Table_1_Metabolomic biomarkers in autism: identification of complex dysregulations of cellular bioenergetics.xlsx

Autism Spectrum Disorder (ASD or autism) is a phenotypically and etiologically heterogeneous condition. Identifying biomarkers of clinically significant metabolic subtypes of autism could improve understanding of its underlying pathophysiology and potentially lead to more targeted interventions. We hypothesized that the application of metabolite-based biomarker techniques using decision thresholds derived from quantitative measurements could identify autism-associated subpopulations. Metabolomic profiling was carried out in a case–control study of 499 autistic and 209 typically developing (TYP) children, ages 18–48 months, enrolled in the Children’s Autism Metabolome Project (CAMP; ClinicalTrials.gov Identifier: NCT02548442). Fifty-four metabolites, associated with amino acid, organic acid, acylcarnitine and purine metabolism as well as microbiome-associated metabolites, were quantified using liquid chromatography-tandem mass spectrometry. Using quantitative thresholds, the concentrations of 4 metabolites and 149 ratios of metabolites were identified as biomarkers, each identifying subpopulations of 4.5–11% of the CAMP autistic population. A subset of 42 biomarkers could identify CAMP autistic individuals with 72% sensitivity and 90% specificity. Many participants were identified by several metabolic biomarkers. Using hierarchical clustering, 30 clusters of biomarkers were created based on participants’ biomarker profiles. Metabolic changes associated with the clusters suggest that altered regulation of cellular metabolism, especially of mitochondrial bioenergetics, were common metabolic phenotypes in this cohort of autistic participants. Autism severity and cognitive and developmental impairment were associated with increased lactate, many lactate containing ratios, and the number of biomarker clusters a participant displayed. These studies provide evidence that metabolic phenotyping is feasible and that defined autistic subgroups can lead to enhanced understanding of the underlying pathophysiology and potentially suggest pathways for targeted metabolic treatments.

自闭症谱系障碍（Autism Spectrum Disorder，ASD，亦称自闭症）是一种表型与病因均具有异质性的疾病。识别自闭症具有临床意义的代谢亚型的生物标志物，有助于加深对其潜在病理生理学机制的理解，并有望催生更具针对性的干预方案。本研究提出假说：采用基于定量测量结果推导得出的判定阈值的代谢物生物标志物分析技术，可识别与自闭症相关的亚群。本研究依托儿童自闭症代谢组计划（Children’s Autism Metabolome Project，CAMP；ClinicalTrials.gov标识符：NCT02548442）开展病例对照研究，共纳入499名自闭症儿童与209名典型发育（TYP）儿童（年龄介于18至48个月之间），并对所有受试者完成代谢组谱分析。本研究采用液相色谱-串联质谱法（liquid chromatography-tandem mass spectrometry），对54种代谢物进行定量检测，这些代谢物涵盖氨基酸代谢、有机酸代谢、酰基肉碱代谢、嘌呤代谢以及微生物组相关代谢物类别。通过定量阈值筛选，本研究鉴定出4种代谢物浓度与149种代谢物比值作为生物标志物，每种标志物均可识别CAMP研究队列中4.5%~11%的自闭症亚群。其中42种生物标志物组成的子集可用于甄别CAMP队列中的自闭症个体，其诊断灵敏度为72%，特异度为90%。多名受试者可被多种代谢生物标志物共同识别。研究基于受试者的生物标志物谱采用层级聚类算法，将生物标志物划分为30个聚类簇。与各聚类簇关联的代谢变化提示，细胞代谢调控异常（尤其是线粒体生物能学调控异常）是本队列自闭症受试者中普遍存在的代谢表型。自闭症严重程度、认知与发育损伤状况，与乳酸水平升高、多种含乳酸的代谢比值以及受试者所涉及的生物标志物聚类簇数量显著相关。本研究证实代谢表型分型具备可行性，明确划分的自闭症亚群有助于深化对其潜在病理生理学机制的理解，并可为靶向代谢治疗提供潜在通路方向。