Identification of potential targets of the curcumin analog CCA-1.1 for glioblastoma treatment : integrated computational analysis and in vitro study

The treatment of glioblastoma multiforme (GBM) is challenging owing to its localization in the brain, the limited capacity of brain cells to repair, resistance to conventional therapy, and its aggressiveness. Curcumin has anticancer activity against aggressive cancers, such as leukemia, and GBM; however, its application is limited by its low solubility and bioavailability. Chemoprevention curcumin analog 1.1 (CCA-1.1), a curcumin analog, has better solubility and stability than those of curcumin. In this study, we explored potential targets of CCA-1.1 in GBM (PTCGs) by an integrated computational analysis and in vitro study. Predicted targets of CCA-1.1 obtained using various databases were subjected to comprehensive downstream analyses, including functional annotation, disease and drug association analyses, protein–protein interaction network analyses, analyses of genetic alterations, expression, and associations with survival and immune cell infiltration. Our integrative bioinformatics analysis revealed four candidate targets of CCA-1.1 in GBM: TP53, EGFR, AKT1, and CASP3. In addition to targeting specific proteins with regulatory effects in GBM, CCA-1.1 has the capacity to modulate the immunological milieu. Cytotoxicity of CCA-1.1 was lower than TMZ with an IC50 value of 9.8 μM compared to TMZ with an IC50 of 40 μM. mRNA sequencing revealed EGFR transcript variant 8 was upregulated, whereas EGFRvIII was downregulated in U87 cells after treatment with CCA-1.1. Furthermore, a molecular docking analysis suggested that CCA-1.1 inhibits EGFR with various mutations in GBM, which was confirmed using molecular dynamics simulation, wherein the binding between CCA-1.1 with the mutant EGFR L861Q was stable. For successful clinical translation, the effects of CCA-1.1 need to be confirmed in laboratory studies and clinical trials.

immune cells. The measurement of immune cell infiltration 10 in GBM is an important tool for predicting clinical outcomes 11,12 , as a prognostic marker and a predictor of therapeutic outcomes.
To address the hurdles limiting effective GBM treatment, we explored new therapeutic compounds related to curcumin (Fig. 1), which has anticancer activity against various aggressive cancers, such as colon cancer, leukemia, and GBM 13 . Curcumin has been shown to increase the sensitivity of GBM cells to cisplatin, etoposide, camptothecin, and doxorubicin 14 . Curcumin exerts therapeutic effects in GBM via multiple pathways, including the suppression of AKT/mTOR and activation of ERK1/2 pathways in human malignant glioma U87-MG and U373-MG with PTEN mutations 15 . Furthermore, the effect of curcumin on the ERK pathway promotes the activation of p21, as observed by Choi et al. 16 The curcumin-induced inhibition of GBM cell proliferation and chemoresistance is mediated by AP-1 and NF-κB 14 . An in vivo study by Perry et al. (2010) revealed that curcumin affects glioblastoma growth and angiogenesis in mice with U87 glioma xenografts 17 . In addition, Facina et al. (2021) demonstrated the anticarcinogenic effect of curcumin alone, and in combination with piperin, in bisphenol A-induced carcinogenesis in gerbil prostates 18 . Moreover, in vitro and in silico studies by Liang et al. (2021) successfully synthesized curcumin and its analog and suggested their potential as EGFR inhibitors, in which curcumin and its analog regulate the expression of EGFR 19 . A recent study has demonstrated that the curcumin analog dimethoxycurcumin promotes apoptosis, autophagy, and ROS production and suppresses cell viability in human gliomas 20 .
Natural products such as Tripterygium wilfordii 21 and Ganoderma 22 have immunomodulatory effects by inhibiting the expression of pro-inflammatory cytokines, and the production of other cytokines and antibodies. Curcumin exerts anticancer activity, in part, by modulating the immune system. A previous study has shown that curcumin increases the efficacy of immunotherapy in melanoma cells 23 . Additionally, curcumin is a promising immunotherapy for GBM 24 . A previous study showed that curcumin can be used in immunotherapy by decreasing the expression of immune checkpoint ligands and restoring the CD8 + cell function in head and neck cancer cells 25 . As discussed in a recent review, curcumin promotes immune function to eliminate cancer cells via several mechanisms 26 , however, its application is limited by its low solubility and bioavailability 27 .
Chemoprevention curcumin analog 1.1 (CCA-1.1), shown in Fig. 1, is a curcumin analog, with a substitution of the ketone group in the cyclopentane structure of PGV-1 (Fig. 1), a former analog, with a hydroxyl group; it has better solubility and stability than those of curcumin and PGV-1 28 . CCA-1.1 also exhibits better anticancer activity than that of PGV-1 in several cancer cells, including luminal A MCF-7, HER2-positive HCC1954, triplenegative 4T1 breast cancer cells, K562 leukemic cells, Caco2, and WiDr colon cancer cells 28 . CCA-1.1 is able to induce cell cycle arrest and senescence 29 , increase the cytotoxicity of doxorubicin 30 , and hamper migration in T47D, estrogen-positive breast cancer cells 31 and in WiDr colon cancer cells 32 . CCA-1.1 also inhibits the migration of triple-negative and HER2-positive breast cancer cells 33 and induces mitotic arrest in triple-negative breast cancer 34 . Bioinformatics studies have explored the target genes of CCA-1.1 in colon cancer 35 and triple-negative breast cancer cells 36 ; however, similar analyses have not been performed for GBM.
In this study, we explored potential targets of CCA-1.1 in GBM (PTCG) by an integrated computational analysis and in vitro study (Fig. 2). Targets of CCA-1.1 were predicted from public databases and further analyzed for the selection of candidates. Our results indicate that CCA-1.1 not only targets certain regulatory genes in GBM but also modulates the immune environment.   43 using the default settings for each database, as described previously 35 . Regulatory genes associated with GBM were obtained by searching against DISGENET https:// www. disge net. org 44 with the keyword human glioblastoma and default settings for the database. Padjajaran, Bandung. The U87 cells were cultured in RPMI medium, containing 10% of fetal bovine serum (FBS, Gibco), 1% of penicillin-streptomycin (Gibco), and maintained in 5% of CO2 incubator. For the cytotoxicity assay, the U87 cells (3,000 cells/ well) were seeded in a 96-well-plate and incubated for 24 h prior to treatment of CCA-1.1, temozolomide (TMZ, purchased from Sigma), or DMSO for the following 72 h. TMZ was used as a control as TMZ is the first choice for GBM treatment 55 . DMSO was used as a co-solvent of CCA-1.1, and TMZ, and as a control at a maximum concentration of 1% (v/v). At the end of incubation, an MTT solution was added and incubated for 3 h prior to addition of 10% of SDS solution. Cell viability was calculated as previously described 56  www.nature.com/scientificreports/ sequences: 3.00; identical query threshold: 90%. A prediction score of five indicates "tolerated. " The FATHMM analysis was performed using the following parameters: cancer-relativity inherited disease; weighted prediction; phenotype association, disease ontology. A prediction score of less than −1.5 indicates a "damaging" mutation. The coding SNP that impacted protein function was predicted using PANTHER with the following interpretations of the probability of deleterious effect (Pdel): "probably damaging" (time > 450 my, corresponding to a false positive rate of ~ 0.2 as tested using HumVar), "possibly damaging" (450 my > time > 200 my, corresponding to a false positive rate of ~ 0.4), and "probably benign" (time < 200 my). Predictions were performed by comparing the mutant to wild-type EGFR (PDB ID: 3NJP).

Molecular docking.
The binding properties of curcumin and its analogs (PGV-1 and CCA-1.1) against EGFR and its mutant forms were predicted by a molecular docking analysis. Before performing the simulations, a template of the EGFR structure (UniProt code P00533) was retrieved from AlphaFold (https:// alpha fold. ebi. ac. uk/) 64 . The structures of mutant EGFR (E709K, T263P, V774M, and L861Q) were manually predicted using the MOE 2010 software, using the default step preparation. Due to the unknown binding site of each compound, the sitefinder in MOE was used to create a dummy site as the possible cavity for docking simulation. MOE 2010 (licensed from the Faculty of Pharmacy UGM) was also used for docking simulations, and the visualization of interactions. PGV-1 and CCA-1.1 structures were drawn using Marvin Sketch, and the curcumin structure was downloaded from PubChem. These structures were then subjected to conformational searches and energy minimization by MOE using the Energy Minimize Menu. For the docking simulation settings, London dG was used for both Rescoring 1 and Rescoring 2. Triangle Matcher was used for the score function and placement setting, and Forcefield was used to refine the docking results from 30 retained poses, as described previously 49 . The conformation with the lowest binding interaction between the ligand and receptor was determined.

Molecular dynamics simulation.
The results of molecular docking were validated using molecular dynamics (MD) simulation. As the representative, we chose the binding pocket of EGFR L861Q in complex with curcumin, PGV-1, and CCA-1.1. The MD simulation was completed in NAMD 2.14 65 and visualized using VMD 1.9.4 66 . Parameterization of the proteins and ligands was prepared using CHARMM36 and CGenFF, available in the CHARMM-GUI web server 67 . For the solvation and neutralization steps, a cubic water box with 20-Å padding was added followed by K + and Cl − ion addition. For equilibration, the complex was minimized for 70 ps and simulated for 1 ns. Further, a 1-ns simulation (NPT ensemble, pressure 1 atm, and temperature 303 K) was conducted to finalize the MD simulation process. The visualization and trajectories of the MD results were analyzed using root-mean-square deviation (RMSD).
Ethical approval. This article does not contain any studies with human participants or animals performed by any of the authors.

Functional annotation.
A Gene Ontology analysis revealed that PTCGs were involved in various biological processes, including the response to stimulus, metabolic process, and cell communication (Fig. 3B). PTCGs were also enriched for cellular components, including the membrane, nucleus, and cytosol. In addition, PTCGs were associated with terms in the molecular functions category, including protein, ion, and nucleotide binding. A KEGG pathway enrichment analysis revealed that PTCGs were involved in several pathways, such as glioma, pathways in cancer, and p53 signaling pathways (Supplementary Table 4).
Protein-protein interaction network and hub gene identification. Using STRING, we constructed a PPI network including 268 nodes and 4597 edges, with an average node degree of 34.3, an average local clustering coefficient of 0.523, and a PPI enrichment p value of < 1.0e − 16 (Fig. 3E). Hub genes were selected using the cytoHubba plugin of Cytoscape as the top ten target genes with respect to degree scores, including AKT1, TP53, ALB, EGFR, SRC, TNF, CASP3, MAPK1, HSP90AA1, and MAPK8 (Fig. 3F, Table 1).

Analysis of genetic alterations in hub genes.
Genetic alterations in the ten hub genes were evaluated based on six studies of GBM using cBioportal (Fig. 4A). TCGA PanCancer Atlas 68 which showed the second highest genetic alterations and the largest number of patients among the GBM studies and was selected for further analysis. We found mutation rates of 0.3-53% in hub genes in the study population, including CASP3 www.nature.com/scientificreports/ www.nature.com/scientificreports/ TP53 (33%), and EGFR (53%) (Fig. 4B). In a mutual exclusivity analysis, three gene pairs were significant, namely TP53-EGFR, ALB-SRC, and TNF-CASP3 (Table 2). A pathway enrichment analysis revealed that several pathways are affected by the observed genetic alterations, including RTK-RAS, TP53, PI3K, and cell cycle pathways (Supplementary Table 5). The RTK-RAS pathway was detected in two queries, EGFR and MAPK1, as well as neighboring genes, including members of the ERBB family, RAS family, and RAF family, which are involved in cellular processes including proliferation, cell survival, and translation ( Fig. 4C). Copy number alterations in ALB, SRC, and TNF were not obvious (Fig. 4D). In AKT1, significant differences in mRNA levels were found between alteration types (i.e., shallow deletion, diploid, and gain); in particular, the expression of AKT1 was highest in cases with copy number gain, followed by diploid, and shallow deletion. The mRNA levels of TP53 in the shallow deletion group were significantly lower than those in the diploid and gain groups. In EGFR, we found that mRNA expression levels in the case of amplification were significantly higher than that those in the diploid and gain groups. mRNA levels of CASP3 and MAPK8 in diploids were significantly higher than those in the shallow deletion group. In addition, mRNA levels of MAPK1 and HSP90AA1 were significantly higher in the case of gain than in the diploid and shallow deletion groups. We then evaluated TP53 and EGFR mutations across patient samples in Liu et al. (2018). We found several mutations in TP53 in the p53 tetramerization domain, p53-DNA binding domain, and p53 transactivation domain (Fig. 4E). EGFR mutations occurred in many domains, such as the receptor-ligand domain, furin-like cysteine-rich region, growth factor receptor domain IV, and protein tyrosine kinase domain (Fig. 4F).
Expression of hub genes in glioblastoma samples. The mRNA levels of the hub genes AKT1, TP53, EGFR, and CASP3 were significantly higher in patients with GBM than in normal brain tissues (Fig. 5A). In addition, mRNA levels of ALB, SRC, TNF, MAPK1, HSP90AA1, and MAPK8 were not statistically significant between GBM and normal brain tissues.
Survival analysis of hub genes. The prognostic value of each hub gene was analyzed using a Kaplan-Meier plot. Among the hub genes, only ALB and MAPK8 levels were significantly associated with the survival of patients with GBM (Fig. 5B). Patients with low levels of ALB had a better overall survival than that of patients in group with high expression (p = 0.0223), whereas patients with high levels of MAPK8 had a better overall survival than that of patients with low expression levels (p = 0.0416).

Correlation between immune cell infiltration and hub genes.
We explored correlations between the expression of hub genes and levels of immune cell infiltration in GBM using the TIMER 2.0 database (Table 3    www.nature.com/scientificreports/ CCA-1.1 performed cytotoxicity and induces the modulation of EGFR on U87 glioblastoma cells. We performed an MTT assay to measure the cytotoxicity of CCA-1.1 and TMZ, and both compounds showed cytotoxicity against U87 cells with an IC50 value of 9.8 and 40 μM, respectively (Fig. 6A,B). To check the molecular mechanism of CCA-1.1 in U87 cells, we performed next generation sequencing between untreated and CCA-1.1 treated U87 cells, and then analyzed the results for DEGs (Fig. 6C, Supplementary Table 6). The raw data of gene expression can be accessed at the Gene Expression Omnibus (GEO, http:// www. ncbi. nlm. nih. gov/ geo/), using accession number GSE206241. Among the potential target genes, only EGFR showed significant results based on differential expression analysis, in which EGFR transcript variant 8 was upregulated in CCA-1.1 treated U87 cells, whereas EGFRvIII was downregulated in U87 cells after treatment with CCA-1.1 (Table 4). These findings confirm the bioinformatic approach which highlights the importance of EGFR as targets of CCA-1.1 in inhibition of GBM.
Prediction of effects of mutations on protein function. We identified EGFR as a promising target of CCA-1.1 for GBM treatment. We further predicted the functional effects of EGFR alterations using several databases, including PolyPhen-2, Provean, SIFT, FATHMM, and PANTHER. We selected 22 EGFR mutations detected in GBM samples by Liu et al. (2018) (TCGA PanCancer); these mutations were located in the growth factor receptor domain, protein kinase-like (PK-like), receptor L domain, growth factor receptor domain IV, furin-like cysteine-rich region, protein kinase-like (PK-like), and protein tyrosine kinase (Table 5, Supplementary Table 7). The EGFR mutations in the protein kinase-like domain, namely E709K, V774M, and L861Q, were predicted to be damaging, deleterious, and cancer-related (Table 5). Another mutation, T263P, located in a furinlike cysteine-rich region, was also predicted to be associated with cancer. The V774M mutation, which occurs in the protein kinase-like domain, was predicted to be damaging and associated with cancer. In addition, L861Q, in the protein tyrosine kinase domain, was predicted to be damaging and related to cancer.

Molecular docking and MD.
We successfully predicted the structures of mutant EGFR using a template from AlphaFold ( Supplementary Fig. 2). Four mutants were selected from previous experiments. Each complex protein (Fig. 7A) was docked against curcumin and its analogues, PGV-1 and CCA-1.1. The molecular docking results showed that in wild-type EGFR, PGV-1 had the lowest docking score of -13.87 kcal/mol and formed one hydrogen bond with Arg686 (Fig. 7B, Table 6). For the E709K and V774M mutant forms of EGFR, curcumin had the lowest binding energy of −11.74 kcal/Mol with three hydrogen bonds (Gly696, Pro699, and Asn700) and −11.94 kcal/Mol with two hydrogen bonds (Asn298 and Arg831), respectively. CCA-1.1 showed the lowest docking scores of −11.29 and −12.62 kcal/Mol in the T263P and L861Q mutant forms of EGFR, respectively. Interestingly, for all mutant forms, CCA-1.1 showed better binding affinity than PGV-1 (Table 6). CCA-1.1 also had much stronger binding activity (ΔG = -12.62 kcal/Mol) for the L861Q mutant than wild-type EGFR, while PGV-1 did no show a difference between mutant and wild-type EGFR. These results show that CCA-1.1 performs better than PGV-1 in the inhibition of mutant EGFR (E709K, T263P, V774, and L861Q). Taken together, these results indicate that CCA-1.1 can inhibit many EGFR variants. The results of molecular docking were validated using MD simulation. As the representative, we chose the binding pocket of EGFR L861Q in complex with curcumin, PGV-1, and CCA-1.1. After a 1-ns simulation, CCA-1.1 displayed a minor change in the position and binding trajectory with mutant EGFR L861Q, which indicates the most stable interaction (Fig. 7C). In the presence of PGV-1, the binding pocket of mutant EGFR L861Q showed more change in position than in CCA-1.1. Further, a more dynamic change was observed with curcumin, which clarified the less stable interaction of curcumin and PGV-1 than that of CCA-1.1 (Fig. 7C). Higher-binding stability of CCA-1.1 compared with that of PGV-1 and curcumin was also demonstrated by the RMSD value of each compound after the 1-ns MD simulation. CCA-1.1 demonstrated a stable RMSD value around 1.8 nm (Fig. 7D). An increase in the RMSD value up to 2.4 and 4.6 nm was shown by PGV-1 and CCA-1.1, respectively, which demonstrates a less stable binding interaction (Fig. 7D). The results of the MD simulation confirmed the validity of the molecular docking study, indicating CCA-1.1 as the most effective EGFR inhibitor.

Discussion
We identified four targets of CCA-1.1 in GBM (i.e., TP53, EGFR, AKT1, and CASP3) by an integrative bioinformatics analysis. TP53 encodes the P53 protein, a tumor suppressor that inhibits cancer cell proliferation and promotes apoptosis 69 . TP53 is frequently mutated in GBM, and these mutations are mainly deletions, affecting P53 function and thereby triggering cancer progression. We also detected copy number gains, suggesting an increase in p53 expression. Both curcumin and PGV-1 compounds have been shown to increase p53 expression in breast cancer cells 70 . Further studies of changes in p53 expression in response to CCA-1.1 treatment in GBM are needed to support the findings of this study.  www.nature.com/scientificreports/ the PI3K/AKT pathway, which regulates cell proliferation and survival 74 . The dysregulation of AKT is common in cancer, with reports of epigenetic modifications, mutations, and overexpression 75,76 . PI3K/Akt is a highly targeted pathway for glioblastoma therapy 77 . Several previous studies have explored the AKT-targeted anticancer effects of curcumin and its analogs. Curcumin may be effective in combination with TMZ in GBM 78 . Yin reported that curcumin increases the effectiveness of temozolomide against U87 glioblastoma cells by increasing ROS levels, inhibiting AKT/mTOR signaling, and promoting apoptosis 79 . Curcumin inhibits GBM via the pRb, p53, JAK/ STST, MAPK, PI3K/Akt, and NF-κB pathways 80 . Another analog of curcumin, C-150, inhibits GBM progression by targeting the NF-κB, Notch, and Akt pathways 81 . Previous research on curcumin and PGV-1 has shown that these compounds inhibit PI3K/AKT signaling in breast cancer cells and colon cancer cells. PGV-1 inhibits NF-κB activation 82 which is related to the PI3K/Akt pathway. Elucidating the mechanism by which CCA-1.1 influences the PI3K/AKT pathway will provide a scientific basis for its utilization as an anti-GBM agent. CASP3 encodes caspase 3, which contributes to the final steps in apoptosis, and is also called an executioner caspase 83 . Increased caspase-3 expression in triggers GBM cell death 84 . The inhibition of caspase-3 in brainresident immune cells promotes GBM progression 85 . Previous studies have shown that both curcumin and PGV-1 trigger apoptosis by increasing caspase expression. The curcumin analogs PGV-0 and PGV-1 stimulate the apoptosis of T47D breast cancer cells by the activation of Caspase-3 86 . Further studies of the effect of CCA-1.1 on caspase 3 expression and activity are needed.
EGFR encodes the human epidermal growth factor receptor, a member of the tyrosine kinase receptor family 87 . Mutations in EGFR activate EGFR signaling, which triggers proliferation and survival in GBM 88 . EGFR mutations have been found in 53% of patients with GBM 68 , including gains or amplifications, suggesting an increase in EGFR expression. Several compounds successfully inhibit EGFR signaling, for example, Higenamine 89 , 20(R, S)-protopanaxatriol, a metabolite from protopanaxatriol ginsenosides 90 , and Tubeimoside-I, which increases the sensitivity of glioblastoma cells towards temozolomide 91 .
Extensive research has focused on the effects of curcumin and its analogs targeting EGFR in cancer cells. Curcumin inhibits EGFR signaling and reduces EGFR expression in cancer cells. Curcumin increases sensitivity to gefitinib by inhibiting EGFR signaling in non-small cell lung cancer 92 . In addition, curcumin enhances the anticancer activity of gefitinib in vitro and in vivo in lung cancer by inducing EGFR degradation 93 . Curcumin downregulates EGFR in colon cancer cells by reducing the transcription factor EGR1 94 . Another study has shown that curcumin inhibits the autophosphorylation of EGFR 95 . Starok et al. showed that curcumin has dual effects on EGFR by inhibiting enzymatic activity of the EGFR tyrosine kinase domain and by entering the lipid bilayer, thus affecting EGFR dimerization 96 . A recent study by Ali et al. has shown that curcumin analog 3c has a greater inhibitory effect on leukemic cells than those of curcumin and gefitinib, and this analog inhibits EGFR activity 97 .
Mutations in the EGFR kinase domain have been shown to cause constitutively active ligand-independent signaling 98 and to affect the sensitivity of glioma cells to temozolomide 99 . E709K is a mutation in EGFR exon 18 responsible for lung cancer cell resistance to gefitinib, erlotinib, AZD9291, and CO1686 100 . It is a rare type of EGFR mutation in lung cancer 101 . The T263P mutation is located in the extracellular domain of EGFR, which leads to ligand-independent signaling activation 102 and tumor progression in GBM 103 . Moreover, the T263P EGFR mutant form has a furin-like cysteine-rich (FU-CR) domain involved in signal transduction, including an important role in promoting Wnt/β-catenin signaling [104][105][106][107] . L861Q is a missense mutation in the EGFR kinase domain of GBM 108 . The L861Q mutation increases kinase activity and tumor progression but does not increase the sensitivity of tumor cells to EGFR tyrosine kinase inhibitors 109 . The EGFR V774M mutation is associated with non-small-cell lung cancer progression 110 and resistance to tyrosine kinase inhibitors 111 . A missense mutation in the EGFR kinase domain, V774M, which leads to amplification, has also been found in Japanese patients with GBM 112 . V774M is considered a functional mutation in lung cancer 113 .
In a molecular docking analysis, CCA-1.1 showed a lower docking score than that of PGV-1 in wild-type and mutant EGFR E709K, T263P, and L861Q and slightly higher docking scores for V774M. The molecular docking results for wild-type EGFR are supported by previous studies. PGV-1 exhibits the weakest interaction with EGFR and HER2 in silico 114 . Interestingly, CCA-1.1 showed a similar or better interaction with EGFR than PGV-1 28 . Further, MD simulation demonstrated a more stable binding interaction of CCA-1.1 during the www.nature.com/scientificreports/ 1-ns simulation compared to the binding of PGV-1 and curcumin. Thus, clarifying the results of the molecular docking study. Therefore, further research on CCA-1.1 targeting EGFR is very important for its development as an anti-GBM agent. www.nature.com/scientificreports/ GBM gene profiling has revealed three GBM subtypes: proneural (TCGA-PN), classical (TCGA-CL), and mesenchymal (TCGA-MES) 115 . GBM subtypes are characterized by abnormalities in platelet-derived growth factor alpha (PDGFRA), isocitrate dehydrogenase1 (IDH1), epidermal growth factor receptor (EGFR), and neurofibromin1 (NF1) 116 . Different subtypes may respond differently to therapies and show differences in the immune microenvironment 117 . Several studies have suggested that mesenchymal GBM is the most immunogenic, proinflammatory subtype, characterized by significant M2 macrophage and neutrophil gene expression 118,119 . Therefore, we expected to observe correlations between the expression of the four hub genes and the level of immune cell infiltration in GBM. In general, immune cell infiltration can be classified into two types: (1) activation of the immune response by pro-inflammatory cells and CD8 + cytotoxic T lymphocytes (CTL) and (2) suppression of the immune response to cancer cells, e.g., by regulatory T cells (Tregs). Considering the complexity of GBM and the presence of the blood-brain barrier, it is plausible that the immune response is strictly regulated, resulting in extensive immune cell infiltration [120][121][122] . Both adaptive and innate tumor-infiltrating immune cells are involved, i.e., B cells, CD8 + , and CD4 + cells as well as macrophages, neutrophils, and dendritic cells (DCs), respectively 123 . AKT1 and EGFR negatively affected CD8 + , while B cells were negatively correlated with CASP3 expression levels (with correlation coefficients of < 0.5). Positive correlations were observed between the expression levels of AKT1, TP53, and EGFR and the frequencies of CD4 + cells and all of the above-mentioned innate immune cells. CASP3 expression was positively correlated with DCs. Despite the low frequency of fibroblasts in the healthy brain, CAFs are found in GBM 124,125 . Here, we found that CAFs are positively related to AKT1, TP53, and CASP3 expression. Mu et al. reported that CD4 + plays a role in angiogenesis and the progression of GBM 126 . We propose that targeting the four newly identified gene candidates may be an effective approach to alter the immune response to cancer.
The cytotoxicity assay of CCA-1.1 and TMZ showed that CCA-1.1 has a better cytotoxicity than TMZ based on the IC50 values, in which the cytotoxicity against U87 cells with an IC50 value are 9.8 uM for CCA-1.1 and 40 uM for TMZ, indicating high potency of CCA-1.1 for GBM therapy. DEGs showed that among the potential target genes, only EGFR showed significant results, in which the EGFR transcript variant 8 was upregulated in CCA-1.1 treated U87 cells, whereas EGFRvIII was downregulated in U87 cells after treatment with CCA-1.1., indicating the important role of EGFR in the cytotoxicity of CCA-1.1. A previous study showed the heterogeneity of EGFR in glioblastoma cells, also referred to as EGFR truncation variants 127 . Moreover, genetic amplification and mutations in EGFR are the most common oncogenic events in GBM 128 . EGFR is encoded by the EGFR gene, producing mRNA transcript EGFR variant 1, which produces isoform a. In addition to isoform a, EGFR produces several alternatively spliced transcript variants 129 . Several mRNA variants encode EGFR isoforms, such as variants 1 and 8. EGFR transcript variant 1 encodes the full-length protein of EGFR, while variant 8 encodes a shorter protein. A previous study stated that all isoforms encoded by all EGFR variants could interact with their ligand, namely epidermal growth factor (EGF) 130 . Furthermore, Weinholdt explained that only the EGFR1 isoform had been widely studied for its biological function 131 . EGFRvII is an oncogenic EGFR that is responsible for sensitivity to tyrosine kinase inhibitors 127 .
EGFRvIII is an interesting therapeutic target in GBM therapy because EGFRvIII is present in 25-30% of the glioblastoma cell population 132 . EGFRvIII undergoes a 6-273 amino acid deletion at exon 2-7, encoding the extracellular domain of EGFR 133 , and EGFRvIII can undergo dimerization via a ligand-independent activation pathway 132 . EGFRvIII differs from mutant EGFR on the extracellular domain, namely due to the deletion of This study had several limitations. First, the protein targets of CCA-1.1 were curated or predicted using public databases based on a particular algorithm. Second, the results of the bioinformatics analyses need to be validated by in vitro and in vivo assays as well as clinical trials. Nevertheless, the results of this study are expected to accelerate the development of drugs for GBM.

Conclusion
Using an integrative bioinformatics approach, four CCA-1.1 targets in GBM were obtained: TP53, EGFR, AKT1, and CASP3. In addition to the potential therapeutic effects of CCA-1.1 mediated by these four proteins and the inhibition of signaling pathways, it also has the potential to modulate the immune environment. A cytotoxicity  www.nature.com/scientificreports/ assay showed that CCA-1.1 has a better cytotoxicity than TMZ with an IC50 value of 9.8 μM compared 40 μM for TMZ. DEGs showed that among the potential target genes, only EGFR showed significant results, in which the EGFR transcript variant 8 was upregulated, whereas EGFRvIII was downregulated in U87 cells after treatment with CCA-1.1. Molecular docking results revealed that CCA-1.1 can inhibit many EGFR mutants in GBM. Further, MD simulation revealed that the binding of CCA-1.1 with the mutant EGFR L861Q is the most stable compared to those of curcumin and PGV-1. These findings require further confirmation with laboratory experiments and clinical trials for the development of GBM therapies.

Data availability
All data produced by this study are disclosed in the manuscript and additional files. The raw data of gene expression can be accessed at the Gene Expression Omnibus (GEO, http:// www. ncbi. nlm. nih. gov/ geo/), using accession number GSE206241.