早期乳腺癌患者唾液蛋白质异常糖基化的研究

发布时间：2018-08-15 16:14

【摘要】：研究背景:乳腺癌是全世界范围内女性发病率最高的一类恶性肿瘤。目前,尚无理想的体外分子标志物可用于乳腺癌的早期诊断。大量研究表明:伴随肿瘤的发生与发展,患者肿瘤组织及体液中蛋白质糖基化水平常发生异常改变。本研究从唾液检测入手,利用高灵敏度且高通量的糖链检测技术——凝集素芯片,对健康女性志愿者(healthy volunteers,HV)、良性乳腺肿瘤/囊肿患者(benign breast diseases patients,BB)、Ⅰ期乳腺癌患者(patients with breast cancer in Ⅰ stage,BC-Ⅰ)以及 Ⅱ 期乳腺癌患者(patients with breast cancer in Ⅱ stage,BC-Ⅱ)的唾液蛋白质糖链结构进行比较分析,期望发现异常的唾液蛋白质糖链结构,讨论其作为检测指标应用于乳腺癌患者早期诊断的可能性。实验方法:本研究首先应用凝集素芯片,对HV、BB、BC-Ⅰ以及BC-Ⅱ混合唾液样本(各30例混合)中的蛋白质糖链结构进行分析,比较四组混合唾液中蛋白质糖链结构的差异,并通过凝集素印迹实验验证;随后,应用凝集素芯片对新收集的139例训练集样本(36例HV,35例BB,36例BC-Ⅰ以及32例BC-Ⅱ)进行个例分析,筛选可用于鉴别诊断BB、BC-Ⅰ和BC-Ⅱ唾液样本的候选凝集素,并利用二元逐步Logistic回归分析构建诊断模型;最后,在127例验证集样本(31例HV,30例BB,30 例 BC-Ⅰ 以及 36 例 BC-Ⅱ)中通过 ROC 曲线分析(receiver operating characteristic curve)评估诊断模型对BB、BC-Ⅰ和BC-Ⅱ的鉴别能力。实验结果:1.混合唾液样本凝集素芯片结果表明,11种凝集素的归一化荧光强度值(normalized fluorescent intensities,NFIs)在 HV、BB、BC-Ⅰ 以及 BC-Ⅱ 混合样本中具有显著差异:其中5种凝集素(DBA、PNA、PHA-E+L、UEA-I和PWM)特异性识别的 αGalNAc、GalNAcαl-3(Fucαl-2)Gal、Galβl-3GalNAcα、平分型 GlcNAc、多天线型N-聚糖、Fucαl-2Ga1β1-4Glc(NAc)以及分支(LacNAc)n糖链在至少一种乳腺疾病患者唾液中较HV表达水平发生改变。另外6种凝集素(MAL-Ⅰ、ECA、NPA、BPL、PTL-Ⅱ 和 BS-Ⅰ)特异性识别的 Galβl-4/3GlcNAc、High-Man、Manαl-6Man、Ga1β1-3GalNAcα-Ser/Thr 以及 Galαl-3/6Gal/Glc 糖链结构表达水平在 BB、BC-Ⅰ或BC-Ⅱ混合唾液样本中具有显著性差异。随后利用4种凝集素(PHA-E+L、NPA、MAL-Ⅰ以及BS-Ⅰ)进行凝集素印迹实验,证明了混合唾液样本芯片结果的准确性。2.经过训练集个例唾液凝集芯片进一步筛选,11种凝集素中的8种凝集素(PNA、PHA-E+L、UEA-Ⅰ、PWM、MAL-Ⅰ、NPA、BS-Ⅰ 和 PTL-Ⅱ)以及 PHA-E 作为候选凝集素进行ROC曲线分析。结果发现仅有PHA-E+L和PWM具有较高的鉴别诊断能力(ROC曲线线下面积(AUC)0.70),表明单一凝集素无法应用于BB、BC-I和BC-Ⅱ样本的鉴别。因此基于9种候选凝集素,通过Logistic回归构建4个诊断模型:Model BD、Model BB、Model BC-Ⅰ及Model BC-Ⅱ用于鉴别不同乳腺疾病患者。经验证集个例样本验证,Model BD获得了较好的诊断效果(AUC:0.902、灵敏度:0.823、特异性:0.839),能够准确鉴别出96例乳腺疾病患者中的79例,31例HV中的26例;Model BB(AUC:0.796、灵敏度:0.727、特异性:0.767)可准确鉴别出30例BB中的23例,66例乳腺癌患者中的48例;此外,Model BC-Ⅰ(AUC:0.781、灵敏度:0.700、特异性:0.864)可准确鉴别出30例BC-Ⅰ组中的21例,其余两个患病组66例样本中的57例;而Model BC-Ⅱ(AUC:0.759、灵敏度:0.861、特异性:0.550)可准确鉴别出36例BC-Ⅱ组中的31例,其余两个患者组60例样本中的33例。验证集中诊断模型的ROC-AUC值均高于0.75,仍能具有较高的鉴别诊断能力,表明乳腺癌患者唾液中蛋白质的异常糖链结构有望成为潜在的体外检测标志物,用于乳腺癌患者的早期筛查。
[Abstract]:Background: Breast cancer is one of the most common malignancies in women worldwide. At present, there is no ideal in vitro molecular marker for early diagnosis of breast cancer. Starting with saliva detection, lectin chip, a highly sensitive and high throughput sugar chain detection technique, was used to detect healthy volunteers (HV), benign breast cancer patients (BB), patients with stage I breast cancer in stage I, BC-I, and stage II breast cancer. The glycan structure of salivary protein in patients with breast cancer in stage II (BC-II) was compared and analyzed. The abnormal glycan structure of salivary protein was expected to be found and the possibility of early diagnosis of breast cancer as a detection index was discussed. The structure of protein glycan in mixed saliva samples (30 cases each) was analyzed, and the difference of protein glycan in mixed saliva was compared among the four groups, which was verified by agglutinin imprinting test. Then, 139 new training samples (36 HV, 35 BB, 36 BC-I and 32 BC-II) were collected by agglutinin chip. A case study was conducted to screen candidate lectins for differential diagnosis of BB, BC-I and BC-II saliva samples, and a diagnostic model was constructed using binary stepwise logistic regression analysis. Finally, ROC curve analysis was performed in 127 validation samples (31 HV, 30 BB, 30 BC-I and 36 BC-II). The results showed that: 1. The normalized fluorescent intensities (NFIs) of 11 kinds of agglutinin were significantly different in the mixed samples of HV, BB, BC-I and BC-II. Five kinds of agglutinin (DBA, PNA, PHA-E) +L, UEA-I, and PWM-specific alpha GalNAc, GalNAc alpha L-3 (Fuc alpha l-2) Gal, Gal beta l-3Gal NAc alpha, bisected GlcNAc, multi-antenna N-glycans, Fuc alpha l-2Ga1 beta 1-4Glc (NAc) and branched (LacNAc) n chains were altered in saliva of at least one breast disease compared with HV expression levels. Six other agglutinins (MAL-I, ECA, BPL, PTL-I I, NPA, and LacNAC) were also altered. S-I-specific recognised levels of Gal-beta-l-4/3G LCNAc, High-Man, Man-alpha-l-6Man, Ga-1-beta-3G alNAc alpha-Ser/Thr and Gal-alpha-l-3/6Gal/Glc sugar chains were significantly different in BB, BC-I or BC-II mixed saliva samples. Four kinds of lectins (PHA-E+L, NPA, MAL-I and BS-I) were subsequently used for lectin imprinting experiments to verify the results. Accuracy of mixed saliva sample microarray results. 2. After further screening of training samples, eight lectins (PNA, PHA-E+L, UEA-I, PWM, MAL-I, NPA, BS-I and PTL-II) and PHA-E as candidate lectins were analyzed by ROC curves. Only PHA-E+L and PWM were found to be highly differentiated. Diagnostic ability (AUC) 0.70 indicated that single lectin could not be used to differentiate BB, BC-I and BC-I I samples. Therefore, based on nine candidate lectins, four diagnostic models were constructed by logistic regression: Model BD, Model BB, Model BC-I and Model BC-I I for differentiating patients with different breast diseases. Model BD (AUC: 0.902, sensitivity: 0.823, specificity: 0.839) can accurately identify 79 out of 96 patients with breast diseases, 26 out of 31 patients with HV; Model BB (AUC: 0.796, sensitivity: 0.727, specificity: 0.767) can accurately identify 23 out of 30 BBs, 48 out of 66 breast cancer patients; In addition, BC-BB (model: 0.796, sensitivity: 0.727, specificity: 0.767) can accurately identify 23 out of 30 BBs and 48 out of 66 breast cancer patients. I (AUC: 0.781, sensitivity: 0.700, specificity: 0.864) could accurately identify 21 out of 30 BC-I patients, 57 out of 66 samples from the other two groups, while Model BC-II (AUC: 0.759, sensitivity: 0.861, specificity: 0.550) could accurately identify 31 out of 36 BC-II patients and 33 out of 60 samples from the other two groups. The ROC-AUC values of the model were higher than 0.75, and the model still had high differential diagnostic ability, indicating that the abnormal sugar chain structure of protein in saliva of breast cancer patients may be potential in vitro detection markers for early screening of breast cancer patients.
【学位授予单位】：西北大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：R737.9

【相似文献】