Identification of DEPRGs in PDR
Based on the GSE102485 dataset, 2,399 up-regulated and 1,185 down-regulated genes were identified in PDR patients (Fig. 1A), with the top 15 up-regulated genes and top 15 down-regulated DEGs (sorted by adj. P) depicted in heatmap (Fig. 1B). After Venn analysis, a total of 168 genes (DEPRGs) were illustrated shared by both DEGs and platelet-related genes (Fig. 1C), composed of 146 (86.9%) up-regulated and 22 (13.1%) down-regulated genes (Fig. 1D). The DEPRGs were selected to visualize the expression patterns and chromosomal locations (Fig. 1E), of which the distribution was shown in all chromosomes except for chromosome Y, with chromosome 1 covering the most proportion.
Functional enrichment analysis of DEPRGs
GO, KEGG and DO enrichment analyses were conducted to determine the functions of the DEPRGs, reporting the “wound healing”, “regulation of body fluid levels”, “blood coagulation”, “hemostasis”, “coagulation”, “platelet activation”, “leukocyte migration”, “myeloid leukocyte migration”, “homotypic cell-cell adhesion” and “platelet aggregation” as top 10 associated GO biological functions (BPs) according to both the adj. P and functioned gene numbers (Fig. 2A). At the same time, results of KEGG analysis illustrated the significant enrichment of DEPRGs in “chemokine signaling pathway”, “PI3K-Akt signaling pathway”, “platelet activation”, “focal adhesion”, “proteoglycans in cancer”, “Rap1 signaling pathway”, “Ras signaling pathway”, “regulation of actin cytoskeleton”, “relaxin signaling pathway” and “Apelin signaling pathway” (Fig. 2B), and “lung disease”, “hematopoietic system disease”, “myocardial infarction”, “coronary artery disease”, “blood coagulation disease”, “interstitial lung disease”, “hemorrhagic disease”, “pulmonary fibrosis”, “blood platelet disease” and “thrombocytopenia” according to DO analysis (Fig. 2C).
Identification and validation of hub genes
A PPI network was generated with an interaction score set to 0.4 by uploading DEPRGs to the STRING online database (Fig. 3A). Afterwards, 9 hub genes (CDC42, GNAI2, LCK, LCP2, LYN, PLCG2, PTPN6, RAC1 and SYK) were eventually identified with the plug-1in MCODE of Cytoscape (Fig. 3B), which were all indicated up-regulated in GSE102485 (Fig. 3C) and GSE60436 (Fig. 3D). Furthermore, the AUCs of hub genes were all over 0.75 in GSE102485 (Fig. 3E) and GSE60436 (Fig. 3F), all hub genes could serve as biomarkers to sensitively and accurately distinguish PDR from normal samples.
Expression correlation analysis and organ-specific expression of hub genes
Correlation analysis revealed the positive correlations between hub genes (Fig. 4A), such as GNAI2 and RAC1 (cor = 0.91, P < 0.05), GNAI2 and SYK (cor = 0.91, P < 0.05), LCP2 and LYN (cor = 0.91, P < 0.05), and LCP2 and SYK (cor = 0.9, P < 0.05). These hub genes were observed specifically expressed in the bone marrow (CDC42, GNAI2 and LCP2), cerebellum (PLCG2), esophagus (RAC1), lymph node (PTPN6), parathyroid gland (SYK), spleen (LYN) and thymus (LCK), while less in the retina (Fig. 4B).
Functional similarity analysis and GSEA of hub genes
To further locate the key genes, the GO semantic similarity of these 9 hub genes was calculated, finding relatively higher functional similarities among LCK, LYN, and PLCG2 (Fig. 5A). To further explore the potential roles of hub genes in PDR, an GSEA analysis of the hub genes was performed, and the results showed the close association of the hub genes to “KEGG_RIBOSOME, KEGG_LYSOSOME”, “KEGG_T_CELL_RECEPTOR_SIGNALING_PATHWAY”, “KEGG_NATURAL_KILLER_CELL_MEDIATED_CYTOTOXICITY”, “KEGG_PATHWAYS_IN_CANCER”, “KEGG_NEUROACTIVE_LIGAND_RECEPTOR_INTERACTION”, “KEGG_CHEMOKINE_SIGNALING_PATHWAY”, “KEGG_UBIQUITIN_MEDIATED_PROTEOLYSIS”, and so on, in which the activation of these hub genes might participate (Fig. 5B).
MiRNA hub gene regulatory network
446 predicted miRNAs targeting hub genes were obtained on miRNet database (Supplementary Table 1). Figure 6 depicted the regulatory network of 25 miRNAs with degree ≥ 3 and 9 hub genes, with the top 4 of hsa-mir-374a-5p, hsa-mir-124-3p, hsa-mir-155-5p and hsa-mir-182-5p, and GNAI2, LYN, PLCG2, PTPN6, RAC1 and SYK as potential targets of hsa-mir-374a-5p. Hsa-mir-124-3p and hsa-mir-155-5p might play the regulatory role in CDC42, GNAI2, LYN, PTPN6 and RAC1. Hsa-mir-182-5p possibly bounded to CDC42, GNAI2, PTPN6, RAC1 and SYK. The most targeted hub genes pointed to RAC1, which was targeted by 156 miRNAs.
TF hub gene regulatory network
For further understanding of the regulatory network between TFs and hub genes, the 9 hub genes were taken to obtain the transcription-regulated network with 46 TFs (Fig. 7 and Supplementary Table 2). The predicted top 5 TFs targeting hub genes were FOXC1, YY1, FOXL1, GATA2 and TP53, regulating 6, 5, 4, 4 and 4 hub genes, respectively. At the same time, the top 5 hub genes were GNAI2, PTPN6, SYK, CDC42, and LCK, which were regulated by 19, 13, 13, 10, and 10 TFs, respectively.
Potential drugs identification of hub genes
139 drug-gene interaction pairs were determined in DGIdb, with 138 drugs and 7 hub genes (CDC42, LCK, LYN, PLCG2, PTPN6, RAC1 and SYK), as depicted in Fig. 8 and Supplementary Table 3, among which LCK, LYN and SYK lied the top 3 hub genes targeted by 52, 40, 28 drugs, respectively. Most potential drugs showed the potential interaction with hub genes as inhibitors in unknown manners or. This outcome may contribute to developing new targets for PDR therapy.
qRTPCR analysis of diabetic mouse retinal tissue and HRMECs treated with HG
qRT-PCR results showed an upregulation in the expression of GNAI2, LCK, LCP2, PLCG2, PTPN6, and RAC1 in HRMECs treated with HG compared to NG. It is worth mentioning that there was a significant downregulation in SYK. Similarly, in diabetic retina (DR) treated with HG, qRT-PCR results indicated an upregulation in the expression of LCP2, LYN, and PTPN6 when compared to expression in the naive retina (NR). However, GNAI2 was found to be significantly downregulated in this context (Figure. 9A and B, p < 0.05).