Brain-derived neurotrophic factor, a new soluble biomarker for malignant pleural mesothelioma involved in angiogenesis

Previous transcriptomic data show an overexpression of BDNF gene expression in MPM cell lines compared to lung adenocarcinoma cell lines (Additional file 1: Figure S1) [3]. To confirm these results, BDNF expression was measured in 179 MPM tumor samples and 26 normal pleura (Additional file 2: Table S1.1). Figure 1a confirms the significant higher expression of BDNF in MPM tumors compared to normal pleura (p = 0.0006). BDNF showed differential expression between MPM subtypes (p = 0.0011) with a lower expression in epithelioid MPM (EM) than in sarcomatoid (SM) and desmoplastic (DM) MPM (Additional file 3: Figure S2A).

BDNF expression and overall survival of patients were related (Fig. 1b and Additional file 4: Table S2). Indeed, patients with high BDNF had a lower survival than patients with low BDNF (15.9 versus 21.1 months, p = 0.0736) and this survival difference is significant at 3 years (p = 0.0401).

These observations were confirmed using TCGA database (Additional file 2: Table S1.2). Expression of BDNF was significantly higher in MPM than in lung squamous carcinoma and lung adenocarcinoma (Fig. 1c). As previously observed, high BDNF was associated with low survival compared to low BDNF (12.4 versus 27.5 months, p < 0.0001) (Fig. 1d and Additional file 4: Table S2). BDNF was already described as overexpressed in several other cancers [4]. In TCGA cohort, we observed that MPM has the highest BDNF expression among 37 tumor types indicating that BDNF gene overexpression is a hallmark of MPM (Fig. 1e and Additional file 5: Table S3). These results were confirmed at the mRNA level and using Immunofluorescence on cancer cell lines and commercial primary mesothelial cells (MC) (Additional file 6: Figure S3A-B).

In our collection of pleural effusions (PE) (Additional file 2: Table S1.3), a significant higher BDNF level was observed in MPM samples (median, 95.26 pg/ml) compared to other neoplasia or benign samples (BPE) (median, 28.08 pg/ml and 8.87 pg/ml) (Fig. 2a) and also to all PE (malignant and non-malignant) (median, 23.33 pg/ml) (Fig. 2b) according to the mRNA results. No significant difference in BDNF level was observed between the MPM subgroups (Additional file 3: Figure S2B).

These results confirmed a preliminary observation by Duysinx and colleagues performed on only 10 MPM PE [4] and can be explained, in part, by the ability of MPM cells to produce high level of BDNF (Additional file 6: Figure S3C). This growth factor can also be produced by a large variety of cells [5] explaining its presence in other PE, but at a lower amount.

Area under the curve (AUC) of BDNF to differentiate MPM from other neoplasia or all PE were similar (AUC = 0.6710 ± 0.04 and AUC = 0.6972 ± 0.038) (Fig. 2c and Additional file 7: Table S4.1). The best specificity and sensitivity for BDNF were ~ 86.05% and ~ 49.51% (Additional file 7: Table S4.2).

The diagnostic value of BDNF (AUC = 0.69) seems slightly lower than the one of SMRP (AUC = 0.76 to 0.87) [6], the best MPM soluble biomarker to date. However, BDNF is expressed by all subtypes of MPM unlike SMRP which is not expressed by SM [2]. Then, an association of these two biomarkers has a strong potential to improve the sensitivity and the specificity of MPM diagnosis. Comparison of BDNF diagnostic value with fibulin-3 is currently complicated due to heterogeneity in the results obtained with this biomarker [2].

In several cancers, BDNF was described as overexpressed in the tumor environment [4, 7] and can be associated with poor survival [8]. Then, we evaluated the prognosis value of BDNF in MPM PE. Interestingly, as in mRNA study, patients with BDNF above median presented a significantly lower survival than the others (8.3 versus 13 months; p = 0.0061) (Fig. 2d and Additional file 4: Table S2). This association between high BDNF and poor survival suggests an implication of this protein in the development of the pathology.

Whereas prognostic value of SMRP remains inconclusive [2], patients with high BDNF have a shorter survival than patients with low BDNF. In PE, this observation is not related to MPM subtype. Indeed, in this cohort, SM, the most aggressive subtype of mesothelioma, only represent 7% of the cases and therefore cannot be responsible for this result. In PE, these characteristics are similar to the prognostic value of Fibulin-3 [2].

Several studies have demonstrated a pro-tumoral autocrine action of BDNF on cancer cells [8]. To evaluate this activity on MPM cells, expressions of BDNF receptors (TrkB and p75NTR) were measured first. Additional file 8: Figure S4A showed a heterogeneous and significant reduced expression of TrkB in MPM cells compared with MC. p75NTR expression was also heterogeneous in MPM cells and similar to MC (Additional file 8: Figure S4B). Figure 3a and b show that BDNF had no effect on MPM cell growth and sensitivity to cisplatin. These results suggest that BDNF has no autocrine action on MPM cells.

BDNF was also described as involved on angiogenesis in different cancer types [9]. We thus studied this property by measuring the induction of HUVEC proliferation. First, we showed that MPM PE induced angiogenesis by leading to an increase of HUVEC tube formation and proliferation (Additional file 9: Figure S5A-B). Figure 3c shows that an anti-BDNF blocking antibody (from rabbit, Abcam) reduced significantly by ~ 31% the MPM PE-induced HUVEC proliferation. A detailed analysis of the results led to the segregation of the MPM PE in a sensitive group to BDNF blocking (n = 11) and in a resistant group (n = 3) (Fig. 3d). These results were confirmed using another anti-BDNF blocking antibody (from chicken, Abcam) (Additional file 9: Figure S5C).

These observations demonstrate the strong implication of BDNF in the PE-induced angiogenesis. However, the resistance of some PE to the blocking antibody demonstrates that BDNF is not the only player participating to this process. This is also supported by the observation that the activity of the blocking antibody is not correlated to BDNF concentrations in PE (Additional file 10: Figure S6). Previous works have shown that, in some cancers, BDNF can induce expression of the vascular endothelial growth factor (VEGF), well known to induce angiogenesis, [9]. Thus, we measured VEGF in MPM PE. No evident correlation between BDNF and VEGF was observed (Additional file 11: Figure S7A). However, we did not observe samples with high BDNF and low VEGF. Moreover, in PE with BDNF higher than median value, a positive correlation with VEGF was observed (Additional file 11: Figure S7B). This suggests that VEGF can be dependent of BDNF in some PE. As observed for BDNF, the activity of the blocking antibody was not correlated to VEGF concentrations (Additional file 11: Figure S7C). These results show that VEGF cannot explain anti-angiogenic effect of the BDNF blocking antibody.

Recently, in the MAPS study, it was shown that the combination pemetrexed/cisplatin in association with bevacizumab (anti-VEGF) improves overall survival of MPM patients [10]. This clinical trial demonstrates the interest of targeting angiogenesis in MPM. Regarding our results, this suggests that BDNF could be an interesting target in MPM due to its implication in this process.