Antagonizing the Hedgehog Pathway with Vismodegib Impairs Malignant Pleural Mesothelioma Growth In Vivo by Affecting Stroma

Malignant pleural mesothelioma (MPM), an aggressive cancer most often caused by asbestos exposure, is still a public health concern. MPM arises from malignant transformation of the mesothelial cell layer lining the pleural cavity (1). About 80% of MPM cases are related to asbestos exposure, further causes are other environmental exposures and genetic predisposition. Incidence rates in Western European nations and Australia are still rising with expected peak around 2020 and beyond (2) due to long disease latency (3). The prognosis of MPM remains poor. Intensive multimodal regimen combining chemotherapy, surgery, and radiotherapy contributed to a median overall survival of 24 months (4). Currently, other treatment options including molecular targeted therapy as well as immunotherapy are being a focus of interest.

Recent findings of the activated Hedgehog (Hh) signaling pathway in MPM have underlined its role as a target for MPM treatment (5–7). The Hh pathway is one of the crucial stem cell signaling pathways that regulate embryonic development and adult tissue repair (8). In the absence of Hh ligand binding, the receptor Patched (PTCH1) represses the transmembrane G protein–coupled receptor Smoothened (SMO). Binding of ligands [Sonic hedgehog (SHH), Indian hedgehog (IHH), and Desert hedgehog (DHH)] to PTCH1 unleashes SMO to induce Hh pathway activation via an intracellular signaling cascade. The transcription factors glioma-associated oncogenes 1–3 (GLI1, GLI2, and GLI3) convey the cytoplasmic signal and induce transcription of proproliferative and antiapoptotic target genes, such as GLI1, PTCH1, Hedgehog interacting protein (HHIP), CyclinD1, or BCL2 (8) depending on cell type and situation. Constitutive pathway activation resulting from mutations of pathway components is frequently detected in sporadic basal cell carcinoma and medulloblastoma (9,10). Furthermore, aberrant upregulation of Hh pathway in ligand-dependent (paracrine or autocrine) manner has been described in hematologic malignancies and several solid tumors (11). Studies in colon and pancreatic adenocarcinoma demonstrated the paracrine activation of Hh signaling (12,13). In these studies, human tumor cells cultured in vitro did not respond to a SMO antagonist. In contrast, the treatment with the SMO antagonist in subcutaneous tumor xenograft resulted in delayed tumor growth. Gene expression analyses suggested that only murine stroma but not tumor epithelium could respond to SMO antagonist explaining the lack of response in vitro (12,13). Coculture experiments further showed the paracrine activation of Hh signaling in lung and pancreatic cancer (14,15). These highlight the importance of Hh signaling in tumor and thus serve as a rational for a treatment targeting cancer (11).

In MPM, increased SHH and GLI1 mRNA expression was detected in tumor tissues compared with normal mesothelium (5). High GLI1 expression (5) and high SMO and SHH protein levels (6) were associated with poor survival of MPM patients. Recent data from The Cancer Genome Atlas demonstrated that the Hh pathway is among the top 10 most deregulated pathways in MPM (reviewed in ref. 16). Altogether, these data underline the importance of the Hh pathway in MPM. Nevertheless, mutations of Hh pathway components in MPM were reported to be relatively rare (17). So far, upregulation of the Hh signaling in MPM has been demonstrated in ligand-dependent autocrine manner, interestingly only mediated by DHH in the primary culture in 3% O₂ without serum (5). It is important to note that 2 of 6 MPM primary cells employed in that study were not responsive to a SMO antagonist (5) suggesting a possible role of paracrine activation in MPM. Indeed, here we provide new evidence regarding paracrine Hh activation in MPM. Using an immunocompetent orthotopic rat MPM model, we demonstrated that Hh pathway activation in stroma plays a role in the maintenance of tumor growth.

Eight diagnostic MPM tumor specimens and two tissue microarrays from 138 patients collected between 1999 and 2009 were employed. The study was approved by the institutional review board (ethical approval numbers: StV 29-2009 and EK-ZH 2012-0094) and all patients received and signed a written inform consent. Formalin-fixed paraffin-embedded tissues were cut into 3-μm thick slices and processed for IHC and scored semiquantitatively as described below.

Rat mesothelioma cell line IL45 was originally produced in rat exposed intraperitoneally to asbestos (18). IL45-expressing luciferase (IL45-luc) was generated by us in 2011 by stable transfection of IL45 with luciferase-expressing vector (19). Both IL45 and IL45-luc were maintained in RPMI1640 containing 1 mmol/L sodium pyruvate, penicillin/streptomycin (P/S), 200 μg/mL Geneticin (G418), and 10% FBS at 37°C with 5% CO₂. NIH3T3, employed for Hh activity assay as described in the previous publication (5), were maintained in RPMI (ATCC modification) supplemented with 10% FBS and 1% P/S at 37°C, 5% CO₂. Rat and mouse cell lines employed in this study were tested for the absence of Mycoplasma but were not authenticated by us. Primary culture of tumors was performed using the previously described method (6); and maintained in medium w/o serum (DMEM: F12, 0.4 μg/mL hydrocortisone, 10 ng/mL rat-EGF, 20 ng/mL rat-basic fibroblast growth factor, 10 μg/mL insulin, 5.5 μg/mL transferrin, 6.7 μg/mL selenium, 1 mmol/L sodium pyruvate, 100 μmol/L β-mercaptoethanol, nonessential amino acids, and 30% own conditioned medium) at 37°C, 5% O₂ and 3% O₂.

Vismodegib was prepared in DMSO and stored at −20°C. Working dilutions were freshly prepared in culture medium containing 0.5% FBS. A viability assay of the IL45-luc monolayer cells was performed using Methylthiazolyldiphenyl-tetrazolium bromide (MTT) after 72-hour treatment. Primary IL45-luc derived from tumors maintained in 3% O₂ without serum formed clumps and spheres. For this reason, the acid phosphatase (APH) assay using ImmunoPure p-nitrophenyl phosphate substrate (20) was employed to measure their viability. Colony formation assay was performed in 0.5% FBS. Cells were treated with vismodegib for 10 days, and stained with crystal violet and analyzed under a light microscope where colonies with ≥5 cells were counted. All results were confirmed in at least three independent experiments.

Supernatant from primary culture of rat tumors were harvested freshly, after centrifugation at 400 g for 5 minutes. Supernatant from a primary human MPM culture was harvested and kept at 4°C until used. NIH3T3 cells were seeded in a 6-well plate at 900,000 cells/well in the complete medium. After overnight incubation at atmospheric oxygen level, the medium were then replaced by supernatant from primary culture mixed with 0.05% DMSO or vismodegib. SMO agonist (SAG: Abcam; Ab142160) was employed as a positive control. SAG powder was dissolved in DMSO and stored at −20°C. Working dilution of SAG was freshly prepared and added to culture medium used for rat primary culture. After receiving the treatment, NIH3T3 were incubated at 37°C, 5% CO₂ and 3% O₂. Cells were harvested at 72 hours later for RNA extraction. The results were confirmed by at least two independent experiments except for the treatment with human MPM supernatant where no material was available for the duplication.

The experimental scheme is depicted in Fig. 3A. The animal experiment was authorized by the veterinary office of canton Zurich, Switzerland (112/2013). Male Fischer 344 rats (n = 12, 200–250 g; Harlan Laboratories) were housed according to the guideline for animal welfare. A sterile 50 μL of 500,000 IL45-luc cell suspension in Dulbecco's PBS (DPBS) was implanted underneath the parietal pleura using a surgical procedure as described previously (21). To start with the same tumor size, the treatment started when the tumor was first detectable by bioluminescence (BLI; between 2 and 6 days after implantation). Vismodegib free base (LC laboratories, V-4050) was dissolved in 0.5% methylcellulose/0.2% Tween (MCT) solution and sonicated. The drug suspension was freshly prepared and administered once daily for 6 days by oral gavage at a dosage of 100 mg/kg (22). Control animals received only MCT. Once the treatment started, BLI was performed every second day until day 6. Magnetic resonance imaging (MRI) was performed after BLI measurement at day 6 prior to euthanasia. Macroscopic tumor volume was measured at autopsy using the ellipsoid volume formula V = 4/3 π × a/2 × b/2 × c/2 (a = length, b = width, c = thickness measured on H&E sections).

MRI was performed using a 4.7 Tesla small-animal MR scanner equipped with a ¹H whole-body rat coil (Bruker 4.7-T PharmaScan 47/16 US). MR data sets were acquired applying T2-weighted fast spin-echo sequences in transverse/sagittal/coronal orientation (repetition time TR 3,500 ms, echo time TE 45 ms). The tumor volume (mm³) was quantified blindly with ImageJ software (NIH, Bethesda, MD) by two observers as we described previously (23).

BLI was performed in the imaging chamber of the IVIS200 (Caliper Life Sciences, Inc.). Prior to the imaging procedure, rats were injected with 150 mg/kg body weight Xenolight d-Luciferin –K+ salt (Perkin Elmer, 122796; dissolved in sterile DPBS). BLI measurement was recorded over 30 minutes with the exposure time of 60 seconds. When the maximum BLI signal was reached (in about 15–25 minutes), region of tumor was measured for BLI signal intensity (total flux; photons/second).

RNA extraction was performed using TRIzol reagent (Ambion, 15596-018) according to the manufacturer's instructions. Five-hundred nanograms of total RNA was used for cDNA synthesis using RT² first strand kit (Qiagen, 205311). Quantitative real-time PCR was performed using Power SYBR Green PCR Mastermix (Applied Biosystems, 4368577). Primers are listed in the Supplementary Table S1 or were described previously (24,25). The reaction was performed in triplicate and normalized with a loading control, Histone H3 for rat cells and Beta actin (Actb) for mouse cells (26). Relative quantitation (RQ) of gene expression was calculated using the 2^−ΔΔC_t method. For in vivo samples, the following relative quantitation was applied RQ = 2^{−(C_t,target-C_t,H3)/(average C_t,target-C_t,H3 of control group)}.

Tumor tissues were collected and immediately fixed in formalin. They were decalcified in neutral EDTA pH 7.0 for 2 to 4 weeks before paraffin embedding. Tissues were cut into 3-μm slices and rehydrated through series of graded alcohol followed by IHC using the following antibodies: GLI1 (Santa Cruz Biotechnology, H-300), Fibronectin (Abcam, ab6328), DHH (Santa Cruz Biotechnology, H-19), PTCH1 (Aviva Systems Biology, ARP44249_P050), HHIP (Santa Cruz Biotechnology, H-280), Ki-67 (Cell Marque, CMC27531021), phospho-histone H3 (ser10; Cell Signaling Technology, #9701).

GLI1 and DHH antigen retrieval was achieved by 20-minute microwave (700 watt) cooking in sodium citrate buffer pH 6.0. After overnight incubation with primary antibodies at 4°C, the secondary antibody was applied followed by Vectastain ABC Reagent (Vector Laboratories). DAB chromogen (Dako) was applied for visualization of the signals and counterstained with hematoxylin. PTCH1 and HHIP were stained with Dako Autostainer Link48 Instrument (Dako Denmark A/S). Antigen retrieval for PTCH1 was performed using target retrieval solution high pH (Dako, K8004; 20 minutes, 97°C). For HHIP, target retrieval solution low pH (Dako, K8005; 20 minutes, 97°C) was employed for rat tissues and target retrieval solution high pH (Dako, K8004; 20 minutes, 97°C) was employed for human tissues. The visualization system consisted of the Dako EnVision Rabbit/HRP/DAB system and hematoxylin as counterstain. TUNEL staining was conducted on the Leica-bond RX automater using digoxigenein-labeled dUTP (Roche, 11570013910) and terminal deoxynucleotidyl transferase (Promega, M1875) after enzymatic retrieval (Sophistolab AG).

Images were captured with a light microscope (Leica DM6000). H-scoring system was applied to semiquantitatively quantify protein expression by three observers (K. Bérard, M. Meerang, and B. Vrugt). Staining intensity was scored 0–3 (negative-strong). Staining frequency was set as 0.1 (<10% positive cells), 0.5 (10%–50% positive cells), 1 (>50% positive cells). The H-score was the multiplication of intensity and frequency (range 0–3). Proliferation, mitotic, and TUNEL-positive indices were calculated by counting positive tumor nuclei per total tumor cells (at least 1,000 total cells).

The differences between means were verified by unpaired two-sided t tests using the GraphPad Prism software. A statistically significant difference was defined as P ≤ 0.05.

Eight MPM tumor specimens including 6 epithelioid, 1 biphasic, and 1 sarcomatoid were analyzed for the expression of GLI1, DHH, PTCH1, and HHIP by IHC (Fig. 1A). In addition, two TMAs consisting of 138 patients (89 epithelioid, 41 biphasic, and 8 sarcomatoid) were analyzed for GLI1 and DHH expression. GLI1 is expressed in both tumor and stromal fractions in all the samples assessed. When comparing tumor and stroma, DHH, PTCH1, and HHIP expression is higher in the tumor fraction (Fig. 1A).

To investigate the role of Hh signaling in MPM stroma, we employed our previously described rat MPM model, generated by the implantation of a rat mesothelioma cell line, IL45-luc, underneath parietal pleura of immunocompetent rats (19). IL45-luc cells cultured in vitro in the presence of 10% FBS expressed high levels of Gli1, Ptch1, and Smo, however, mRNA levels of all Hh ligands were very low or undetectable. Interestingly, a robust increase of Dhh mRNA level was detected in in vivo tumor compared with in vitro cell culture (Fig. 2A). Expression levels of other Hh pathway components, Gli1, Gli2, and Gli3 were reduced in tumor derived from IL45-luc compared with in vitro. Hhip, Hh pathway target gene and its negative regulator, became detectable only in vivo (Fig. 2A). Primary cells isolated from IL45-luc–derived tumors, cultured in 3% O₂ without serum continued to express Dhh but expressed only low levels of Gli1 when compared with in vivo or IL45-luc cultured in the presence of 10% FBS (Fig. 2A).

Histologic assessment of the orthotopic tumors revealed that tumor cells were surrounded by stroma including extracellular matrix (ECM)/lymphocytes/fibroblasts/blood vessels (Fig. 1B). Similar to human MPM, immunohistochemical analysis revealed predominant cytoplasmic localization of DHH in tumor cells but not in the stroma (S), whereas GLI1 was expressed in both tumor (T) and stroma fractions (Fig. 1B).

In terms of the sensitivity of these cell lines to Hh pathway inhibition in vitro, IL45 and IL45-luc responded to vismodegib at the same extent after 72-hour treatment (Fig. 2B). Growth-inhibitory effect was observed when using > 20 μmol/L of vismodegib in both cell lines. This is in line with the literature reporting the effective concentration of vismodegib in vitro (7,27). However, this concentration is much higher than the effective concentration reported for Hh-responsive cell lines (13,27). Vismodegib could not suppress the expression of Hh target genes, Gli1 and Ptch1 (Fig. 2C). We therefore employed colony formation assay to assess long-term treatment effects (10 days). Although, a lower amount of vismodegib was required to inhibit colony formation of IL45 and IL45-luc cells (Fig. 2B) compared with short-term treatment, this effective concentration was still higher than that required for Hh pathway inhibition (27). The unresponsiveness of IL45 and IL45-luc to vismodegib in terms of Hh target genes downregulation in vitro implied the lack of constitutive activation of Hh pathway in these cell lines.

Thus, we investigated the effects of the SMO antagonist, vismodegib, in the rat MPM model to further elaborate the role of Hh signaling. The experimental scheme is depicted in Fig. 3A. SMO antagonist, vismodegib, was administered with the effective dosage applied in preclinical studies (22). Six days later, we observed that tumor volume measured by MRI was significantly reduced in the treated group compared with control (3-fold difference, P = 0.03; Fig. 3B). Consistently, significant reduction in macroscopic tumor volume (P = 0.03) of the treated group was detected (Fig. 3D). Tumor growth monitored by BLI seems to be reduced in the treated group compared with control (see images in Fig. 3E). Nevertheless, the overall growth suppression was not statistically significant comparing control with treatment group, which may be in part due to the high standard deviation of the measurement (Fig. 3E).

Robust reduction of mRNA levels of Gli1 Hhip and Ptch1 was observed in vivo after vismodegib treatment compared with control group (Fig. 4A). No significant change in the expression of Dhh, Gli2, and Gli3 was detected (Fig. 4A). The analysis in skin samples also showed strong downregulation of Gli1 (Fig. 4A), confirming the efficacy of vismodegib in the inhibition of the Hh pathway in this rat model. We further analyzed GLI1, PTCH1, HHIP, and DHH expression by IHC. Tumor fraction was clearly distinguishable based on large cell size, big nuclei, less spindeloid shape than fibroblast (Fig. 1B). Slightly reduced GLI1 expression in tumor cells was observed (Fig. 4B and C). Interestingly, we observed significant downregulation of GLI1 and HHIP in the stromal fraction of treated rats compared with controls (Fig. 4B and C). We did not detect any change in PTCH1 expression on the protein level, may be because the change in mRNA expression levels was small as compared with GLI1 or HHIP (Fig. 4A). DHH expression levels remained unchanged in the treated rats (Fig. 4B).

Immunohistochemical analysis of Ki-67 and p-H3 (phospho-histone H3) indices was performed to assess proliferation and mitosis of tumor cells, respectively (Fig. 4D). Significant reduction of both Ki-67 and p-H3 index in the treated group versus control was detected. TUNEL staining showed a very low percentage of dead cells without central necrosis in both control and treated tumors. No significant difference of TUNEL positivity was observed in the treated group compared with controls (Fig. 4D).

We previously observed that primary human MPM cells cultured in the physiologic condition [in own conditioned medium with growth factors (see Materials and Methods) and 3% O₂] could sustain stemness and DHH secretion (5). Therefore, we isolated primary cells from the orthotopic rat tumors and cultured them in this condition. We indeed detected increased Dhh expression in primary cells cultured in 3% O₂ without serum compared with IL45-luc cultured in 10% serum and 20% O₂ (Fig. 2A). This finding is consistent in primary cells isolated from three different animals (Supplementary Fig. S1). We could not observe any correlation between Gli1 and Dhh expression in these cells (Supplementary Fig. S1). The response of primary cells to vismodegib is identical to what we observed in IL45-luc cultured in serum (Fig. 5A). No difference in response to vismodegib was detected in cells isolated from control and vismodegib-treated rats (Fig. 5A). No downregulation of Gli1 or Ptch1 was observed under this culture condition in response to SAG or vismodegib treatment (Fig. 5B and Supplementary Fig. S2). The response of tumor cells to vismodegib (determined by IC₅₀) did not correlate with the expression levels of Hh pathway components (Dhh, Smo, Ptch1, Gli1; Fig. 5C).

We could not observe downregulation of other canonical Hh target genes, namely, CyclinD1 and Abcg2 (Fig. 6A; refs. 28,29). As in our model, we observed that vismodegib treatment induced stronger GLI1 downregulation in the stromal compared with the tumor compartment, we hypothesized that Hh pathway activation occurred primarily in a paracrine fashion. We therefore focused on the effects of vismodegib in stroma to understand the mechanisms leading to the reduced proliferation hence tumor growth delay. Previous experiments suggest that antagonization of Hh pathway in fibroblasts reduces the production of ECM, cytokines, and growth factors (13–15). Therefore, we analyzed mRNA levels of the ECM molecules Fibronectin (Fn1) and Collagen (Col1a2). We did not detect differences in Collagen mRNA levels, may be due to the fact that the expression of collagen is too abundant (Col1a2 Ct∼18 vs. loading control Histone H3 Ct∼22) to allow us to observe any difference. Nevertheless, we could detect a downregulation of Fn1 mRNA levels in the treated group (Fig. 6A). We further analyzed Fibronectin expression by IHC and observed significantly reduced levels of Fibronectin in the treated group compared with controls (Supplementary Fig. S3). Thus, one possibility is that the difference in tumor volume between control and vismodegib-treated tumor might be in part due to reduced ECM content. In addition, SHH was shown to stimulate the secretion of VEGF and LIF (14) from fibroblasts. Indeed, in agreement with this finding, we observed here in vivo, a significant reduction on Vegf and Lif mRNA in the treated group compared with control (Fig. 6A).

To confirm paracrine activation of Hh signaling between tumor cells and stroma, we employed in vitro assay using mouse embryonic fibroblast cells (NIH3T3). NIH3T3 were treated with the supernatant collected from primary cells derived from a rat tumor. Seventy-two hours after the incubation with rat MPM supernatant in 3% O₂, we observed a strong upregulation of Gli1 and Ptch1 in NIH3T3 cells which was reduced by the treatment with vismodegib. The treatment with 50 nmol/L SAG induced slight Gli1 upregulation but not Ptch1 (Fig. 6B). This may be explained by the fact that Ptch1 expression is not strongly induced by the Hh pathway activation compared with Gli1 [e.g., culture supernatant of primary cells induced four times Gli1 upregulation but only 1.8 times Ptch1 upregulation (Fig. 6B)]. NIH3T3 treated with SAG and rat supernatant showed upregulation of Fn1 which was suppressed by vismodegib treatment. Similar to the effect seen with the rat MPM cell culture supernatant, NIH3T3 treated with primary human MPM cell culture supernatant showed upregulation of Gli1, Ptch1, and Fn1 (Supplementary Fig. S4).

In this study, we provide new evidence regarding the importance of Hh signaling in MPM stroma in the maintenance of MPM tumor growth. The Hh pathway was activated in both tumor and stroma of MPM human tissues and the orthotopic immunocompetent rat model. After continuous treatment with the SMO antagonist, vismodegib, in vivo, pronounced and significant downregulation of the well-known Hh target genes, GLI1 and HHIP was observed only in the stromal fraction. The tumor fraction was less responsive to vismodegib in vivo. This is further demonstrated in vitro where we could not see the effect of vismodegib in the downregulation of Gli1 in primary cells isolated from the rat MPM model. The outcomes of daily treatment with vismodegib targeting Hh pathway in the stroma included reduced tumor volume and suppressed tumor growth without inducing cell death. The treatment with vismodegib repressed indirect Hh-responsive genes previously reported in fibroblasts, that is, Fn1, Vegfa, and Lif. Reduced levels of these factors may thus be responsible for tumor growth delay observed in this study.

We detected the expression of Hh pathway components GLI1, HHIP, and PTCH1 in human MPM specimens in both tumor and stromal fractions. Hh ligand, DHH was weakly expressed and was predominantly localized in the tumor fraction. Similar to our finding, previous studies also showed that another Hh ligand detected in human MPM, SHH, is also weakly expressed in only 47% of patients (7). The similar pattern of GLI1, PTCH1, HHIP, and DHH expression was observed in orthotopic immunocompetent rat mesothelioma model. IL45-luc cells that were implanted orthotopically in syngeneic rats started to express Dhh but not the other Hh ligands. Nevertheless, Gli1, Gli2, and Gli3 expression levels were reduced in vivo. One most possible explanation is the presence of 10% FBS in in vitro culture medium. We observed that when the FBS was withdrawn from IL45-luc, Gli1 expression was strongly reduced (unpublished observations). The upregulation of Gli1 in the presence of FBS is presumably mediated by Hh independent mechanism as its expression could not be suppressed by the treatment with vismodegib. The expression of Dhh was still maintained when isolated and cultured without serum in 3% O₂. This observation was also documented in primary culture of human MPM cells cultured in 3% O₂ without serum in which DHH was the only ligand expressed (5). So far, there is no obvious explanation and it remains to be elucidated why the primary culture and this in vivo MPM models upregulate only the expression of Dhh but not other ligands.

Target genes responding to Hh pathway stimulation have been shown to be dependent on cell type and context. SHH has been shown to indirectly induce angiogenesis in juvenile and adult mice by upregulating angiogenic factors including Vegf (30). In this study, it was also demonstrated in vitro that only fibroblasts could secrete VEGF upon Hh pathway stimulation. Another finding showed in vitro that lung fibroblasts upregulated LIF and VEGF expression when treated with cell culture supernatant of lung cancer cell lines that secrete SHH (14). In line with these studies, we also observed here, in vivo, downregulation of Lif and Vegfa expression following Hh pathway suppression by vismodegib. In addition, cell culture supernatant of primary rat and human MPM cells could stimulate Gli1, Ptch1, and Fn1 expression in NIH3T3 when incubated at 3% O₂, 37°C for but not at atmospheric O₂ level. Under this culture condition, we did not detect an increase in Lif and Vegfa expression in NIH3T3 cells treated either with cell culture supernatant or SAG (unpublished observations). One possible reason is that the tested in vitro condition does not fully recapitulate in vivo tumor microenvironment. Although Fn1, Lif, and Vegfa can also be produced by some MPM cells, (31–33), it is not known whether these are Hh-responsive genes in tumor. On the basis of our results being (i) stroma was the main part affected by vismodegib (ii) Fn1, Lif, and Vegfa have been so far described to be regulated by the Hh pathway only in fibroblast, we presume that the effects of vismodegib that we observed herein results primarily from the modulation of fibroblasts. Interestingly, Vegfa promoters do not contain GLI1 binding consensus sequence (14). Thus, the effect of vismodegib on Vegfa expression could mostly stem from indirect regulation. One mediator could be hypoxia as a study in pancreatic cancer demonstrated that SHH secreted from tumors stabilized HIF1α, a known transcription factor of VEGF, in pancreatic fibroblasts (34).

In vitro, high concentration of vismodegib (>20 μmol/L) was required to suppress the growth of IL45 and IL45-luc cells. This concentration is in the range reported for most human cancer cell lines having ligand-dependent activation as well as human MPM cells (IC₅₀ = 15–20 μmol/L; refs. 7,27). Nevertheless, the concentration used in these studies was far higher than the effective concentration (100–500 nmol/L) of vismodegib required to completely inhibit Hh pathway activity of responsive cell lines (13,27). In the contrast to these studies, we could not observe a significant reduction of Gli1 expression after exposing IL45 cells to various concentrations of vismodegib. Similarly, a study in various other cancer cell lines showed that the treatment with Hh antagonist as well as ligand-blocking antibody did not induce downregulation of GLI1, the well-described Hh target gene in vitro (13). Thus, it is still a matter of debate whether the growth inhibitory effect of SMO antagonist observed in vitro is only due to off-target effects when applied at high concentration. Vismodegib was shown to inhibit ABCG2 drug transporter (35) as well as cellular calcium homeostasis (36) when applied at a high concentration. Thus, the lack of Gli1 downregulation after treatment with vismodegib in vitro suggested that the growth inhibitory effect observed herein may result from off-target effects. Nonetheless, given the heterogeneic nature of MPM, the effect may depend on the cell lines employed. Indeed, Shi and colleagues (5) described that only 4 of 6 cell lines cultured in 3% O₂ without serum showed downregulation of GLI1 in response to SMO antagonist treatment.

GLI1 has been shown to be activated in Hh-independent fashion, via cross-talking with other oncogenic pathways such as KRAS, EGFR, and mTOR (8,37). A recent study demonstrated that GLI1 was highly expressed in some MPM tumors of which SHH expression was absent (7), suggesting a possible role of GLI1 activation in Hh signaling independent manner (13). A recent study suggests a better proliferation inhibitory effect of GLI1 inhibitor (GANT61) compared with vismodegib on MPM cells (7). Nevertheless, the role of noncanonical ligand-independent Hh activation in MPM has to be explored in more detail. Besides, agents targeting GLI1 have not yet been available for clinical use.

We have the advantage of using an immunocompetent preclinical model that contains clinically relevant stromal components including ECM, fibroblasts, and immune cells. However, there are limitations of our study including short treatment and observation period; the major reason for this limitation being the aggressiveness of the model that influences animal well-being. Vismodegib was shown to inhibit the drug transporter, ABCG2 with the IC₅₀ of more than 1 μmol/L (35). ABCG2 was important for the efflux of BLI substrate luciferin, thus the treatment with vismodegib caused false-positive BLI signal (38). We indeed observed the effect of vismodegib in interfering luciferin efflux in IL45-luc at concentration of >1 μmol/L in vitro, [sustained BLI decline rate, Supplementary Fig. S5). A large proportion of circulating vismodegib [98% in rats (39)] is bound to serum alpha 1-acid glycoprotein (AAG). An in vivo study in mice receiving 100 mg/kg demonstrated a C_max of 35 μmol/L for total vismodegib (22). Thus, we estimate to have approximately 0.7 μmol/L of unbound vismodegib in the rat serum. This concentration still does not interfere with BLI. Moreover, persisting BLI signal should be observed when ABCG2 was inhibited (38), but BLI kinetics after the substrate administration in control and treated animals are similar (Supplementary Fig. S5). These data ruled out the possible interference of vismodegib on the detection of tumor by BLI in this study.

Here, we employed an immunocompetent preclinical model of MPM to demonstrate that the Hh pathway inhibition in MPM affects the production of ECM and cytokines, factors crucial for tumor progression. Our study also provides novel information regarding the role of Hh signaling in MPM in stroma.

No potential conflicts of interest were disclosed.

Conception and design: M. Meerang, E. Felley-Bosco, D. Kenkel, R.A. Stahel, W. Weder, I. Opitz

Development of methodology: M. Meerang, K. Bérard, A. Boss, D. Kenkel, I. Opitz

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): M. Meerang, K. Bérard, O. Lauk, B. Vrugt, A. Boss, D. Kenkel, A. Broggini-Tenzer, I. Opitz

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): M. Meerang, E. Felley-Bosco, A. Boss, D. Kenkel, I. Opitz

Writing, review, and/or revision of the manuscript: M. Meerang, K. Bérard, E. Felley-Bosco, B. Vrugt, A. Boss, R.A. Stahel, I. Opitz

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): M. Meerang, D. Kenkel, A. Broggini-Tenzer, S. Arni

Study supervision: M. Meerang, E. Felley-Bosco, D. Kenkel, W. Weder, I. Opitz

The authors thank Prof. Martin Pruschy for providing the IVIS machine; Dr. Vet. Med. Margarete Arras, Dr. Vet. Med. Nikola Cesarovic, and Dr. Vet. Med. Thea Fleischman for their kind assistance in designing the anesthetic and analgesic protocols. The authors also thank Dr. Michaela Kirschner for thoroughly reading and providing insightful comments for the manuscript.

The project is financed by the Swiss National Science Foundation to I. Opitz (grant number: PP00P3_159269).

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.