【作者】 高碧霞；

【导师】 李学旺； 李明喜；

【作者基本信息】 北京协和医学院 ， 内科学， 2011， 博士

【摘要】 目的优化弱阳离子交换磁珠(MB-WCX)联合基质辅助激光解析电离飞行时间质谱(MALDI-TOF-MS)建立尿液多肽谱的实验方法,建立代谢综合征(MS)早期肾损害尿液多肽谱。方法(1)探讨尿液采集方式、溶解温度、pH值、样本上样量、样品靶和质谱数据采集方式对磁珠联合基质辅助激光解析电离飞行时间质谱(MB-MALDI-TOF-MS)建立尿液多肽谱的影响,评估实验方法的重复性；(2)尿液样本来源于2008～2009年北京平谷地区“代谢综合征肾脏损害”流行病学研究。MS按美国国家胆固醇教育计划的成人治疗专家组Ⅲ(ATPⅢ)诊断标准,MS患者20μg/min≤尿白蛋白排泄率(UAE)<200μg/min和估算肾小球滤过率(eGFR)≥60ml/min.1.73m2则为早期肾损害。入选者分为健康对照(组Ⅰ)、MS合并正常白蛋白尿组(组Ⅱ)和MS合并微量白蛋白尿组(组Ⅲ)。应用MB-MALDI-TOF-MS方法建立三组样本尿液多肽谱。结果(1)MB-MALDI-TOF-MS建立尿液多肽谱的优化实验流程包括：收集过夜段尿标本、常温溶解、上样量30μL、点靶Polished steel target及手动模式采集数据；该实验方法日内变异系数为7.7%～14.2%,日间变异系数为7.9%～23.0%。(2)165例入选者中组Ⅰ65例、组Ⅱ54例、组Ⅲ46例,采用优化的MB-MALDI-TOF-MS建立了三组尿液多肽谱。结论本实验建立了灵敏度较高、稳定性较好的MB-MALDI-TOF-MS实验流程,适合高通量的临床蛋白质组学研究。应用该技术建立了正常人、MS合并正常白蛋白尿及MS合并微量白蛋白尿的尿液多肽谱。目的应用生物信息学方法筛选差异多肽峰并建立MS早期肾损害的尿液诊断模型,对尿液多肽标志物进行序列鉴定。方法两种方法进行数据分析：(1)三组样本分别分为training组和testing组,ClinProTools 2.1中统计学方法筛选差异多肽峰,遗传算法(GA)分别对组Ⅰ和组Ⅲ、组Ⅱ和组Ⅲ的training组数据构建诊断模型,采用10倍交叉验证评估模型的诊断能力；用testing组数据进行外部验证评估模型的预测能力；(2)Matlab7.10.0中的随机森林(RF)算法筛选差异多肽峰,支持向量机(SVM)算法对组Ⅰ和组Ⅲ、组Ⅱ和组Ⅲ及三组数据分别构建诊断模型,采用10倍交叉验证和ROC曲线评估模型的诊断能力。应用线性离子阱静电场轨道阱质谱仪(LTQ Orbitrap Velos)对尿液差异多肽峰分别进行鉴定,分析多肽标志物的生物学功能。结果(1)组Ⅰ和组Ⅲ多肽谱比较,GA算法构建诊断模型10倍交叉验证敏感性100%、特异性92.1%、准确性95.9%,外部验证敏感性76.2%、特异性80%、准确性78.4%；SVM算法构建模型的敏感性82.0%、特异性90.9%、准确性87.3%, ROC曲线下面积(AUC)为0.924。两个模型共同包含四个多肽峰：m/z 2755.97、3016.72、9076.41和11728.45；(2)组Ⅱ和组Ⅲ多肽谱比较,GA算法构建诊断模型10倍交叉验证敏感性100%、特异性87.5%、准确性93.4%,外部验证敏感性71.4%、特异性73.1%、准确性72.3%；SVM算法构建模型的敏感性89.2%、特异性81.1%、准确性85.5%,AUC为0.911。两个模型共同包含四个多肽峰：m/z 2755.97、3016.72、9076.41、10052.09；(3)三组多肽谱比较,SVM算法构建诊断模型准确性64.5%,包含8个多肽峰：m/z 2048.72、2562.67、2755.97、8779.30、9076.41、10052.09、10530.43和11728.45。(4)三个多肽峰m/z 1884.33、2562.67和2661.41均为纤维蛋白原α链的肽段。m/z 2562.67在组Ⅱ和组Ⅲ表达上调,m/z 1884.33和2661.41在组Ⅱ表达上调。结论通过两种生物信息学方法构建的模型均具有较好的诊断能力。鉴定得到了差异肽段m/z 1884.33、2562.67和2661.41的序列并鉴定为Fibrinogen alpha chain的多肽片段,可能是MS早期肾脏损害尿液多肽标志物,参与了MS及MS肾脏损伤的致病过程。更多还原

【Abstract】 Objective To optimize the peptidome analysis of magnetic bead-based weak cation exchange chromatography (MB-WCX) based matrix-assisted laser desorption/ionization time-of-flight mass spectrometry (MALDI-TOF-MS) method and to generate urine peptidome profiling of metabolic syndrome with early renal injury. Methods (1) Six preanalytical variables (urine collection methods, urine thawing temperature, pH values, urinary protein concentration, MALDI-TOF targets and spectra acquisition modes) were investigated on the influence of urine peptidome profiling, and the precision of this experiment method was evaluated. (2) Urine samples were collected from epidemiologic study of MS and renal involvement in Pinggu district, Beijing during the period of 2008 and 2009. MS was diagnosed by ATPⅢcriteria, while MS patients with early renal injury were defined as 20μg/min≤urinary albumin excretion(UAE)<200μg/min and estimated glomerμLar filtration rate (eGFR)≥60ml/min.1.73m2.Participants were divided into three group:group I (healthy subjects), group II (MS patients with normoalbuminuria) and group III (MS patients with microalbuminuria). ResμLts (1) An optimized method for urine peptidome profiling by MB-MALDI-TOF-MS included:overnight urine collection, thawing at room temperature, applying 30μL urine per sample, using polished steel target and acquire spectra data by manual mode. Within-day and between-day coefficient of variation (CV) ranged from 7.7% to 14.2% and from 7.9% to 23.0% respectively. (2) One hundred and sixty-five subjects were enrolled into our study(sixty five subjects in groupⅠ, fifty-four subjects in group II and forty-six subjects in groupⅢ) and their urinary peptidome spectra were generated separately by the optimized MB-MALDI-TOF-MS method. Conclusion We have established a high accurate and reproducible analytic platform for urine peptidome profiling, it is suitable for high-throughput clinical proteomics research technology. Appling this approach, we established urinary peptidome patterns of healthy subjects, MS with normoalbuminuria and MS with microalbuminuria. Objective To explore potential urine biomarkers and to generate diagnostic models by bioinformatics tools, to identify urinary peptide biomarkers of MS with early renal injury. Methods Two bioinformatics software had been applied to analyze urinary peptidome profiling mass spectrometry data:(1) Subjects in groupⅠ, groupⅡand groupⅢwere divided into training set and testing set. Statistical tests in ClinProTools 2.1 software were adopted to screen differential peptide peaks of urinary peptidome in training set of groupⅠversus groupⅢ, groupⅡversus groupⅢand comparison of three groups separately. Genetic algorithm (GA) was used to establish diagnostic models of groupⅠversus groupⅢ, groupⅡversus groupⅢ.10-fold cross validation in training set was used to evaluate recognizability and external validation in testing set was used to assess prediction ability of diagnostic models. (2) Random forests (RF) algorithm in Matlab7.10.0 software was used to screen differential peptide peaks, then combined support vector machine (S VM) algorithm to generate diagnostic models of group I versus group III, groupⅡversus groupⅢand three groups comparison separately.10-fold cross validation and receiver operating characteristic curve (ROC) were used to evaluate recognizability of diagnostic models. Differential peptide peaks were identified by linear ion trap-orbitrap mass spectrometry (LTQ Orbitrap Velos) and biologic function of these identified peptide biomarkers were analyzed. ResμLts (1)GroupⅠversus Group III:GA based model showed 100% sensitivity,92.1% specificity and 95.9% accuracy by 10-fold cross-validation in training set in identifying MS patients with early renal injury, and it revealed 76.2% sensitivity, 80% specificity and 78.4% accuracy in testing set; Correspondingly, SVM algorithm based model reported 82.0% sensitivity,90.9% specificity and 87.3% accuracy. Area under curve(AUC) value of receiver operating characteristic curve (ROC curve) was 0.924. Four peptide peaks were included in two diagnostic models with m/z 2755.97、3016.72.9076.41 and 11728.45;(2)GroupⅡversus GroupⅢ:GA based model showed 100% sensitivity,87.5% specificity and 93.4% accuracy by 10-fold cross-validation in training set, and it revealed 71.4% sensitivity,73.1% specificity and 72.3% accuracy in testing set; Correspondingly, SVM algorithm based model reported 89.2% sensitivity,81.1% specificity and 85.5% accuracy, AUC value was 0.911. Four peptide peaks were included in two diagnostic models with m/z 2755.97, 3016.72,9076.41,10052.09;(3)Three groups comparison:SVM algorithm based model showed 64.5% overall accuracy, the model included eight peptide peak:m/z 2048.72,2562.67,2755.97,8779.30,9076.41,10052.09,10530.43 and 11728.45.(4) Three differential peaks of m/z 1884.33,2562.67 and 2661.41 were identified as peptide derived from fibrinogen alpha chain. Fragment of m/z 2562.67 was up-regμLated in urine in patients of MS and MS with early renal injury, while fragments of m/z 1884.33 and 2661.41 were up-regμLated in MS patients without renal injury. Conclusion Diagnostic models based on GA and SVM showed highly accuracy and class prediction in discriminating MS patients with early renal injury. Three peptide peaks identified based on MS/MS spectra were all peptide fragments of fibrinogen alpha chain, suggesting fibrinogen alpha chain might be urinary peptide biomarkers of MS with early renal injury and participated in pathogenesis of MS and MS with renal injury更多还原