation, TEM/ western blot, qRT-PCR	63.7	86.0	0.810	Zheng et al. (2021) [65]
		miR-195-5p, miR-211-5p	88/88	Plasma	miRNeasy Serum/Plasma Kit, NTA/TEM/western blot, qRT-PCR	NP	NP	0.830	Yang et al. (2021) [66]
		miR-4741, miR-32, miR-3149, miR-6727	120/57	Plasma	ExoQuickTM, NTA/ western blot, qRT-PCR			0.855, 0.946, 0.768, 0.892	Tang et al. (2022) [67]
	Exosomal lncRNAs	lncUEGC1	51/60	Plasma	Ultracentrifugation, NTA/TEM, qRT-PCR	88.2	88.3	0.876 (0.850*)	Lin et al. (2018) [68]
		lncRNA-HOTTIP	126/120	Serum	qRT-PCR	69.8	85.0	0.827	Zhao et al. (2018) [69]
		lncRNA PCSK2-2:1	63/29	Serum	HiPureExosomekits, TEM, qRT-PCR	84.0	86.5	0.896	Cai et al. (2019) [70]
		lnc-GNAQ-6:1	43/27	Serum	Particle size analysis, TEM/western blot, qRT-PCR	83.7	55.6	0.732	Li et al. (2020) [71]
		lncRNA-GC1	522/219	Serum	NanoSight particle-tracking analysis, TEM/western blot, qRT-PCR	84.8 (88.7*)	85.0 (80.8*)	0.898 (0.861*)	Guo et al. (2020) [72]
	Exosomal circRNAs	hsa_circ_0065149	39/41	Plasma	qRT-PCR	48.7*	90.2*	0.640*	Shao et al. (2020) [73]
		hsa_circ_0015286	60/30	Plasma	ExoQuick, TEM/NTA, qRT-PCR	82.1	65.7	0.778	Zheng et al. (2022) [74]
	Exosomal piRNAs	piR-019308, piR-004918, piR-018569	70/60	Serum	exoRNeasy Serum/Plasma Kits, NTA/TEM/western blot, qRT-PCR	NP	NP	0.820, 0.754, 0.732	Ge et al. (2020) [64]
	Proteins	TFF1, TFF2, TFF3	183/280	Serum	ELISA	80.9	81.0	0.890	Aikou et al. (2011) [75]
		VEGF, ADAM8, IgG to H. pylori, serum pepsinogen I, pepsinogen II	285/238	Serum	Luminex multiplex panel	88.6	83.2	0.850	Tong et al. (2016) [76]
		CDH17, TFF3	111/44	Plasma	ELISA	66.7, 62.2	61.4, 56.8	NP	Choi et al. (2017) [77]
		TXNRD1	896/228	Plasma	Ultraviolet spectrophotometry	95.6	76.3	0.945	Zhu et al. (2022) [78]

GC = gastric cancer; AUC = area under the curve; CTC = circulating tumor cell; CD = cluster of differentiation; cfDNA = cell-free DNA; qPCR = quantitative polymerase chain reaction; dPCR = digital polymerase chain reaction; NP = not presented; miRNA = microRNA; qRT-PCR = quantitative reverse transcription polymerase chain reaction; NTA = nanoparticle tracking analysis; TEM = transmission electron microscopy; ddPCR = droplet digital polymerase chain reaction; lncRNA = long ncRNA; circRNA = circular RNA; piRNA = PIWI-interacting RNA; ELISA = enzyme-linked immunosorbent assay.

^*Only early gastric cancer

Previous studies involving screening using cfDNA have traditionally measured cfDNA levels [21,22,23]; however, cfDNAs are also released from normal hematopoietic cells, not tumor-specific, and have low accuracy. Recent studies have used tumor-specific mutations or cfDNA methylation for the early diagnosis of patients with GC [24,25,26,27,28,29]. The CancerSEEK blood test uses a unique, non-invasive, multi-analyte test that simultaneously evaluates the presence of mutations and eight cancer-associated protein biomarkers in the blood [24]. CancerSEEK was used in 1,005 patients previously diagnosed with stage I–III esophageal, gastric, colorectal, liver, pancreatic, breast, lung, and ovarian cancers. Using this strategy, they identified patients with eight common cancer types with approximately 70% and 99% sensitivity and specificity, respectively. However, the accuracy of the predictions varied by tumor type and was high for colorectal, ovarian, and pancreatic cancers (up to 80%) but markedly low for gastric, liver, and lung cancers (up to 40%). EpiPanGI Dx consists of a panel of three distinct differentially methylated regions and has a diagnostic performance, with an area under the curve (AUC) of 0.850–0.950 for most gastrointestinal cancers (0.900 for GC) in cfDNA specimens [25]. Recently, several methylated cfDNAs (RUNX3, RPRML, SFRP2, H19, and MALAT1) have shown diagnostic performance in patients with GC than that in healthy participants [26,27,28,29].

ncRNAs are the most studied blood-based diagnostic biomarkers in patients with GC because of their abundance, stability, widespread circulation, and tumor-specific expression. Shin et al. have reported that the diagnostic value of the three-miRNA signature panel (miR-627, miR-629, and miR-652) was significantly higher in patients with GC than that in healthy participants, with 86.7% sensitivity, 85.5% specificity, and 0.941 AUC [32]. In a multicenter cohort of 5,248 participants from Singapore and Korea, the diagnostic value of the 12-miRNA panel had a sensitivity of 87.0%, specificity of 68.4%, and AUC of 0.848 [45]. Several studies have reported circRNA-based diagnostic biomarkers for GC, and an eight-circRNA panel was recently reported as a potential diagnostic biomarker for early GC detection [61]. Exosomal ncRNAs have recently been actively reported as diagnostic biomarkers for GC because exosomes can protect RNA from degradation by endogenous RNases, thereby enhancing exosomal ncRNA stability in peripheral blood circulation. Exosomal lncRNAs, including lncUEGC1 and lncRNA-GC1, are highly sensitive and stable biomarkers for detecting EGC [68,72].

Recurrence monitoring

Surgical resection is the primary curative treatment for locoregional GC. Although adjuvant chemotherapies, such as S-1 or capecitabine plus oxaliplatin, have improved relapse-free survival and OS, the 5-year OS rate for localized stages is approximately 60%, and approximately 40% patients with GC experience recurrence [79]. Standard methods for monitoring recurrence after surgical resection in patients with GC include radiological imaging and serum tumor marker assessment. Routine clinical tests such as endoscopy and computed tomography (CT) cannot reliably detect post-operative MRD or micrometastases after surgical resection. Commonly used serum markers, including CEA and CA19-9, detect only approximately 40% recurrences with poor sensitivity and specificity [80]. Therefore, MRD evaluation in patients with GC after surgical resection may provide an opportunity to reduce recurrence rates in resected disease if appropriate action can be taken for MRD-positive patients.

ctDNA is a reliable biomarker for detecting MRD in patients with breast, colon, and lung cancer. Recent studies have demonstrated the feasibility of using ctDNAs to detect MRD and monitor disease recurrence in patients with GC (Table 2) [83,84,85,86]. ctDNA levels can be affected by systemic and local treatments such as systemic anticancer therapy, radiotherapy, and other concurrent inflammatory processes or trauma. Therefore, blood sampling should ideally be performed at least two weeks after surgery to detect post-operative MRD. Approaches to MRD assessment are categorized into targeted tumor-informed and untargeted tumor-agnostic assessments. Tumor-informed approaches begin with genomic sequencing of the primary tumor, whereas tumor-agnostic approaches are uninformed about mutations in the primary tumor. Tumor-informed approaches are preferable because they determine MRD with high sensitivity and specificity by identifying tumor-specific variants with a remarkably low variant allele frequency (0.01%). Several studies have evaluated the role of ctDNAs in post-operative blood samples and their association with clinical outcomes in resected GC [83,84,85,86]. The presence of post-operative ctDNA significantly correlate with cancer recurrence and poor survival, and precede the radiographic detection of recurrence. Most of these studies performed targeted tumor-informed NGS and then applied them to longitudinal ctDNA monitoring after surgical resection. However, till date, few prospective large-scale studies have been conducted and consistent data are missing. Therefore, the feasibility of its routine clinical use remains challenging.

Table 2. Blood-based biomarkers for recurrence monitoring of GC.

Biosources		Biomarkers	No. of patients with GC	Methods	Clinical significance	Reference
Cellular components
	CTCs	FR+ CTCs	132	CytoploRare Kit	Prediction of peritoneal metastasis	Zeng et al. (2022) [81]
Acellular components
	cfDNAs	Postoperative long-fragment LINE-1	99	HELP	High risk of recurrence	Ko et al. (2021) [82]
	ctDNAs	Postoperative ctDNA (+)	25	WGS (Illumina)	High risk of recurrence	Kim et al. (2019) [83]
				Personalized cancer-specific rearrangements
		Postoperative ctDNA (+)	1,630	73-gene NGS (Guardant360 test)	Inferior DFS	Maron et al. (2019) [84]
		Postoperative ctDNA (+)	46	Targeted NGS (Illumina Hiseq 3000)	High risk of recurrence	Yang et al. (2020) [85]
		Postoperative ctDNA (+)	35 (22 curative)	Targeted NGS	Poor RFS	Openshaw et al. (2020) [86]

GC = gastric cancer; CTC = circulating tumor cell; cfDNA = cell-free DNA; ctDNA = circulating tumor DNA; FR = folate receptor; NGS = next-generation sequencing; HELP = HpaII tiny fragment enrichment by ligation-mediated polymerase chain reaction; WGS = whole-genome sequencing; DFS = disease-free survival; RFS = relapse-free survival.

Prognostic prediction

Prognostic biomarkers provide information on long-term outcomes irrespective of treatment, identifying patients with more aggressive tumors. It aims to predict the prognosis and adjust the treatment intensity for optimal survival rates through individual treatment for GC. Several studies have predicted the prognosis of patients with GC using blood-based biomarkers (Table 3) [33,40,69,74,82,86,88,89,90,91,92,93,94,95,96,97,98,99,100,101].

Table 3. Blood-based prognostic biomarkers of GC.

Biosources		Biomarkers	No. of patients with GC	Methods	Clinical significance	Reference
Cellular components
	CTCs	>5 CTCs/7.5 mL	65	OBP-401	Poor OS	Ito et al. (2016) [88]
		≥2 CTCs/7.5 mL	106	CellSearch	Poor PFS and OS	Pernot et al. (2017) [89]
		>3 CTCs/7.5 mL	59	CellSearch, CTC-Biopsy system	Poor PFS and OS	Ning et al. (2021) [90]
		CD44+ CTCs	228	Immunofluorescent double staining	Poor OS	Szczepanik et al. (2019) [91]
		CSV+PD-1+ CTCs	70	Microbead selection method	Poor OS	Liu et al. (2020) [92]
		TWIST+ CTCs	31	Centrifugal microfluidic system	Poor OS	Jhi et al. (2021) [93]
	PBMCs	High CD3+CD8+ T-cells, low CD4+CD25+Foxp3+ Tregs	105	FACS	Increased OS	He et al. (2017) [94]
		TIM-3+CD8+ T-cells	47	FACS	Increased OS	Ma et al. (2023) [95]
		Early increase in PD-1+CD8+ T-cells after chemotherapy	68	FACS	Increased OS	Shin et al. (2023) [96]
Acellular components
	cfDNAs	cfDNA level	277	TaqMan qPCR	Peritoneal recurrence and poor OS	Fang et al. (2016) [97]
		Pre-surgical low methylation levels of LINE-1	99	HELP	Poor RFS and OS	Ko et al. (2021) [82]
		Methylation of RASSF1A, SOX17 and Wif-1	70	Methylation-specific PCR	Poor PFS and OS	Karamitrousis et al. (2021) [98]
	ctDNAs	ctDNA level	277	TaqMan qPCR (68 mutations in 8 genes)	Peritoneal recurrence and poor OS	Fang et al. (2016) [97]
		ctDNA level	35 (13 palliative)	Targeted NGS	Poor OS	Openshaw et al. (2020) [86]
		TP53 mutation and MET amplification	23	Multigene NGS-panel technology	Poor OS	Li et al. (2021) [99]
		SFRP2 hypermethylation	148	Methylation-specific dPCR	Poor PFS and OS	Yan et al. (2021) [100]
	miRNAs	High miR-23b level	138	qRT-PCR	Poor clinical outcome	Zhuang et al. (2016) [33]
		Low miR-212 level	110	qRT-PCR	Poor OS	Shao et al. (2020) [40]
		Low miR-148a level	132	qRT-PCR	Poor OS	Komatsu et al. (2021) [101]
	Exosomal lncRNAs	High lncRNA-HOTTIP level	126	qRT-PCR	Poor OS	Zhao et al. (2018) [69]
	Exosomal circRNAs	High hsa_circ_0015286 level	60	ExoQuick, TEM/NTA, qRT-PCR	Poor OS	Zheng et al. (2022) [74]

GC = gastric cancer; CTC = circulating tumor cell; OS = overall survival; PFS = progression-free survival; CD = cluster of differentiation; PD-1 = programmed cell death-1; PBMC = peripheral blood mononuclear cell; FACS = fluorescence-activated cell sorting; TIM-3 = T-cell immunoglobulin and mucin-domain containing-3; cfDNA = cell-free DNA; qPCR = quantitative polymerase chain reaction; HELP = HpaII tiny fragment enrichment by ligation-mediated polymerase chain reaction; RFS = relapse-free survival; NGS = next-generation sequencing; dPCR = digital polymerase chain reaction; miRNA = microRNA; qRT-PCR = quantitative reverse transcription polymerase chain reaction; lncRNA = long non-coding RNA; circRNA = circular RNA; TEM = transmission electron microscopy; NTA = nanoparticle tracking analysis.

CTCs are traditional prognostic biomarkers of liquid biopsy and have been reported for the longest time to predict prognosis in several cancer types. CTC count significantly correlates with poor OS in patients with GC. In a meta-analysis of 579 patients with GC from 7 studies, CTC detection in peripheral blood is significantly associated with poor OS and progression-free survival (PFS) [87]. However, the detection rate and cut-off value of CTCs differed in each study [88,89,90]. CTCs that express specific molecules are potential prognostic biomarkers in patients with GC [91,92,93]. CTCs expressing stem cell markers (CD44) [91] or mesenchymal markers (cell-surface vimentin or TWIST) [92,93] are significantly associated with poor OS in patients with GC.

cfDNAs or ctDNAs can reflect the tumor status in almost real time because they have a short half-life and can serve as prognostic biomarkers for patients with GC. High cfDNA or ctDNA levels are associated with a poor prognosis [86,97]. The methylation status of several genes in the cfDNA (LINE-1, RASSF1A, SOX17, and Wif-1) and ctDNA (SFRP2) is significantly associated with PFS and OS [82,98,100]. Only few studies on miRNAs or exosomal ncRNAs have investigated the correlation between ncRNAs and survival of patients with GC. High miR-23b, exosomal lncRNA-HOTTIP, and exosomal hsa_circ_0015286 levels [33,69,74] and low miR-212 and miR-148a levels [40,101] are associated with poor OS in patients with GC.

Since the tumor immune microenvironment is closely related to the prognosis of patients with cancer, studies have focused on predicting the prognosis of patients with cancer by analyzing immune cells and tumor-related components [94,95,96]. Patients with high CD3⁺CD8⁺ T cells and low CD4⁺CD25⁺Foxp3⁺ regulatory T cells after neoadjuvant chemotherapy in PBMCs had significantly increased OS in patients with GC receiving neoadjuvant chemotherapy [94]. Ma et al. [95] reported that the percentage of T-cell immunoglobulin and mucin-domain containing-3 (TIM-3)⁺CD8⁺ T cells is an independent protective factor in patients with GC. Shin et al. [96] reported that an early increase in programmed cell death-1 (PD-1)⁺CD8⁺ T cells potentially predicts favorable prognosis and durable responses in patients with GC receiving palliative platinum-based chemotherapy. Therefore, high expression of immune checkpoint receptors such as PD-1 or TIM-3 in tumor-infiltrating lymphocytes reflects T-cell exhaustion, whereas an elevated proportion of T cells expressing immune checkpoint receptors in PBMCs may reflect systemic immune activation.

Treatment guidance

The area showing the greatest change in standard treatment for GC is the development of novel anticancer drugs. In addition to conventional cytotoxic chemotherapy, molecular targeted therapy and immunotherapy have significantly improved the survival of patients with advanced or metastatic GC [2]. However, the limitations of systemic anticancer therapy include primary resistance, which does not respond from the start of treatment, and secondary resistance, which acquires resistance while maintaining tumor response. Additionally, the overall tumor status may change over time after systemic anticancer therapy is administered. Therefore, dynamic monitoring through sequential liquid biopsies is essential for predicting drug response or resistance, especially in patients with GC receiving palliative anticancer systemic therapy (Table 4) [102,103,104,105,106,107,108,109,110,111,112,113,114].

Table 4. Blood-based predictive biomarkers for the treatment guidance of GC.

Biosources		Biomarkers	No. of patients with GC	Methods	Clinical significance	Reference
Cellular components
	CTCs	≥4 CTCs at 2 and 4 weeks	52	CellSearch system (Veridex)	Determination of tumor response to S-1-based or paclitaxel regimens	Matsusaka et al. (2010) [102]
		≥2 HER2+ CTCs/6 mL	115 (56 HER2+)	Metafer-iFISH Cytelligen system	Resistance to trastuzumab	Li et al. (2018) [103]
		CTC-PD-L1+	32	CanPatrol® CTC enrichment technique	Guidance of immunotherapy	Cheng et al. (2019) [104]
		EMT progression of single CTC	24	Size selective method	Detection of metastasis and drug resistance for systemic therapy	Negish et al. (2022) [105]
				Single-cell RNA sequencing
	PBMCs	CD103+PD-1+CD8+ T cells	29	FACS	Prediction of anti-PD-1 therapy	Nose et al. (2023) [106]
Acellular components
	cfDNA	cfDNA MSI-H	16	Pan-cancer MSI detection using Guardant360 (74-gene panel)	Concordant with tissue MSI	Willis et al. (2019) [107]
					Response prediction of immunotherapy
		cfDNA level	106	SuperbDNA™ hybridization	Monitoring of the efficacy of systemic therapy	Zhong et al. (2020) [108]
	ctDNA	ctDNA level 3 months after the start of chemotherapy	30	GeneQuant RNA/DNA Calculator-Amersham Pharmacia Biotech (Biochrom) Ltd.	Correlated with DFS for chemotherapy	Normando et al. (2018) [109]
		ctDNA levels 6 weeks post-treatment	61	NGS (73-gene sequencing panel, Guardant360)	Prediction of response to anti-PD-1 and PFS	Kim et al. (2018) [110]
		Genomic alterations	55	Targeted NGS (54–73 gene panel)	Targetable alterations by an FDA-approved agent	Kato et al. (2018) [111]
		Chromosomal instability assessed by CNI score of ctDNA	55	WGS	Prediction of the tumor response of systemic therapy	Chen et al. (2019) [112]
		Higher mTBI (≥1%) in pretreatment ctDNA	39 HER2+	NGS	Poor PFS	Wang et al. (2019) [113]
		HER2 SCNA, PIK3CA, ERBB2/4, NF1 mutations	24 HER2+	Targeted NGS	Prediction of trastuzumab resistance	Wang et al. (2019) [114]

GC = gastric cancer; CTC = circulating tumor cell; HER2 = human epidermal growth factor receptor 2; PD-L1 = programmed death-ligand 1; EMT = epithelial-mesenchymal transition; PBMC = peripheral blood mononuclear cell; PD-1 = programmed cell death-1; FACS = fluorescence-activated cell sorting; cfDNA = cell-free DNA; MSI = microsatellite instability; ctDNA = circulating tumor DNA; DFS = disease-free survival; NGS = next-generation sequencing; FDA = Food and Drug Administration; CNI = copy number instability; WGS = whole-genome sequencing; mTBI = molecular tumor burden index; PFS = progression-free survival.

The number of CTCs, EMT progression of single CTCs, ctDNA levels, and chromosomal instability of ctDNA are related to tumor response or survival in patients with GC receiving systemic anticancer therapy [102,105,108,109,112]. Recent studies have reported predictive biomarkers that can predict the response to anti-PD-1 and trastuzumab, an anti-human epidermal growth factor receptor 2 (HER2) monoclonal antibody, therapy using blood-based liquid biopsy. The HER2 phenotype and chromosome 8 aneuploidies in CTCs, high molecular tumor burden index or HER2 somatic copy number alteration, PIK3CA, ERBB2/4, and NF1 mutations in ctDNA are associated with trastuzumab resistance [103,113,114]. Since anti-PD-1 therapies, such as nivolumab or pembrolizumab, increase survival in locally advanced unresectable or metastatic GC, the importance and need for biomarkers to predict tumor response to anti-PD-1 therapy is rapidly increasing. Kim et al. [110] have reported that changes in ctDNA levels 6 weeks after treatment predicts tumor response and the decreased ctDNA is associated with improved PFS in patients with GC treated with pembrolizumab as a salvage treatment. Programmed death-ligand 1 (PD-L1) expression and microsatellite instability (MSI) status in tumor tissues are well-known predictive biomarkers for immunotherapy. PD-L1 on CTCs or the MSI status of cfDNA can predict the tumor response to immunotherapy in GC [104,107]. Nose et al. [106] have reported that the tissue-resident marker CD103 present in peripheral blood PD-1⁺CD8⁺ T cells 2 weeks after nivolumab treatment in patients with GC is significantly associated with PFS, suggesting the importance of immune cell-related predictive biomarkers for immunotherapy.

Therefore, for patients with AGC who need to be treated with different anticancer drugs depending on the molecular status of their tumor, systemic anticancer therapy is expected to be increasingly selected based on liquid biopsy, which can reveal the entire tumor state at the time of drug administration, rather than the tumor tissue at the time of initial diagnosis.

Non-blood-based liquid biopsy in GC

Although most studies have focused on blood-based specimens as the standard concept for liquid biopsy, non-blood-based body fluids, such as saliva, gastric juice, urine, stool, and ascitic fluids, can be potential cancer biomarker sources (Tables 5 and 6) [34,115,116,117,118,119,120,121,122,123,124,125,126,127,128,129,130,131,132,133,134,135,136,137,138,139,140,141,142,143,144,145,146]. However, higher heterogeneity was observed in non-blood-based body fluids than in blood-based body fluids. Therefore, the specimen collection and processing methods should be standardized to increase the comparability and clinical implementation of non-blood-based body fluids as biomarkers in patients with GC.

Table 5. Non-blood-based diagnostic biomarkers of GC from saliva, gastric juice, urine, and stool.

Biosources		Biomarkers	No. of participants (GC/control)	Methods	Sensitivity (%)	Specificity (%)	AUC	Reference
Saliva
	miRNAs + mRNAs	3 mRNAs (SPINK7, PPL, and SEMA4B) and 2 miRNAs (miR140-5p and miR301a)	163/131	qRT-qPCR	75.0	83.0	0.810	Li et al. (2018) [115]
	mRNAs	mRNAs (SPINK7, PPL, SEMA4B, SMAD4)	200/200	Luminex xMAP assay + qRT-PCR	94.0	91.0	NP	Xu et al. (2020) [116]
	Proteins	CSTB, TPI1, and deleted in DMBT1	40/40	TMT technology	85.0	80.0	0.930	Xiao et al. (2016) [117]
	Protein glycosylation	SBA and VVA	87/134	Lectin microarrays	96.0	80.0	0.890	Shu et al. (2017) [118]
	Microbiota	Bacterial genera	99/194	16S rRNA gene sequencing	NP	NP	0.910	Huang et al. (2021) [119]
Gastric juice
	miRNAs	miR-129-1-3p, miR-129-2-3p	42/99	qRT-PCR	68.7	71.9	0.656	Yu et al. (2013) [120]
		miR-21, miR-106a	42/99	qRT-PCR	85.7, 73.8	97.8, 89.3	0.969, 0.871	Cui et al. (2013) [121]
		miR-133a	62/142	qRT-PCR	85.9	84.8	0.907	Shao et al. (2016) [122]
	lncRNAs	AA174084	39/92	qRT-PCR	46.0	93.0	0.848	Shao et al. (2014) [123]
		RMRP	39/92	qRT-PCR	56.4	75.4	0.699	Shao et al. (2016) [124]
		ABHD11-AS1	39/92	qRT-PCR	41.0	93.4	0.650	Yang et al. (2016) [125]
	piRNAs	piR-1245	66/66	qRT-PCR	90.9	74.2	0.885	Zhou et al. (2020) [126]
	Exosomal DNAs	BARHL2 methylation	20/10	ExoQuick-TC, Bisulfite PCR	90.0	100.0	0.923	Yamamoto et al. (2016) [127]
Urine
	DNA	8-OHdG and 8-OHG	60/70	SPE and UPLC-MS/MS	NP	NP	0.777 and 0.841	Chen et al. (2020) [128]
	miRNAs	miR-376c	65/108	qRT-PCR	60.0	64.0	0.700	Hung et al. (2017) [34]
		miR-6807-5p, miR-6856-5p, and Helicobacter pylori	153/153	qRT-PCR	76.9	88.9	0.885	Iwasaki et al. (2019) [129]
		MMP-9/NGAL and ADAM12	35/35		NP	NP	0.825	Shimura et al. (2015) [130]
	Proteins	Endothelial lipase	90/57	Western blotting	79.0	100.0	0.970	Dong et al. (2013) [131]
		TFF1, ADAM12, H. pylori (male)	144/138	Quantitative proteomics analysis	NP	NP	0.858 (male), 0.893 (female)	Shimura et al. (2020) [132]
		TFF1, uBARD1, H. pylori (female)
Stool
	Microbiome	Veillonella, Megasphaera, and Prevotella 7 genus and Streptococcus salivarius subsp. Salivarius, Bifidobacterium dentium, and Lactobacillus salivarius	134/58	GC-MS	NP	NP	0.694–0.837	Wu et al. (2020) [133]
		Desulfovibrio, Escherichia, Faecalibacterium, or Oscillospira	38/35	qPCR	NP	NP	0.900–0.920	Liu et al. (2021) [134]

GC = gastric cancer; AUC = area under the curve; miRNA = microRNA; mRNA = messenger RNA; qRT-qPCR = quantitative reverse transcription quantitative real-time polymerase chain reaction; qRT-PCR = quantitative reverse transcription polymerase chain reaction; NP = not presented; TMT = Tandem Mass Tags; lncRNA = long non-coding RNA; piRNA = PIWI-interacting RNA; PCR = polymerase chain reaction; SPE = solid phase extraction; UPLC-MS/MS = ultra performance liquid chromatography-tandem mass spectrometry; GC-MS = gas chromatography-mass spectrometry; qPCR = quantitative polymerase chain reaction.

Table 6. Non-blood-based GC biomarkers from peritoneal lavage or ascitic fluid.

Biosources		Biomarkers	No. of patients with GC	Methods	Clinical significance	Reference
Peritoneal lavage fluid
	CTCs	Positive CTCs	136	CellSearch	Poor PFS	Okabe et al. (2015) [135]
	Single cell suspension	Phenotypes of lymphocyte and macrophage	122	FACS	Peritoneal metastasis	Takahashi et al. (2022) [136]
	cfDNAs	Methylated THBS1	92	Quantitative methylation-specific PCR	Poor DFS	Hu et al. (2021) [137]
		BNIP3, CHFR, CYP1B1, MINT25, SFRP2, and RASSF2 methylation	107	Quantitative methylation-specific PCR	Peritoneal recurrence	Hiraki et al. (2011) [138]
		MINT2 methylation	92	Real-time methylation-specific PCR	Prediction of peritoneal metastasis	Han et al. (2014) [139]
					Poor DFS
	Exosomal miRNAs	High miR-21 and miR-1225-5p level	24	Ultracentrifugation, Micro BCA Protein Assay kit, qRT-PCR	Peritoneal recurrence	Tokuhisa et al. (2015) [140]
		Low miR-29s level	85	Nanosight LM10, NTA/western blot, qRT-PCR	Peritoneal recurrence	Ohzawa et al. (2020) [141]
					Poor OS
	Cytokines	IL-6	145	ELISA	Peritoneal metastasis	Yang et al. (2022) [142]
Ascites
	Single cell suspension	CD45RO+CD8+ T cells, HLA-DR+CD8+ T cells, PD-1+CD3+ T cells, FoxP3+ Tregs	55	FACS	Poor OS	Park et al. (2019) [143]
	DNA, RNA	SEs at the ELF3, KLF5, and EHF loci, TGF-β pathway activation through SMAD3 SE activation and high TEAD1 expression	98	WGS, RNA sequencing	Molecular guided therapy by two distinct molecular subtypes of peritoneal metastases	Tanaka et al. (2021) [144]
	RNA	12-gene prognostic signature	15	Single cell RNA sequencing	Prognostic signature of gastric dominant and GI-mixed type	Wang et al. (2021) [145]
	Exosomal miRNAs	Low miR-181b-5p level	92	miRCURY Exosome Isolation Kit, qRT-PCR	Diagnosis of GC-associated malignant ascites (AUC 0.846)	Yun et al. (2019) [146]

GC = gastric cancer; CTC = circulating tumor cell; PFS = progression-free survival; FACS = fluorescence-activated cell sorting; cfDNA = cell-free DNA; PCR = polymerase chain reaction; DFS = disease-free survival; miRNA = microRNA; qRT-PCR = quantitative reverse transcription polymerase chain reaction; NTA = nanoparticle tracking analysis; OS = overall survival; ELISA = enzyme-linked immunosorbent assay; WGS = whole-genome sequencing; GI = gastrointestinal; AUC = area under the curve.

Saliva

Saliva is a complex fluid containing proteins, metabolites, DNA, mRNA, ncRNAs, and microbiota that can be used as biomarkers. Using saliva as a biosource for biomarker analysis is another notable approach in patients with GC because its collection is minimally invasive and the salivary glands may be stimulated by mediators that can be released by GC. In a previous study, a salivary panel of 3 mRNAs (SPINK7, PPL, and SEMA4B) and 2 miRNAs (miR-140-5p and miR-301a) showed diagnostic performance with 75.0% sensitivity, 83.0% specificity, and 0.810 AUC in a cohort of 163 and 131 patients with and without GC, respectively [115]. Using another panel of four mRNAs (SPINK7, PPL, SEMA4B, and SMAD4), patients with GC were discriminated from healthy participants with 94.0% sensitivity and 91.0% specificity using saliva samples from 200 patients with GC and 200 healthy participants [116]. Huang et al. [119] have reported a distinct salivary microbiota as a diagnostic biomarker and demonstrated that microbiota profiles in saliva have diagnostic potential, with 0.910 AUC, in patients with GC.

However, saliva-based GC diagnosis remains unsatisfactory because of unstable quality control, oral microorganism disturbance, and salivary sample contamination with other molecules.

Gastric juice

Gastric juice is a potential specimen for GC screening and diagnosis because it exists only in the stomach and can be collected directly from the lesion area. Moreover, it can be easily obtained during diagnostic endoscopy to screen healthy participants or monitor patients for benign or precancerous lesions.

As DNA is easily degraded due to the acidity of gastric juice, most biomarkers identified in gastric juice are ncRNAs and proteins. Several miRNAs (miR-129-1-3p, miR-129-2-3p, miR-21, miR-106a, and miR-133a) have shown diagnostic performance in distinguishing GC patients from healthy participants [120,121,122]. Other ncRNAs, including lncRNAs, circRNAs, piRNAs, and exosomal RNAs, are diagnostic biomarkers for GC because of their stability in gastric juice [123,124,125,126].

Urine and stool

Urine or stool are notable biological specimens used to identify new non-invasive biomarkers for GC. Compared to blood, they can be obtained easily, in large amounts, and without invasion. Iwasaki et al. [129] found that an urinary miRNA (miR-6807-5p and miR-6856-5p) panel in combination with serum Helicobacter pylori status could distinguish patients with GC from healthy controls, with 76.9% sensitivity, 88.9% specificity, and 0.885 AUC. Recently, fecal microbiome alterations have been associated with GC and have shown good diagnostic value for distinguishing patients with GC from healthy participants [133,134].

However, the concentration of biomarkers in urine or stool has a high degree of variability and is likely to be influenced by other conditions, such as an individual’s hydration status, medications, and other comorbidities, making it difficult to determine whether these biomarkers in urine or stool are GC-specific.

Peritoneal lavage fluid and ascites

Peritoneal metastasis has an extremely poor prognosis. Therefore, accurate peritoneal metastasis diagnosis and peritoneal recurrence prediction are extremely important in patients with GC. Currently, cytological analysis of the ascitic fluid or radiological tests, such as abdominal and pelvic CT, are performed to diagnose peritoneal metastasis. However, owing to the low sensitivity of these traditional diagnostic methods, patients with GC sometimes require invasive procedures such as laparoscopy or laparotomy. Therefore, more sensitive and non-invasive diagnostic methods are required to detect peritoneal metastasis or malignant ascites in patients with GC.

Peritoneal lavage during surgical GC resection is useful for detecting free cancer cells in the peritoneal cavity. Patients with positive cytology are classified as having M1 or stage IV cancer [2]. However, cytology has low sensitivity, and several patients with GC and negative peritoneal cytology experience peritoneal recurrence after radical surgery. Previous studies have attempted to improve peritoneal recurrence prediction using RT-PCR for CEA mRNA in combination with other mRNAs. Recent studies have reported that the GC-specific cfDNA or exosomal miRNA, which are stable in the peritoneal environment, predicts peritoneal recurrence and is associated with prognosis [137,138,139,140,141,146]. Furthermore, several studies are being actively conducted to analyze malignant ascites and understand peritoneal metastasis with the development of high-quality techniques that can analyze immune cell characteristics and genomic or epigenomic changes. Park et al. [143] reported that the proportions of CD8⁺ T cells with memory markers (CD45RO) and activation markers (human leukocyte antigen [HLA]-DR), CD3⁺ T cells with PD-1, and Foxp3⁺ regulatory T cells profiled by FACS were significantly higher in malignant ascites than those in blood and were independent prognostic factors [143]. A comprehensive multiomic analysis of malignant ascites from 98 patients with GC revealed two distinct molecular subtypes: one displaying active super enhancers (SEs) at the ELF3, KLF5, and EHF loci, and the second displaying transforming growth factor-β pathway activation through SMAD3 SE activation and enhanced transcriptional enhancer factor-1 expression [144]. Single-cell transcriptome profiling of malignant ascites cells from 15 patients with GC divided them into two subtypes prognostically independent of clinical variables, and a 12-gene prognostic signature was derived [145].

CHALLENGES AND FUTURE PERSPECTIVES

Liquid biopsy is a powerful and innovative tool that can potentially change the diagnostic and treatment paradigm for patients with GC through early diagnosis before clinical detection, representative molecular analysis of the entire tumor, MRD detection after surgical resection, improved risk stratification, and systemic therapy guidance. However, practical clinical application of liquid biopsy in GC faces several challenges.

One of the major challenges in developing liquid biopsy for GC is its low sensitivity and specificity. Moreover, single omics information often cannot reflect the entire molecular characteristics, considering tumor heterogeneity. The integration of multiple omics information can partially complement each other and improve our understanding of tumor molecular characteristics. However, multiomics studies on liquid biopsy samples are challenging because of the remarkably low analyte content in body fluid samples. Therefore, various highly sensitive GC-specific detection methods must be developed along with liquid biopsies. In addition, specimen processing and analysis procedures should be standardized to ensure reproducibility and comparability of liquid biopsy results. Previous studies and review articles have focused on tumor-related components and blood samples. Therefore, more studies on tumor microenvironment-related components or non-blood-based samples are required. Although there are several notable biomarkers in liquid biopsies of patients with GC, the best strategy may be to combine different biomarkers and clinical and biochemical parameters to increase diagnostic or predictive performance. This approach may be one of the most feasible future strategies. Finally, clinical validation with large-scale clinical trials is necessary for its widespread use in clinical practice.

Although evidence supporting the clinical application of liquid biopsy in GC is not as robust as that in other tumor types, this field is advancing rapidly, with several active clinical trials incorporating liquid biopsy. Furthermore, liquid biopsy may be valuable for highly heterogeneous tumors, such as GC, by overcoming tumor heterogeneity and providing comprehensive molecular characterization. Taken together, liquid biopsy is an emerging tool that can play an important role in improving personalized medicine for patients with GC.