Deciphering the mechanism of Indirubin and its derivatives in the inhibition of Imatinib resistance using a “drug target prediction-gene microarray analysis-protein network construction” strategy

We identified indirubin and its derivatives from two sources: first by searching the PubChem database and then by manually searching PubMed to augment the data. PubChem (https://​pubchem.​ncbi.​nlm.​nih.​gov) is a public repository for information on chemical substances and their biological activities. As of September 2015, it contained more than 157 million depositor-provided chemical substance descriptions, 60 million unique chemical structures and 1 million biological assay descriptions, covering approximately 10 thousand unique protein target sequences [13]. We searched the PubChem database with “indirubin” as the key word to identify indirubin and its derivatives, downloaded the compound 2D structures and finally downloaded the “smile” format. In order to increase the comprehensiveness of the data, we manually searched the relevant literature in the PubMed database for titles dealing with indirubin derivatives.

It requires much manpower, material and financial resources to Identify targets of indirubin and its derivatives through experimentation. Therefore, we used a computerized virtual platform to screen for targets and then validated the targets by molecular docking or experimental verification. Swiss Target Prediction (http://​www.​swisstargetpredi​ction.​ch/​), a web server to accurately predict the targets of bioactive molecules based on a combination of 2D and 3D similarity measures with known ligands, was used to predict the putative targets of the indirubin and its derivatives. Predictions can be carried out in five different organisms, and mapping predictions by homology within and between different species is enabled for close paralogs and orthologs [14]. The “smiles” formats of indirubin and its derivatives were imported into Swiss Target Prediction to predict their putative targets of action. It is noteworthy that the predicted putative target was limited to Homo sapiens. In order to improve the reliability of predictions goal, only high-probability targets were selected. All putative targets Identified were sent to the Therapeutic Target Database (TTD) (http://​bidd.​nus.​edu.​sg/​group/​cjttd/​, 2015-09-10), the Comparative Toxicogenomics Database (CTD) (http://​ctdbase.​org/​, 2017-12-05) and the PharmGKB (https://​www.​pharmgkb.​org/​) to verify whether these putative targets had some connection to CML.

Gene expression profiling analysis is a useful method with broad clinical application in the identification of tumour-related genes in various types of cancer, from molecular diagnosis to pathological classification, from therapeutic evaluation to prognosis prediction, and from drug sensitivity to neoplasm recurrence [15]. Gene expression profile GSE2810 was downloaded from the Gene Expression Omnibus (GEO) database, GSE2810 data is based on the GPL2531 (Novusgene type 3 Hematology/Oncology TMU 667 array) platform,including 4 samples (2 imatinib-resistant samples and 2 imatinib-sensitive samples). It was submitted by Ohyashiki JH [16]. Quality control of gene expression data was performed using gene-specific probes. The analysis was carried out by using GEO2R, an online analysis tool for the GEO database, based on R language. We applied the analysis to classify the sample into two groups that had similar expression patterns in imatinib-sensitive and imatinib-resistant. We defined genes as differentially expressed (DEGs) when logFC > 1 or logFC < − 1(FC:Fold Change,the difference in the amount of gene expression in the sample). A p value < 0.05 was considered statistically significant. To further study the characteristics of DEGs and their functions, we analysed the DEGs with Gene Ontology and KEGG Pathway. Gene Ontology annotates and classifies genes by Molecular Function (MF), biological process (BP) and cellular component (CC). The pvalue of the GO term of the DEGs was calculated, and the most likely related GO term of the differential gene was located [17].KEGG is an online biochemical energy database that contains a set of genomic and enzymatic methods and is an information resource for the systematic analysis of gene functions and associated high-level genomic functions [18]. ClueGo, a plugin for Cytoscape 3.5.1 software, provides systematic and comprehensive biologically functional annotation of high-throughput gene expression [19]. Therefore, ClueGo online tools were employed for GO and KEGG pathway analysis. P < 0.05 was considered significant.

We Identified indirubin and 41 derivatives from the database and downloaded “smiles” format and 2D structures. The putative targets of indirubin and its derivatives were predicted through structural similarities. Indirubin and 41 derivatives and putative targets are shown in Table 1.

After gene chip data analysis, we obtained a heat map of the differentially expressed genes of the gene chip G2810 (Additional file 1: Fig. S1), we Identified a total of 125 DEGs with imatinib resistance (Fig. 1), of which 66 were up-regulated and 59 were down-regulated. According to FC,the top 10 significantly up-regulated DEGs and down-regulated DEGs are shown in Table 2. Go analysis and KEGG analysis of DEGs, we found that DEGs of imatinib resistance were closely related to biological processes including immune responses, regulation of protein modification process, regulation of phosphorylation, and regulation of cellular protein metabolic processes. DEGs were mainly involved in cytokine-cytokine receptor interaction pathways.

CCL13, the first significantly up-regulated chemokine, is a chemotactic factor that attracts monocytes, lymphocytes, basophils and eosinophils [24]. MAPK11, the second significantly up-regulated chemokine, plays an important role in the cascades of cellular responses evoked by extracellular stimuli, including proinflammatory cytokines and physical stress leading to direct activation of transcription factors. The study of Huang J et al. showed that the ERK signalling pathway was more activated in epirubicin treated triple-negative breast cancer (TNBC), possibly contributing to epirubicin resistance, suggesting that the ERK pathway could be used as a novel candidate for targeting therapy in refractory and relapse TNBC [25]. MLH1, the first significantly down-regulated DEG, has been shown to play an important role in haematologic malignancies. The novel mutation was also revealed to be a somatic aberration occurring prior to the initiation of the blast phase in a chronic myelogenous leukaemia (CML) patient. Among the possible MLH1 partners involved in signalling MMR or apoptosis is the proto-oncogene c-MYC, closely associated with cellular proliferation [26]. BCL10, the second significantly down-regulated chemokine, was involved in adaptive immune responses. Proliferation of NIK and IKK cells is promoted by pro-caspase-9 maturation and NF-κB activation.

To further explain the function of differentially expressed genes, we performed functional enrichment analysis of all differential genes based on GO analysis, and performed passway enrichment analysis of all differential genes based on KEGG analysis. we chose significantly up-regulated and down-regulated GO categories based on functional enrichment, The analysis results are shown in Figs. 2 and 3. Through GO analysis, we reached the following conclusions: up-regulated differentially expressed genes were primarily involved in the regulation of cell apoptosis, including immune responses, regulation of apoptosis, regulation of programmed cell death, regulation of cell death, regulation of transcription, cell death, death and DNA binding. The down-regulated DEGs were primarily related to cellular structures, such as cytoplasm, nucleus, extracellular space, positive regulation of transcription from the RNA polymerase II promoter, transcription factor activity and sequence-specific DNA binding growth factor activity. We performed pathway enrichment analysis of differentially expressed genes to Identify the biological pathways. Up-regulated differentially expressed genes were primarily involved in cytokine-cytokine receptor interaction, chemokine signalling pathways, the Toll-like receptor signalling pathway, the neurotrophin signalling pathway, leukocyte transendothelial migration, the MAPK signalling pathway, haematopoietic cell lineage, apoptosis, the T cell receptor signalling pathway and the JAK-STAT signalling pathway. Pathways dramatically altered among down-regulated genes were the cytokine-cytokine receptor interaction, Toll-like receptor signalling pathway, Jak-STAT signalling pathway, pathways in cancer, the NOD-like receptor signalling pathway, apoptosis, cell cycle and the p53 signalling pathway.

To identify the relationship between the putative targets of indirubin and its derivatives and DEGs of imatinib resistance, we constructed a P-P network of putative targets and DEGs (Fig. 4). The T-T network consisted of 171 nodes and 1082 edges. The major hubs in the hub interaction network were determined by calculating four features: degree,betweenness,closeness and K-coreness. We showed the major hubs in Fig. 3. After screening, we identified a total of 62 major hubs (Table 3), including 11 (EGFR, JAK2, ERBB2, CHUK, CDK5, KIF11, DRD2, CDK3, HTR1A, JAK3 and TYK2) indirubin and derivative targets and 51 DEGs for imatinib resistance. These 11 major hubs were closely related to DEGs that were resistant to imatinib. Indirubin and its derivatives may inhibit imatinib resistance through the regulation of these genes.

We manually screened out small molecule compounds that affected 11 major hubs in the putative target. After screening, a total of 15 small molecule compounds affected these putative targets, including 1, 3, 4, 5, 6, 8, 11, 14, 21, 24, 26, 33,36, 40, 41. These derivatives may all inhibit imatinib resistance. To further verify this conclusion, we evaluated docking of small molecule compounds and their putative targets that were included in the major hubs. The docking results are shown in Table 4.