Biomarkers of prognosis and pazopanib response in soft tissue sarcoma: An exploratory analysis

Abstract

Background

Advanced soft tissue sarcoma (STS) is a group of rare and heterogeneous malignancies, requiring better therapeutic strategies. Pazopanib, a widely used second-line treatment, has been suggested to exert immunomodulatory effects, which may be relevant for enhancing response. Specifically, pazopanib reduces immunosuppressive cells like myeloid-derived suppressor cells (MDSC) and regulatory T cells (Treg) while enhancing the function of dendritic cells, T cells, and NK effectors. Based on this knowledge, we hypothesized that proteins involved in immunomodulation could serve as predictive biomarkers of response to pazopanib.

Objectives

To explore immune-related biomarkers of prognosis and pazopanib response in advanced soft tissue sarcoma using paired pre- and post-treatment tumour samples and complementary preclinical models.

Design

Exploratory translational study integrating differential gene expression profiling of clinical specimens with in silico prognostic evaluation and in vitro functional assays in soft tissue sarcoma cell lines treated with pazopanib.

Methods

HTG transcriptome-direct profiling was performed on paired FFPE tumour samples from pazopanib-treated patients, followed by bioinformatic identification of differentially expressed genes and survival correlations, and subsequent validation of candidate biomarkers, at RNA and protein levels in sarcoma cell lines exposed to pazopanib.

Results

Using HTG transcriptomics, we identified 38 differentially expressed genes in post-pazopanib STS samples. Among them, SERPINE1 was the most consistent dynamic biomarker, showing a 4.19-fold increase (FDR = 0.016) and correlating with poor survival (progression-free survival (PFS), p = 0.033; disease-free survival (DFS), p = 0.004). In vitro, SERPINE1 upregulation was observed in STS cell lines CP0024 (24h, 1.47-fold; 48h, 1.71-fold), ICP059 (48h, 1.15-fold), and 93T449 (72h, 2.29-fold), with earlier expression in more sensitive cell lines. Protein validation confirmed these results after 48h of treatment.

Conclusions

Our results support SERPINE1 as a biomarker for poor prognosis and suggest its role as an exploratory, hypothesis-generating candidate marker for pazopanib response, warranting further validation in larger cohorts.

Introduction

Soft-tissue sarcoma (STS) is a rare and heterogeneous group of mesenchymal histologies, with an estimated annual incidence of 6 per 100,000 people, and they can occur throughout childhood (7.4% of all paediatric malignancies) and adulthood (1.5% of all malignant tumours in adults).^1,2 The standard treatment for localized STS is surgery (with or without radiation therapy) for low-risk cases, with the addition of adjuvant/neoadjuvant chemotherapy in high-risk patients. On the other hand, in the past 30 years, there has been no substantial progression regarding first-line treatment for patients with advanced STS, which mainly consists of single-agent anthracycline (usually doxorubicin) or anthracycline-based combinations. Despite the efforts to improve the activity of these treatments, it seems that their effectiveness has reached the ceiling.^3–5 For second-line treatment of advanced STS, a few therapeutic options are available, including gemcitabine combinations, eribulin, trabectedin, and pazopanib. There are no studies that indicate the best sequence for using second-line treatments in sarcomas. Additionally, the response rates for these drugs are less than 10%, with median progression-free survival (PFS) typically under 4-5 months.^6,7 Therefore, advancements in predicting effectiveness are highly desirable in the field of second-line treatments for individuals suffering from advanced STS.

Pazopanib is an oral tyrosine kinase inhibitor (TKI) that targets a wide spectrum of tyrosine kinase receptors (TKRs), including vascular endothelial growth factor receptors 1, 2, and 3 (VEGFR-1, VEGFR-2, VEGFR-3), platelet-derived growth factor receptors α and β (PDGFRα-β), fibroblast growth factor receptor 1 and 3 (FGFR1 and FGFR3), and c-KIT, among others. These targets are normally overexpressed in STS, and they are associated with worse prognosis, metastasis, chemotherapy resistance, and higher histologic grade. ⁸ Although pazopanib is active in some patients, there is a clinical need in terms of predictive biomarkers for this drug in STS. Even the most robust phase III trial evaluating pazopanib across multiple STS subtypes, the PALETTE study, failed to identify reliable predictive biomarkers of treatment response. ⁹ Notably, pazopanib seems to have a key immunomodulatory role as part of its mechanisms of action. Pazopanib has been shown to reduce immunosuppressive cells (myeloid-derived suppressor cells (MDSC) and regulatory T cells (Treg)) and increase the performance of T cells and NK effectors in the context of renal cell carcinoma. ¹⁰ In addition, pazopanib can prime dendritic cells in this type of tumor. ¹¹ Together, these findings suggest that the integration of this agent with immunotherapy may be feasible and that immune-related biomarkers may serve as predictors of the efficacy of pazopanib.

Although the knowledge about the pazopanib mechanisms of action is growing, robust biomarkers of pazopanib response in STS are still unknown. In the pivotal study of positioning pazopanib in STS, the authors reported an absence of correlation between the expression of angiogenic factors and the efficacy. ⁹ In this study, we analyze paired pre- and post-treatment samples along with in vitro studies to explore potential immune-related biomarkers of pazopanib efficacy in STS, generating hypotheses for further validation.

Materials and methods

Study population

Patients’ formalin-fixed paraffin-embedded (FFPE) tumour samples were selected and collected from six Spanish hospitals, with a long experience in STS management. The cases had to meet the following criteria: received pazopanib at any treatment line during the disease, with available FFPE blocks at diagnosis and after pazopanib treatment (either surgery or biopsy), and a minimum follow-up of 3 months. The clinical-pathological data collected in the study were: birth and diagnosis date, age at diagnosis, gender, presentation at diagnosis, tumour location, staging, tumour size, data on systemic treatment(s) if any, and histologic subtype. The status at the last follow-up (i.e., alive with or without disease, dead (with or without disease), date and type of recurrence, and any administered systemic treatment was obtained by direct follow-up from medical oncologists. Clinical data was updated and checked on a queries-based task between researchers and medical oncologists from each participating centre. The histologic subtypes were reviewed by a pathologist with expertise in STS. Given the limited number of available paired pre- and post-pazopanib samples (n=11), this study was designed as an exploratory, hypothesis-generating analysis. For 3 patients, the pre- and post-samples were collected from the same site, whereas for the other 8 patients, samples were collected from localized and metastatic disease sites. The small sample size limits statistical power. However, the paired study design partially mitigates inter-patient variability and allows the identification of dynamic transcriptional changes associated with pazopanib exposure. The reporting of this study conforms to the STROBE checklist statement ¹² (Supplementary File S1).

HTG molecular EdgeSeq immuno-oncology assay

The HTG EdgeSeq Immuno-Oncology Assay was used for the transcriptomic analysis of the immune response to cancer. The HTG EdgeSeq Immuno-Oncology assay quantitatively measured the expression of 549 genes involved in the innate and adaptive immune response to cancer. This specific panel was chosen in light of the reported immunomodulatory effects of pazopanib, allowing targeted exploration of biologically relevant immune-related pathways while limiting the multiple testing burden. When selecting the samples for the assay, only those with at least 70% of tumour area were directly processed. The samples that had less than 70% tumour or more than 20% necrotic tissue underwent macro-dissection. HTG EdgeSeq Chemistry was used for the RNA-Seq library synthesis, with all samples randomized before inclusion in this system to reduce potential biases in the run. Each sample, after hybridization, was used as a template to set up PCR reactions. Tags were designed to share common sequences that are complementary to 5′- end and 3′- sequences of the probes, common adaptors required for cluster generation on an Illumina sequencing platform, and a unique barcode used for sample identification and multiplexing. Once the PCR ended, a clean-up procedure was performed using Agencourt AMPure XP (Beckman Coulter, Beverly, MA, USA).

KAPA Library Quantification (Roche) was used to quantify the library. Triplicates for all samples and controls, and a no-template control were included in every run. 2N NaOH followed by 2N HCl was added to the library for denaturation, and PhiX was spiked in at a 5% (concentration of 12.5 pM). A demultiplexed FASTQ file per sample was obtained from the sequencer. HTG EdgeSeq host software performed the alignment of the FASTQ files to the probe list, processed the results, and provided a read counts matrix.

Bioinformatics analysis

RNA-Seq analysis was performed using the miARma-Seq pipeline from the demultiplexed FASTQ file. Reads were pre-processed with Cutadapt and then mapped to the Human Genome version 38, using Hisat2 aligner. Differential expression analysis was carried out with the EdgeR package (R version 3.4.2) with a paired sample design. The log₂ fold change (logFC) was calculated as log2(Baseline/Post-pazopanib). Consequently, negative logFC values indicate upregulation in post-treatment samples compared to baseline, while positive values indicate downregulation. Linear fold changes were calculated using the formula 2^-logFC for upregulated genes.

Since only a very small subset of genes is significant according to the adjusted (FDR) p-values, pathway enrichment was not executed. Besides, in silico bioinformatics analysis was performed with the cBioPortal software. The cBioPortal for Cancer Genomics provides visualization, analysis, and download of large-scale cancer genomics data sets.^13,14 The analysis was executed using the Cancer Genome Atlas (TCGA) sarcoma database. This database included 261 cases of sarcoma, which were sequenced by RNA-Seq. The mRNA Expression z-Score used was ± 2.0. The database included 105 leiomyosarcomas, 58 liposarcomas, 49 undifferentiated pleomorphic sarcomas (UPS), 25 myxofibrosarcomas, 10 synovial sarcomas, 9 malignant peripheral nerve sheath tumours (MPNST), 2 pleomorphic liposarcomas, 2 desmoid tumours, and 5 other STS.

Biostatistics

For the IC₅₀ proliferation index, a non-linear regression was performed, and the concentration at which 50% of the proliferation is inhibited (IC₅₀) was estimated for each cell line. Transcriptional activity was analysed with multiple t-tests with Holm-Sidak statistical significance, considering α=0.05. To analyse protein variation, unpaired t-tests were carried out. Statistical significance was considered at p-value<0.05. Statistical analysis was performed with GraphPad Prism 8.0.0 Software. Given the limited sample size, the statistical power of the analysis is constrained. However, the paired pre- and post-treatment design reduces inter-patient variability, and the application of FDR correction helps control for multiple testing.

Cell lines

Cell lines of leiomyosarcoma CP0024 (established at the Martin-Broto lab), liposarcoma 93T449 (ATCC® CRL-3043™; ATCC, Old Town Manassas, VA, USA) and angiosarcoma ICP059 (established at the Martin-Broto lab) were cultured in RPMI Medium 1640 + Glutamax^TM (Gibco, Thermo Fisher Scientific, Inc., Waltham, MA, USA) supplemented with 10% Fetal Bovine Serum (FBS) (Gibco), 1% of Penicillin and Streptomycin (P/S) (Sigma Aldrich) and 1% Fungizone (F) (Sigma Aldrich). Leiomyosarcoma cell line AA (kindly provided by Dr. Amancio Carnero of the Institute of Biomedicine of Seville, CSIC, US, HUVR; Seville, Spain) was maintained in F-10 Medium Nut Mix 1X (Ham) with L-glutamine (Gibco) supplemented with 10% FBS, 1% P/S, and 1% F. No FBS was added to cultures before the protein precipitation protocol. Cultures were incubated in a 5% CO₂ chamber at 37°C, and the medium was freshly replaced twice a week. Cells were counted using a Neubauer chamber and Trypan Blue (Sigma-Aldrich).

Determination of IC₅₀ proliferation index

Pazopanib was prepared in a DMSO vehicle at a 50 mM stock solution. To determine IC₅₀ values, cell lines were cultured in 96-well plates (each well with 2.5 × 10³ cells) and treated 24 h later with different Pazopanib concentrations (10^-7 M to 10^-4 M). These concentrations were prepared in 100µl of cell medium for each well. Control cells were treated with the vehicle. After 72 h of treatment, cellular viability was assessed adding 20µL MTS (3-(4,5-dimethylthiazol-2-yl)-5-(3-carboxymethoxyphenyl)-2-(4-sulfophenyl)-2H-tetrazolium) (“Cell Titer 96® Aqueous One Solution Cell Proliferation Assay”, ref. G3582, Promega, Wisconsin, EEUU) per well and absorbance at 490 nm was measured using iMarkTM Microplate Absorbance Reader (Bio-Rad Laboratories, Inc., Hercules, CA, USA).

In vitro analysis of transcriptional activity

The transcriptional profile in response to pazopanib was studied by real-time quantitative reverse polymerase chain reaction (qRT-PCR) from total RNA extracted from treated STS lines. The genes of interest were selected based on the fold change and p-value obtained from the bioinformatics analysis performed with HTG molecular data. RNA was extracted after 24 h, 48 h, and 72 h of pazopanib treatment (or DMSO for control) with PureLink™ RNA Mini Kit (ref. 12183020, Invitrogen, Thermo Fisher Scientific), following the manufacturer’s instructions. The High-Capacity cDNA Reverse Transcription Kit (Applied BiosystemsTM, Thermo Fisher Scientific) was used for the retrotranscription of the extracted RNA and of a universal RNA Pool (Universal Human Reference RNA, ref. 740000, Agilent, Santa Clara, CA, USA). The following TaqMan probes (TaqMan® Gene Expression Assays, Applied Biosystems) were used: MMP9 (Hs00957562_m1), SERPINE1 (Hs01126606_m1), CXCL8 (Hs00174103_m1), and GAPDH (Hs03929097_g1). The qPCR was carried out in a 7900HT Fast Real-Time PCR System thermal cycler (Applied Biosystems), with 3 technical replicates for each sample/probe. GAPDH was used as the endogenous housekeeping reference gene. The ^ΔCt and ^ΔΔCt methods were used to determine absolute and relative gene expression, respectively.

Western blot of media proteins

Protein samples were obtained from the culture medium through TCA-DOC precipitation, as the target proteins are usually found in the extracellular matrix. Protein levels were quantified through the Bradford (Bio-Rad) assay. For SDS-PAGE/western blot, acrylamide gels with two different running concentrations were used: 10% acrylamide for larger proteins (PAI-1 and MMP9) and 15% acrylamide for smaller proteins (IL-8). Electrophoresis was carried out at 90V and electroblot transference at 200 mA for 2 h, while membrane blocking was performed with 5% BSA/Tris-Buffered Saline, 0.1% Tween® (TBS-T). Subsequently, membranes were incubated overnight at 4° C, with the following primary antibodies: anti-MMP9 (ab38898, Abcam), anti-PAI-1 (SC-5297, Santa Cruz), and anti-CXCL8 (AF-208-NA, RD Systems). Secondary antibodies were incubated for 1 h, at room temperature: Goat/anti-rabbit IgG antibodies (ab6721, Abcam) and IgG anti-mouse Peroxidase (a9044, Sigma-Aldrich). ChemiDoc Imaging System (Bio-Rad) was used for signal detection.

Results

Patients

A series of 11 patients diagnosed with STS, who had received pazopanib at any line of advanced disease and with a post-pazopanib tumour FFPE sample available, were included in the study population. With a median age at diagnosis of 41 years old, 6 males and 5 females were included. Patients were diagnosed between 2005 and 2015. The median follow-up of this series was 73 months, with 8 tumors initially localized and 3 metastatic. However, 7 metastatic events were recorded during follow-up. The STS subtypes included were the following: solitary fibrous tumour (n=3), synovial sarcoma (n=3), leiomyosarcoma (n=2), liposarcoma (n=2), and UPS(n=1). The median size was 10.15 cm (ranging from 4 to 30 cm), and primary tumour locations were extremities (n=3), retroperitoneum (n=2), dorsolumbar (n=1), eye orbit (n=1), abdominal wall (n=1), superior maxilla (n=1), and uterus (n=1).

The median overall survival (OS) of the study population was 72.83 (4.63-144.40) months. There were 5 events of death among the 11 cases included in the study. Patients were treated with pazopanib at any line of treatment for advanced disease: 2 cases received pazopanib as first-line treatment, 1 case as second-line treatment, 5 cases were treated with pazopanib as third-line treatment, and 3 received pazopanib as fourth-line treatment. Regarding sampling timelines, all pre-treatment specimens were uniformly obtained from the treatment-naive tumor at the time of initial diagnosis. Notably, to ensure biological consistency across the cohort, all post-treatment samples were systematically collected immediately following the completion or cessation of pazopanib therapy. The two patients who received pazopanib as first-line systemic therapy were diagnosed with dedifferentiated liposarcoma (DD-LPS) and were treated within the GEIS-30 clinical trial (NCT01692496), which evaluated the activity and tolerability of pazopanib in patients with advanced and/or metastatic liposarcoma for whom no standard therapeutic options were available. The median PFS was 5.57 (1.40-24.37) months (Table 1).

Table 1.

Demographics and clinical-pathologic information.

Variable	N
Age: median (range)	41 (21-65)
Sex: male/female (%)	6 (54.55%)/5 (45.45%)
Median follow-up (months)	73
Median size of primary tumours (cm; range)	10.15 (4-30)
Primary tumor presentation:
Localized (%)	8 (72.73%)
Metastatic (%)	3 (27.27%)
Metastatic events (%)	7 (63.64%)
Primary tumor site:
Extremities (%)	3 (27.27%)
Retroperitoneum (%)	2 (18.18%)
Dorsolumbar (%)	1 (9.10%)
Eye orbit (%)	1 (9.10%)
Abdominal wall (%)	1 (9.10%)
Superior maxilla (%)	1 (9.10%)
Uterus (%)	1 (9.10%)
Kidney (%)	1 (9.10%)
Soft-tissue sarcoma Subtype:
Solitary fibrous tumour (%)	3 (27.27%)
Synovial sarcoma (%)	3 (27.27%)
Leiomyosarcoma (%)	2 (18.18%)
Liposarcoma (%)	2 (18.18%)
Undifferentiated pleomorphic sarcoma	1 (9.10%)
Median OS (Range)	72.83 (4.63-144.40)
Median PFS (Range)	5.57 (1.40-24.37)
Pazopanib treatment-line:
First-line	2 (18.18%)
Second-line	1 (9.10%)
Third-line	5 (45.45%)
Fourth-line	3 (27.27%)

Differential gene expression analysis

The bioinformatics analysis carried out identified 38 genes significantly and differentially expressed between baseline and post-pazopanib tumour samples (Table 2). Among them, the statistical significance after multiple comparisons (FDR) was maintained only for: CD68 (logFC=-1.512; FDR=0.007), SERPINE1 (logFC= -2.068; FDR=0.016), C1QA (logFC=-1.563; FDR= 0.027), and CXCL8 (logFC=-1.192; FDR=0.044) genes. These genes were all overexpressed in post-pazopanib samples, compared to baseline samples. Specific values for fold changes and p-values of the 38 differentially expressed genes can be found in Table 2.

Table 2.

Paired analysis between pre- and post-Pazopanib tumour samples.

​	Length	logFC*	p-Value	FDR
CD68	2189	-1.512	1.450E-5	0.007
SERPINE1	3190	-2.068	6.950E-5	0.016
C1QA	1272	-1.563	1.770E-3	0.027
CXCL8	2274	-1.192	3.827E-4	0.044
CCR1	2731	-0.967	1.701E-3	0.130
ITGAM	5665	-1.128	1.707E-3	0.130
CTSS	4228	-1.162	2.368E-3	0.151
TLR9	3870	1.511	2.647E-3	0.151
TNFSF11	2733	1.488	3.159E-3	0.160
MYC	3001	-1.082	4.939E-3	0.221
MMP9	2336	-1.656	5.829E-3	0.221
CD4	4583	-0.907	6.440E-3	0.221
HAVCR2	3804	-0.773	6.750E-3	0.221
HMOX1	2405	-1.165	7.014E-3	0.221
HLA-B	3051	-1.050	7.274E-3	0.221
PAX5	9333	1.226	9.272E-3	0.247
CYBB	4423	-1.033	9.372E-3	0.247
CSF1R	5151	-0.859	9.781E-3	0.247
ITGB2	6797	-0.998	1.098E-2	0.263
CD24	3097	1.825	1.268E-2	0.284
MSR1	7812	-0.807	1.359E-2	0.284
CD44	10057	-0.961	1.373E-2	0.284
XCL2	565	2.008	1.455E-2	0.288
CD74	3184	-1.293	1.640E-2	0.311
IL7	3392	0.901	1.800E-2	0.328
CD14	1886	-0.788	1.927E-2	0.333
ITGA5	7577	-0.984	1.978E-2	0.333
C3AR1	2145	-0.690	2.141E-2	0.348
CXCL2	1500	-0.948	2.395E-2	0.376
NCAM1	12737	0.953	2.613E-2	0.396
NFATC1	9068	-0.688	2.971E-2	0.433
IL20	1567	1.040	3.043E-2	0.433
CD86	3433	-0.715	3.324E-2	0.458
CRP	2015	0.755	3.609E-2	0.483
CX3CR1	3655	0.691	3.873E-2	0.504
CCL2	1986	-0.903	4.459E-2	0.550
SPP1	2321	-1.247	4.472E-2	0.550
CSF1	5418	-0.715	4.914E-2	0.588

LogFC is calculated as log2(Baseline/Post-treatment). Negative logFC values represent gene upregulation post-treatment, whereas positive values represent downregulation.

*p-value <0.05.

SERPINE1 overexpression correlates with worse PFS and DFS in silico

Based on FDR values and/or logFC, 3 genes were selected for in silico and in vitro analysis: SERPINE1, CXCL8, and MMP9. For MMP9, although not significant by FDR, this gene was selected according to the fold change value, which was the second-greatest one. The cBioPortal platform for Cancer Genomics ¹⁴ was used, with data from the TCGA sarcoma database ¹⁵ and with a z-Score=±2.0. The association of SERPINE1 overexpression with poorer patient survival (Supplementary Figure S1) was significantly confirmed in PFS (p=0.033) and disease-free survival (DFS; p=0.004) compared to normal levels of SERPINE1 expression in STS patients. Regarding OS, there was no statistically significant difference between the two groups of patients (p=0.0998). In the case of both MMP9 and CXCL8, we could not find any association of overexpression with worse OS (p=0.283, 0.369), worse PFS (p=0.480, 0.957), or worse DFS (p=0.262, 0.636) in STS patients.

IC₅₀ proliferation index values

IC₅₀ proliferation index values were determined in STS cell lines after 72 h of Pazopanib treatment. The IC₅₀ values calculated were the following: 1.7 µM for CP0024, 2.5 µM for ICP059, 6.3 µM for 93T449 and 12.3 µM for AA (Supplementary Figure S2).

Gene expression of SERPINE1, MMP9, and CXCL8 after pazopanib treatment

RNA expression of SERPINE1, MMP9, and CXCL8 was determined in our cell lines after treating them with 10 µM pazopanib or vehicle control at three different time conditions: 24 h, 48 h, and 72 h. MMP9 had increased RNA expression after pazopanib treatment in CP0024 at 24 h (FC=1.972; p=0.003), CP0024 at 48 h (FC=1.498; p=0.025), and ICP059 at 48 h (FC=1.206; p=0.016) (Figure 1, Supplementary Table S1). SERPINE1 was the gene that had the most significant changes in RNA expression induced by pazopanib, throughout both treatment time and cell lines. Specifically, it was increased in CP0024 at 24 h (FC=1.469; p=0.002), CP0024 at 48 h (FC=1.710; p<0.001), ICP059 at 48 h (FC=1.154; p=0.007) and 93T449 at 72 h (FC=2.288; p=0.007). There was also a significant decrease in 93T449 at 24 h (FC=0.291; p=0.012). Regarding CXCL8, the results were more variable. At 48 h, there was a significant increase in CP0024 (FC=4.422; p=0.031) and ICP059 (FC=3.779; p<0.001), as well as in AA at 72 h (FC=3.203; p=0.020). However, there were also several significant decreases, mostly in 93T449 at all time conditions (24 h: FC=0.055, p<0.001; 4 h: FC=0.207, p<0.001; 72 h: FC=0.163, p<0.001) and also in ICP059 at 72 h (FC=0.504; p=0.019) (Figure 1, Supplementary Table S1). In general, gene expression is triggered later in cell lines that show greater resistance (IC₅₀) to pazopanib. Accelerated Failure Time (AFT) regression was applied to time-to-event data to estimate how the IC₅₀ values affected the time to significant gene expression induction, while handling censored data (no induction observed for AA). The IC₅₀ coefficient was 0.197, with a p-value of 0.002. However, the results should be interpreted with caution due to the low number of samples.

Figure 1.

MMP9, SERPINE1, and CXCL8 mRNA expression after pazopanib treatment. Cells were collected after 24 h, 48 h, or 72 h of pazopanib or vehicle (DMSO) treatment. mRNA levels were evaluated by RT-qPCR. Increases in gene expression seem to occur at longer times in cell lines with higher IC50 values for pazopanib. Unpaired t-student tests: (*) stands for p-value < 0.05, (**) for p-value < 0.01, and (***) for p-value < 0.001. Bars are indicative of means ± SD. Independent experiments were performed (N=4).

Changes in SERPINE1, MMP9, and CXCL8 after pazopanib treatment are validated at the protein level

RNA expression patterns were consecutively validated at the protein level, 48 h after pazopanib treatment. Supernatant-purified proteins were used for western blot; for protein quantification, and in the case of PAI-1 (SERPINE1), only the active band (∼47 kDa) was analyzed. Analogous to the RNA expression studies, we observed a significant increment after treatment in PAI-1 (1.47-fold, p=0.003), MMP9 (1.89-fold, p<0.001), and IL-8 (2.34-fold, p=0.011) protein expression in CP0024 compared to control. Likewise, the protein levels of PAI-1 (1.30-fold, p=0.042) and IL-8 (3.12-fold, p=0.018) were significantly increased in ICP059 compared to control. However, MMP9 levels did not significantly change between ICP059 pazopanib-treated and control cells. In 93T449, only PAI-1 expression levels were significantly higher (1.69-fold, p=0.013) after treatment. In AA cells, no significant changes in protein expression were detected (Figure 2).

Figure 2.

MMP9, PAI-1, and IL-8 extracellular protein expression after pazopanib treatment. Serum-free medium from cultured cells was collected after 48 h of pazopanib or vehicle (DMSO) treatment. Proteins were precipitated by the TCA-DOC method, and subsequently, the samples were used for WB analysis. Statistically significant increase in extracellular protein expression was observed for MMP9 in CP0024 cells; PAI-1 in CP0024, ICP059, and 93T449 cells, and IL-8 in CP0024 and ICP059 cells. Unpaired t-student tests: (*) stands for p-value < 0.05, (**) for p-value < 0.01, and (***) for p-value < 0.001. Error bars are indicative of means ± SD. Independent experiments were performed (N=3).

Discussion

In our study, 38 genes were significantly modulated by pazopanib treatment, with SERPINE1 emerging as the most reliable dynamic biomarker. Its expression increased after treatment, but this response varied among cell lines: it appeared earlier in more sensitive lines and was delayed or absent in resistant ones. This was also confirmed at the protein level after 48 hours.

SERPINE1 encodes PAI-1, the main inhibitor of tissue plasminogen activator (tPA) and urokinase (uPA), playing a key role in fibrinolysis. ¹⁶ It has been linked to poor prognosis, metastasis, and cancer progression in multiple cancers, including lung, gastric, and breast cancer, as well as osteosarcoma.^17–21 Our in silico analysis supports this association with prognosis, showing a correlation between SERPINE1 and poor survival in STS patients (PFS, p=0.033; DFS, p=0.004). HTG analysis of STS tumors further confirmed its overexpression in over 50% of cases. Earlier SERPINE1 increase in sensitive cells could seem counterintuitive since it could mean that SERPINE1 works as a poor prognosis biomarker in STS cancer progression, but as a better response predictor for pazopanib treatment. This dual role is not unprecedented, as SERPINE1 has been proposed as a predictive biomarker for standard treatment response in rectal cancer, where higher expression correlated with better treatment outcomes, ²² despite its known association with poor prognosis.

The link between SERPINE1 and pazopanib response may relate to the drug’s mechanism of action. Pazopanib inhibits VEGFR-1, -2, and -3, blocking VEGF signaling and suppressing angiogenesis. Since SERPINE1 may regulate VEGFA, as its knockdown results in VEGFA suppression and reverses paclitaxel resistance in triple-negative breast cancer, ²³ its overexpression after pazopanib treatment could be a compensatory response to VEGFR inhibition. In resistant cell lines, where pazopanib is less effective at blocking VEGFRs, this compensatory mechanism might not be triggered. Additionally, it has been described that SERPINE1 inhibits activation signalling downstream of VEGFR-2 by inhibiting its phosphorylation and favouring its translocation to perinuclear endosomes. ²⁴ This could mean that SERPINE1 overexpression might eventually enhance the pazopanib antiangiogenic effect. It should be noted that the in vitro cell line models used in this study do not capture the vascular or angiogenic effects of pazopanib. Therefore, observations related to VEGF signaling and vascular modulation are inferred from previously published evidence and should not be interpreted as direct functional validation of vascular effects. Crucially, while our in silico TCGA analysis firmly establishes the prognostic relevance of SERPINE1 expression in STS, its potential predictive value regarding pazopanib response remains an indirect, hypothesis-generating finding from our exploratory cohort. Consequently, these results must be interpreted with caution, and their definitive role as a predictive biomarker requires validation in prospective, uniformly treated cohorts.

As part of our exploratory, hypothesis-generating observations, while MMP9 and CXCL8 showed some promising trends similar to SERPINE1, such as RNA expression increases, we found insufficient evidence to support their prognostic or predictive roles. However, these three genes appear to be functionally connected in certain processes. SERPINE1 has been reported to induce cancer-associated mesothelial cells, which subsequently increase CXCL8 and CXCL5 expression, promoting ovarian cancer metastasis in a feedback loop. ²⁵ This suggests that SERPINE1 upregulation leads to CXCL8 upregulation, meaning that monitoring both genes could serve as a double-check in post-pazopanib response, although our CXCL8 results remain inconclusive. In breast cancer patients, there is a positive association between MMP9-CXCL8 and SERPINE1-CXCL8 as indicators of poor prognosis. ²⁶ Similarly, both MMP9 and SERPINE1 have been identified as prognostic factors in esophageal cancer. ²⁷ Additionally, MMP9 positively regulates VEGFA, much like SERPINE1, ²⁸ suggesting that both genes may counteract pazopanib’s anti-angiogenic effects. Their similar behavior in the pazopanib response further supports this functional analogy. Moreover, SERPINE1, CXCL8, and MMP9 have been linked as poor-prognosis biomarkers in multiple malignancies, including gastric adenocarcinoma, bladder cancer, breast cancer, and cardio-cerebral vascular diseases.^26,29–31 This interconnection may explain their similar expression patterns following pazopanib treatment. Further studies are needed to clarify the roles of MMP9 and CXCL8 in pazopanib response.

Beyond individual gene dynamics, our findings hold significant translational relevance when framed within the broader antiangiogenic–immune interface in STS. VEGFR-targeted TKIs, such as pazopanib, are increasingly recognized not just as pure antiangiogenic agents, but as vital tools for immune remodeling within the tumor microenvironment.^32,33 By targeting the VEGF/VEGFR axis, these TKIs can alleviate tumor-induced immunosuppression—reversing T-cell exhaustion, decreasing Tregs and MDSCs, and enhancing dendritic cell maturation. ³⁴ This dual mechanistic action provides a powerful biological rationale for combining TKIs with immune checkpoint inhibitors (ICIs) to achieve a synergistic effect. Indeed, recent prospective clinical trials exploring TKI plus anti-PD-1/PD-L1 combinations in advanced STS have demonstrated encouraging antitumor activity and prolonged PFS compared to historical TKI monotherapy benchmarks.^35–40 Within this therapeutic framework, our identification of SERPINE1 as a highly dynamic, treatment-associated biomarker takes on a predictive utility. Because SERPINE1 upregulation represents a potential biomarker of VEGFR inhibition—and is intimately tied to macrophage infiltration and microenvironmental crosstalk ⁴¹ —it stands as a promising candidate biomarker for future trials. Specifically, monitoring SERPINE1 expression could prove invaluable for tracking drug-induced immune remodeling, identifying early adaptive resistance, and selecting patients most likely to benefit from rational antiangiogenic–immunotherapy combinations in STS.

Our study has several limitations that should be considered. First, the small number of patients with paired samples could impact the transcriptomic analysis. While collecting 11 paired samples was a significant achievement given the extreme rarity of STS specimens post-pazopanib treatment, our findings should be validated in a larger case series of paired samples. Second, the heterogeneity of STS subtypes, primary tumor locations, and metastatic events poses a challenge for obtaining significant data. Future observational or clinical studies with anti-angiogenic agents should aim for more homogeneous patient cohorts, where post-treatment biopsies could help generate stronger data. Third, the timing of post-pazopanib sample collection varied among patients in terms of line of treatment (first to fourth line), even if all patients received pazopanib and were sampled immediately upon treatment cessation. While all patients received pazopanib, the line of treatment sequence may have influenced both clinical outcomes and transcriptional profiles due to the cumulative effects of prior systemic treatments. This variability represents an additional confounding factor that should be considered when interpreting the results. A clean cohort with baseline and end-of-treatment samples could address this issue, though such paired samples remain rare and difficult to collect unless a prospective study with mandatory biopsies before and after pazopanib treatment is designed. Fourth, the prolonged inclusion period required to accumulate this rare cohort introduces potential temporal bias due to historical variations in clinical management, treatment sequencing, and pre-analytical tissue handling. While all samples met strict quality control thresholds, these long-term variations remain an inherent confounding factor in retrospective tissue collections. Lastly, differences in disease progression stages among patients may explain the higher expression of certain genes in post-pazopanib samples. However, our in vitro results replicated the upregulation of some genes following pazopanib treatment, suggesting that pazopanib plays a major role—if not the primary role—in driving these gene expression changes.

Most of the selected genes are linked to macrophages, suggesting that the STS tumor microenvironment influences their expression. Moreover, the dynamic crosstalk between antiangiogenic therapy and the immune microenvironment is increasingly recognized as a critical determinant of treatment response. Emerging data indicate that antiangiogenic agents can normalize or remodel tumor vasculature and immune cell infiltration, thereby reshaping the tumor immune microenvironment, modulating immune escape and drug resistance, and ultimately influencing the efficacy of targeted therapies, including pazopanib.^32,42 MMP9 and IL-8 are secreted by both tumor and immune cells, including macrophages.^43–45 Other macrophage-related genes, such as C1QA, CD68, and TLR9, were identified in previous studies but not selected for this one. ⁴⁶ Notably, CD68 expression correlates with M2 macrophage density in pazopanib-treated melanomas. ⁴⁷ These findings indicate a strong connection between post-pazopanib gene overexpression and the immune microenvironment. Immunohistochemistry of macrophage markers in pazopanib-treated patient samples could clarify whether macrophages drive these differences, but our RNA and protein analyses suggest that SERPINE1 upregulation occurs independently of the microenvironment, as it was induced by pazopanib in monolayer cultures.

Future research should include in vivo studies in immunocompetent STS models to validate the impact of pazopanib on gene expression. Additionally, co-culture experiments with immune cells could determine their role in the tumor’s response to pazopanib. A key step would be investigating whether these gene expression patterns, particularly SERPINE1, can be detected in liquid biopsies. This could provide a minimally invasive method to assess pazopanib response, helping identify responders within 24–72 hours of treatment.

Conclusions

In conclusion, this exploratory study highlights the highly dynamic nature of the STS transcriptomic landscape during systemic therapy, uncovering key microenvironmental changes associated with pazopanib treatment. Our findings position SERPINE1 as a promising, hypothesis-generating candidate biomarker for pazopanib response. However, given the limited sample size and exploratory nature of this cohort, these results must be interpreted with caution. Ultimately, large-scale, prospective validation in independent and uniformly treated patient cohorts is strictly required to definitively establish the predictive and clinical utility of SERPINE1, as well as the roles of other emerging candidates such as MMP9 and CXCL8, before any translational application can be considered.

ORCID iDs

Jose L. Mondaza-Hernandez https://orcid.org/0000-0002-5642-3796

Josefina Cruz https://orcid.org/0000-0001-6398-1641

Javier Martin-Broto https://orcid.org/0000-0001-7350-6916

Ethical considerations

The study was approved by the Ethics Committee for Research with Medicines of Hospital Universitario Virgen del Rocío (CEIm-HUVR, Seville, Spain) and conducted in accordance with the Declaration of Helsinki.

Consent to participate

All patients provided written informed consent to participate in the translational study.

Data Availability Statement

The datasets generated and/or analyzed during the current study are not publicly available due to patient privacy and ethical restrictions, but are available from the corresponding authors on request.*