Phosphatidylinositol-3-kinase (PI3K)/Akt Signaling is Functionally Essential in Myxoid Liposarcoma

Myxoid liposarcoma (MLS) accounts for approximately 5%–10% of all soft-tissue sarcomas, representing about 20% of all malignant adipocytic tumors (1). MLS constitutes the most frequent liposarcoma subtype in patients below the age of 20 years. A high rate of local recurrence and particular propensity to develop distant metastases has been reported in approximately 40% of MLS patients (2). MLS exhibit a morphologic spectrum ranging from myxoid tumors with low cellularity to hypercellular, round-cell sarcomas associated with an aggressive clinical course (3). Genetically, 95% of MLS are characterized by a chromosomal t(12;16)(q13;p11) translocation, juxtaposing the FUS and DDIT3 genes. The remaining 5% display an alternative chromosomal t(12;22)(q13;q12) rearrangement leading to an EWSR1-DDIT3 gene fusion (4). The resulting FUS-DDIT3 and EWSR1-DDIT3 fusion proteins have been shown to play an essential role in MLS tumorigenesis, acting as pathogenic transcriptional (dys-)regulators (5–8). Current therapeutic approaches in high-grade MLS complement radical surgery and adjuvant radiation and/or conventional chemotherapy based on anthracyclines and ifosfamide, recently supplemented by agents such as trabectedin or eribulin (9–11). Although MLS displays a higher chemosensitivity than other liposarcoma subtypes, the substantial frequency of recurrence and metastasis in MLS underlines the urgent need for novel, biology-guided therapeutic strategies. In principle, counteracting the effects of the aberrant FUS–DDIT3 transcription factor represents the most promising strategy to selectively target MLS tumor cells; however, the therapeutic interference with (chimeric) transcription factors in vivo represents a particular challenge.

In line with the situation in other translocation-associated soft-tissue and bone sarcomas driven by specific gene fusions, such as, Ewing sarcoma, synovial sarcoma, or alveolar rhabdomyosarcoma (12–14), MLS are genomically “stable” tumors with few mutations beyond the pathognomic gene fusion. However, two recent studies described activating gain-of-function mutations in the PIK3CA gene encoding the catalytic PI3K subunit in a subset of MLS. Demicco and colleagues provided evidence of PTEN losses occurring as further structural alteration of crucial PI3K/Akt pathway components (12, 15). Beyond that, it has been reported that EGFR-, MET-, VEGFR-, RET-, and PDGFRB signaling sustained by autocrine/paracrine transduction and receptor tyrosine kinase (RTK) cross-talk may lead to an increase in PI3K-dependent signals (16, 17). The PI3K/Akt cascade is an elementary hub in the signal transduction of diverse RTK stimuli involving different growth-controlling regulators such as GSK-3β as well as the cell-cycle effectors Cyclin D1 and p21^WAF (18–21).

Overall, previously published results indicate a fundamental relevance of PI3K/Akt signals with respect to the specific liability of MLS tumor cells, accomplished either by genetic PIK3CA and/or PTEN alterations or complex signaling networks (8, 12, 15–17, 22) resulting in aberrant activation of PI3K activity. This preclinical study was conducted to explore the functional importance of PI3K/Akt signals in MLS tumorigenesis, and to establish a molecularly targeted strategy employing small-molecule PI3K inhibitors.

MLS tumor specimens were selected from the archive of the Gerhard-Domagk-Institute of Pathology (Münster University Hospital, Münster, Germany). All diagnoses were reviewed by three experienced pathologists (E. Wardelmann, S. Huss, W. Hartmann) based on morphologic criteria, DDIT3 break-apart FISH, or RT-PCR analysis in accordance with the current WHO Classification of Tumours of Soft Tissue and Bone (1). Evaluation of the round-cell component (3, 23) and tissue microarray (TMA) sampling/construction was performed as described previously (8). In total, 56 MLS tumor specimens were included (24/42.9% female, 32/57.1% male). Median patient age at diagnosis was 48 years (range 24–73 years of age). Median tumor size was 10 cm (range 1.5–29 cm). Clinicopathologic characteristics of the cohort are summarized in Table 1 and Supplementary Table S1. The study was approved by the Ethics Review Board of the University of Münster (2015-548-f-S), and experiments were conformed to the principles set out in the World Medical Association Declaration of Helsinki and the United States Department of Health and Human Services Belmont Report. Written informed consent from patients was not requested by the Ethics Review Board of the University of Münster (2015-548-f-S).

The following primary antibodies were applied: phospho-Akt (S473, monoclonal rabbit, D9E, 1:50, catalog no. 4060), Cyclin D1 (monoclonal rabbit, 92G2, 1:50, catalog no. 2978), phosphor-GSK-3α/β(S21/9, polyclonal rabbit, 1:50, catalog no. 9331), p21 (monoclonal rabbit, 12D1, 1:1000, catalog no. 2947; all purchased from Cell Signaling Technology). IHC staining was performed with a BenchMark ULTRA Autostainer (VENTANA/Roche) on 3 μm tumor sections as described previously (8). IHC results were evaluated by a semiquantitative approach (H-score; ranging from 0 to 300) determining the percentage of cells at each staining intensity level, using the following formula: [1 × (% cells 1+) + 2 × (% cells 2+) + 3 × (% cells 3+)]. Immunoreactivity was assessed defining the staining intensity (0, negative; 1+, weak; 2+, moderate; 3+, strong) in the positive control (breast cancer; no special type) as strong. TMA tissue cores with H-score >50 were considered positive. The discriminatory threshold (positive = semiquantitative H-score >50) was prespecified without prior analyses of the clinical course. Loss of PTEN protein was analyzed by IHC as described previously (8, 15). The IHC readers were blinded to the outcome data.

PIK3CA gene amplification was examined by FISH analyses using the SPEC PIK3CA/CEN 3 Dual Color Probe (ZytoVision). At least 40 cells of each MLS tumor specimen were assessed. PIK3CA gene amplification was defined by a PIK3CA/centromere 3 (CEN3) ratio ≥2.

A customized GeneRead DNAseq Panel (Mix-n-Match V2, Qiagen) was applied to assess the complete exonic region of 23 genes that are frequently mutated in various cancer types (summarized in Supplementary Table S2). Target enrichment was performed by means of the GeneRead DNAseq PCR V2 Kit (Qiagen). All end repair, A-addition, and DNA barcode ligation steps were processed applying the GeneRead DNA Library I Core Kit (Qiagen). Amplification of adapter-ligated DNA was performed using the GeneRead DNA I Amp Kit (Qiagen). NGS was conducted with the Illumina MiSeq system. The Quantitative Multiplex FFPE Reference Standard (Horizon Discovery, catalog no. HD200) was applied as isogenic quality control (11 somatic mutations verified at 0.8%–24.5% allelic frequency in genomic DNA) for routine performance monitoring and evaluation of NGS workflow integrity (preanalytical DNA extraction, NGS workflow, and postanalytical bioinformatics). NGS data were analyzed by means of the CLC Biomedical Genomics Workbench Software (CLC bio, Qiagen). Validation by Sanger sequencing was performed using the BigDye Terminator v3.1 Cycle Sequencing Kit (Life Technologies).

The functional context of NGS-detected variants in the coding regions was predicted using following in silico tools: PolyPhen-2 (v2.2.2r398; ref. 24), Protein Variation Effect Analyzer (PROVEAN; v1.1.3; ref. 25), Sorting Intolerant From Tolerant (SIFT; Ensembl 66; ref. 26), Mutation Assessor (release 3; ref. 27), and Combined Annotation Dependent Depletion (CADD; v.1.3; ref. 28). In brief, the PolyPhen-2 method utilizes evolutionary and physical comparative considerations to predict amino acid substitutions on protein structure and function. Scoring ranges from 0 (neutral) to 1 (deleterious) and functional significance is categorized into benign, possibly damaging, and probably damaging. The web-based PROVEAN algorithm classified each single-nucleotide variant (SNV) as either neutral or deleterious (cutoff −2.5). SIFT was used with the default settings classifying variants from 0 (damaging) to 1 (tolerated). CADD is a trained method to differentiate 14.7 million high-frequency human-derived alleles for objective integration of diverse annotations into a single measure (C score). Scoring correlates with allelic diversity, annotations of pathogenicity, disease severity, experimentally measured regulatory effects, and complex trait associations. Calculated C scores rank a pathogenic variant relative to all possible substitutions of the human genome. To increase the prediction accuracy and the level of confidence, a combination of in silico tools based on evolutionary information, protein structure, and functional parameters was used. We combined the individual output from ≥3 of 5 in silico prediction tools and defined a single-consensus outcome summarized in Supplementary Table S3.

The human MLS cell lines MLS402-91 (FUS–DDIT3 exon 7-2; type 1), MLS2645-94 (FUS–DDIT3 exon 5-2; type 2), and MLS1765-92 (FUS-DDIT3 exon 13-2; type 8) were described before (29). For the purpose of cell line authentication, presence of the MLS specific FUS–DDIT3 fusion gene transcripts was confirmed by RT-PCR and Sanger sequencing. All monolayer cell cultures were maintained in Roswell Park Memorial Institute Medium 1640 (RPMI), supplemented with 10% FBS (Life Technologies), and incubated under standard conditions (humidified atmosphere, 5% CO₂, 37°C). Cells were passaged for a maximum of 20–30 culturing cycles and Mycoplasma testing was performed quarterly by standardized PCR. To study the biological effects of treatment with the class I PI3K inhibitor BKM120 (buparlisib) (30, 31) and the prototypic pan-PI3K inhibitor LY294002 (32, 33), MLS cells were cultured in medium supplemented with 2% FBS. For experimental PI3K modulation, MLS402-91 cells were transfected with pCMV2-Tag 2A plasmids encoding either the mutant ΔH1047R PIK3CA (Addgene, catalog no. 16639) or wild-type PIK3CA (Addgene, catalog no. 16643) coding sequence (34).

Following primary antibodies were used: Akt (monoclonal rabbit, C67E7, 1:1,000, catalog no. 4691), phospho-Akt (S473, monoclonal rabbit, D9E, 1:1,000, catalog no. 4060), Cyclin D1 (monoclonal rabbit, 92G2, 1:1,000, catalog no. 2978), GSK-3β (monoclonal rabbit, 27C10, 1:1,000, catalog no. 9315), phospho-GSK-3β (S9, polyclonal rabbit, 1:1,000, catalog no. 9336), PI3K p110α (rabbit, 1:1,000, catalog no. 4255), p21 (monoclonal rabbit, 12D1, 1:1,000, catalog no. 2947; all purchased from Cell Signaling Technology), and β-actin (loading control; monoclonal mouse, AC15, 1:10.000, catalog no. A5441 Sigma-Aldrich). Secondary antibody labeling (Bio-Rad Laboratories) and immunoblot development was performed using an Enhanced Chemiluminescence Detection Kit (SignalFire ECL Reagent; Cell Signaling Technology) and the Molecular Imager ChemiDoc System (Image Lab Software; Bio-Rad Laboratories).

MTT assays (Roche) were performed as described previously (35–37). In brief, 2.5 × 10³ MLS402-91, MLS2645-94, and MLS1765-92 cells were seeded in 96-well cell culture plates (100 μL of medium supplemented with 2% FBS) and exposed to increasing concentrations of the class I PI3K inhibitor BKM120 (buparlisib) (1.25-5 μmol/L) and the prototypic pan-PI3K inhibitor LY294002 (3.125-50 μmol/L) for 72 hours. DMSO was included as vehicle control.

For in vivo confirmation, we used the chick chorioallantoic membrane (CAM) model as previously reported and validated for anticancer agents (8, 38, 39). Because of the presence of vascular supply and the absence of an immune response from the graft, the CAM enables the transplantation of human cancer cells and the subsequent development of solid tumor xenografts in a three-dimensional in vivo microenvironment. The CAM model matches the 3R recommendations to reduce mammalian animal experiments and is regarded as reproducible, reliable, and effective (40).

Seven days after fertilization, MLS402-91 and MLS1765-92 cells (1–1.5 × 10⁶ cells/egg; dissolved in medium/Matrigel 1:1, v/v) were xenografted onto the CAM and incubated with 60% relative humidity at 37°C. Topical treatment with the class I PI3K inhibitor BKM120 (buparlisib) (1–2 μmol/L), the prototypic pan-PI3K inhibitor LY294002 (10 μmol/L), or vehicle control (0.2% DMSO in NaCl 0.9%) was initiated on day 8 and recapitulated for two consecutive days. Three days after treatment initiation, tumor-bearing CAM xenografts were imaged, explanted, and fixed (5% PFA). Tumor volume (mm³) was calculated according to the formula: TV = length (mm) × width² (mm) × π/6. The in vivo studies were approved by an Institutional Animal Care and Use Committee (84-02-04-2016-A195) and conducted in accordance with recognized standards of the National and European Union guidelines.

The class I PI3K inhibitor BKM120 (buparlisib/NVP-BKM120; C₁₈H₂₁F₃N₆O₂; CAS#: 944396-07-0; refs. 30, 31) was purchased from Selleck Chemicals and dissolved in DMSO (Sigma-Aldrich). The prototypic pan-PI3K inhibitor LY294002 (C₁₉H₁₇NO₃; CAS#: 154447-36-6; refs. 32, 33) was purchased from Cell Signaling Technology and dissolved in DMSO. Final DMSO concentration did not exceed 0.2% (v/v) for all in vitro and in vivo experiments.

To determine the involvement of PI3K/Akt signaling in MLS tumorigenesis, expression of phospho-Akt (S473), phospho-GSK-3α/β (S21/9), p21, and Cyclin D1 was analyzed in a set of 56 MLS tumor specimens by IHC (Fig. 1A and B; Supplementary Table S1). On the basis of the calculated H-score, 62.0% of MLS tumor specimens were positive for phosphorylated Akt (S473) and 38.5% for phosphorylated GSK-3α/β (S21/9). No significant Akt and/or GSK-3β phosphorylation levels were detected in 23.2% of cases. 32.7% of MLS specimens were positive for p21, while expression of Cyclin D1 was shown in 48.1% (Fig. 1A and B). Loss of PTEN protein expression was demonstrated in 5 of 56 (8.9%) MLS tumor specimens (Supplementary Table S1). Phosphorylation and expression levels of PI3K/Akt signaling components did not correlate with tumor size, FUS-DDIT3 translocation subtype, patients' age, and/or gender. In accordance with the IHC results in MLS tumor specimens, protein expression (PI3K p110α, p21, and Cyclin D1) and phosphorylation levels (phospho-Akt S473 and phospho-GSK-3β S9) were demonstrated in total protein extracts of MLS402-91 (FUS–DDIT3 type 1), MLS2645-94 (FUS–DDIT3 type 2), and MLS1765-92 (FUS-DDIT3 type 8) cells (Fig. 1C, top). Expression of the specific FUS–DDIT3 fusion transcript was confirmed in all three MLS cell lines by RT-PCR (Fig. 1C, bottom). These findings provide evidence that activation of PI3K/Akt signaling is a common feature of MLS.

As pathogenic PIK3CA mutations can be responsible for constitutive activation of the PI3K/Akt signaling cascade, we examined the entire PIK3CA coding region by targeted NGS followed by Sanger sequencing for validation purposes. In 12 of 56 (21.4%) MLS cases, genetic alterations in PIK3CA (most frequently H1047R) could be detected (Fig. 2A; Supplementary Table S1). Overall, 55.6% (5/9) of the observed variants in the PIK3CA coding region (N345K, C378Y, C901F, H1047R, and N1068fs) were predicted to have a deleterious impact by ≥3 independent in silico tools (summarized in Fig. 2B and Supplementary Table S3). No PIK3CA gene mutations were identified in all three MLS cell lines. PIK3CA gene amplification was examined by FISH and could be excluded for all analyzed MLS cases. In addition, the coding sequence of AKT1 and PTEN was analyzed. Few deleterious variants were identified in both genes: AKT1 (E17K and W80R) and PTEN (V119F and I253fs). Notably, round-cell MLS (3/6; 50.0%) were more likely than myxoid MLS tumors (5/13; 38.5%) to harbor a deleterious mutation in the PIK3CA gene. In contrast, all deleterious mutations in AKT1 (E17K and W80R) and PTEN (V119F and I253fs) were exclusively identified in myxoid tumors. In Kaplan–Meier correlations, MLS patients with round-cell morphology (Fig. 2C, left; **, P = 0.0014) and activating alterations in PI3K/Akt signaling components (Fig. 2C, right; **, P = 0.0071) displayed a significantly reduced overall/event–free survival. In addition, all coding exons of 20 further cancer-associated genes were analyzed with another 11 actionable nonsynonymous mutations detected in 10 MLS cases (summarized in Supplementary Tables S2 and S3): CTNNB1 (P321H), ERBB2 (R138W), FGFR3 (T311M), GNAQ (M59L), GNAS (P533R), and TP53 (V272G; allelic frequency: 10.5%–90.3%). In summary, 28 nonsynonymous somatic variants were identified in 25 of 56 (44.6%) MLS tumor specimens. The overall concordance between all 5 in silico prediction tools was 52.2% and independent from the deleterious/damaging or neutral/tolerated status. At least 4 of 5 in silico predictions were in agreement in 34.8% of variants and in only 13.0% of variants the prediction was based on 3 of 5 in silico tools (summarized in Supplementary Table S3). Correlating the (global) presence of gene mutations with MLS tumor morphology, no significant trend concerning the mutational burden could be determined comparing the round cell to the myxoid MLS subtype, which has been previously established as a prognostically relevant histologic parameter (1, 3, 23). In total, gene mutations were detected in 9 of 21 (42.8%) round-cell MLS versus 16 of 35 (45.7%) myxoid tumors. Overall, 85.7% of MLS patients with alterations in PIK3CA, AKT1, or PTEN were positive for phosphorylated Akt (S473) and/or GSK-3α/β (S21/9) as examined by IHC and thus may be regarded as indicators of an activation of PI3K/Akt signaling. In the subset of patients with PIK3CA/AKT1/PTEN wild-type MLS, the fraction of tumors immunohistochemically displaying either phosphorylation of Akt and/or GSK-3β was substantially lower (69.4%).

To confirm the biological importance and to analyze the functional role of PI3K/Akt signaling by a nonpharmacologic approach, three MLS cell lines (MLS402-91, MLS2645-94, and MLS1765-92) were transfected with two different prevalidated siRNAs (to exclude unspecific off-target effects) directed against human PIK3CA. Consistently, specific knockdown of PIK3CA led to reduced phosphorylation levels of Akt (S473) and GSK-3β (S9), combined with diminished Cyclin D1 downstream target expression (Fig. 3A). In MTT assays, all MLS cell lines displayed a significant reduction of cell viability (***, P < 0.001) in comparison with nontargeting control siRNA (Fig. 3B). These results imply that PI3K/Akt–mediated signals may be of major functional relevance in the pathogenic context of MLS.

Our results suggested that aberrant PI3K/Akt signaling might be a therapeutic target in MLS. We therefore evaluated the biological effects of pharmacologic PI3K inhibition. In MTT assays, three MLS cell lines were exposed to increasing concentrations (1.25–5 μmol/L) of the class I PI3K inhibitor BKM120 (buparlisib). Cell viability and growth were significantly suppressed in all three MLS cell lines with IC₅₀ values ranging from 1.01 to 1.84 μmol/L (BKM120, buparlisib), indicating a dose-dependent mode of action (Fig. 3C; Table 2). Analogue experiments with the prototypic pan-PI3K inhibitor LY294002 (3.125–50 μmol/L) showed comparable effects with IC₅₀ values ranging from 9.57 to 12.02 μmol/L (Supplementary Fig. S1A)

To examine the functional basis of PI3K/Akt suppression, we analyzed the specific effects of treatment with the class I PI3K inhibitor BKM120 (buparlisib) on the activation status of PI3K/Akt signaling components by immunoblotting. The growth-inhibitory effects were associated with a significant dose-dependent reduction of phosphorylation levels for Akt (S473) and GSK-3β (S9) in all three MLS cell lines (Fig. 3D). Analogue effects were observed with the prototypic pan-PI3K inhibitor LY294002 (Supplementary Fig. S1B).

Performing flow cytometric analyses, poly-adenosine diphosphate (ADP)-ribose polymerase (cPARP; Asp214) cleavage was employed as a marker of apoptosis. After treatment with increasing concentrations of BKM120 (buparlisib) (1–2 μmol/L; 72 hours), all three MLS cell lines showed a significantly and dose-dependently increased rate of apoptosis (Fig. 3E and F).

Antitumor efficacy of the class I PI3K inhibitor BKM120 (buparlisib) on tumor growth was tested in a previously established in vivo model of human MLS (8). We xenografted MLS402-91 and MLS1765-92 cells onto chick CAMs to initiate MLS tumor formation (n ≥ 5 eggs/group). A significant reduction of MLS tumor volume through topical administration of BKM120 (buparlisib) (1–2 μmol/L) compared with the DMSO vehicle-treated control group (Fig. 3G) was demonstrated for two different MLS cell lines. Analogue experiments with the prototypic pan-PI3K inhibitor LY294002 (10 μmol/L) resulted in comparable effects (Supplementary Fig. S1C). Collectively, these results showed that inhibition of the PI3K/Akt signaling cascade impairs the maintenance of MLS tumors in vivo, further supporting the concept that PI3K/Akt pathway activation represents a novel, molecularly based target for therapeutic interventions in patients with advanced high-grade MLS.

To assess the biological effects of ΔH1047R-mutated PIK3CA in response to BKM120 (buparlisib) treatment, MLS402-91 cells transfected with pCMV2-Tag 2A plasmid DNA encoding different PI3K p110 catalytic subunit alpha variants were exposed to BKM120 (buparlisib) (2 μmol/L; DMSO was employed as vehicle control). Significantly increased rates of apoptosis were detected applying two different markers of apoptosis: (i) caspase 3/7 activity (Fig. 4A) and (ii) PARP cleavage (cPARP; Asp214; Fig. 4B). Consistently, MLS402-91 cells expressing ΔH1047R-mutated PIK3CA showed increased rates of apoptosis upon treatment with BKM120 (buparlisib) compared with cells expressing wild-type PIK3CA.

MLS are malignant tumors of lipogenic differentiation with a high rate of local recurrence and particular propensity to develop distant metastases. High histologic grade, defined as round-cell component >5% is the primary predictor of an unfavorable clinical outcome (3, 23). Although there is an established role for conventional cytotoxic and radiation-based therapies in MLS (10), molecularly targeted therapies are not implemented so far. Almost all MLS are characterized by a FUS-DDIT3 gene fusion, encoding an oncogenic transcription factor with the potential to transform mesenchymal stem cells to form MLS in mice (5). As in other sarcomas driven by specific gene fusions, such as, Ewing sarcoma, synovial sarcoma, or alveolar rhabdomyosarcoma (12–14), MLS are genomically “stable” tumors with a low mutational burden. However, two recent studies described recurrent PIK3CA gene mutations in a subset of 14%–18% MLS cases beyond the pathognomonic FUS–DDIT3 gene fusion (12, 15). In addition, PTEN loss and IGF-IR overexpression have been described, which both represent alternative mechanisms of PI3K/Akt signal transduction activation (8, 15, 41). Beyond that, it has been reported that EGFR-, MET-, VEGFR-, RET-, and PDGFRB signaling sustained by autocrine/paracrine transduction and RTK cross-talk involving the extensive tumor vasculature of MLS may lead to an additional increase in PI3K/Akt signaling pathway activity (16, 17). Both, IGF-IR overexpression and PIK3CA mutations appear to be overrepresented in the round-cell variant of MLS (15, 41). Given the fact that (i) round cell phenotype, (ii) presence of an activating PIK3CA mutation, and (iii) IGF-IR overexpression are associated with an unfavorable clinical course, overactivation of PI3K/Akt signals may represent a crucial factor for a more aggressive biology of MLS (3, 12, 15, 41). Unfortunately, the limited size of our MLS tumor set and the retrospective nature of our analysis prevented a definitive evaluation of the contribution of each of these factors to MLS prognosis; however, this interesting issue might be worth to be systematically analyzed in future prospective trials.

We previously provided substantial evidence that therapeutic interference with a cell-autonomous stimulation loop involving IGF-IR-dependent signals may represent a highly effective approach associated with therapeutic benefit, which is specifically found in the tumor's biology involving FUS-DDIT3-mediated overexpression of IGF-II (8). However, this therapeutic strategy applies only to those tumors in which PI3K/Akt signaling is not activated through mutations of central components as PIK3CA, AKT1, and/or PTEN. We therefore hypothesized that central pathway interference through PI3K inhibition might constitute a rational therapeutic alternative since it covers different modes of pathogenic PI3K/Akt activation.

Our genomic study, identifying activating gain-of-function PIK3CA mutations in 14.2% of MLS concordantly confirms published results (12, 15), while genomic amplifications of PIK3CA could be excluded. In line with data presented by Barretina and colleagues (12), we identified PIK3CA mutations to be associated with a negative prognostic impact. Notably, round-cell MLS (50.0%) were more likely than myxoid MLS tumors (38.5%) to harbor a deleterious mutation in the PIK3CA gene. In addition, we detected deleterious AKT1 (E17K and W80R) and PTEN (V119F and I253fs) alterations, which were bioinformatically predicted to be of functional relevance; these were exclusively detected in myxoid tumors. In agreement with Demicco and colleagues (15), we found 8.9% of the analyzed cases to be negative for PTEN, pointing either to a genetic loss or epigenetic silencing through promotor hypermethylation (8, 42, 43). All deleterious mutations found in PIK3CA, AKT1, and PTEN were mutually exclusive, resulting in a total fraction of 26.8% of MLS tumors with evidence of an activated PI3K/Akt signaling pathway caused by genetic/epigenetic alterations. The few additional mutations in classic oncogenic drivers detected in our genomic screen were randomly spread among all MLS cases and appear to be secondary passenger alterations rather than crucial oncogenic drivers in MLS tumorigenesis. Given further kinase signaling input in MLS, for example, via the IGF-IR system as well as the effector loop discussed by Negri and colleagues (8, 17), it is not surprising that the fraction of tumors displaying either phosphorylation of Akt and/or GSK-3β (both regarded as indicators of an activation of PI3K/Akt signaling) in IHC analysis was substantially higher than 26.8%. With regard to the prognostic impact of PIK3CA mutations as shown in our study and the previous works published by Barretina and Demicco (12, 15) it appears improbable that these alterations represent mere passenger mutations; however, it remains to be elucidated in which way aberrant PI3K signaling contributes to MLS pathogenesis, which essentially depends on the FUS–DDIT3 gene fusion. Given our findings, it appeared rational to extend the study to preclinically evaluate the therapeutic potential of a PI3K/Akt–directed approach, which up to now has not been assessed in the context of MLS.

Considering the therapeutic accessibility of PI3K/Akt signaling, PI3K enzymes themselves represent the central hub of the signaling cascade as they serve as integrators and (enhancing) distributors of diverse signaling cues, and the development of inhibitory molecules is certainly most advanced for PI3K as therapeutic target. Combining pharmacologic studies and RNA interference (RNAi)-based approaches, our study convincingly shows in vitro that proliferation and viability of MLS cells significantly depend on PI3K-mediated signals. Importantly, we were able to show that expression of the ΔH1047R-mutated PIK3CA variant in MLS402-91 was associated with an increased sensitivity to BKM120 (buparlisib). Consistent with our in vitro findings, administration of BKM120 (buparlisib) and LY294002 to xenografted MLS402-91 or MLS1765-92 cells led to a significant suppression of MLS tumor growth in vivo. Although these findings are consistent and imply a potential therapeutic role of PI3K inhibitory approaches in MLS, we are aware of the limitations of the employed avian CAM model as an in vivo tool in the preclinical transfer. However, apart from a published patient-derived xenograft (PDX) model which was not available to us (44), establishment and propagation of cell line–based MLS xenotransplants have not been successful to our best knowledge, which led to our choice of the (artificial) CAM system. It would, of course, be rational to try to transfer our findings to a mammalian in vivo system and to elaborate on combinations of diverse therapeutic agents including PI3K inhibitors. The results presented by Qi and colleagues showing efficacy of the PI3K inhibitor PF-04691502 in a PIK3CA-mutated MLS PDX model (44) support the relevance of our findings.

Given its particular dependence on kinase signals, MLS somehow resembles a genetically different mesenchymal neoplasia as gastrointestinal stromal tumors or dermatofibrosarcoma protuberans. These tumors are either driven by activating mutations in the receptor tyrosine kinases KIT/PDGFRA or by translocation-dependent overexpression of the PDGFR ligand PDGFB (45, 46). However, PI3K-directed therapeutic approaches were not established in these tumors as effective inhibition of the pathogenic signaling pathway could be achieved by interference with the activated KIT and PDGFR RTKs employing the tyrosine kinase inhibitor imatinib, resulting in significant clinical benefit (46–48). In the context of MLS, the situation is more complex, as the primary tumor-driving oncogene is the chimeric FUS-DDIT3 transcription factor, and only a subset of about 30% MLS cases is characterized by additional activating alterations in central components of the PI3K/Akt signaling pathway. However, given our previously reported finding of a FUS–DDIT3–driven cell-autonomous stimulation of MLS cells involving an IGF-IR/-(PI3K/Akt) transactivation loop, a much larger fraction of MLS appear to in fact represent a PI3K/Akt–dependent neoplasia. Given the diversity of activation modes of PI3K/Akt signal transduction during MLS tumorigenesis and regarding the fact that IGF-IR–driven tumors involve signaling cascades in parallel to PI3K/Akt with higher probability, the identification of appropriate predictive biomarkers once more becomes the crucial issue in the establishment of molecularly targeted therapeutic approaches. In MLS, PIK3CA, AKT1, and PTEN mutational testing as well as PTEN IHC/FISH may help to identify those patients that most probably may clinically benefit from a PI3K-directed therapeutic approach. In patients without alterations in these components a (combined) IGF-IR–directed therapeutic approach might be more beneficial as documented IGF-IR-driven signals also involve PI3K/Akt–independent intracellular signaling pathways as the MEK/ERK cascade (8, 41). Despite promising preclinical data, clinical trials testing diverse PI3K inhibitory approaches in different solid tumors showed a limited efficacy of monotherapy inhibition (49, 50). This may be due to insufficient target inhibition, which may be increased by optimized small-molecule structures, but also due to the fact that in many solid tumors, activating mutations in components of the PI3K/Akt signaling cascade occur as one genetic alteration among many others, entering the discussion on driver and passenger mutations. In MLS, the situation differs from the vast majority of cancers in so far as activating mutations in components of the PI3K/Akt signaling cascade appear to represent the outstanding alteration apart from the FUS-DDIT3 gene fusion against the background of genetic stability. This characteristic feature of MLS underlines the potential of PI3K inhibitory approaches and should promote systematic biomarker-guided clinical trials.

In conclusion, the results of this study imply that (apart from the pathognomonic FUS-DDIT3 gene fusion) activation of PI3K/Akt signaling is an essential pattern in MLS tumorigenesis which is realized, at least in part by genetic alterations in PIK3CA, AKT1, or PTEN. Interference with PI3K-mediated signals via small-molecule compounds is an effective therapeutic strategy for advanced high-grade MLS that could be exploited for clinical benefit. The current preclinical testing of a PI3K-targeted therapeutic approach indicates potent effects both in vitro and in vivo, qualifying the PI3K/Akt signaling pathway as a molecularly based target in MLS cancer therapy.

K. Steinestel has received speakers bureau honoraria from Boehringer Ingelheim and is a consultant/advisory board member for Novartis. E. Wardelmann has provided expert testimony for Novartis Oncology, Milestone, Menarini, PharmaMar, Roche, Nanobiotix, Bayer, and Lilly. No potential conflicts of interest were disclosed by the other authors.

Conception and design: M. Trautmann, S. Huss, W. Hartmann

Development of methodology: M. Trautmann, M. Cyra, I. Isfort, B. Jeiler, A. Krüger, B. Altvater, W. Hartmann

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): I. Grünewald, B. Altvater, C. Rossig, S. Hafner, E. Wardelmann, S. Huss, W. Hartmann

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): M. Trautmann, M. Cyra, I. Isfort, B. Jeiler, A. Krüger, K. Steinestel, B. Altvater, S. Hafner, J. Becker, S. Huss, W. Hartmann

Writing, review, and/or revision of the manuscript: M. Trautmann, M. Cyra, I. Isfort, I. Grünewald, K. Steinestel, B. Altvater, C. Rossig, S. Hafner, J. Becker, P. Åman, E. Wardelmann, S. Huss, W. Hartmann

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): K. Steinestel, P. Åman, E. Wardelmann, W. Hartmann

Study supervision: M. Trautmann, W. Hartmann

Others (designed and supervised CAM experiments): T. Simmet

The authors thank Charlotte Sohlbach, Inka Buchroth and Christian Bertling for excellent technical support. The study was supported by grants from the Deutsche Forschungsgemeinschaft (DFG, HA4441/2-1, to W. Hartmann and M. Trautmann;), the Wilhelm-Sander-Stiftung (2016.099.1, to W. Hartmann, M. Trautmann, and K. Steinestel;), and the “Innovative Medical Research” funding program of the University of Münster Medical School (IMF; TR121716 and I-TR221611, to M. Trautmann; I-HU121421, to M. Trautmann and S. Huss).

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.