1. Introduction
The different members of the transforming growth factor-β (TGF-β) superfamily, which include TGF-β, activins and bone morphogenetic proteins (BMPs), are regulatory cytokines involved in a multitude of biological processes [
1]. In the liver, the prototype of this family, i.e., TGF-β1, is produced in parenchymal and non-parenchymal cells. During liver insults pivotal targets of TGF-β1 are hepatic stellate cells (HSC) and portal fibroblasts which become activated and acquire a myofibroblast-like phenotype, thereby inducing the profibrogenic program [
2]. Due to the pleiotropic effects, the function of this cytokine is strictly regulated. TGF-β1 is stored inactively as a large latent complex in the extracellular matrix. It is proteolytically activated and released and its biological activity is modulated by many soluble factors [
3,
4]. Once released, the ligand docks to receptors located in the plasma membrane, a process which is tightly controlled by secreted proteins such as the connective tissue growth factor (CTGF) [
5].
TGF-β signaling is initiated by binding of the ligand to a receptor complex, which contains type I (TβRI), type II (TβRII) cell surface receptors [
6]. The TβRIs (ALK1 or ALK5) and TβRIIs are serine/threonine kinases with a similar structure containing a cysteine-rich extracellular domain, a short hydrophobic transmembrane domain, and a cytoplasmic region containing the kinase domain. Upon binding to TβRII, TβRI is recruited and becomes trans-phosphorylated by TβRII. Subsequently, the activation of the type I receptor results in endocytosis of the receptor complex, followed by phosphorylation of substrate proteins by TβRI. In general, there are two different ways of internalization involving clathrin coated pits (ccp, EEA-1 positive) or Caveolin coated pits (lipid rafts, Caveolin-1 positive). It is assumed that localization to ccp leads to Smad activation, whereas routing to lipid rafts leads to degradation of receptors or activation of non-Smad signaling [
7,
8]. Activated Smads form complexes with the common Smad4, which then translocate to the nucleus. In association with positive and negative transcription factors, Smads regulate a wide spectrum of target genes [
9].
In addition to the signaling receptors TβRI and TβRII, there are two type III receptors, i.e., betaglycan and endoglin (Eng, also known as CD105), which modulate binding to the signaling receptors. Both receptors are single membrane spanning, glycosylated accessory receptors for members of the TGF-β superfamily and have no intrinsic kinase activity [
10,
11]. In case of Eng, defective
N-glycosylation has been shown to interfere with membrane localization similar to TβRII [
12,
13]. For betaglycan it has been shown that changes in the type of glycosylation affect the signaling outcome [
14]. Betaglycan is broadly expressed and has been shown to positively regulate ALK5/Smad2/3-signaling [
15,
16]. In contrast, Eng expression is more restricted, being highly expressed on endothelial cells and dampening ALK5/Smad2/3, while enhancing ALK1/Smad1/5 signaling [
17]. Differential splicing of Eng creates a shorter variant and imposes an even higher degree of fine tuning to the signaling cascade [
18]. Nevertheless, the exact mechanism(s) on how Eng governs this fine tuning is not known, but it is known to involve a direct interaction of Eng with TβRI and TβRII. We could previously demonstrate this interaction of Eng and TβRII in HSC [
19]. In endothelial cells, Eng association with TβRI (ALK5 and ALK1) and TβRII, as well as the consecutive phosphorylation events on Eng, have been analyzed in detail. The corresponding consequences for response outcomes have been described and affect Smad and non-Smad signaling [
20]. However, if Eng modulates those responses by affecting TβRI and TβRII routing is not known, since Eng trafficking modalities have been only sparsely analyzed so far [
21,
22].
The in vivo importance of Eng for signaling of individual ligands of the TGF-β family becomes evident by the human disease hereditary hemorrhagic telangiectasia type-1 (HHT-1) caused by Eng haploinsufficiency/and dominant negative effects caused by truncation, nonsense and missense mutants of Eng affecting endothelial cell biology [
23,
24]. On a mechanistical basis, mutations cause varying defects which abolish Eng function. Some missense mutations are not capable of ligand binding, while others are dominant negative by sequestering wild type Eng in a heteromeric complex inside the cell [
12,
25,
26]. Several missense mutations cause misfolding, leading to premature proteins defective in final glycosylation most likely due to retention in the ER by the ERAD system [
12,
26].
In addition to the membrane bound receptor, soluble Eng (sol-Eng) has been implicated in HHT-1 [
27]. The extracellular domain of Eng (sol-Eng) can be released from the cell by a regulated shedding mechanism involving matrix metalloprotease-14 (MMP-14) [
28]. The shedded receptor has been initially shown to contribute to the pathogenesis of pre-eclampsia and the sol-Eng content in the serum of pre-eclamptic women is increased [
29]. Although it is claimed that Eng only binds ligands in the presence of signaling receptors, sol-Eng is secreted, can dimerize, is able to directly bind to BMP9 and BMP10, and affects TGF-β1 signaling [
29,
30].
Recently, it was demonstrated that exosomes isolated from the serum of pre-eclampsia patients contain abundant sol-Eng, contributing to the damage of vascular functions and complications in pre-eclampsia patients [
31]. Exosomes are nano-sized membrane vesicles in size from 30–150 nm formed by the inward budding of late endosomes that are released into the extracellular environment. They can contain a diverse array of signaling molecules (e.g., proteins, mRNAs, and microRNAs) and are key players in mediating cell paracrine effects by transferring genetic materials and proteins to target cells [
32]. In regard to the liver, exosomes represent one of the communication routes between the parenchymal cells (PC, hepatocytes) and mesenchymal cells (HSC, pMF) and vice versa [
33]. HSC-derived exosomes contain components, which induce activation of quiescent HSC and metabolic activity of Kupffer cells (KC) and liver sinusoidal endothelial cells (LSEC) to promote fibrosis [
34].
In addition to LSEC in the liver, HSC produce comparable amounts of Eng and its expression during hepatic fibrogenesis is upregulated—especially in HSC [
19,
20,
35]. Genetic depletion of Eng in HSC leads to an aggravation of fibrosis, implying Eng as an antifibrotic receptor [
36]. However, to our best knowledge, there are no reports, showing that full-length Endoglin (FL-Eng) is cargo of exosomes in general and especially in HSC.
We show here that Eng is enriched in Caveolin-1 containing fractions (lipid rafts). In addition, in this study we isolated exosomes from different cellular primary sources of the liver and a number of different immortalized hepatic cell lines. We could show that exosomes derived from primary rat HSC or immortalized Col-GFP cells derived from primary mouse HSC contain large quantities of FL-Eng. Likewise, the human cell line LX-2, originally established by spontaneous immortalization in low serum conditions, had the potential to direct Eng to exosomes when transiently infected with an adenoviral vector expressing FL-Eng or sol-Eng. Moreover, exosome localization did not depend on endogenous Eng expression, since hepatocyte-derived exosomes were positive for heterologously expressed Eng—although PC lack endogenous Eng. Finally, we show that block of N-glycosylation does not inhibit sol-Eng or FL-Eng dimerization but does impede sol-Eng secretion and FL-Eng exosomal localization.
4. Discussion
TGF-β1 is a key player in the pathogenesis of liver disease involving the activation of HSC and pMF as well as the induction of their fibrogenic program, leading to a wound healing reaction. Aside the TGF-β signaling receptors TβRI and TβRII, which are both indispensable for signaling, FL-Eng representing an accessory receptor for TGF-β1 is highly expressed in HSC and upregulated during activation of HSC [
35]. In addition, the overexpression FL-Eng in HSC cell lines causes a higher expression of the activation marker α-SMA and the profibrogenic protein CTGF [
19,
42]. In contrast, genetically manipulated mice lacking FL-Eng in HSC show a higher fibrogenic response in the setting of liver intoxication (carbon tetrachloride) and cholestasis (bile duct ligation). However, siRNA-based or adenovirally-expressed Cre-mediated knockdown of FL-Eng in primary isolated HSC results in reduced CTGF expression [
36].
To solve these discrepant findings, we tried to understand the mechanism of Eng function in more detail. Although membrane localization of Eng has been addressed in a plentitude of studies concerning with hereditary hemorrhagic telangiectasia type 1 (HHT-1), there is only very limited information available about Eng trafficking in liver cells. Since the localization of receptors to membrane subdomains (i.e., caveolae or ccp) is vital for signaling outcome, we first sought to locate Eng in those subdomains via sucrose density gradient analysis. In murine endothelial cells, this analysis revealed a coupling of Eng and eNOS in caveolae to regulate eNOS stability and activity. In these studies, Eng was exclusively localized to lipid rafts [
53]. In contrast, our analysis showed a distribution of Eng in lipid rafts as well in non-raft fractions. However, there is a higher abundance of Eng in raft fractions compared to non-raft fractions (
Figure 1). The importance of Caveolin-1 (and therefore lipid rafts) for HSC activation and biology has been emphasized in recent study by others, showing that overexpression of Caveolin-1 in a HSC cell line (GRX) affected divers functions of HSC including activation, cell cycle, profibrogenic behavior and migration [
54]. Therefore, we are currently investigating whether Eng overexpression or depletion in HSC as well as depletion of cholesterol changes the localization and distribution of the signaling receptor TβRI and TβRII and finally TGF-β-signaling in those subdomains.
Extracellular vesicles are lipid-bilayer enclosed vesicles secreted by virtually all cell types contributing to cellular communication by shuttling of a large variety of cargo molecules [
55]. They play key roles in both normal and disease processes and can influence biological functions in cells located in close vicinity or at distant locations via transfer through the systemic circulation [
55].
In our study, we prepared “exosomes” which have a typical size of 30–100 nm and were sedimented at 100,000×
g [
45]. However, it should be critically mentioned that the usage of the term “exosome” is somewhat vague and should be used with caution [
56]. Based on the suggestions published by the International Society for Extracellular Vesicles (ISEV) the term exosome can be used in three different ways [
57]. Some researchers argue that the term exosome refers to vesicles that bud into endosomes and are released when the resulting multivesicular bodies fuse with the plasma membrane. Others argue that exosomes are secreted vesicles that may have a physiological function, while the third researchers classify exosomes as particles that only sediment after centrifugation at ~70.000–100.000×
g.
In regard to liver, several biological functions have been attributed to exosomes. Hepatocyte-derived exosomes from primary human hepatocytes were recently shown to promote liver immune tolerance [
58]. Another report has recently shown that cholangiocyte-derived exosomes enriched by the noncoding RNA-H19 enhance transdifferentiation of cultured mouse primary HSC and promote progression of cholestatic liver fibrosis [
59]. Similarly, the delivery of exosomes released from activated HSC can provoke metabolic switches in nonparenchymal liver cells affecting glucose metabolism by delivery of glycolysis-related proteins [
34]. Exosomes derived from HCC cell lines include many proteins, microRNAs, long noncoding RNAs, mRNAs, and DNAs [
50]. Therefore, it was proposed that some exosomes may be potential diagnostic biomarkers for early-stage hepatocellular carcinoma (HCC) [
60].
On the other hand, there is growing evidence that exosomes function as conduits for the intercellular transfer of components to induce resolution of hepatic fibrosis by inhibiting macrophage activation, cytokine secretion, modulation of extracellular matrix, and inactivation of HSC [
33]. Experimentally, it was shown that delivery of miRNA targeting CTGF can suppress fibrogenic signaling in human HSC [
61]. It is obvious that exosomes are therefore potentially of fundamental importance for the therapy of fibrotic liver lesions and to interfere with processes relevant in the pathogenesis of HCC.
This is underpinned by the finding that microvesicles derived from Eng-positive cancer stem cells can confer an activated angiogenic phenotype to normal human endothelial cells and stimulate proliferation and vessel formation [
62], which is also a key process in HCC.
In our study, we tested whether individual liver cells form exosomes containing endogenously expressed Eng or have the capacity to direct overexpressed Eng to the exosomal compartment. Therefore, we have cloned adenoviral expression constructs expressing either FL-Eng or sol-Eng (
Figure 2 and
Figure 3,
Supplementary Figure S2) and tested for the functionality of FL-Eng and sol-Eng in comparison with our previously published results and literature data. In addition, we tested if proper glycosylation of Eng is a mandatory need to be secreted for the artificial sol-Eng and targeted to exosomes for FL-Eng. In respective experiments, we either prevented proper glycosylation in liver cells by treatment with tunicamycin which inhibits
N-linked glycosylation by preventing core oligosaccharide addition to nascent polypeptides, or alternatively by overexpressing FL-Eng in the Chinese hamster cell line CHO-Lec 3.2.8.1, allowing to produce glycoproteins in which all
N-linked carbohydrates in the Man5 oligomannosyl form and O-linked carbohydrates are truncated to a single GalNAc [
44].
We first isolated and characterized exosomes in regard to the existence of typical exosomal marker proteins from several primary cultured liver cells and immortalized derivatives thereof. Interestingly, the subset of exosomal markers (CD81, Alix, Cav-1) usually recognized as exosomal markers were not found in all liver cells. In contrast to a previous report [
63], CD81, which belongs to the group of endosome-specific tetraspanins having a broad tissue distribution and known to be enriched in exosomal membranes [
64] was not detected in HepG2 cells in our study (
Figure 3,
Table 2). However, other cells tested (LX-2, Col-GFP, rat HSC, mouse HSC, rat pMF) expressed large amount of this marker. Alix, which is involved in exosome biogenesis and exosome sorting and common to most exosomes, was found in all liver cells tested (
Figure 4,
Figure 5 and
Figure 6 and
Figure 8). Cav-1, a multifunctional membrane protein typically upregulated in the final stages of cancer and associated with promotion of migration and invasion [
65] was found to be expressed in LX-2, Col-GFP and rat pMF, while in primary HSC from rat and mouse, as well as human HepG2, this scaffolding protein was virtually absent.
Primary rat HSC, Col-GFP representing a cell line derived from murine HSC, and rat pMF showed high endogenous expression of Eng, confirming our previous reports [
19,
20,
35]. Nevertheless, viral-mediated overexpression leads to an even higher abundance of FL-Eng in these cells (
Figure 2C–E). Similarly, sol-Eng is highly expressed in LX-2 and HepG2 upon viral transduction. Respective cells were also capable to target FL-Eng and sol-Eng to exosomes (
Figure 5A and
Figure 6E).
In contrast, HepG2 cells, which lack endogenous Eng, were able to direct adenovirally expressed FL-Eng and sol-Eng to exosomes (
Figure 5). This suggests that the endosomal sorting complex required for transport (ESCRT) during exosome biogenesis [
57] of all liver cells have the capacity to sort Eng into exsomes. However, although we could show exosomal localization of sol-Eng by overexpression of the truncated FL-Eng, we were not able to detect shedded sol-Eng originating from the expressed FL-Eng (
Figure 5). This is unexpected, since the effector protease is expressed in the analysed cells and is co-localized with FL-Eng in exosomal fraction (
Figure 4 and
Figure 5) [
36,
66].
In previous work from other groups, defined glycosylation signatures identified by glycomic analysis were shown to be enriched in exosomes [
67]. Based on the fact, that Eng contains four verified
N-linked glycosylation sites and a potential O-linked glycosylation site [
11,
12], we next tested if glycosylation of Eng is mandatory for exosome targeting. We found that the blockade of
N-glycosylation by tunicamycin resulted in a significant molecular shift of FL-Eng in gel electrophoresis indicative of lower glycosylation in primary mouse HSC and LX-2 (
Figure 8). Due to the experimental setting, both fully and low glycosylated species are detected. However, only fully glycosylated FL-Eng is found in the exosomal fraction, but not the lower molecular weight species (
Figure 8). Lux et al. showed that all four
N-glycosylation sites are used, and that eliminating one or two sites does not affect membrane targeting, but loss of three sites nearly abolishes surface expression [
12]. Although exosomal localization of FL-Eng is lost when not properly glycosylated, the dimerization of this receptor still occurs (
Figure 7E). Nevertheless this previous report has not given details about surface localization of tunicamycin treated FL-Eng. Based on our experiments, we cannot exclude that the unglycosylated FL-Eng does not reach the plasma membrane. However, a general arrest of proteins in the ER induced by tunicamycin can be excluded as at least LCN2 is secreted in both glycosylated and none glycosylated forms [
45].
In line with the missing localization in exosomes of deglycosylated FL-Eng we were not able to demonstrate secretion of deglycosylated sol-Eng—neither in COS-7, nor in HSC Col-GFP (
Figure 7C,G). This points strictly towards an intracellular sequestration of the deglycosylated Eng and a loss of function. For the signaling receptor TβRII, loss of proper
N-glycosylation is associated with abolished receptor function [
13], which we assume also for Eng. Nevertheless, the tunicamycin-induced unfolded protein response does not necessarily block TGF-β1 signaling as shown for HSC Col-GFP (
Figure 7E,
Supplementary Figure S3B). Moreover, it can also activate those responses in a cell type-specific manner [
68]. In the case of COS-7 cells, those immature proteins potentially have a low half-life since it was difficult to display these lower migrating species. In contrast to COS-7 cells, the glycosylated and deglycosylated Eng forms are present in equal amounts after tunicamycin treatment, which implies that those immature receptors are stabilized in HSC Col-GFP.
In our view, the finding that Eng is part of extravesicular vesicles is extremely important. These particles have gained interest due to their roles in cell-to-cell communication and potential roles of exosomes are discussed for many liver diseases, including alcoholic fatty liver disease/non-alcoholic fatty liver disease [
69], hepatitis infection [
70], hepatocellular carcinoma [
60], and liver metastasis [
71]. Several aspects in these malignancies such as immune cell infiltration and fibrogenesis are majorly driven by members of the TGF-β superfamily. In addition, it is well accepted that exosomes isolated from women with preeclampsia can effectively transfer sol-Eng to endothelial cells, where it provokes significant cell damage [
31]. In regard to liver, such mechanisms were not described before, but the eminent importance of Eng and exosomes in the setting of hepatic diseases is undoubtedly. Even more, there are no reports available, which systemically analyze different liver cells for their potential to secrete Eng-containing exosomes.
However, our study has some limitation because we have no data pointing to the function of exosomal Eng. The function of cargo-loaded Eng will of course depend on its absolute concentration. Other factors are the potential target cells, the physiological or pathological state of the organism or organ, concentrations of Eng ligands being present, the capacity to degrade them, and many other factors. We are aware of the fact that these questions must be answered in future studies.
So in summary, our study showed that: (i) Eng is present in lipid rafts; (ii) all liver cells have the capacity to direct Eng to exosomes, irrespectively if they express endogenous Eng; (iii) proper Eng N-glycosylation is mandatory for trafficking of Eng leading to secretion of sol-Eng and exosome targeting.