Neurons produce FGF2 and VEGF and secrete them at least in part by shedding extracellular vesicles

Introduction

We previously found, by using a transwell co-culture system, that neurons and astrocytes have synergistic effects on reduction of brain capillary endothelial cell permeability to radioactive sucrose, used as a marker of paracellular flux [1–5].

A further point was that in the three-cell type system of culture, neurons, astrocytes and endothelial cells were not in physical contact. Therefore, the effects discovered should be due, at least in part, to molecular mechanisms not based on cell–cell contacts. As independent experiments showed that transformed glial cells (oligodendroglioma cells) are able to shed membrane vesicles [6], we wondered whether extracellular vesicles could be also the way adopted by neurons (this paper) and/or astrocytes (manuscript in preparation) to address inductive signals to endothelial cells.

Many eukaryotic cell types have been indeed shown to release into the extracellular environment vesicles of different sizes and composition which have been differently defined [for recent reviews, see: 7–9]. The first population of exovesiscles to be described seemed to originate from the cell plasma membrane through a mechanism similar to virus budding [10]. By electron microscopic analyses, vesicles shed by circulating human lymphocytes [10] and by in vitro cultured human breast carcinoma cells [11] were found to be large and heterogeneous in size (100–1000 nm). Extracellular vesicles of this kind, called microvesicles (MVs), are released by several kinds of viable cells through an active process that require RNA and protein synthesis [10, 12] and seem to originate from domains of the plasma membrane selectively enriched in specific membrane components, such as β1 integrin and membrane-bound gelatinase B/MMP-9 [see, e.g. 13]. Other vesicles, known as exosomes, are smaller (30–90 nm) and seem to originate from the endosomal system by exocytosis of cell structures, called multivesicular bodies [MVBs; 14, 15]; exosomes are released by most cells of the haemopoietic lineage and contain several cytosolic proteins (including heterotrimeric G protein Gi2α, annexin II and heat shock cognate protein Hsc73), as well as different integral and peripheral membrane proteins [15]. Both MVs and exosomes are involved in cell-to-cell communications. The phenomenon of shedding was first demonstrated in tumour cells, where shedding was related to the invasive potential of the tumour [16–19]. The presence in shed vesicles of numerous extracellular matrix (ECM)-degrading proteolytic enzymes can also have an effect on angiogenesis [20]. However, vesicles promote angiogenesis [21, 22] also because they vehicle numerous pro-angiogenic signalling molecules. For example, fibroblast growth factor 2 (FGF-2) was demonstrated to be externalized from the cells as a component of shed membrane vesicles [12]; and vascular endothelial growth factor (VEGF), perhaps the most effective angiogenic factor known at present, also appears to be present in vesicles [23].

FGF-2, also known as basic FGF, is a non-glycosylated 18 kD polypeptide with very high affinity for heparin, that is secreted in spite of lacking a conventional secretory peptide and being independent of the classical endoplasmic reticulum-Golgi pathway [12]. It is a potent inducer of blood vessel formation, involved in the pathogenesis of several diseases characterized by abnormal neovascularization [24] and in a cancer [25, 26], as well as in development and differentiation of various tissues [27, 28], including the nervous system [28]. In addition to the 18-kD peptide, larger isoforms of FGF-2, ranging in size from 22 to 34 kD, and with different sub-cellular localization have been described [28]. While the 18-kD isoform is cytoplasmic and secreted, the larger isoforms, that derive from unusual translation initiation at CUG codons, have been reported to be predominantly intracellular and, in particular, nuclear [28].

Like FGF-2, VEGF, also known as vascular permeability factor (VPF), is a potent mitogen of endothelial cells, involved both in pathological angiogenic processes and in the embryonic development and differentiation of the vascular system [29]. VEGF expression is regulated not only at the transcriptional level but also at a post-transcriptional level, by alternative splicing, which generates in human beings 5 isoforms (of 121, 145, 165, 189 and 206 residues, respectively). The major isoform is VEGF₁₆₅, a basic, heparin-binding glycoprotein, existing as a homodimer of 45 kD. It has been reported that VEGF₁₂₁ is a freely soluble protein, while VEGF₁₆₅, even if also secreted, tends to remain “bound” to the ECM [for review, see 29]. Interestingly, VEGF₁₈₉ and VEGF₂₀₆ seem to be almost completely sequestered in the ECM [30]. In addition to these shorter isoforms, a further high molecular weight isoform (large VEGF), containing an N-terminal extension of about 200 residues and with a monomeric mass of about 45 kD, has been demonstrated in human beings and mouse [31, 32]. This isoform is generated by an unusual translation initiation process, which, like in the case of FGF-2 larger isoforms, occurs at one of four potential CUG translation initiation codons, positioned in between two ribosome internal entry sites (IRES), and in frame with the classical AUG start codon [31, 32]. Complexity of this family of growth factors is further increased by proteolitic cleavage of the longer isoform into N-terminal and C-terminal fragments [31, 32] and glycosylation [31, 33].

Several lines of evidence indicate that VEGF influences early stages of vascularization of the developing brain and that VEGF inactivation in the nervous system results in abnormalities of vessel density and cell number in the cerebral cortex. Moreover, it was reported that reduced levels of VEGF caused ALSlike motoneuron degeneration in mice [reviewed in 34]. This latter finding suggested that, in addition to its specific angiogenic action, VEGF might also have neuroprotective functions.

FGF-2 was also demonstrated to play a role in hippocampal neurogenesis [35] and to have neurotrophic functions both in vitro and in vivo[see, e.g. 36–38]. Interestingly, there are also evidence of a cross-talk between FGF-2 and different members of the VEGF family during angiogenesis and vasculogenesis [39].

In this paper, we report that neurons synthesize growth factors FGF-2 and VEGF and release them at least in part via extracellular vesicles.

Materials and methods

Results

We previously found that, in a three-cell type system of culture, containing neurons, astrocytes and endothelial cells, both neurons and astrocytes influence the ability of endothelial cells to form a barrier with properties resembling those of BBB [3, 4]. As in that system cells were not in physical contact, we deduced that the effects discovered should be due, at least in part, to molecular mechanisms not based on cell-to-cell contacts. Moreover, independent experiments showed that oligodendroglioma cells shed membrane vesicles [6]. Thus, we asked whether extracellular vesicles could be also produced by neurons to address inductive signals to endothelial cells. A further question was which kind of signals neurons deliver to them. As two of the major factors acting on endothelial cells (i.e. FGF-2 and VEGF) have been reported to be released through extracellular vesicles in other cell systems, we looked for the possible existence of extracellular structures, produced by neurons, carrying the two factors. As shown in Fig. 1A, there is evidence that neuronal cultures contain vesicular structures with sizes ranging from 400 to even 2000 nm. Most of these vesicles are in the range of size (100–1000 nm) already reported for the extracellular structures produced by several kinds of viable cells [10–12] and called MVs. As some of the structures observed are very large, we also looked for the presence of whole nuclei or nuclei fragments in them. We found that only intact nuclei, but not the putative MVs, were stained by DAPI (Fig. 1Ab).

We then purified the vesicular fraction (Fig. 1B) from the medium conditioned by neurons, according to the protocol used by Dolo et al. (1994) for vesicles released by tumour cells. This fraction, together with total neuronal lysates, was analyzed by Western blot. One example of this analysis is reported in Fig. 1C, which shows that bands of the expected range of size for both VEGF- (Ca: arrowheads) and FGF-2- isoforms (Cb: arrows) are present both in cell lysates and vesicles. Interestingly, the patterns of proteins are not exactly identical when cell lysates and vesicles are compared. In more detail, among the main bands immunostained by anti-VEGF antibodies, two bands of about 21 and 26 kD, respectively, together with one of about 50 kD, are more concentrated in vesicles (lane 2) than in lysates (lane 1), while, on the contrary, a protein with apparent mass of 42 kD is more concentrated in total cell lysates (lane 1) than in vesicles (lane 2). On the other hand, anti-FGF-2 antibodies recognize three main proteins, with apparent masses of about 18, 22 and 32 kD. Among these proteins, two (18 kD and 32 kD) are apparently present only in vesicles (lane 2), while one (22 kD) is present both in cell lysates and vesicles. In order to investigate the fate of vesicle proteins from biosynthesis to appearance in vesicles, neurons at the 15^th day of culture were also metabolically labelled with ³⁵S-methionine. After labelling for 24 hrs, the ³⁵S methionine- containing medium was washed away and neurons were cultured for further 24 or 48 hrs, in unlabelled complete Maat Medium. Conditioned medium was finally collected and centrifuged to obtain labelled vesicles. When the last conditioned medium was collected, cells were also lysed to obtain in parallel total labelled cell proteins. As we did not find any difference in the pattern of radioactive proteins, either in cell lysates or in vesicles, at the different times at which media had been collected, only radioactive proteins recovered 48 hrs after incubation with ³⁵S-methionine are shown (Fig. 1Cc). Interestingly, at least one of the putative VEGF- (arrowhead) and one of the putative FGF-2- (arrow) isoforms, respectively, seem to be radioactive (Fig 1Cc).

The fact that neurons produce structures resembling MVs has been also confirmed by scanner electron microscopy (Fig. 2). As shown in Fig. 2B, some of these structures are very large, around 2000 nm (see below).

MVs have been reported to originate from domains of the plasma membrane selectively enriched in specific membrane components, such as β1 integrin and membrane-bound gelatinase B/MMP-9 [13]. Therefore we investigated, by double immunofluorescence and confocal microscopy, the possible colocalization of both FGF-2 and VEGF with β1 integrin. The results confirmed, first of all, that neurons produce both FGF-2 and VEGF. As shown in Fig. 3, indeed, neurons were immunostained by both anti-FGF-2 antibodies (b, b*) and anti-VEGF antibodies (e, e*). Second, we found that both molecules are at least in part present in round extracellular structures, emanating from neurons, also stained by anti-β1 integrin antibodies (a, a*; d, d*). The sizes of the extracellular structures, visible in Fig. 3, are comparable with the sizes of those visible in Figs. 1 and 2. We noticed in some cases a tendency of vesicles to cluster around neurites (see, e.g.Fig. 3f*, structures indicated by arrows).

As we know that neuronal cultures are contaminated by some rare astrocytes (Fig. 4), we also looked by immunofluorescence at colocalization in the same cell of different neuron-specific cytoskeletal proteins such as NF68 and microtubule-associated protein 2, MAP2) and either FGF-2- or VEGF- immunoreactivity. One example of this kind of analysis was reported in Fig. 5, that shows neurons immunostained for both MAP2 (A: red fluorescence) and VEGF (B: green fluorescence). Extracellular vesicles are clearly seen, attached to MAP2-positive cells.

Finally, as neurons in our cell culture system are post-mitotic and are fed with a serum-free medium, the possibility existed that vesicle production was the effect of cell death. This possibility was also suggested by the large size of vesicles. In order to compare vesicles produced by vital cells with vesicular structures (such as apoptotic bodies), produced by dying cells, we treated neurons with campothecin, an S-phase-specific topoisomerase I inhibitor that has been shown to induce significant, dose-dependent cell death of post-mitotic rat cortical neurons in vitro[42]. Camptothecin-induced neuronal death was already shown to be apoptotic, as characterized by chromatin condensation, cytoplasmic shrinking, plasma membrane blebbing, and fragmentation of neurites [42]. Camptothecin induced massive cell death also in our cell culture system (Fig. 6): in most fields, neurons were fragmented into putative apoptotic bodies (indicated by arrows in Fig. 6B and D). These cell fragments appeared orange/red when cells were stained with a mixture of the fluorescent DNA-binding dyes AO and EO (Fig. 6B). On the contrary, nuclei of control neurons (Fig. 6A) appeared intact and bright green (as expected for alive cells). More important, vesicular structures deriving from cell death were also stained by DAPI (Fig. 6D: structures indicated by arrows), while the putative extracellular vesicles identified by us in normal cultures were not stained (Figs 1A and 6C).

Discussion

Many independent lines of evidence suggest that shedding of extracellular vesicles is a general route, largely adopted to mediate cell-to-cell communications [7–9]. Although the biochemical mechanisms underlying the phenomenon are largely unknown, some proteins, among which β1 integrin, have been consistently localized to extracellular vesicles [13]. Moreover, it has been suggested that extracellular vesicles can mediate secretion of proteins that lack standard signal sequence and cannot be sorted the endoplamic reticulum. One of the secreted factors that lack a signal sequence is FGF-2, which has been reported to be included into extracellular vesicles shed by tumour cells [12].

We previously reported that RBE4.B brain capillary endothelial cells, cultured on collagen IV, form a barrier with BBB permeability properties, if co-cultured cultured with astrocytes and/or neurons [2–4]. In the co-culture systems, neurons, astrocytes and endothelial cells were not in physical contact. Therefore, we assumed that the effects discovered were due, at least in part, to molecular mechanisms not based on cell–cell contacts.

All of these considerations prompted us to investigate: (i) the possibility that neurons produce extracellular factors responsible for addressing signals to endothelial cells; (ii) the putative presence, among these factors, of angiogenic proteins, such as FGF-2 and VEGF and (iii) possible release of these factors through extracellular vesicles. The study described in this paper demonstrates by different techniques that neurons do shed extracellular vesicles of different sizes, from about 400 nm to over 2000 nm. Because of their dimensions, but also because they contain β1 integrin, these structures should be ranked as MVs.

Their very large size, however, induced us to investigate possible relationships between these structures and structures arising from dead cell fragmentation (e.g. apoptotic bodies). In order to directly compare extracellular vesicles and dead cell fragments, we treated neurons with campothecin, a topisomerase I inhibitor already known to induce apoptosis in postmitotic rat cortical neurons in culture [42]. We then looked by immunofluorescence at both morphological integrity of cell structures (first of all, nuclei and neurites) and the availability of nuclear DNA to be stained by different dyes. We did not find any evidence of DNA presence in extracellular vesicles observed in control cultures, not treated with camptothecin. Moreover, these cells appeared quite intact, with unfragmented nuclei and intricate branches of neurites. Extracellular vesicles shed from these wellshaped, alive cells, were not found to be stained by DAPI (Fig. 1). On the contrary, apoptotic and possibly necrotic cells entirely dissolve into structures that are stained by DAPI. In other words, in this case, there is no evidence of extracellular structures pinching off from otherwise intact cells. These findings corroborate the idea that the phenomenon of shedding associates with viable cells, in accord with previous reports demonstrating that cells are more active in producing extracellular vesicles when they are fed with serum-rich media [7].

We also demonstrate that the extracellular vesicles shed by neurons contain proteins of different sizes which react with either anti-VEGF- or anti- FGF-2-antibodies. Unfortunately, most commercially available antibodies are rinsed against portions of the two growth factors that are present in all the putative isoforms of them; this fact, together with the generally low concentration of the factors and their great variability, results in a complex pattern in the western analyses. Importantly, however, even if some bands of unknown origin are present, the main observable bands show apparent molecular masses compatible with those of VEGF and FGF-2, respectively. Moreover, some of these isoforms are clearly enriched in vesicles respect to cell lysates; curiously, we found in vesicles not only isoforms, such as the 18 kD FGF-2, normally secreted [30] but also isoforms that have been reported to be retained in the cells.

Vesicles released by neurons could interact and/or fuse with other vesicles and/or dock to specific components of ECM, thus releasing the factors to target cells only slowly. Actually, the fate of vesicles in the extracellular environment is not yet clear: they might fuse with target cells, thus directly delivering their contents into the cells or, alternatively, bind to specific membrane receptors or co-receptors, thus eliciting a signal transduction pathway. In any case, tendency of neurons to release extracellular vesicles add new strategies to their communication skills. One could indeed suppose that, beside influencing target cells through traditional transmission by neurotransmitters and neuromodulators at the level of synapses, neurons can also modulate their own survival and functioning by releasing growth factors, that can either directly act on neurons themselves or influence differentiation and functioning of other cell types, first of all astrocytes and endothelial cells. Shedding of growth factor-containing extracellular vesicles might allow protection of the factors in the ECM and/or their modulated release to target cells. In contrast with traditional localized neurotransmission, in the course of our analyses, we observed vesicles pinching off from different sites of neurons, from the cell body to different levels of neurites. It is important to underline that neurons analyzed in this paper are fetal and it cannot be excluded at the moment that their ability to shed extracellular vesicles is bound to their developmental stage. In the future, it will be of great interest to investigate the phenomenon in vivo both in developing and in mature brain, under physiological as well as pathological conditions, such as those that normally involve BBB modifications and neurodegeneration. In order to face these more complex problems, we think that the first step is to understand the fate of secreted VEGF and FGF-2 when vesicles shed by neurons are added to other cell cultures: neurons themselves, astrocytes and/or endothelial cells. An important second step will be to study the effect of extracellular vesicles released by neurons in a model of bloodbrain barrier.

Aknowledgements

The first two authors equally contributed to this paper. We wish to thank Dr. F Carfi Pavia for helpful discussions, advices and assistance with scanner electron microscopy, and Mr. S Agnello for technical assistance with confocal microscopy.

This work was supported by a special grant by Serono S.p.A., by the Italian Ministero dell’Università e della Ricerca (PRIN 2004) and by the University of Palermo (fondi di Ateneo, Università degli Studi di Palermo). P. Proia was supported by a Ph.D. fellowship of the Università degli Studi di Palermo, Palermo, Italy. G. Schiera was supported by a post-doctoral fellowship by Università degli Studi di Palermo, Palermo, Italy.