The aim of the current study was to gain insight into the molecular function and biological roles of EWSR1. Accordingly, we performed global proteome profiling of the brain tissue from Ewsr1 KO and WT littermate control mice. Both three Ewsr1 KO and three WT mice were generated and gene KO was validated by PCR, and their brain proteomes were compared. For quantitative proteomic analysis, we used 6-plex TMT isobaric labeling, with high-pH fractionation to increase protein identification efficiency. Protein samples of brain tissue from each mouse were individually labeled and analyzed by MS. The data were analyzed using protein sequence database search and various bioinformatics tools (Fig. 1).
3 − 1. Quantitative proteome analysis of Ewsr1 KO versus WT mouse brain tissue
We compared the global proteome of Ewsr1 KO mouse brain tissue with that of WT mice using isobaric TMT labeling with LC-MS/MS. PCA of MS data showed good separation of Ewsr1 KO and WT mouse brain samples (Fig. 2A). Overall, 9115 proteins were identified and 9031 proteins were quantified with the protein and peptide level FDR below 1% (FDR Q-value < 0.01) using the multiplexed TMT approach, with a stable coefficient of variance (CV) of raw peptide abundance during the analysis (Fig. 2B and Table S2).
To identify proteins that are potentially involved in the Ewsr1 mediated cellular process, we used three different strategies, i.e., quantifying of peptide abundance based on intensities of the reporter tags using width adjustment or total amount normalization at protein-level, and quantifying of peptide abundance from each protein at PSM-level using normalization at spectrum- and protein-level by MSstatsTMT 23 (Figure S1). The box plots show that the three different methods perform as expected and make the centroids of the global distributions more similar (Figure S2). In addition, we compared CV distributions for the original data, after with adjustment normalization, after total amount normalization, or after MSstatsTMT. The averages CVs are 14.19% and 15.02% for WT and KO groups in the original data, respectively (Fig. 2B). Three normalization methods dramatically lower the average CVs - to 60–93%, respectively (Figure S3). Interestingly, the MSstatsTMT showed a very good behavior with a lowest CV value, whereas the width adjustment normalization displayed a greater dispersion of data compared with other methods (Figure S3). However, all three methods performed well, showing relatively lower CV values.
During comparison, we selected proteins that were significantly up-regulated (p-value < 0.05 and fold-change > 1.2) or down-regulated (p-value < 0.05 and fold-change < 0.84). This revealed 158 (up = 73, down = 85) and 163 DEPs (up = 84, down = 79) identified by width adjustment normalization and total abundance normalization, respectively, and 175 DEPs (up = 80, down = 95) identified by MSstatsTMT analysis. All of the detected DEPs were visualized using volcano plots (Fig. 2C–E). All DEPs are listed in Table S3-S5. Figure 2F also shows the lists of differentially expressed proteins found using the same criteria for the data with different normalization procedures. There was a good overlap between the normalization procedures, and 118 DEPs were common to all the analyses. In other words, irrespective of normalization methods (with or without background normalization for TMT signal intensity, with or without quantile normalization of the summed protein intensity, and with or without spectrum-level normalization) these 118 proteins were differentially expressed in Ewsr1 KO mouse brain compared to wildtype mice.
3 − 2. Physiological process and molecular function of Ewrs1 KO mouse brain proteome
To uncover the physiological or functional characteristics associated with specific changes in the Ewsr1 KO mouse brain, we first concentrated on the 118 DEPs that identified in all three statistical analysis. We first generated a hierarchical clustering map of the 118 DEPs using normalized protein intensities that calculated from MSstatsTMT and visualized the protein expression patterns in the brain tissue (Fig. 3A). The hierarchical clustering of 118 DEPs using normalized protein intensities obtained from width adjustment and total peptide normalization is also shown in Figure S4. Sample clusters by column were clearly separated into two large clusters, Ewsr1 KO or WT, and protein expression patterns illustrated by the colorimetric scheme by row also indicated two major clusters (Fig. 3B). The first cluster contained proteins that were significantly down-regulated in Ewsr1 KO mouse brain and the second cluster contained significantly up-regulated proteins in Ewsr1 KO mouse brain compared to that of WT mice. DEPs from each cluster were subjected to functional classification analysis using the web-based DAVID annotation tool (Table S6 and S7). The first cluster proteins were mostly categorized into metabolism-related biological processes, such as organic acid, carboxylic acid, and cellular amino acid metabolic processes. The major pathways determined by the KEGG and Reactome pathway analysis also indicated metabolic pathways, metabolism of amino acids and derivatives, and ABC-family protein-mediated transport pathways. In addition, the modulation of synaptic transmission and neurotransmitter transport were attained. Proteins up-regulated in Ewsr1 KO brain tissue were also related to carboxylic acid and organic acid metabolic processes. Specifically, positive regulation of bone resorption and remodeling functions were detected (Fig. 3C). This could be related to the physiological traits of EWSR1, a primary bone sarcoma that commonly occurs in the proximal long tubular bone or around the growth plate 31.
Furthermore, we used EnrichmentMap 26 and gene ontology (GO) to analyze all DEPs to investigate functionally relevant proteins, coherence pathways, and network interactions by implementing statistically significant (p-value < 0.05) DAVID terms (Fig. 4). We visualized the protein network and integrated it within the Cytoscape tool 32. The identified GO terms were categorized into 11 clusters. The majority of the enriched functions were correlated with metabolic pathways, such as the phosphorous metabolic pathway, which was the most statistically significant enriched term. This could be correlated with the results of a microRNA study, which showed that Ewsr1 KO in the spinal cord leads to the deregulation of G-protein signaling and affects metabolic pathways33. Among these pathways, the regulation of neurotransmitter secretion and bone remodeling were enriched, which were correlated with the DAVID analysis outcomes. The neurotransmitter secretion cluster includes the regulation of presynaptic processes, signal release from synapses, and neurotransmitter level processes. By contrast, regulation of bone remodeling, cell–matrix adhesion, and wound healing functions were classified into tissue remodeling: bone clusters. Therefore, we concentrated on these two physiological processes and extracted the proteins allocated to these clusters to investigate their expression patterns. Five and ten proteins were classified into neurotransmitter secretion and tissue remodeling: bone processes (Table 1), respectively. The expression patterns and protein interactions were visualized using the Cytoscape tool (Fig. 5). Overall, proteins involved in the tissue remodeling: bone function were up-regulated and those involved in the regulation of neurotransmitter secretion process were down-regulated in the brain of Ewsr1 KO mice.
Table 1
Biological process and gene categorized in tissue remodeling bone and neurotransmitter regulation process.
Enrichment
|
Term
|
Count
|
%
|
p-value
|
Genes
|
Fold Enrichment
|
GO
|
Tissue remodeling bone
|
wound healing
|
8
|
5.517
|
0.021
|
SERPINE2, CPQ, PECAM1, NF1, HRG, SERPING1, SYT7, ITGB3
|
2.869
|
GO:0042060
|
|
positive regulation of tissue remodeling
|
3
|
2.069
|
0.017
|
TFRC, HRG, ITGB3
|
15.064
|
GO:0034105
|
|
regulation of bone remodeling
|
4
|
2.759
|
0.006
|
TFRC, NF1, SYT7, ITGB3
|
10.590
|
GO:0046850
|
|
cell-matrix adhesion
|
5
|
3.448
|
0.032
|
SORBS1, PECAM1, NF1, HRG, ITGB3
|
4.161
|
GO:0007160
|
|
regulation of bone resorption
|
3
|
2.069
|
0.036
|
TFRC, NF1, ITGB3
|
9.928
|
GO:0045124
|
regulation of neurotransmitter secretion
|
signal release from synapse
|
4
|
2.759
|
0.044
|
CPLX2, PTPRN2, NF1, SYT7
|
5.065
|
GO:0099643
|
|
neurotransmitter secretion
|
4
|
2.759
|
0.044
|
CPLX2, PTPRN2, NF1, SYT7
|
5.065
|
GO:0007269
|
|
presynaptic process involved in chemical synaptic transmission
|
4
|
2.759
|
0.049
|
CPLX2, PTPRN2, NF1, SYT7
|
4.854
|
GO:0099531
|
|
regulation of neurotransmitter levels
|
5
|
3.448
|
0.035
|
CPLX2, PTPRN2, NF1, COMT, SYT7
|
4.023
|
GO:0001505
|
3–3. Interactome analysis of Ewrs1 KO mouse brain proteome
Although we have identified a few enriched biological processes for the differentially expressed proteins, we still could not rule out key biological process and molecules in Ewsr1 KO mouse brain proteome. For this, we conducted an interactome analysis with 156 DEPs that identified in two statistical methods or more to unravel key functional modules using the STRING version 11 25. We used the protein-protein interaction (PPI) with co-expression, co-occurrence, neighborhood, and embedded databases. The required PPI confidence was calculated using the BottleNeck ranking method, embedded in cytoHubba 28 and top 20 hits with the shortest path between each protein were filtered using a colorimetric scheme. The layout was refined using Prefuse force-directed layout. The expanded sub-network of interacting proteins (Fig. 6A) and the top 20 interactome proteins were plotted, showing that Ewsr1 directly interacted with Nrf1, C4b, and Acat3 (Fig. 6B). Interestingly, the GO enrichment showed that the Ewsr1 sub-network is associated with CNS development and regulation of bone resorption (Fig. 6C). Especially, Syt7 (synaptotagmin 7) and Nf1 (neurofibromin) showed a high rank of interaction correlation, and both proteins were downregulated in the brains of Ewsr1 KO mice (Table S3-S5).
3–4. Validation of global proteome experiments
Data-independent acquisition (DIA)-MS was used for orthogonal validation of our finding from TMT quantification and interactome analysis 34. DIA-MS analysis was performed in the independent samples consisting of three wild-type and three knock-out mice. Significantly differentially regulated targets in Ewsr1 KO relative to WT showed good agreement between TMT and DIA datasets (Figure S5A and Table S8). Interestingly, nine (Ewsr1, Syt7, Nf1, Prkdc, Comt, C4b, Tfrc, Acat3, and Cbs) out of 20 top interactome highlighted in Fig. 6B were validated by DIA-MS analysis in the independent samples, indicating the reliability of our analysis (Figure S5B). In addition, two proteins (Map2 and Syt7) were validated by western blotting. We selected Map2 as proteins that were highly downregulated in Ewsr1 KO mouse brain compared with other DEPs (Table S3-S4). Syt7 was selected because it is involved in both terms of tissue remodeling bone and regulation of neurotransmitter secretion. Furthermore, Syt7 and Map2 were associated with the top 20 interacting proteins with Ewsr1. The analysis revealed that all these proteins were down-regulated in the brain of Ewsr1 KO mice (Fig. 7 and Figure S6). This confirmed the findings of the proteomics approach. The t-test results and abundance of each blotted band are presented in Table S9.