1. Introduction
The formation of new blood vessels from existing vessel in the postnatal life, referred to as angiogenesis, is an important process in both physiological and pathological conditions. It is a tightly regulated process involving the interplay of a number of pro- and antiangiogenic factors. Dysregulation of the balance between these factors leads to excess or inhibited angiogenesis, contributing to different pathological conditions [
1,
2,
3]. Tumors cannot grow beyond a certain size unless they are vascularized to supply oxygen and much-needed nutrients for their growth. Angiogenic growth factors, such as Vascular Endothelial Growth Factor (
VEGF), Fibroblast Growth Factor (
FGF), Transforming Growth Factor (
TGF), and Epidermal Growth Factor (
EGF) [
4], play an important role in promoting tumor angiogenesis and growth. Among them,
VEGF is the key endothelial cell-specific mediator of angiogenesis. It induces angiogenesis by increasing the endothelial permeability, Endothelial cell (EC) proliferation, migration, survival, cell–cell contact and lumen formation.
VEGF exerts its effects through binding with the cell surface receptors, a family of trans-membrane tyrosine kinase receptors. Its interaction with the receptor on the cell surface triggers the activation of intracellular signaling pathways and expression of a number of genes that modulate different cellular events critical to angiogenesis [
5,
6]. Since tumor angiogenesis is vital for growth, targeting it is considered a potential therapeutic strategy to inhibit tumor growth and development [
1,
7].
Bevacizumab, a humanized immunoglobulin G1 (IgG1) monoclonal antibody that selectively binds with a high affinity to human
VEGF and neutralizes its biological activity, is one such anticancer agent [
8]. The mechanisms of such anticancer effects include the direct inhibition of tumor-associated angiogenesis. However, recently, it has been noted that tumors develop a resistance to such anti-
VEGF therapy and form capillaries, apparently through some alternative mechanism. This may be due to the activation of other pathways that have a proper connection with the downstream signaling of
VEGF-mediated angiogenesis [
9,
10].
Glioblastoma is a common type of aggressive malignant brain tumor in adults characterized by histopathologic features involving necrosis and endothelial proliferation. These tumors arising from glial cells may be grouped as (i) grade I—pilocytic astrocytomas, pleomorphic xantho astrocytomas, and subependymal giant cell astrocytomas; (ii) grade II—oligodendrogliomas and astrocytomas; (iii) grade III—anaplastic oligodendrogliomas, anaplastic astrocytomas, anaplastic oligoastrocytomas, anaplastic ependymomas; and (iv) grade IV—the glioblastoma multiforme (GBM) [
11]. Even though enormous therapies have been developed, the survival rate of GBM patients has not remarkably changed, and the five-year survival rate is 5.1% [
12].
GBM is associated with excessive and aberrant angiogenesis, and it is characterized by rapid angiogenesis-dependent (re)growth, cell heterogeneity and extensive local tissue infiltration. With regard to treatment strategies, radiotherapy becomes ineffective, since GBM infiltrates the surrounding tissues and its complete resection is impossible. Further, the blood–brain barrier makes treatment more difficult and tumor cells found in the areas of hypoxia are resistant to radiotherapy. Bevacizumab, apart from inhibiting tumor angiogenesis by blocking VEGF, caused a disruption of the glioma stem cell microvascular niche and improved vascular normalization. However, glioma is quite often refractory to anti-VEGF therapy, and the molecular mechanisms underlying the development of drug resistance in GBM patients are not well-understood.
In recent years, high-throughput approaches have been developed to capture differentially expressed genes in various conditions, including drug resistance. Microarray-based gene expression profiling and sequence-based techniques like the RNA-seq analysis provide useful information about the differentially expressed genes, key pathways and the signature genes with respect to different conditions. Most of these datasets are now publicly available. Therefore, gene expression data-based computational approaches can be employed to characterize the genetic alterations at the genome level, which helps to identify differentially expressed genes and their possible physiological or pathological relevance. In the present study, the computational approach of expression data analysis is employed for identifying the potential genes responsible for the resistance to anti-VEGF therapy in glioblastoma. A number of studies have been conducted to examine the gene expression profiles of GBM patients compared with healthy controls and are made available in databases like the NCBI-GEO (National Centre for Biotechnology Information-Gene Expression Omnibus Database).
In this study, we analyzed the microarray datasets downloaded from the NCBI-GEO database of glioblastoma xenografts that developed a resistance against bevacizumab treatment and compared them with glioblastoma xenografts without bevacizumab treatment to examine whether the gene expression differed during the development of resistance to anti-
VEGF therapy. Orthologous xenograft models are suitable to study GBM formation, progression, and investigation for potential therapeutics [
13]. The data from three generations of glioblastoma xenografts were used to examine the changes in gene expression relating to tumor growth and angiogenesis in bevacizumab resistance tumors and, also, to understand whether the changes were similar, different or further changes occur as the tumor progresses, sustaining the resistance to therapy. The DEGs (differentially expressed genes) were identified from bevacizumab-treated and untreated samples. Analyses of Gene Ontology (GO) enrichment, protein–protein interaction (PPI) network, Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway of DEG, and survival helped screen hub genes and their possible involvement in anti-
VEGF resistance in GBM.
4. Discussion
Tumor angiogenesis, being a critical process for tumor growth and progression targeting, is recognized as an important therapeutic strategy. Though anti-VEGF therapy using inhibitors of VEGF action has been reported to arrest tumor growth in several types of cancers, in certain cases, resistance to antiangiogenic treatment, particularly against anti-VEGF therapy, has been reported. Glioblastoma is one such tumor that shows resistance to anti-VEGF therapy. In our effort to understand the molecular basis of resistance to anti-VEGF therapy in glioblastoma, we analyzed the gene expression data in xenografts from anti-VEGF-resistant GBM, using bioinformatics tools, and the results suggested that the cells adapt to such conditions by changing gene expression and restoring angiogenesis. This is evidenced by the following observations: (a) The analysis of the microarray data from fourth generation xenografts of anti-VEGF-resistant GBM patients showed the upregulation of 359 genes and downregulation of 514 genes, indicating differences in gene expression during the development of anti-VEGF resistance. (b) The GO function and pathway enrichment analysis of DEG showed significant enrichment in the biological processes such as cell proliferation, cell migration and angiogenesis, indicating the ability to acquire angiogenic phenotypes. A further analysis of the DEGs showed enrichment in the molecular functions such as receptor binding and growth factor activity and the signaling pathways such as TNF signaling pathway, PI3-AKT pathway, and cytokine receptor pathway, particularly in upregulated DEGs. (c) The PPI network analysis showed enrichment in the key angiogenic pathways, such as the HIF1 pathway, PI3-AKT pathway and cell cycle pathway, critical in angiogenesis and cancer development. (d) Among the DEGs, several hub genes, including IL6, VEGFA, and SRC were identified. The survival analysis showed that the high expression of three hub genes were associated with a shorter overall survival time of GB patients.
Identification of DEGs in the resistance condition provides valuable information about the mechanism of resistance. In the present study, the gene expression profile dataset, (anti angiogenic therapy resistance condition) GSE81465 from GEO was analyzed to obtain DEGs. The GO biological process analysis revealed that several of the upregulated DEGs were functionally enriched in the process of cell proliferation, migration, cell adhesion and angiogenesis, confirming that the cells were resistant to the bevacizumab therapy and equipped to develop new vessels needed for tumor growth. Further, the downregulated DEGs were enriched in tumor suppressive pathways that regulate cell cycle and signal transduction by
p53, indicating that the tumor growth was not arrested by anti-angiogenic therapy. Previous studies reported that anti
VEGF therapy only seizes
VEGF and it does not block other molecules involved in angiogenesis pathway that leads to the cell proliferation, migration and survival [
22]. It was observed that 18 angiogenesis related genes were upregulated, among them
VEGF and
TGFA are growth factors and
NRP2 is a receptor for
VEGFA. Several of these DEGs encode proteins that are reported to affect growth and characteristics of the GBM. For instance,
EPAS1/HIF2A is a hypoxia responsive transcription factor, the over expression of it in glioblastoma enhances the tumor aggressiveness [
23]. Many aggressive aspects of GBM such as cell proliferation and poor prognosis are highly correlated with the expression of
PTGS2 [
24,
25,
26] and it is overexpressed in radiation resistance glioma [
27]. High expression of
TNFRSF12A has been reported in GBM [
28] and is also involved in glioma cell migration, invasion, and resistance to chemotherapeutic agents. Temozolomide- resistant GBM shows high expression of
TNFRSF12A and greater migratory capacity [
29].
CXCL8/IL8 is a multifunctional cytokine which enhances the vascular permeability in GBM [
30,
31]; high expression of both
VEGFA and
CXCL8 can reduce the overall survival rate of GBM patients [
32].
ADGRG1/GPR56 is a
GPCR involved in adhesion signaling and
HIF1A is a transcription factor, which has a critical role in GBM survival, resistance and invasion [
33]. Recent studies showed that
SRPX2 promotes epithelial to mesenchymal transition in GBM, and it’s over expression induced TMZ resistance in GBM [
34]. In GBM,
EREG enhances the phosphorylation of
EGFR, thus activates
EGFR signaling and directs cancer cell proliferation [
35].
CLIC4, which is a key element in the apoptotic response to oxidative stress, is highly expressed in GBM [
36].
Further analysis of the DEGs identified 15 upregulated genes associated with growth factor activity, including
CSF3,
IL6,
OSGIN2,
FGF13,
IL11,
TIMP1,
LIF,
BDNF,
EREG,
CLCF1,
VEGFA,
TGFA,
HBEGF,
NRG1 and
FGF2. Among these,
CSF3,
IL6,
IL11,
LIF and
CLCF1 are cytokines. In that,
FGF2 and
FGF13 showed high expression in GBM samples. Previous studies indicated that
FGF13 regulates GBM cell invasion and bevacizumab-induced glioma invasion [
37,
38,
39]. Seven proto-oncogenes (
FYN,
MLLT11,
PDGFRB,
BCL6,
SRC,
CBLB and
CRKL) were also upregulated in GBM, agreeing with the previous studies [
40,
41,
42,
43,
44,
45,
46,
47].
Pathway enrichment analysis showed that the upregulated DEGs were enriched in the cancer related pathways, suggesting that the genes involved in these pathways might be responsible for the formation of resistance to anti
VEGF therapy in GBM. Sixteen genes in
PI3K-Akt pathway were upregulated to suggest that the pathway was activated in GBM, possibly causing the suppression of cell death and increasing cell survival [
48,
49,
50,
51,
52,
53]. Twenty genes in cancer pathways were upregulated; these genes are known to have important role in biological processes such as angiogenesis, cell invasion, cell proliferation, apoptosis and mobility [
52,
53]. In the context of the reported role of cytokines in the glioma formation, data showing upregulation of 14 genes in the cytokine receptor interaction might be important in the induction of resistance. Upregulation of 13 genes encoding components of
MAPK network, which is severely altered in GBM [
54], has also been observed. Ten genes involved in focal adhesion were also significantly upregulated during the drug resistance condition.
Further analysis of the DEGs in terms of the nature and distribution of the proteins encoded by these genes revealed that several of them belonged to classes of glycoproteins, secretory proteins, membrane associated proteins and intracellular proteins. Alterations in glycoproteins, particularly changes in their nature and distribution have been known to play a key role in tumor development as well as resistance to drug treatment [
55]. In the current study we observed that 115 glycoproteins were upregulated, most of which are present on the cell surface that may also act as a ligand for the cell surface receptor. These glycoproteins are involved in several cellular processes such as cell growth, cell-cell recognition, and cell migration, critical in angiogenesis. A possible alteration of the structure and therefore, their function was indicated by the identification of 56 DEGs that encode enzymes related to glycoprotein metabolism including 5 glycosyl transferases.
VEGF receptor, a glycoprotein, may recognize other proteins in the absence of
VEGF and triggers the downstream signaling. Furthermore, studies also reported that interaction of galectin-1 and
VEGFR2 activate
VEGF-like signaling in tumor angiogenesis [
56]. In this context, it is also pertinent to note that glioma cells employ different metabolic strategies including aerobic glycolysis, pentose phosphate pathway, one carbon metabolism, fatty acid metabolism which contribute to energy production in glioma cells and several bioenergetics pathways are linked to oncogenic signals such as
AMPK and
MTOR pathways [
57,
58].
One of the signaling pathways altered during the development of resistance to anti-
VEGF therapy in GBM appears to be
BMP signaling pathway. It is a complex network of receptors, ligands and antagonists which may dynamically impact GBM growth, maintenance and progression.
GREM1 is an antagonist of
BMP signaling. Glioma stem cells secrete
GREM1 to promote tumorigenesis through inhibition of
BMP signaling. Studies reported that the secretion of
GREM1 contribute to treatment resistance by maintaining cellular proliferation and cellular hierarchies within the tumor, and also increasing resistance to differentiation therapy [
59]. The data presented here showed that gene encoding
GREM1 is upregulated about 3-fold, though
BMPs and
TGFB were slightly downregulated. However, certain other genes which are known to modulate angiogenesis did not show any significant change. For instance,
TSGA10 (testis specific Gene Antigen 10), which acts as a tumor suppressor in many types of human cancers [
60] and inhibits
VEGF-induced angiogenesis [
61] was not differentially expressed in anti-
VEGF resistant condition. Further, recent RNA seq analysis of anti-
VEGF resistant ovarian cancer model showed upregulation of apelin/
APJ receptor signaling pathway [
62]. However, this receptor-ligand pair gene expression was not altered in anti-
VEGF resistant glioma described in the present study, probably suggesting that mechanisms underlying anti
VEGF resistance are different in different tumors.
Further analysis of DEGs in anti-
VEGF resistant microarray data sets revealed that 19 ligand receptor pairs were differentially expressed. The receptors
CD44,
F3,
IL6ST,
ITGB1,
NRP2,
PLAUR and
EGFR were upregulated.
CD44 is a trans-membrane glycoprotein receptor of hyaluronic acid which is overexpressed in GBM and enhances the GBM invasion, proliferation and therapy resistance [
63]. It is also involved in epithelial-mesenchymal transition, angiogenesis, proliferation, invasion, and migration [
64]. Genes encoding the ligands for
CD44 such as
SPP1,
HBEGF and
FGF2 were also upregulated. Another important signaling molecule involved in GBM is
EGFR whose ligands such as
EREG,
FGF13,
HBEGF,
TGFA and
VEGFA were also upregulated.
IL6 and
TFPI, ligands of
F3 receptor were also upregulated. Four ligands of
IL6ST and
ITGB1 receptors were upregulated and in the case of receptors
NRP2 and
PLAUR, one ligand each was upregulated. These 12 ligands included factors with growth factor (10 genes) and cytokine activity (5 genes). In this context, earlier data on alteration in
NRP1 expression and activation of
TGFB signaling restoring angiogenesis in anti-
VEGF resistant GBM is particularly relevant [
65]. Therefore, it appears that instead of the principal ligands of several of these receptors, the upregulated ligands were alternate ligands, suggesting development of alternate mechanisms for angiogenesis and tumor growth. In this context it is important to note that, though not all glioblastoma patients are resistant to anti-
VEGF therapy, the possibility of angiogenesis-independent tumor progression by diffuse invasion of single tumor cell in brain, as reported recently [
66].
PPI network was developed, including both up- and downregulated genes, to verify the interaction between these genes and how they are coordinately involved in the formation of resistance. Many genes with high connectivity in the PPI network were enriched in pathways in cancer,
PI3K-Akt signaling pathway, Proteoglycans in cancer and
MAPK signaling pathway. We have identified 21 hub genes with hybrid centrality score > 12, among which 10 were up- and 11 downregulated. The possible role of these hub genes in the development of anti
VEGF resistance in GBM was suggested from the data showing interaction of 18 of these hub genes with 58 genes of different network modules in the
VEGF- mediated angiogenesis signaling pathway [
6,
67]. In this context, our previous report on multiple phytochemicals of a poly-herbal formulation targeting multiple components of
VEGF-
VEGFR2 pathway and inhibiting angiogenesis is particularly significant [
68].
KEGG pathway analysis and GO enrichment analysis also demonstrated that these hub genes were associated with pathways in cancer and significantly involved in positive regulation of angiogenesis and negative regulation of apoptosis.
VEGF pathway analyses revealed that nine upregulated hub genes (
IL6,
EGFR,
VEGFA,
SRC,
CXCL8,
PTGS2,
IDH1,
APP and
SQSTM1) and five downregulated hub genes (
POLR2H,
RPS3,
UBA52,
CCNB1 and
UBE2C) are linked with several of the
VEGF signaling pathway components. Across all the generations six (
IL6,
CXCL8,
PTGS2,
IDH1,
POLR2H,
UBA52) hub genes and in fourth and nine generation 11 (
IL6,
CXCL8,
PTGS2,
IDH1,
APP,
SQSTM1,
POLR2H,
RSP3,
UBA52,
CCNB1,
UBE2C) hub genes were differentially expressed, and the maximum fold change was observed in fourth generation. Studies also reported that high expression of
IL6 [
69], and
EGFR [
70], had worst survival outcome than low expression. Chang et al. reported that GBM patients with lower
IL6 expression showed longer survival time and a few patients with longer survival time did not show significant expression of
IL6. [
71].
PTGS2 is another hub gene which was enhanced in radiation resistant glioma cells [
27]. Ribosomal protein S3 is suggested to be a substrate for induction of radio- resistance in glioblastoma [
72].
Further evidence linking these hub genes with the development of resistance, was provided by survival analysis which revealed that out of the 10 upregulated hub genes, expression of three genes (
VEGFA,
CXCL8,
IDH1) was statistically significant and were associated with a worse prognosis among patients with GBM. However, expression levels of none of the downregulated hub genes, including six genes whose downregulation is known to relate with low survival and five genes whose downregulation is associated with longer survival [
20], showed any statistically significant association with survival. Altered expression of these hub genes has been reported in GBM.
CXCL8 was upregulated in GBM compared to diffuse astrocytoma and its expression levels were positively associated with progression and poor prognosis of glioma [
73]. As discussed before
VEGF plays important role in angiogenesis and its expression is high in GBM patients compared to the healthy subjects [
74]. Hub gene
IDH1 mutation in GBM patients showed a longer survival rate compared to the wild-type [
75]. However, the potential role of these predicted hub genes need to be further examined experimentally. The lack of any clinically established biomarker in glioblastoma, unlike in other tumors makes it difficult to follow response to anti-angiogenic therapy and survival of malignant glioma. The expression of angiogenic target molecules and also patterns of tumor vascularization did not predict response to bevacizumab [
76] highlighting the need for reliable predictive biomarkers. The results presented here predict that the hub genes associated with the GBM resistance to bevacizumab may be a potential therapeutic target or biomarker of anti-
VEGF resistance of GBM.