Functional Mapping of Receptor Tyrosine Kinases in Myxoid Liposarcoma

Myxoid liposarcomas (MLS) account for more than one third of all liposarcomas, which represent ∼10% of all adult sarcomas. Histologically, MLSs consist of primitive nonlipogenic mesenchymal cells in myxoid glycosaminoglycan-rich stroma and have characteristic branching vessels. There are two variants: The pure or usual variant is a stroma-rich paucicellular tumor containing numerous capillaries arranged in a plexiform pattern, whereas the round cell or cellular variant is highly cellular with little or no intervening stroma and has a vascular pattern that is difficult to visualize. Diagnosis of the round cell variant requires >5% hypercellular areas (1). Both variants harbor a specific translocation leading to the fusion of FUS/CHOP t(12;16) (q13;p11) and more rarely to EWS/CHOP t(12;22) (q13;q12) genes (1).

The main treatment for localized MLS is surgery with or without radiotherapy. In advanced disease, standard anthracycline/ifosfamide-based chemotherapy offers some benefit, and more recently, it has been found that trabectedin (Yondelis, ET-743, PharmaMar) leads to prolonged tumor control (2, 3). However, resistance to both treatments may develop.

Published studies of receptor tyrosine kinases (RTK) in surgical samples have shown that RET transcripts are selectively expressed in MLS rather than in normal adipose tissue (4), and that RET [together with insulin-like growth factor-I receptor (IGF-IR) and IGF-II] is more expressed in MLS than in other soft tissue tumors (5). High levels of platelet-derived growth factor receptor B (PDGFRB) and epidermal growth factor receptor (EGFR) protein and mRNA expression have also been reported in MLS (6). Interestingly, at preclinical level, the upregulation of glial cell–derived neurotrophic factor (GDNF) family receptor α1 (GFRα1; RET coreceptor), PDGFA, MET, and its ligand hepatocyte growth factor (HGF) has been described in mesenchymal progenitor cells expressing the FUS/CHOP fusion gene and growing in mice as human-like MLSs (7).

The RET activation induced by dimerization depends on binding to a complex consisting of GDNF family ligands (GFL) and GFRα. The former include GDNF, neurturin, artemin, and persephin, and the latter include specific GFRα proteins that have unique binding affinities for each GFL: GFRα1 for GDNF, GFRα2 for neurturin, GFRα3 for artemin, and GFRα4 for persephin (8). The presence of glycosaminoglycans (of which the stroma of usual type MLS is rich) favoring the concentration of GDNF in the vicinity of GFRα1 and/or RET/GFRα1 complexes may contribute to RET activation (8). RET activating mutations and amplifications have also been rarely described (9, 10).

The dimerization-induced activation of RET leads to intracellular signaling and MET overexpression (11, 12), and it has been shown that GDNF activates MET indirectly via Src family kinases by binding GFRα1 or RET/GFRα1 (13), which evokes extracellular signal-regulated kinase 1/2 (ERK1/2) phosphorylation. Finally, it has been shown that GDNF expression [in cooperation with the expression of vascular endothelial growth factor A (VEGFA)] leads to the activation of both RET and VEGFR2 through physical interactions between VEGFR2 and RET that stimulate vascular morphogenesis in ureteric bud cell line cultures (14).

As far as MET is concerned, gene amplification (15, 16) and activating mutations (17, 18) have been described, but the most frequent cause of constitutive MET activation in human tumors is increased protein expression due to transcriptional upregulation and autocrine or paracrine ligand-dependent regulatory loops. It has also been shown that, in addition to RET (11, 12), various activated oncogenes such as RAS and ETS induce MET overexpression (16), and it is interesting to note that, even in the absence of genetic alterations, MET can act as an “oncogene expedient” by potentiating the effect of other oncogenes and fostering malignant progression, plays a role in tumor angiogenesis, and shows context-dependent oncogene addiction (19).

In the light of these findings, we used biochemical, molecular, and molecular-cytogenetic techniques, together with immunohistochemistry and confocal microscopy, to analyze RTKs and their downstream signaling in a series of molecularly profiled naive and treated MLS (but not by RTK inhibitors) with the aim of acquiring a comprehensive picture of RTK profiles that might also be useful in the development of targeted treatments.

The case material consisted of samples taken from 14 surgically resected MLS: 7 naive and 7 treated with trabectedin or conventional chemotherapy/radiotherapy. Six were histologically diagnosed as usual MLS and eight as round cell MLS (cell component, 5% to >90%). The diagnoses were confirmed by means of fluorescence in situ hybridization (FISH) analysis of fixed material, which revealed the presence of t(12:16), and by reverse transcription-PCR (RT-PCR) searching for the transcript type in cryopreserved tissue checked by H&E-stained frozen section (2, 3). Table 1A summarizes the patient data about gender, age, pTNM, tumor site, variant, transcript type, and treatment/response. The follow-up of the treated patients showed that one died of the disease, three showed no evidence of disease, and three were alive with disease.

Written informed consent was obtained from all of the patients.

The proteins were extracted from tissue samples stored at −80°C (20).

The Proteome Profiler Array kit (R&D Systems) was used as previously described (21).

For the immunoprecipitation analyses, equal amounts (1 mg) of protein lysates were precipitated by incubation with protein A–Sepharose (Amersham Biosciences) and specific antibodies (anti-PDGFRB, anti-EGFR, anti-RET, and anti-MET) as previously described (20, 21).

Western blotting was carried out using anti-phosphotyrosine antibody (Upstate) to reveal the presence or absence of activated/phosphorylated PDGFRB, EGFR, and MET; anti–phospho-RET (Santa Cruz Biotechnology) was used for activated RET. The filters were stripped and then incubated with the specific antibody to establish the degree of receptor expression.

To detect the coimmunoprecipitated proteins, the filter of immunoprecipitated anti-PDGFRB was stripped again and incubated first with anti-EGFR and then with anti-MET, the filter of immunoprecipitated anti-RET was incubated with anti-MET or anti-VEGFR2 (Santa Cruz Biotechnology), the filter of immunoprecipitated anti-MET was incubated with anti-EGFR, and the filter of immunoprecipitated anti-EGFR was incubated with anti-MET.

The expression of GFRα3 was detected using anti-GFRα3 antibody (clone AF670; R&D Systems).

The expression and activation of downstream targets were detected by means of Western blotting as previously described (21), and the results were normalized by incubating the filters with anti-actin (Sigma).

The antibodies and experimental conditions used to detect PDGFRB, EGFR, RET, MET, VEGFR2, and GFRα3 are shown in Table 1B (part 1).

Total RNA was extracted from fresh-frozen tissue and reverse transcribed as previously described (20).

The presence or absence of mRNA of the RET ligands (GDNF, artemin, neurturin, and persephin) and MET ligand (HGF) was verified by means of RT-PCR using the conditions described in Table 1B (part 2), and the PCR products were sequenced. The presence or absence of PDGFB, EGF, TGFα, VEGFA, and artemin mRNA was verified by means of real-time PCR using specific probes and the conditions described in Table 1B (part 2).

Mutation analysis was carried out on exons 9 to 11 and 14 to 17 of RET, and exons 2, 14 to 16, and 17 to 20 of MET, using the primers and PCR conditions shown in Table 1B (part 2), and on exons 2 to 9 of PTEN using primers and conditions that have been previously published (22) or are shown in Table 1B (part 2). The PCR products underwent automated DNA sequencing (ABI Prism 377; Applied Biosystems).

DNA from frozen material was digested by proteinase K and then extracted using the QIAamp DNA Mini Kit 250 (Qiagen) in accordance with the manufacturer's instructions. Mutation analysis was carried out on exons 10, 12, 14, and 18 of PDGFRB (20), exons 18 to 21 of EGFR (21), exons 9 and 20 of PI3KCA (23), exons 2 and 5 of PTEN (23), exons 1 and 2 of KRAS (23), and exons 11 and 15 of BRAF (23), as previously described. The PCR products underwent automated DNA sequencing.

FISH analyses were used to investigate the status of PDGFRB, EGFR, RET, MET, and PTEN as previously described (21, 23, 24).

Representative 5-μm sections were obtained from the fixed samples, and the first part of the immunofluorescence protocol was done as previously described for immunohistochemistry (21). Anti-MET, anti-GFRα3, anti-RET, and anti-VEGFR2 antibodies (Table 1B, part 1) were incubated overnight at 4°C, and the slides were washed three times with 0.01% (v/v) Triton X-100 in PBS buffer. In the case of the GFRα3/MET colocalization experiment, after being incubated with anti-GFRα3 antibody, the slides were incubated for 30 minutes at room temperature with rabbit anti-goat biotinylated antibody (dilution 1:100; E0466; DAKO) followed by Alexa Fluor 546 streptavidin (dilution 1:1,000; Invitrogen). Thereafter, the slides already incubated with anti-MET antibody were incubated with Alexa Fluor 488 anti-rabbit antibody (dilution 1:1,000; Invitrogen) for 1 hour at room temperature. For the RET/MET colocalization experiment, after being incubated with anti-MET antibody, the slides were incubated for 30 minutes at room temperature with goat anti-rabbit biotinylated antibody (dilution 1:100; E0432; DAKO) followed by Alexa Fluor 546 streptavidin (dilution 1:1,000). Thereafter, the slides already incubated with anti-RET antibody were incubated with Alexa Fluor 488 anti-mouse antibody (dilution 1:1,000; Invitrogen) for 1 hour at room temperature. In the RET/VEGFR2 colocalization experiment, the slides were first incubated with anti-VEGFR2 antibody and then incubated for 30 minutes at room temperature with goat anti-rabbit biotinylated antibody (dilution 1:100; E0432) followed by Alexa Fluor 546 streptavidin (dilution 1:1,000). RET labeling was done as above. After washing three times with 0.01% (v/v) Triton X-100 in PBS buffer, the nuclei were stained with DRAQ5 (1:1,000; Biostatus Ltd.) for 30 minutes at room temperature. Subsequently, the slides were mounted using SLOW-FADE anti-fade gold (Invitrogen) and analyzed by means of a confocal microscope (Microradiance 2000; Bio-Rad Laboratories, Inc.) equipped with Ar (488 nm), HeNe (543 nm), and red diode (637 nm) lasers. The images (512 × 512 pixels) were obtained using a 60× oil immersion lens and analyzed using ImagePro 6.3 software. The shown images represent extended depth of field frames in stack, and the focus regions were selected for maximum intensity. The pinhole diameter was adjusted on the basis of the value suggested by the acquisition software to obtain the maximum power of resolution.

Samples of four usual and three round cell MLS were analyzed; the results are summarized in Supplementary Table S1.

We used phospho-RTK arrays to make an exploratory analysis of two usual (nos. 2 and 4) and two round cell cases (nos. 6 and 7) for which sufficient frozen material was available. This revealed heterogeneous activation profiles in which EGFR and PDGFRB were highly activated, whereas IGF-insulin receptor (IR), IR, VEGFR2, macrophage-stimulating protein receptor (MSPR; RON), PDGFRA, macrophage colony-stimulating factor receptor (M-CSFR), and fibroblast growth factor receptor 2α were less activated (data not shown). Immunoprecipitation/Western blotting was then used to extend the analysis to all seven cases, concentrating on PDGFRB and EGFR, but also adding MET and RET because published data indicate they are biomarkers of MLS (4, 5, 7); we have found activated RON (a member of the MET family of receptors) in our arrays (25), and there is a difference in the efficacy of the antibodies used in the arrays and those used for Western blotting (21).

Cumulatively, the results showed that PDGFRB, RET, and MET were expressed and phosphorylated in all of the analyzed cases (no frozen material was available for MET immunoprecipitation in case no. 1; Fig. 1A), whereas, although expressed in all cases, EGFR was phosphorylated to different extents (from no phosphorylation in no. 2 to strong phosphorylation in no. 5).

Furthermore, given the reported heterodimer formations of PDGFRB/EGFR (26, 27), EGFR/MET (28–30), and RET/VEGFR2 (14), we did coimmunoprecipitation experiments, which were also used to investigate RET/MET and PDGFRB/MET. The results showed the coimmunoprecipitation of PDGFRB/EGFR, PDGFRB/MET (Fig. 1B, a), RET/MET (Fig. 1B, c), and RET/VEGFR2 [the interaction between RET and VEGFR2 was shown in the three cases for which material was available (nos. 2-4); see below and Fig. 3D], whereas EGFR and MET did not coimmunoprecipitate (Fig. 1B, b).

This analysis was complemented by immunohistochemical antibody studies of surgical samples, which confirmed that all seven cases showed EGFR cytoplasmic tumoral cell immunoreactivity, and that PDGFRB, MET, and VEGFR2 immunodecorated tumoral branching vessels and tumoral cells, whereas RET was restricted to the vascular component.

We investigated whether the RTKs were activated by means of activating mutations, gene amplification, or an autocrine/paracrine loop.

As no evidence of activating mutations or genomic alterations were found in any of the analyzed receptors (PDGFRB, EGFR, RET, and MET), we used RT-PCR and real-time PCR to investigate the presence of their respective ligands.

RT-PCR analysis showed that among the RET ligands, only artemin was expressed, whereas HGF (the MET ligand) was expressed in all cases but nos. 2 and 6. Real-time PCR showed that all of the cases expressed PDGFB (the PDGFRB ligand), VEGFA (the VEGFR2 ligand), as well as artemin (see Supplementary Data); TGF-α was expressed, but EGF (the other investigated EGFR ligand) was not. Finally, immunofluorescence microscopy confirmed the expression of HGF, artemin, PDGFB, and VEGFA in three samples (data not shown).

Taken together, the above findings were consistent with an autocrine/paracrine activation loop of PDGFRB, EGFR, RET, and MET and transactivation between the investigated RTKs (PDGFRB/EGFR, PDGFRB/MET, MET/RET, and RET/VEGFR2).

However, the unexpected distribution of RET and MET immunoreactivity in the vascular and cellular tumoral components and the expression/activation of VEGFR2 and RET at the biochemical level (validated by means of immunohistochemistry as the expression of the two biomarkers involving branching tumoral vessels) deserved further investigation.

To explain why RET immunoreactivity was restricted to tumoral vessels (Fig. 2A, a), whereas MET decorated both vessel and tumoral cells (Fig. 2A, c), we searched for the expression of GFRα3 by means of immunohistochemistry and Western blotting.

We reasoned that if GDNF signaling did not necessarily need the coexpression of RET and GFRαs (RET coreceptors; ref. 13), RET-null tumoral cells should be immunoreactive to one of the four RET coreceptors. As each GFL has unique binding affinity for each GFRα protein and all of our cases expressed artemin (Fig. 2A, f), it could be expected that GFRα3 (the coreceptor that specifically binds artemin) was involved.

In line with this, immunohistochemistry showed that all of our samples expressed GFRα3 (Fig. 2A, b). Western blotting analysis of matched pairs of cryopreserved samples confirmed the immunohistochemical results (Fig. 2A, e).

Interestingly, GFRα3 was positive in both branching vessels and tumoral cells, a finding that supports the hypothesis that artemin and GFRα3 concurred in activating MET in tumoral cells even in the absence of RET expression (i.e., GFRα3/artemin-mediated MET activation; ref. 13). Confocal microscopy confirmed the coexpression of MET and GFRα3 in the tumoral cells (Fig. 2A, d).

The restriction of RET and MET colocalization to the tumoral vascular component (Fig. 2B, b) and MET expression to tumoral cells (Fig. 2B, d) was confirmed by confocal microscopy, which was also used to explain the apparent discrepancy between the biochemical analysis showing RET/MET coimmunoprecipitation (Fig. 1B, c) and the immunohistochemical evidence of RET-null tumoral cells. The analysis showed that the tumoral component contributing to the biochemical result was restricted to the vessel walls (Fig. 2B, b and d).

Interactions between VEGFR2 and RET were shown by means of coimmunoprecipitation experiments using three cases for which material was available (Fig. 3D). The colocalization of RET and VEGFR2 in the vascular tumoral component was confirmed by confocal microscopy (Fig. 3E). Furthermore, the presence of VEGFA (Fig. 3F) and artemin transcripts (Fig. 3G), together with GFRα3 immunoreactivity (Fig. 3C) in all of the matched frozen samples/fixed sections, supports the notion that VEGFR2/RET interactions may also be mediated by a VEGFA/GFRα3 (RET coreceptor)/artemin complex stimulating RET signaling, which subsequently evokes MET (i.e., VEGFA/GFRα3/artemin-mediated RET/VEGFR2 interaction).

Investigation of the expression/activation status of AKT, ERK1/2, and mammalian target of rapamycin (mTOR) showed that AKT was always expressed, but its level of phosphorylation/activation varied from none (in usual MLS) to strong (in round cell MLS). ERK1/2 was expressed and phosphorylated in all cases, and mTOR was highly phosphorylated in all but no. 7 (Fig. 4). Sequencing did not reveal any mutations in the PI3KCA, PTEN, KRAS, or BRAF genes, and FISH showed that PTEN was disomic in all cases.

Taken together, the segregation of AKT activation with the round cell variant of MLS was the most important finding.

MLS samples taken from four patients treated with trabectedin and three treated with conventional chemo-radiotherapy were analyzed, and their pathologic responses were evaluated on the basis of previously published criteria (2). The details are shown in Table 1A, all of the results are given in Supplementary Table S1, and the most relevant are shown in Fig. 5.

PDGFRB and EGFR showed decreased phosphorylation levels (data not shown) in all cases, whereas RET and MET were expressed and phosphorylated to the same extent as in the untreated samples (Fig. 5A). RET/MET coimmunoprecipitation was also observed (Fig. 5A, a).

No relevant differences were observed between the treated and naive cases (see Supplementary Data for details).

The RTK downstream signaling in the treated cases mirrored that observed in the naive cases, and remarkably, the restriction of AKT activation to the round cell variant was also retained (Fig. 5B).

Mutational analysis showed that PI3KCA, KRAS, and BRAF were wild-type, and FISH showed that PTEN was disomic. PTEN sequencing revealed heterozygous inactivating mutations in two cases: no. 12 showed a 1512-nucleotide deletion (exon 5) leading to an R161E change and subsequent stop codon, and no. 14 showed a GC insertion between nucleotides 1164 and 1165 (exon 2) leading to a V45G change and a stop codon. To the best of our knowledge, neither mutation has been previously reported.

We investigated RTKs and their downstream signaling pathways in a series of tumors from naive and treated MLS patients, and identified some particular traits in the RTK activation profiles.

Our combined analyses showed that PDGFRB, EGFR, MET, RET, and VEGFR2 were all activated in the naive cases. We also found the presence of the corresponding cognate ligands in all but two cases in which no HGF was detected, which is otherwise dispensable for MET activation (31). As all of the analyzed receptors and downstream effectors were wild-type and had a disomic gene pattern, these findings are consistent with a ligand-dependent activation mechanism.

We also found evidence of cross-talk between PDGFRB and EGFR (26, 27), as well as between MET and RET, and MET and PDGFRB. This suggests that RET and MET activation is one of the major RTK-related factors in MLS tumorigenesis and that its inhibition may be a suitable therapeutic target. The immunohistochemical analysis showed the expected results for PDGFRB, EGFR, MET, and VEGFR2 receptors but highlighted unusual RET immunoreactivity: RET immmunostaining was restricted to tumoral branching vessels, whereas MET immunodecorated both tumoral cells and vessels (Fig. 2A, a and c).

In an attempt to explain this finding, and given that GFLs activate RET via different specific GFRαs (RET coreceptors) and that all of our cases expressed artemin (which has unique binding affinity for GFRα3), we used immunohistochemistry and found the coreceptor expressed in both the vascular and tumoral cell components (Fig. 2A, b). Bearing in mind that GDNF can induce MET phosphorylation in RET-deficient/GFRα1-positive cells (8), our immunohistochemical, Western blotting, and confocal assay evidence that tumoral cells express GFRα3 (Fig. 2A, b, e, and d) in the presence of artemin (Fig. 2A, f) supports the idea of RET-independent, GFRα3-dependent/artemin-induced activation of MET.

In addition to confirming that RET expression was restricted to tumoral vessels, confocal microscopy helped to explain the apparent discrepancy between the immunohistochemical and coimmunoprecipitation results (i.e., RET-null tumoral cells but RET/MET coimmunoprecipitation; Fig. 1B, c) by showing that RET/MET coexpression was restricted to the vascular wall, whereas the tumoral cells expressed MET (Fig. 2B).

Moreover, as VEGFR2 and RET, which were highly expressed in the vascular component of our specimens (Fig. 3A and B), proved to be activated at phospho-RTK array (VEGFR2; data not shown) or biochemical analysis (RET; Fig. 1A, c), and as RT-PCR (Fig. 3G) and Western blotting/immunohistochemistry (Figs. 2A, e, and 3C) showed that artemin and GFRα3 were expressed in all of our samples/sections, we attempted to verify whether VEGFR2 and RET contribute to vessel branching morphogenesis in a similar manner to that described in recent reports about ureteric bud branching morphogenesis in kidney (14). To this end, we did coimmunoprecipitation experiments and confocal assays that respectively showed that VEGFR2 and RET were physically associated (Fig. 3D) and coexpressed in vascular walls (Fig. 3E). We also used real-time PCR to assess VEGFA, which is known to stimulate the branching of endothelial tubules, and found that it was expressed in all cases (Fig. 3F). These findings support the existence of functional interactions between RET and VEGFR2 and suggest that, in the presence of the VEGFA ligand, the artemin/GFRα3 (RET coreceptor) complex may have an addictive effect on RET signaling that may contribute to vascular morphogenesis (14).

Taken together, the results summarized in Figs. 2 and 3 indicate a particular cell distribution of RET and MET in MLS. The RTKs in tumoral vessels are activated by means of cross-talk between the RET/MET, VEGFR2/RET, and VEGFA/artemin signaling pathways, whereas, in tumoral cells, MET seems to be activated not only by means of a ligand activation loop mediated by HGF but also as a result of RET-independent, GFRα3-dependent/artemin-induced signaling evoking MET. It is interesting to note that our findings are in line with the reported observation of RET transcripts restricted to the tumoral component of MLS surgical samples (4) and that they also closely mirror preclinical data showing that, in addition to MET and RET, GFRα1 belongs to the transcriptome of mesenchymal progenitor cells expressing FUS/CHOP fusion protein (7).

The most relevant factor in downstream signaling in our untreated cases was that AKT phosphorylation was restricted to the more aggressive round cell variant of MLS.

Although we analyzed only a few treated cases, various comments can be made. First, there were no differences in RTK or downstream activation profiles between the two types of treatment. Second, there were no posttreatment changes in the expression or phosphorylation of RET or MET despite the dramatic decrease in the vascular and cellular components observed in the histologic analysis of surgical samples taken from the responding cases (data not shown), and the same was true for AKT in the round cell variant. These last two findings seem to be relevant in at least two ways: They suggest that none of these biomarkers represents a target for the treatments used, and they broaden the number of observations and thus reinforce the pivotal role of RET, MET, and AKT in MLS.

Two cases treated with conventional chemo-radiotherapy showed a heterozygous stop codon PTEN mutation that does not preserve the phosphatase domain (32), thus suggesting that the loss of PTEN may contribute to tumor growth by enhancing AKT/PI3K signaling and tumor angiogenesis (33).

Taking all of our findings together, RET and MET activation seems to be the driving mechanism in usual MLS, with AKT activation being the hallmark of the round cell variant. Neither of the therapies used led to any relevant changes in the activation profiles, and as all of these patients can be expected to progress eventually, the combined inhibition of MET, AKT, and ERK1/2 might be considered in their further management. Finally, our results and recent data about the structure of the GFRα3/artemin complex suggest that an alternative therapeutic approach might be based on small GFRα-binding molecules whose exclusive interaction with GFLs would have the advantage of being highly specific (34, 35).

No potential conflicts of interest were disclosed.

We thank Dr. Maria Grazia Borrello (Department of Experimental Oncology, Operative Unit Molecular Mechanisms of Cancer Growth and Progression, IRCCS Istituto Nazionale dei Tumori Foundation, Milan, Italy) for her precious help with the RET biochemical analysis.

Grant Support: Associazione Italiana per la Ricerca sul Cancro (S. Pilotti).

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.