A tumor-promoting role for soluble TβRIII in glioblastoma

Purpose Members of the transforming growth factor (TGF)-β superfamily play a key role in the regulation of the malignant phenotype of glioblastoma by promoting invasiveness, angiogenesis, immunosuppression, and maintaining stem cell-like properties. Betaglycan, a TGF-β coreceptor also known as TGF-β receptor III (TβRIII), interacts with members of the TGF-β superfamily and acts as membrane-associated or shed molecule. Shed, soluble TβRIII (sTβRIII) is produced upon ectodomain cleavage of the membrane-bound form. Elucidating the role of TβRIII may improve our understanding of TGF-β pathway activity in glioblastoma Methods Protein levels of TβRIII were determined by immunohistochemical analyses and ex vivo single-cell gene expression profiling of glioblastoma tissue respectively. In vitro, TβRIII levels were assessed investigating long-term glioma cell lines (LTCs), cultured human brain-derived microvascular endothelial cells (hCMECs), glioblastoma-derived microvascular endothelial cells, and glioma-initiating cell lines (GICs). The impact of TβRIII on TGF-β signaling was investigated, and results were validated in a xenograft mouse glioma model Results Immunohistochemistry and ex vivo single-cell gene expression profiling of glioblastoma tissue showed that TβRIII was expressed in the tumor tissue, predominantly in the vascular compartment. We confirmed this pattern of TβRIII expression in vitro. Specifically, we detected sTβRIII in glioblastoma-derived microvascular endothelial cells. STβRIII facilitated TGF-β-induced Smad2 phosphorylation in vitro and overexpression of sTβRIII in a xenograft mouse glioma model led to increased levels of Smad2 phosphorylation, increased tumor volume, and decreased survival Conclusions These data shed light on the potential tumor-promoting role of extracellular shed TβRIII which may be released by glioblastoma endothelium with high sTβRIII levels. Supplementary Information The online version contains supplementary material available at 10.1007/s11010-021-04128-y.


Introduction
Glioblastoma is one of the most common malignant intrinsic brain tumors [1]. Prominent biological features of glioblastoma include excessive migratory, invasive and angiogenic potential, and suppression of anti-tumor immune surveillance. Glioma-derived transforming growth factor (TGF)-β is thought to be fundamental in these processes. Self-renewing, highly tumorigenic glioma-initiating cells (GIC) have been proposed to be more resistant to therapy than the tumor bulk [2], to have invasive properties and to be involved in angiogenesis [3][4][5]. TGF-β signaling has been proposed to be a key regulator in glioma vasculature [6] and the maintenance of stem cell-like properties and tumorigenic activity of GIC [7,8]. Thus, targeting TGF-β may affect glioma vessels and GIC, thereby inhibiting tumor growth and sensitizing tumors to conventional therapies [9]. Proteins of the TGF-β superfamily interact with TGF-β receptors I-III (TβRI-III). TβRI, also known as activin receptor-like kinases (ALK), and TβRII form heteromeric complexes upon binding of TGF-β family ligands [10]. Subsequently, canonical and non-canonical TGF-β signal transduction is activated. Canonical TGF-β signal transduction is mediated by Smad transcription factors resulting in the phosphorylation of the receptor-regulated Smads, Smad 1, 2, 3, 5, and 8 [11]. TGF-β receptors may also directly interact with or phosphorylate non-Smad proteins initiating parallel signaling that cooperates with the Smad pathway in downstream responses [12]. The TGF-β receptor core complex is not only constituted by TβRI and TβRII, but also by the TGF-β signaling coreceptor TGF-β receptor type III (TβRIII), a ubiquitously expressed accessory TGF-β receptor [13]. TβRIII is a transmembrane protein with a large extracellular domain (ECD) with glycosaminoglycan groups [41]. By its ECD, it may bind multiple members of the TGF-β family such as TGF-β1-3, activin-A, bone morphogenetic proteins (BMP)-2, BMP-4, BMP-7, and growth differentiation factor (GDF) 5 as well as inhibin [14,15]. TβRIII undergoes ectodomain shedding from the cell surface to generate soluble forms of the receptor [16]. The ECD is then capable of binding TGF-β ligands [17]. Overall, the role of TβRIII is complex and difficult to predict as the balance of cell surface and shed TβRIII, which is also called soluble TβRIII (sTβRIII), may differentially regulate TGF-β superfamily signaling [13,18]. Given the putative central role of TGF-β superfamily signaling in glioblastoma, the present study focused on the role of TβRIII in the regulation of TGF-β pathway activity in this disease.

Cell culture and reagents
Details are summarized in supplementary note 1.

Real-time PCR (RT-PCR)
RT-PCR was performed as previously described [19] and is described in supplementary note 2.

Immunoblot analyses
Immunoblot analysis was performed as previously described [20], and reagents are listed in supplementary note 1.

Enzyme-linked immunosorbent assay (ELISA)
Experimental details are listed in supplementary note 3.

Flow cytometry
LTCs were seeded at confluency and flow cytometry was performed 48 h after serum deprivation [19].

Animal studies
Experimental details are described in supplementary note 4.

Immunohistochemistry
Experimental steps were conducted as described in supplementary note 5.

Immunofluorescence microscopy
Experimental details are provided in supplementary note 6.

Single-cell real-time polymerase chain reaction (scRT-PCR) of reverse-transcribed RNA
Experimental details are described in supplementary note 7.

Statistical analysis
Details are provided in supplementary note 8.

TβRIII expression in gliomas in vivo
Immunohistochemistry was used to assess TβRIII levels in normal human brain and in glioblastoma tissue samples. Normal brain of 13 individuals who were not diagnosed with a glioblastoma was tested using a tissue microarray (TMA). The TMA included normal brain hippocampus tissue punches of 12 different individuals and three punches of a gyrus temporo-occipitalis tissue of one individual. TβRIII in glioblastoma was examined in 52 newly diagnosed and nine recurrent tissue samples (Supplementary Table 1). TβRIII protein levels were analyzed separately in the tumor and normal brain tissue versus the endothelium. Quantitative assessment of non-endothelial regions of tumor and normal brain tissue separately from the respective endothelial regions revealed higher TβRIII levels in the vasculature of glioblastoma (p < 0.001) and of normal brain (p < 0.001) than in the tumor tissue proper or the parenchymal brain cells (Fig. 1a). Representative paraffin sections of four glioblastoma and four normal brain patients stained for TβRIII are shown in Figure  S1. Survival data were available for all 52 glioblastoma patients assessed for TβRIII staining in the tumor cells and for 48 patients assessed for TβRIII staining in the endothelial cells. Comparison of the median overall survival (OS) of glioblastoma patients with high versus low TβRIII levels (cut-off defined by median TβRIII levels) revealed no differences, neither for patients with high (median OS 17.7 months, 95% confidence interval (CI) 11.8-23.5 months) versus low (median OS 15.8 months, 95% CI 5.8-25.8 months) staining in the tumor cells (p = 0.88) ( Figure S2A) proper nor for patients with high (median OS 19.0 months, 95% CI 8.3-29.7 months) versus low (median OS 17.7 months, 95% CI 0.0-53.5 months) staining in the glioblastoma vasculature (p = 0.36) ( Figure  S2B).
Representative normal brain and glioblastoma tissue sections with the endothelial marker vWF and TβRIII stained by co-immunofluorescence confirmed a predominant vascular localization both within the normal brain and the glioblastoma sections (Fig. 1b). TβRIII levels in tumor cells positively correlated with TβRIII levels in endothelium in  Table 2). In a previous study, we found a localization of CD31 restricted to the tumor blood vessels in glioblastomas [21], defining CD31 as an endothelial marker in glioblastoma. As a confirmation that TβRIII is more abundant in tumor endothelium than in the tumor cells, we used the endothelial cell marker CD31 and compared matched CD31 + and CD31 − cell populations from four freshly dissociated primary glioblastoma tissues. In three out of four tumors, TβRIII expression was higher in the CD31 + cell fraction (Fig. 1c). Endothelial-like (CD31 + ) cell lines previously established in our laboratory [22] also released sTβRIII into the supernatant (Fig. 1d). TβRIII mRNA expression with molecules associated with TGF-β superfamily signaling such as all three TGF-β isoforms, TGF-βRII, the TGF-β target gene serpine1 and molecules associated with angiogenic signaling such as VEGF, and associated receptors such as vascular endothelial growth factor receptor (VEGFR)1, VEGFR2, neuropilin (NRP)1, and NRP2. In the CD31-positive subpopulation, there was a positive correlation of TβRIII and further molecules linked to TGF-β signaling such as ALK-1, aryl hydrocarbon receptor (AhR), the proTGF-β processing enzymes proprotein convertase subtilisin/kexin type (PCSK) 5, PCSK 7 and furin, and the TGF-β target genes Id (inhibitor of DNA binding) 1 and Id3 (Fig. 2b).

TβRIII expression and release in human glioma cell lines
We assessed TβRIII levels in a panel of human long-term cell lines (LTC), glioma-initiating cell (GIC) cultures, and cultured human brain-derived microvascular endothelial cells (HCMEC). TβRIII was consistently expressed in LTC and GIC on mRNA (Fig. 3a) and protein levels as assessed in total cellular lysates (Fig. 3b). By flow cytometry, we ensured the localization of TβRIII on the cell surface (Fig. 3c). Moreover, sTβRIII was detected in the supernatant (Fig. 3d) at levels correlating with the levels in cellular lysates, indicating constitutive shedding of the receptor in the cell lines investigated (r = 0.69; p = 0.0059). Next, we evaluated the effect of rhTGF-β2 and of TGF-β signaling blockade using the TGF-βRI inhibitor SD-208 [23]. Upon TGF-β2 stimulation, TβRIII mRNA expression was reduced in a concentration-dependent manner in LN-229 cells.
To analyze the specific effects of sTβRIII, we overexpressed a mutated form of TβRIII comprising only the extracellular domain of TβRIII. This had no effect on baseline constitutive pSMAD levels, but increased pSmad2 levels in response to TGF-β2 in LN-229 cells (Fig. 4c, left panel lanes 2 versus 5); furthermore, increased baseline and TGF-β2-evoked pSmad2 levels were observed in ZH-161 cells overexpressing sTβRIII (Fig. 4c, right  panel, lanes 2 versus 5 and 1 versus 4). Overexpression  (Fig. 4c, left panel, lanes 3 versus 6), whereas baseline and TGF-β2-or BMP-4-evoked pSmad1/5 levels were increased in ZH-161 cells overexpressing sTβRIII (Fig. 4c, right panel, lanes 3 versus 6). To investigate a possible differential modulating effect of sTβRIII on TGF-β1-versus TGF-β2-mediated signaling, we costimulated with TGF-β1 when supernatant from LN-229 cells with shed TβRIII was transferred onto wildtype LN-229 cells. Still, an increase in pSmad2 also occurred when costimulation with TGF-β1 was performed (Fig. 4d). In LN-308 shed, TβRIII did not modulate Smaddependent signaling when we transferred supernatant with high levels of shed TβRIII onto this cell line (Fig. 4d, right   f TβRIII protein levels in whole cell lysates of LN-229 cells were detected by immunoblot following the treatments described in e. g and h Shed TβRIII levels in LN-229 glioma cell supernatants following the treatments as described in e were determined by ELISA (g) and by immunoblot (h), ponceau S staining is shown as loading control in h panel), consistent with the lack of effect when TβRIII was depleted (see Fig. 4a).

(Soluble) TβRIII promotes experimental glioma growth in vivo
Finally, we studied whether the biological effects of altered TβRIII availability translated into altered glioma growth in vivo. In vitro, LN-229 overexpressing sTβRIII showed no differences in proliferation or clonogenicity (data not shown). LN-229 cells stably overexpressing sTβRIII were implanted into the right striatum of athymic CD1 nude mice. Cell lines generated ex vivo from micebearing LN-229 sTβRIII or control tumors at the time of sacrifice confirmed stable overexpression of sTβRIII (Fig. 5a). Assessment of tumor volumes of three mice of each group when the first clinical symptoms occurred showed increased tumor volumes (Fig. 5b) and increased pSmad2 levels in the tumors whereas pSmad1/5 levels were unaffected (Fig. 5c). Survival of nude mice with  orthotopically implanted tumors overexpressing sTβRIII was significantly decreased (Fig. 5d).

Discussion
TGF-β and its receptors TβRI and TβRII have been well studied in glioblastoma [24,25]; however, little is known about the coreceptor TβRIII. TβRIII is ubiquitously expressed on most cell types, however, commonly not on endothelial cells [26]. TβRIII is commonly considered a tumor suppressor [26], since its expression is lost during progression of several tumor entities, e.g., breast cancer [27], non-small cell lung cancer [28], and prostate cancer [29]. Sequestering of TGF-β by shed TβRIII inhibiting downstream signaling may be a mechanism of action [13,18,27,28,30]. However, a tumor-promoting role for TβRIII has been deduced from the reduction of tumorigenicity when TβRIII expression was reduced in metastatic breast cancer cells [31] and in human breast cancer stroma [32]. In high-grade non-Hodgkin`s lymphomas [33], colon cancer [34], and B-cell chronic lymphatic leukemia [35], TβRIII expression is increased and not lost during cancer progression and may increase tumorigenicity [34]. A dichotomous role of TβRIII has been described in the context of lung cancer where an extracellular mutant of TβRIII with enhanced ectodomain shedding reduced tumorigenicity of the respective tumor cells but increased their growth rate in vitro and in vivo [36]. A study revealed reduced levels of TβRIII in the tumor stroma. Further, there were distinct paracrine roles of sTβRIII in the tumor microenvironment depending whether it was derived from normal or cancer tissue what emphasizes the complex regulation of availability of cytokines in the ECM [32]. Gene therapy with sTßRII and sTßRIII pointed towards a possible therapeutic role of sTβRII and sTβRIII [37], what led us to perform the current profound study on TβRIII in glioblastoma. TβRIII levels in newly diagnosed glioblastoma samples and of normal brain showed predominantly endothelial localization (Fig. 1a, b). Analysis of freshly dissociated glioblastoma tissues revealed higher mRNA levels in the endothelial fraction of the tumors (Fig. 1c), and endothelial cell lines isolated from freshly dissected glioblastoma samples also secreted the shed form of TβRIII (Fig. 1d). Single-cell RT-PCR from freshly dissociated tissue of six glioblastoma patients confirmed this prevalence in the vascular compartment. Both in the vascular (CD31-and αSMA-positive cells) and in the non-vascular compartment, TGF-βRIII correlated with molecules associated with TGF-β superfamily signaling such as LTBP and FN which are both involved in the process of extracellular activation of TGF-β signaling [20,21] (Fig. 2b). In line with previous studies on embryonic mouse cells [38], levels of cellular and shed TβRIII correlated in our cell line panel (Fig. 3b, d). As reported for ovarian and breast cancer cell lines [39], TGF-β negatively regulated TβRIII expression in a TβRI-dependent manner (Fig. 3e-h). Here, TGF-βRI inhibition increases membrane-associated TβRIII levels (Fig. 3f), but levels of sTβRIII were unaffected (Fig. 3g,  h). In light of this finding, we speculate that in addition to negatively regulating TβRIII expression, TGF-β exerts a posttranscriptional effect on TβRIII by increasing shedding and, thus, increasing its release from the cell membrane. Interestingly, expression of TIMP2, an inhibitor of TβRIII shedding [17], is reduced in human glioma cell lines exposed to TGF-β2 [40].
In many models, TβRIII is described as a dual modulator of TGF-β signaling with cell surface TβRIII enhancing signaling and shed TβRIII acting as an antagonist by ligand sequestration [41]. More recently, shedding-independent inhibition of signaling exerted by the cytoplasmic domain, which sequestered type I and type II TGF-β receptors, has been described [42]. Depletion of TβRIII by transient (Fig. 4a) or stable (Fig. 4b) knockdown exerted an inhibitory effect on Smad signaling in the glioma cell line LN-229 as well.
To better distinguish effects of membrane associated versus soluble TβRIII, we overexpressed sTβRIII in LN-229 and ZH-161 cells. Surprisingly, sTβRIII activated Smad2 signaling in both models (Fig. 4c, d) and decreased survival in LN-229 tumor-bearing mice (Fig. 5). In the glioma cell line LN-308, we observed neither an effect on Smad signaling upon depletion of TβRIII (Fig. 4a) nor upon exposure to sTβRIII (Fig. 4d). Since LN-308 is a TGF-β-driven cell line as defined by high levels of furin, active TGF-β2, and pSmad2 [19,43], it might not require additional coreceptor activity. Indeed, in a classical TGF-β activity assay with the TGF-β reporter cell line Mv1Lu (ATCC® CCL-64), sTβRIII enhances TGF-β bioactivity, too. Here, a fragment of sTβRIII comprising one fourth of the ECD which is closest to the membrane-spanning segment increased binding of TGF-β to TGF-β receptor II [44]. Similarly, soluble endoglin, a TGF-β type III receptor similar to TβRIII, does not inhibit TGF-β superfamily signal transduction but binds to circulating BMP-9 and induces signaling on endothelial cells [45]. There may be a strong dependency of TβRIII`s effects on the cellular context, e.g., presumably including the presence of other TGF-β superfamily ligands and receptors. Indeed, while sTβRIII inhibited BMP-4/Smad1/5 signaling in LN-229 cells, it activated BMP-4/Smad1/5 signaling in ZH-161 cells and did not modulate signaling in LN-308 cells (Fig. 4). The important role of the balance of shed and cell surface TβRIII for Smad1/5 signaling has been studied in breast cancer models [13,18]. We observed an activating role for sTβRIII for Smad2 phosphorylation both in TGF-β1and TGF-β2-mediated signaling in glioma cells (Fig. 4d). The potency of sTβRIII as a protumorigenic factor in glioma cells was confirmed in vivo using xenografts of LN-229 control versus LN-229 sTβRIII-overexpressing cells in nude mice. Orthotopically implanted sTβRIII-transfectants formed larger tumors (Fig. 5b) and had increased levels of pSmad2 in their tumors (Fig. 5c). In line, the animals had shorter survival compared to the controls (Fig. 5d). Previous work has pointed towards a role of TGF-β pathway activity in glioma vessels [6,25,46], and glioblastomas are one of the most vascularized tumors [24,47,48]. Endothelial cells might provide large amounts of sTβRIII within the tumor to promote TGF-β2/Smad2 signaling in glioma cells and thereby promote tumorigenicity. Our study may have implications for approaches of pharmacological targeting of TGF-β superfamily ligands using TGF-β binding traps derived from TGF-β (co-)receptors. Although systemic administration of a sTβRIII ECD has been shown to inhibit tumor growth in prostate and breast cancer models [49,50], an opposite effect is indicated by our glioma model.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http:// creat iveco mmons. org/ licen ses/ by/4. 0/.