Combination of PPARγ Agonist Pioglitazone and Trabectedin Induce Adipocyte Differentiation to Overcome Trabectedin Resistance in Myxoid Liposarcomas

Trabectedin is a marine alkaloid originally extracted from a Caribbean tunicate and now prepared synthetically. It binds the minor groove of DNA, causing a distortion of the double helix that bends toward the major groove. Part of the molecule protrudes from the DNA interacting with DNA binding proteins. Trabectedin interacts with DNA-repair pathways and directly affects trans-activated transcription. It also modulates cytokine and chemokine production by cancer cells and tumor-associated macrophages. Due to these unique mechanisms of action, trabectedin acts against both tumor cells and the tumor microenvironment (1–3). In 2007, the drug was approved by the European Medicines Agency (EMA) for the treatment of adult patients with advanced soft-tissue sarcomas (STS) after the failure of anthracyclines and ifosfamide or for patients who cannot be given these medicines (4). In 2015, trabectedin also received FDA approval for the treatment of patients with unresectable or metastatic liposarcoma and leyomiosarcoma who received a prior anthracycline-containing regimen (5). Within the different histologic subtypes of STS, leyomiosarcomas and liposarcomas seem to benefit more than other subtypes from treatment with trabectedin. In particular, in myxoid liposarcomas (MLS)—an STS histotype generally associated with a good chemosensitivity—trabectedin has shown a very good activity. A retrospective study enrolling 51 pretreated MLS patients reported two complete responses and 24 partial responses, with an overall responses rate of 51% and a progression-free survival at 6 months of 88%. In most of the responding patients, changes in tissue density were observed during the radiologic evaluations before tumor shrinkage. Histologic analysis conducted after trabectedin treatment on surgically resected tumors showed major pathologic responses characterized by cellular depletion, disappearance of the vascular network, deposition of myxoid stroma, and appearance of monovacuolated adipoblasts (6). Further studies confirmed the exquisite sensitivity of MLS to trabectedin even in the neoadjuvant setting (7–9). Unlike anthracyclines, trabectedin has a good toxicity profile and does not cause cumulative toxicity, allowing prolonged treatment that lasts until tumor progression (10). At this stage, effective therapeutic options are not available for trabectedin-resistant MLS patients.

From a molecular point of view, MLS is characterized by the chromosomal translocation t(12;16) (q13;p11), resulting in the FUS–CHOP fusion gene. Rarely, the translocation t(12;22)(q13;q12) occurs, giving rise to the EWS–CHOP chimera. These translocations are recognized as being the pathogenic event that leads to MLS development (11–15). The aberrant transcription factor represses the expression of the master regulator of adipogenesis, PPARγ-2 and cEBPα, thus causing the inhibition of the late stages of adipogenesis (16, 17) with the accumulation of immature adipoblasts that proliferate without undergoing terminal differentiation.

The unusual mechanism of action of trabectedin in MLS was investigated in a panel of patient-derived xenografts (PDXs; refs. 18, 19). In these preclinical models as well as in a human biopsy, trabectedin was able to displace the protein FUS–CHOP from the promoters of its target genes. The inactivation of the chimera allows the reactivation of adipogenesis with consequent differentiation of adipocytes (19). The ability of trabectedin to detach FUS–CHOP from DNA appeared to be reduced in PDX models of MLS that are resistant to trabectedin treatment and do not undergo adipocyte maturation (19, 20).

A central role in adipogenesis is carried out by the peroxisome proliferator-activated receptor gamma (PPARγ). PPARγ activates the expression of cEBPα and vice versa in a positive loop. PPARγ and cEBPα cooperate in the regulation of the expression of several genes involved in adipocytic maturation (21). Thiazolidendiones are a class of antidiabetic drugs that act as PPARγ agonists. They have shown anticancer properties in several preclinical models, both in vitro and in vivo, where cell-cycle arrest, inhibition of angiogenesis, and differentiation and apoptosis were observed (22–26). In the clinic, no or very few responses were observed in phase II trials with this class of compounds in solid tumors (27–29). Nevertheless, Demetri and colleagues described dramatic tissue responses, with lipid accumulation in the cell cytoplasm and reduction of the proliferation marker Ki-67 in two cases of MLS and one pleomorphic liposarcoma patient, who received the PPARγ agonist troglitazone (30). A subsequent phase I study conducted with efatutazone in patients with advanced solid malignancies showed a prolonged partial response in a patient with metastatic MLS (31). These data suggest that PPARγ agonists can overcome the block of adipocyte differentiation induced by the fusion protein.

The possibility of combining trabectedin with PPARγ agonists was first hypothesized by Charytonowicz and colleagues (32). In a transgenic mouse model of MLS that was sensitive to trabectedin, they observed that rosiglitazone enhanced trabectedin-induced adipogenesis with a considerable improvement in mouse survival.

In this work, we tested the combination of trabectedin with the PPARγ agonist pioglitazone in various MLS PDXs characterized by different sensitivity to trabectedin. The efficacy results together with the obtained pathologic and molecular data provide a strong rationale for the clinical development of this combination.

Six- to 8-week-old female Athymic nude mice were obtained from Envigo. Animals were housed and handled under specific pathogen-free conditions in the Institute's Animal Care Facilities, which meet international standards; they are regularly checked by a certified veterinarian who is responsible for health monitoring, animal welfare supervision, experimental protocols and procedures revision.

Procedures involving animals and their care were conducted in conformity with the following laws, regulations, and policies governing the care and use of laboratory animals: Italian Governing Law (D.lgs 26/2014; Authorization n.19/2008-A issued March 6, 2008, by the Ministry of Health); Mario Negri Institutional Regulations and Policies providing internal authorization for persons conducting animal experiments (Quality Management System Certificate—UNI EN ISO 9001:2008—Reg. No. 6121); the NIH Guide for the Care and Use of Laboratory Animals (2011 edition) and EU directives and guidelines (EEC Council Directive 2010/63/UE), and in line with guidelines for the welfare and use of animals in cancer research (33).

Experimental protocols have been reviewed and approved by the IRFMN Animal Care and Use Committee, which includes members “ad hoc” for ethical issues, and by the Italian Ministry of Health.

Trabectedin (Yondelis) was provided by PharmaMar, S.A.; it was dissolved in water and further diluted in saline immediately before use.

Pioglitazone (Takeda) was dissolved in 10% DMSO and diluted with methocell 0.5% added with Tween 80 0.5%.

ML006 and ML017 patient–derived MLS xenografts were obtained from biopsies of patients suffering from round cell (RC) variant of MLS and maintained through serial transplantation in mice as previously described (18). ML006 was characterized by an innate resistance to trabectedin, whereas ML017 was very sensitive. ML017/ET was obtained from ML017 through exposition at repeated in vivo cycles of trabectedin acquiring a resistant phenotype (20). The histologic features of the tumors grown in mice were verified after each passage, and compared with that of the original human sample in order to maintain the clinical relevance of these models.

When tumor burden reached about 300 to 400 mg, mice bearing ML006, ML017, or ML017/ET xenografts were randomized to receive trabectedin 0.15 mg/kg i.v., every 7 days for three times (q7d×3), pioglitazone 150 mg/kg p.o. daily for 28 days or their combination. Tumor growth was measured using Vernier caliper, and tumor weights were calculated by the formula: length × (width)²/2.

To perform molecular and pathologic studies, tumor-bearing mice were treated as described above. Fourteen days after the last dose of trabectedin (3 hours after the last dose of pioglitazone), mice were sacrificed. Tumor samples were collected, frozen in dry ice or formalin-fixed and paraffin-embedded for hematoxylin/eosin staining. At least three biological replicates were used for each experimental condition.

Each tumor weight (TW) measure was normalized to the tumor weight of the same mouse at the start of treatment, and treatment efficacy was evaluated in the normalized tumor weight curve of individual mice using three independent parameters: tumor growth (usually referred to as “growth inhibition,” GI) during treatment, tumor weight at nadir (TW_nadir), and absolute growth delay (AGD).

The percentage of GI, indicative of the short-term antiproliferative effect, measures the relative tumor growth between the start (day 0) and the end (day X) of treatment and was calculated adapting the NCI definition (33, 34) to the case:

%GI_0-X = [(TWT_X – TWT₀)/(<TWC_X> − <TWC₀>)] × 100 when TWT_X ≥ <TWC₀>;

%GI_0-X = [(TWT_X – TWT₀)/<TWC₀>] × 100 when TWT_X < <TWC₀>,

where (TWT_X – TWT₀) is the increment of the tumor weight between day 0 and day X of the treated (T) tumor under analysis, (<TWC_X> – <TWC₀>) is the same increment as averages in the control (C) group (both TWT₀ and <TWC₀> are equal 1 by effect of the previous normalization).

TW_nadir, indicative of the extent of tumor shrinkage (usually in the middle-term after treatment), is the minimum reached by the normalized tumor growth curve, from the start of treatment. In controls and when no tumor regression was observed, TW_nadir was equal to 1 (the normalized weight at the start of treatment).

AGD, indicative of the long-term delay of tumor regrowth, was calculated as the difference (in days) between the time to reach a target size in a treated tumor and the median time to reach the same size in the control group (35, 36). Depending on tumor growth curves, AGD was calculated at four or six times the size at the start of treatment (AGD4 and AGD6, respectively).

The histologic criteria applied to define MLS histotype and its usual and RC subtypes were the ones described in the WHO classification (38).

RNA was extracted using the miRNeasy Mini kit (QIAGEN), according to the manufacturer's protocols. Using the Low Input Quick Amp Labeling Kit (Agilent Technologies), 150 ng of total RNA was reverse transcribed into Cy3-labeled cRNA and then hybridized onto commercially available array platforms as previously described (39). For each treatment, at least three biological or technical replicates were used. Raw data from Agilent Feature Extraction version 11 were preprocessed, removing features marked as unreliable by the scanning software. Arrays were normalized using the “quantile” method (40), and a batch correction was applied to normalized data. Raw data are available on the ArrayExpress database, under accession ID E-MTAB-8632.

A linear model for microarray analysis (41) was used to determine differentially expressed genes applying a correction for technical replicates and for batch bias (42) and setting a log fold-change cutoff at ±1, and false discovery rate corrected P value less than or equal to 0.05 (43). Each comparison was performed setting treated samples versus untreated control.

Gene Set Enrichment Analysis (44) was used for functional enrichment analysis comparing gene-expression data from each treatment with untreated controls, using default parameters. Gene sets used were biological states or processes defined as hallmark (50 gene sets; ref. 45) and Biological Process from the Gene Ontology (46), both retrieved from the Molecular Signature Database (MSigDB, version 6.1). EnrichmentMap application version 3.0.0 of Cytoscape version 3.6.0 was used for plotting, setting a P < 0.01, q-value <0.05, and using the overlap parameter set at 0.20.

Total RNA (250 ng) was reverse transcribed using the High-Capacity cDNA Reverse Transcription Kit according to the manufacturer's instructions (Applied Biosystems). Real-time PCR was performed in triplicate for each case by using specific primer for five selected genes (belonging to the adipogenesis pathway) found to be differentially expressed in treated tissue samples compared with untreated control. These genes were LIPE, ADIPOQ, LPL, FABP4, and PLIN1. All the reactions were carried out on the 7900HT Fast Real-Time PCR System (Applied Biosystems) using QuantiFast SYBR Green PCR Master Mix (Qiagen). Data were normalized using geometric mean of two selected invariant genes (CARNMT1 and G6PD). Analysis was performed by using the 2^−ΔΔCt protocol and expressed as fluorescence intensity arbitrary unit.

Proteins, extracted from the frozen specimens, were homogenized in protein lysis buffer, loaded on SDS-PAGE and immunoblotted as previously described (44, 47). Odissey FC Imaging System (LI-COR) was used for the acquisition. Primary antiadiponectin (1:1,000, cat. No. ab22554) was purchased from Abcam, and actin (1:500, cat. No. sc-1616) was purchased from Santa Cruz Biotechnology.

Treatment effect on xenograft tumor growth curves has been formally tested using a nonparametric approach. For each mouse, the partial tumor growth rates (k) between every time interval were calculated as follows: k = [log(TW_ti+1) − log (TW_ti)]/(t_i+1 − t_i). The experimental groups were compared two by two using a Wilcoxon rank-sum test, stratified by time intervals, on the obtained k values.

The Student t test for unpaired samples was used to compare %GI, TW_nadir, and AGD parameters and real-time PCR data between different treatment groups.

The antitumor efficacy of trabectedin in combination with pioglitazone was studied in three PDX models of RC MLS characterized by different sensitivities to trabectedin. ML017 is very sensitive, whereas ML006 shows an innate partial resistance to the drug. ML017/ET was obtained from ML017 through the administration of 10 cycles of trabectedin in vivo. Once obtained, the resistant phenotype was stable, being maintained for several passages in mice without further treatment. Tumor growth curves are shown in Fig. 1. In all xenograft models, the combination of trabectedin with pioglitazone significantly improved tumor response compared with single-agent treatments (P < 0.001 in ML006, P < 0.05 in ML017, and P < 0.01 in ML017/ET).

A more in-depth analysis of the response to treatment was made considering short, intermediate, and long-term characteristics of the tumor volume versus time curve for each individual tumor. A score of the short-term response to treatment was obtained based on the growth inhibition measured on the increase/decrease of tumor volume at the end of treatment (or alternatively on the day of sacrifice because of ethically acceptable endpoint, that is, TW ≥ 1.5 g had been reached), compared with the volume at the start of treatment (%GI_0-X). The intermediate-time score was provided by the minimum value of the tumor weight (TW_nadir) relative to the weight at treatment start (TW₀). Absolute growth delay (AGD4) was calculated by the time to reach a target relative volume (4 times the size at the start of treatment) minus the median time to reach the same target in the control group.

As shown in Fig. 1A, the growth of untreated ML006 tumors was very slow, with doubling times of Td = 37.8 ± 5.0 days (N = 10 tumors). In this model, trabectedin and the combination with pioglitazone induced a strong GI, while a modest GI was detected in the pioglitazone alone group.

Some variability was observed in the group of mice treated with trabectedin alone: after initial growth inhibition during treatment (average %GI = 32%), prolonged tumor growth arrest was achieved in four of 10 tumors, while the others regrew with 52 days' average AGD4.

A closer inspection of the time course of TW indicates that pioglitazone alone progressively reduced the tumor growth rate with complete arrest in week 3. The pioglitazone-induced growth arrest, without shrinkage, continued for a long time after treatment discontinuation, up to the end of follow-up (6 months) in four of 10 ML006 tumors. The other tumors eventually regrew, with 55 days' average AGD4.

The combination enabled us to consolidate the response achieving robust (>40%) tumor shrinkage in seven of nine tumors and minor shrinkage in two, with 0.39 average TW_nadir. Complete responses were achieved in three of nine tumors with undetectable residual mass at sacrifice, and partial responses in four of nine mice, with tumor size remaining unchanged from the end of treatment. A very slow regrowth (98 days AGD4) was observed in only one tumor. One mouse died for undiagnosed reasons 1 week after the last trabectedin dose, its body weight loss was ∼10%.

ML017 tumors were characterized by a doubling time of 10.9 ± 3.3 days (N = 15), shorter than that of the ML006 tumor. As a consequence, untreated tumors reached maximal size within 3 weeks, and mice had to be killed for ethical reasons (Fig. 1B). Trabectedin-induced growth inhibition was already observed in the first week of treatment and tumor shrinkage occurred in the second, leading to a negative %GI on day 13. Tumor shrinkage continued during trabectedin treatment and for some days after its discontinuation, reaching a nadir (TW_nadir = 0.58 ± 0.06) on day 30 (median value). Thereafter, all tumors regrew, with an average AGD4 of 46 days. Pioglitazone alone did not inhibit tumor growth. Tumors in pioglitazone-treated animals grew even faster than controls, and mice were sacrificed on day 13. This rapid growth might have prevented the observation of a posttreatment inhibition, as observed in the ML006 model. The effects of the treatment combination were significantly superior to those of trabectedin alone, reaching a very low nadir (TW_nadir = 0.23 ± 0.03). The combination delayed regrowth (average AGD4: 69 days), which occurred in two mice only after 4 months.

In mice bearing trabectedin-resistant ML017/ET tumors, tumor growth was faster than in animals with their sensitive counterpart (DT 7.4 ± 0.4 days, N = 11). In these mice, pioglitazone-treated tumors could be measured up to the third week (Fig. 1C), allowing detection of both an initial fast growth phase (GI_0-7 = 163%) and a subsequent decreased growth rate in week 3 (GI_0-21 = 91%). As expected, trabectedin was much less effective in these mice than in the ML017 tumor–bearing mice without observable tumor growth arrest or shrinkage. AGD6 was only 12 days on average.

In mice bearing the ML017/ET tumor, the combination was more efficacious than trabectedin alone with values of 35% GI_0-21 and 23 days AGD6. Tumor regression was not observed, and all tumors regrew in the fifth week. Nevertheless, the addition of pioglitazone to the treatment regimen caused a reduction in growth rate (Td = 16.7 ± 1.1 days, range, 12–24 days). This in turn caused prolongation in murine survival, i.e., the period within which maximal TW was achieved (1.5–2 g), by about 20 days compared with mice treated with trabectedin only.

The efficacy of the combined treatment was further confirmed in the ML004 model, another PDX partially resistant to trabectedin (Supplementary Fig. S1). Both trabectedin and pioglitazone were well tolerated with minimal or no body weight loss and absence of clinical signs of distress. Likewise, the combination did not increase toxicity (Fig. 2A–C).

Morphologically, in samples from all three xenografts from untreated mice, the observed pattern of growth was consistent with the RC variant of MLS. However, ML006 tumors presented with more intervening stroma than ML017 and ML017/ET tumors, consistent with the differences in doubling times. ML006 tumors in mice that received trabectedin showed a decrease of cellularity and vascular supply, in line with the known effect of the drug, but did not show a clear evidence of adipocytic differentiation (Fig. 3A). Adipocytic differentiation, mainly represented by univacuolated lipoblasts harboring hyperchromatic scalloped nuclei, was evident after administration of pioglitazone alone or in combination. Maturation effects persisted and increased after drug discontinuation as visible in a tumor that had regressed for a long time. These observations closely paralleled tumor shrinkage. In mice that received trabectedin, ML017 tumors exhibited a decrease of cellularity with persistence of vascular network, evidence of lipoblastoma-like adipocytic maturation, and occasional white foci of myxoid material in the background (Fig. 3B). All of these findings are consistent with the prolonged growth arrest observed. Pioglitazone administration induced a diffuse microvesicular lipid accumulation without subsequent changes in nuclear morphology. Inconsistent with the observation in ML006 tumors, the nuclei of ML017 tumor cells retained their primitive mesenchymal-like morphology and the maturation effects are not associated with a tumor growth arrest. The combined treatment showed mixed changes, including some effects on the tumor vasculature (Supplementary Fig. S2) that correlated with drug activity eventually leading to delayed tumor regrowth (Fig. 3B).

In mice bearing ML017/ET tumors, trabectedin treatment did not cause morphologic changes when compared with controls, consistent with the lack of tumor growth arrest. After pioglitazone, the immature nonlipogenic RCs showed evidence of lipid accumulation similar to that observed in ML017 tumors. In addition, they exhibited a zonal pattern enriched with signet ring cells. This cell pattern was more prevalent in mice on the combination treatment, consistent with the prolonged AGD (Fig. 3C).

Figure 4 shows a selection of significant pathways engaged by the treatments. As shown in Fig. 4A, in ML006 tumors trabectedin did not activate adipogenic processes, while pioglitazone regulated the adipogenic pathway together with processes related to it, such as fatty acid metabolism and lipid storage. The treatment combination induced adipogenic differentiation, probably due to the preponderance of the effect of pioglitazone. The molecular data are consistent with the H&E staining (Fig. 3A).

Figure 4B and C show the results of the gene-expression array analysis, illustrating networks related to ML017 and ML017/ET tumors: trabectedin alone activated apoptosis, adipogenesis, and several related pathways in the ML017 tumor, while apoptosis and adipogenesis were not significantly modulated in the resistant counterpart. However, processes related to adipogenesis such as lipid storage and positive regulation of lipid transport were seen in ML017/ET tumors. These data are consistent with the morphologic observations (Fig. 3B and C). Pioglitazone alone and in combination activated adipocytic differentiation in both tumor types.

The heat map in Supplementary Fig. S3 shows the levels of expression of major genes implicated in adipogenesis, such as PPARγ, CEBPα, FABP4, PLIN1, LEP, AGPAT2, GLUT4, LPL, and ADIPOQ. As expected, these genes, especially those regulated downstream in the adipogenic process, had higher expression values where the adipogenesis was significantly activated in the network analysis (Fig. 4). Among the genes, ADIPOQ (Adiponectin) seemed to be the one of the most expressed. This finding was validated by real-time PCR (Fig. 5A), and this protein was selected as a marker in Western blot studies. Figure 5B shows adiponectin levels confirming its expression where adipogenesis is activated.

In this article, we demonstrate that the resistance against trabectedin of PDX of MLS can be overcome by concomitant treatment with the PPARγ agonist pioglitazone. This effect was demonstrated in MLS models of both innate and acquired resistance, and it is associated with the reactivation of adipocytic differentiation.

The overexpression of CHOP is a typical feature of endoplasmic reticulum stress where it is associated with the negative regulation of genes related to lipid homeostasis such as cEBPα (48, 49). Nevertheless, its overexpression in transgenic mice did not cause tumors, while the expression of FUS–CHOP led to the development of liposarcomas, meaning that the FUS domain of the chimera is required for tumorigenesis (13).

The biochemical role of the FUS–CHOP chimera in preventing adipocytic differentiation was previously described (16, 17). They demonstrated downregulation of PPARγ2 and CEBPα expression in different in vitro systems, that is, mouse embryonic fibroblast and human liposarcoma cell lines carrying the chimera. These genes are crucial for terminal adipogenesis. This is particularly the case for PPARγ2, which can activate adipogenesis even in the absence of cEBPα. Our group has previously demonstrated, in a panel of MLS PDX, the ability of trabectedin to interfere with the binding of FUS–CHOP to DNA promoters, thus blocking its transcriptional activity (19, 50) and reactivating the differentiation process leading to the maturation of adipoblasts in adipocytes. Adipocytic differentiation following trabectedin treatment was previously reported both in preclinical systems and in the clinic. In MLS patients, after several courses of trabectedin, structural changes assessed by CT and MRI scanning were reported before tumor shrinkage. Biopsies of residual tumor masses confirmed adipocytic maturation of the neoplastic tissue (6, 32), thus indicating that the striking antitumor activity of trabectedin in MLS is related to the activation of adipocytic differentiation. Consistent with previous preclinical (18) and clinical evidence we have observed adipocytic maturation only in the trabectedin-sensitive models such as ML017. In this model, differentiation is associated with increased expression of cEBPα and with the activation of the adipogenic pathway. In contrast, according to both pathologic and molecular analyses, differentiation was not observed in the ML006 or ML017/ET tumors, the first of which harbors innate resistance, and the second of which has acquired resistance against trabectedin. These observations further confirm our previous data according to which lipidic maturation and PPARγ2 expression were observed in the trabectedin-sensitive ML017 and ML015 xenografts, but not in the resistant ML004 (19). These data support the hypothesis that the block of adipocytic maturation could be one of the mechanisms underlying the resistance of MLS against trabectedin and suggest that restoring adipogenesis may represent a rational strategy to overcome it.

Molecular studies performed in human samples showed that PPARγ is expressed at high levels—comparable with normal fat—in most liposarcoma histotypes but not in other STS subtypes (24). The treatment of primary cultures of liposarcomas obtained from surgically resected sarcomas treated with pioglitazone showed lipid accumulation and morphologic changes characteristic of mature adipocytes. Differentiation was not observed in primary cultures of different STS histotypes that did not express PPARγ. These data suggest that the block of differentiation in liposarcomas as a consequence of different molecular reasons may be overcome by stimulation of PPARγ (23).

Clinical studies report tumor responses after treatment with PPARγ agonists in patients affected by MLS. Demetri and colleagues observed “dramatic histologic changes” in two cases of MLS and one pleomorphic liposarcoma patient treated with troglitazone (29). Posttreatment samples were characterized by lipid accumulation in the cytoplasm and increased cell volume without changes in the morphology of the nuclei. MRI scans performed after 6 weeks of treatment showed a moderate increase in tumor mass with changes in fat density signals compared with pretreatment images. These findings are consistent with those observed in our xenograft models after treatment with pioglitazone as reported here. Histologically, they showed diffuse lipid accumulation in tumor cells exhibiting microvesicular cytoplasm coupled with nuclear footprints of stemness, and rapid growth that was associated with an initial tumor progression. These similarities between clinical and preclinical observations corroborate the potential clinical relevance of our data.

More recently, a prolonged partial response lasting 690 days was observed in a patient with MLS enrolled in the phase I study of efatutazone. Subsequent surgical resection demonstrated the absence of viable disease in three of four remaining neoplastic lesions and the patient remained tumor free at the last reevaluation 3 years after treatment start (31).

Collectively, these data are consistent with the notion that within the myxoid–RC continuum, both trabectedin-sensitive and -resistant RC variants retain the machinery to activate the adipocytic program that may be targeted pharmacologically.

The possibility to combine PPARγ agonists and trabectedin was first proposed by Charytonowicz and colleagues (32). They generated genetically modified mice expressing FUS–CHOP under the control of the mesoderm-specific promoter Prx1 and crossed them with p53-null mice to obtain spontaneous tumor formation. The resultant TCp53-null mice were treated with trabectedin, the PPARγ agonist rosiglitazone, or with the combination of both after the emergence of sarcoma. Trabectedin-treated mice displayed increased survival associated with dramatic adipocytic differentiation and these were even more marked in mice on the combination. Rosiglitazone alone did not affect outcomes compared with untreated mice. Interestingly, all untreated and rosiglitazone-treated mice died of sarcoma progression, whereas mice that received the combination and half of the mice treated with trabectedin died of secondary lymphoma related to the loss of p53. This observation suggests that the combination of trabectedin and PPARγ agonist may not only improve trabectedin efficacy but may eventually lead to the cure of the original sarcoma.

In our work, we demonstrated that the combination of trabectedin and pioglitazone can increase the efficacy of trabectedin alone not only in MLS with acquired resistance after many courses of trabectedin treatment, but also in ab initio–resistant liposarcomas. Notably, we observed in ML006 tumors with innate resistance against the drug complete regressions in three of nine mice and partial responses, with minimal residual disease, in four of nine mice. At the end of the observation period, tumor regrowth was observed in only one mouse.

Notably, pioglitazone itself did not elicit toxicity, and the toxicity of trabectedin in terms of weight loss was not increased when combined with pioglitazone.

The preclinical findings reported here as well as some previous preclinical and clinical observations reported in the literature provide a strong rationale to undertake a clinical study aimed at assessing if treatment with a PPARγ agonist—such as pioglitazone—increases the efficacy of trabectedin in patients with MLS, leading possibly to a clinically significant prolongation of PFS or even to the cures in some patients. We surmise that the rationale for a clinical study on the combination of PPARγ agonists with trabectedin in MLS is compelling. Yet, there is no evidence that this is also the case for other liposarcomas, even though it is known that all liposarcomas express high levels of PPARγ receptors. This issue is currently under preclinical evaluation using appropriate preclinical models of liposarcomas other than MLS.

One aspect that needs to be considered in the design of a clinical study of the combination of a PPARγ agonist with trabectedin is that in our preclinical models the efficacy of pioglitazone became evident only after prolonged chronic treatment lasting several weeks. Designers of clinical trials who aim at testing this combination should bear the requirement of prolonged treatment with the PPARγ agonist in mind.

Therefore, the efficacy of coadministration of trabectedin with pioglitazone should be explored early on during treatment, when tumor growth is still controlled by trabectedin, definitely before the tumor has acquired robust resistance against trabectedin leading to its rapid progression.

P.G. Casali reports receiving speakers bureau honoraria from PharmaMar, Eisai, Pfizer and Eli Lilly, and is a consultant/advisory board member for Bayer, Deciphera, Eisai, and Nektar Pharm. A. Gronchi reports receiving speakers bureau honoraria from Novartis, Pfizer, Bayer, Lilly, PharmaMar, Nanobiotix and SpringWorks, and reports receiving commercial research grants from PharmaMar. No potential conflicts of interest were disclosed by the other authors.

Conception and design: R. Frapolli, A. Gronchi, S. Pilotti, M. D'Incalci

Development of methodology: R. Frapolli, E. Bello, P. Ubezio, L. Porcu

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): R. Frapolli, E. Bello, M. Ponzo, I. Craparotta, S. Ballabio, L. Carrassa, S. Brich, R. Sanfilippo, P.G. Casali, A. Gronchi, S. Pilotti

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): R. Frapolli, L. Mannarino, S. Marchini, L. Carrassa, P. Ubezio, L. Porcu, R. Sanfilippo, P.G. Casali, S. Pilotti

Writing, review, and/or revision of the manuscript: R. Frapolli, E. Bello, L. Mannarino, P. Ubezio, L. Porcu, R. Sanfilippo, A. Gronchi, S. Pilotti, M. D'Incalci

Study supervision: R. Frapolli, S. Pilotti, M. D'Incalci

This work was supported by the Italian Association for Cancer Research grant to M. D'Incalci (Project Number 19189).

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.