Principal component analysis and differentially expressed protein (DEPs) identification
To explore the expression differences between glioma cells and normal cells, the proteomic data of 12 glioma samples were compared with that of 4 cerebral cortex cell types, using the principal components analysis (PCA) (Fig. 2). This analysis summarized the variations in glioma and cerebral cortex samples, based on the expression differences in proteins, and generated plots that separated samples. Twelve glioma samples were indicated as yellow triangles, and 4 cerebral cortex cell types were indicated as green dots. Glioma and cerebral cortex samples were clustered respectively, and 64.6% of variance was explained by the first two principal components. We found that glioma samples were separated from normal cerebral cortex samples, which revealed that glioma and cerebral cortex samples were expressional indistinguishable at the overall proteomic level (Fig. 2a), and expressional differences between them could be used for further analysis.
In order to explore the occurrence and development mechanism of glioma, differentially expressed proteins were identified between glioma and 4 normal cerebral cortex samples, with established criteria (P < 0.05 and FDR level < 0.01). A total of 1480 differentially expressed proteins were obtained, which consisted of 226 up regulated proteins and 1254 down regulated proteins in glioma (Additional file 1: Table S4, Fig. 2b).
Oxygen and carbon contents in all proteins expressed in glioma and normal cerebral cortex cells
After retrieving proteomic data of glioma and four normal cerebral cortex samples, we calculated and compared oxygen contents and C:O ratios of proteins. The average oxygen content of all expressed proteins was 0.482 in glioma, and 0.483, 0.484, 0.482, 0.481 in four normal cerebral cortex samples (Additional file 1: Table S2). No significant differences of oxygen contents and C:O ratios among the selected groups were detected (Fig. 3a-b, Additional file 1: Table S2). The average oxygen content in proteins expressed in glioma was not significantly different from that of the normal cerebral cortex samples (P > 0.05, Fig. 3c). The average C:O ratio of proteins expressed in glioma was also similar to that of the normal cerebral cortex samples (P > 0.05, Fig. 3d). Since no differences of oxygen contents were found in this step, we only analyzed oxygen contents in proteins. We did not associate it with the protein’s expression levels.
Patterns of oxygen contents in highly and lowly expressed proteins
To better explore the association between a protein’s oxygen content and its expression level, we further calculated and compared the oxygen contents in the highly and lowly expressed proteins in glioma and four kinds of normal cerebral cortex cells. To ensure the result reliability, the highly and lowly expressed proteins were screened out based on the top/bottom 1% proteins, top/bottom 3% expressed proteins, the top/bottom 5% expressed proteins and proteins with threshold expression score of ≥maximum/≤0.1. Meanwhile the C:O ratio in each protein was calculated and used as a control for oxygen content comparisons. As the principal element of organisms, carbon is a suitable control for element composition analysis [49].
Our results showed that, in both glioma and 4 kinds of normal cerebral cortex cells, the oxygen contents of highly (proteins with maximum threshold expression score) and lowly (proteins with threshold expression score of ≤0.1) expressed proteins were significantly different (Fig. 4, Additional file 1: Table S3). Average oxygen content of highly expressed proteins is 6.65% higher than that of lowly expressed proteins in glioma, and this trend was less pronounced in endothelial, glial, neuronal, and neuropil cells (with 2.43% averaged) (Additional file 1: Table S3). Of the 5 samples considered, the distributions of oxygen content of the highly and lowly expressed proteins are entirely nonoverlapping (0.477 in the lowest highly expressed proteins vs. 0.475 for the highest lowly expressed proteins; Additional file 1: Table S3). All results were agreed for proteins expressed at the top/bottom 1, 3% or 5%, or proteins with a preset threshold score, and significant difference of oxygen contents could be detected in the majority (Fig. 4, Additional file 1: Figures S2, S3, S4, S5, S6, S7, S8, S9, S10, S11, S12, S13, S14, S15, S16).
Meanwhile the C:O ratio in each protein was calculated and used as another indicator for oxygen content changes. As carbon-to-nitrogen ratios had been used as an indicator for nitrogen limitation of plants and other organisms [50], this method could reduce the deviation caused by side chain lengths. Our results showed that the C:O ratio of highly expressed proteins (5.820, 6.291, 6.129, 6.153 and 6.117) was 4.50% lower than that of lowly expressed proteins (6.488, 6.362, 6.375, 6.390 and 6.339) in both glioma and 4 kinds of normal cerebral cortex cell types, and the C:O ratio of highly and lowly expressed proteins in both glioma and 4 normal samples were significantly different (p = 2.47e-10,1.30e-2, 1.01e-13, 1.39e-9 and 6.93e-3, Ks test, Additional file 1: Table S3) and their distributions were largely separated (Fig. 4).
All together, these results suggested that there was an association existed between proteins’ oxygen contents and their expression levels in glioma and four normal cerebral cortex samples. On average, oxygen content was 3.27% higher in the highly expressed proteins than that in the lowly expressed proteins.
Patterns of oxygen contents in up regulated and down regulated proteins in glioma
After observing the fact that oxygen contents of the highly expressed proteins were higher than that of the lowly expressed proteins in glioma and four cerebral cortex samples, we checked whether this phenomenon existed between up regulated and down regulated proteins. After screening the differentially expressed proteins, we identified 226 up regulated and 1254 down regulated proteins in glioma. We then calculated and compared the oxygen contents in these proteins. The average oxygen content (0.481) of up regulated proteins was 2.56% higher than that (0.469) of the down regulated proteins in glioma, and the difference was statistically supported (p = 0.0195, Ks test, Additional file 1: Table S5). The distribution of oxygen contents in up and down regulated proteins was largely separated (Fig. 5).
Meanwhile the carbon contents were calculated and used as a control for oxygen content calculations. The average carbon contents of DEPs in glioma and cerebral cortex tissues weren’t significantly different (p = 0.335, Ks test, Additional file 1: Table S5). As it was shown, the distribution of carbon contents in the up and the down regulated proteins virtually overlapped, which indicated no significant carbon content differences between glioma and four cerebral cortex proteomes (Fig. 5).
Localization of differentially expressed genes on genome
As described above, the oxygen contents of up regulated proteins were higher than that of the down regulated proteins in glioma, which is of great significance for deciphering the mechanism of oxygen element usage bias. Meanwhile, differentially expressed genes (DEGs) could help us to better understand the mechanism of glioma. Those genes coding the up or down regulated proteins were commonly considered as key disease-associated genes. We further determined the genome location of genes encoding these up and down regulated proteins. 226 up regulated genes and 1254 down regulated genes were respectively located on human genome (Fig. 6). Most differentially expressed genes were distributed on chromosome 1, with 25 up regulated and 108 down regulated genes respectively, following with 17 up, 68 down regulated genes on chromosome 2, and 15 up, 67 down regulated genes located on chromosome 19. Fewest differentially expressed genes were distributed on chromosome Y, with only 1 gene being located.
Among these key genes, ZSWIM5 was the most up regulated, which plays a possible role in nerve conduction formation [51], and was considered as high cytoplasmic expression gene of interest for human glioma [52]. In addition, CCR9, ADGRE5 and MEOX1 were significantly up regulated. CCR9 associated with T lymphocyte development when bound to its specific ligand, and is highly expressed in variety of cancers [53]. MEOX1 plays key role in regulating somite development, which associated with cancer progression [54]. ADGRE5 was closely related with tumor cell adhesion, migration, angiogenesis, and apoptosis [55]. These key genes play important roles in glioma proliferation.
Functional enrichment and dissection
After locating the key genes on human genome, we further analyzed functional enrichments of these key genes, which could help us to efficiently examine large gene lists in a network context. KEGG enrichments were performed respectively for genes of up regulated proteins and down regulated proteins to check the specific pathways enriched by these two gene groups. A total of 14 pathways were enriched by genes of up regulated proteins, 62 pathways were enriched by genes of down regulated proteins, with FDR value less than 0.05 and p value less than 0.05 as the threshold for enrichment analysis.
The top 10 enriched pathways of up regulated proteins were illustrated in Fig. 7a and Additional file 1: Table S6. The most enriched pathway was cell cycle, which were enriched by 12 proteins. Other proteins were enriched in MAPK signaling pathway (7 proteins), pathways in cancer (7 proteins), p53 signaling pathway (5 proteins), VEGF signaling pathway (5 proteins), oocyte meiosis (5 proteins), DNA replication (4 proteins), peroxisome (4 proteins), progesterone-mediated oocyte maturation (4 proteins), apoptosis (4 proteins), arginine and proline metabolism (3 proteins), arachidonic acid metabolism (3 proteins), glycolysis/gluconeogenesis (3 proteins) and alpha-Linolenic acid metabolism (2 proteins). In addition, the top ten enriched pathways of down regulated proteins were presented in Fig. 7a and Additional file 1: Table S6. Twenty-eight proteins were enriched in Focal adhesion, and 26 proteins were enriched in regulation of actin cytoskeleton and neuroactive ligand-receptor interaction. Other proteins were enriched in MAPK signaling pathway (25 proteins), tight junction (21 proteins), calcium signaling pathway (21 proteins), endocytosis (20 proteins), leukocyte transendothelial migration (17 proteins), axon guidance (17 proteins) and ECM-receptor interaction (14 proteins).
These pathways enriched by up regulated proteins and down regulated proteins could help us to explain the pathogenetic mechanisms of glioma. Peroxisome is an important site for biological oxidation and energy metabolism. Up-regulation of this pathway increases the energy metabolism of glioma cells [56], which supplies adequate energy to the rapid proliferation of glioma cells. Peroxisomes are also major oxygen users and the oxygen produced from hydrogen peroxide is used within the organelle [57]. In addition, p53 signaling pathway and apoptosis have also be determined to be closely related to the occurrence of glioma. Pathways enriched by down regulated proteins suggested that neuroactive-ligand receptor interaction pathway and axon guidance were associated with the occurrence of glioma. As having been reported, the interruption of neuroactive-ligand receptor interaction pathway could cause some neuron related diseases [58]. Proper function of axon guidance is essential for avoiding neurological disorders [59]. So the disorders of neuroactive ligand-receptor interaction and axon guidance inevitably lead to glioma.
Further, we dissected these pathways enriched by genes of up regulated proteins for understanding which kind of up regulated proteins consumed more oxygen in the hypoxic microenvironment of glioma. The functions of those proteins with oxygen content higher than the average (of five proteomes, 0.482) were checked one by one manually (Fig. 7b, Additional file 1: Table S7). We found that, most of these proteins with high oxygen contents were involved in cell cycle (33.33%), followed by signaling (23.08%), cyclins (17.95%), apoptosis (12.82%), metabolism (7.69%), and transcription factor (5.13%) (Fig. 7c).