Data Sheet 1_Multi-omics exploration of chaperone-mediated immune-proteostasis crosstalk in vascular dementia and identification of diagnostic biomarkers.xlsx

IntroductionVascular dementia (VaD), the second most prevalent form of dementia globally, remains insufficiently understood in terms of its molecular mechanisms and diagnostic biomarkers. This study aims to elucidate the regulatory network and diagnostic potential of the molecular chaperone system in VaD through the integration of multi-omics data and machine learning algorithms.
MethodsTranscriptomic data from frontal and temporal cortex (GSE122063, n=15)and white matter (GSE282111, n=8) samples were obtained from the GEO database. Differentially expressed genes (DEGs) were identified using the limma package (log2FC>0.656, p<0.05). Protein-protein interaction (PPI) networks were constructed using the STRING database. Biomarker validation was performed through cross-validation using LASSO, SVM-RFE, and Random Forest algorithms. Immune microenvironment analysis was conducted using CIBERSORT, while single-cell transcriptomics was analyzed within the Seurat framework.
ResultsA total of 897 DEGs were identified, with functional enrichment analysis revealing significant involvement in T cell activation (p=2.84×10-3), neuroactive ligand-receptor interaction (p=6.01×10-4), and osteoclast differentiation (NES=2.83). PPI network analysis identified HSP90AA1, HSPA1B, and DNAJB1 as core hub genes (degree centrality >20). Machine learning validation demonstrated their combined exceptional diagnostic efficacy (AUC=0.963, F1 = 0.88). Immune analysis revealed that this molecular chaperone axis modulates neuroinflammation by suppressing naive B cell differentiation (61% reduction) and activating Tregs (55.53% increase). Single-cell resolution analysis showed HSP90AA1 to be specifically overexpressed in oligodendrocytes (72.23%), significantly correlating with glial depletion (4.56% decrease in oligodendrocytes, p<0.01) and aberrant neuronal proliferation (144.23% increase, p=0.0032). In vivo experiments utilized a bilateral common carotid artery stenosis (BCAS) mouse model to simulate human vascular dementia (VaD), with further validation through Morris water maze testing. The BCAS group exhibited significantly upregulated mRNA expression of HSP90AA1, HSPA1B, and DNAJB1, consistent with integrated bioinformatics analysis results.
ConclusionThis study elucidates the HSP90AA1-HSPA1B-DNAJB1 network as a key driver of VaD pathogenesis through dual mechanisms of protein homeostasis and immune reprogramming. The diagnostic performance of this network significantly surpasses traditional biomarkers (ΔAUC≥14.3%), offering novel targets for precision diagnostics and therapeutics. However, further validation with larger cohorts is necessary to assess its clinical translational potential.

引言：血管性痴呆（Vascular dementia, VaD）是全球范围内第二大常见的痴呆类型，但其分子机制与诊断生物标志物仍未得到充分阐明。本研究旨在通过多组学数据整合与机器学习算法，阐明VaD中分子伴侣系统的调控网络及其诊断潜力。
方法：从GEO数据库获取了额叶和颞叶皮层（GSE122063，n=15）以及脑白质（GSE282111，n=8）样本的转录组数据。使用limma包（log2FC>0.656，p<0.05）筛选差异表达基因（Differentially expressed genes, DEGs）；依托STRING数据库构建蛋白质相互作用（Protein-protein interaction, PPI）网络。采用最小绝对收缩与选择算子（Least Absolute Shrinkage and Selection Operator, LASSO）、支持向量机递归特征消除（Support Vector Machine-Recursive Feature Elimination, SVM-RFE）以及随机森林（Random Forest）算法开展交叉验证，以完成生物标志物验证；使用CIBERSORT进行免疫微环境分析，并依托Seurat框架完成单细胞转录组学分析。
结果：共筛选得到897个差异表达基因，功能富集分析显示其显著富集于T细胞激活（p=2.84×10^-3）、神经活性配体-受体相互作用（p=6.01×10^-4）以及破骨细胞分化（归一化富集得分，NES=2.83）通路。蛋白质相互作用网络分析筛选出HSP90AA1、HSPA1B与DNAJB1作为核心枢纽基因（度中心性>20）。机器学习验证结果显示，三者联合具有优异的诊断效能（受试者工作特征曲线下面积（Area Under Curve, AUC）=0.963；F1分数=0.88）。免疫分析表明，该分子伴侣轴通过抑制初始B细胞分化（水平降低61%）并激活调节性T细胞（Regulatory T cells, Tregs，占比提升55.53%）来调控神经炎症。单细胞分辨率分析显示，HSP90AA1在寡突胶质细胞中特异性高表达（占比72.23%），且其表达与神经胶质细胞耗竭（寡突胶质细胞占比减少4.56%，p<0.01）以及神经元异常增殖（增殖水平提升144.23%，p=0.0032）显著相关。体内实验采用双侧颈总动脉狭窄（Bilateral Common Carotid Artery Stenosis, BCAS）小鼠模型模拟人类血管性痴呆（VaD），并通过莫里斯水迷宫实验完成进一步验证；BCAS模型组小鼠的HSP90AA1、HSPA1B与DNAJB1的mRNA表达水平显著上调，与整合生物信息学分析结果一致。
结论：本研究阐明，HSP90AA1-HSPA1B-DNAJB1调控轴通过蛋白质稳态与免疫重编程双重机制，成为VaD发病机制的关键驱动因素。该调控轴的诊断性能显著优于传统生物标志物（ΔAUC≥14.3%），为精准诊断与治疗提供了全新靶点。不过，仍需通过更大规模的队列研究开展进一步验证，以评估其临床转化潜力。