The retrospective study included 311 patients with primary gastric cancer undergoing surgery for the first time, all of whom had lymph node dissection at Hanoi Medical University Hospital, Vietnam, between October 2019 and June 2021. Patients with recurrent or secondary gastric cancer were not included in this study. The patients in this study were all from the Northern region of Vietnam. Retrospective data on imaging studies, such as scintigraphy or histopathological evaluation of surgical tumor specimens with or without accompanying metastatic specimens, were used to evaluate distant metastases. This study received permission from the Ethics Committee of Hanoi Medical University (decision number: 4397/QD-DHYHN), complies with the ethical standards set forth in the Declaration of Helsinki, and was revised by the World Medical Association’s General Assembly in Seoul, South Korea, in October 2008.

Data on patient age, gender, lymph node status, distant metastasis, histological type, and primary tumor location were obtained from digitized archival records. H&E-stained slides and corresponding tissue samples enclosed in paraffin blocks were also obtained to study E-Cadherin expression. Histological gastric carcinoma variants were identified based on the 2019 World Health Organization (WHO) histological classification of gastrointestinal cancer1. The tumors’ pT and pN classifications were determined using the eighth edition of the American Joint Committee on Cancer classification system1.

Tumor budding was evaluated using the International Tumor Budding Consensus Conference (ITBCC) 2016 protocol11. Tumor budding was counted in the “hotspot” field (tumor tissue areas with the most tumor buds) at 20× magnification (corresponding to an evaluation area of 0.785mm²). The “hotspot” areas are usually at the edge of the tumor. Ten microscopic fields at invasive areas were scanned at 10× magnification to identify “hotspot” areas. Two pathologists counted the tumor buds in the “hotspot” areas at 20× magnification. The number of tumor buds was calculated by counting in the "hotspot" field and then dividing by the conversion factor to determine the number of tumor buds/0.785 mm². Tumor budding was graded as low (0–4 tumor buds/0.785 mm²), moderate (5–9 tumor buds/0.785 mm²), and high (≥ 10 tumor buds/0.785 mm²). Differences in tumor budding grades between the two pathologists were discussed and re-evaluated.

After the steps of routine microscopic techniques, immunostaining was performed using a Ventana Benchmark XT automatic stainer and an anti-E-cadherin (mouse monoclonal primary antibody) antibody kit (cat. no.: 760-500/05269806001; diagnostics Roche, USA). Two pathologists, blind to the cases’ clinical data, semi-quantitatively analyzed the E-cadherin staining results for gastric adenocarcinoma tissue samples from 311 cases. The cellular membrane is the main location of E-cadherin immunoexpression, and sometimes this expression extends into the cytoplasm. In each tumor, the rate of immunostained cells was determined by counting 500 tumor cells. Scoring of E-cadherin expression is based on immunostaining intensity (0=negative; 1=low; 2=moderate; 3=strong) and the percentage of immunoreactive cells (0=none; 1=1%–9%; 2=10%–49%; 3=50%–79%; 4=80%–100%) and grading: 0=negative score; 1+ =score of 1–4; 2+ =score of 5–8; 3+ =score of 9–1216.

E-cadherin expression and tumor budding were analyzed relative to several clinical pathological parameters: histopathological type (WHO and Lauren’s classification), grade, lymph node metastasis, stage pT, age, and sex. E-cadherin expression was classified as low (0–1+) and high (2+–3+). Tumor budding was classified as low grade (0–4 tumor buds/0.785mm²) and high grade (≥5 tumor buds/0.785mm²). In E-cadherin expression, the original group was named “EcadhGOC”, consisting of 4 levels: 0, 1+, 2+, and 3+; we regrouped them into low (0–1+) and high (2+–3+) titled “phanlopE1”. Similarly, in tumor budding, the original group was named “Budding”, consisting of 3 levels: low (0–1+), medium (2+), and high (3+); we regrouped them into low (0–1+) and high (2+–3+) titled “phanlopB2”. For the purpose of comparing the predictability of lymph node status of the respective groups, the receiver-operating characteristic (ROC) curve was used in the analysis.

SPSS 20 for Windows software (SPSS, Inc., Chicago, IL, USA) was used to analyze the data. The χ² test was used to compare proportions, and Fisher's exact test was used for cases with expected frequencies < 5. The associations of various variables with a binary dependent variable were determined using multivariable logistic regression analysis. Statistical significance was determined when the difference was p < 0.05. The receiver-operating characteristic curve (ROC) and the area under the curve (AUC) were used to investigate the ability of two parameters to predict lymph node metastasis, including tumor budding and E-cadherin expression. Good predictability is indicated when the AUC is ≥ 0.8.

Table 1

Clinicopathologic features		Total n = 311 (100%)
		n (%)
Ages
	40-49	24 (7.7%)
	50-59	81 (26%)
	60-69	144 (46.3%)
	70-79
	80-89	6 (1.9%)
Sex
	Nam	203 (65.3%)
	Nữ	108 (34.7%)
Lauren's histological type
	Intestinal	211 (67.8%)
	Mixed	52 (16.7%)
	Diffuse	48 (15.4%)
WHO histological type
	Tubular	206 (66.2%)
	Poorly cohesive	50 (16.1%)
	Mucinous	3 (1%)
	Mixed	51 (16.4%)
	With lymphoid stroma	1 (0.3%)
Grade (WHO)
	High	142 (45.7%)
	Moderate	93 (29.9%)
	Poor	76 (24.4%)
Stage
	T1a	3 (1%)
	T1b	13 (4.2%)
	T2	54 (17.4%)
	T3	217 (69.8%)
	T4a	20 (6.4%)
	T4b	4 (1.3%)
Lymph nodes metastasis
	Yes	173 (55.6%)
	No	138 (44.4%)
E-Cadherin expression
	Weak	136 (43.7%)
	Strong	175 (56.3%)
Tumor budding
	Low	108 (34.7%)
	High	203 (65.3%)

Table 2

Clinicopathologic parameters		Lymph nodes metastasis
		No	Yes	p
Ages
	40-49	9	15	0.470
	50-59	41	40
	60-69	65	79
	70-79	20	36
	80-89	3	3
Sex
	Nam	81	122	0.030
	Nữ	57	51
Lauren's histological type
	Intestinal	126	85	<0.001
	Mixed	9	43
	Diffuse	3	45
WHO histological type
	Tubular	124	82	<0.001
	Poorly cohesive	3	47
	Mucinous	0	3
	Mixed	10	41
	With lymphoid stroma	1	0
Grade (WHO)
	High	100	42	<0.001
	Moderate	35	58
	Poor	3	73
Stage
	T1a	3	0	<0.001
	T1b	10	3
	T2	41	13
	T3	83	134
	T4a	1	19
	T4b	0	4
E-Cadherin expression
	Weak	10	126	<0.001
	Strong	128	47
Tumor budding
	Low	96	12	<0.001
	High	42	161

Table 3

Clinicopathologic parameters		Lymph nodes metastasis
		No	Yes	OR	95%CI	p
Lauren's histological type
	Intestinal	126	85	2.067	1.125-3.799	0.019
	Mixed	9	43
	Diffuse	3	45
Stage
	T1a	3	0	2.233	1.165-4.281	0.016
	T1b	10	3
	T2	41	13
	T3	83	134
	T4a	1	19
	T4b	0	4
E-Cadherin expression
	Weak	10	126	0.043	0.017-0.108	<0.001
	Strong	128	47
Tumor budding
	Low	96	12	14.59	5.895-36.09	<0.001
	High	42	161

OR: odds ratio, 95%CI: 95% confidence intervals

Table 4

Clinicopathologic parameters		Lymph nodes metastasis
		No	Yes	OR	95%CI	p
WHO histological type
	Tubular	124	82	1.028	0.722-1.463	0.879
	Poorly cohesive	3	47
	Mucinous	0	3
	Mixed	10	41
	With lymphoid stroma	1	0
Grade (WHO)
	High	100	42	2.032	1.081-3.820	0.028
	Moderate	35	58
	Poor	3	73
Stage
	T1a	3	0	2.329	1.204-4.504	0.012
	T1b	10	3
	T2	41	13
	T3	83	134
	T4a	1	19
	T4b	0	4
E-Cadherin expression
	Weak	10	126	0.048	0.019-0.121	<0.001
	Strong	128	47
Tumor budding
	Low	96	12	12.73	4.980-32.53	<0.001
	High	42	161

OR: odds ratio, 95%CI: 95% confidence intervals

Figure 1

Figure 2

Figure 3

Figure 4

The average age of the 311 patients in this study was 62.99 ± 8.64 years (with an age range of 42–85 years), and the male/female ratio was 1.9/1. Most cases belonged to pT3 (69.8%), followed by pT2 (17.4%), pT4 (7.7%), and pT1 (5.1%). The overall node metastasis rate was 55.6%. Incidence of node metastasis was 18.7% for pT1, 25.9% for pT2, 47.0% for pT3, and 91.7% for pT4. The number of nodes dissected in each case ranged from 0 to 28. The patient's clinical-pathological manifestations are summarized in Table 1.

Univariate analysis (Table 2) found lymph node metastasis associated with many parameters: sex (p = 0.001), Lauren’s histological type (p < 0.001), WHO histological type (p < 0.001), WHO histological grade (p < 0.001), pT stage (p < 0.001), tumor budding (p < 0.001), and E-cadherin expression (p < 0.001). However, it was not involved in age (p = 0.408). Among the 173 (55.6%) cases with node metastasis, tumor budding (Figure 1) was 3+ (high) in 79.3% (161/203) and 1+ (low) in 11.1% (12/108; p<0.001). Similarly, E-cadherin immunoexpression (Figure 2) was weak in 92.6% (126/136) and strong in 26.8% (47/175; p < 0.001).

Through multivariate logistic regression analysis, tumor budding (odds ratio [OR] = 14.59, 95% confidence interval [CI] = 5.89–36.09, p < 0.001), E-cadherin expression (OR = 0.043, 95% CI = 0.017–0.108, p < 0.001), stage (OR = 2.233, 95% CI = 1.165–4.281, p = 0.016), and Lauren type (OR = 2.067, 95% CI = 1.125–3.799, p = 0.019) were independent factors of lymph node metastasis in the Lauren histology group (Table 3).

In the WHO histological group, multivariate logistic regression analysis identified tumor budding (OR=12.73, 95% CI = 4.980–32.530, p < 0.001), E-cadherin expression (OR=0.048, 95% CI=0.019–0.121, p < 0.001), stage (OR = 2.329, 95% CI = 1.204–4.504, p = 0.012), and grade (OR = 2.032, 95% CI = 1.081–3.820, p = 0.028) as independent factors of lymph node metastasis (Table 4).

In terms of tumor budding, the AUC=0.89 (95% CI: 0.85-0.93, p < 0.001) in the original group (budding) and AUC=0.81 (95% CI: 0.76-0.87, p < 0.001) in the regrouped tumor budding (phanlopB2) both revealed high prognostic ability for node metastasis (Figure 3). For E-cadherin immunoexpression, the AUC=0.87 (95% CI: 0.83-0.91, p<0.001) in the original group (EcadhGOC) and the AUC=0.83 (95% CI: 0.78-0.88, p<0.001) in regrouped E-cadherin expression (phanlopE1) both showed a high predictive ability of lymph node metastasis (Figure 4).

The rate of node metastasis in stomach cancer patients increased with pT stage in our study. Many other works have also found that the more advanced the gastric cancer, the deeper the invasion and the higher the rate of node metastasis17, 18, 19. The association of node metastasis with tumor budding and E-cadherin immunoexpression was confirmed by univariate analysis (Table 2). EMT of tumor cells may be the cause of these phenomena. EMT has been involved in the metastasis of stomach cancer and many other carcinomas20. This transformation has been described to downregulate E-cadherin expression (an immunomarker for epithelial cells) and upregulate vimentin expression (the mesenchymal marker)21, and be associated with increased tumor budding22. EMT may have a significant effect on the mechanism of tumor invasion, which has been suggested in some studies14, 15. Bronsert et al. have noticed that cells of tumor budding in various cancers (colorectal, pancreatic ductal carcinoma, lung adenocarcinoma, and invasive ductal breast carcinoma) had lost their polarity (more round and spindle shape), decreased E-cadherin staining intensity, decreased membrane E-cadherin staining, and increased nuclear zinc finger E-box binding homeobox 1 (ZEB1) staining23.

Human E-cadherin, a transmembrane glycoprotein mainly involved in epithelial cell adhesion, is encoded in the cadherin 1 (CDH1) gene of chromosome 16q22.124. E-cadherin binds to cytoskeleton actin to maintain cell structure stability, inhibits individual cell migration, and is involved in cell signaling25, 26. Increased abilities such as cell-cell interactions, migration, invasion, and metastasis of tumor cells are significantly influenced by aberrant E-cadherin immunoexpression27, 28.

The intensive (normal) expression rate of E-cadherin in cell membranes and cytoplasm was higher than the weak (abnormal) expression rate. Multivariate regression analysis demonstrated that E-cadherin immunoexpression is an independent factor associated with tumor grade (OR = 0.23, 95% CI = 0.12–0.43, p < 0.001), tumor budding (OR = 10.61, 95% CI = 2.73–41.16, p = 0.001), and lymph node metastasis (OR = 0.04, 95% CI = 0.012–0.134, p < 0.001). Several other studies have found abnormalities of E-cadherin immunoexpression in 25%29 or 46.7% – 48.6%30, 31 of cases and are closely related to lymph node metastasis. Several meta-analyses have suggested a significant association between E-cadherin expression and several variables, including invasion (p < 0.001), node metastasis (p < 0.001), and distant spread (p < 0.001), resulting in low five-year survival (p < 0.001)32, 33, 34, 35. Numerous studies have discovered E-cadherin as a crucial tumor suppressor in several cancers, including gastric carcinoma35, 36. One study found that 33% – 50% of incident gastric cancers had somatic E-cadherin inactivating mutations, and 63.6% of patients with signet ring cell carcinoma had decreased E-cadherin immunoexpression37. Expression dysfunction of E-cadherin is caused by multiple molecular mechanisms, for instance CDH1 mutations38, DNA hypermethylation32, and non-coding microRNAs. Reducing immunoexpression of E-cadherin at the cell membrane causes an attenuation or complete loss of cell-cell interactions and leads to the inhibition of the activation of transcription factors snail homolog 1 (SNAIL), twist family bHLH transcription factor 1 (TWIST), and ZEB1, leading to EMT33, 39. These reasons have caused differences in E-cadherin expression in different studies. CDH1 gene mutations correlate with many factors (geography, race, and especially eating habits) that are known to influence cancer risk in general.

Tumor budding should be distinguished from poorly differentiated cell clusters. A cell cluster with ≥ 5 cancer cells invading the stromal tissue and not forming glands is called poorly differentiated clusters40. The tumor budding concept has been mentioned for a long time, but only recently has it garnered interest in further research. It has been discovered that tumor budding has the potential to increase node metastasis, distant spread, tumor stage, and survival prognosis in numerous cancers, including colon cancer41, esophageal squamous carcinoma42, breast cancer43, and lung cancer44. ITBCC confirmed that tumor budding can predict the prognosis of node status in stage pT1 colorectal cancer and the survival ability of patients with stage II colorectal cancer11.

At the molecular level, the association between tumor budding and E-cadherin immunoexpression has been interpreted as tumor budding correlates with the high CpG island methylator phenotype (CIMP-H), which occurs in 40% of cancer cases. Microsatellite instability and CIMP-H cases often present with similar clinical and molecular features: tumor location, poor differentiation, and B-Raf proto-oncogene, serine/threonine kinase (BRAF) mutations45. E-cadherin loss in gastrointestinal carcinoma may be related to methylation of the O6-methylguanine-DNA methyltransferase (MGMT) promoter, which is part of the EMT pathway that induces Wnt pathway activation and nuclear beta-catenin translocation, thereby affecting E-cadherin expression41.

Predicting node status in stomach cancer is very useful because it affects the selection of appropriate surgical methods and treatment combinations. This study found that tumor budding and E-cadherin immunoexpression are independent factors and may be strong candidates for predicting node metastasis of primary stomach cancer based on surgical specimens. According to the conventional concepts, small biopsy samples often do not have enough necessary information such as pT stage, vascular, and nerve invasion, so predicting lymph node status using traditional methods is impossible. Therefore, this study can be considered a pilot work to predict lymph node metastasis before applying it to small biopsy samples of gastric cancer.

AUC: Area Under the Curve, BRAF: B-Raf Proto-Oncogene, Serine/Threonine Kinase, CDH1: Cadherin 1, CI: Confidence Interval, CIMP-H: CpG Island Methylator Phenotype - High, EMT: Epithelial-Mesenchymal Transformation, GLOBOCAN: Global Cancer Observatory, H&E: Hematoxylin and Eosin, ITBCC: International Tumor Budding Consensus Conference, MGMT: O6-Methylguanine-DNA Methyltransferase, OR: Odds Ratio, ROC: Receiver-Operating Characteristic, SNAIL: Snail Homolog 1, SPSS: Statistical Package for the Social Sciences, TWIST: Twist Family bHLH Transcription Factor 1, WHO: World Health Organization, ZEB1: Zinc Finger E-Box Binding Homeobox 1

We would like to thank the University for the study’s approval.

NVH, NTT: Conceptualization, Methodology, Writing-Original draft preparation; NTT, TVC: Visualization, Methodology, Software; TVC, NVH, DTL: Data curation, Writing-Original draft preparation; DTL, NTT, TVC: Validation, investigation, Supervision. All authors read and approved the final manuscript.

None.

Data and materials used and/or analyzed during the current study are available from the corresponding author on reasonable request.

This study received permission from the Ethics Committee of Hanoi Medical University (decision number: 4397/QD-DHYHN), complies with the ethical standards set forth in the Declaration of Helsinki, and was revised by the World Medical Association’s General Assembly in Seoul, South Korea, in October 2008.

Not applicable.

The authors declare that they have no competing interests.