Combined MEK and PI3K/p110β Inhibition as a Novel Targeted Therapy for Malignant Mesothelioma Displaying Sarcomatoid Features

Malignant mesothelioma (MM) arises mainly from the pleural (pl) and peritoneal (pe) mesothelium and less frequently from other sites, and it is strongly associated with asbestos exposure. There are three main histologic subtypes: epithelioid (EMM), accounting for approximately 60% of MM; sarcomatoid (SMM), constituting approximately 20%; and biphasic (BMM). BMM contain a variable proportion of tumor cells with epithelioid and sarcomatoid features. Patients with MM have an extremely poor prognosis, especially when the tumor displays sarcomatoid features (pure or biphasic), in part due to its aggressiveness and chemoresistance. Even with multimodality treatment, median survival is 14 to 18 months (1). There is currently no standard second-line treatment for MM, and bevacizumab is the only approved targeted therapy, in combination with cisplatin and pemetrexed (2). This is due, in part, to the limited understanding of the molecular pathogenesis of MM. Genetic analyses have allowed classifying MM and identifying druggable pathways (3–7). Sarcomatoid MM cases are underrepresented in all reported cohorts, constituting 3% to 9% of all specimens, hampering their molecular characterization and the identification of genetic alterations contributing to their poor prognosis and chemoresistance. Recently proposed molecular classifications of MM agree on the existence of a poor prognosis subtype, which includes sarcomatoid tumors. TP53 alterations and PI3K-mTOR and RAS/MAPK upregulation were proposed to associate with the nonepithelioid clusters (6–8).

The PI3K/AKT/mTOR pathway promotes cell proliferation and survival and is activated in most MM (9–11). The mechanisms leading to PI3K/AKT/mTOR activation have not been completely elucidated. Point mutations in genes coding for PI3K pathway components are rare in MM (6, 12), but loss of expression of PTEN, the lipid phosphatase that dephosphorylates the main PI3K product [PtdIns(3,4,5)P3, PIP3], appears much more frequent (16%–62%) when measured by IHC assays (13–16). Genetic MM mouse models, sometimes combined with asbestos exposure, have confirmed the key role of PI3K/mTOR activation and Trp53 inactivation in MM initiation and progression (17–19). Indeed, combined Pten and Trp53 deletion in mouse pleural or peritoneal mesothelium leads to sarcomatoid and biphasic MM (19).

PI3K members belong to three classes; class I PI3K are subdivided into IA and IB based on their structure and the activating mechanisms. Class IA PI3K members are heterodimers of a catalytic (p110) and a regulatory subunit (p85). Four genes (PIK3CA, PIK3CB, PIK3CD, and PIK3CG) code for individual p110 isoforms (p110α, p110β, p110γ, p110δ). p110α and p110β are ubiquitous, while p110δ and p110γ expression is enriched in hematopoietic cells. p110α signals downstream of receptor tyrosine kinases (RTK), whereas p110β has a prominent role downstream of G protein-coupled receptors (GPCR; refs. 20, 21). PIP3 activates AKT, which, in turn, signals to mTOR via TSC1/2 phosphorylation.

Here, we report that concurrent PI3K activation and Trp53 deletion in the peritoneal mesothelium lead to aggressive sarcomatoid MM, confirming previous findings (19). We find that Pten deletion drives more aggressive tumors than Pik3ca mutational activation. We confirm that human SMM display low PTEN expression. These mouse model–based strategies allowed identifying the mechanisms responsible for the aggressiveness of SMM, which revealed combined p110β/MEK inhibition as a novel targeted therapy for MM with sarcomatoid features.

Pten^lox/lox and Trp53^lox/lox strains were previously reported (22, 23). To generate Pik3ca^H1047R knockin mice, the Pik3ca region encompassing exon 18 to 3′-untranslated region (3′UTR), comprising exons 18 to 20, was substituted by homologous recombination by a cDNA including wild-type (WT) exons 18–20–3′UTR, followed by PGK-NEO for neomycin selection. These sequences were flanked by LoxP sites; following the 3′ LoxP site, the recombinant sequence included the cDNA of exons 18 to 20 with the 3141A<G substitution, coding for the H1047R mutation, followed by IRES GFP-Luc, and the 3′UTR (Fig. 1A). The construct was electroporated into embryonic stem cells (ESC), which were selected with neomycin. The recombinant sequence was analyzed by PCR. Positive cells were transferred into pseudo-pregnant female mice. ESC electroporation, selection, and transfer followed standard protocols. Mice born after ESC transfer were genotyped by Southern blot and PCR, crossed with WT C57BL/6 mice, and pups were genotyped to verify germ line transmission. Mice from all the indicated strains (mixed background) were bred and maintained in specific pathogen-free conditions. All animal procedures were approved by the Ethics Committee for Research and Animal Welfare of Instituto de Salud Carlos III and the General Guidance of the Environment of Madrid Community and performed following guidelines for Ethical Conduct in the Care and Use of Animals as stated in The International Guiding Principles for Biomedical Research involving Animals, developed by the Council for International Organizations of Medical Sciences.

Mouse tissues were dissected, fixed in 10% buffered formalin, and paraffin-embedded. Human MM tissue microarrays (TMA) and associated data were obtained through the National Mesothelioma Virtual Bank (University of Maryland, College Park, MD; University of Pittsburgh Medical Center, Pittsburgh, PA; and Roswell Park Cancer Institute, Buffalo, NY). Antigen retrieval was performed by boiling in citrate buffer pH 6.0; endogenous peroxidase was inactivated with 3% H₂O₂ in methanol. Sections were blocked, rinsed, and incubated with specific antibodies (Supplementary Table S1) and species-specific secondary horseradish peroxidase (HRP)–conjugated antibodies (EnVision Systems, DAKO-Agilent Technologies). Reactions were developed using DAB (DAKO-Agilent Technologies). Histoscores of human TMAs were calculated as the product of the proportion of reactive cells (0%–100%) and intensity (0–3). The average of the histoscores from each patient was used for analysis. RNAscope ISH for Pd-l1/Cd274 and Ppib (used for quality control) on formalin-fixed paraffin-embedded mouse tumors was performed using the RNAscope 2.5 HD Assay-RED (Advanced Cell Diagnostics) according to the manufacturer's instructions. For mouse IHC and RNAscope quantification, hematoxylin and antigen-specific signals were detached into separated channels by Colour Deconvolution plugin of Image J software and channel-specific pixel area quantified using the same software. DAB area was normalized to the hematoxylin area; Pd-l1/Cd274 RNAscope signal was normalized to Ppib signal.

Frozen tissues and cultured cells were lysed in radioimmunoprecipitation buffer (20 mmol/L Tris-HCl, pH 8.0, 137 mmol/L NaCl, 1 mmol/L MgCl₂, 1 mmol/L CaCl₂, 10% glycerol, 1% NP40, 0.5% sodium deoxycholate, and 0.1% SDS) containing protease and phosphatase inhibitors. Proteins fractionated by SDS-PAGE were transferred to nitrocellulose membranes, which were incubated with primary antibodies (Supplementary Table S1) and HPRP-conjugated secondary antibodies (DAKO-Agilent Technologies). Reactions were developed using Luminata Classico HRP substrate (Merck-Millipore). Phospho-specific and total protein antibodies were either applied independently to different membranes, and processed in parallel, or used sequentially with the same membrane, after stripping with Restore Western Blot Stripping Buffer (Thermo Scientific). Phospho-RTK array (R&D Systems) and RAS Activation Assay kits (Merck-Millipore) were used following manufacturer's guidelines. Image J software was used for Western blotting signal quantification.

Total RNA was isolated from frozen tissues and cultured cells using GenElute Mammalian Total RNA Miniprep Kit (Sigma-Aldrich). DNA was eliminated using the DNA-free Kit (Ambion-Life Technologies). Reverse transcription was performed using TaqMan reverse transcription reagents (Roche-Life Technologies); for PCR, an ABIPRISM 7900HT instrument (Applied Biosystems-Life Technologies) and the SYBR Green PCR Master Mix (Applied Biosystems-Life Technologies) were used. Changes in gene expression were calculated using the quantitative ΔΔC_t method and normalized against Hprt. To assay for expression of mutant Pik3ca mRNA, a region of the gene encompassing exons 19 and 20 was amplified by PCR from cDNA. PCR products were treated with ExoSAP-IT (Affymetrix-Thermo Fisher Scientific), and the WT and the H1047R alleles were identified using a specific primer and the ABI PRISM SNaPshot Multiplex Kit (Applied Biosystems-Life Technologies). All primer sequences are provided in Supplementary Table S2.

Total RNA (2 μg) from Pik3ca*;Trp53-null and Pten;Trp53-null mice (n = 6 each) was used. RNA integrity numbers ranged from 7.5 to 9.4, assayed on an Agilent 2100 Bioanalyzer. PolyA+ RNA was extracted and randomly fragmented, converted to double-stranded cDNA, and processed through enzymatic treatments of end-repair, dA-tailing, and adapter ligation following "TruSeq RNA Sample Preparation Guide." The library was produced by limited-cycle PCR with Illumina PE primers (8 cycles). The purified cDNA library was applied to an Illumina flow cell for cluster generation (TruSeq cluster generation kit v5) and sequenced on the Genome Analyzer IIx with SBS TruSeq v5 reagents.

Image analysis and per-cycle base-calling were performed with Illumina Real Time Analysis software (RTA1.13). Conversion to FASTQ read format with the ELAND algorithm (v2e) was performed with CASAVA-1.8 (Illumina Quality check was done via fastqc v0.9.4, Babraham Bioinformatics). Raw reads were aligned with Tophat5 (version 2.0.4) to mouse genome GRCm38/mm10. Gene expression normalized counts (TPMs) and differential expression (Pik3ca*;Trp53-null vs. Pten;Trp53-null mice) was done with DeSeq2 (version 2.0.2). Raw and processed sequencing data are available in The Gene Expression Omnibus repository (GSE138389).

RNA-seq data (6) were downloaded from the European Genome-phenome Archive under accession code EGAS00001001563; data were normalized by the Upper Quartile method. Normalized data from The Cancer Genome Atlas (TCGA) were downloaded from cBioPortal for Cancer Genomics (24).

For mouse tumor cell isolation, tumors were digested in NB8 collagenase (1.5 mg/mL in HBSS; LabClinics) and maintained in DMEM supplemented with 10% FBS, 1 mmol/L Na pyruvate, and nonessential amino acids (Gibco-Life Technologies), and incubated at 37°C in a 5% CO₂ and 3% O₂ atmosphere. All experiments were performed at <15 passage. Human primary mesothelioma cultures (MPM.04, 12, 24, 28, 29, 34, 35, 37, 47, 59, and 60) were established at INSERM U.1138. MESO.49, 62, 80, 84, 92, 103, 161 and VMC.40, 48 were established at the Medical University of Vienna. Cells were grown either in RPMI 1640 (VMC40 and Meso161) or in DMEM supplemented with Glutamax (Gibco-Life Technologies) and 10% FBS at 37°C in a 5% CO₂ atmosphere. Human primary cultures were used at low-passage numbers (<20) and tested for Mycoplasma contamination using either MycoAlert Detection Kit (Lonza) or a home-made qPCR assay developed by the CNIO Genomics and Monoclonal Antibodies Units. Cell lines were authenticated based on specific gene mutations.

PD0325901 MEK inhibitor (1408) was purchased from Axon-Medchem, TGX-221 (S1169) was obtained from Selleckchem, and Pertussis toxin (P7208) and Cisplatin (C2210000) were from Sigma-Aldrich. Selumetinib and AZD8186 were kindly provided by AstraZeneca.

In vitro invasion assays were performed using BioCoat Matrigel Invasion Chambers (Corning) following the manufacturer's guidelines. Cells were fixed with 4% PFA, and nuclei were stained with DAPI (0.5 μg/mL in PBS). Cells were visualized using a Leica TCS SP5 WLL confocal microscope. IMARIS software v5.0 (Bitplane) was used for 3D image reconstruction and nuclear quantification. Mouse cell proliferation assays were performed in 96-well plates (Greiner Bio-one). Twenty-four hours after plating at 5 × 10³ cells/well (triplicates), inhibitors or vehicle was added; cells were incubated until fixation with 4% PFA. Visualization and image acquisition of DAPI-stained nuclei were performed using the confocal microscopy-based PerkinElmer's Opera high content screening system. Nuclei were counted using Acapella software (PerkinElmer). For proliferation assays, human primary cells were seeded at 5 to 7 × 10³ cells/well in triplicate in 96-well plates (Corning, Falcon). Cells were treated for 72 hours with increasing concentrations of single, or both, drugs; viability was quantified by CellTiter 96 AQueous One Solution Cell Proliferation Assay (MTS; G3582) or CellTiter Glo Luminescent Cell Viability Assay (G7573; Promega). The combination ratios for selumetinib and AZ8186 are specified in Supplementary Table S3. AUC and GI₅₀s for single drugs were calculated using GraphPad software 5.0 (Prism). Combination drug GI₅₀ and combination index (CI) were determined by Compusyn software (http://www.combosyn.com/).

Gαi2 was knocked down in mouse mesothelioma cells using MISSION shRNA Gαi2-shRNAs (SHC002, SHCLNG-NM_008138, and control sh; Sigma Aldrich). Targeting sequences are provided in Supplementary Table S2. Lentiviruses were produced in 293T cells using calcium phosphate transfection. Viral supernatant was collected 48 and 72 hours after transfection. Lentiviruses were transduced into mouse mesothelioma cells (0.2 × 10⁶ cells/6 well-plate) followed by Puromycin selection (1 μg/mL, 48 hours). Cells were harvested 24 hours later.

CT tumor–bearing Pten;Trp53-null mice were randomized into control and treatment groups. Selumetinib (10 mg/kg) and AZD8186 (40 mg/kg) were administered by oral gavage, twice daily, 5 days/week. Control mice received vehicle (0.5% methylcellulose, 0.2% Tween80).

Mice were anesthetized, and the abdominopelvic area was imaged using the eXplore Vista micro-CT scanner (GE Healthcare) without contrast. Micro-CT image acquisition consisted of 400 projections collected in one full rotation of the gantry in approximately 10 minutes using 80 kV and 450 μA X-ray tube settings. Reconstructed images were viewed and analyzed using AMIDE software.

Sample size was estimated on the basis of prior, or pilot, experiments, and no formal sample size analysis was performed. The Mann–Whitney U test was used to compare differences between two independent groups when the data did not follow a normal distribution. The Kruskal–Wallis one-way ANOVA was used when more than two groups were compared. The Log-rank (Mantel–Cox) test was used to compare survival distributions. The Fisher exact test was used in the analysis of the distribution of categorical values within two groups. GraphPad software 5.0 (Prism) was used. Human data statistics were performed by R software, ANOVA test was used to compare differences between group means, and Tukey test was applied as post hoc test. Two-sided P values <0.05 were considered significant.

Peritoneal MM developed in 8- to 12-week-old Pten^lox/lox;Trp53^lox/lox mice (referred to herein as Pten;Trp53-null) following inoculation of adenoviruses coding for Cre-recombinase into the bladder wall, as described (Fig. 1B; Supplementary Information; ref. 18). Recombination was evaluated using R26-LSL-LacZ reporter mice 2 weeks later (Fig. 1C; Supplementary Information). All inoculated Pten;Trp53-null mice (n = 26), but only 2 of 11 Trp53^lox/lox mice, developed tumors in the peritoneal aspect of the bladder wall, frequently invading contiguous fat and muscle (Fig. 1D). In addition, 15 of 26 (57%) mice developed peritoneal lesions. Tumor histology and latency recapitulate previously described Trp53;Pten-null nonepithelioid MM (19). Tumors were pleomorphic, enriched in spindle cells, and showed mild infiltration by lymphocytes and granulocytes. IHC showed heterogeneous but consistent expression of WT1 (9/9) and vimentin (12/12) and diffuse KRT5 expression in 4 of 12 tumors (Fig. 1E). These features, together with the histology, are characteristic of SMM.

Given the relevance of the PI3K/AKT/mTOR pathway in MM cell proliferation and survival (9, 10, 18), we assessed more broadly the impact of PI3K activation in a Trp53-null context. We analyzed Pik3ca^wt/H1047R;Trp53^lox/lox mice (referred to as Pik3ca*;Trp53-null), in which the H1047R hotspot mutation was conditionally knocked in the Pik3ca locus (Fig. 1A). After adeno-Cre inoculation, all Pik3ca*;Trp53-null mice developed tumors histologically undistinguishable from those of Pten;Trp53-null mice (Fig. 1E). However, the median survival of Pik3ca*;Trp53-null mice was significantly longer than that of Pten;Trp53-null mice (17.6 vs. 10.9 weeks after adenoviral inoculation, P < 0.0001; Fig. 1F). Importantly, <20% of Pik3ca*;Trp53-null mice (n = 24) developed macroscopic metastases (intraperitoneal, spleen, kidneys, liver, and gastrointestinal tract), compared with 50% of Pten;Trp53-null mice (n = 14; P = 0.024, Fisher exact test; Supplementary Fig. S1).

IHC analysis revealed that Pten;Trp53-null tumors displayed higher proliferation (phospho-Histone 3) and increased inflammation (macrophages and T cells) than Pik3ca*;Trp53-null tumors, with no differences in apoptosis (cleaved caspase-3; Fig. 2A and B). A refined analysis revealed increased Cd8⁺ cells, Ym1⁺ M2 macrophages, and Cd4⁺ FoxP3⁺ T regs in Pten;Trp53-null tumors (Fig. 2C and D). RNAscope analysis showed discrete areas of Pd-l1–expressing cells in Pten;Trp53-null tumors, but not in Pik3ca*;Trp53-null lesions (Fig. 2E and F). These data indicate that Pten inactivation associates with a more immunosuppressive stroma and metastatic phenotype than Pik3ca activation.

Given the relevance of Trp53 inactivation and PI3K activation in mouse SMM, we interrogated the PI3K pathway and upstream regulators in human SMM. PTEN loss, assessed by IHC, has been reported in 12% to 62% of unselected MM (11, 13–15), but its relevance to the sarcomatoid phenotype is unknown. PTEN expression was analyzed in TMAs including 149 clinically annotated specimens (National Mesothelioma Virtual Bank; Supplementary Table S4; ref. 25). PTEN immunostaining scores were categorized as negative (0), weak (1), moderate (2), and strong (3; Fig. 3A): 15.4% tumors scored negative (<20 immunoscore). PTEN-negative cases were enriched in SMM (P = 0.017, Fisher exact test; Fig. 3B). PTEN immunoscores were significantly lower in SMM than in epithelioid and biphasic tumors (Fig. 3C). Given the heterogeneity of biphasic MMs, we subdivided them according to the sarcomatoid or epithelioid components. Biphasic MMs with <50% epithelioid cells displayed lower PTEN levels, but differences were not significant when compared with those with >50% cells (P = 0.134). At the transcriptomic level, expression of PTEN was significantly lower in the sarcomatoid group of Bueno and colleagues (6), whereas expression of other PI3K effectors was increased. Despite following the same trend, differences were not significant when considering histology groups (Fig. 3D). A similar tendency was observed in the TCGA cohort (7) when clustered similarly (Fig. 3E; ref. 6). These data demonstrate that PTEN downregulation is common in SMM and validate the Pten;Trp53-null mouse model as a valuable tool.

To identify the mechanisms contributing to the aggressiveness of Pten;Trp53-null tumors, we analyzed PI3K and MAPK/ERK activation, known to cooperate in tumor progression. Akt Ser473 and Thr308 phosphorylation was significantly higher in Pten;Trp53-null tumors compared with Pik3ca*;Trp53-null tumors. Remarkably, pErk1/2 (Thr202 and Tyr204) and pMek1/2 (Ser217/221) were also higher in Pten;Trp53-null tumors (Fig. 4A–C). We used RAS-GTP pull-down assays to assess upstream signaling and did not find strong evidence of RAS hyperactivation in Pten;Trp53-null tumors, compared with Pik3ca*;Trp53-null tumors (Supplementary Fig. S2A and S2B). Downstream, we did not find consistent differences in phosphorylation of GSK3β Ser9 nor in the mTOR targets S6K and 4Ebp1 in Pten;Trp53-null tumors (Supplementary Fig. S2C and S2D).

To study the mechanisms involved in MEK/ERK activation and their contribution to the aggressiveness of Pten;Trp53-null tumors, we established murine MM cell lines (Fig. 4D). Expression of mutant Pik3ca mRNA in Pik3ca*;Trp53-null cells was confirmed using a SNaPshot assay (Supplementary Fig. S3A). Trp53 and Pten deletions were also confirmed (Fig. 4E; Supplementary Fig. S3B). We found higher pAkt levels in Pten;Trp53-null cells than in Pik3ca*;Trp53-null cells (Fig. 4E; Supplementary Fig. S3C), but the differences in pERK levels in exponentially growing cells from both genotypes were less apparent than in tumor tissues. MEK1/2 inhibition with PD325901 led to decreased pErk in cells of both genotypes (Supplementary Fig. S4A). MEK1/2 inhibition led to reduced in vitro proliferation and invasiveness of Pten;Trp53-null cells, whereas only proliferation was affected in Pik3ca*;Trp53-null cells (Fig. 4F and G). These results indicate that MAPK pathway activation contributes to the aggressiveness of Pten;Trp53-null tumors, although additional signaling pathways must be implicated.

MAPK activation resulting from PTEN loss has been reported, but no underlying unifying mechanism has been identified (26). A role of RTK—such as EGFR, EpHB4, MET, and AXL—in MAPK activation has been proposed in MM (10, 27). To explore if such mechanisms are involved in the differential activation of MAPK in Trp53;Pten-null mice, tumor lysates were analyzed by p-RTK dot blot arrays. We did not find strong evidence for a role of hyperphosphorylation of any of the 39 RTK tested. By contrast, selected receptors were preferentially hyperphosphorylated in Pik3ca*;Trp53-null tumors (Fig. 5A and B). Altogether, these results do not support RTK/RAS as the predominant mechanism responsible for MAPK activation in aggressive Pten;Trp53-null MM.

Next, we assessed whether GPCRs may participate in MAPK activation. To explore this notion, we used Pertussis toxin (PTX), which induces ADP-ribosylation of the αi subunits of heterotrimeric G proteins and impaired interaction with GPCRs, thus preventing downstream pathway activation (28). Erk phosphorylation was markedly reduced in PTX-treated Pten;Trp53-null cells but not in Pik3ca*;Trp53-null cells, whereas Akt phosphorylation was only modestly affected (Fig. 5C). The Gαi family consists of Gαi1-3, Gαo, Gαt1-3, and Gαz. RNA-seq data showed that Gαi2 is the main Gαi subunit expressed in mouse SMM, regardless of the host genotype (Fig. 5D). Gαi2 expression was confirmed at the protein level in all tested Pten;Trp53-null (n = 8) and Pik3ca*;Trp53-null (n = 9) MM by IHC, with a heterogeneous expression pattern (Fig. 5E). To confirm the role of GPCRs signaling in MAPK activation in Pten;Trp53-null cells, we targeted Gαi2 using shRNAs. Gαi2 knockdown efficiently and consistently reduced Erk and Akt phosphorylation in Pten;Trp53-null, but not in Pik3ca*;Trp53-null cells (Fig. 5F). These results provide novel evidence on the involvement of GPCRs in MAPK and PI3K activation in Pten;Trp53-null MM cells. To further assess the relevance of this pathway in human MMs, we interrogated expression in the series from Bueno and colleagues (6), finding that GNAI2 mRNA is significantly upregulated in sarcomatoid transcriptome- and histology-defined subgroups (Fig. 5G).

p110β has been shown to play a crucial role in PTEN-null tumors in vitro and in vivo (29, 30) and is mainly activated by GPCRs (21). We hypothesized that p110β and Mek1/2 cooperate driving the aggressive phenotype of Pten;Trp53-null tumors. TGX-221, a selective p110β inhibitor, reduced PI3K-dependent Akt phosphorylation (Thr308) to a greater extent in Pten;Trp53-null cells than in Pik3ca*;Trp53-null cells. Surprisingly, mTORC2-dependent phosphorylation of Akt (Ser473) was inhibited in all cell lines (Supplementary Fig. S4B). TGX-221 had no effect on pErk, suggesting that combined inhibition of p110β and MEK1/2 might be more active in Pten;Trp53 cells.

As single agents, TGX-221 and PD325901 had modest antiproliferative effect. However, their combination completely suppressed the proliferation of Pten;Trp53-null cells. Both inhibitors decreased in vitro invasiveness (Fig. 6A and B). Based on the in vitro efficacy of simultaneous MEK and p110β inhibition, we tested the combination in vivo using drugs undergoing clinical investigation. Selumetinib is a selective MEK1/2 inhibitor in phase III trials for NSCLC, thyroid cancer, and metastatic uveal melanoma, and AZD8186 is a p110β/p110δ inhibitor undergoing phase I investigation (https://clinicaltrials.gov). Pten;Trp53-null tumor growth was followed by weekly micro-CT; early tumor detection was facilitated by bone differentiation, a common feature of MM (Fig. 6C). Once tumors were detected (6–7 weeks after Cre inoculation), mice (n = 6/group) were randomized to receive daily oral treatment with vehicle, selumetinib, AZD8186, or selumetinib/AZD8186 for 7 weeks. Single compound treatment did not affect survival. In contrast, a significant lifespan increase (P = 0.031) was observed with the combination (Fig. 6D). Accordingly, CT images showed increased size and number of lesions in vehicle- and single compound-treated mice, and reduced tumor progression upon drug combination (Fig. 6E). The dependency of tumor shrinkage on treatment was shown upon its discontinuation: within the following 5 weeks, tumors grew in 5 mice, but 1 surviving mouse sacrificed 26 weeks later showed a complete pathologic response (Supplementary Fig. S5A and S5B). In an independent experiment, CT tumor–bearing Pten;Trp53-null mice were randomized to receive vehicle or selumetinib/AZD8186 (n = 11). Treatment was maintained for 12 weeks without toxicity. Median survival of treated Pten;Trp53-null mice was 13.4 weeks, compared with 9.3 weeks for control mice (P = 0.005; Fig. 6F). Of 2 surviving mice, one had a minimal peritoneal tumor and the other displayed only microscopic tumor foci at autopsy (Supplementary Fig. S5C–S5E).

We next analyzed the effects of this drug combination on 20 human primary pl-MM cultures [EMM (n = 8), BMM (n = 7), and SMM (n = 5; refs. 4, 31–33]. Unfortunately, no primary pe-MM were available for these studies. Response was defined as a >50% reduction of cell viability and combination GI₅₀ < 10 μmol/L (calculated as the sum of GI₅₀s of each drug). Six of 12 BMM/SMM and 1 of 8 EMM cultures responded to the combination (Fig. 7A and B; Supplementary Fig. S6A); in 6 of the responding cultures, selumetinib and AZD8186 were synergistic, displaying a CI <1 (Supplementary Table S5). BMM and SMM also displayed lower GI₅₀s to selumetinib/AZD8186 combination (Supplementary Fig. S6B and S6C). AUC responses to increasing drug concentrations revealed that BMM and SMM were more sensitive to AZD8186 (P = 0.015), and to the combination (P = 0.049), than EMM (Fig. 7B and C). To further define predictive markers of response to the selumetinib/AZD8186 combination, we assessed vimentin, Gαi2, TP53, and MAPK and PI3K pathways status in lysates from primary human MM cells (Fig. 7D) and found that low vimentin, Gαi2, and AKT levels identify the subgroup of resistant cells (Fig. 7E). Finally, to explore the potential clinical relevance of these findings, we compared the effects of selumetinib/AZD8186 combination to cisplatin (Supplementary Fig. S7): responder cells displayed higher sensitivity to selumetinib/AZD8186 than to cisplatin (P = 0.038; Fig. 7F). Altogether, these results point to selumetinib/AZD8186 as a novel therapy alternative to cisplatin for patients with pl- an pe- nonepithelioid MM, high expression of vimentin, AKT, and Gαi2 being candidate biomarkers predictive of response.

There is an urgent need to understand the molecular factors linked to sarcomatoid MM subtype and their aggressiveness to define therapeutic improvements. EMT may, in part, explain the poor prognosis of SMM, conferring both high invasiveness and chemoresistance (4, 6, 7). Recent reports highlight genetic alterations and molecular mechanisms associated to EMT in SMM, such as TP53 inactivation, LATS2 mutations, CDKN2A homozygous deletions, E2F target upregulation, lncRNA expression, and PI3K-mTOR and RAS/MAPK activation (4, 6, 7, 34). Here, we confirmed the relevance of concomitant Trp53 loss and PI3K activation as drivers of mouse SMM. A few genetic models of MM have previously been reported based on the inactivation of Nf2, Bap1, Ink4a/Arf, Trp53, Tsc1, and Pten—with or without the combined administration of asbestos. In most cases, the three major histologic subtypes of MM developed but only Nf2;Rb and Pten;Trp53 combined genetic loss leads to exclusively nonepithelioid (17, 19, 35). In our hands, adenovirus-Cre–mediated Pten;Trp53 inactivation resulted in SMM development with a median latency and histology similar to those reported by Sementino and colleagues, but we did not observe biphasic MM, as they did. These differences might reside in the genetic background used.

PI3K/AKT/mTORC1 and ERK pathway activation occurs in most human MM specimens, thought to be a consequence of RTK activation (9, 10, 18), although anti-RTK therapies have shown limited effectiveness (16, 36). The role of PTEN, and its implication in PI3K pathway activation in MM, is controversial as few genetic alterations have been identified in this gene. However, PTEN can be inactivated through multiple mechanisms (37). Thus, PTEN loss is reported in 16% to 62% of cases using IHC (11, 13–15), and a recent case report associates PTEN-hamartoma-tumor-syndrome with MM development pointing to PTEN loss as an MM driver (38). Most studies relating PTEN and MM did not discriminate between histologic MM subtypes. We show, using human MM TMAs, that PTEN protein loss is frequent in SMM and that this subtype displays lower PTEN expression, compared with epithelioid and biphasic MM. The analysis is limited by the low number of sarcomatoid samples in our cohort (n = 7), but PTEN mRNA levels were lower in transcriptomic defined sarcomatoid tumors when compared with the other MMs. PTEN expression in biphasic MMs is paradoxal, with higher histoscores in biphasic than in epithelioid tumors. Biphasic MM constitutes a highly heterogeneous group with variable proportions of epithelioid and sarcomatoid cells. These differences were highlighted in the transcriptomic biphasic-E and biphasic-S clusters defined by Bueno and colleagues (6). In addition, the relevance of intratumor heterogeneity regarding epithelioid and sarcomatoid components was recently underscored by Blum and colleagues (8), indicating that a more precise definition of the biphasic group is required. The reported low PTEN levels in sarcomatoid MMs, plus the reported association of TP53 mutations with nonepithelioid MM, underline the relevance of Trp53;Pten-null mice as an MM preclinical model (6, 8, 19).

Our data also revealed that Pten loss has stronger oncogenic properties than Pik3ca mutations, favoring metastases and an immunosuppressive environment enriched in Cd8 T cells, T regs, M2 macrophages, and PD-L1–expressing cells that have been associated with poor prognosis in human MM (39–41). Remarkably, nonepithelioid human MMs exhibit higher PD-L1 expression and T-cell infiltration (6–8, 41). These observations are in agreement with the fact that PTEN antagonizes further PI3K isoforms implicated in the activation of pathways other than mTOR, and that it exerts lipid phosphatase–independent functions (42). Biochemical analysis of Pik3ca*;Trp53-null and Trp53;Pten-null tumors revealed strong Akt and Mek/Erk activation without major evidence for a role of upstream RTKs. This allowed us to uncover a novel role of GPCR/Gαi2 signaling in Erk phosphorylation specifically in Trp53;Pten-null, but not in Pik3ca*;Trp53-null, cells. Importantly, Gαi2 mRNA levels are upregulated in human sarcomatoid MMs when compared with the other subgroups, and its signaling has been implicated in the proliferation and migration of ovarian and prostate cancer cells (43, 44).

Several observations led us to focus on p110β as a crucial mediator of the Trp53;Pten-null tumor phenotype. p110β has been shown to be essential for tumorigenesis in a PTEN-null context (29, 30). In addition, p110β can be activated by both RTKs and GPCRs, but genetic data suggest a minor impact of p110β deletion, or its lipid kinase inactivation, on RTK signaling (21). In vitro, only a fraction of PTEN-null cancer cells are sensitive to p110β inhibitors, and in vivo studies have not consistently shown tumor regression (45, 46). As single agent, the dual PI3K and mTOR inhibitor apitolisib (GDC0980) showed modest antitumor activity in a phase I trial that included patients with advanced MM (47), indicating that PI3K inhibitors need to be combined with other drugs to achieve greater antitumor effects. The MAPK pathway activation observed in Trp53;Pten-null tumors pointed to the combination of p110β and MEK inhibitors. Indeed, we observed that p110β (TGX221) and MEK (PD325901) inhibitors show synergy on the growth of Trp53;Pten-null tumor cells. More importantly, the combination of two drugs that are currently under clinical investigation and target MEK and p110β revealed strong antitumor effect in vivo in established Trp53;Pten-null tumors, including induction of complete pathologic responses. Several early-phase clinical studies using MEK and PI3K inhibitors are ongoing with significant toxicity (48), but we did not encounter toxicity in mice. Recent data point that the use of isoform-selective PI3K inhibitors would allow optimizing combined therapy and reducing toxicity (49). Accordingly, p110β/MEK inhibition has been proposed in preclinical settings for PTEN-deficient/BRAF-mutated melanoma (50).

One major reason for the inability to reproduce promising results of preclinical studies in clinical trials is the lack of adequate patient stratification. Here, we provide evidence pointing to the nonepithelioid histology and high vimentin, AKT1/2, and Gαi2 protein levels as predictive markers of response to selumetinib/AZD8186 therapy by testing this combination on 20 primary cultures derived from epithelioid, biphasic, and sarcomatoid human pl-MM. Responder cells presented significantly lower GI₅₀s to combined therapy than cisplatin. We propose that this combination be tested in patients with advanced MM with biphasic/sarcomatoid features, a subgroup for which therapeutic options—including surgery—are scarce. Should the acceptable toxicity observed in mice hold in patients, the combination of selumetinib/AZD8186 with cisplatin would also merit consideration.

In summary, we reveal a novel druggable molecular pathway driven by GPCRs/MEK/p110β signaling, which has a major impact on MM progression using human-relevant genetic mouse models and represents a novel targeted therapy for patients with MM whose tumors display sarcomatoid features.

S.T. Barry is Senior Principal Scientist at AstraZeneca. No potential conflicts of interest were disclosed by the other authors.

Conception and design: M. Marqués, F.X. Real

Development of methodology: M. Marqués, N. del Pozo

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): M. Marqués, R. Tranchant, B. Risa-Ebrí, M.L. Suárez-Solís, L.C. Fernández, N. del Pozo, C. Meiller, C. Pirker, B. Hegedus, A. Carnero, W. Berger, F.X. Real

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): M. Marqués, R. Tranchant, B. Risa-Ebrí, E. Carrillo-de-Santa-Pau, J. Martínez de Villarreal, Y. Allory, Y. Blum, C. Pirker, W. Berger, D. Jean, F.X. Real

Writing, review, and/or revision of the manuscript: M. Marqués, R. Tranchant, B. Risa-Ebrí, M.L. Suárez-Solís, L.C. Fernández, E. Carrillo-de-Santa-Pau, N. del Pozo, J. Martínez de Villarreal, C. Meiller, Y. Allory, Y. Blum, C. Pirker, B. Hegedus, S.T. Barry, A. Carnero, W. Berger, D. Jean, F.X. Real

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): F.X. Real

Study supervision: D. Jean, F.X. Real

Other (Provision of reagents): S.T. Barry

Other (Obtained funding): F.X. Real

We thank E. Andrada, N. Malats, M. Márquez, M. Rava, V.J. Sánchez-Arévalo Lobo, X. Langa, F. Díaz, Y. Cecilia, and F. Larousserie for valuable contributions; the Imaging, Genomics, Transgenics, and Histopathology CNIO Units for help with CT and histology interpretation; and M. Barbacid, R. García, X. Bustelo, A.C. Carrera, P. Smith, and A. Nebreda for valuable comments to earlier versions of the article. Astra-Zeneca provided drug (as required by Astra-Zeneca publication policy). We thank the National Mesothelioma Virtual Bank for providing TMAs.

This work was supported, in part, by grants from Asociación Española Contra el Cáncer (F.X. Real), Spanish Ministry of Economy and Competitivity, Plan Estatal de I+D+I 2013-2016, ISCIII (FIS PI15/00045 to A. Carnero), RTICC (Instituto de Salud Carlos III, grants RD12/0036/0034 to F.X. Real and A. Carnero, respectively), and CIBERONC (CB16/12/00453 and CD16/12/00275 to F.X. Real and A. Carnero, respectively), cofunded by FEDER from Regional Development European Funds (European Union) and Inserm (Institut national de la santé et de la recherche médicale). M. Marqués was supported by a Sara Borrell Fellowship from Instituto de Salud Carlos III. CNIO is supported by Ministerio de Ciencia, Innovación y Universidades as a Centro de Excelencia Severo Ochoa SEV-2015-0510.

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.