目的

利用粪便代谢组学研究人胎盘来源间充质干细胞（hPMSCs）治疗对克罗恩病模型小鼠的代谢影响并筛选疗效评估生物标志物。

方法

将25 只自发产生慢性回肠炎的SAMP1/Yit 小鼠作为克罗恩病模型，分为治疗组（15 只）和疾病模型组（10 只）；另将10 只AKR/J 小鼠作为健康对照组。治疗组腹腔注射hPMSCs，疾病模型组及健康对照组腹腔注射等体积磷酸盐缓冲液。在治疗的第0、3、7、14 天，分别取三组小鼠的粪便样本，使用化学同位素标记液相色谱-质谱法（CIL LC-MS）进行代谢组学分析。

结果

CIL LC-MS 检测到了3 504 个峰对，其中77.77%可被鉴定或推定。根据代谢物质谱数据进行偏最小二乘判别分析发现，克罗恩病小鼠的代谢紊乱在治疗后得到了一定程度的恢复。对各组小鼠代谢物丰度进行统计分析发现，共有35 个代谢物在疾病模型组与健康对照组之间、以及治疗组与疾病模型组之间均发生了显著变化（P<0.05，倍数变化>1.2）。逐步优化标志物组合，通过正交偏最小二乘判别分析和基于逻辑回归算法的机器学习模型，最终确定了5 个具有最佳分类性能和生物学意义的标志物，即5-羟基吲哚丙酮酸、苏氨酰脯氨酸、对-辛弗林、4-(2-氨基苯基)-2,4-二氧代丁酸异构体1 和3-羟基-L-脯氨酸。这些标志物组合的曲线下面积达0.920，准确率为87.0%。

结论

基于CIL LC-MS 的粪便代谢组学分析筛选出5 个代谢物作为生物标志物组合，可以较高准确率评估hPMSCs 治疗克罗恩病小鼠模型的疗效。

Objective

To investigate the metabolic alterations of human placental mesenchymal stem cell (hPMSC) therapy on Crohn's disease model mice based on fecal metabolomics, and to screen biomarkers related to efficacy evaluation.

Methods

A total of 25 SAMP1 / Yit mice with spontaneous chronic ileitis were used as Crohn's disease models and divided into a treatment group (n= 15) and a disease group (n = 10). Additionally, 10 AKR /J mice were used as healthy controls. Mice in the treatment group were injected intraperitoneally with hPMSC, and mice in the disease group and the healthy control group were injected intraperitoneally with an equal volume of phosphate buffered saline at the same time points. On the 0th, 3rd, 7th, and 14th days after treatment, fecal samples of mice in the three groups were obtained, and chemical isotope labeling liquid chromatography-mass spectrometry (CIL LC-MS) was used for metabolomic analysis.

Results

The fecal metabolomic analysis based on CIL LC-MS detected 3 504 mass spectral peak pairs, of which 77.77% could be positively or putatively identified. Partial least squares-discriminant analysis (PLS-DA) was performed according to the mass spectrometry data of the metabolites, and it was found that the metabolic disorders of Crohn's disease mice were restored to a certain extent after treatment. Statistical analysis of the metabolite abundance of mice in each group found that a total of 35 metabolites had significant changes between the disease group and the healthy control group, and between the treatment group and the disease group (P < 0.05, fold change > 1.2). Through stepwise optimization of biomarker combinations, five biomarkers with the best classification performance and biological significance were finally determined by the orthogonal partial least squares-discriminant analysis(OPLS-DA) and machine learning model based on logistic regression algorithm, namely 5-Hydroxyindolepyruvate, p-Synephrine, Isomer 1 of 4-(2-Aminophenyl)-2,4-dioxobutanoic acid, LMetanephrine, and Threonylproline. The combination of these biomarkers achieved an area under the receiver operating characteristic curve of 0.920 with an accuracy of 87.0%.

Conclusion

Five metabolites identified through CIL LC-MS-based fecal metabolomic analysis can serve as a biomarker panel to assess the efficacy of hPMSC in the treatment of Crohn's disease with high accuracy.