Data-Driven Cluster Analysis of Cerebrospinal Fluid Proteome and Associations with Clinical Phenotypes in Systemic Lupus Erythematosus.

OBJECTIVE

To explore the cerebrospinal fluid (CSF) proteome in systemic lupus erythematosus (SLE) and the associations between the CSF proteomic patterns and clinical manifestations.

METHODS

CSF samples from 29 female outpatients with SLE were analyzed with label-free liquid chromatography tandem mass spectrometry. Inclusion and CSF collection were conducted irrespective of clinical manifestations and disease duration. Proteomic data were used for sample clustering and analyzed for clinical variance. Proteins were clustered using Weighted Gene Co-expression Correlation Network Analysis. Modules were biologically characterized and analyzed for correlation to the clinical dataset.

RESULTS

Three patient clusters were identified. Cluster 1 was characterized by the highest frequency of nephritis, depression, and cognitive dysfunction. Cluster 2 showed the highest frequency of alopecia and Sjogren disease antigen A-antibodies (anti-SSA) and a low frequency of cognitive impairment. Cluster 3 had a higher frequency of autonomic neuropathy and headache. Six protein modules were identified (module 1 [M1]-M6). Modules were characterized by nervous tissue proteins (M1), central nervous system (CNS) lipoproteins (M2), macrophage proteins (M3), plasma proteins (M4), Ig (M5), and intracellular metabolic proteins (M6). M1 and M2 proteins were most abundant in cluster 1 and correlated with nephritis, depression, and cognitive impairment. Increased abundance of M4 and M5 proteins were most distinct in cluster 2 and inversely correlated to cognitive impairment and brain atrophy.

CONCLUSION

Patients clustered by their CSF proteomic pattern had different disease phenotypes. Nephritis and neuronal damage defined the group with higher levels of neuronal proteins in CSF, which may suggest shared pathogenetic pathways in SLE affecting the kidney and CNS.