Adenosine A2B Receptor Promotes Tumor Progression and Metastases in Undifferentiated Pleomorphic Sarcoma

Abstract

Purpose:

Undifferentiated pleomorphic sarcoma (UPS) is an aggressive subtype of soft tissue sarcoma with poor outcomes, particularly in metastatic cases. The mechanisms driving metastasis in UPS remain poorly understood, limiting therapeutic advances.

Experimental Design:

A multi-omics approach was used to analyze paired primary and metastatic UPS tumor samples. Spatial transcriptomics, bulk RNA sequencing, and deconvolution analyses were performed to identify molecular pathways and immune microenvironment alterations associated with metastasis. Functional assays using CRISPR-Cas9 knockout (KO) UPS cell lines, alongside in vivo models, were used for functional validation experiments.

Results:

Transcriptomic analyses on 13 patients with UPS revealed significant upregulation of hypoxia, epithelial–mesenchymal transition, and immune-suppressive pathways in metastatic UPS. ADORA2B was identified as a key driver of these processes, with elevated expression correlating with poor disease-free survival in patients with UPS. Functional studies confirmed that ADORA2B promotes proliferation, migration, invasion, and matrix remodeling via metalloprotease regulation. In vivo, ADORA2B KO reduced primary tumor growth and metastatic dissemination in UPS models.

Conclusions:

This study identifies ADORA2B as a critical regulator of metastatic progression in UPS, implicating it as a promising therapeutic target. Ongoing clinical trials targeting adenosine pathways further support the translational potential of ADORA2B inhibition to disrupt metastasis and improve outcomes for patients with UPS.

Translational Relevance.

Our findings not only identify ADORA2B as a critical regulator of metastasis-associated pathways but also suggest its potential as a promising therapeutic target in this aggressive cancer subtype.

Introduction

Undifferentiated pleomorphic sarcoma (UPS) is one of the most common and aggressive subtypes of soft tissue sarcomas (STS), accounting for approximately 16% of cases (1). Initially identified in 1964 as malignant fibrous histiocytoma (2), UPS is defined by a lack of specific differentiation and a pleomorphic histologic pattern (3). Today, UPS is primarily a diagnosis of exclusion, in which a tumor is classified as UPS when no specific lineage differentiation can be determined after comprehensive histopathologic and molecular analyses (4, 5).

Most patients with UPS are diagnosed at a localized stage; however, despite optimal treatment strategies, approximately 40% of those with high-grade disease will develop distant metastases. Once metastases occur, the prognosis is poor, with a median overall survival of 20 months and a 5-year survival rate of about 30% (6). First-line treatment options, including anthracycline-based regimens, provide only limited benefit, yielding a median progression-free survival of 6 months (7), and other treatments like ifosfamide, gemcitabine, and pazopanib have demonstrated modest efficacy at best (8).

Despite advances in the understanding of sarcomas, the mechanisms driving metastatic progression in UPS remain poorly understood, creating a significant gap in the development of effective therapeutic strategies aimed at preventing metastasis. Metastatic dissemination remains the primary cause of mortality in patients with UPS, and the biological processes behind this progression need to be fully elucidated to develop new therapeutic interventions.

In this study, we aim to address this knowledge gap by conducting a multi-omics analysis of paired primary and metastatic UPS tumor samples. By examining the molecular differences between these two stages, we seek to identify key pathways and potential therapeutic targets involved in the metastatic process. Our ultimate goal is to contribute to the development of new, more effective treatments for UPS by unveiling the mechanisms that promote metastasis.

Materials and Methods

Patients and clinical sample processing

For RNA sequencing (RNA-seq) and immunohistochemistry analyses, we first interrogated our database of patients with STS treated at the Bergonié Institute from 1990 to 2020. We selected patients for whom histological samples from both primary and metastatic tumors were available, after obtaining written informed consent and Institutional Review Board (IRB) approval. Following these criteria, we identified 70 patients with STS with matched primary and metastatic tumor samples (UPS: n = 20, leiomyosarcomas (LMS): n = 17, liposarcomas (LPS): n = 19, others: n = 14). All histologic samples were reviewed by a single expert pathologist blinded to sample identity. To ensure homogeneity, we stratified patients by histologic subtype and focused on patients with UPS for this study.

After quality control of the sequencing data, we retained RNA-seq data from 11 primary UPS tumors and 8 metastases (1 liver and 7 lung metastases) derived from 13 patients with UPS. The clinical characteristics of these patients are presented in Table 1.

Table 1.

Clinical characteristics of patients with UPS for whom we sequenced primary and/or metastatic tumors.

Variable	N. of patients	% of patients
Age	​	​
<65 years	4	30.8
≥65 years	9	69.2
Sex	​	​
Male	7	53.8
Female	6	46.2
Topography	​	​
Superficial	1	7.7
Deep	12	92.3
Localization	​	​
Upper limb	2	15.4
Lower limb	8	61.5
Trunk wall	2	15.4
Abdominal	1	7.7
Size	​	​
<5 cm	4	30.8
5–10 cm	4	30.8
>10 cm	5	38.4
Grade FNCLCC	​	​
1	0	0
2	3	23.1
3	10	76.9

Spatial transcriptomics and deconvolution analysis

For spatial transcriptomic analysis, we used an independent cohort of 4 patients with coupled primary and metastatic samples. Tissue sections (5 µm) were prepared from formalin-fixed paraffin-embedded (FFPE) blocks and processed for the NanoString Whole Transcriptome Atlas (WTA) assay, covering 18,695 protein-coding genes, following the manufacturer's instructions (NanoString, MAN-10150-03). The tissue was prepared using standard paraffin removal, rehydration, and heat-induced epitope retrieval (CC1 solution, 32 minutes at 99°C) on a Ventana Discovery instrument Ventana Discovery Ultra (RRID:SCR_025097). Enzymatic digestion was performed using 1 μg/mL proteinase K for 16 minutes at 37°C, followed by overnight hybridization with the human WTA probe mix.

Six to eleven regions of interest (ROI), depending on sample size, were selected within areas containing both tumor and stromal cells on the GeoMx platform (NanoString GeoMx Digital Spatial Profiler (RRID:SCR_021660) and were segmented further using CD45 staining to define areas of illumination (AOI) to separate tumor (CD45-) and stromal (CD45⁺) compartments (Supplementary Table S1). RNA-specific oligonucleotide tags were released via UV light and collected for each AOI. Libraries were constructed, amplified using Illumina i5 and i7 dual-indexing primers, purified using AMPure XP beads, and assessed for concentration [Thermo Fisher Qubit fluorimeter (RRID:SCR_018095)] and quality [Agilent 4150 TapeStation System (RRID:SCR_019393)]. The sequencing was performed on the Illumina NovaSeq6000 platform (RRID:SCR_016387). One patient was excluded from subsequent analyses because of outlier. The resulting fastq files were processed using NanoString’s DND pipeline, and raw count data were normalized using the GeomxTools R package [NanoString GeoMx Tools (RRID:SCR_023424)]. Differential gene expression was analyzed using LIMMA in R (RRID:SCR_010943), and cell abundance was estimated using SpatialDecon (safeTME matrix). The list of genes differentially expressed in tumor and stroma ROIs of patients with UPS between primitive and metastasis tumor are listed in Supplementary Tables S2 and S3.

Deconvolution analysis estimating the immune cell composition of analyzed segments was conducted using the Spatial Decon R package. Differences between primary and metastatic tumors were then calculated using the Wilcoxon test.

Multiplex immunohistofluorescence

Multiplex immunohistofluorescence (mIHF) was performed on an independent cohort of 5 matched primary and metastatic UPS samples. mIHF staining was performed using a 7-plex protocol on the Ventana Discovery platform (RRID:SCR_025097). Primary antibodies included (Agilent cat. #M7103, RRID:AB_2075537), CD68 (Agilent cat. #M0876, RRID:AB_2074844), CD14 (Cell Marque cat. #114R, RRID:AB_2827391), cMAF (Abcam cat. #ab225416, RRID:AB_2868542), CD20 (Ventana Medical Systems cat. #760-2531, RRID:AB_2335956), and CD45 (Cell Marque cat. #145M-96, RRID:AB_1158594). Detection was achieved using OmniMap anti-rabbit and anti-mouse horseradish peroxidase kits (Ventana), followed by TSA Opal fluorophores (Opal 480, 520, 570, 620, 690, and 780, Akoya Bioscience). Slides were counterstained with spectral DAPI (Akoya), scanned using the PhenoImager HT system (Akoya), and images unmixed using spectral libraries from monoplex-stained images (inForm Advanced Image Analysis software; RRID:SCR_019155). Tumor areas were defined using PhenoChart, and cell segmentation was conducted with inForm software, using DAPI and membrane markers to define cell boundaries. Signal intensities were normalized using the GaussNorm function (flowstat R package; RRID:SCR_000399).

RNA-seq

RNA-seq was performed on an independent cohort of paired primary and metastatic samples of 13 patients with UPS. RNA was extracted from frozen or FFPE samples using the QIAamp RNeasy Mini Kit (Qiagen) or Maxwell RSC RNA FFPE Kit (RRID:SCR_025867), respectively, as per manufacturer protocols. RNA purity and concentration were assessed using the NanoDrop spectrophotometer [Thermo Scientific (RRID:SCR_018042)]. RNA-seq was performed by Novogene on the NovaSeq6000 S2 platform [Illumina, (RRID:SCR_016387)] using poly-A capture and a paired-end 150×2 reads protocol. Sequences were quality-controlled using standard tools, and curated reads were obtained for transcriptome analysis. Transcript counts were normalized using the VOOM method, and differential gene expression was assessed using LIMMA (RRID:SCR_010943). Significantly upregulated/downregulated transcripts were defined by an FDR <0.05 and a fold change >2. Gene set enrichment analysis was conducted using Molecular Signatures Database (RRID:SCR_016863) and FGSEA (RRID:SCR_020938).

Real-time quantitative PCR analysis

Total RNA was extracted from cell lines using the RNeasy Plus Mini Kit (Qiagen) according to the manufacturer’s instructions. Reverse transcription of RNA into cDNA was performed using the High-Capacity cDNA Reverse Transcription Kit (Applied Biosystems). To quantify gene expression, real-time quantitative PCR (RT-qPCR) was performed using the Power SYBR Green PCR Master Mix (Applied Biosystems) and specific primers for MMP1, MMP2, and VEGFA (sequences provided in Supplementary Table S4). Reactions were conducted in triplicate on a StepOnePlus Real-Time PCR System (RRID:SCR_015805; Applied Biosystems). The expression of target genes was normalized to ACTB (β-actin) expression, and the 2-ΔΔCt method was used to calculate relative gene expression.

Cell culture and generation of knockout cells

UPS cell lines (IB144 and IB202) were derived from human UPS surgical specimens, with patient consent and IRB approval, as previously described (9). The cells were grown in RPMI 1640 medium supplemented with 10% fetal bovine serum and 1% penicillin/streptomycin at 37°C in a humidified 5% CO₂ atmosphere. Cell lines were maintained between passages 25 and 60, with genomic integrity confirmed by array-comparative genomic hybridization every 10 passages.

For knockout (KO) generation, multi-guide single-guide RNAs (sgRNA) targeting ADORA2B (sgRNA #1: AGC​UGG​CCC​GGC​CAU​GCU​GC; sgRNA #2: GGU​CAU​CGC​CGC​GCU​UUC​GG; sgRNA #3: GAC​GCC​CAC​CAA​CUA​CUU​CC), along with spCas9 2NLS nuclease, were purchased from Synthego. Cells were transfected with sgRNA-Cas9 ribonucleoprotein complexes using Lipofectamine CRISPRMAX (Thermo Fisher). After 7 days of growth, cells were clonally expanded, and KO clones were confirmed by PCR and Sanger sequencing. For pharmacologic inhibition, M1069, a dual inhibitor of ADORA2A/2B purchased from MedChemExpress (MCE; cat. #M1069).

Proliferation and wound healing assays

Proliferation was measured by seeding 1,000 cells/well in six-well plates, and cells were counted at days 3, 7, and 9 using flow cytometry [MACSQuant10, Miltenyi Biotec (RRID:SCR_008984)]. For wound healing assays, 300,000 cells/well were seeded in 12-well plates, and a scratch was made in the cell monolayer. Images were captured at 0 and 24 hours (IB144) or 18 hours (IB202). Wound area was quantified using the ImageJ (RRID:SCR_003070) plugin “Wound_healing_size_tool.”

Invasion assays

Invasion was assessed using Matrigel-coated transwell inserts (Corning BiocatTM). Cells (5 × 10⁴) were seeded into the upper chamber, and complete medium (with 10% FBS) was added to the lower chamber as a chemoattractant. After 24 hours of incubation at 37°C, noninvading cells were removed, and invading cells were fixed, stained with crystal violet, and counted.

Zymography

Gelatin zymography was used to evaluate MMP2 activity. Protein supernatants (2 µg) were electrophoresed on 7.5% acrylamide gels containing 1 mg/mL gelatin, followed by incubation in reaction buffer (Triton X-100, Tris-HCl, CaCl2, ZnCl2) for 24 hours at 37°C. Gels were stained with Coomassie blue, and MMP2 activity was quantified using ImageJ (RRID:SCR_003070).

Animal studies

Animal experiments were performed in compliance with French and European guidelines for animal experimentation (R.214-87 to R.214-126 and 2010/63/UE). Approval was obtained from the institutional animal use and care committee (APAFIS #38231-2022071215405233). IB144 and IB202 cells [wild-type (WT) or KO)] were injected orthotopically or intravenously into Rag2-/- γc-/- mice (n = 10 per condition). Tumor growth and metastasis were onitored weekly by bioluminescence imaging (Photon IMAGER, Biospace Lab). Mice received intraperitoneal injections of D-Luciferin (75 mg/kg) 10 minutes before imaging. When study endpoints were reached, tumors and metastatic tissues were harvested for further analysis. Growth curves were analyzed using two-way ANOVA with Sidak multiple comparisons test, and survival curves were assessed with the log-rank (Mantel–Cox) test using GraphPad Prism (RRID:SCR_002798).

Prognostic analysis and survival outcomes in STS

For survival analysis, gene expression and clinical data from STS primary tumor samples were obtained from different databases. We first interrogated the institutional database of Predictive Oncology laboratory collected from multiple public datasets (10), including 1,432 STS primary tumor samples, notably 330 UPS – 199 with informed disease-free survival (DFS). From the UCSC Xena (RRID:SCR_018938) transcriptome database, which includes the normal samples of the Genotype-Tissue Expression (RRID:SCR_013042) project, The Cancer Genome Atlas (RRID:SCR_003193) tumor samples used for normalization purpose, we extracted 1,795 normal connective tissue samples (adipose tissue, muscles, nerves, and blood vessels). The data sets of origin for STS and normal samples are listed in Supplementary Table S5, and the clinicopathologic characteristics of patients with STS are summarized in Suppl. Table 6. The preanalytic processing was done as previously described (10). The prognostic impact of ADORA2B gene expression was analyzed. Kaplan–Meier survival curves were generated from the time of initial diagnosis to the disease relapse (DFS) or the most recent follow-up. Statistical significance was assessed using the log-rank test, and P values <0.05 were considered significant for survival differences between patient subgroups. All statistical analyses were performed using GraphPad Prism software (RRID:SCR_002798).

Results

Molecular and immune landscape of primary and metastatic UPS reveals ADORA2B as a potential driver of metastatic progression

To investigate the molecular and immune differences between primary and metastatic UPS, we conducted both spatial transcriptomics and bulk RNA-seq to gain a comprehensive view of the gene expression alterations associated with metastasis.

Using the GeoMX Digital Spatial Profiler with the WTA assay (see Methods), we quantified gene expression within distinct ROIs, enabling a spatially resolved analysis of more than 18,000 genes across both the tumor and stromal compartments (Fig. 1A). This analysis was conducted on three paired primary tumor and metastatic samples (one patient was excluded because of outlier) in which sufficient material was available, allowing for a comprehensive examination of gene expression differences between these two stages of disease. In metastatic samples, we observed upregulation of key pathways related to hypoxia, glycolysis, and epithelial–mesenchymal transition (EMT), suggesting their involvement in metastatic progression (Fig. 1B and C).

Figure 1.

Molecular and immune landscape of primary and metastatic UPS. A, GeoMX workflow, with representative picture of ROI and AOI selection. B and C, ROI gene set enrichment in tumor (B) and stromal compartment (C) on 3 paired primary and metastatic tumors using the GeoMX DSP for the WTA assay. Upregulated or downregulated pathways in metastases were identified using MsigDB and FGSEA. D, Deconvolution analysis estimating difference in the immune and endothelial cell composition between primary and metastatic tumors, using the SpatialDecon R package. E, Immune cell enrichment in primary tumor compared with matched metastasis in mIHF, using the Ventana Discovery platform. F, CD8⁺ representative images of mIHF unmixed by inForm Advanced Image Analysis software [nucleus staining DAPI (blue), membrane marker CD8 (green)].

We also analyzed the tumor microenvironment through deconvolution analysis, revealing the enrichment of endothelial cells in metastases, indicating increased angiogenesis compared with primary tumors (Fig. 1D). Furthermore, the analysis of immune microenvironment showed that primary tumors had significantly higher infiltration of CD8⁺ T cells, NK cells, and memory B cells, particularly in the stromal regions. In contrast, metastatic samples exhibited a substantial reduction in these immune populations, indicative of an immune-suppressive microenvironment (Fig. 1D). This was further validated by mIHF on an independent cohort of five coupled primary and metastatic cases, confirming reduced infiltration of CD8⁺ T cells in metastases, although the cohort size limited statistical significance (Fig. 1E and F).

To complement these findings, we performed bulk RNA-seq on an independent cohort of paired primary and metastatic tumor samples from an independent cohort of 13 patients to capture global transcriptomic changes. This revealed an enrichment of pathways associated with metastasis, including hypoxia, xenobiotic metabolism, TNF-α signaling via NF-kB, inflammatory response, and complement activation (Fig. 2A). These results, aligned with those from spatial transcriptomic analysis, reinforced the critical role these pathways play in UPS metastases.

Figure 2.

Comparative analysis of primary tumors versus metastases in patients with UPS. A, Gene set enrichment analysis of RNA-seq in B. MsigDB and FGSEA were used to identify upregulated or downregulated pathways in metastatic samples. B, Volcano plot of differential gene expression analysis after RNA-seq of 11 primary tumors versus 8 metastatic samples from patients with UPS. A total of 690 genes were found significantly upregulated (red) or downregulated (blue) in metastases (FDR threshold of 0.05 and fold change of 2). C, Box plots of ADORA2B gene expression in STS, UPS, LMS, and LPS. D, Kaplan–Meier curves of DFS according to ADORA2B gene expression in 199 patients with UPS. E, Immune cell composition of patient samples was estimated using ConsensusTME that benchmarks different deconvolution algorithms (Bindea, CIBERSORT, xCell). Differences between conditions were then calculated using paired t tests (**, P < 0.01; *, P < 0.1).

The differential gene expression analysis identified 690 genes significantly altered between primary and metastatic tumors (Fig. 2B). Among the genes overexpressed in metastases, we prioritized those already known in literature to be implicated in metastatic processes of other solid tumors and with potential as therapeutic targets. ADORA2B emerged as one of the key candidates, owing to its druggability and the fact that inhibitors targeting it are currently in clinical development. This makes ADORA2B a highly promising candidate for therapeutic intervention in UPS.

ADORA2B, a G protein–coupled receptor involved in stimulating adenylate cyclase activity, has been previously implicated in the metastatic progression of several epithelial cancers (11–14), but its role in sarcomas, particularly in UPS, had not been explored until now.

Analyzing the database of the Predictive Oncology laboratory, we found that ADORA2B is significantly overexpressed in STS as compared with normal connective tissues (P < 0.001; Fig. 2C), with heterogeneous expression within a range of >15 unites in log₂ scale. Expression was higher in UPS samples than in LMS samples (P = 0.001) and LPS samples (P = 0.045; Fig. 2C). Moreover, we evaluated the prognostic impact of ADORA2B overexpression in a cohort of 199 patients operated for UPS, defining overexpression as expression above the median expression level in the 1,432 STS. This ADORA2B expression cutoff, not based upon disease-free survival (DFS) data, allowed to avoid any overfitting in the prognostic analysis. Interestingly, in this analysis, ADORA2B overexpression was associated with shorter DFS, with a 61% 5-year DFS (95% CI, 52–72) in the ADORA2B-high subgroup versus 72% (95% CI, 61–85) in the ADORA2B-low subgroup (P = 0.029; Fig. 2D).

Finally, deconvolution analysis of the RNA-seq data from metastatic UPS samples, stratified by ADORA2B expression, revealed that ADORA2B-low samples exhibited a trend toward immune signature enrichment, suggesting a role for ADORA2B in creating an immune-suppressive microenvironment in metastatic UPS (Fig. 2E).

All these findings position ADORA2B as a promising target for therapeutic intervention in UPS, with the potential to disrupt metastatic progression and modulate the immune microenvironment.

ADORA2B KO reveals its critical role in metastasis-associated pathways in UPS

To investigate the role of ADORA2B in UPS progression, we utilized two patient-derived UPS cell lines, IB144 and IB202, and generated ADORA2B-KO models using CRISPR-Cas9 technology. KO efficiency was confirmed by PCR (Fig. 3A) and Sanger sequencing, which revealed large deletions in the target region, leading to a nonfunctional protein (Supplementary Fig. S1A and S1B).

Figure 3.

Effects of ADORA2B KO in UPS cells. A, PCR on genomic DNA of IB144 and IB202 WT/ADORA2B-KO (sgA2B) cells with specific primer flanking ADORA2B exon 1. The lower bands in the sgA2B conditions show a deletion in exon 1 following the CRISPR lipofection of cells confirmed by Sanger sequencing in Supplementary Fig. S1A and S1B. B, Pathways enrichment analysis in WT cell lines versus ADORA2B-KO (sgA2B) cell lines using FGSEA after RNA-seq. Three independent RNA extractions of each cell line were used. WT conditions (n = 6) were pooled [IB144 WT (n = 3) + IB202 WT (n = 3)] as well as sgA2B conditions (n = 6). C, Proliferation assay of WT versus sgA2B cells for IB144 and IB202 cell lines (n = 3; ****, P < 0.0001; two-way ANOVA; Tukey multiple comparisons test). D, Representative pictures of wound healing assay showing the migration capacity of IB144 and IB202 WT versus sgA2B cells. Histograms represent the percentage of the wound closure relative to the WT condition (n = 3; **, P < 0.01; t test). E, Representative pictures of invasive cells after 24 hours of incubation in Boyden chambers. Histograms represent the percentage of invasive cells relative to the WT condition (n = 3; *, P < 0.05; t test). F, RT-qPCR of MMP1, MMP2, and VEGFA genes in IB144 and IB202 WT/sgA2B cells (n = 3 independent RNA extractions; ****, P < 0.0001; ***, P < 0.001; t test). G, Representative zymogram gel of MMP2 activity in IB144 WT and sgA2B cells. Histograms represent the densitometric quantitation of 3 independent experiments (n = 3; **, P < 0.01; t test). H, Representative Western blot of MMP2 protein in IB144 and IB202 WT and sgA2B cells. Histograms represent the densitometric quantitation of three independent experiments (n = 3, t test).

Then, we performed RNA-seq on both the parental and KO models to identify differentially expressed pathways influenced by the loss of ADORA2B. In the ADORA2B-KO cells, we observed downregulation of key pathways involved in metastasis, including the EMT pathway, as well as inflammatory pathways mediated by IFN-α and -γ responses, TNF-α signaling via NF-kB, apoptosis, and angiogenesis pathways, which are implicated in cancer cell migration and survival in distant tissues (Fig. 3B).

ADORA2B KO reduces proliferation, migration, and invasion in UPS models

To further investigate the role of ADORA2B in UPS progression, we utilized our two ADORA2B-KO UPS cell line models to test the three primary mechanisms involved in metastasis: proliferation, migration, and invasion. KO of the ADORA2B gene significantly reduced the proliferation capacity in both cell lines (Fig. 3C). The scratch assay further demonstrated that ADORA2B contributes to cell migration, as the KO models exhibited reduced migration compared with the parental line, with the reduction being statistically significant in the IB144 model (Fig. 3D). Additionally, the transwell invasion assay demonstrated that ADORA2B-KO cells had a diminished ability to invade the Matrigel-coated membrane in the IB202 model, although this effect was not observed in the IB144 line (Fig. 3E). These findings suggest that distinct mechanisms may drive the prometastatic behavior in each cell line, confirming ADORA2B's involvement in key processes associated with metastatic progression. This prompted us to further explore the underlying mechanisms.

Previous studies have implicated the regulation of metalloproteases and angiogenesis in ADORA2B’s effects on tumor progression (12, 15). Our transcriptomic analysis of the UPS models supported this, showing that ADORA2B-KO cells significantly underexpressed metalloprotease 2 (MMP2), and in the IB144 model, also MMP1, which was not expressed in the IB202 parental line. These findings were validated through RT-qPCR, confirming reduced MMP1 and MMP2 expression in the IB144 ADORA2B-KO model. Whereas MMP1 was absent in the IB202 parental line, the same trend in MMP2 expression was observed in the KO cells (Fig. 3F). Moreover, zymography assays and Western blot revealed reduced MMP2 activity in the IB144-KO model (Fig. 3G and H).

We also explored the angiogenesis pathway by analyzing the vascular endothelial growth factor (VEGF) expression in the KO cell lines. Although RNA-seq data and literature evidence from epithelial tumors suggest reduced VEGF expression, the difference in VEGF levels between KO and parental lines was not statistically significant in our models (Fig. 3F).

In vivo models confirm ADORA2B’s role in tumor growth and metastasis in UPS

Our previous findings prompted us to conduct in vivo experiments to confirm the impact of ADORA2B on UPS progression and metastasis. We utilized two models based on ADORA2B-KO IB144 and IB202 cells.

In the first model, we established a patient-derived orthotopic xenograft model by injecting both parental and KO IB144 and IB202 cells orthotopically into Rag2-/- γc-/- mice. In this model, tumors formed from ADORA2B-KO cells exhibited halted proliferation, resulting in significantly smaller and nonproliferating tumors compared with those derived from ADORA2B WT cells, which showed exponential tumor growth in both the IB144 and IB202 lines (Fig. 4A, B, E, and F). Moreover, in the IB144 WT model, 8 of 10 mice developed distant metastases, with 5 mice exhibiting lung metastases and 3 mice also showing liver metastases. In contrast, only 3 of 10 mice with ADORA2B-KO tumors developed lung micrometastases (Supplementary Fig. S2A).

Figure 4.

Antitumoral and antimetastatic activities of ADORA2B KO. A, IB144 WT and sgA2B cells (expressing luciferase reporter gene) were xenografted orthotopically into the lower limb of Rag2^−/− γc^−/− mice. Tumor growth was monitored once a week by bioluminescence imaging (n = 10 mice per condition; ****, P < 0.0001; two-way ANOVA; Sidak multiple comparisons test). RLU, relative luminescence unit. B, Images of mice in A with bioluminescent signal taken on day 68. C, IB144 WT or sgA2B cells (expressing luciferase reporter gene) were injected directly into the tail vein of Rag2^−/− γc^−/− mice (n = 10 mice per condition). Metastases were monitored once a week by bioluminescence imaging. When the endpoints were reached, mice were euthanized and survival curve was analyzed [log-rank (Mantel–Cox) test]. D, Images of mice in C with bioluminescent signal taken on day 89. The two missing mice in the WT condition had already reached the endpoints by day 89. E and G, IB202 WT and sgA2B cells were xenografted orthotopically (E) or injected into the tail vein (G) of Rag2^−/− γc^−/− mice as for IB144 model. Tumor growth and metastases were monitored once a week by bioluminescence. F, Images of mice in E with bioluminescent signal taken on day 77. H, Images of mice in G with bioluminescent signal taken on day 77.

In the second model, we used a forced metastasis model by injecting tumor cells intravenously via the tail vein. Both IB144 and IB202 cells were capable of producing lung and liver metastases in Rag2-/- γc-/- mice. However, ADORA2B-KO cells resulted in significantly fewer metastases compared with WT cells, translating into improved survival rates. In the KO arm, 100% of the mice survived, compared with 0% in the control (WT) arm for both cell lines (Fig. 4C, D, G and H). Additionally, in ADORA2B WT models, metastases appeared earlier and were significantly larger compared with those in the KO models (Supplementary Fig. S2B and S2C).

These results demonstrate that ADORA2B plays a key role in both primary tumor growth and metastatic spread in UPS, making it a potential target for therapeutic intervention.

The pharmacologic inhibitor also reduces proliferation and invasion in UPS models

To confirm the results obtained with CRISPR-Cas9 and validate the druggability of our target, we used a pharmacologic inhibitor. For this purpose, we obtained the M1069, a dual ADORA2A/ADORA2B inhibitor (MCE; cat. #M1069).This compound is of particular interest as it is already being evaluated in early-phase clinical trials in solid tumors.

We repeated our in vitro proliferation and invasion assays and confirmed the efficacy of M1069 not only in the UPS patient-derived cell lines IB144 and IB202 (previously used for CRISPR-Cas9 KO) but also in two additional patient-derived lines, the JR588 and KN473 (Fig. 5A and B).

Figure 5.

Antitumoral activity of M1069, a dual inhibitor ADORA2A/2B. A, Proliferation assay of four UPS cell lines treated with or without M1069 for 10 days (n = 3, two-way ANOVA, Sidak multiple comparisons test). M1069 treatment was renewed every 3 days. B, Representative pictures of invasive cells in Boyden chambers treated or not with M1069 at 100 µmol/L for 24 hours. Histograms represent the percentage of invasive cells relative to the DMSO condition (n = 3, t test). ****, P < 0.0001; ***, P < 0.001; **, P < 0.01; *, P < 0.05.

Discussion

We present the largest investigation to date using high-throughput analysis of patient samples to identify genes potentially involved in the metastatic progression of UPS. This comprehensive database represents a valuable resource for exploring and validating additional therapeutic targets, with the potential to enhance the current treatment arsenal and advance personalized care for patients with UPS, particularly in this challenging and rare disease.

UPS is one of the most common, highly aggressive and less characterized histologic types of STS (1), with a metastatic evolution in nearly half patients despite optimal treatment in localized stages. Although primary UPS characteristics are less known, metastases, responsible for the high lethality of this histology, are even less known. In this work, we wanted to explore key factors regulating UPS metastases, which seem necessary to achieve new targets to potentially improve the therapeutic arsenal.

Our findings from comparative analyses of transcriptomic data of coupled primary and metastatic UPS samples showed that effectively some differences between primary and metastatic tumors exist and are responsible of biological processes involved in tumor evolution; indeed, metastases overexpressed genes involved in inflammation, hypoxia, and EMT pathways. Even if EMT is not a process described in sarcomas due to their mesenchymal origin, the genes responsible of EMT overexpression were mainly implicated in cell migration, which is coherent with the metastatic context. Cross-referencing the literature and transcriptomic analysis, we focused our attention on potentially druggable genes showing a real interest in metastatic process and with a prognostic value in sarcoma, notably ADORA2B.

Although ADORA2B was not the most highly overexpressed gene identified in our differential expression analysis of metastatic versus nonmetastatic UPS tumors, it was prioritized based on its well-documented role in promoting metastatic progression in various solid tumors and, importantly, the availability of pharmacological inhibitors currently in early-phase clinical trials. This combination of biological relevance and clinical feasibility makes ADORA2B a compelling target with high translational potential. In the context of sarcomas, where treatment options remain limited and patient outcomes for metastatic disease are poor, identifying targets that can be rapidly moved into clinical testing is a major advantage. While the novelty of ADORA2B as a therapeutic target may be reduced by ongoing studies in other cancer types, its potential impact in UPS justifies its investigation in this setting.

Most solid tumors grow in hypoxic conditions (16), and previous studies have shown a link between hypoxia and ADORA2B expression (17). Hypoxia stimulates angiogenesis by increasing ADORA2B expression (18). Overexpression of ADORA2B in various epithelial tumors suggests that this adenosine receptor plays a significant role in cancer pathogenicity and progression. For instance, ADORA2B expression in oral cancer and in lung adenocarcinoma positively correlates with tumor stage (17, 19); in patients with breast cancer, it represents a negative prognostic factor (20); whereas in prostate cancer, the activation of this receptor supports tumor cell survival (21).

There is growing evidence that adenosine receptors are attractive therapeutic targets for several conditions, including certain cancer types (22). Even more, analyzing a very large sarcoma database, we highlighted for the first time that ADORA2B is overexpressed in sarcomas, notably in UPS, also having a prognostic value.

Our gene expression analysis and KO experiments indicate that ADORA2B plays a critical role in promoting metastatic and progression-related processes in UPS. The observed reduction in EMT and angiogenesis pathway activity in ADORA2B-KO cells, both essential for cancer cell migration and survival in distant tissues, underscores ADORA2B's potential involvement in driving the aggressive metastatic behavior of UPS cells. These findings further support the hypothesis that targeting ADORA2B could disrupt multiple mechanisms facilitating UPS metastasis.

To validate the role of ADORA2B in the metastatic process and assess its potential as a therapeutic target, we conducted functional test. We confirmed that ADORA2B contributes to the proliferative, migratory, and invasive activity of UPS cells, by modulating metalloproteases expression and function for matrix remodeling, as well as angiogenesis for tumor progression. We observed in vivo, in two different UPS models, that ADORA2B depletion inhibited primary tumor growth and considerably reduced metastasis rate. Even more, the pharmacologic inhibition by the M1069 confirmed the interest and the druggability of this target.

In conclusion, a model emerges from our study, in which ADORA2B acts as regulator of several steps of the metastatic cascade in UPS cells, paving the way for clinical studies exploring pharmacologic inhibition of adenosine receptors in patients with UPS.

One limitation of our study is the use of immunocompromised mouse models, which precludes the evaluation of immune-mediated effects of ADORA2B KO. This is particularly relevant in light of our transcriptomic data, which revealed that many of the most downregulated genes in ADORA2B-KO tumor cells are immune-related (Fig. 3B). Furthermore, analysis of patient tumor samples showed a significant correlation between high ADORA2B expression and immunosuppressive gene signatures (Fig. 2G), suggesting that ADORA2B may play a key role in shaping the tumor-immune microenvironment.

Whereas our current work focused on delineating the tumor-intrinsic roles of ADORA2B using human-derived xenograft models, future studies using immunocompetent murine models, such as the KrasG12D; Trp53−/− (KP) UPS model (23), will be critical to assess how genetic or pharmacologic inhibition of ADORA2B influences immune cell infiltration and function. Such models will provide a more comprehensive understanding of the immunomodulatory role of ADORA2B and its potential as a therapeutic target in combination with immunotherapies.

Currently, approximately 106 active clinical trials are targeting adenosine or its receptors, including ADORA2A, ADORA2B, and CD73 inhibitors. These trials primarily aim to explore how adenosine pathway inhibitors can enhance the effectiveness of existing therapies, particularly immune checkpoint inhibitors. Among these, around 10 clinical trials are specifically evaluating ADORA2B inhibitors, focusing on their therapeutic potential in solid tumors and their effects on tumor progression and immune responses. Notably, the CURATE trial (NCT05272709), a phase Ib/II study of the ADORA2B receptor antagonist TT-702, seeks to determine its efficacy in treating advanced solid tumors. This trial underscores the increasing interest in targeting ADORA2B to enhance antitumor immunity by mitigating the immunosuppressive effects of elevated adenosine levels in cancer. Together, these efforts highlight the therapeutic promise of ADORA2B inhibition, further supporting its potential as a target for disrupting the metastatic progression of UPS and improving outcomes in this rare and aggressive cancer.

Supplementary Material

Supplementary Table S1, S2, S3

Table S1, S2, S3

Supplementary Table S4

Table S4

Supplementary Table S5, S6

Table S5, S6

Supplementary Table S7

Table S7

Figure S1

Figure S1

Figure S2

Figure S2

Data Availability

The data generated in this study are available in the Supplementary Information. Raw data (including full blots and source data for graphs) are available from the corresponding author upon reasonable request. Availability of cell lines and PDX models should be discussed with the corresponding author and the Bergonié Institute Biological Resource Center.

Authors’ Disclosures

M. Spalato Ceruso reports other support from Deciphera outside the submitted work; this work was carried out in collaboration with Explicyte and the INSERM Unit U1312 BRIC, within the framework of the CONDOR Program. J.-P. Guégan reports employment with Explicyte. A. Bessede reports other support from Explicyte during the conduct of the study. S. Verbeke reports nonfinancial support from Explicyte during the conduct of the study and grants and nonfinancial support from Merck Healthcare KGaA and Xenothera outside the submitted work. A. Italiano reports grants from Bayer, MSD, Merck, Roche, Bristol Myers Squibb, and AstraZeneca outside the submitted work. No disclosures were reported by the other authors.

Authors’ Contributions

M. Spalato Ceruso: Conceptualization, formal analysis, supervision, validation, investigation, visualization, methodology, writing–original draft, writing–review and editing. J.-P. Guégan: Data curation, formal analysis, visualization, methodology. A. Bourdon: Data curation, formal analysis, visualization, methodology. V. Chaire: Investigation, methodology. Y. Valverde Timana: Investigation, methodology. A. Giese: Investigation, methodology. C. Rey: Investigation, methodology. A. Bessede: Resources, data curation, formal analysis. R. Perret: Formal analysis. L. Vanhersecke: Formal analysis. V. Velasco: Methodology. P. Finetti: Data curation, formal analysis, visualization. F. Bertucci: Formal analysis, investigation, visualization. S. Verbeke: Conceptualization, formal analysis, validation, investigation, visualization, methodology, writing–original draft. A. Italiano: Conceptualization, resources, supervision, funding acquisition, writing–original draft, project administration, writing–review and editing.