研究目的

基于我国上消化道癌高发区内镜筛查项目，量化评估不同贲门粘膜病变的长期癌变风险，评价血清胃蛋白酶原、胃泌素及幽门螺杆菌（Helicobacter pylori, H. pylori）抗体分型等胃粘膜“血清学活检”指标用于贲门癌（gastric cardia adenocarcinoma，GCA）及癌前病变的诊断价值，并在蛋白质组学层面探索与贲门癌发生发展相关的蛋白标志物，以期为我国贲门癌内镜筛查方案的优化提供数据参考及新思路。

材料与方法

1. 我国上消化道癌高发区贲门癌及癌前病变发病风险的前瞻性队列研究：本研究为基于自然人群的前瞻性队列研究。以2005-2009年在我国上消化道癌高发区河南林州的筛查队列为基础，选取队列中参与内镜筛查并有明确病理诊断的40-69岁人群作为目标人群，整理收集目标人群一般人口学信息、流行病学暴露信息及内镜病理诊断信息。根据基线病理诊断，将目标人群分为正常（健康）、非萎缩性贲门胃炎（non-atrophic carditis, NAC）、萎缩性贲门胃炎（atrophic carditis, AC）、低级别上皮内瘤变（cardia low-grade dysplasia, CLGD）、高级别上皮内瘤变（cardia high-grade dysplasia, CHGD）组。通过主动随访与被动随访相结合的方式，调查目标人群贲门癌发病情况与死亡情况。队列随访截止至2017年12月。以人年为单位计算队列人群贲门癌标化发病率、死亡率。将随访期间贲门癌发病人数除以随访总人数计算贲门癌1年、3年、5年累积发病率，Log-Rank法比较不同组间发病/死亡风险之间的差异。采用COX比例风险模型计算基线贲门粘膜病变状态对贲门癌发病/死亡的风险比及95%的置信区间，并按照性别、年龄组展开亚组分析。

2. 胃功能血清学指标对贲门癌及癌前病变诊断价值的研究：本研究为一项以人群为基础的病例对照研究。研究对象来自2017-2019年河南林州及山东肥城参与过食管癌专病队列研究的普通人群。收集队列中基线病理诊断为早期GCA（粘膜内腺癌）、CHGD及CLGD患者的受检者，从中选择符合条件的患者作为病例组。以早期GCA患者为参照，根据年龄及性别，按照1:2的匹配原则选择队列中健康人群作为通用对照组。使用酶联免疫吸附实验对两组研究对象的血清胃蛋白酶原（pepsinogen, PG）I/II、胃泌素17（gastrin-17, G-17）水平做定量检测，通过免疫印迹法对血清H. pylori抗体做分型检测。描述各组间PG I/II、胃泌素及H. pylori抗体分型的分布情况，结合Logistic回归模型、受试者工作特征曲线 （receiver operating characteristic curve，ROC）及曲线下面积（area under curve，AUC）分析上述胃功能血清学指标在贲门癌及癌前病变诊断中应用价值。

3. 基于蛋白质组学的贲门癌生物标志物探索性研究：本研究为一项以人群为基础的病例对照研究。研究以经病理确诊的GCA、CHGD、CLGD及健康者为研究对象。血浆蛋白提取后通过LC-MS进行分离，在数据依赖性采集（data-dependent acquisition, DDA）模式建立扫描图谱库的基础上，采用数据非依赖性采集（data-independent acquisition, DIA）模式进行质谱数据采集。通过主成分分析（principal component analysis, PCA）及偏最小二乘法判别分析（partial least squares discriminant analysis, PLS-DA），观察各组间蛋白表达谱的整体分类趋势。通过火山图分析健康、GCA、CHGD、CLGD四组间两两比较时的差异蛋白。通过加权基因共表达网络分析（weighted correlation network analysis, WGCNA）识别与GCA发病相关的关键模块，对关键模块进行功能GO分析及KEGG/ WikiPathways通路分析。采用随机森林（random forest, RF）和套索（least absolute shrinkage and selection operator, LASSO）回归浓缩关键模块中的特征蛋白。结合蛋白组织特异性，筛选潜在蛋白标志物，构建诊断Panel，通过ROC曲线下面积AUC评价其对贲门癌及癌前病变诊断能力。

研究结果

1. 2005-2009至2017年间，队列中共有9,740人纳入最终的随访分析。随访期间，共有123名新发贲门癌患者并有31人因贲门癌死亡。队列中位随访时间为10年，目标人群贲门癌年龄标准化发病率和死亡率分别为128.71/10万人年和35.69/10万人年。健康组、NAC组、AC组、CLGD组和CHGD组贲门癌累积发病率分别为0.57%（27/4728）、0.71%（22/3099）、1.58%（20/921）、3.05%（27/884）和25%（27/108）。贲门癌长期发病风险与死亡风险均随病变级别的升高而升高。此外，亚组分析结果表明，50-69岁年龄组的贲门癌累积发病率和死亡率分别是40-69岁年龄组的4.4倍和4.7倍；男性贲门癌累积发病率是女性的1.8倍，死亡率是女性的2倍。同正常组相比，AC组累积发病率在随访第五年后才呈现上升趋势，而CLGD组在随访前3年便已出现显著差异。

2. CHGD、GCA组中PGII水平高于对照组（P<0.05），疾病相关风险随着PGII水平升高而升高（P _trend<0.05）。PGR水平与CHGD（OR=2.42, 95%CI: 1.19-4.92）、GCA（OR=2.04, 95%CI: 1.00-4.15）及全疾病（OR=1.89, 95%CI: 1.08-3.29）发病呈负相关。各病例组PGI及G-17水平与对照组相比无显著差异（P>0.05）。对照组、CLGD组、CHGD组及GCA组H. pylori感染率分别为53.33%、71.67%、71.88%及71.67%。H. pylori感染与CLGD （OR=2.23, 95%CI:1.11-4.47）、CHGD（OR=2.27, 95%CI:1.17-4.39）、GCA（OR=2.21, 95%CI:1.14-4.32）的发生有关，且趋势性分析结果表明，病变程度与H. pylori感染率之间呈正相关（P _trend<0.05）。Type I及Type II均阳性者与贲门癌前病变发生密切相关（OR=2.67, 95%CI: 1.57-4.54），其中，CagA (+)/VagA (+)（OR=2.64, 95%CI: 1.44-4.84）以及CagA (-)/VagA (+)（OR=7.98, 95%CI: 1.71-32.21）均可观测到较高的发病风险。诊断试验结果表明，五种胃粘膜“血清学活检”指标联合诊断可发现人群中66.7%的癌前病变患者，联合诊断对于CLGD的诊断效果最高（AUC=0.72, 95%CI: 0.65 - 0.79），对CHGD（AUC=0.67, 95%CI: 0.62 - 0.72）及GCA（AUC=0.64, 95%CI: 0.56 - 0.71）的诊断效果欠佳。剔除PGI及G-17后，诊断效果未发生显著变化。

3. PCA及PLS-DA分析显示，健康组、CLGD组、CHGD组及GCA组血浆蛋白存在显著的分类趋势，观察到变异的主要来源可用分组差异来解释；且任意两组进行比较时，蛋白表达谱均存在一定程度的差异，为差异蛋白的筛选提供了先验证据。WGCNA对蛋白表达谱与疾病进程进行了关联分析，42个蛋白分子与GCA发展正相关，正相关中蛋白基因在微量元素硒相关信号通路、泛素蛋白酶体通路、金黄色葡萄球菌感染以及雄激素受体信号通路中富集；165个蛋白分子与GCA发展负相关，负相关中蛋白基因在肌蛋白调节、自噬信号通路、粘着斑信号通路、脂质代谢、病毒致癌作用、雌激素相关信号通路、谷胱甘肽信号通路中富集。RF得到的Top200蛋白标志物与WGCNA关键模块取交集，最终得到67个蛋白分子。利用LASSO回归结合蛋白组织特异性，发现了LY6G6F、CSRP1、PRDX1、GSTP1及RAB27B等5个潜在蛋白质生物标志物。交叉验证结果显示，构建的Panel对贲门癌及癌前病变具有较好的诊断效果，AUC为0.96（95%CI:0.77-1.00）。

研究结论

1. CHGD患者癌变风险较高，一经发现应及时接受治疗，拒绝治疗者每半年随访一次；CLGD患者应在初次内镜后1-3年内接受复查；AC/CIM患者的内镜随访间隔延长到5年1次。从筛查成本的角度出发，可将高发区内镜筛查起始年龄推迟到50岁。本研究结果提供了各级别癌前病变的贲门癌发病的绝对风险与相对风险，为我国上消化道癌高发区内镜筛查方案的优化提供了循证医学依据，并为下一步内镜筛查方案的推广提供了数据支持。

2. 相比于PGI及G-17，血清PGII及PGR水平更能反映贲门粘膜的病变进展情况，I型H. pylori感染，尤其是VagA型H. pylori感染与贲门癌前病变发生关联最为密切。PGII、PGR联合血清H. pylori感染情况对于仅累及贲门部的贲门癌及癌前病变有着较好的诊断应用前景，但仍需要更多研究证据加以支持。

3. 肿瘤恶病质所致的肌损失、硒水平及性激素相关通路与GCA进展相关，在疾病发病的纵贯视角发现的5个蛋白质标志物对于贲门癌及癌前病变诊断具有显著效果。通过蛋白质组学发现潜在标志物，为识别和浓缩贲门癌高危个体提供了新的研究方法及思路，但基于小规模人群所筛选的生物标志物在应用于人群筛查前，还需要再大规模、前瞻性研究中加以验证。

Objectives：

Based on the endoscopic screening program in high-risk areas of upper gastrointestinal cancer in China, the present study assessed and quantified the long‐term risk of gastric cardia precancerous lesions, evaluated the diagnostic value of "serological biopsies" indicators such as serum pepsinogen (PG), gastrin and Helicobacter pylori (H. pylori) for gastric cardia adenocarcinoma (GCA) and precancerous lesions, and explored novel blood plasma protein biomarkers for GCA by proteomic techniques based on mass spectrometry, in order to provide reference data and new ideas for the optimization of endoscopic screening scheme for GCA in China.

Materials and Methods：

    1. A Prospective cohort study on long-term gastric cardia adenocarcinoma risk of precancerous lesions in a high incidence area of upper gastrointestinal cancer in China. This part was a prospective cohort study based on asymptomatic healthy population. Based on the screening cohort in Linzhou, Henan Province, a high incidence area of upper gastrointestinal cancer in China from 2005 to 2009, the population aged 40-69 years who participated in endoscopic screening and had definite pathological diagnosis was selected as the target population, and the general demographic information, epidemiological exposure information and endoscopic pathological diagnosis of the target population were collected. According to the baseline pathological diagnosis, the target population was divided into Normal, non-atrophic carditis (NAC), atrophic carditis (AC), cardia low-grade dysplasia (CLGD), and cardia high-grade dysplasia (CHGD). The incidence and mortality of GCA in the target population were investigated by active follow-up and passive follow-up. The deadline for follow-up was on December 31st, 2017. The standardized incidence and mortality of GCA in the cohort population were expressed by person-year. The incidence of GCA during follow-up was divided by the total number observed at baseline to calculate the 1-year, 3-year, 5-year and the total cumulative incidence of GCA. Log-Rank method was used to compare the differences in the risk of morbidity/mortality among different groups. The COX proportional hazards model was used to estimate the hazard ratio (HR) and 95% confidence interval (CI) of baseline mucosal lesion status to cardia cancer incidence/death, and subgroup analyses were performed by sex and age group.

    2. The diagnostic value of serological tests related to gastric function for GCA and precancerous lesions. This part was a multicenter, population-based case-control study. The subjects were from the general population who participated in the esophageal cancer-specific disease cohort study in Linzhou, Henan Province and Feicheng, Shandong Province, from 2017 to 2019. Subjects with baseline pathological diagnosis of early GCA, CHGD and CLGD were collected from the cohort. With early GCA patients as reference, the healthy population in the cohort was selected as the general control group according to age and sex according to the 1: 2 matching principle. The serum levels of PG I/II and gastrin were quantitatively detected by enzyme-linked immunosorbent assay (ELISA), and the serum antibodies of H. pylori were typed by immunoblotting. To describe the distribution of PGI/II, gastrin and H. pylori antibodies typing among each group, and analyze the value of the above-mentioned serum indexes of gastric function in the diagnosis of cardia cancer and precancerous lesions in combination with Logistic regression model and ROC curve.

    3. Plasma biomarkers in GCA by mass spectrometry-based proteomics. This part was a case-control study. Plasma samples from patients with pathologically confirmed GCA, CLGD, CHGD and Normal cardia were studied. The extracted plasma proteins were separated by LC-MS, and the mass spectrometric data were collected in DIA mode on the basis of establishing a scanned spectrum library in DDA mode. The overall classification trend of protein expression profiles among the groups was obtained by principal component analysis (PCA) and partial least squares discriminant analysis (PLS-DA). The differential proteins of Normal, GCA, CHGD and CLGD were analyzed by volcano plots. The key modules related to the pathogenesis of GCA were identified by weighted correlation network analysis (WGCNA). Functional GO analysis and KEGG/WikiPathways pathway analysis were performed on the key modules. Random forest (RF) and the least absolute shrinkage and selection operator (LASSO) were used to regress the characteristic proteins in the key modules. In combination with the tissue specificity of the protein, potential protein markers were screened, and a diagnostic Panel was constructed, and its ability to diagnose cardia cancer and precancerous lesions were evaluated by area under the ROC curve AUC.

Results：

    1. Between 2005-2009 and 2017, a total of 9,740 subjects were included in the final cohort follow-up. During the follow-up period, there were 123 patients diagnosed with GCA and 31 died from GCA. During a median follow-up time of 10.0 years, the age-standardized incidence and mortality rates of GCA in the target population were 128.71 and 35.69 per 100,000 person-years, respectively. The cumulative incidences of GCA were 0.57% (27/4728), 0.71% (22/3099), 1.58% (20/921), 3.05% (27/884) and 25% (27/108) in the Normal, AC, CLGD and CHGD groups, respectively. Compared with the Normal group, the cumulative incidence in the AC group showed an increasing trend after the fifth year of follow-up, whereas the CLGD group had a significant difference in the first 3 years of follow-up. The risk of long-term morbidity and mortality of GCA increased with the increase of lesion grade. In addition, the results of subgroup analysis showed that the cumulative incidence of GCA in the 50-69 years age group was 4.4 times higher than that in the 40-49 age group, and the mortality rate was 4.7 times higher than that in the 40-49 age group; the cumulative incidence of GCA in male subjects was 1.8 times higher than that in women, and the mortality rate was 2 times higher than that in women.

    2. The levels of PGII in CHGD and GCA groups were higher than those in the control group (P < 0.05), and the disease-related risk increased with the increase of PGII levels in the CHGD group (P for trend = 0.13), GCA group (P for trend <0.05) and the whole disease group (P for trend <0.05). PGR levels were inversely associated with CHGD (OR = 2.42, 95%CI: 1.19-4.92), GCA (OR = 2.04, 95%CI: 1.00-4.15), and total disease (OR = 1.89, 95%CI: 1.08-3.29). The levels of PGI and G-17 in each case group were not significantly different from those in the control group (P > 0.05). The infection rates of H. pylori in control group, CLGD group, CHGD group and GCA group were 53.33%, 71.67%, 71.88% and 71.67%, respectively. H. pylori infection was associated with the occurrence of CLGD (OR = 2.23, 95%CI: 1.11-4.47), CHGD (OR = 2.27, 95%CI: 1.17-4.39), GCA (OR = 2.21, 95%CI: 1.14-4.32), and the trend analysis showed that there was a positive correlation between the degree of H. pylori infection (P for trend<0.05). Type I and Type II H. pylori infection was closely related to the occurrence of cardia precancerous lesions (OR = 2.67, 95%CI: 1.57-4.54). There were 66.7% of the patients with precancerous lesions in the population could be detected by the combined diagnosis of five gastric mucosal "serological biopsy" indicators. The combined diagnosis had the highest diagnostic effect for CLGD (AUC = 0.72, 95%CI: 0.65-0.79) and the lowest for GCA (AUC = 0.64, 95%CI: 0.56-0.71). After excluding PGI and G-17, the diagnostic effect did not change significantly.

    3. PCA and PLS-DA analysis showed that there was a significant classification trend in serum proteins among the four groups, and the main source of variation could be explained by group differences; there were some differences in protein expression profiles between any two groups, which provided a priori evidence for the screening of different proteins. WGCNA was used to analyze the correlation between protein expression profile and disease progression. Forty-two proteins were positively correlated with the development of GCA, and positive correlation protein genes were enriched in trace element selenium-related signaling pathway, ubiquitin-proteasome pathway, Staphylococcus aureus infection and androgen receptor signaling pathway; 165 protein molecules were inversely associated with GCA development, and protein genes were enriched in myosin regulation, autophagy signaling pathway, focal adhesion signaling pathway, lipid metabolism, viral carcinogenesis, estrogen-related signaling pathway, and glutathione signaling pathway. RF-derived Top200 protein markers were crossed with those associated with GCA pathogenesis in WGCNA analysis, resulting in 67 key protein molecules. Five potential protein biomarkers, such as LY6G6F, CSRP1, PRDX1, GSTP1 and RAB27B, were identified by LASSO regression combined with tissue specificity. Cross-validation results showed that the constructed Panel had a good diagnostic effect on GCA and precancerous lesions, with an AUC of 0.96 (95% CI: 0.77-1.00).

Conclusion：

    1. Based on the above results, we suggest that patients with CHGD should receive treatment once detected, and those who refuse treatment should be followed up every six months; patients with CLGD should be reexamined within 1-3 years of initial endoscopy; the endoscopic follow-up interval for patients with AC/CIM was extended to 5 years. Our results also suggest that the age of endoscopy initiation in high-incidence areas can be postponed to 50 years from the perspective of screening costs. The results of this study provide the absolute and relative risks of gastric GCA at all levels of precancerous lesions, provide evidence-based medical evidence for the optimization of endoscopic screening programs in high-risk areas of upper gastrointestinal cancer in China, and provide data support for the popularization of endoscopic screening programs in the next step.

    2. Compared with PGI and G-17, serum PGII and PGR levels could reflect the progression of gastric cardia mucosa. Type I H. pylori infection, especially VagA H. pylori infection, was most closely associated with the occurrence of precancerous lesions of the cardia. PGII, PGR combined with serum H. pylori infection has a good prospect in the diagnosis of early GCA and precancerous lesions involving only the cardia, but more research evidence is still needed to support it.

    3. Muscle loss, selenium levels and sex hormone-related pathways caused by tumor cachexia are associated with the mechanism of GCA progression. The 5 protein markers found in the longitudinal viewpoint of disease onset have significant effects on the diagnosis of GCA and precancerous lesions. The discovery of potential markers by proteomics provides a new idea for identifying high risk group of GCA, but the biomarkers found based on small populations need to be validated in further large-scale, prospective studies before they are applied to population screening.