GDF-15 Predicts Epithelioid Hemangioendothelioma Aggressiveness and Is Downregulated by Sirolimus through ATF4/ATF5 Suppression

Abstract

Purpose: Epithelioid hemangioendothelioma (EHE), an ultra-rare sarcoma, poses therapeutic challenges because of limited efficacy of conventional chemotherapy in advanced cases, necessitating exploration of new treatment avenues and identification of novel aggressive biomarkers. This study aimed at (i) utilizing a patient-derived xenograft model of EHE and its associated cell line to assess the efficacy of sirolimus and (ii) analyzing two distinct patient cohorts to pinpoint circulating biomarkers of EHE aggressiveness.

Experimental design: A patient-derived xenograft model and corresponding cell line were established from a patient with advanced EHE, demonstrating consistency with the original tumor in terms of histomorphology, WWTR1::CAMTA1 fusion presence, and genomic and transcriptomic profiles. Two independent patient series were employed to investigate the association between growth/differentiation factor 15 (GDF-15) serum levels and EHE aggressiveness.

Results: ELISA analyses on EHE cell culture medium and blood from EHE-carrying mice revealed the release of GDF-15 by EHE cells. Sirolimus exhibited markedly higher antitumor activity compared with doxorubicin, concurrently reducing GDF-15 expression/release both in vivo and in vitro. This reduction was attributed to the drug-induced inhibition of phosphorylation/activation of 4E-BP1 and subsequent downregulation of the GDF-15 transcription factors ATF4 and ATF5. Blood sample analyses from two independent patient series showed a significant correlation between GDF-15 and EHE aggressiveness.

Conclusions: This study identifies GDF-15 as a novel biomarker of EHE aggressiveness and underscores the superior efficacy of sirolimus compared with doxorubicin in our experimental models. The observed inhibition of GDF-15 release by sirolimus suggests its potential as a biomarker for monitoring the drug's activity in patients.

Figure 1.

Characterization of EHE PDX and paired cell line. Representative pictures of the EHE clinical sample and corresponding EHE PDX model. A, The histology was assessed on hematoxylin and eosin–stained slides. Scale bar, 100 µm. WWTR1::CAMTA1 dual-color dual-fusion FISH pattern in diploid tumor cells showing one single fusion, consistently with the genomic profile. B, Genomic profile of EHE clinical tumor (left) and PDX (right). C, Loss (blue) and gain (red) were depicted in the top. LOH (green) was reported in the bottom (scale bar, 10 µm). D, Scatter plot of significant correlation between the transcriptome of the clinical tumor and the paired PDX (Spearman correlation, r_s = 0.916; P < 0.001). E, Morphology of the EHE cell line derived from the PDX growing as monolayer (left, scale bar, 250 µm) and confirmation of WWTR1::CAMTA1 translocation at FISH analysis (right, scale bar, 10 µm). H&E, hematoxylin and eosin; LOH, loss of heterozygosity; RNA-seq, RNA sequencing.

Figure 2.

GDF-15 was released by tumor cells in patient-derived models of EHE. A, Assessment of released cytokines in the culture medium of EHE cells using the Human XL Cytokine Array. Detection of GDF-15 by ELISA in culture medium of EHE cell lines and cell lines of PLPS and DDLPS-1 and -2. B, Data were normalized as amount (pg) of released GDF-15 to total released (mg) proteins. RT-qPCR and ELISA results are reported as mean ± SD of three independent experiments. C, Detection of GDF-15 by ELISA in the plasma of healthy mice and mice carrying EHE PDX or PLPS PDX. D, siRNA-mediated downregulation of GDF-15 in the EHE cell line as detected at the mRNA level by RT-qPCR (left), protein level by Western blotting (middle), and as cytokine released in cell culture medium by ELISA 3 days after transfection (right). ELISA data were normalized as amount (pg) of released GDF-15 to total released (mg) proteins and reported as mean ± SD of three independent experiments. E, GSEA (RRID: SCR_003199) was employed on Hallmark (H) collection of the Molecular Signature Database, showing a limited number of modulated pathways following GDF-15 knockdown. DDLPS, dedifferentiated liposarcoma; PLPS, pleomorphic liposarcoma.

Figure 3.

Sirolimus inhibited the growth of patient-derived models of EHE. A, Cell growth inhibition curves obtained after exposure to different doses of doxorubicin or sirolimus. Data are reported as the percentage of drug-treated cells compared with control cells and represent mean ± SD of three independent experiments. B, Growth curves reporting the RTW (mean ± SEM) in control and doxorubicin- or sirolimus-treated mouse groups (nine mice/group), in which 1 indicates the tumor weight at the beginning of the treatment. The arrows indicate when drugs were administered. C, Histomorphologic evaluation of tumors obtained from untreated and drug-treated mice. D, Ki67 immunostaining of tumors obtained from untreated and drug-treated mice (top) and quantification of Ki67 index (bottom). Symbols reported in the represent counted fields. Histomorphologic analysis and Ki67 immunostaining were performed on tumors excised from mice at the end of drug treatment. Scale bar, 100 μm. Data are reported as means ± SD of three independent experiments. E, Western blot analysis of downstream mTOR pathway in untreated cells and cells treated with different sirolimus concentrations for 3 days (left) and in tumors removed from untreated and sirolimus-treated mice at the end of treatment with different drug doses (right). Cropped images of selected proteins are shown. RTW, relative tumor weight.

Figure 4.

Sirolimus downmodulated GDF-15 release in patient-derived models of EHE through the inhibition of ATF4 and ATF5. A, GDF-15 released in culture medium of control cells and cells exposed for 3 days to doxorubicin or different sirolimus concentrations as detected by ELISA. Data were normalized as amount (pg) of released GDF-15 to total released (mg) proteins and reported as mean ± SD of three independent experiments. B, GDF-15 released in the blood collected from untreated mice and after the end of treatment with doxorubicin or different sirolimus doses. Data were normalized as amount (pg) of released GDF-15 to tumor weight (g) and reported as mean ± SD of three mice/experimental group. C, Western blot analysis of GDF-15, ATF4, and ATF5 expression in untreated cells (−) and cells treated with different sirolimus concentrations for 3 days (left) and in tumors removed from untreated (−) and sirolimus-treated mice after the first round of treatment with different drug doses (right). Western blot data of GDF-15 expression obtained on tumors from three untreated mice (−) and three mice exposed to 2.5 mg/kg sirolimus. Cropped images of selected proteins are shown. D, siRNA-mediated downregulation of ATF4 (left) and ATF5 (right) in the EHE cell line as detected at the mRNA level by RT-qPCR. E, Effects of ATF4 and ATF5 downregulation on GDF-15 protein expression as detected by level by Western blotting (left) and on cytokine release in cell culture medium as measured by ELISA (right). ELISA data were normalized as amount (pg) of released GDF-15 to total released (mg) proteins and reported as mean ± SD of three independent experiments. Cropped images of selected proteins are shown. F, The working model of sirolimus-induced GDF-15 downregulation. Sirolimus-induced mTORC1 inhibition led to reduced 4E-BP1 phosphorylation/activation followed by decreased ATF4 and ATF5 expression, thus resulting in GDF-15 downregulation. (F, Created with BioRender.com.)

Figure 5.

Circulating GDF-15 levels correlated with disease aggressiveness in patients with EHE. A, Analysis of circulating cytokines in plasma samples of 16 patients with EHE and four healthy donors using the Human XL Cytokine Array showed higher GDF-15 levels in patients compared with healthy donors. B, GDF-15 levels, as detected by ELISA and expressed in pg/mL plasma, in patients with EHE compared with healthy donors in a training (retrospective; 8 higher-risk and 12 lower-risk patients and 32 healthy donors) and a testing (prospective; 6 higher-risk and 15 lower-risk patients and 32 healthy donors) cohort and in both the cohorts. C, Two clinical cases are reported showing levels of GDF-15 in a lower-risk patient (top) with stable disease at follow-up and a higher-risk patient (bottom) who had an initial tumor response and levels of GDF-15 reduced after being started on the mTOR inhibitor sirolimus followed by an increase in GDF-15 before the CT scan that showed disease progression. PD, progressive disease; SD, stable disease.