基于NMR的代谢组学技术预测大鼠脓毒症预后的准确性

Accuracy of nuclear magnetic resonance spectroscopy-based metabonomic technique in predicting prognosis in septic rats

目的评价基于核磁共振波谱仪（NMR）的代谢组学技术预测大鼠脓毒症预后的准确性。方法实验一取48只雄性sD大鼠，周龄6～8周，体重220～250g，随机分为3组：对照组（C组，n=8）、假手术组（SI组，n=8）和脓毒症组（CLPI组，n=32）。采用盲肠结扎穿孔法制备大鼠脓毒症模型。C组和SI组于术后12h时，CLP组于术后6、12、18、24h时各取8只大鼠，采集颈内动脉血样3ml，进行血气分析，并测定肝、肾功能；然后放血处死大鼠，取左下肺组织和左肾组织，光镜下观察病理学结果。实验二另取65只雄性sD大鼠，周龄6～8周，体重220—250g，随机分为2组：假手术组（SⅡ组，n=20）和CLPⅡ组（n=45）。采用盲肠结扎穿孔法制备大鼠脓毒症模型。术后12h经尾静脉采血样1．2ml，取血清待行^1H NMR分析。根据6d内生存状况，将CLPⅡ组大鼠分为3个亚组：生存组（存活期〉6d）、死亡组（存活期24h-6d）和废弃组（存活期〈24h）。采用κ-最近邻算法（κ-NN）和径向基函数神经网络算法（RBFNN）预测脓毒症预后。结果正交偏最小二乘-判别分析法可完全区分3组大鼠的生理特征。与脓毒症预后有关的6个标志物为8-羟丁酸、乳酸、丙氨酸、醋酸和乙酰乙酸和甲酸。采用κ-NN和RBFNN均可早期预测脓毒症大鼠预后，RBFNN预测效能优于κ-NN（P〈0．05）。结论脓毒症后在各器官发生轻微损伤时，机体糖、蛋白质、脂肪和核酸代谢异常，利用基于NMR的代谢组学技术可早期、有效地预测脓毒症的预后。

Objective To evaluate the accuracy of metabonomic technique based on nuclear magnetic resonance （NMR） spectroscopy in predicting the prognosis in septic rats. Methods Male SD rats aged 6-8 weeks weighing 220-250 g were used in this study. The experiment was performed in 2 parts. In part Ⅰ 48 male SD rats were randomly divided into 3 groups ： group Ⅰ control （C, n = 8） ; group Ⅱ sham operation （ S, n = 8） ; group Ⅲ sepsis （CLP, n = 32）. The animals were anesthetized with intraperitoneal 10% chloral hydrate 1 ml. Sepsis was produced by cecum ligation and puncture （CLP）. Arterial blood samples were obtained at 6, 12, 18, 24 h （ n = 8 each） after CLP for blood gas analysis and liver and kidney function tests. The animals were then killed and the left lung and kidney were removed for microscopic examination. In part H 65 male SD rats were randomly divided into 2 groups ： group Ⅰ sham operation （ S, n = 20） and group Ⅱ CLP （ n = 45）. Group Ⅲ was further divided into 2 subgroups according to the survival time after CLP： subgroup Ⅰ survival time 〉 6 days （CLPI） and subgroup Ⅱ survival time between 24 h-6 d. The animals were excluded if survival time after CLP was less than 24 h. Venous blood samples were obtained at 12 h after CLP for measurement of serum metabolites, κ-Nearest neighbar （κ-NN） and radial basis function neural net work （RBFNN） were used to predict prognosis of sepsis. Results Orthogonal partial least square discriminant analysis （O-PLS-DA） showed clustering according to groups indicating that NMRspectroscopy-based metabonomic technique could reveal pathologic characteristics of different groups. There were significant changes in six markers including lactate, alanine, acetate, acetoacetate, β-hydroxybutyrate and formate in septic rats. Both κ-NN and RBFNN could predict the prognosis in septic rats. RBFNN was superior to κ-NN in the prediction of prognosis. Conclusion NMR spectroscopy-based metabonomic approach combined with pattern recognition permits accurate prediction of the outcome of septic rats in the early stage when organs are slightly damaged.