Comparative proteome analysis of the spinal dural arteriovenous stula arterial draining vein with label-free quantitative proteomics

Background: Spinal dural arteriovenous stula (SDAVF) is the most common spinal vascular shunt lesion. Although pathological changes in the SDAVF draining vein (SDAVF-DV) have been elucidated, protein changes remain enigmatic. We investigated protein changes in the SDAVF-DV. Methods: Three SDAVF-DV samples were collected, and supercial temporal artery (STA) and supercial temporal vein (STV) samples were used as controls. After quantication and enzymolysis of the proteins, label-free quantitative proteomics was performed, and the peptide mixture was fractionated and analysed by liquid chromatography tandem mass spectrometry (LC-MS/MS) to identify the differentially expressed proteins. Bioinformatics analysis of the differentially expressed proteins was also performed using Gene Ontology (GO), Kyoto Encyclopedia of Genes and Genomes (KEGG) and protein-protein interaction (PPI) network analyses. Results: Compared with the STA, the SDAVF-DV had 195 upregulated proteins and 303 downregulated proteins. GO analysis showed that the most differential GO terms in each category were the adenylate cyclase-modulating G protein-coupled receptor signalling pathway, U6 snRNP and SH3 domain binding. KEGG pathway analysis showed that the most differentially expressed protein pathway was focal adhesion. Compared with the STV, the SDAVF-DV had 158 upregulated proteins and 362 downregulated proteins. GO analysis showed that the most differential GO terms in each category were lamellipodium assembly, U6 snRNP, and SH3 domain binding. KEGG pathway analysis showed that the most differentially expressed protein pathway was dilated cardiomyopathy. The PPI analysis revealed PPIs among the top 300 proteins. Conclusions: We demonstrated that the SDAVF-DV showed specic protein expression changes under long-period venous hypertension. The results of the present study will provide insights into the pathogenesis of SDAVF formation at the protein level. The proteomic results provide a scientic foundation for further study to explore the pathophysiological mechanism of SDAVF.


Background
Spinal dural arteriovenous stula (SDAVF) is the most common spinal vascular shunt lesion characterized by an abnormal connection between a radicular meningeal artery and a radicular medullary vein. As venous connections drain to radicular veins, the draining vein shows gradual arterialization.
Because of venous hypertension, clinical presentations and progressive myelopathy can be assessed.
In SDAVF, venous drainage is provided by longitudinal spinal veins linked together and to the epidural network 1,2 . In many clinical case reports, the arterialized SDAVF draining vein (SDAVF-DV) was identi ed easily after opening the dura during an operation. The pathology of the arterialized SDAVF-DV was mentioned in a previous study. However, the protein changes in this arterialized vein under high intravascular pressure remain enigmatic.
In the present study, we used quantitative proteomics to compare the SDAVF-DV with the super cial temporal artery (STA) and super cial temporal vein (STV) to show different protein expression levels under venous hypertension. The results of our present study might provide insights into the pathogenesis of SDAVF formation at the protein level.

Ethics statement
The current study was examined and approved by the Ethics Committee of Huashan Hospital, Fudan University. Each participant provided their written informed consent to participate in this study.

Patients and tissue sample preparation
Three SDAVF-DVs were removed after microsurgery ligation. Three STAs and three STVs were obtained from patients with intracranial tumours via the extended pterional approach 3,4 . We used the samples from each group for the comparative proteomics analysis. The tissues used for the proteomics analysis were immediately frozen in liquid nitrogen and stored at -80°C. The SDAVF-DVs, STAs and STVs were homogenized in a 4% SDS, 100 mM Tris-HCl and 100 mM DTT solution. Then, a uorescence assay was conducted to determine the total protein concentration. Approximately 200 μg of total protein from the tissues was proteolysed on a 10-kDa lter (PALL Life Sciences, Shanghai, China) using a Filter Aided Sample Preparation (FASP) protocol as described in detail elsewhere 5 . The peptide solution was transferred to a Solid Phase Extraction Cartridge (Empore 7 mm/3 mL) for desalting and clean-up. The peptide samples were resuspended in water with 0.1% formic acid (v/v), and the protein content was estimated by UV light spectral density at 280 nm 6 prior to analysis by nano-liquid chromatography tandem mass spectrometry (N-LC-MS/MS).
Label-free quantitative analysis and data processing Trypsin-digested peptides from the tissues were analysed by LC-MS/MS; each sample was analysed twice. All raw Xcalibur les acquired from the MS runs were analysed using the default settings of MaxQuant software (version 1.3.0.5) with minor modi cations as previously described 7 . Hierarchical clustering was performed with MEV software (v4.6, TIGR). The differentially expressed proteins (p < 0.05) were analysed by hierarchical clustering to identify potential markers capable of classifying all samples.
The clustering pattern and expression analyses and volcano plots were based on R software according to standardized data. Venn diagrams of the characteristics of each of the three groups of differentially expressed proteins were generated.
The Gene Ontology (GO) and enrichment analyses of the dysregulated proteins in this experiment were based on the publicly available databases DAVID 6.7 (http://david.abcc.ncifcrf.gov/) and QuickGO (http://www.ebi.ac.uk/QuickGO/). The genomic, chemical and systemic functions of the dysregulated proteins were analysed and enriched by Kyoto Encyclopedia of Genes and Genomes (KEGG) analysis (http://www.kegg.jp/kegg/pathway.html). The signi cance of differential protein enrichment in each pathway entry was calculated using the hypergeometric distribution test and is expressed as the p value. Predicted protein-protein interaction (PPI) networks for these differentially expressed proteins were constructed using the STRING database (http://string.embl.de/) and Cytoscape software (http://www.cytoscape.org/).

Statistical analysis
The statistical analysis was performed with IBM SPSS, and the graphs were generated with GraphPad Prism software. The signi cance of differences between two groups in the proteomics analysis was assessed using one-way analysis of variance (ANOVA). Proteins were de ned as signi cantly differentially expressed when the ratio was ³2 or £0.5 in the SDAVF-DV compared with normal tissues (p <0.01).

Results
Identi cation of differentially expressed proteins in the SDAVF-DVs, STAs and STVs Three paired SDAVF-DV, STA and STV tissue samples were analysed in the initial discovery phase.
An equal amount of protein from each tissue was digested. Then, the peptides were analysed by N-LC-MS/MS. Using MaxQuant (version 1.3.0.5), we identi ed 2829 non-redundant proteins with a local false discovery rate (FDR) < 1% and at least two unique peptides per protein. The label-free quanti cation (LFQ) intensity ratios for the 2829 proteins were calculated, and signi cant differences in the protein expression levels between two tissues were determined using a t-test (p< 0.05). Compared with the STA, the SDAVF-DV had 195 signi cantly upregulated proteins and 303 signi cantly downregulated proteins. Compared with the STV, among the 520 proteins that exhibited signi cant differences, 158 were signi cantly upregulated and 362 were signi cantly downregulated in the SDAVF-DV ( Figure 1A). When the three groups were combined, 480 differentially expressed proteins were identi ed (shown in the heatmap in Figure 1B). Venn analysis showed the variation and commonalities of different proteins in each group. A total of 1026 proteins were expressed in all groups, and 150 proteins were identi ed only in the SDAVF-DV ( Figure 1C).

GO analysis
We performed GO analysis to analyse the differentially expressed proteins. When comparing the SDAVF-DV with the STA, most of differential GO terms expressed in each category were the adenylate cyclasemodulating G protein-coupled receptor signalling pathway, U6 snRNP and SH3 domain binding. We examined the top ten up-and downregulated GO terms in the biological processes, cellular components and molecular functions categories with 2.0-fold (p<0.05) differential gene expression (Table 1).
When comparing the SDAVF-DV with the STV, and the most differential GO terms expressed in each category were lamellipodium assembly, U6 snRNP, and SH3 domain binding.
The top ten up-and downregulated GO terms based on the comparison of the SDAVF-DV and STV are also listed in Table 2. In GO classi cation, 93 differentially enriched GO terms were found between the SDAVF-DV and STA: 60 GO terms were upregulated, and 33 were downregulated (Figure 2A). Compared with the STV, the SDAVF-DV had 109 differentially enriched GO terms: 31 terms were upregulated, and 78 were downregulated ( Figure 2B).

KEGG pathway analysis
The KEGG pathway analysis of these differentially expressed proteins also demonstrated related pathways. Figure 3 shows the number of proteins in each KEGG pathway and the p value of the top 20 pathways. Compared with the STA, the top three differentially expressed protein pathways were focal adhesion, the PI3K-Akt signalling pathway and the extracellular matrix (ECM)-receptor interaction. Compared with the STV, the top three differentially expressed pathways were dilated cardiomyopathy, hypertrophic cardiomyopathy and adrenergic signalling in cardiomyocytes.

PPI analysis
We used the STRING database to analyse the differentially expressed proteins, obtain the interactions/relationships among the differentially expressed proteins and calculate the combined score. We selected the top 300 proteins and found signi cant PPIs among them (Figures 5 and 6). Compared with the STA and STV, the SDAVF-DV showed up-and downregulated proteins, and the top three interaction proteins are listed in Table 3.

Discussion
SDAVF is a common arteriovenous shunt located inside the dura mater close to the spinal nerve root 8 .
Venous hypertension, which induces medullar venous out ow disturbances, results in chronic hypoxia and congestive myelopathy 9 . The direct intraoperative measurement of the vascular pressure in the stula can be as high as 74% of the systemic arterial pressure 10,11 . This nding may explain why, in some patients, symptoms become worse during physical activity with a concomitant increase in arterial pressure 12,13 . Under long-period venous hypertension, draining vein arterialization begins.
A clear understanding of the mechanism of SDAVF development is still lacking. Our study was the rst to perform a comparative proteome analysis and show the differential expression of proteins in arterialized SDAVF-DVs compared with normal arteries and veins. In general, most of the proteins were the same between the three groups. Because of its special pathophysiology, the SDAVF-DV showed speci c protein expression compared with the STA and STV.
In the intraoperative observation, the SDAVF-DV showed arterial morphology. A. Thron proposed a hypothesis based on spine arteriovenous shunt anatomy 9 . Keisuke Takai showed that the vessel wall of the proximal subarachnoid portion of the intradural draining vessels was irregularly thickened by collagen and exhibited elastic brosis and was without a continuous internal elastic lamina and a regular smooth muscle layer. The diameter of the vessels was signi cantly enlarged 14 .
After GO analysis, the SDAVF-DV showed a decrease in smooth muscle contractile bres, which might indicate smooth muscle cell dysfunction. This might be induced by long-range venous hypertension stretching on the SDAVF-DV. Stretch plays an important role in maintaining smooth muscle cell function and regulating in ammation. A former study showed that mechanical stretch-induced endoplasmic reticulum stress, apoptosis and in ammation contribute to thoracic aortic aneurysm and dissection 15 . In our research, we also identi ed that mechanical stretching could induce smooth muscle cell changes from a contract phenotype to an in ammatory phenotype 16,17 . The regulation of in ammatory factors is related to the hypothesis on SDAVF formation. In THE KEGG and PPI analyses, the ECM and focal adhesion showed obvious changes. Degeneration of the ECM is primarily induced by the secretion of in ammatory cytokines and cell in ltration in cerebral vascular disease 18, 19 . During SDAVF formation, inner vessel wall in ammation might contribute to an insu cient ECM and trigger changes in pathological proteins.

Conclusions
To our knowledge, few studies have focused on the SDAVF-DV. However, several researchers have investigated its pathological characteristics 14,20 . We rst examined protein changes to determine whether the lesion vessel was an artery, vein or vein-to-artery transition.
Most previous studies have revealed the histology and anatomy of the SDAVF-DV. Based on intraoperative ndings, we demonstrated protein changes in the arterial SDAVF-DV. Our study adds new information on the formation of SDAVF to the realm of protein changes in the draining vein using proteomics. This nding may shed light on the mechanism of SDAVF formation. Availability of data and materials: All the materials and data are freely available.
Competing interests: The authors declare that they have no competing interests.      Visualization of PPIs between the SDAVF-DV and STA. Visualization of PPIs for the top 300 proteins using STRING analysis. Red represents upregulated proteins, and green represents downregulated proteins.
Page 17/17 Visualization of PPIs between the SDAVF-DV and STV. Visualization of PPIs for the top 300 proteins using STRING analysis. Red represents upregulated proteins, and green represents downregulated proteins.