MDM2 derived from dedifferentiated liposarcoma extracellular vesicles induces MMP2 production from preadipocytes

Abstract

Dedifferentiated liposarcoma (DDLPS) is frequently diagnosed late and patients typically respond poorly to treatments. DDLPS is molecularly characterized by wild-type p53 and amplification of the MDM2 gene which results in overexpression of MDM2 protein, a key oncogenic process in DDLPS. In this study, we demonstrate that extracellular vesicles derived from patients with DDLPS or from DDLPS cell lines are carriers of MDM2 DNA that can be transferred to preadipocytes, a major and ubiquitous cellular component of the DDLPS tumor microenvironment, leading to impaired p53 activity in preadipocytes and increased proliferation, migration, and production of matrix metalloproteinase 2; treatment with MDM2 inhibitors repressed these effects. Overall, these findings indicate that MDM2 plays a crucial role in DDLPS by enabling crosstalk between tumor cells and the surrounding microenvironment and that targeting vesicular MDM2 could represent a therapeutic option for treating DDLPS.

INTRODUCTION

Mesenchymal origin liposarcoma (LPS) is the most common human sarcoma, comprising 24% of extremity and 45% of such lesions in the retroperitoneum, respectively (1). Dedifferentiated liposarcoma (DDLPS) poses a remarkable clinical challenge due to frequent large growth before clinical detection and the lack of any new and effective therapeutics. Since the early 1970s treatments consist of radical surgery, adjacent organ radiotherapy of undefined overall and disease-free survival impact, with the potential to damage adjacent organs, and untargeted toxic, marginal efficiency, systemic therapies. DDLPS is especially concerning given its propensity for primarily local and occasionally distant recurrence, accounting for an overall survival rate of only 10% at 10 years (2) with approximately 1500 new diagnoses of this lesion annually in the U.S. alone. Moreover, almost 60% of retroperitoneal DDLPS ultimately recur as synchronous multifocal tumors, even after initial margin negative resection; such multi-centric failures are typically beyond meaningful therapeutic interventions other than palliation (3,4). This unique pattern of multifocal loco-regional failure remains a key problem in DDLPS and the main cause of death; however, the underlying molecular mechanisms driving these multifocal recurrence processes have not been extensively explored, hampering the development of DDLPS-specific therapeutics.

In addition, no validated DDLPS patient-associated molecular biomarkers have been identified to inform prognosis, facilitate early detection of DDLPS progression or recurrence, or possibly predict therapeutic resistance (5). The American Joint Committee on Cancer (AJCC) staging prognostic algorithms remain defined by the same pre-therapeutic clinical parameters originally introduced in the 1970s. MRI or CT scanning is used for post-therapy surveillance; these modalities frequently cannot detect or resolve early recurrence versus scarring from previous resection, causing delays in subsequent treatment initiation. At the molecular level, DDLPS is characterized by MDM2 gene amplification with expression of wt p53; this unusual pattern is observed in approximately 10% of all human cancers, approximately 20% of soft tissue sarcomas, and in almost 100% of de-differentited (DDLPS) liposarcomas (6). Although the role of MDM2 as an oncogene has focused on its inhibition of wt p53, several studies have suggested that MDM2 may also have p53-independent roles, perhaps in sarcoma (7), and involved pathways have yet to be extensively examined (8).

Interactions between malignant and non-transformed cells can occur within the tumor microenvironment (TME); the DDLPS microenvironment contains preadipocytes (P-a), adipocytes, macrophage (9), and other cell types. Communication between tumor and TME cells is crucial in both normal and pathological circumstances; extracellular vesicle (EV) trafficking has emerged as one such process of tumor:microenvironment cell-cell communication (10). EVs are extruded nanoparticles involved in intercellular communication from donor to recipient cells via transfer of protein, nucleic acids, and other biologically active molecules (11). Tumor cell-derived EVs can influence non-cancer cells to generate pre-metastatic niches that facilitate tumor dissemination and growth (12). Studies demonstrate that uptake of cancer cell EV proteins and RNA molecules can induce phenotypic changes in recipient neighboring TME cells (13–17), thereby contributing to pre-metastatic niche formation as sites prone to foster metastasis via tumor cell colonization. Steps in pre-metastatic niche formation can include the acquisition of a pro-inflammatory phenotype by the stroma of the metastatic niche as well as extracellular matrix remodeling through matrix metalloproteinases (MMPs) (18). However, to date, processes potentially contributing to DDLPS pre-metastatic niche formation have not yet been identified.

Against that backdrop, we evaluated MDM2 in DDLPS-derived EVs isolated from both patient serum and DDLPS cell lines. DDLPS EV bearing MDM2 cargo induced preadipocytes to produce MMP2, a process potentially relevant to establishing the DDLPS loco-regional pre-metastatic niche, and thereby enabling multifocal failure in this disease.

MATERIAL AND METHODS

Patients and clinical samples

Blood samples of LPS patients (n=16) were collected from OSU James Cancer Medical Center, written informed consent was received from participants prior to inclusion in the study, in accordance with the Helsinki Declaration whose protocols have been approved by The Ohio State University Wexner Medical Center institutional Review Board. Patient venous blood (12 ml) was collected in Vacutainer® Plus whole blood tubes with K2 EDTA (BD, Franklin Lakes, NJ). Blood serum was retrieved from the whole blood samples via centrifugation at 1900 g x 10 min at 4C°, then aliquoted and stored at −80 C° until analysis. Healthy donor blood used in the discovery and in the validation sets was purchased from ZenBio. The detailed characteristics of patient and healthy control participants are summarized in Supplemental Table 1 and 2. Prior to any therapy, patient pathology was confirmed using surgically resected sarcomas, and graded as per standard FNCLCC criteria.

RNA/DNA isolation and RT-PCR

Total RNA from cellular samples and from EVs was isolated by using Norgen kit and following the provided instructions (Norgen BioTek). For cell line-derived EVs, RNA was isolated by using Norgen kit as described above. Total DNA derived from tissues, cell lines and EVs was isolated by using Qiagen kit following manufacture protocol.

The expression level of an individual gene starting from RNA preparation was determined using RNA sequence specific probes (MDM2-Hs01066930_m1; GAPDH-Hs00266705_g1 ThermoFisher) as per quantitative real-time RT-PCR-based detection methodology. Total RNA was reverse transcribed by using TaqMan® Advanced mRNA cDNA Synthesis Kit (ThermoFisher), according to the manufacturer’s protocol. GAPDH (Hs00266705_g1, ThermoFisher) and/or ACTB (Hs99999903_m1, ThermoFisher) was used to normalize quantitative Real-Time PCR on RNA cellular samples. The expression level of an individual gene starting from a DNA preparation was determined using DNA sequence specific probes (MDM2-Hs00540450_s1, ThermoFisher). As for the RT of MDM2-DNA from cellular EVs, the same quantity of vesicles was used (calculated by nanosight), and the results were normalized on GAPDH (Hs03929097_g1, ThermoFisher). Determination of number of molecules of MDM2 in the serum-EVs was performed using standard curve methodology (Supplemental Figure 1A–C). For the RT-PCR on the DNA-EVs of the serum, the normalization was volumetrically performed. All samples were run in triplicate.

Cell Culture

Human LPS cell lines Lipo246, Lipo863 and Lipo224 were established in our laboratory as previously reported (19). SW872 cells were obtained from ATCC. Cells were maintained using standard conditions and were grown in DMEM (Gibco), supplemented with 10% (vol/vol) FBS. Two sources of human P-a were used. All the experiments were performed using both cell lines, unless indicated othwerwise. Human preadipocytes (XA15A1) were purchased from Lonza and maintained following the manufacturer’s instructions. SGBS P-a inastead (20) were cultured in DMEM/Ham`s F12 (1:1) containing 33 μM biotin, 17 μM pantothenate, antibiotics (serum-free, basal medium), and 10 % FBS. All the cell line used in this study were acquired within the past 5 years and authenticated by STR on 7/14/18. All cell lines were tested for mycoplasma.

Extracellular-vesicle isolation and treatments

EVs were isolated according to He W. and Calore F. et al. (21). Serum-derived vesicles were isolated by using ExoQuick (System Biosciences), following manufacturer’s protocol. The quality and size of isolated particles was assessed through nanosight and EV size assessment was performed with nanosight, while we verified the purity of isolated particles by WB (Supplemental Figure 2A–C). For treatments with GW4869 (Sigma), Lipo246 cells were incubated with GW4869 5 μM diluted in FBS-depleted medium for 48h (as in Casadei et al. 2017), then EVs were isolated through ultracentrifugation. For all cellular treatments, P-a were seeded in a 12 well plate; after 24h they were treated with isolated EVs for 72 or 96h; SAR405838 (Sanofi-Aventis) was added at a final concentration of 0.2 μM as proposed by Bill et al. (22).

Western blotting

For immunoblotting analysis cells were lysed with ice-cold NP-40 Cell Lysis Buffer (Invitrogen) supplemented with protease inhibitors (Roche) for 30 min at 4ºC. Equivalent amounts of protein were first mixed with sample buffer, then loaded on a Criterion Tris-HCl 4-20% pre-cast gel (Bio-Rad) and transferred to PVDF or nitrocellulose membranes. Membranes were incubated overnight at 4°C with commercially available antibodies as indicated per experiment: anti -p53 (MA5-14516, Invitrogen) -p21 (#SC-756, Santa Cruz); -MDM2 (#MA1-113, Invitrogen); -GAPDH (#SC-48167, Santa Cruz); -βActin (#SC-1616, Santa Cruz) -Calnexin (#C7617, Sigma); -CD9 (#D8O1A, Cell Signaling); -Alix (#SAB4200476, Sigma); -TSG101 (#T5701, Sigma). The proteins of interest were detected through chemi-luminescence reaction. The band density of proteins was quantified using densitometric software (Odyssey, Li-Cor Biosciences) or ImageJ.

MTS, Migration assays and Cell Cycle

Cell proliferation was performed as described previously (9). Cell migration was assessed by using transwell migration chamber (Corning). Briefly, P-a were diluted in serum-free medium and seeded in the transwell upper chamber following different conditions (EV-depleted medium, Lipo246-derived EVs, Lipo863-derived EVs, EVs with or without SAR405838, EVs derived from the serum of healthy donors or from DDLPS patients). The lower chamber was filled with medium supplemented with 10% FBS. After 72h, filters were washed, fixed and stained with Coomassie Brilliant Blue (Sigma-Aldrich Corp.). Migrated cells in the lower surface of the filter were analyzed using image J. For SAR405838 treatment, isolated EVs were suspended in medium (without FBS) where SAR405838 at a concentration of 0.2 μM was added (22). For cell cycle analysis, cells were harvested, washed and fixed and then stained with 50 μg/mL propidium iodide (PI, Sigma Aldrich Co.) for 30 min. Cells were analyzed in a FACSCalibur, data were analyzed with ModFitLT v3.1 software (Verity Software House).

Copy Number Variation (CNV) Assay

Assessment of copy number variation quantified genomic MDM2 amplification in DDLPS tissues. Isolated genomic DNA samples were measured for concentration and quality using the Cytation 3 spectrophotometer (BioTek Instruments). Samples were diluted to 5 ng/μL with nuclease-free water, and assessed using the MDM2 copy number probe (item# Hs06365580_cn, cat# 4400291, Thermo Fisher Scientific) with the associated TaqmMan® Copy Number Assay kit (Thermo Fisher Scientific) in the StepOnePlus^™ Real-Time PCR System (Applied Biosystems).

Molecular number variation

We used RT-PCR to calculate the number of molecules of MDM2 DNA in the EVs of the serum, following the methodology as per Dubois et al. 2011 (23). First, we performed serial dilution of MDM2 synthetic oligo and calculated the number of molecules of MDM2 that correspond to each different concentration. Then a RT-PCR using MDM2 probe (Hs03929097_g1, ThermoFisher) was performed using these diluted synthetic oligo samples (Integrated DNA technologies). A standard curve was then constructed in which a specific number of molecules was assigned based on the corresponding Ct value (Supplemental Figure 1A–C).

Gelatin Zymography

Protein content of the isolated conditioned medium (CM) samples was quantified by μBCA according to manufacturer’s protocol (Thermo Fisher Scientific). Gelatin zymography was performed as described by Deshmukh and Toth (24,25). Briefly, 30 μg of protein collected from the conditioned medium (CM) were loaded onto precast gelatin zymography gels (10% polyacrylamide, 0.1% gelatin; Thermo Fisher Scientific). Samples were electrophoresed for 2 h at 120 V at 4°C. Gels were then washed three times for 10 min in renaturing buffer containing 2.5% Triton X-100 in 50 mM Tris-HCl (pH 7.5), to allow the proteinases to renature. Gels were then transferred to a developing buffer solution containing 1% Triton X-100, 1 μM ZnCl₂ and 5 mM CaCl₂ in 50 mM Tris-HCl (pH 7.5) overnight at 37°C, under gentle agitation. After 24 h, gels were stained with 0.5% (w/v) Coomassie Brilliant Blue (Thermo Fisher Scientific) in a solution of deionized water, methanol and acetic acid (50/40/10 v/v) for 2 h at room temperature. Gels were de-stained in the same solution, but without Coomassie Blue, for 10 min. Images were captured using a FluorChem E gel imager (Protein Simple) using the UV transilluminator (365 nm). Band intensities were quantified using ImageJ software as described by Ren (26) and normalized with respect to untreated condition.

DNA plasmids, virus production and transduction

Non-targeting control vector plasmid and shRNA targeting endogenous human MDM2 transcript were obtained from Origene. Both plasmids were packaged in Lenti-X^™ 293T cell line (Clontech) by transfection with Lenti-X^™ Packaging Single Shots (VSV-G) (Clontech). Lipo863 cells were then transduced with lentiviral particles in the presence of polybrene 8μg/mL (Sigma). Medium was replaced 24 hours after transduction.

Sequencing

MDM2 sequencing was analyzed by polymerase chain reaction (PCR) amplification and subsequent DNA sequencing of exon 1, 6, 10 (using primers built on the introns before and after each exon, primers description in Supplemental Table 3). PCR products were purified with the QIAquick PCR Purification Kit (Qiagen) according to manufacturer’s specifications. DNA sequencing was performed by the Genomic Shared Resource (GSR) at the OSUCCC Cancer Center.

Statistics

Differentially amplified genes and differentially expressed mRNAs between comparison groups were determined by two-sided t-tests and fold changes using log-transformed values. Unpaired t test with Welch’s correction was applied in serum and tissues samples analysis. A one-way ANOVA with Dunnett’s multiple comparison test was applied to the analysis of gelatin zymography. We also calculated the area under the ROC (AUC) of each ROC curve. AUC is the average sensitivity of the biomarker over the range of specificities that used as a summary statistic representing the overall performance of the biomarker. AUC of a biomarker with no predictive value would be 50%, whereas a biomarker with an AUC of 100% would indicate perfect ability to predict disease.

RESULTS

1. DDLPS patient serum-derived EVs contain high levels of MDM2 DNA.

The molecular hallmark of DDLPS are high levels of MDM2 concomitant with wt p53, a finding observed in nearly 100% of DDLPS tumors (6). However, the content of DDLPS EVs has not yet been assessed, thus we examined EVs isolated from DDLPS patient serum for the presence of MDM2 DNA as compared to normal individual control serum-derived EVs (N= 16 DDLPS patients and 6 healthy controls). Isolated EVs were characterized by nanosight and showed particle sizes in the characteristic 30-100 nm range (Supplemental Figure 3; details regarding patient and healthy control characteristics are described in Supplemental Table 1 and 2). The standard method to assess the DNA level of MDM2, is by the determination of Copy Number Variation (CNV) (27). However, the lack of a calibrator sample (a gene of known and stable copy number) contained within the EVs meant this CNV could not be used to determine the levels of MDM2 DNA in the DDLPS EVs. Consequently, RT-PCR incorporating a standard curve methodology was used to calculate a specific threshold cycle (Ct value) that corresponded to the number of EV MDM2 DNA molecules (Supplemental Figure 1) within the EVs of both DDLPS patients and healthy-control groups. Our results showed that the number of MDM2 DNA molecules present in DDLPS patients was significantly higher versus healthy counterparts (p≤0.001; Figure 1A). Interestingly, this increase was also concordant with the MDM2 copy number variation (CNV) as measured in DDLPS tissues (N=14) compared to normal adjacent tissues (N=5; p≤ 0.0045; Figure 1B; details of patients and normal controls are described in Supplemental Table 4 and 5). EV DNA sequencing of the entire exons 1, 6, 10 of MDM2 (using primers built on the introns before and after each exon; primer descriptions in Supplemental Table 3) showed the presence of MDM2 DNA within the isolated serum EVs (Figure 1C).

Figure 1. DDLPS patient serum-derived EVs contain high level of MDM2 DNA.

(A) RT-PCR representing the number of molecules of MDM2 in DDLPS patient serum EVs (n=16) compared to normal healthy controls (n=6) (p≤ 0.001). (B) MDM2 copy number variation (CNV) measured in DDLPS tissues (n=14) compared to normal adjacent tissues (n=5) (p≤0.0045). (C) EVs derived from DDLPS patient serum contain MDM2 by DNA Sequencing on the entire exons 1, 6, 10 of MDM2. (D) ROC curve analysis to estimate the sensitivity and specificity for circulating EV-MDM2 in discriminate DDLPS patients from controls. Results are presented as average ±SEM. Statistical analysis were performed using unpaired t-test with Welch’s correction. ** 0.001 ≤p ≤0.01; *** ≤ 0.001.

Receiving Operating Characteristic (ROC) curve analysis was conducted on the serum-EVs data obtained from RT-PCR, to estimate the sensitivity and specificity of circulating EV-MDM2 to discriminate DDLPS patients from controls (Figure 1D). The area under the curve (AUC) for MDM2 was 95.8% with a 95% confidence interval from 86.9% to 100%, indicating robust separation between the DDLPS and healthy controls.

2. DDLPS cells constitutively shed EVs enriched in MDM2 DNA.

Previously we showed that DDLPS cell lines release EVs (9). To verify the DDLPS tumor origin of MDM2 EVs isolated from the serum of DDLPS patients, we collected conditioned medium (CM) from different DDLPS cell lines, isolated the EVs by ultracentrifugation, and assessed MDM2 content. EV DNA sequencing of the entire exons 1, 6, 10 of MDM2 (using primers built on the introns before and after each exon; primer descriptions in Supplemental Table 3) demonstrated the presence of MDM2 DNA within isolated EVs (Figure 2A). EV size assessment was performed with nanosight, demonstrating particle sizes in the characteristic 30-100 nm range (Supplemental Figure 2A), while we verified the purity of isolated particles through the detection of typical EV proteins by WB (Supplemental Figure 2B, C). The low expression levels of cell normalizers such as RNU48, RNU6 and RNU44 within isolated vesicles confirmed the absence of cell contamination in the EVs preparations (9) (Supplemental Table 6). When measured using RT-PCR, the level of MDM2 DNA EVs secreted by DDLPS cell line (Lipo863, Lipo246 and Lipo224) demonstrated consistent and significant up regulation compared to the level of MDM2 DNA in P-a-derived EVs and LPS SW872 derived EVs (p≤ 0.01; Figure 2B). Furthermore, the level of MDM2 DNA in EVs was proportional to the level of MDM2 in the cells of EV origin (p≤0.0001; Figure 2C). The quantity of MDM2 DNA was calculated by RT-PCR, using the same amount of EVs for each cell line (calculated by nanosight) and normalized for GAPDH DNA. Taken together, these data indicate that DDLPS cells release EVs with high levels of MDM2 DNA that reflect the levels of MDM2 in the DDLPS cells. We then wanted to make sure that the increased level of MDM2 considered its DNA form, rather than mRNA or protein. So we measure the quantity of mRNA and proteins also in the isolated DDLPS EVs. While we were able to detect a high level of MDM2 in DDLPS EVs at the DNA level, the amount of MDM2 mRNA was much lower and the amount of MDM2 protein was undetectable.

Figure 2. DDLPS cells constitutively release EVs carrying MDM2 DNA.

(A) EVs derived from Lipo246 contain MDM2 by DNA Sequencing on the entire exons 1, 6, 10 of MDM2. (B) Level of MDM2 (calculated by RT-PCR) in DDLPS secreted vesicles (Lipo863, Lipo246, Lipo224) is consistently and significantly up regulated compared to the level of MDM2 DNA in P-a secreted EVs (p≤0.01). The level of MDM2 in EVs is proportional to the level of MDM2 in the originating cells. (C) Level of MDM2 in different LPS cell lines calculated by CNV (p≤0.0001). Results are presented as average ± SD. Statistical analysis were performed using t-test. ** 0.001 ≤p ≤0.01; *** ≤ 0.001.

3. DDLPS cell EV cargo MDM2 DNA is transferred to normal P-a

Since DDLPS predominantly arise in fat bearing areas of the retroperitoneum in which the TME is enriched for P-a, we wanted to determine whether MDM2 DNA was transferred from DDLPS-EVs to P-a. P-a were incubated with Lipo246-derived EVs for 72h. When we assessed the expression level of MDM2 mRNA within the recipient P-a cells, we observed that it was significantly increased compared to P-a incubated with EV depleted medium (p≤0.001; Figure 3A). To further demonstrate the transfer of MDM2 DNA from DDLPS to P-a, we treated the latter with increased amount of Lipo-246-derived EVs and determined the MDM2 mRNA expression level in the recipient cells. As shown in Figure 3B, the level of MDM2 in recipient P-a increased in proportion to the amount of Lipo246 EVs added. Finally, to verify whether the MDM2 DNA transfer led to an increased production of MDM2 protein within recipient P-a, we performed a western blot analysis of P-a lysates derived from cells incubated for 72 and 96h with Lipo-246-derived EVs (Figure 3C and Supplemental Figure 4). These studies showed that the level of MDM2 protein increased in incubated P-a in a time-dependent manner, reaching a three-fold incremental change compared to untreated P-a (P-a treated with EV depleted medium). Taken together, these data indicate that MDM2 DNA was transferred from DDLPS EVs to P-a as a biologically active molecule capable of being translated into MDM2 protein within the P-a cells.

Figure 3. DDLPS cell EVs cargo MDM2 DNA is transferred to recipient P-a and affects intracellular pathway downstream of MDM2.

(A) Level of MDM2 (measured by RT-PCR) in recipient P-a increases at mRNA level, when P-a are treated with Lipo246 EVs for 72h (p≤0.04). (B) Recipient P-a treated with Lipo246 EVs for 72h, show an increased level of MDM2-mRNA in a dose-response manner (p≤0.002). (C) When P-a are treated with Lipo246 EVs, the protein level of MDM2 increases 3 fold after 96h. Results are presented as average ± SD. Statistical analysis were performed using t-test. ** 0.001 ≤p ≤0.01; *** ≤ 0.001. (D) When P-a are treated with Lipo246-EVs (for 96h) they show a decreased level of p53 and p21 compared to untreated P-a (lane three). When P-a are treated with Lipo246-EVs together with MDM2 inhibitor (SAR405838, 0.2 μM), the inhibitory effect of EVs on p53 and p21 is rescued (lane four). Treatment of P-a with EVs isolated from Lipo863 (whose MDM2 levels are lower compared to Lipo246), produces results analogous to the treatment of P-a with SAR405838 (lane five). When P-a are treated with EVs isolated from Lipo863 where MDM2 is overexpressed (OE), the level of p53 and p21 decreases again (lane six).

4. Transfer of EV MDM2 DNA leads to downregulated P-a p53 activity.

We next examined whether MDM2 DNA, upon translation into MDM2 protein in recipient P-a, could downregulate P-a p53 activity (Fig. 3D). Using western blot, we determined the amount of p53 and p21 protein in P-a exposed to DDLPS EVs for 96h under different conditions. As shown in Figure 3D, incubation of P-a with Lipo246 EVs for 96h led to decreased p53 and p21 protein levels (lane three) compared to treatment with EV depleted medium (lane one). Next, we considered whether MDM2 was at least partially responsible for these changes. The selective MDM2 inhibitor SAR405838 blocks p53:MDM2 interaction at the protein level by occupying the MDM2 p53 binding site (28) causing an increase in both P-a p53 and p21 expression (lane two). To confirm the mechanism of action of the MDM2 inhibitor (SAR405838), a WB showing a concomitant increase in MDM2 protein expression after treatment with the MDM2 inhibitor is also provided (Supplemental Figure 5). When P-a were treated with Lipo246 EVs and SAR405838 for 96h, the EV inhibitory effect on p53 and p21 was abrogated (lane four) compared to P-a treated with Lipo246 EVs (in lane three), suggesting that uptake of MDM2 DNA by P-a inhibited p53 and p21 expression. P-a treatment with EVs isolated from Lipo863, a DDLPS cell line whose MDM2 levels are much lower than these of Lipo246 (see Figure 2, B and C), produced p53 and p21 results more closely resembling P-a treated with Lipo246 EV+ SAR405838 (lane five). Interestingly, when P-a were incubated with EVs derived from Lipo863 transduced with lentiviral particles for the overexpression of MDM2 (see material and methods and Supplemental Figure 6), the level of p53 and p21 decreased again resembling P-a treated with Lipo246 EVs (lane six). These data suggest that DDLPS EVs induce downregulated p53 activity which was due to MDM2 DNA transfer from DDLPS to recipient P-a per se.

5. EV-origin transferred MDM2 DNA confers oncogenic features in normal recipient P-a.

Next, we asked whether the transfer of MDM2 DNA within DDLPS-derived EVs conferred DDLPS cell pro-oncogenic features to normal P-a cells. P-a were incubated for 72h with EVs isolated from Lipo246 cells; when proliferation and migration was assessed, P-a exhibited enhanced proliferation and migration (p≤0.0001), compared to cells incubated with EV depleted medium or media alone (Figure 4, A and B). When P-a were incubated for 72h with EVs in the presence of the MDM2 inhibitor SAR405838, the rate of proliferation and migration of recipient cells was significantly impaired compared to EV treatment only, suggesting that this characteristic was dependent on MDM2 DNA transfer. Furthermore, P-a treated with Lipo863 EVs (whose MDM2 levels are lower compared to Lipo246; Figure 2, B and C) displayed impaired proliferation (p≤ 0.03) and migration (p≤ 0.001) compared to Lipo246-EV treatment. To confirm these results, we treated P-a with EVs isolated from pooled DDLPS patient serum (N=8) and compared proliferation and migration versus P-a treated with EVs isolated from pooled healthy donor serum and P-a treated with media alone (N=3; patient and healthy donor clinical information are presented in Supplemental Tables 8 and 9; RT-PCR was used to assess the number of molecules of MDM2 in the EVs of both patient and normal pooled serum; Supplemental Figure 7). As depicted in Figure 4, C and D, incubation for 72h with patient serum-derived EVs significantly increased P-a proliferation and migration compared to cells incubated with EVs derived from normal controls and media alone (p≤0.005 and p≤0.0001 respectively). To mechanistically explain the increase in cellular proliferation following MDM2 transport to the P-a, cell cycle analysis by FACS was performed. As depicted in Figure 4E and Supplemental Table 7, incubation of P-a with Lipo246 EVs induces a three-fold increase in S phase compared to normal medium or depleted medium P-a treatment (p≤0.01). Taken together, our data suggest that DDLPS EV MDM2 can contribute to oncogenic features such as enhanced P-a proliferation and migration.

Figure 4. MDM2 cargo confers oncogenic features in normal recipient P-a.

(A) EVs increase P-a proliferation. When P-a are treated with Lipo246 EVs, for 72h, they show increased proliferation (p≤0.0001) compared to P-a treated with EV depleted medium (Pa+CM) and media alone (P-a). (B) Lipo246 EVs promote migration in recipient P-a. When P-a are treated with Lipo246-EVs they show an increased migration compared to P-a treated with EV depleted medium (Pa+CM) and media alone (P-a). Furthermore, P-a treated with Lipo246-EVs together with MDM2 inhibitor (SAR405838, 0.2 μM), show a decrease migration compared to P-a treated with Lipo246-EVs without drug (p≤0.05). P-a treated with drug alone don’t have a significant change on migration compared to P-a treated with EV depleted medium. P-a treated with Lipo863 EVs (whose MDM2 levels are lower compared to Lipo246) have decresed migration compared to P-a treated with Lipo246 EVs. P-a treated with EVs isolated from a pool of serum derived from DDLPS patients (N=8), have increased proliferation (p≤0.0001) (C) and migration (p≤0.005) (D) compared to P-a treated with EVs isolated from a pool of normal serum (N=3) and compared to P-a treated with media alone. (E) Cell cycle analysis of P-a by FACS. Treatment with Lipo246 EVs induces increase in S phase compared to normal medium or EV-depleted medium. Results are presented as average ± SD. Statistical analysis were performed using t-test. ** 0.001 ≤p ≤0.01; *** ≤ 0.001.

6. Transfer of EV-origin MDM2 DNA promotes production of MMP2 by P-a

EVs contribute to facilitate pre-metastatic niche establishment and maintenance (29). The nearly 60% rate of DDLPS multifocal loco-regional recurrence is remarkably high among all solid tumors. However, the reason for this extremely high rate remains unknown. Several factors, including matrix metalloproteinases (MMPs), have been identified as contributing to pre-metastatic niche formation. Therefore we wanted to further explore the consequences of DDLPS EV interaction with P-a to see if other processes relevant to multi-focal DDLPS loco regional recurrence might so induced.

To verify whether EVs stimulated P-a to release active MMPs, we incubated P-a with Lipo246-EVs and Lipo224-EVs for 96h and then performed gelatin zymography analysis with the resultant CM. We found that active MMP2 (62 kDa) was significantly overproduced in the CM derived from EV-treated P-a versus P-a treated with EV depleted medium (Figure 5). To verify that EV-derived MDM2 was the driver of the enhanced MMP2 activity, we also treated P-a with DDLPS-derived EVs (Lipo246 EVs) in the presence of the MDM2 inhibitor, SAR405838 (0.2 μM for 72h); this treatment strongly impaired the release of active MMP2 (Figure 5). When P-a were treated with EVs isolated from Lipo246 previously treated with GW4869 (a drug that blocks EV secretion; (30)), the release of active MMP2 was impaired. Moreover, when P-a were treated with Lipo863 EVs (whose MDM2 level are lower compared to Lipo246, Figure 2, B and C) active MMP2 release was strongly impaired (Figure 5, A and B). In contrast, when P-a were treated with EVs isolated from Lipo863 transduced with lentiviral particles for the overexpression of MDM2 (see material and methods), active MMP2 release was strongly increased (Figure 5, A and B). Of note, the appearance of an active pro-MMP2 in the zymograms is an artifact of the electrophoretic process per se; this species would be inactive under physiological conditions. SDS-containing zymography buffer results in denaturation of the proteins during electrophoresis. Upon removal of SDS during the zymogram development phase the proteins renature, partially refold and become active. As such both the higher molecular weight pro-MMP2 (72 kDa) band as well as the active MMP2 (62 kDa) band can be visualized for some preparations in the zymograms depicted in Figure 5A. Finally, it is also pertinent to note that gelatin zymography is a semi-quantitative process and is constrained in its detection limits for MMP2; it is possible that other preparations could generate MMP2 activity that is below the threshold of detectability using gelatin zymography.

Figure 5. Uptake of MDM2-EVs by P-a induces MMP-2 activation.

Zymography showing increase of active MMP-2 released by P-a after EV incubation for 96h. MMP-2 activity (active MMP-2, 62 kDa) was significantly enhanced in the medium of Lipo246-EVs and Lipo224-treated P-a compared to P-a treated with EV depleted medium (P-a+CM) and P-a alone (P-a). The treatment with the MDM2 inhibitor SAR405838 (0.2 μM) impaired the release of active MMP-2. When P-a are treated with Lipo246-EVs after incubation of DDLPS cells with GW4869 (a drug that blocks EVs generation), the active MMP-2 released decreases as well as when P-a are treated with EVs isolated from Lipo863 (whose MDM2 level are lower compared to Lipo246). The level of active MMP-2 are rescued when P-a are treated with Lipo863 EVs where MDM2 is overexpressed. Representative images, experiments performed at least 3 times. In B results are presented as average ± SD. Statistical analysis were performed using a one-way ANOVA with Dunnett’s Multiple Comparisons Test. **** p ≤ 0.0001 and ** p ≤ 0.01.

Taken together, our findings support the premise that P-a uptake of DDLPS EV derived MDM2 DNA increases MMP2 secretion in recipient P-a cells, a potential factor contributing to the establishment of multifocal loco-regional pre-metastatic niches, especially given the widespread P-a presence throughout retroperitoneal and abdominal fat-bearing areas of DDLPS patients.

7. EV exposed P-a media promote DDLPS proliferation

After establishing that DDLPS EVs promote an oncogenic phenotype in P-a, we wanted to verify also whether the induction of this oncogenic phenotype, together with MMP2 activation, had any implication regarding disease progression. Therefore, proliferation assays were performed using DDLPS cells treated with media collected from P-a previously exposed to EVs. As shown in Figure 6, when Lipo246 and Lipo224 cells were treated with Lipo246-EV and Lipo224-EV exposed P-a medias for 48h respectively, they showed increased proliferation (p≤ 0.001) compared to untreated Lipo246 and Lipo224 as well as when compared to DDLPS cells incubated with P-a derived CM originating from P-a treated with Lipo246 or Lipo224 EV-depleted media. On the contrary, incubation of Lipo863 with Lipo863-EV exposed P-a media did not change Lipo863 growth, suggesting that the observed increased growth is possibly due to changes induced by DDLPS EV derived from cell lines bearing high levels of MDM2.

Figure 6. EV exposed P-a’s media reciprocally promote DDLPS proliferation.

(A) When Lipo246 and Lipo224 are treated with EV exposed P-a’s media for 48h, they show increased proliferation compared to Lipo246 and Lipo224 treated with normal media, but also compared to each cell line treated with P-a’s exposed media (CM). At the contrary, incubation of Lipo863 with Lipo863-EV exposed P-a’s media has no effect in Lipo863 growth. Results are presented as average ± SD. Statistical analysis were performed using t-test. ** 0.001 ≤p ≤0.01; *** ≤ 0.001.

DISCUSSION

The vast majority of DDLPS contain wt p53 whose tumor suppressor function is impaired by the marked overproduction of MDM2 at both the DNA and at protein level. Highlighting this specific onco-biology, FISH assessment of MDM2 is currently the definitive diagnostic methodology for DDLPS. Almost 60% of retroperitoneal DDLPS ultimately recur as synchronous multifocal tumors, even after initial margin negative resection; this deleterious loco-regional metastatic outgrowth, whose mechanism is unknown, remains the major cause of DDLPS lethality (3,4).

Here we examined the oncobiologic significance of DDLPS EV-derived MDM2 in the circulation, demonstrating that DDLPS patients produce significantly increased amounts of MDM2 DNA in their EVs compared to normal controls. To date, no validated DDLPS patient-associated molecular biomarkers have been identified; consequently, this discovery suggests that circulating EV MDM2 may serve as a biomarker, perhaps informing prognosis and facilitating early detection of DDLPS progression or recurrence, or possibly even predicting therapeutic resistance.

We also demonstrated that DDLPS EV cargo MDM2 could be transferred to recipient P-a (one of the most prominent cells in the DDLPS microenvironment); P-a treated with DDLPS EV-origin MDM2 expressed both increased MDM2 mRNA as well as increased amounts of MDM2 protein in a dose-dependent manner. This discovery is consistent with other studies showing that tumor-secreted EVs, along with their cargos, can be internalized by other cell types in the primary tumor microenvironment as well as in recipient pre-metastatic niche cells where they can exert profound effects (31–33). However, to the best of our knowledge, it has not been previously shown that MDM2 can be released from tumor-derived EVs with subsequent transfer into cells that populate the tumor microenvironment or other recipient normal cells. This study is also one of the first to demonstrate that P-a can serve as potential recipients of EV cargo, and the effect of tumor-secreted EVs on P-a has also apparently not been reported to date. Since MDM2 is amplified in more than forty different types of malignancies including sarcomas, other solid tumors, and leukemias (34), our findings may be relevant to several different malignant diseases. Likewise, other diseases in which TME P-a are prominent (e.g. breast cancer) could possibly have comparable clinically relevant oncobiologies.

After establishing that EVs isolated from DDLPS patient serum and cell lines both contain increased level of MDM2 DNA levels, which can be uptaken by P-a, we showed that EV MDM2 cargo induced MMP2 activity in recipient P-a, a previously not described relationship potentially relevant to pre-metastatic niche formation. MMP2 (together with MMP9) is particularly effective in degrading type IV collagen (35,36) the major structural component of basement membranes, thus facilitating tumor invasion and metastasis. MMP2 has also been studied for its contribution to angiogenesis (37). Importantly, MMP2 have /has been implicated in key processes of pre-metastatic niche development via break down of collagen into peptides that can act as chemoattractant for circulating tumor cells (38). The role of MMP2 has also been suggested in other disseminating diseases (39–44). In the context of liposarcoma, MMP2 and MMP9 expression has been correlated with cell invasiveness (45), metastasis (P = 0.008 and P = 0.005, respectively), and grade (P = 0.001 and P = 0.04 respectively, (46). Among MMP2 and MMP9, we focused on MMP2 because our results with MMP9 were inconsistent and did not achieve significance due to difficulties in MMP9 detection.

MMP2 has been shown to be enhanced by MDM2 in the context of breast cancer (47); not yet in sarcoma, where MDM2 is the key driver, the correlation between MDM2 and MMP2 has never been shown. Moreover, Bradbury et al. (29) describe the regulation between MDM2 and MMP2 to occur within breast cancer cells, whereas we demonstrate that MDM2, as a EV cargo secreted from DDLPS cells, is able to induce MMP2 production in normal P-a, a major TME component.

Our results, summarized in Figure 7, demonstrate that DDLPS EV-origin MDM2 induces P-a production of active MMP2, an initial demonstration of a possible regulatory relationship between tumor-derived EV MDM2 and matrix metalloproteinases in normal TME component cells. We are performing studies focusing on the underlying mechanism of these MDM2-MMP2 interactions, hopefully leading to improved awareness of the genetic controls underlying this process.

Figure 7. Effects of EV-origin MDM2 DNA transfer in recipient P-a.

Our study establish the presence of high level of MDM2 DNA in DDLPS EVs derived from both DDLPS cell lines and patient serum samples. We also show that DDLPS EV MDM2-cargo can be transfer to recipient P-a in the DDLPS TME. This DDLPS EV-origin MDM2 leads to downregulated P-a p53 activity conferring oncogenic feature in normal recipient cells. Importantly, MDM2-cargo promote the release of active MMP-2 in normal recipient P-a thereby possibly contributing to subsequent loco-regional multifocal DDLPS dissemination.

In conclusion, we have established the presence of MDM2 in DDLPS EVs derived from both DDLPS cell lines and also DDLPS patient serum samples. We showed that DDLPS EV MDM2 cargo can be transferred to recipient P-a cells, leading to downregulated P-a p53 activity. Importantly, MDM2 cargo promote release of active MMP2 in normal recipient P-a, thereby possibly contributing to a loco-regional milieu favoring multifocal DDLPS dissemination.

Significance:

Extracellular vesicles derived from dedifferentiated liposarcoma cells induce oncogenic properties in pre-adipocytes.