February     Selected CNS literature

Large-scale proteomic analysis reveals new pathophysiological mechanisms and potential biomarkers of Parkinson's disease

Working together, Wei Cheng's team at Fudan University Shanghai Medical College and Jin-Tai Yu's team analysed Olink plasma protein data for 2,920 proteins from 51,804 participants in the UK Biobank and, over 14.45 years of follow-up, identified 859 new-onset Parkinson's disease (PD) cases. The study identified 38 proteins associated with PD and validated 6 of the top 10 in the Parkinson's Progression Markers Initiative (PPMI) cohort. In particular, ITGAV, HNMT and ITGAM showed significant and consistent associations (hazard ratios: 0.11 to 0.57; P values from 6.90×10^-24 to 2.10×10^-11). The study found that lipid metabolism dysfunction was already evident 15 years before PD onset, while the levels of BAG3, HPGDS, ITGAV and PEPD declined continuously before diagnosis; changes in these proteins were closely linked to prodromal symptoms and brain measures. Mendelian randomisation analysis suggested that ITGAM and EGFR may be risk factors for PD. Based on these findings, the team used machine learning to build a prediction model combining the top 16 protein markers with demographic information, achieving prediction within 5 years (AUC = 0.887) and long-term prediction beyond 5 years (AUC = 0.816), significantly outperforming models based on demographic information alone. The model was also validated in the independent PPMI cohort (AUC = 0.802). The study highlights the importance of early peripheral pathophysiological changes in PD, provides new insight for the development of early biomarkers and precision treatment strategies, and marks important progress in the early diagnosis and prediction of Parkinson's disease.

Lipoprotein metabolism is markedly altered in patients with chronic kidney disease (CKD), but the relationship between subfraction profiles and mortality in haemodialysis patients remains unclear. Xiaoqiang Ding's team and Xiaoyan Zhang's team from the Department of Nephrology, Zhongshan Hospital, Fudan University, together with Huiru Tang's team from the School of Life Sciences, Fudan University, performed NMR (nuclear magnetic resonance) metabolomic analysis on baseline plasma samples from 368 maintenance haemodialysis patients and explored the association of lipoprotein subfractions with all-cause and cardiovascular disease (CVD) mortality. Over a mean follow-up of 45.1 months, the study recorded 144 all-cause deaths and 67 CVD deaths. After adjusting for 14 covariates, 18 lipoprotein parameters were associated with all-cause mortality and 35 with CVD mortality. Notably, total low-density lipoprotein cholesterol (LDL-C) and total high-density lipoprotein cholesterol (HDL-C) showed no significant association with mortality, but specific lipoprotein subfractions did: triglycerides in large very-low-density lipoprotein (VLDL) particles were positively associated with all-cause mortality; lipids (triglycerides, cholesterol and phospholipids) in medium VLDL particles, cholesterol/total lipids in intermediate-density lipoprotein (IDL), and free cholesterol/total lipids in small HDL particles were positively associated with both all-cause and CVD mortality. Adding lipoprotein parameters to a conventional risk factor model significantly improved the accuracy of mortality risk prediction: the AUC for all-cause mortality rose from 0.811 to 0.842 (p = 0.020), and for CVD mortality from 0.806 to 0.854 (p = 0.005). The study shows that specific lipoprotein subfractions are closely associated with mortality risk in maintenance haemodialysis patients, and that prediction models based on these lipoproteins significantly outperform conventional models, providing an important basis for personalised treatment of CKD patients.

Arja M. Kullaa's team at the Institute of Dentistry, Faculty of Health Sciences, University of Eastern Finland, conducted a clinical longitudinal study assessing changes in the salivary metabolic profile of patients with head and neck cancer (HNC) before, during and one week after radiotherapy (RT), aiming to reveal the metabolite changes induced by radiotherapy and the associated oral biological pathways. Unstimulated whole mouth saliva samples (USWMS) were collected from 45 HNC patients and 30 healthy controls, and 24 metabolites were identified and quantified by NMR (nuclear magnetic resonance) spectroscopy. The results showed that 17 salivary metabolites declined continuously during and after treatment. Compared with healthy controls, HNC patients had lower salivary proline and higher fucose and 1,2-propanediol, suggesting disturbed redox balance and abnormal fucosylation. Spearman correlation analysis showed that the relationships between pyruvate and metabolites such as alanine, trimethylamine, choline, taurine and succinate changed significantly during radiotherapy. Pathway enrichment analysis (MetaboAnalyst 6.0) further revealed five key metabolic pathways significantly affected (pyruvate metabolism; glycolysis/gluconeogenesis; glycine, serine and threonine metabolism; glyoxylate and dicarboxylate metabolism; and alanine, aspartate and glutamate metabolism), with pyruvate metabolism scoring highest for pathway impact. The study shows that radiotherapy significantly alters the salivary metabolic profile of HNC patients, indicating disturbed redox homeostasis, increased oxidative stress and inflammatory response. Non-invasive monitoring of salivary metabolites opens a new direction for personalised medicine, with the prospect of improving treatment outcomes and quality of life through early detection of treatment-related toxicity, monitoring of treatment response and tailored interventions.

Tubulointerstitial fibrosis is a hallmark feature of chronic kidney disease (CKD), and metabolic disturbance plays a key role in the pathogenesis of CKD. Tyrone L. R. Humphries' team at the Kidney Disease Research Collaboration and the Institute for Molecular Bioscience, University of Queensland, Brisbane, Australia, investigated the chemical changes in the course of renal fibrosis using two-dimensional 1H-nuclear magnetic resonance (2D-COSY) metabolomics. The team established an in vitro TGF-β1-induced renal fibrosis model using human kidney 2 (HK-2) proximal tubular epithelial cells (PTECs) and validated the model by quantitative PCR and Western blotting. 2D-COSY analysis confirmed morphological and functional changes associated with fibrosis. The results showed that in TGF-β1-treated cells, the expression of fibronectin, type IV collagen, smooth muscle actin and oxidative stress enzymes increased significantly (p < 0.05). NMR metabolomics revealed changes in metabolites related to glycolysis and glutamine metabolism, including decreases in myo-inositol and choline, reduced metabolites of the oxidative phase of the pentose phosphate pathway, and increases in glucose and glucuronic acid. These changes suggest that metabolic reprogramming of PTEC cells may support the rapid fibrotic process and its high energy demands. The study shows that 2D-COSY metabolomics can effectively identify metabolite changes associated with renal fibrosis. These results will help develop fibrosis biomarker panels detectable by clinical magnetic resonance spectroscopy, providing new tools for the diagnosis and management of CKD, supporting early detection of disease progression and the design of personalised treatment plans, and ultimately improving treatment outcomes and quality of life for patients.

Liver cirrhosis is one of the leading causes of death in patients with chronic liver disease (CLD). Huang Kai's team from the Department of Cardiovascular Surgery, Sun Yat-sen Memorial Hospital, Sun Yat-sen University, together with Wei Ye's team from the Department of Gastroenterology, Hangzhou Hospital of Traditional Chinese Medicine affiliated to Zhejiang Chinese Medical University, used proton nuclear magnetic resonance (1H-NMR) serum metabolomics data from the UK Biobank (UKB) combined with an elastic net-regularised Cox proportional hazards model to explore the role of metabolomics in risk stratification for liver cirrhosis in CLD patients. The study combined metabolomics data with the aspartate aminotransferase to platelet ratio index (APRI) and the fibrosis-4 score (FIB-4) to build a liver cirrhosis risk prediction model, and evaluated its performance in derivation and validation cohorts. A total of 2,738 patients were enrolled, and multiple metabolites were independently associated with incident cirrhosis (after full adjustment, 21 of 168 metabolites were significant). The combination of metabolomics with FIB-4 significantly improved predictive performance (Harrell's C: 0.717 vs. 0.696, ΔC = 0.021, 95% CI 0.014–0.028; NRI: 0.504 [0.488–0.520]). Similarly, metabolomics combined with APRI outperformed APRI alone (Harrell's C: 0.747 vs. 0.718, ΔC = 0.029, 95% CI 0.022–0.035; NRI: 0.378 [0.366–0.389]). Key metabolites included branched-chain amino acids (BCAAs), lipids and oxidative stress markers, and pathway enrichment analysis revealed the central role of lipid and amino acid metabolism disturbance in the progression of cirrhosis. The study shows that 1H-NMR serum metabolomics significantly improves the prediction of cirrhosis risk. The APRI + metabolomics model showed strong discriminatory power, providing a powerful tool for early screening of cirrhosis risk. This approach not only helps identify high-risk patients at an early stage, it also provides an important basis for designing personalised treatment strategies and is expected to improve patient prognosis and quality of life.

Apostolia M. Tsimberidou's team at The University of Texas MD Anderson Cancer Center reviewed relevant studies to demonstrate the potential of artificial intelligence (AI) and machine learning (ML) in digital pathology, particularly in diagnosis, treatment and personalised management.

The following experiments were mainly introduced:

1. Using AI/ML technologies to analyse patient samples to improve diagnostic accuracy. For example, AI-assisted tools automate immunohistochemistry (IHC) scoring, reducing human error and improving the accuracy and consistency of diagnosis.

2. Using AI/ML technologies to predict biomarkers and molecular features. For example, convolutional neural network (CNN) models predict genetic mutations and microsatellite instability (MSI) status in tumour patients from H&E-stained whole-slide images.

3. Using AI/ML technologies to analyse patient samples to improve individualised treatment plans. For example, CNN models classify tumour and non-tumour cells to determine tumour burden more precisely and predict treatment response.

The study shows that AI and ML technologies hold great promise in cancer, but digital pathology faces challenges in data management and storage and in standardised practice. Future research needs to address these challenges in order to realise the full benefits of digital pathology.