Potential biomarkers of spinal dural arteriovenous fistula: C4BPA and C1QA

Abstract

Background and purpose: A major challenge in spinal dural arteriovenous fistula (SDAVF) is timely diagnosis, but no specific predictive biomarkers are known.

Methods: In the discovery cohort (case, n = 8 vs. control, n = 8), we used cerebrospinal fluid (CSF) and paired plasma samples to identify differentially expressed proteins by label-free quantitative proteomics. Further bioinformatics enrichment analyses were performed to screen target proteins. Finally, it was validated by ELISA in two of the new cohorts (case, n = 17 vs. control, n = 9), and univariate analysis, simple linear regression, and receiver operator characteristic (ROC) curve analysis were performed to evaluate the diagnostic potential.

Results: In the discovery cohort, the most overexpressed proteins were APOB and C4BPA in CSF samples of patients. The GO/KEGG enrichment analysis indicated that the upregulated proteins were mainly involved in the acute inflammatory response and complement activation. Hub-gene analysis revealed that APP might be the key protein in the molecular interaction network. In the validation cohort, C4BPA and C1QA were significantly overexpressed in the CSF of patients, averaging 3046.9 ng/ml and 2167.2 ng/ml, respectively. Simple linear regression demonstrated that levels of C1QA and C4 were positively correlated with total protein in CSF (R² = 0.8021, p = 0.0005; R² = 0.7447, p = 0.0013). The areas under the ROC curves of C4BPA and C1QA were 0.86 and 1.00, respectively.

Conclusions: This study was the first to identify C4BPA and C1QA as potential biomarkers for the diagnosis of SDAVF and revealed that complement pathway activation might be one of the molecular mechanisms for venous hypertension myelopathy.

Keywords: Biomarkers; C1QA; C4BPA; Spinal dural arteriovenous fistula; Venous hypertension myelopathy.

Fig. 1

Study design flowchart. The discovery cohort was followed for at least 1 year, except one lost for the wrong contact detail. The validation cohort with at least a 6-month follow-up was divided into two cohorts due to the consumption of CSF samples in the exploration of experimental conditions. * The two validation cohorts had some duplicated patients (n = 2) and controls (n = 1) but were different from the discovery cohort

Fig. 2

Imaging findings of a patient with misdiagnosis. A 69-year-old woman complained of lower-extremity progressive weakness and numbness over one year. Magnetic resonance imaging showed a multisegmental intramedullary T2 hyperintensity (A) with diffuse contrast enhancement (B). There was no evidence of obvious fluid voids. C DSA revealed a fistula (white arrow) at the right T6 level, fed by the radicular artery rising from the segmental artery

Fig. 3

Preliminary analysis of DEPs. Volcano plot of CSF (A) and paired plasma samples (B) comparing SDAVF patients with the controls. The red dots refer to significantly overexpressed proteins, and significantly downregulated proteins are coloured blue. Cluster heatmap of CSF (C) and paired plasma samples (D) comparing SDAVF patients with controls. The ‘red’ blocks refer to overexpression and downregulated proteins are coloured ‘blue’. The colour intensity indicates the degree of fold change. E The Venn diagram summarizes the relationship between the two datasets. F GO term/KEGG pathway enrichment analysis of overexpressed proteins in CSF samples. The size of the nodes denotes the number of proteins, and the colour represents the adjusted p value

Fig. 4

PPI network and hub gene analysis. Five interactional subclusters (A) and the top 10 hub genes (B) were obtained from 98 DEPs in CSF samples. The node score cut-off was set as 0.2. Continuous colour variation from deep to shallow reveals the score (A) or rank (B) from high to low. C Enrichment and visualization of GO/KEGG annotation based on the subclusters and hub gene list. The size of the nodes denotes the number of proteins, and the colour represents the adjusted p value

Fig. 5

Quantitative analysis and clinical relevance of potential biomarkers. A–C The specific protein levels were measured in CSF samples via ELISA kits. The concentration (Y-axis) was normalized to the total protein in CSF. Generally, data that passed the normality test were described by the mean with SD and compared by unpaired t test. *p < 0.05, “ns” refers to p > 0.05. D, E Simple linear regression was used to assess the correlation between clinical factors and proteins levels. Solid lines are the fit of linear regression, and dotted lines represent the 95% confidence interval. F ROC analysis of multiple variables. The diagonal dashed line reflects a random prediction (AUC = 0.5)