Addressing inter-individual variability in CSF levels of brain-derived proteins across neurodegenerative diseases

Background Accurate diagnosis and monitoring of neurodegenerative diseases requires reliable biomarkers. Cerebrospinal fluid (CSF) proteins hold promise for reflecting brain pathology, yet their utility may be compromised by natural variability between individuals, weakening their association with disease and impacting their diagnostic accuracy.

Method We measured the levels of 69 proteins in CSF from 499 individuals using an antibody-based suspension bead array technology. In this multi-disease cohort including participants with Alzheimer"s disease, behavioural variant frontotemporal dementia, primary progressive aphasias, amyotrophic lateral sclerosis, corticobasal syndrome, primary supranuclear palsy and healthy controls, we investigated inter-individual variability in CSF protein levels. Further, we explored protein profiles across the diseases, after adjusting for this variability using linear modelling.

Results Correlation and PCA analysis on the measured CSF proteins revealed the presence of inter-individual variability in overall CSF levels of brain-derived proteins, which could not be explained by disease associations. Using linear modelling, we show that adjusting for median CSF levels of brain-derived proteins increases the diagnostic accuracy of many proteins, including those previously identified as altered in CSF in the context of neurodegenerative disorders. We further demonstrate a simplified approach for the adjustment using pairs of correlated proteins with opposite alteration in the diseases. With this approach, the proteins adjust for each other and further increase the biomarker performance through additive effect. Moreover, we find that many CSF proteins show similar alteration across the studied diseases, indicating that these proteins likely reflect shared processes.

Conclusion CSF protein profiles are affected by inter-individual variability which partially hinders the diagnostic potential of CSF proteins. Thus, we propose that this variability should be accounted for in future studies. Further, we demonstrate high overlap in CSF protein patterns across neurodegenerative diseases, highlighting the need for multi-disease studies in order to identify biomarkers that are specific to or common between different diseases.