Pathophysiological evolutions in early-stage Alzheimer’s disease (AD) are not well understood. We used data of 2923 Olink plasma proteins from 51,296 non-demented middle-aged adults. During a follow-up of 15 years, 689 incident AD cases occurred. Cox-proportional hazard models were applied to identify AD-associated proteins in different time intervals. Through linking to protein categories, changing sequences of protein z-scores can reflect pathophysiological evolutions. Mendelian randomization using blood protein quantitative loci data provided causal evidence for potentially druggable proteins. We identified 48 AD-related proteins, with CEND1, GFAP, NEFL, and SYT1 being top hits in both near-term (HR:1.15–1.77; P:9.11 × 10⁻⁶⁵–2.78 × 10⁻⁶) and long-term AD risk (HR:1.20-1.54; P:2.43 × 10⁻²¹–3.95 × 10⁻⁶). These four proteins increased 15 years before AD diagnosis and progressively escalated, indicating early and sustained dysfunction in synapse and neurons. Proteins related to extracellular matrix organization, apoptosis, innate immunity, coagulation, and lipid homeostasis showed early disturbances, followed by malfunctions in metabolism, adaptive immunity, and final synaptic and neuronal loss. Combining CEND1, GFAP, NEFL, and SYT1 with demographics generated desirable predictions for 10-year (AUC = 0.901) and over-10-year AD (AUC = 0.864), comparable to full model. Mendelian randomization supports potential genetic link between CEND1, SYT1, and AD as outcome. Our findings highlight the importance of exploring the pathophysiological evolutions in early stages of AD, which is essential for the development of early biomarkers and precision therapeutics.

早期阿尔茨海默病 （AD） 的病理生理学演变尚不清楚。我们使用了来自 51,296 名非痴呆中年人的 2923 个 Olink 血浆蛋白的数据。在 15 年的随访期间，发生了 689 例 AD 事件。应用 Cox 比例风险模型来识别不同时间间隔的 AD 相关蛋白。通过与蛋白质类别相关联，蛋白质 z 分数的变化序列可以反映病理生理学演变。使用血液蛋白定量位点数据的孟德尔随机化为潜在的可成药蛋白质提供了因果证据。我们鉴定了 48 种 AD 相关蛋白，其中 CEND1、GFAP、NEFL 和 SYT1 是近期的热门化合物 （HR：1.15–1.77;P：9.11 × ^10-65-2.78 × ^10-6）和长期 AD 风险 （HR：1.20-1.54;P：2.43 × ^10-21-3.95 × ^10-6）。这四种蛋白在 AD 诊断前 15 年增加并逐渐增加，表明突触和神经元早期和持续功能障碍。与细胞外基质组织、细胞凋亡、先天免疫、凝血和脂质稳态相关的蛋白质表现出早期紊乱，随后出现代谢功能障碍、适应性免疫以及最终突触和神经元丢失。将 CEND1、GFAP、NEFL 和 SYT1 与人口统计学相结合，产生了对 10 年 （AUC = 0.901） 和 10 年以上 AD （AUC = 0.864） 的理想预测，与完整模型相当。孟德尔随机化支持 CEND1、SYT1 和 AD 之间的潜在遗传联系作为结局。我们的研究结果强调了探索 AD 早期病理生理演变的重要性，这对于早期生物标志物和精准治疗的开发至关重要。