Molecular biomarker‐guided anti‐angiogenic targeted therapy for malignant glioma

Abstract Despite aggressive multimodality treatment, the prognosis of glioma, especially malignant glioma, remains very poor. After decades of effort, anti‐angiogenic therapy has become an important method of cancer treatment in addition to surgery, radiotherapy and chemotherapy. Although the performance of anti‐angiogenic therapy in colorectal cancer is good, its performance in malignant glioma remains unsatisfactory. Several phase III clinical trials showed no overall survival benefits. To solve this problem, the division of patients into groups based on their molecular biomarkers is an important step. This paper provides current insights into anti‐angiogenic drugs undergoing clinical trials and discusses the potential of molecular biomarkers to guide glioma diagnosis.


| INTRODUC TI ON
Glioblastoma (GBM) is the most common and deadly brain tumour, with an annual incidence of 3.19/100 000 1 Currently, the preferred treatment for GBM is maximal safe surgical resection followed by a combination of radiation therapy and chemotherapy with temozolomide (TMZ). Nevertheless, GBM still has a grim prognosis, with a median overall survival (mOS) ranging from 14.6 to 20.5 months. [2][3][4][5] The outcome is much worse in elderly patients, who have an average survival of less than 8.5 months. 6 In 2004, the FDA commissioner announced that 'antiangiogenic therapy can now be considered the fourth modality of cancer treatment' in addition to surgery, radiotherapy and chemotherapy. 7,8 Angiogenesis is necessary for the growth of solid tumours and their escape from a hypoxic environment. 9 Tumours that acquired a blood supply were observed to undergo a 19 000-fold burst in tumour volume, which is consistent with the hypothesis of an 'angiogenic switch'. 10 Compared with other human neoplasms, GBM was suggested to be uniquely susceptible to angiosuppression because of its exceedingly high extent of neovascularization. 11 VEGF and the VEGF pathway are the first targets for angiogenesis-directed therapeutics. 12 Vitro experiments have demonstrated that anti-angiogenic drugs bind specifically to VEGF and prevent its interaction with its receptors. Thereby these drugs destroy preexisting tumour blood vessels and cutting off the supply of oxygen and other nutrients required for tumour cell growth as well as inhibiting tumour neovascularization, which in turn inhibits the growth and metastasis of tumour cells and effectively improving the transport of chemotherapy drugs. 13 The array of clinically useful angiogenic inhibitors is now expanding beyond the VEGF pathway to include inhibitors of placental growth factor, integrin and other molecules.
Although anti-angiogenic therapy has exhibited excellent efficacy against other human tumours, such as colorectal cancer and non-small cell lung cancer and GBM showed high extent of neovascularization, its performance in GBM therapy is not satisfactory. 14,15 For newly diagnosed GBM, two multi-centre, randomized phase III clinical trials (the AVAglio study and the Radiation Therapy and Oncology Group [RTOG] 0825 trial) showed benefits only in progression-free survival (PFS), not in overall survival (OS). [16][17][18] These results suggest that not all patients benefit from anti-angiogenic therapy, and furthermore, there are no definitive biomarkers predicting the benefits of anti-angiogenic therapy in GBM. Therefore, finding biomarkers that can identify patients who are more likely to benefit from anti-angiogenic treatment is very important.
This article will summarize potential biomarkers that can predict the benefits of anti-angiogenic treatment for GBM and provide diagnostic information and will describe our expectations for the near future.

| B I OMARK ER S FOR B E VACIZUMAB
In 1993, bevacizumab (BEV) was synthesized for the first time. 19 As an anti-angiogenic monoclonal antibody, BEV can slow the growth of new blood vessels in tumours by inhibiting VEGF-A. 20 As discussed above, two phase III trials combining BEV with radiotherapy and chemotherapy displayed no OS benefit. 17 Molecular markers that can predict the effect of BEV on GBM therapy are urgently needed.

| Isocitrate dehydrogenase 1 wild-type proneural subtype
Better clarification of the roles of key genes has helped to classify gliomas into different molecular subtypes based on their molecular markers, providing new ideas for the clinical diagnosis and prognosis of gliomas.
Due to the heterogeneity of gliomas, patient outcome may vary across different subtypes. For example, among the four GBM subtypes in Phillips's classification, 21 the proneural subtype was originally considered the subtype with the longest OS, but this result was later attributed to a small subset of patients with an isocitrate dehydrogenase 1 (IDH1) mutation. 22 In addition, IDH1 wild-type proneural tumours had the worst prognosis among all GBM subtypes in the placebo arm.
Based on these facts, a retrospective study of the AVAglio trial compared the effects of BEV treatment on proneural GBM and three other subtypes of GBM (including only wild-type IDH1). 16,23 The results showed beneficial effects on PFS in the proneural (9.

| IGS-18
In addition to IDH1 wild-type proneural patients, IGS-18 patients (as defined by Gravendeel) also benefited from BEV treatment. 26  In GBM patients treated with a combination of BEV and CCNU, a higher FM04/OSBPL3 expression level was associated with a significantly increased mOS (6.1 vs 12.4 months, P < 0.0001) ( Table 1). 27

| Matrix metalloproteinases
Matrix metalloproteinases (MMPs), also known as matrixins, are calcium-dependent zinc-containing endopeptidases, which play major roles in cell behaviours such as proliferation, migration, differentiation, angiogenesis, apoptosis and host defence. 28 A report published in Neuro-Oncology 2013 showed the predictive value of serum matrix metalloproteinase 2 (MMP2) levels by investigating the relationship between recurrent high-grade glioma (HGG) and serum MMP2 levels (Table 1). 29 In the initial cohort (cohort 1), patients treated with BEV and irinotecan were divided into two groups according to their response. Most patients with increased serum MMP2 levels were found to be responders (10/12

| MicroRNA
MicroRNA expression plays an important role in the tumourigenesis, infiltration and deterioration of glioma. The role of microRNA in glioma physiology and its high stability in clinical samples indicate that microRNA may be a primary candidate as a predictive biomarker. 31 Many researchers have sought to determine the predictive value of microRNA.
A 2016 study used TCGA data to investigate the relationship between microRNA expression and the effect of BEV treatment in GBM patients. 32 The study identified and validated the predictive value of an 8-microRNA profile for the therapeutic effect of BEV.
The responder group was defined as having a response score >0, while the non-responder group had a score <0.
The defined cut-off response score of 0 was used to separate a total of 37 test set samples into two groups, responders and non-responders. The OS of the responder group was significantly longer than that of the non-responder group (mOS 21 vs 15 months, HR = 0.34, 95% CI = 0.11-1.01, P = 0.026). The researchers also calculated the response scores for all 473 GBM patients in the TCGA database (treated with various regimens excluding BEV) and found no significant difference between the responder group and the non-responder group (Table 1). These results indicated that the predictive value of the 8-mi-croRNA algorithm was BEV specific. Unfortunately, the 8-microRNA spectrum did not significantly correlate with PFS and the validation of this assay will require assessment in larger cohorts.
In addition, recent evidence has shown the predictive value of microRNA-21 and microRNA-10b. 33 A meta-analysis of previous

| Peripheral neutrophil count
Preclinical studies have shown that neutrophils may promote tumour neovascularization. Therefore, a high neutrophil count was hypothesized to be associated with a better response to anti-VEGF therapy. 35 To address this possibility, researchers investigated the predictive role of the peripheral blood neutrophil count before BEV treatment on the efficacy of BEV. 36 A total of 256 GBM patients have been included in the analysis since 2006 and the best cut-off for the baseline neutrophil count was found to be 6000/mm 3

| Angiotensinogen and human leucocyte antigen class II
Angiotensinogen (AGT) and all components of the renin-angiotensin system are expressed in GBM. 37 In addition, increased AGT expression was associated with a higher level of vascular proliferation.
In a study published in 2016, gene expression in tumour tissue was analysed in recurrent GBM (rGBM) patients who were responsive to BEV/irinotecan combination therapy. The analysis was conducted using a platform covering 800 genes to identify predictive biomarkers for BEV response in rGBM patients. Multivariant logistic analysis and Cox regression analysis were also performed for candidate genes with possible predictive value. 38 The results showed that  Table 1).
In addition to AGT, single-variant analysis showed that high gene expression of human leukocyte antigen (HLA) class II was significantly associated with prolonged OS (P = 0.03) and better treatment response (twofold increase in HLA class II: OR = 1.22, 95% CI = 1.01-1.47, P = 0.04) but was not associated with PFS (P = 0.16).
However, this association was not detected by multivariant analysis.  (Table 1). However, this study was limited by its small sample size and its results should be further verified by larger studies.

| Predictive value of hypertension after BEV treatment
VEGF, via binding to VEGF receptor (VEGFR), can stimulate endothelial cells to produce NO, which leads to vessel dilatation and a decrease in arterial blood pressure. BEV can inhibit VEGF signalling and indirectly lead to an immediate increase in blood pressure.
Therefore, hypertension can give indirect information about the effect of BEV and might be a prognostic factor for treatment.
An article published in Cancer 2014 reported the interesting finding that drug-induced hypertension might have predictive value for the effect of BEV treatment in rGBM patients. 40 A total of 82 rGBM patients who received BEV therapy after standard treatment were included in the study. Patients with no history of hypertension were divided into two groups: patients with post-BEV treatment systolic pressure >140 mm Hg or diastolic pressure >90 mm Hg were placed in the hypertensive group and others in the normotensive group. The PFS and OS for the two groups showed a marked difference (PFS: hypertensive 6.7 vs normotensive 2.5 months, P < 0.001; OS: hypertensive 11.7 vs normotensive 4.9 months, P < 0.001) ( Table 1). This result suggested that drug-induced hypertension was associated with better outcome after BEV treatment, which was supported by the results from other malignancies.

| B I OMARK ER S FOR ENZ A S TAURIN
Enzastaurin, a cyclic bisindole maleimide, is an oral serine/threonine kinase inhibitor that specifically inhibits the protein kinase C and phosphatidylinositol 3-kinase and protein kinase B (PI3K/AKT) signalling pathways, leading to cell apoptosis, the inhibition of cell proliferation and anti-tumour-induced angiogenesis. 45,46 In 2006, the EU and FDA approved enzastaurin for the treatment of GBM. 47 In 2010, a phase I/II clinical trial showed that enzastaurin exhibited some anti-tumour activity against recurrent gliomas, but it cannot be used for monotherapy. 48

| The predictive value of phosphorylated glycogen synthase kinase-3β
Phosphorylated glycogen synthase kinase-3β (pGSK3β), which may be an effective biomarker for enzastaurin, was discovered by the detection of GSK-3 phosphorylation in peripheral blood mononu- However, many problems remain to be solved in using molecular biomarkers to predict the therapeutic effects of anti-angiogenic drugs. The first problem is the heterogeneity and dynamic changes in tumours. Ideally, the therapeutic effect can be predicted by examining tumour tissue specimens or biomarkers in the circulating blood before treatment. However, not only primary tumours may be different from metastatic tumours, but the progress and treatment of a tumour may lead to changes in the tumour's biological characteristics. In addition, tumours, especially gliomas, are highly heterogeneous: different regions of the tumour may differ in their molecular characteristics. Therefore, dynamic biomarkers must be established, which leads to another problem-repeated biopsy.
Fortunately, as imaging development and research continue, the use of imaging markers, such as relative cerebral blood volume, K trans and others, to predict the effect of treatment is also under development.
Some low-cost biopsies can also address this problem through the analysis of markers in the circulating blood, the urine or a combination. In addition to the problems involving markers themselves, problems in the research methods cannot be ignored. The current studies are all retrospective studies of well-designed clinical trials, lacking independent validation. In addition, more attention should be paid to molecular markers, such as peripheral blood neutrophils and plasma MMPs. Baselines must be determined to obtain unified criteria to guide follow-up studies.
In addition to the above problems, it is also important to account for the differences among studies. The predictive weights of different markers may be different, so a formula must be established to determine the weights of different markers.
In conclusion, although many problems remain, the use of molecular biomarkers can prolong the OS of patients and reduce their financial burden. We should pay more attention to the study of biomarkers and establish a predictive formula.

CO N FLI C T S O F I NTE R E S T
None.