Constitutive Secretion of Protease Nexin-1 by Glial Cells and Its Regulation by G-Protein-Coupled Receptors

Introduction

Glial cells and, in particular, astrocytes play a critical role in neuronal nutrition, guidance, homeostasis, and detoxification (Barres, 2003). They provide a structural support and a source of regulatory molecules (Araque and Perea, 2004), neurotrophic factors, and neuropeptides (Lafon-Cazal et al., 2003; Slezak and Pfrieger, 2003). Moreover, astrocytes actively control synaptogenesis (Slezak and Pfrieger, 2003; Ullian et al., 2004), synaptic efficacy, and plasticity (Pfrieger and Barres, 1997; Murai et al., 2003). In this regard, it has been proposed that glia-derived soluble proteins are necessary for the synapse formation and function (Blondel et al., 2000; Slezak and Pfrieger, 2003). However, little is known about the components of the intracellular secretory pathways through which these proteins transit and ultimately exit the astrocytes.

Astrocytes possess machinery for vesicular fusion that transport and secrete these neuroactive proteins (Volknandt, 2002). Cultured hippocampal astrocytes secrete secretogranin II (Calegari et al., 1999), and cortical astrocytes discharge atrial natriuretic peptide (Krzan et al., 2003) in a Ca²⁺-dependent manner, suggesting that these cells possess a regulated secretory pathway. Dense core and clear electron-lucent vesicles have also been identified in cultured astrocytes (Calegari et al., 1999). This raises the possibility that astrocytes contain several types of vesicles involved in the release of various neurotransmitters, peptides, or proteins (Bezzi et al., 2004).

Glial and neuronal cells possess their own serine protease-signaling system, including many components of the blood coagulation and fibrinolysis systems (i.e., thrombin or plasmin) as well as unique members of the serpin class (i.e., neuroserpin) (Turgeon and Houenou, 1997; Gingrich and Traynelis, 2000). Cumulative evidence has shown that the balance between extracellular serine proteases and their cognate serpins may be involved in the development (Turgeon and Houenou, 1997), plasticity (Tomimatsu et al., 2002), and pathology of the nervous system (Turgeon and Houenou, 1997; Molinari et al., 2003). Moreover, these proteases and their inhibitors may also play a role when the blood–brain barrier is compromised (Gingrich and Traynelis, 2000).

Protease nexin-1 (PN-1) or serpinE2 is a 43 kDa glycoprotein synthesized and released by both glial and neuronal cells (Reinhard et al., 1994). PN-1 belongs to the serpin family (Gloor et al., 1986) and inhibits several serine proteases, including thrombin (Baker et al., 1980; Knauer et al., 2000). In cell cultures, PN-1 promotes neurite outgrowth by preventing the cleavage-induced activation of thrombin receptors (Gurwitz and Cunningham, 1988; Suidan et al., 1992). Together with thrombin, PN-1 plays a role in neuronal degeneration or protection (Smith-Swintosky et al., 1995; Vaughan et al., 1995). In mice, both overexpression and inactivation of PN-1 alter hippocampal long-term potentiation (Luthi et al., 1997), and overexpression leads to a progressive neuronal and motor dysfunction (Meins et al., 2001).

To better understand the mechanisms that underlie the trafficking and secretion in astrocytes, we focused on the traffic and exocytosis of PN-1 using a green fluorescent protein (GFP)-tagged-PN-1 (GFP/PN-1). The post-Golgi biosynthetic traffic of GFP/PN-1 was studied by time-lapse imaging, fluorescence and biochemical quantitations, and dual imaging in C6 glioma cells and cultured astrocytes.

Materials and Methods

Plasmid constructions. The GFP/PN-1 construction was obtained by inserting GFP between the signal peptide (SP-PN-1; aa 1–20) and the mature PN-1 (aa 21–408) in a three-step process. First, SP-PN-1 was amplified using appropriate primers. Second, the mature PN-1 sequence was amplified and linked to the 3′ of the SP-PN-1. Third, the GFP sequence was amplified from pEGFP-N1 (Clontech, Cambridge, UK) with two glycine residues and HindIII restriction sites flanking either ends (5′-HindIII-GlyGly-EGFP-GlyGly-HindIII 3′). The modified sequence of GFP was then incorporated into the PN-1 sequence.

The β-tubulin/DsRed construction was obtained by fusing the red fluorescent protein (DsRed) from Discosoma sp. in pDsRed vector (Clontech) to the 3′ of the β5-tubulin. Tubulin was amplified by PCR without the stop codon and linked to the 5′ of the full open reading frame of DsRed. The fused product was then subcloned in the pEGFP-N1 vector lacking the EGFP gene. All of the constructs were confirmed by sequencing.

Cell culture and transfections. C6 glioma cells were cultured as described previously (Lagriffoul et al., 1996). For transient or stable transfections, subconfluent cells were incubated with different plasmid constructions and Exgen 500 reagent (Euromedex, Strasbourg, France) following the instructions of the manufacturer. Cells were used 48–72 h after transfection. To create stable cell lines, transfected cells were selected with 0.4 mg/ml G418 (Invitrogen, Cergy Pontoise, France), and neo-resistant populations expressing GFP/PN-1 were enriched by cell sorting using a fluorescence-activated cell sorting analyzer (Beckman Instruments, Villepinte, France). Thereafter, they were cultured in medium supplemented with 0.4 mg/ml G418.

Striatal astrocytes were cultured as described previously (el-Etr et al., 1989) and transfected with lipofectamine plus reagent (Invitrogen) following the instructions of the manufacturer and maintained in medium for 48–72 h before experiments.

Antibodies and reagents. Mouse monoclonal antibodies against PN-1 (4B3) were a kind gift from Dr. D. Monard (Friedrich Miescher Institut, Basel, Switzerland) and have been described previously (Reinhard et al., 1994). For detection of GFP, we used monoclonal antibodies from Roche Diagnostics (Meylan, France) To stain F-actin, we used rhodamine-labeled phalloidin, and to label microtubules, we used anti-tubulin antibodies (all were obtained from Sigma, St. Louis, MO). Monoclonal antibodies against trans-Golgi network-38 (TGN38) and γ-adaptin were obtained from BD (Le Pont de Claix, France). To disrupt microtubule, we used nocodazole (Fluka, Saint Quentin Fallavier, France), and to induce F-actin disassembly, we used cytochalasin D (Sigma) or latrunculin B (Calbiochem, La Jolla, CA). Jasplakinolide was obtained from Invitrogen. Mastoparan (Mp) was synthesized, as described previously (Lagriffoul et al., 1996), or supplied from Calbiochem. Plasmid encoding for enhanced yellow fluorescent protein (EYFP)-tagged β-actin (EYFP/actin) was obtained from Clontech. Vasoactive intestinal peptide (VIP) and pituitary adenylate cyclase-activating polypeptide with 38 amino acids (PACAP-38) were obtained from Neosystem (Strasbourg, France), and Bordetella pertussis toxin (PTX) was obtained from Sigma.

Drug treatment and imaging. Before analysis, stable transfected cells were incubated in the presence of 5 mm sodium butyrate (Sigma) for 18 h to increase GFP/PN-1 expression by enhancing transcription of cytomegalovirus promoter-controlled genes (Dorner et al., 1989). Blocks of 20°C were performed in the medium supplemented with 20 mm HEPES, pH 7.4, and 50 μg/ml cycloheximide, a concentration sufficient to block protein synthesis by >95%. Shifting temperature after the 20°C block was performed by changing medium and incubating cells at 37°C in the medium, without HEPES. Imaging of live cells was performed on a Zeiss (Le Pecq, France) Axiovert 100 TV inverted microscope at 34°C to slightly reduce vesicle motion. The microscope was equipped with a software-driven filter wheel, a 63× oil immersion objective [1.3 numerical aperture (NA); Zeiss], or a 25× oil immersion objective (0.8 NA; Zeiss). A homemade temperature-regulated chamber perfused by a syringe pump (Harvard Apparatus, Holliston, MA) at 100 μl/min and a MicroMax 1300 charge-device camera (Princeton Instruments, Evry, France) driven by MetaMorph imaging software (version 4.17; Universal Imaging Corporation, West Chester, PA) were used for cell imaging. Filter sets for conventional fluorescein imaging and neutral density filters were used for GFP/PN-1. For β-tubulin/DsRed, we used filters (Omega Optical, Brattleboro, VT) with a 540 ± 25 nm bandpass and a 575-nm-long bandpass for excitation and emission, respectively, and for YFP/actin, we used 500 ± 25 and 545 ± 35 nm bandpass filters for excitation and emission, respectively.

For time-lapse imaging of both GFP/PN-1 and β-tubulin/Dsred, molecules were excited with a 500 ± 25 nm bandpass filter and imaged with a 530 nm long-pass emission filter. For time-lapse imaging of both GFP/PN-1 and EYFP/actin, molecules were excited with a 475 ± 40 nm bandpass filter and imaged with a 535 ± 45 nm filter. Exposure times were 300–500 ms for GFP/PN-1, 0.5–1 s for DsRed, and 500 ms for YFP/actin. Images from the MicroMax camera were converted to TIFF files that were edited using NIH Image 1.62 and compiled into QuickTime (Apple Computers, Cupertino, CA) movies with Premiere 6.0 (Adobe Systems, San Jose, CA).

Quantitative fluorescence measurements. The rate of GFP/PN-1 exit from the Golgi/TGN area was determined by measuring GFP/PN-1-integrated fluorescence intensity in the TGN over time (10 min intervals; 120 min duration). Using the 25× objective, we measured the entire cell fluorescence intensity by imaging in a single focal plane and applied thresholds to delineate Golgi-associated GFP/PN-1, as described previously (Kreitzer et al., 2000). Threshold values selected at the first time point in each recording were maintained constant for all images. Data were normalized as a percentage of TGN-associated fluorescence, and at time 0 was set to 100. Loss of fluorescence associated with the Golgi/TGN region was not the result of photobleaching, because the total fluorescence over time remained constant when identical imaging conditions were used in fixed cells.

Vesicle number was calculated on fixed cells by segmenting the 12-bit image and determining the number of objects using a macro routine of the MetaMorph software. Objects having an area between two and seven pixels were considered to be vesicles.

Vesicle elementary speeds were determined using a macro routine of MetaMorph software. Data regarding X- and Y-position coordinates and vesicle displacement were logged to disk and treated with Excel Software (Microsoft, Seattle, WA).

F-actin distribution was determined using a 25× objective after F-actin staining by rhodamine phalloidin (RP) and nucleus staining by Hoechst 33258. Images of equatorial slices were quantified by radial line scan analysis. Two to three lines per cell (on at least 30 cells) were obtained, each extending from the exterior to the nucleus. The nucleus was normally >8 μm from the cell edge and, therefore, not included in the analysis. The RP fluorescence intensity profiles for each experiment were pooled and averaged to form a mean profile.

Deconvolution. Stacks of 12-bit files from time-lapse acquisition and images captured by conventional or confocal microscopy were restored with Huygens (Scientific Volume Imaging, Hilversum, The Netherlands). Restored stacks and images were then further processed with Imaris (Bitplane, Zurich, Switzerland).

Indirect immunofluorescence. Cell cultures generated on 12 mm glass coverslips were dealt with as described previously (Lagriffoul et al., 1996). For PN-1 antibodies, cells were permeabilized with 0.4% SDS in 1.5 m Tris-HCl, pH 8.8, to allow antibody access to GFP/PN-1 into the vesicle lumen, as described previously for PN-1 (Reinhard et al., 1994). Cells were viewed on a Zeiss Axiophot 2 microscope or on a BioRad (Hercules, CA) MRC 1024 confocal laser-scanning microscope equipped with an argon/krypton ion laser. Colocalization analysis was processed by the colocalization Bio-Rad Laser-sharp software on individual confocal optical sections using the facilities of the Centre Régional d'Imagerie Cellulaire (Montpellier, France).

Subcellular fractionation. Cells subjected to the temperature-induced block (2 h at 20°C and 30 min at 37°C) were washed and homogenized in 0.25 m sucrose containing 20 mm HEPES, pH 7.4, 1 mm EGTA, and 1 mm magnesium acetate supplemented with a mixture of protease inhibitors (1 μg/ml aprotinin, 2 μm leupeptin, 60 μm benzamidine, and 3 μm antipain) using a cell Cracker (EMBL, Heidelberg, Germany). After centrifugation for 10 min at 1000 × g, the supernatant was loaded onto a 0.3–1.2 m continuous sucrose gradient in 20 mm HEPES, pH 7.4 (velocity gradient), and centrifuged for 15 min at 106,000 × g in a swing-out rotor (SW41; Beckman Instruments). The first four fractions from the top of the gradient were loaded onto a 0.5–1.6 m continuous sucrose gradient in 20 mm HEPES, pH 7.4 (density gradient), and centrifuged for 20 h at 106,000 × g. Fractions were collected and centrifuged for 1.5 h at 150,000 × g in a TL-100 centrifuge (Beckman Instruments). Aliquots of resuspended pellets were analyzed by SDS-PAGE and immunoblotting.

Quantitative analysis of PN-1 secretion. Cultured cells in 12-well plates were subjected to the temperature block (20°C) in serum-free DMEM supplemented with 20 mm HEPES, 500 mg/L chicken ovalbumin, and 50 μg/ml cycloheximide. Bordetella pertussis toxin cell treatments were performed as described previously (Lagriffoul et al., 1996). Secretion experiments were performed at 37°C in the same medium but without HEPES. Media and cells from triplicate wells were harvested, centrifuged, and frozen at –80°C. Aliquots of 30 μl from C6 cells (15 μg and 20–30 μg of protein for media and lysates, respectively) and 70 μl from striatal astrocytes (35 μg protein, precipitated for 2 h with 10% ice-cold trichloroacetic acid) were subjected to SDS-PAGE analysis and immunoblotting with monoclonal anti-GFP or anti-PN-1 antibodies as described previously (Lagriffoul et al., 1996; Lafon-Cazal et al., 2003). Detection was performed by chemiluminescence, and signals were quantified by densitometric analysis. Data were processed for mean ± SEM and Student's t test.

Results

Discussion

In the present study, we analyzed by several methods, including time-lapse videomicroscopy, the mechanism by which glial cells secrete and control PN-1 traffic. To address these questions, we constructed a recombinant PN-1 protein that fused its mature N terminus with GFP (GFP/PN-1) and used this reporter protein as a probe to study post-Golgi exocytotic trafficking and secretion in glial cell cultures. We described for the first time PN-1 trafficking and found that PN-1 release is controlled by reorganization of F-actin cytoskeleton following activation of G-protein-coupled receptors, such as VPAC receptors.

Several serpins have been detected in the brain (Osterwalder et al., 1998). Some of them are mainly secreted by neurons such as neuroserpin (Osterwalder et al., 1998) and others, phosphatidyl-ethanolamine-binding protein by glial cells (Lafon-Cazal et al., 2003) and PN-1 by both cell types (Reinhard et al., 1994). Neuroserpin is associated to dense-core granules and could follow a regulated secretory pathway (Osterwalder et al., 1998). To date, there is no evidence for a regulated secretion of serpins in glial cells. However, hippocampal astrocytes appear to express two classical secretory pathways, constitutive and regulated, respectively (Calegari et al., 1999; Volknandt, 2002).

We found that GFP/PN-1 secretion followed a constitutive secretory pathway, operating via small vesicles, which were not stored along the plasma membrane. This contrasts with the classical regulated secretory pathway in which dense-core vesicles are stored until a specific signal evokes their exocytosis. However, GFP/PN-1 secretion is regulated at its final stage by a G-protein-coupled receptor-dependent step, indicating that this secretion belongs to receptor-mediated constitutive secretions. Evidence of a receptor-mediated regulation of the constitutive secretory pathway has been described previously (Luini and De Matteis, 1993). Our data suggest that the PN-1 secretory pathway follows this type of secretion.

GFP/PN-1 traveled through the secretory pathway (reticulum, Golgi/TGN), as expected for a secreted glycoprotein possessing a signal peptide. GFP/PN-1-containing carriers appeared only as small vesicles in both transfected astrocytes in primary culture and glial cell lines. This contrasts with results obtained with vesicular stomatitis virus glycoprotein/GFP (Hirschberg et al., 1998; Toomre et al., 1999) or GFP-tagged membrane proteins in neuronal axons (Nakata et al., 1998; Kaether et al., 2000) that appeared, in majority, as tubules. These differences could be attributable to either the cell type or the location of the protein (membrane-bound vs lumenal proteins) (Toomre et al., 1999).

GFP/PN-1-containing vesicles were transported from the TGN to the plasma membrane along tracks in a discontinuous movement. This movement is composed of rapid runs, pauses, and changes in direction. Movement occurs at maximal and average speeds of ∼1 and ∼0.4 μm/s, respectively. These values are consistent with that of 0.3–1 μm/s from other studies (Hirschberg et al., 1998; Nakata et al., 1998; Toomre et al., 1999; Kaether et al., 2000). Moreover, these values are compatible with speed of motors such as kinesins, which drive vesicles along microtubule tracks (Hirokawa, 1998; Goldstein and Yang, 2000).

Depolymerization of microtubules induced changes in the profile of vesicle elementary speeds by decreasing either duration of long runs or maximal speed. This disassembly was followed by a decrease in GFP/PN-1 secretion. This suggested that microtubules are used as tracks to transport GFP/PN-1 containers and that translocation was not a simple random diffusion. However, disruption of microtubules is known to induce both fragmentation and redistribution of Golgi stacks (Storrie et al., 1998) that may either modify kinetics from TGN sorting or decrease GFP/PN-1 secretion.

To tackle this problem, we used a more direct approach by labeling microtubules with β-tubulin/DsRed and imaging of double-transfected cells (β-tubulin/DsRed and GFP/PN-1) to directly visualize vesicle movements along microtubules. Using this technique, we showed that vesicles moved along microtubules, stopped, remained associated or close, and switched tracks during movement. Thus, GFP/PN-1 vesicle transport proceeds along microtubules, probably using motors (kinesins and dyneins) driving a predominant vesicular flow from the center to the cell periphery, despite its stochastic appearance.

In addition to microtubules that served as tracks for long-range transport, actin cytoskeleton and myosins are also implicated in vesicular transport. In this case, displacements are shorter, and actin-based transport is slower (∼0.1 μm/s) than translocation along microtubules (Wu et al., 2000). In astrocytes, during endocytotic and recycling events, membrane-bound type II deiodinase is transported by myosin Va along actin tracks (Stachelek et al., 2001). In contrast, time-lapse imaging did not show any clear movements of GFP/PN-1-containing vesicles along F-actin fibers in cells expressing both EYFP/actin and GFP/PN-1. Finally, the myosin inhibitor, BDM did not modify the secretion of GFP/PN-1. This confirms that myosin motors are not involved in GFP/PN-1 post-Golgi trafficking. Moreover, we detected no bursts of actin polymerization at vesicles carrying GFP/PN-1, such as previously reported during endocytic events in cultured mast cells (Merrifield et al., 1999), cell sorting of viruses (Rietdorf et al., 2001), and propulsion of Listeria in infected cells (Cossart, 2002). Thus, the actin cytoskeleton is not used as a track or a propulsion force to drive transport of GFP/PN-1-containing vesicles.

Imaging on fixed and live cells shows an accumulation of GFP/PN-1-containing vesicles on cortical F-actin, suggesting that F-actin at the cell cortex acts as a physical barrier that hinders vesicular movement. This proposal of an F-actin cortical barrier has been reported previously for dense-core secretory granules in the regulated secretion of chromaffin cells (Trifaro et al., 1992) and mast cells (Norman et al., 1994). In contrast to the regulated secretory pathway, there is little evidence for the involvement of actin in the constitutive secretory pathway in mammal cells (Hirschberg et al., 1998). Generally, F-actin disruption failed to affect constitutive secretion and did not affect kinetics of post-Golgi transport (Salas et al., 1986).

In contrast, GFP/PN-1 secretion was enhanced by cytochalasin D and latrunculin B treatment. Furthermore, we showed that mastoparan was able to disrupt cortical F-actin-promoting exocytosis without affecting vesicular budding or transport. This effect was prevented by jasplakinolide, an analog of phalloidin that stabilizes F-actin bundles. Similar effects were obtained by activation of VPAC receptors coupled to Gi/Go-proteins. Together, these results indicate that cortical F-actin hinders movement of vesicles and behaves as an obstacle to vesicle docking and fusion. Removal of this obstacle by cytochalasin D, latrunculin B, VPAC receptor activation, and mastoparan allows GFP/PN-1-containing vesicles to dock and fuse to plasma membrane-enhancing secretion of GFP/PN-1. Thus, cortical F-actin acts as a barrier in both constitutive and regulated secretions. However, in constitutive secretion, vesicles are not stored and embedded in the cortical F-actin network, like granules in regulated secretion (Lang et al., 2000).

PN-1 is the most potent thrombin inhibitor in the brain, but it can also form complexes with plasmin, tissue, and urokinase plasminogen activators and trypsin (Reinhard et al., 1994; Gingrich and Traynelis, 2000). Apart from regulation by de novo biosynthesis (Vaughan and Cunningham, 1993), PN-1 secretion appears to be regulated by activated G-coupled receptors, such as VIP receptors (Festoff et al., 1996). Here, we show that activation of VPAC receptors coupled to heterotrimeric Gi/Go-proteins is able to control exocytosis of GFP/PN-1 in C6 glioma cells. Furthermore, PACAP induced PN-1 secretion in both cultures of C6 glioma cells and striatal astrocytes, indicating that glial cells possess a receptor-mediated regulation of PN-1 constitutive secretion that is triggered after VPAC receptor activation. The receptor-mediated regulation of secretion provides a way to adapt serpin release and modulate proteolytic activity in the glial cell environment. In particular, this regulation could represent an efficient mechanism to control extracellular thrombin activity and activation of protease-activated receptors, such as PAR-1 that trigger important changes in actin polymerization (Gingrich and Traynelis, 2000). Thus, this mechanism might operate in the regulation of dynamic changes during neurite outgrowth and in the extracellular matrix components during neural plasticity in developing or mature brain (Turgeon and Houenou, 1997; Gingrich and Traynelis, 2000). Furthermore, PN-1 is localized in the perivascular endfeet of glial cells, suggesting that PN-1 is well positioned to counteract extravasated thrombin (Gingrich and Traynelis, 2000). Such a receptor-mediated regulation of secretion might promote a rapid release of anti-proteases, counteracting the deleterious effects of extravasated proteases after brain trauma.

In conclusion, our study provides for the first time an extensive characterization of a regulated secretory pathway involved in the release of PN-1 (and possibly other serpins). This process may contribute to the adaptation of extracellular proteolytic activity in response to synaptic activity.