The Association of Trefoil Factors with Gastric Cancer and Premalignant Lesions: A Cross-Sectional Population-Based Cohort Study

Although the incidence of gastric cancer has continuously decreased, it remains the third leading cause of cancer-related deaths worldwide (1). In 2018, the incidence (32.05/100,000) and mortality (27.42/100,000) rates of gastric cancer in China were 2–3 times higher than the world average (13.54/100,000 and 10.25/100,000, respectively). Furthermore, the incident cases and deaths accounted for 44.1% and 49.8% of the total gastric cancer occurrences worldwide, respectively (1, 2). The high incidence and mortality rates of gastric cancer in China are serious public health issues and impose immense burdens on the health care system. Patients with gastric cancer have a poor prognosis primarily because the majority of patients are diagnosed at advanced stages. Thus, early diagnosis is essential for reducing mortality due to gastric cancer.

Endoscopy is currently the standard modality for the diagnosis and clinicopathologic evaluation of gastric cancer. Studies have reported 47% and 67% reductions in the gastric cancer mortality due to gastroscopy screening in Korea and Japan, respectively (3, 4). However, population-based gastric cancer screening with endoscopy as the preferred screening modality in China is a formidable challenge due to China's vast territory and large population. Thus, the rate of early diagnosis in China remains low (5). Furthermore, endoscopy is highly invasive and relies heavily on the availability of endoscopic instruments and professional skills. These limitations make endoscopy less appropriate for large-scale population screening, especially in economically underdeveloped and low-risk areas. Therefore, simple, efficacious, and noninvasive biomarkers to screen high-risk populations, followed by subsequent endoscopic examination of identified at-risk individuals would be a reasonable strategy for mass screening with respect to gastric cancer.

Although pepsinogen (PG) and anti-Helicobacter pylori immunoglobulin G antibody (Hp-IgG) serum screening methods for gastric cancer have been in clinical practice for a substantial amount of time (6, 7), limitations in terms of sensitivity and positive predictive value have been reported, and this panel has low reliability (8, 9). Recent studies have reported on the pivotal role of the trefoil factor (TFF) family in the oncogenic transformation, growth, and metastatic extension of common human solid tumors, including gastric cancer (10, 11). The combination of serum TFF and PG can improve gastric cancer screening (12, 13). However, to the best of our knowledge, previous research on TFFs has been limited to patients with clinically confirmed gastric cancer, and the relationship between TFF levels and stages in Correa's cascade of gastric cancer development has not been investigated. The purpose of gastric cancer screening is to identify early gastric cancer and to monitor premalignant lesions (PML), as early detection provides a meaningful and valuable reduction in the incidence and mortality of gastric cancer.

The Wuwei region in China has consistently reported high incidence and mortality rates for gastric cancer (14). Thus, this region was selected as the setting for a large-scale endoscopic gastric cancer screening program between March 2013 and April 2016. The survey results from the Wuwei Cohort indicate a high number of PMLs, with 7,592 cases in total, including 3,444 cases of chronic atrophic gastritis (CAG), 1,788 cases of intestinal metaplasia (IM), and 2,283 cases of dysplasia, and 1.7% cases of gastric cancer were reported by pathology (167/9,623; ref. 15). Due to the large number of premalignant lesions in Wuwei, the Wuwei Cohort Study provides an opportunity to analyze the association between serum levels of TFFs and the different stages in Correa's cascade of gastric cancer development. We aimed to analyze the association of TFFs with gastric cancer and PMLs and undertake an initial exploration of the possibility of TFFs as biomarkers to predict the risk of PMLs [i.e., CAG, IM, low-grade intraepithelial neoplasia/dysplasia (LGD)] and gastric cancer in screening for high-risk gastric cancer populations.

This population-based screening study evaluated 25,000 adults ages 35–70 years residing in Wuwei, China (Wuwei Cohort) using a cluster sampling method (Fig. 1). Residents who (i) initially declined consent (n = 1,544), (ii) physically unable to provide consent (n = 85), (iii) pregnant (n = 17), or (iv) mentally ineligible (n = 8) were excluded from this study. Of the remaining 23,346 participants, 2,001 either declined participation or were ineligible for gastroscopy, and 21,345 underwent gastroscopy (15). We randomly selected 9,396 participants to undergo serologic testing for Hp-IgG, PGI, PGII, and high-sensitivity C-reactive protein (hs-CRP) using an automated biochemical analyzer. Among them, 4,069 subjects with pathology reports of gastric biopsy were extracted for TFFs, and 83 subjects with a medical history of malignancy were excluded. Therefore, 3,986 subjects who completed the survey and had complete baseline data were evaluated in the final analysis.

The Ethics Committee of The First Hospital of Lanzhou University approved this study (approval number: LDYYLL2012001), which was conducted in accordance with the principles of the Declaration of Helsinki. All of the subjects provided written informed consent before participation.

First, personal information, including a family history of gastric cancer, past medical history, lifestyle, dietary habits, and smoking habits, was collected using a self-reported questionnaire survey. Second, a 10-mL blood sample was drawn from a peripheral vein. The sample was immediately centrifuged, and the serum stored at −80°C. No freeze–thaw cycles were performed before the assays. Finally, a certified gastroenterologist performed endoscopic examination (GIF-H260Z/H290Z, Olympus), and the diagnosis of gastric cancer was established independently by three experienced pathologists.

The Wuwei Cohort database was established and managed using Stata software (StataCorp LLC). Because there was no loss of glands in the cascade stage of nonatrophic gastritis, the patients with nonatrophic gastritis were combined with the normal patients served as control group. The participants were classified into five groups based on Correa's cascade of gastric cancer development: control group (n = 773), CAG (n = 746), IM with CAG (n = 1,002), LGD accompanied by CAG and/or IM (n = 1,334), and high-grade intraepithelial neoplasia/dysplasia (HGD) or gastric cancer (n = 131). The PML groups include CAG, IM, and LGD groups. HGD was defined as equivalent to carcinoma in situ (16, 17). Gastric inflammation and atrophy were graded according to the Updated Sydney System (18), and a biopsy diagnosis of gastric epithelial neoplasia was established based on the World Health Organization classification (17).

The serum levels of TFF1, TFF2, and TFF3 (TFF1–3), pepsinogen I (PGI), pepsinogen II (PGII), anti-H. pylori IgG antibody titer, and hs-CRP were measured, and a ¹⁴C-urea breath test (¹⁴C-UBT) was performed. For TFF1–3, measurements were conducted using an enzyme-linked immunosorbent assay (ELISA) using a commercial kit (R&D Systems), following the manufacturer's instructions. Duplicate negative and positive controls were included in each 96-well plate. Each TFF antibody reacted specifically and showed no cross-reactivity with other TFFs.

The status of H. pylori infection was determined by measuring the serum levels of Hp-IgG and ¹⁴C-UBT (Shenzhen Zhonghe Headway Bio-Sci & Tech Co., Ltd). Positive infection was defined as Hp-IgG values >15 AU/mL. Hp-IgG (Wantai DRD), PG (Wantai DRD), and hs-CRP (Beijing Strong Biotechnologies) were measured via immunoturbidimetry on an automated clinical chemistry analyzer (Beckman Coulter AU5831). Samples that yielded implausible values according to referent intervals (i.e., the normal range of values) were retested.

For the descriptive statistics of the baseline characteristics, the study population was divided into five groups (control group, CAG, IM, LGD, and gastric cancer groups), and the participant characteristics were presented as mean ± standard deviation (SD), median [25th–75th percentiles (interquartile ranges)], or numbers and percentages as appropriate. The groups were compared using parametric or nonparametric test for continuous variables and a chi-square test for categorical variables. Due to a highly skewed distribution, the biomarkers were log-transformed and then standardized to have a mean of 0 and one SD of 1.

For the baseline TFFs and outcomes, three models were generated for evaluating odds ratio (OR) per SD and 95% confidence intervals (95% CI) in the following stages: (i) Model 1 was constructed after adjusting age and sex for each biomarker. (ii) After using forward stepwise procedure with P values for entry of 0.05 and P values for removal of 0.10, the following covariates were selected to create model 2, which was adjusted for age, sex, education, occupation, income, body mass index (BMI), smoking, alcohol consumption, eating hot food, eating of fried food, eating of fruits/vegetables, and medical history (gastritis, peptic ulcer, polyp, and high blood pressure). (iii) Finally, model 3 was constructed after additionally adjusting for all of the aforementioned biomarkers except PG II in model 2. Pearson partial correlations of TFFs with the continuous PGI, PGII, PGI/II ratio, and Hp-IgG were also reported.

All of the statistical analyses were performed using Stata software (version 15.0). P values < 0.05 were considered to indicate statistical significance. This was a pilot study; therefore, sample size and power calculations were not necessary.

The data used for this study, though not available in a public repository, will be made available to other researchers upon request.

The participants' average age was 51.8 years, and the majority of the participants (54.7%) were male (Table 1). In total, 60.3% of the participants had H. pylori infection as confirmed by serum Hp-IgG testing. The severity of gastric mucosal lesions gradually worsened within the disease pathogenesis stages from the control group to the gastric cancer group. The lowest and highest average age, males, and smokers’ proportions of 50.7 versus 57.1 years, 46.2% versus 74.8%, and 33.9% versus 48.9%, occurred in the control group versus gastric cancer group, respectively. Farmers accounted for 91.1% of the population, with an average annual income of 20,000–28,000 RMB Yuan per family, representing a typical low-income status. The rate of alcohol consumption also increased from the control group to the LGD group. However, lower alcohol intake was noted in the gastric cancer group. More than 60% of the subjects in the gastric cancer and PML groups had a usual habit of eating hot food. Only 6.7% of the subjects regularly consumed fruits and vegetables for six months, which may weaken the role of vegetables and fruits as a protective factor for gastric cancer. The medical history of gastritis, peptic ulcers, and gastric polyps was statistically significant between the PML and gastric cancer groups (P < 0.05).

The serum biomarker distributions in the five groups are shown in Table 1. The serum levels of TFF1 and TFF2 showed an incremental trend from the control group, CAG, IM, and LGD groups to the gastric cancer group [TFF1 median (Q1–Q3) pg/mL: from 42.2 (15.2–99.2) in the control group to 134.0 (49.2–239.5) in the gastric cancer group; TFF2 median (Q1–Q3) pg/mL: from 1,811.0 (1,056.6–2,991.4) in the control group to 3,647.9 (2,485.9–6,164.2) in the gastric cancer group; P < 0.001]. However, the distribution of serum TFF3 levels differed across the five groups and was significantly higher in the IM group [median (Q1–Q3) pg/mL, 7,195.1 (5,650.0–9,211.9)]. Serum levels of Hp-IgG gradually increased from the control group [median (Q1–Q3) AU/mL, 24.0 (3.6–54.3)] to the LGD group [median (Q1–Q3) AU/mL, 34.4 (10.0–62.0)]. However, serum level of H. pylori decreased in the gastric cancer group [median (Q1–Q3) AU/mL, 29.8 (12.7–54.3); P < 0.001]. However, the rate of H. pylori–positive infection as confirmed via Hp-IgG titer gradually increased from the control group (57.3%) to the LGD group (69.6%), and the highest value was observed in the gastric cancer group (72.5%). The serum levels of PGI differed among the groups (P = 0.025). The PGI/II ratio showed a declining trend from the control group to the gastric cancer group, whereas the serum levels of PGII showed a reversing trend (P < 0.001). No significant association was detected between serum hs-CRP level and the histologic grade (P = 0.84).

The correlation coefficient analysis results among the evaluated biomarkers were shown in Table 2. After adjusting for age and sex, we observed a strong positive partial correlation between PGI and PGII (r > 0.50, P < 0.001) and between PGII and Hp-IgG (r > 0.50, P < 0.001). We likewise observed moderate significant correlations between Hp-IgG and TFF1 or TFF2, PGI and TFF2, PGII and TFF1 or TFF2, and TFF1 and TFF2 (0.20 < r ≤ 0.50, P < 0.001). We found weak correlations between TFF1 and TFF3, TFF2 and TFF3, and PGI and TFF1 (r ≤ 0.20, P < 0.001).

Table 3 shows the results of comparing the CAG, IM, LGD, and gastric cancer groups with the control group, as well as analyzing the association of each biomarker with the gastric cancer and PML groups. The referent OR for the control group was defined as 1. In the three models, the ORs for TFF1, TFF2, and TFF3 were statistically significantly higher than those for PGI, PGII, the PGI/II, and Hp-IgG in the PMLs (i.e., CAG, IM, and LGD) and gastric cancer groups (P < 0.05). It is worth noting that the ORs (95% CI, P value) for TFF3 in models 1–3 of the IM group were particularly prominent; the ORs and associated CIs were 2.40 (2.08–2.76, P < 0.001), 2.36 (1.76–3.15, P < 0.001), and 1.92 (1.64–2.25, P < 0.001), respectively. Compared with PG and Hp-IgG, TFF1, TFF2, and TFF3 had greater odds ratios with respect to the gastric cancer group. The ORs (95% CI, P value) for TFF1 TFF2, and TFF3 in model 3 for the gastric cancer group were 1.67 (1.27–2.20, P < 0.001), 2.66 (2.01–3.51, P < 0.001), and 1.32 (1.00–1.74, P = 0.050), respectively. In CAG, IM, LGD, and gastric cancer groups, the OR (95% CI) for Hp-IgG were 1.14 (1.02–1.26), 1.66 (1.38–2.01), 2.02 (1.67–2.45), and 1.40 (1.11–1.77; all P < 0.05), respectively, in model 2, these ORs were approximately 1 (all P > 0.05) in model 3 after multivariate adjustment. Similarly, in the PML and gastric cancer groups, the ORs for the PGI and PGI/II ratio were marginally close to 1 in model 3 (P > 0.05).

Gastric carcinogenesis is a multifactorial, multistep process normally described by a series of events known as Correa's cascade. This process begins with chronic active gastritis and progresses through a chain of mucosal changes including gastric atrophy, IM, and intraepithelial neoplasia/dysplasia (also known as PML), finally developing into gastric cancer (16, 19). Due to the high detection rate for PMLs (78.9%) and gastric cancer (1.7%) in the Wuwei Cohort, there is an urgent need to explore the association of biomarkers with PMLs and gastric cancer in resource-limited areas (15). Studies have reported that the detection sensitivity of the serum TFF3 and PG test for gastric cancer is 80.0%–80.4% and 33.3%–39.5%, respectively (12). Serum TFF3 is a better gastric cancer screening marker than PG (12, 13). In particular, serum levels of TFF3 remain stable for the status and eradication of H. pylori infection (20). These studies have motivated us to investigate the effectiveness of the TFF family (TFF1–3) as biomarkers for assessing the risk of gastric cancer and PMLs in a timely and accurate manner among the primary population in a high-risk area for gastric cancer. This group of biomarkers could function as an essential strategy for early inventions and help reduce the mortality and morbidity of gastric cancer. TFF peptides have a stable triple-loop structure that may have apparent resistance to protease hydrolysis, acid digestion, and heat treatment. Thus, TFFs can maintain biological activity in the complex environment of the digestive tract (21, 22). Commonly, TFF1 and TFF2 are expressed mainly in the mucosal epithelial cells of the gastric body and antrum, and TFF3 is expressed in the goblet cells of the small intestine and the colon (23, 24). However, under pathologic conditions, the site-specific expression of TFFs is absent, and TFFs can be detected in any damaged mucosae as their expression is upregulated to participate in gastrointestinal epithelial reconstruction and mucosal repair process (25, 26). Moreover, in the development of gastric cancer, TFF expression can be dynamically escalated (27, 28). This is expected since the outline of the carcinogenesis process, as described in Correa's cascade process, is characterized by gastric mucosal lesions and gland atrophy (16, 29). In this study, from PMLs (i.e., CAG, IM, and LGD) to gastric cancer, the serum levels of TFF1 and TFF2 changed with an incremental trend (P < 0.05; Table 1), and the TFF3 level was remarkably high in the IM group.

We found statistically significant associations between the severity of gastric mucosal lesions when progressing from PMLs to gastric cancer and sex, age, occupation, smoking, and a medical history of gastritis, peptic ulcer, and polyps. Subsequent analyses were conducted based on a complete profile combining all of the aforementioned biomarkers, and these meaningful baseline characteristics were adjusted to eliminate mutual influence and to explore relatively better independent biomarkers for the clinical screening of high-risk populations.

In Correa's cascade, H. pylori infection initiates gastric inflammation that may advance into PG-related gastric atrophy (30). PG and Hp-IgG can be regulated by more than one trigger or by mutual interactions. In Table 2, we present results of a strong positive partial correlation between PGII and Hp-IgG (r > 0.50, P < 0.001). Interestingly, after adjusting for multiple factors and biomarkers in Table 3, the ORs for the Hp-IgG, PGI, and PGI/II ratio in each group of models 3 were all near 1 (all P > 0.05), indicating that Hp-IgG and PG are not independent markers for high-risk gastric cancer. In this study, we observed trends toward a stepwise increase in the PGII level and a reduction in the PGI/II ratio when progressing from the PML to gastric cancer groups. The anti-H. pylori IgG antibody titer for the gastric cancer group was lower than that for the PML group, which may be due to extensive atrophy that could lead to the loss of H. pylori infection or eradication of H. pylori. Additionally, lower antibody titers in advanced gastric cancer may be partly attributable to a diminished immune response (29). Studies have shown that a false-negative result of H. pylori infection in patients with anti-H. pylori IgG antibody titer of 3–9.9 U/mL (as a negative-high titer) causes the misdiagnosis of high-risk gastric cancer populations (31); hence, additional examinations, including ¹³C-UBT tests, stool antigen tests, and endoscopy, should be performed to overcome this limitation. However, these cumbersome examinations cannot provide simple and effective population-based screening for gastric cancer. Individuals with positive anti-H. pylori IgG antibodies include those with a current or former H. pylori infection or Helicobacter infections other than H. pylori (32, 33). This is consistent with our finding that the positivity rate of H. pylori infection detected by Hp-IgG was higher than that detected via ¹⁴C-UBT (Table 1). Hence, if the anti-H. pylori IgG antibody was used as the screening marker, the false-positive rate in the screening population may increase.

The genes of three TFFs are adjacently located on chromosome 21q22.3 and share five regulatory sequences to coordinate expression (34, 35). Taupin and colleagues demonstrated that TFF peptides could function as the instant early controllers fit for auto- and cross-acceptance of other trefoil peptides in gastric cell lines (36). Additionally, diminished expression of TFF1 and TFF2 has been found in TFF3 knockout mice, and a decrease in TFF2 levels has been observed in TFF1 knockout mice (36, 37). The cross-enlistment of TFFs requires activation by means of phosphorylation of the epidermal growth factor receptor, which is actuated by intestinal trefoil factor (TFF3; ref. 38). TFF3 is normally expressed in the goblet cells of the small and large intestine; however, TFF3 is also expressed in IM of the stomach. Dysplasia is usually manifested by incomplete components of the deep metaplastic glands, pseudostratified nuclei, and large irregular cells. Eventually, the gland structure disappears, tumor cells appear, and progression to gastric cancer occurs (39). In our study, the ORs of TFF3 (Table 3) in the IM group were higher than those in the control group, with each 1-SD increase in TFF3, the risk of IM increased by 92%. If intestinal metaplasia were to be considered an intermediate step in the development of gastric cancer, the induction of TFF3 expression could be anticipated during the progression from a metaplastic epithelium to cancer (40). The increased expression of TFF3 may be an early indicator of IM and intestinal gastric cancer. Therefore, the value of TFF3 lies in its ability to predict the risk of IM. However, in this study, we did not stratify our groups because of the small sample size, which could have caused an underestimation of the predictive value of TFF3 for gastric cancer.

Dhar and colleagues reported that the TFF2 mRNA and its protein product displayed a high level of expression, as well as a correlation with the clinicopathologic stage and/or prognosis of gastric cancer (41). In addition, Aikou and colleagues reported that the area under the curve for TFF3 and TFF1 for patients with gastric cancer was higher than those of the healthy subjects as control group; it has also been reported that the combined serum PGI/II ratio and TFF3 could improve gastric cancer screening (13). However, to the best of our knowledge, the association of the TFF family with PMLs has not been reported. In our study, statistically significant associations between TFFs and premalignant lesions as well as gastric cancer were observed after adjusting for the risk factors and the biomarkers. Compared with those for the PG and Hp-IgG, the ORs of TFFs were larger (Table 3), indicating that TFFs are more reliable predictors of premalignant lesions and gastric cancer than PG and Hp-IgG. In fact, the risk of gastric cancer increases in connection with the severity of premalignant gastric lesions. A successful prevention strategy for gastric cancer depends on the recognition and understanding of the different stages of the premalignant cascade. It was speculated that TFFs had considerable advantages in assessing the risk of PMLs and gastric cancer.

The TFF family has a protective effect against mucosal damage in the gastrointestinal tract, and these biomarkers play a crucial role in the progression of gastric cancer. TFFs influence disease pathogenesis by altering normal mucosal recovery, thus aggravating the ongoing inflammation (42). Additionally, TFFs are also involved in controlling the coordination of proliferation, apoptosis, and differentiation of gastrointestinal cells (43). TFF overexpression is associated with increased cell migration and possibly increased gastric cancer invasion and angiogenesis (41, 44). These effects may be due to TFF gene mutations (45), chromatin remodeling, loss of heterozygosity, or promoter methylation (46). The function of TFFs as regulatory factors for gastrointestinal cell differentiation has been well demonstrated in prior research. The presence of TFFs may help maintain the balance between proliferation and apoptosis of gastrointestinal cells during early tumorigenesis. Therefore, TFFs play an important role in the development and progression of gastric cancer, and TFF levels may serve as a risk indicator for the progression from premalignant to cancerous lesions. TFF expression status also appears to be a promising prognostic marker in patients with gastric cancer.

In addition to the substantial strengths of this investigation, our study has some limitations. For example, the cohort was recruited from a single-center population, and the sample size of the control group was limited. The generalizability of our results is unclear, and our results need to be verified in multicenter studies. Nevertheless, this study is the first to demonstrate that TFF1, TFF2, and TFF3 have good associations with gastric cancer and PMLs and that TFF1–3 have the potential to assess the risks of PMLs and gastric cancer based on a large-scale population. In the future, prospective studies are needed to validate the association between TFFs expression and clinical characteristics and the outcomes of gastric cancer, as well as the role of TFFs expression in cancer progression.

In conclusion, all of the biomarkers evaluated in this study were associated with gastric cancer and PML, whereas the associations were more reliable for TFFs than for PG and Hp-IgG, suggesting that TFFs could be used to help guide the determination of which populations may be the better candidates for gastric cancer earlier screening.

No disclosures were reported.

X. Zan: Conceptualization, resources, data curation, investigation, writing–original draft, writing–review and editing. Z. Chen: Resources, validation, investigation, and project administration. Q. Guo: Resources, supervision, investigation, and project administration. Y. Wang: Resources, investigation, and project administration. Z. Zhang: Resources, investigation, and project administration. R. Ji: Resources and investigation. Y. Zheng: Data curation, software, formal analysis, validation, and methodology. J. Zhang: Resources and investigation. Z. Wu: Resources and investigation. M. Li: Formal analysis and investigation. X. Wang: Resources and investigation. Y. Ye: Resources and investigation. X. Li: Resources and investigation. F. An: Resources and investigation. C. Xu: Resources and investigation. L. Lu: Resources and investigation. P. Fan: Resources and investigation. J. Zhang: Resources and investigation. Q. Guan: Resources and investigation. Q. Li: Resources and investigation. M. Liu: Resources and investigation. Q. Ren: Resources and investigation. X. Hu: Software and formal analysis. H. Lu: Resources and investigation. Y. Wang: Resources and investigation. H. Zhang: Resources and investigation. Y. Zhao: Resources and investigation. X. Gou: Resources and investigation. X. Shu: Resources and investigation. J. Wang: Resources and investigation. Z. Hu: Resources and investigation. R. Liu: Resources and investigation. H. Yuan: Resources and investigation. J. Liu: Data curation, software, formal analysis, and methodology. L. Qiao: Conceptualization, methodology, and project administration. Y. Zhou: Conceptualization, resources, supervision, investigation, methodology, and project administration.

The authors are grateful to all the study participants and staff from the collaborating hospitals. We thank the interviewers from the CDC of Wuwei City and Department of Epidemiology and Statistics at the School of Public Health at Lanzhou University. Y. Zhou received a grant from Ministry of Science and Technology of the People's Republic of China (2012GS620101).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.