Background

Glioblastoma (GBM) remains the most common and deadliest primary malignant brain tumor in adults (Ostrom et al. 2018). Histological characteristics as tumor cell anaplasia, microvascular proliferation and necrosis are hallmarks of GBM (Louis et al. 2016). The well-known pronounced vascular permeability and the high expression of vascular endothelial growth factor (VEGF) present in the abnormal vasculature, highlights the importance of angiogenesis in the tumor pathogenesis and therefore, constitute a rationale to target therapy (Lu-Emerson et al. 2015; Hundsberger et al. 2017).

Although great efforts have been implemented to develop new treatment strategies for GBM patients, the standard therapy is based on a multimodal approach comprising surgery, radiotherapy and chemotherapy, mostly centered on Temozolomide (TMZ) (Stupp et al. 2005). Due to its critical role in tumor biology, VEGF signaling pathways was proposed as an attractive target in cancer therapy since the 70’s (Folkman 1971). Thus far, and despite its established importance in GBM pathogenesis, only one anti-angiogenic therapy – the monoclonal anti-VEGF-A antibody, Bevacizumab is FDA approved despite its unreliable results (Friedman et al. 2009; Kreisl et al. 2009; Chinot et al. 2014; Diaz et al. 2017; Hundsberger et al. 2017; Wick et al. 2017). Indeed, as the complexity of angiogenesis has been unveiled and other VEGF family members were also implicated, alternative signaling pathways were considered as potential targets, fueling the development of new therapies, such as multitarget inhibitors – pan-vascular endothelial growth factor receptors (VEGFR) and tyrosine kinase inhibitors (TKIs). However, the benefits are unclear (Lu-Emerson et al. 2015).

The complexities and biological obstacles curb the use of angiogenesis as a potential therapeutic target in GBM patients. This might be partially explained by the intricate interaction of distinct signaling pathways and the higher number of involved molecules. Primarily, seven VEGF subunits are well distinguished, VEGF-A, −B, −C, −D, −E, −F and placental growth factor (PGF) – which can bind to their common receptors VEGFR-1, − 2, and − 3 and trigger an elaborate process of angiogenesis and lymphangiogenesis. Theses processes are invariably present in embryological, physiological and pathological conditions, such as tumorigenesis and tumor progression (Roy et al. 2006).

For instance, VEGF-A – the prototype member of the VEGF family – mediates VEGFR-1 and VEGFR-2 activation and predominantly regulates the process of angiogenesis in the central nervous system (CNS) (Carmeliet and Jain 2011). VEGF-C is a precursor protein that, in its cleaved form, has high affinity for both VEGFR-2 and VEGFR-3 that, when upregulated, may be associated with more unfavorable prognosis in GBM patients (Grau et al. 2007; Kessler et al. 2015; Zhao et al. 2016). Interestingly, VEGF-C, was initially described as primarily involved in the formation and maintenance of lymphatic vessels (Veikkola et al. 2001), which were classically thought to be absent from the CNS (Antila et al. 2017). However, increasing evidence indicates that VEGF-C is also expressed in the brain and may regulate several biological processes in physiological and pathological settings (Jenny et al. 2006; Grau et al. 2007; Shin et al. 2008; Zhao et al. 2016). From structural and functional perspective, VEGF-D is closely related to VEGF–C (Achen et al. 1998; Stacker and Achen 2018) and binds to the same receptors, i.e., VEGFR-2 and -3 (Roy et al. 2006) and thus, exerting an overlap role by inducing angiogenesis and lymphangiogenesis (Fagiani et al. 2016).

Emerging data has pointed out an increasing importance of VEGF-D, along with -C, as an alternative pro-angiogenic pathway in cancer (Grau et al. 2011; Yang et al. 2015; Stacker and Achen 2018), although their role in CNS tumors is barely understood (Debinski et al. 2001; Moffat et al. 2006; Grau et al. 2007; Ramani et al. 2012). Also, the precise implications of VEGF-B in cancer cells are not fully known (Roy et al. 2006; Lal et al. 2018) and VEGF-E and VEGF-F have not been reported in mammals (Roy et al. 2006). Additionally, only few studies have described the patterns of expression of angiogenic factors in human GBM, most of them just focusing on VEGF-A (Xu et al. 2013; Chen et al. 2015; Kessler et al. 2015; Baumgarten et al. 2016), VEGF-C and VEGF–D.

Therefore, this study aimed to evaluate the expression of VEGF-A, −C, −D and its receptors, VEGFR-1, − 2, and − 3 in a series of GBMs as well as whether any correlation exists between immunostaining and clinical features. Also, it was investigated the potential prognostic value of distinct VEGF family members.

Methods

We retrospectively review consecutive adult patients (> 18 years old) with newly diagnosed GBMs according to the current World Health Organization Classification of Tumors of Central Nervous System (WHO, 2016) (Louis et al. 2016). These patients underwent to neurosurgical procedures, aiming maximal safe resection whenever possible, between January 2003 and December 2011, from two Brazilian hospital databases: Hospital Israelita Albert Einstein (HIAE) and Hospital São Paulo - Universidade Federal de São Paulo (HSP-UNIFESP). All tumor specimens were overnight fixed in buffered formalin and subsequently embedded in paraffin blocks.

Tumor samples from 97 patients were available and carefully reviewed by a reference neuropathologist (LN). Samples with predominantly necrotic tissue or small sample sizes as those from patients with inconsistent clinical and/or therapeutic data were excluded. Thus, the study group comprised tumor tissue from 70 GBM samples (HSP-UNIFESP, n = 57; HIAE, n = 13).

Tissue microarray (TMA) construction

At least two different and of the most representatives areas of each tumor were selected for analysis. Cylindrical cores of 2-mm were removed and used in the construction of tissue microarray (TMA) blocks, as previously described (Saggioro et al. 2014). Eleven TMA blocks were constructed using a Beecher tissue array instrument™ (Beecher Instruments, Silver Spring, MD, USA) in accordance to the manufacturer’s instructions. For each immunostaining, 4 μm slices were cut from TMA blocks by using a Leica microtome and placed on glass charged slides. Slides were cut consecutively to minimize the influence of tissue heterogeneity when comparing the expression of the different VEGF family members within each patient tumor sample. As control, a normal prostate tissue prostate section was added in each TMA block.

Immunohistochemistry assays and evaluation

The immunohistochemical (IHC) procedures were performed on 4-μm-thick TMA sections and mounted on charged slides. Briefly, for immunostaining, the slides were deparaffinized, and rehydrated through a graded ethanol series. The antigen retrieval was done using Diva Decloaker solution (Biocare Med™, Concord, CA, USA), for 40 min at 98 °C in a steamer chamber. The slides were incubated with the pre-diluted antibody overnight at room temperature and subsequently washed with Tris-buffered saline with Tween 20, as shown in Table 1.

Table 1 Summary of antibodies’ clones and titration
Full size table

The immunohistochemistry test was chosen as the preferable evaluation method because of its practical approach, cost-effectiveness, widely availability, replicability, and has a proven usefulness in previous studies (Debinski et al. 2001; Jenny et al. 2006; Grau et al. 2007).

The IHC stains of VEGF and its cognate receptors were blindly evaluated at ×200 magnification using a DX-51 Olympus microscope, according to the following semiquantitative grading score: negative (0), absence or up to 10% IHC staining of the core area; score 1+, focal expression in > 10% to up to 25% of the core area; score 2+, intermediate expression in > 25% to up to 75% of the core; score 3+, diffuse expression in more than 75% of core area. At least a quarter of the core area had to be evaluable for scoring and faint immunostains were not considered. For each case (patient), the definitive IHC score was determined by the mean IHC stain obtained from the cores evaluated (up to 3 cores per case). The intensity of IHC stains was not considered. Finally, for each VEGF member the cellular location of IHC staining was noted, i.e., membrane, cytoplasmic and/or nuclei staining as well as the neoplastic and non-neoplastic endothelial cells.

The isocitrate dehydrogenase-1 (IDH-1) status was evaluated by IHC using the anti-IDH-1 mutant antibody for R132H (H09 clone, Dianova™, Hamburg, Germany) at 1:20, according to manufacturer’s recommendations.

Clinical profile and outcomes

The medical records of 70 patients were reviewed for the following data: gender, age at diagnosis, Karnofsky Performance Status (KPS), date of the first symptom, date of death or last evaluation, date and extent of resection and radiation therapy (RT) and chemotherapy regimens.

Neurosurgery was performed to attain the maximum safe and feasible resection according to the guidelines of both institutions, which have suitably equipped surgical centers. The extension of tumor resection was defined based on immediate (48 h) postoperative imaging (CT or MR), as radical resection (absence of residual contrast enhancement), partial resection or biopsy. Patients underwent 3D localized external beam RT delivered to the contrast-enhancing lesion shown at CT/T1-weightedimagesor T2/fluid attenuated inversion recovery sequence MR. The RT dose was prescribed according to the guidelines of the International Commission of Radiological Units fields, once daily at 2 Gy per fraction, 5 days a week, for a total of 60 Gy. The treatment protocols and personnel varied over time and between centers. Chemotherapy regimens also varied between centers. At HIAE, all patients were treated with concomitant and adjuvant Temozolomide according to the European Organization for Research and Treatment of Cancer–National Cancer Institute of Canada protocol (Stupp et al. 2005). At HSP-UNIFESP until 2008, patients received 200 mg/m2 carmustine (bis-cloroethylnitrosourea [BCNU]) at 6-week intervals starting 6 weeks after RT. Since 2009, TMZ has been available and patients could be treated with the EORTC-NCIC protocol. The patients who underwent chemotherapy treatment according to the EORTC-NCIC protocol were categorized as “RT concurrent with chemotherapy.” The patients who received BCNU were defined as having “RT and sequencing chemotherapy.” The patients who did not received RT and/or chemotherapy were defined as having “best supportive care”.

Statistical analyses

Data were described using absolute and relative frequencies for categorical data. Quantitative data were described using median and range, due to skewness. Overall survival (OS) was calculated from date of surgery until death or last follow-up and the cut-off date was November 30th 2018.

Survival curves were constructed according to the Kaplan-Meier method and comparison between groups by using log-rank test to explore relationships between well-recognized prognostic factors (age, KPS, extent of tumor resection, adjuvant treatments) and survival time in the univariate analysis. A conditional stepwise proportional hazard analysis (Cox-regression model) was used to identify independent predictors of survival. The level of significance was 0.05 (p < 0.05). [p ≤ 0.05].

For statistical analyses, only cases with consistent IHC expression (scores 2+ and 3+) were considered “positive or high expression”, i.e., cases with at least > 25% of IHC staining in the core area at definitive score. Cases showing < 25% IHC staining were considered as “negative or low expression”. Plausible associations among the IHC score of multiple VEGFs and their receptors were assessed using Spearman rank correlation coefficient.

The statistical analysis was performed using the statistical software R (R Core Team, 2017 added to the Car graphic package (Fox and Weisberg 2011) and survival (Therneau 2015).

Results

Clinicopathological features and patient outcomes

The median age of the patients at surgery was 59.5 years (range, 18–78 years), with 44 males (62.8%, male/female ratio of 1.7:1) (Table 2). The median follow-up time of all survivors was 8.21 months (range, 1.2–95.8 months) and the estimated OS was 7.25 months (range, 5.33–14.08 months).

Table 2 Demographics, clinical and treatment data of GBM patients (univariate analyses)
Full size table

VEGF-A, −C, −D and VEGFR-1, −2, −3 IHC expression and patient outcomes

The overwhelming majority of GBM cases showed consistent VEGF expression, regardless of its subunit or receptor, as shown in Table 3.

Table 3 Number of cases that expressed (“positive cases”, 1+ to 3+) and those with “high expression” (> 25%, 2+ or 3+) for each VEGF subunit and cognate receptora
Full size table

Even though several TMA cores were lost, of which the IHC score was not performed, mostly in regards to VEGFR-3 (28.6%) and VEGF-C (20.0%), the majority of GBM cases showed “high expression” for all VEGF subunits and their cognate receptors (> 25%, 2+ or 3+), as shown Table 3. Interestingly, VEGF-C showed conspicuous nuclear immunostaining (repeated three times) (Fig. 1). The remaining VEGF subunits and receptors showed cytoplasmic staining (data not shown).

Fig. 1
figure 1

Morphological and immunohistochemical features of VEGF-C and –D in GBM tissue microarray (TMA) sections. A) Cylindrical core of 2-mm from a representative GBM area (H&E, × 100). B) TMA design in a 4-um tissue section used for VEGF-C immunostaining (lower magnification). C to E) VEGF-C staining. Note the nuclear IHC expression of VEGF-C in tumor cells and in neoplastic endothelium of glomeruloid vessels (1C, arrows). F) VEGF-D expression in cytoplasm of GBM tumor cells and endothelial cells (× 400)

Full size image

Overall, there were no significant differences in OS regardless the VEGF staining score on tumor cores (Table 4).

Table 4 Univariate analysis of overall survival according to the VEGF and VEGFR immunohistochemical score in GBMs
Full size table

Additionally, other univariate analyses was performed using patients grouped according a binary IHC category: “no or low expression” (up to 1+) and “high expression” (2+ or 3+) for each VEGF subunit and its cognate receptors. The model was created to statiscally strengthen groups and overcome the hurdles as a result of a small sample size due to the missing TMA cores. However, no significant differences were found between OS and these IHC categories (Table 5).

Table 5 Univariate model for analysis of the immunohistochemical expression according a binary category (absent/low vs high expression) for each VEGF subunit and receptors in GBMs
Full size table

Finally, regarding the IDH-1 mutation 6 out 70 patients harbor the canonical R132H mutation that is detected by IHC (8.6%). However, the IHC expression of VEGF subunits and cognate receptors did not differ from those with wtIDH-1, which constitutes the overwhelming majority of casuistry.

Evaluation of the relationship between distinct VEGF subunits and receptors

Table 6 shows the Spearman rank correlation coefficients among the IHC expression of VEGFs. The VEGF and its cognate receptors showed a weak and monotonous correlation pattern. A moderate correlation was only found for the VEGF-A and VEGFR-2 (Spearman = 0.478), as well as for VEGF-D and VEGFR-2 (Spearman = 0.456).

Table 6 Spearman rank correlation coefficients among VEGF subunits and their receptors
Full size table

Discussion

In the current study, VEGF-C and -D along with its receptors VEGFR-2 and -3 were overexpressed in the majority of GBM tumor samples, ranging from 87.3% (VEGF-D) to 98.5% of the cases (VEGFR-1). However, there were no significant differences in the OS according to the VEGFs and their receptors and the immunostaining on tumor cells. This finding was expected since the VEGFs were overexpressed and our casuistry was overwhelmed by cases of GBMs with wtIDH-1 (only 6 out of 70 showed the IHD-1 R132H mutation).

Our results emphasized the hurdles for establishing angiogenic prognostic and predictive biomarkers in CNS tumors, which is in line of unclear benefits no matter the anti-angiogenic drug prescribed for unselected CNS tumor patient population (Winkler et al. 2018). Hopes persist to rest on the discovery of predictive biomarkers on tumor tissue, blood and/or radiographic parameters, which could merit the broad use of angiogenesis inhibitors, particularly for GBM patients (Lu-Emerson et al. 2015; Winkler et al. 2018). On the other hand, VEGF-C, −D, VEGFR-2 and -3 were not fashionable research targets due to their established role in primarily promoting the growth and remodeling of lymphatic vessels, which were considered absent in CNS, until recently. However, a considerable amount of evidence suggests the existence of a complex lymphatic system in CNS, mainly within the meninges, which drains cerebrospinal fluid into the deep cervical lymph nodes (Aspelund et al. 2014, 2015; Louveau et al. 2015; Weller et al. 2016; Antila et al. 2017). The development and modeling of that specific lymphatic system are closely related to the interactions between VEGF-C and VEGFR-3 (Antila et al. 2017). The emerging data brought upfront the exploration of VEGF-C and its receptors – VEGFR-2 and -3 – as its partially homologous VEGF, the VEGF-D (Achen et al. 1998; Grau et al. 2011; Michaelsen et al. 2018). To the best of our knowledge, this study represents the first to focus on the immunohistochemical expression of the expanded axis of proangiogenic factors simultaneously – VEGF-C, −D and VEGFR-2, − 3 – and that theoretically relate to lymphangiogenesis in GBM. However, a nuclear VEGF-C staining was found in GBM tumor cells, as was similarly previously described (Cai et al. 2012). Whether this nuclear VEGF-C staining means an abnormal cell reprograming or reflects the functional status of VEGF-C in the tumor cell is unknown. This finding deserves further investigations.

The accumulating data from experimental models and resected human tumor samples further described the expression of various VEGFs and their cognate receptors in glial tumors (Machein and Plate 2000; Huang et al. 2005). The first description of VEGF-D immunoreactivity in GBM was reported by Debinksi et al. (2001) who used GBM cell lines and ten tissue sections. They demonstrated that nearly all the GBM cells produce VEGF-D what may be partially implicated in oncogenic transformation and appeared to be an attractive target for novel treatment strategies. Following these results, Jenny et al. (2006) observed VEGF-C, −D and VEGFR-3 expression in normal brain tissue and in most brain tumors, as such glioblastomas and hemangioblastomas. The expression of VEGF-C and -D has been demonstrated to be diverse in GBM tissue samples, since the VEGF-C is found to be overexpressed compared to VEGF-D (Grau et al. 2011; Michaelsen et al. 2018).

Consistently, our results indicate that VEGF-C is proportionally overexpressed (94.6%), but is without any significant influence in patient prognosis. In contrast, low expressions of the VEGF-D (HR of 4.27 (CI 95%, 1.30 – 13.95) impaired the patient’ OS significantly (p = 0.016). Weickhardt et al. 2015 related that the anti-angiogenic therapy (i.e. Bevacizumab) might exhibited superior efficacy when VEGF-D is low expressed in other clinical scenario. On the other hand, despite its wide variability, previous reports suggested that VEGF-D, when upregulated, is associated with higher tumor grade of malignancy (Grau et al. 2011; Jiang et al. 2018). It is important to note that when the VEGF-D evaluations were grouped in a binary category (“absence/low expression” and “high expression”) no significant OS influence was found (HR of 1.22 (CI 95% 0.52–2.88), p = 0.64). We should emphasize that caution must be taken when interpreting this study’s VEGF-D findings, which might be partially explained by the restricted number of tumor samples with weak VEGF-D IHC expression (8.6%). In the same way, it was not found any significant differences on OS for the expression of others VEGF factors or cognate receptors. This finding might be partially explained by the relatively small number of patients in our series.

Substantial data support the correlation of the expressions of VEGF receptors, such as VEGFR-3 and the tumor grading (Grau et al. 2007, 2008). Baumgarten et al. (2016) have shown an overexpression of VEGFR-1, − 2 and, − 3 in GBM. Moreover, the expression of VEGFR-3 trends to turn positive as the pathologic grade of malignancy increases (Jiang et al. 2018). These previous findings were compatible with our results, as it was found overall higher immunoexpression of VEGFRs (VEGFR-1, 92.9%; − 2, 90%; − 3, 70%). However, in the present study it was found only a weak and monotonous correlation pattern between VEGF and its cognate receptors.

Overall, it is still unknown whether VEGFs and their receptors expression profiles are plausible biological markers to predict response and/or patient prognosis. Nevertheless, the are indications that VEGF-C, −D and VEGFR-2, − 3 in particular may have implications on alternative signaling pathway of primary and acquired resistance to anti-VEGF therapy (Moffat et al. 2006; Grau et al. 2011; Li et al. 2014; Michaelsen et al. 2018). Our results did not support this hypothesis since no recurrent or progressive GBM tumor samples were investigated. However, the overexpression of the entire axis of VEGF-C, −D, VEGFR-2 and -3 might be interpreted as an investigational target to clarify that theory, considering the recent data that suggests that there is a complex lymphatic system in the CNS (Aspelund et al. 2014, 2015; Louveau et al. 2015; Choy and Rahul Jandial 2016; Antila et al. 2017) and its development and modeling may be closely related to the interactions between VEGF-C and VEGFR-3 (Antila et al. 2017).

Conclusions

VEGF-C and -D and its receptors were overexpressed in the overwhelming majority GBMs but their expressions did not correlate with patient’s OS. The overexpression of the axis of VEGF-C, VEGF-D, VEGFR-2 and VEGFR-3 might be interpreted as a potential target for further studies, particularly the interactions between VEGF-C and VEGFR-3 as well the mild nuclear VEGF-C expression. Whether this nuclear staining means an abnormal cell reprograming or reflects the functional status of VEGF-C in the GBM cell is unknown. Additional studies with more extensive series of GBM are still necessary to better evaluate the roles of VEGF-C, VEGF- D and their cognate receptors, VEGFR-2 and VEGFR-3, in GBMs.