Nintedanib Is Active in Malignant Pleural Mesothelioma Cell Models and Inhibits Angiogenesis and Tumor Growth In Vivo

Malignant pleural mesothelioma (MPM) is a rare but aggressive thoracic malignancy that originates from the pleural mesothelium. MPM is most often caused by asbestos but causal links to ionizing radiation have also been suggested (1). Notably, germline mutations in a tumor suppressor gene, BAP1 (BRCA1 associated protein-1), were shown in families with high MPM incidence (2).

There are three main histological subtypes of MPM: epithelioid, sarcomatoid, and biphasic. The latter two subtypes are especially refractory to therapy and thus patients with sarcomatoid or biphasic MPM have poorer outcomes than those with epithelioid tumors (3). Although local invasion and recurrence are pathognomonic features of MPM, a study on a large postmortem series demonstrated that extrathoracic metastases are common (4).

Because MPM presents with advanced stage in most cases, many MPM patients receive systemic therapy either alone or as part of multimodality treatment. The combination of cisplatin with pemetrexed is the standard first-line systemic therapy (5). The recently published phase III MAPS study, however, demonstrated that the addition of bevacizumab, a VEGF (vascular endothelial growth factor) neutralizing antibody, to this combination significantly prolongs patients' overall survival (OS) (6). Accordingly, the latest NCCN guideline recommends bevacizumab/cisplatin/pemetrexed as a first-line therapy option for MPM patients who are not candidates for curative-intent surgery (7).

Besides the recent success with the addition of bevacizumab to platinum-based chemotherapy, accumulating preclinical evidence suggests that angiogenesis has a crucial role in MPM progression. For example, high tumor microvessel density (MVD) was shown to be an independent poor prognostic factor in human MPM (8). Moreover, among patients with solid tumors, those with MPM had the highest circulating VEGF levels and these elevated levels were associated with poor clinical outcome (9, 10). In support of this, neutralizing antibodies against either VEGF or its receptors were shown to inhibit MPM cell proliferation in vitro, indicating that—besides inducing angiogenesis—VEGF can also support MPM growth in an autocrine manner (9).

In addition to VEGF family members, other angiogenic factors and receptors, including fibroblast growth factors (FGF), FGF receptors (FGFR; refs. 11, 12), platelet-derived growth factors (PDGF), and PDGF receptors (PDGFR; ref. 13), are also expressed in MPM. Accordingly, various angiogenesis inhibitors have been tested in MPM in first-line setting or in pretreated patients. Unfortunately, none of these drugs could provide a significant survival benefit for patients with MPM (14, 15).

Nintedanib (BIBF1120) is a small molecule receptor tyrosine kinase inhibitor (RTKI) of VEGFR 1–3, FGFR 1–3, and PDGFR α/β, as well as members of the Src family (16). The antitumor potential of nintedanib has been demonstrated in preclinical models of liver (17), head and neck (16), renal (16), lung (16, 18), pancreatic (18), colon, ovarian, and prostate cancers (16). Nintedanib has demonstrated clinical activity in ovarian (19) and renal cancers (20) and, moreover, it has been approved in Europe for the second-line treatment of patients with locally advanced, metastatic or locally recurrent lung adenocarcinoma in combination with docetaxel (21). Furthermore, nintedanib was approved in 2014 for use in patients with idiopathic pulmonary fibrosis (22), a devastating lung disease characterized by progressive scarring of the lung parenchyma mediated in part by profibrotic factors including PDGF and FGF. These clinical findings, taken together with the aforementioned preclinical data, provided the rationale for the recent LUME-Meso phase II trial, which demonstrated that the addition of nintedanib to cisplatin/pemetrexed significantly prolongs progression-free survival (PFS) in patients with unresectable MPM (23). Based on the promising phase II data, the LUME-Meso study has been extended and it is currently enrolling patients as a confirmatory phase III trial (24). However, the preclinical anti-MPM activity of nintedanib has not been studied in detail so far. Therefore, in the current study, we evaluated the potential of nintedanib against human MPM cell lines, primary cell cultures, and tumor xenografts.

SPC111, SPC212, and M38K cells were established from human biphasic MPMs and kindly provided by Prof. R. Stahel (SPC111 and SPC212, University of Zurich, Zurich, Switzerland) and by Prof. V.L. Kinnula (M38K, University of Helsinki, Helsinki, Finland). The I2 epithelioid MPM cell line was kindly provided by Prof. A. Catania (University of Milano, Milano Italy). p31 (epithelioid MPM) was kindly provided by Prof. K. Grankvist (University of Umea, Umea, Sweden). The VMC and Meso primary cell cultures were established at the Medical University of Vienna. The nonmalignant mesothelial cell line Met5A (ATCC CRL-9444) was purchased from the American Type Culture Collection. The NP normal mesothelial cell cultures were established from pleura tissue samples of patients operated on for spontaneous pneumothorax. Histological subtypes, drug sensitivity data, target RTK copy-number changes, and mutational status of tumor suppressor genes of MPM cell lines are presented in Supplementary Tables S1 and S2. Cells were maintained in RPMI-1640 or DMEM supplemented with 10% fetal calf serum (FCS) (Sigma Chemical Co.), 100 U/mL penicillin and 10 mg/mL streptomycin (Sigma Chemical Co.) and regularly screened for Mycoplasma contamination. Cell line authentication was performed as previously described (12). All cell lines were maintained at 37°C in a humidified incubator with 5% CO_2.

MPM cells (2 × 10⁵) were seeded into 6-well plates and incubated with nintedanib or solvent after 24 hours of recovery. Cells were harvested in RIPA buffer containing protease inhibitor cocktail (both from Fisher Scientific), and proteins were separated by SDS-PAGE, blotted onto nitrocellulose membranes, and immunostained with the following primary antibodies (all from Cell Signaling Technology): rabbit polyclonal pErk1/2 (#9101), Erk1/2 (#9102), pAkt (#4058), Akt (#9272), pS6 (#2215), S6 (#2217), β-actin (#4970) at a dilution of 1:1,000, followed by a horseradish peroxidase-coupled secondary antibody and developed by ECL Reagent (GE Healthcare). Films were scanned and phosphorylation was quantified using the ImageJ software.

To establish orthotopically growing tumors, 2 × 10⁶ MPM cells were inoculated into the thoracic cavity of 8-week-old SCID mice. Once (based on our preliminary experiments) MPM nodules reached a macroscopically visible size (28 days after tumor implantation), mice (n = 9/group) were randomized into treatment and control groups. For the survival experiments, animals with human p31 MPM tumors growing in the thoracic cavity received (i) nintedanib per os (PO, 50 mg/kg), (ii) nintedanib intraperitoneally (IP, 50 mg/kg), (iii) solvent PO, or (iv) solvent IP. Animals were weighed three times a week and euthanized when they showed significant morbidity. Nintedanib was dissolved either in methylcellulose (PO treatment) or in DMSO (IP treatment).

For the tumor growth experiment, p31 or SPC111 cells were injected intrathoracically as described above. The animals were then randomized into the following groups, each consisting of 10 mice: (i) solvent IP, (ii) cisplatin (3 mg/kg, dissolved in 0.9% NaCl IP) and pemetrexed IP (30 mg/kg, dissolved in 0.9% NaCl), (iii) nintedanib IP (50 mg/kg, dissolved in DMSO), (iv) bevacizumab IP (10 mg/kg), (v) cisplatin and pemetrexed in combination with nintedanib IP, (vi) cisplatin and pemetrexed in combination with bevacizumab IP.

The treatments started 21 and 12 days after p31 and SPC111 tumor cell inoculation, respectively. In both sets of experiments, cisplatin was applied once weekly, pemetrexed and nintedanib were administered five times a week on consecutive days, while bevacizumab was injected twice weekly. Body weight was checked thrice a week. The experiments were terminated on the 28th day (p31 tumors) and 16th day (SPC111 tumors) of treatments due to signs of distress in the control groups. Two hours before the mice were sacrificed, 200 mg/kg BrdUrd in 0.9% NaCl was injected intraperitoneally. Tumor nodules were harvested, weighed, and frozen in liquid nitrogen. All animal experiments were carried out in accordance with the ARRIVE guidelines (25) and with the animal welfare regulations of the host institutes (permission number: PEI/001/2574-6/2015).

Consecutive 10-μm-frozen sections were prepared and fixed in methanol (for hematoxylin, CD31, and BrdUrd staining) or 4.5% Histofix (for cleaved caspase-3 staining). For microvessel labeling, the slides were incubated with rat monoclonal anti-mouse CD31 antibody (Clone SZ31, Dianova, dilution 1:50) followed by Alexa 555-conjugated anti-rat IgG (Cell Signaling Technology). To detect apoptotic cells, sections were incubated with cleaved caspase-3 antibody (Cell Signaling, dilution 1:400) and anti-rabbit FITC (Jackson). To determine the proliferation rate of tumor cells, anti-BrdUrd monoclonal antibody (Becton Dickinson, Bioscience, dilution 1:50) and FITC-conjugated anti-mouse IgG (Jackson) were applied. Nuclei were stained with DAPI (Thermo Fisher Scientific). Slides were scanned by TissueFAXS and ImageJ was used to analyze the images as described recently (26). Relative microvessel areas (MVA) were calculated by counting the number of CD31-positive pixels in the total tumor area. The percentages for cleaved caspase-3 and BrdUrd-positive pixels were also determined.

Distribution of the data was tested with the Shapiro–Wilk test. For parametric data, t test was used when comparing two groups, and one-way ANOVA with Tukey multiple comparison test was used for the comparison of more than two groups. Nonparametric data were analyzed with Mann–Whitney U test, and the Kruskal–Wallis test with the Dunn multiple comparison post hoc test was used for analyzing more than two groups. The correlations between different parameters were calculated by Spearman correlation test. Kaplan–Meier curves for animals' survival were evaluated and the log-rank test was used to establish the significance of the difference. All statistical analysis was performed by using GraphPad Prism 7.0 software (GraphPad Inc.). P values are given as two sided and were considered statistically significant below 0.05.

Details of the procedures that were used for RNA isolation, reverse transcription, real-time PCR, magnetic resonance imaging, array comparative genomic hybridization (array CGH), exome sequencing, ELISA, and in vitro chemosensitivity, clonogenic, proliferation, apoptosis, and migration assays can be found in Supplementary Materials and Methods.

Transcript levels of the key molecular targets of nintedanib were determined in 20 MPM cell lines and, as controls, in the Met5a immortalized mesothelial cell line and in three primary mesothelial cell cultures. While PDGFRA, VEGFR1-3, FGFR2, and FGFR3 were expressed only by some of the cell lines and at relatively low levels, PDGFRB and FGFR1 mRNAs could be detected in each tumor cell line and in most of the nonmalignant control cell lines, although with various expression levels (Fig. 1A). Importantly, all of the MPM cell lines were double positive for PDGFRB and FGFR1. Moreover, FGFR1 expressions were elevated in MPM cells compared with control mesothelial cells (Fig. 1A; P < 0.05). The copy-number changes of PDGFRA, PDGFRB, VEGFR2, and FGFR1 are shown in Supplementary Table S1. The mutations identified in tumor suppressor genes BAP1, NF2, and CDKN2A are presented in Supplementary Table S2. Notably, none of these data were associated with nintedanib sensitivity (raw data are shown in Supplementary Tables S1 and S2). The mRNA levels of PDGFRB and FGFR1 in MPM cells according to their histological subtypes are shown in Supplementary Fig. S1. Neither these two RTKs (Supplementary Fig. S1) nor the other target receptors (data not shown) exhibited a histotype-specific expression profile. While sarcomatoid and biphasic MPM cells showed similar expression patterns, the epithelioid group was more heterogeneous.

Mesothelial and MPM cells were exposed to different concentrations of nintedanib for 72 hours. Cell viability was determined by SRB assay, dose–response curves were plotted to determine IC₅₀ values for each cell line. IC₅₀ values for the Met5a and NP3 control cell lines were 1.1 and 4.1 μmol/L, respectively. MPM cells showed a wide range of sensitivity with IC₅₀ varying from 1.6 to 5.9 μmol/L, with no association to histological subtype (Fig. 1B; Supplementary Table S1) or mRNA expression profile of the target RTKs (data not shown). In long-term growth assays, where MPM cells were treated with nintedanib for 10 days, nintedanib was able to effectively inhibit clonogenicity at much lower concentrations (Fig. 1C).

Because RTKs targeted by nintedanib were shown to interfere with chemotherapy in other malignancies (27–29), we next analyzed whether nintedanib has an impact on chemosensitivity in two different human MPM cell lines, the p31 and SPC111. SPC111 cells are more resistant to nintedanib than p31 cells (IC₅₀s are 5.9 vs. 1.9 μmol/L, respectively; Supplementary Table S1). With respect to their cisplatin sensitivity, p31 is more resistant than SPC111 (IC₅₀s are 4.3 vs. 0.7 μmol/L, respectively; Supplementary Table S1). FGFR1 expression levels are similar in the two cell lines, while PDGFRB transcript levels are higher in p31 cells. The other target RTKs of nintedanib cannot be detected on mRNA level in either cell lines (Supplementary Fig. S2). Notably, at certain concentrations, we observed an additive effect between nintedanib and the established chemotherapeutic agent (cisplatin) in both cell lines (Fig. 1D). However, no synergism between nintedanib and cisplatin was evident in p31 and SPC111 cells (Fig. 1D) and, furthermore, there was no correlation between cisplatin and nintedanib sensitivities in a panel of 20 human MPM cell lines (raw data are shown in Supplementary Table S1).

BrdUrd incorporation assay was performed in five different MPM cell lines to investigate the effect of nintedanib on tumor cell proliferation. The drug exhibited a significant antiproliferative effect in each MPM cell line in a dose-dependent manner (Fig. 2A).

To study apoptosis induction, TUNEL assays were performed (Fig. 2B), which enable the detection of DNA fragmentation during apoptosis. Significantly elevated apoptosis rates upon nintedanib treatment could be confirmed only in SPC212 and VMC40 cells (Fig. 2B).

To study the antimigratory activity of nintedanib, five human MPM cell lines with different sensitivities to the drug (Supplementary Table S1) were used. Nintedanib treatment (1 or 3 μmol/L) was able to significantly reduce the migratory activity of each cell line. Average displacements of SPC111 cells are plotted, for various elapsed time periods, in Fig. 2C. Time-lapse microscopy demonstrating the effect of nintedanib treatment on the migratory activity of the SPC212 cells is shown in Supplementary Video S1.

Time-course experiments on nintedanib's effects on Erk, Akt, and S6 phosphorylation were carried out to investigate the molecular mechanisms behind the antitumor effects of the drug. MPM cells were exposed to 0.5 μmol/L nintedanib for different times ranging from 5 minutes to 24 hours. Activation of the abovementioned proteins was determined by Western blot analysis. As shown in Fig. 2D; Supplementary Fig S3, the phosphorylation of Erk1/2 was efficiently blocked as early as 20 minutes in each cell line. Probably due to changes in cell density and in growth factor concentrations, Erk activation declined during the 24-hour incubation also in the solvent controls. Nevertheless, the inhibition of Erk phosphorylation was apparent in p31 and VMC40 cells even at this later time point. Only a minor inhibitory effect was seen in SPC111 and M38K cells, while no effect was observable in SPC212 cells. Regarding Akt phosphorylation, an inhibitory effect could be observed in p31, SPC212, M38K, and VMC40 cells, although in the latter two cell lines only at later time points (24 hours). Nintedanib (at 0.5 μmol/L) had no effect on the activation of Akt in the SPC111 cells, which was the most nintedanib-resistant cell line in the whole panel. The drug's impact on S6 phosphorylation was negligible.

To examine the in vivo efficacy of nintedanib, we used an orthotopic MPM model where human p31 or SPC111 MPM cells were injected into the thoracic cavity of SCID mice.

In the first set of these experiments, 50 mg/kg nintedanib was administered either PO or IP, starting from the 21st day after inoculation of p31 cells. Based on our preliminary experiments (data not shown), by this time, several tumor nodules are already visible macroscopically on the pleura of mice. The p31 cells highly express PDGFRB and FGFR1 receptors (Supplementary Fig. S2), they are relatively responsive to nintedanib (Fig. 1) but insensitive to cisplatin (Supplementary Table S1). Nintedanib was well tolerated without any signs of toxicity during the treatment period. Survival of PO-treated animals showed a favorable trend, but no significant benefit (P = 0.059 vs. PO control; Fig. 3A). However, nintedanib significantly prolonged the survival of mice when it was administered IP (P = 0.0008; Fig. 3A). Importantly, nintedanib IP also significantly inhibited the relative weight loss of the animals (P = 0.0337; Fig. 3B), in accordance with their better overall condition at the end of the experiment. Our interpretation of these data is that nintedanib IP not only prolongs the survival of mice with orthotopically growing human MPM tumors but also interferes with MPM-induced cachexia. To further corroborate the antitumor effects of nintedanib, in vivo MPM growth was also examined by MR imaging. Although control animals had significant intrathoracic tumor burdens on day 25 after tumor implantation, a marked reduction of tumor mass in nintedanib-treated animals was evident (Fig. 3C).

In the second set of orthotopic MPM xenograft experiments, the addition of nintedanib to standard-of-care chemotherapy was investigated by implanting two different human MPM cell lines, the p31 and the SPC111 (Fig. 4). Nintedanib was administered IP, as in the aforementioned set of in vivo experiments, this route of administration proved to be more effective than PO treatment. We could demonstrate significantly reduced tumor burden in mice treated with cisplatin/pemetrexed chemotherapy alone or nintedanib alone in both cell models when compared with untreated controls (Fig. 4A). In comparison with untreated control tumors, combined chemo- and antiangiogenic regimens (including either nintedanib or bevacizumab) also exhibited significant in vivo tumor growth-inhibitory potential in both cell models (Fig. 4A). Importantly, in accordance with a previous study from Li Q and colleagues (30), bevacizumab alone was effective only against p31 tumors with high baseline VEGF-A levels (vs. control) and could not provide therapeutic benefit in the SPC111 model where tumor cells had markedly low baseline VEGF-A levels, as measured by ELISA (Fig. 4A and B).

Another key observation in this second set of in vivo experiments was that adding nintedanib to standard chemotherapy produces significantly higher responses than chemotherapy alone. These responses were comparable (P31 tumors) or superior (SPC111 tumors) to those achievable by combined bevacizumab and chemotherapy (Fig. 4A). It is also important to mention here that unlike bevacizumab alone, nintedanib alone proved to be a more effective inhibitor of in vivo MPM growth than standard chemotherapy in p31 tumors with higher baseline VEGF-A expression (Fig. 4A and B).

Morphometric analysis using CD31 as an endothelial marker revealed a strong tendency for increased microvessel areas (MVAs) in control p31 tumors with high baseline VEGF-A expression as compared with MVAs in untreated low baseline VEGF-A SPC111 tumors (2.7% ± 1.2% vs. 1.2% ± 0.8%, respectively; P = 0.067). In line with the potent in vivo MPM growth inhibitory effect of nintedanib, significantly lower MVAs were present in tumors treated with nintedanib (alone or in combination with chemotherapy) in both cell models (vs. controls; Fig. 5A). Interestingly though, we did not observe a significant reduction in MVAs of p31 or SPC111 tumors treated with bevacizumab with or without chemotherapy (Fig. 5A). The potent antivascular effect of nintedanib was accompanied by increased intratumoral necrosis in both cell models. This was most prominent in the combined nintedanib/chemotherapy groups (Fig. 5B). Nintedanib as a single agent resulted in significantly increased MPM cell apoptosis (P = 0.0317; vs. control; Fig. 5C) and decreased proliferation in p31 tumors (P = 0.0341 vs. control; Fig. 5D). However, none of the other treatment groups showed significant changes in tumor cell apoptosis and proliferation rates and, furthermore, we failed to identify any obvious associations between these parameters (i.e., MPM cell apoptosis and proliferation) and the net in vivo tumor growth inhibitory effect of nintedanib presented in Fig. 4.

The current first-line systemic standard-of-care therapy for MPM is cisplatin plus pemetrexed. The recent phase III MAPS trial demonstrated that the median OS of unresectable MPM patients from adding bevacizumab to this chemotherapeutic regimen can be increased from 16.1 to 18.8 months (6), indicating that angiogenesis inhibition might be an effective anti-MPM strategy and, accordingly, that additional antiangiogenic agents need to be evaluated in human MPM.

Here, by utilizing different in vitro and in vivo models of human MPM, we examined the antitumor activity of nintedanib, a small molecule RTKI targeting VEGFR 1–3, FGFR 1–3, and PDGFR α/β signaling (16). Although the biological impact of these angiogenic pathways has already been reported in MPM, this is the first preclinical study investigating the therapeutic efficacy of nintedanib in this malignancy.

First, we show that all of the studied 20 MPM cell lines express relatively high FGFR1 and PDGFRB levels. PDGFRA and VEGFR2 expressions were also detected in certain cell lines while the mRNA levels of other nintedanib target RTKs were rather low or absent. This is in line with previous studies of our group and others (12, 13, 31). Further corroborating the role of PDGFRA and VEGFR2 signaling in MPM, a recent retrospective study analyzing tumor samples of advanced-stage MPM cases revealed that one-third of the patients carried mutations of nintedanib target genes (32), although these findings have not been confirmed in other MPM genomics studies (33, 34).

Production of FGFR, PDGFR, and VEGFR ligands might enable MPM cells to promote their own growth and survival via distinct autocrine signaling loops. In support of this, Strizzi and colleagues showed that VEGF induces in vitro MPM growth via the phosphorylation of VEGFR1-2 (9). In the same study, members of the VEGF/VEGFR axis were found to be expressed in human MPM tissue and treatment of MPM cells with neutralizing antibodies against VEGF or VEGFR2 resulted in decreased tumor cell proliferation (9). There is also a substantial volume of literature on the autocrine growth-promoting effects of PDGF/PDGFR and FGF/FGFR pathways in MPM (11–13) and in other solid tumors as well (35). These reports, together with our observation that nintedanib dose-dependently inhibits in vitro MPM growth, allow us to conclude that—besides its antiangiogenic potential—nintedanib has a direct in vitro anti-MPM effect.

We have reported recently that MPM cells exhibit an exceedingly high migratory potential, even when compared with malignant cells of highly metastatic tumors such as lung cancer or melanoma (36). Studies examining the regulation of MPM cell chemotaxis have shown that the PDGF/PDGFR axis promotes MPM cell migration via Erk and PI3 kinase/Akt signaling (37, 38) and, furthermore, that activation of FGFR1 by FGF2 induces MPM cell migration and Erk activation (12). In the current study, we found that nintedanib potently blocks MPM cell migration. In line with this, nintedanib rapidly inhibited the phosphorylation of Erk in all the five investigated MPM cell lines and, moreover, decreased the activation of Akt in those MPM cell lines that were found to be relatively sensitive in the in vitro viability assays. Of note, nintedanib has also been shown to inhibit the migration of fibroblasts in idiopathic pulmonary fibrosis (39).

Previous studies from our group and others have reported that FGFR1 blockade enhances the effect of chemotherapy in MPM (12) and in other malignancies (e.g., in breast cancer; ref. 40). Also, PDGFR targeting has been shown to synergize with paclitaxel and doxorubicin in breast cancer in vitro and in vivo (28). Moreover, copy-number gains of VEGFR2 induced platinum resistance in lung cancer cells that could be reduced by knocking down VEGFR2 signaling by using small interfering RNA (27). In this study, we observed an additive effect between nintedanib and cisplatin in human MPM cells in vitro.

As for the in vivo anti-MPM potential of nintedanib, we demonstrated that upon IP administration, the drug significantly prolongs the survival and reduces body weight loss of mice bearing orthotopic human MPM xenografts. Interestingly, although animals treated with nintedanib PO also had longer survival times, this tendency did not reach statistical significance. One possible explanation for this difference is that IP administration of nintedanib results in higher intrathoracic drug concentrations. In support of this, the peritoneal and the thoracic cavities have been reported to be directly connected via the lymphatic capillary system of the diaphragm in both mice and humans (41, 42). Although the mechanism and efficacy of this potential transdiaphragmal drug transfer need to be further investigated, chances are that although nintedanib is an orally available and approved drug, in MPM it might exert a more potent antitumor effect when administered IP.

Based on the above-described considerations, in the further combination experiments, both nintedanib and chemotherapy were applied IP. Perhaps, the most striking findings of this second set of animal experiments were that there is a significant efficacy advantage in combining standard chemotherapy with nintedanib, that nintedanib and bevacizumab have comparable in vivo anti-MPM potential in tumors with high baseline VEGF-A expression and, moreover, that nintedanib is superior to bevacizumab in human MPM growth inhibition in tumors with low baseline VEGF-A expression. Our data, thus, suggest that nintedanib is as effective as bevacizumab, the reference angiogenesis inhibitor in clinical practice (7, 43), in human MPM xenografts with high baseline VEGF-A expression. Moreover, we also conclude that from the efficacy perspective, nintedanib in combination with chemotherapy can be considered as the preferred antiangiogenic regimen of choice for patients with tumors expressing low VEGF-A levels. This is in accordance with a previous study from Li Q and colleagues who demonstrated that bevacizumab inhibited the orthotopic growth of EHMES-10 MPM cells with high baseline VEGF-A expression but not that of “VEGF–low-producing” MSTO-211H MPM cells (30). However, it has to be kept in mind that baseline VEGF levels detected in our MPM cell lines may have changed during in vivo tumor development and, moreover, that the clinical translation of these results is also hampered by the lack of VEGF-A related or other biomarkers that have been validated in patients with MPM or other solid tumors to predict antiangiogenic treatment efficacy (44). Nevertheless, in further support of the more potent antiangiogenic and antitumor activity of nintedanib, we observed significantly reduced vascularization in nintedanib-treated tumors (vs. controls, with or without concurrent chemotherapy and formed by both p31 and SPC111 lines), but not in the bevacizumab-treated p31 or SPC111 tumors. Yet, it is noteworthy that there was a strong tendency for reduced vascularization in bevacizumab-treated p31 tumors with high baseline VEGF-A levels.

The abovementioned findings, taken together with our observation that the IC₅₀ values established in most of our in vitro assays correspond to higher drug levels than pharmacologically achievable nintedanib tissue concentrations in mammals (i.e., 200–450 nmol/L (16)), suggest that the key component of nintedanib's in vivo effect is its antivascular potential. In further support of this, our previous study revealed nintedanib IC₅₀ values in endothelial cultures about one order of magnitude lower than the IC₅₀s reported in the present study for MPM cells (16). It is important to mention here, though, that the antivascular and direct antitumor effects of antiangiogenic RTKIs might have partly opposing impacts on tumor progression. As we have reported recently, because antiangiogenic RTKIs destroy the tumor microvasculature, blood capillaries get separated by longer distances resulting in inadequate tumor tissue RTKI penetration and thus limited drug delivery to RTK-expressing tumor cells (26).

We also demonstrate in this study that nintedanib is similarly active in human MPM xenografts as a single agent and in combination with cisplatin/pemetrexed, indicating that—although they can have additive effects in vitro—the addition of nintedanib to chemotherapy does not provide a further in vivo MPM growth inhibitory effect. In accordance with our current data, Awasthi and colleagues showed that while nintedanib synergizes with gemcitabine in pancreatic cancer in vitro and exhibits potent antitumor effect in vivo, combined nintedanib/gemcitabine therapy does not provide synergistic tumor growth reduction in mice bearing human pancreas cancer xenografts (35).

It is currently not clear whether and how antiangiogenic drugs can improve the delivery of chemotherapy. According to the “vessel normalization theory” of Jain and colleagues, bevacizumab “normalizes” the chaotic tumor vasculature and thus increases chemotherapeutic drug delivery (45). However, recent clinical data raised serious concern that bevacizumab reduces, rather than improves, chemotherapy uptake by solid tumors (46). Also in contrast to the vessel normalization concept, Kutluk-Cenik and colleagues reported decreased capillary pericyte coverage and doxorubicin delivery in human pancreatic cancer xenografts following nintedanib therapy (18). One can also assume, therefore, that antiangiogenic RTKIs—which (in contrast to bevacizumab) are typically approved as monotherapies in the clinic (e.g., in renal and liver cancers)—may not promote normalization but, instead, might destabilize the vasculature and thus interfere with drug delivery by destroying the pericyte coverage of the tumor capillaries. This is because the PDGF/PDGFR axis is crucial for pericytes and the pericyte layer has a key role in the maintenance of vessel wall stability (45). In other studies, however, nintedanib “normalized” the vasculature in human A549 lung cancer xenografts (47) and also in an in vivo model of pulmonary fibrosis (48).

In summary, our current study provides evidence that nintedanib's target RTKs are (co)expressed in MPM cells and that this triple angiokinase inhibitor inhibits MPM growth and migration in vitro. Moreover, our data suggest that nintedanib exhibits potent VEGF-A expression-independent MPM growth inhibition in vivo thereby prolonging the survival of MPM-bearing animals. These preclinical findings support the concept of the LUME-Meso trial evaluating nintedanib in combination with cisplatin/pemetrexed in patients with unresectable MPM.

B. Hegedus and B. Dome report receiving commercial research support from Boehringer-Ingelheim. No potential conflicts of interest were disclosed by the other authors.

Conception and design: V. Laszlo, T. Klikovits, M. Grusch, F. Hilberg, B. Hegedus, B. Dome

Development of methodology: V. Laszlo, J. Tovari, B. Hegedus, B. Dome

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): V. Laszlo, Z. Valko, I. Kovacs, M.A. Hoda, T. Klikovits, D. Lakatos, A. Stiglbauer, M. Groger, C. Pirker, W. Berger, B. Hegedus

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): V. Laszlo, Z. Valko, J. Ozsvar, M.A. Hoda, T. Klikovits, A. Czirok, T. Garay, M. Groger, J. Tovari, C. Pirker, W. Berger, F. Hilberg, B. Hegedus, B. Dome

Writing, review, and/or revision of the manuscript: V. Laszlo, M.A. Hoda, T. Klikovits, A. Czirok, T. Garay, A. Stiglbauer, T.H. Helbich, C. Pirker, M. Grusch, W. Berger, F. Hilberg, B. Hegedus, B. Dome, W. Klepetko

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): V. Laszlo, M.A. Hoda, T. Klikovits, T. Garay, B. Hegedus

Study supervision: V. Laszlo, M. Grusch, B. Hegedus, B. Dome

This work was supported by Boehringer Ingelheim RCV GmbH, Vienna, Austria, the Herzfelder Family Foundation, the Initiative for Cancer Research of the Medical University of Vienna and from the Austrian National Bank (OeNB, 16912), (M. Grusch), the Hungarian National Research, Development and Innovation Office (AIK 12-1-2013-0041, K109626, K108465, KNN121510 and SNN114490, B. Dome; K116295, 16-1-2017-0439, J. Tovari), the Semmelweis University Start-Up grant (40148-11658, B. Dome), the Austrian Science Fund (I2872, B. Hegedus), and the Vienna Fund for Innovative Interdisciplinary Cancer Research (B. Dome). T. Garay is the recipient of a postdoctoral fellowship of the Hungarian Academy of Sciences.

The authors thank the excellent technical assistance of Barbara Dekan.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.