The small gtp-binding protein, Rab6p, is associated with both golgi and post-golgi synaptophysin-containing membranes during synaptogenesis of hypothalamic neurons in culture

Non-hydrolysable analogues of GTP, such as GTPγS, irreversibly block many steps of protein transport along the biosynthetic/secretory and endocytic pathways in eucaryotic cells (Goud and McCaffrey, 1991). The targets of GTPγS appear to be the numerous GTPases, which have been recently identified as key components of the cellular machinery responsible for budding, targeting and fusion of the transport vesicles moving between intracellular compartments (Goud and McCaffrey, 1991; Pfeffer, 1992). The best characterized of these GTPases are the so-called small GTP-binding proteins of the Sec4/Ypt1/Rab family, which now form a distinct branch of the p21^ras superfamily. Mutations in Ypt1p and Sec4p interrupt the yeast Saccharomyces cerevisiae secretory pathway between the endoplasmic reticulum and the Golgi apparatus and between the Golgi apparatus and the plasma membrane, respectively (Bacon et al., 1989; Baker et al., 1990; Goud et al., 1988; Salminen and Novick, 1987; Segev et al., 1988; Segev, 1991).

The role of Sec4 and Ypt1 proteins would be to monitor the accuracy of targeting/fusion of transport vesicles with their acceptor membranes. Mammalian rab proteins share in common with Sec4p and Ypt1p highly conserved domains, including a putative effector region (Chavrier et al., 1990a; Touchot et al., 1987; Zahraoui et al., 1989). One of the rab proteins (Rab1p) is the mammalian counterpart of Ypt1p (Haubruck et al., 1989). Although the function of most of these proteins is presently unknown, recent evidence indicate that several Rab proteins play a role at various steps of intracellular transport (Goud and McCaffrey, 1991; Pfeffer, 1992).

Different rab proteins have been localized to distinct exocytic (Chavrier et al., 1990b; Darchen et al., 1990; Fischer von Mollard et al., 1990; Plutner et al., 1991) and endocytic compartments (Chavrier et al., 1990b; Van der Sluijs et al., 1991). One of them, Rab6p, is associated with medial and trans-Golgi cisternae, as well as with the trans-Golgi network (TGN) (Goud et al., 1990; Antony et al., 1992). In addition, Rab6p also associates with distinct post-Golgi vesicles in Torpedo marmorata electrocytes (Jasmin et al., 1992). The trans side of the Golgi apparatus is a complex region responsible for sorting of proteins to secretory vesicles of the constitutive and regulated pathway and to endosomes and lysosomes (Griffiths and Simon, 1986). Therefore, these localizations of Rab6p make it a good candidate for a small GTP-binding protein involved in the sorting and/or targeting of post-Golgi vesicles.

In developing neurons, one of the late steps of synaptogenesis consists of the polarized targeting of small synaptic vesicles (SSVs) to a specialized domain of the plasma membrane, the presynaptic membrane. However, the biogenesis of SSVs and the mechanism of their targeting is still a matter of speculation (De Camilli and Jahn, 1990; Huttner and Dotti, 1991; Kelly, 1988; Lindstedt and Kelly, 1991). The extensive molecular characterization of several SSV membrane components has provided specific immunological tools to investigate this. Among these components synaptophysin (Sy) (Jahn et al., 1985; Wiedenmann and Franke, 1985) is the most prominent and its level augments during brain synaptogenesis in vivo (Knaus et al., 1986) and in vitro (Tixier-Vidal et al., 1992). We have previously followed its distribution during the development in culture of mouse hypothalamic neurons taken at an early postmitotic stage and which achieve in vitro full synaptogenesis (Tixier-Vidal et al., 1986, 1988). Sy is first expressed in presumably trans-Golgi cisternae and vesicles, several days before the accumulation of Sy immunostained vesicles in nerve endings. Then, the accumulation of SSVs in nerve endings is accompanied by a decreased immunostaining of Golgi membranes, which could be restored upon microtubule disruption (Tixier-Vidal et al., 1988). Moreover, we have shown that in mature neurons in culture, Sy is segregated from a secretory protein, Secretogranin II, upon exit from the TGN (Kagotani et al., 1991). Thus, these findings suggest that Sy is released from the TGN in a vesicular form, as SSV precursors, and is then transported to nerve endings by a mechanism that requires integrity of microtubules. Sy vesicles might acquire SSV configuration either during transport or in axon terminals via interaction with cytoplasmic proteins. Another alternative has been presented from work performed in PC12 cells showing that newly synthesized Sy leaves the TGN via constitutive secretory vesicles to the plasma membrane before being incorporated into synaptic-like microvesicles via early endosomes (Regnier-Vigouroux et al., 1991). In any case, the mechanism of post-Golgi transport of Sy is central to synaptogenesis.

In the present work, we have investigated the possible involvement of Rab6p. We have localized and quantified Rab6p during development in culture of hypothalamic neurons and compared its subcellular distribution and level of expression to that of Sy using various morphological and fractionation procedures.

Hypothalamic cells were mechanically dissociated from 15-day mouse fetus brain and cultured in serum-free medium as previously described (Faivre-Bauman et al., 1984; Tixier-Vidal et al., 1986). They were seeded either on glass coverslips in plastic 4-well trays (2×10⁵ cells/500 µl) for immunofluorescence localization, or in plastic culture dishes (35 mm or 60 mm, Lux Corporation, CA) (1.5 ×10⁶ cells/2 ml, 5 ×10⁶ cells/4 ml, respectively) for electron microscopic immunoperoxidase or fractionation experiments. The medium was renewed after 5 d and replaced by a medium containing cytosine arabinoside (10^-6 M) which was renewed every 2-3 d.

Rabbit anti-Rab6p antibodies (A-Rab6p) have been previously characterized (Goud et al., 1990). Furthermore, these antibodies do not recognize Rab3Ap (see Results, Fig. 1). The monoclonal antibody against Sy (HPLC-purified IgG1) (A-Sy) is monospecific for Sy (Wiedenmann and Franke, 1985). Monoclonal antibody against protein SV2 (A-SV2) (Buckley and Kelly, 1985) was kindly provided by Dr E. Schweitzer (UCLA, USA). Rabbit polyclonal anti-TGN 38 was kindly provided by Dr P. Luzio (Luzio et al., 1990). Monoclonal antibody against microtubule associated protein 2 (MAP2) was purchased from Biomakor (Israël).

Monolayers on glass coverslips were fixed in 4% paraformaldehyde in Sorensen phosphate buffer for 1 h at room temperature, permeabilized with 0.01% Triton X-100 and immunostained as previously described (Tixier-Vidal et al., 1988). For double localization, affinity-purified A-Rab6p (1/5) or A-TGN 38 (1/200) was mixed with A-Sy (1 µg/ml). As second antibodies goat anti-rabbit IgG conjugated to either rhodamine or fluorescein-isothiocyanate (Nordic) were mixed with goat anti-mouse IgG conjugated to either fluorescein isothiocyanate or rhodamine (Nordic). After several washings, coverslips were mounted on glass slides with Mowiol (Hoeschst). They were examined using a Leitz epifluorescence microscope equipped with appropriate filters.

Confocal laser scanning microscopy (CLSM) was performed using a Leica system (Leica Laser Technik, Heidelberg, Germany), based on a Leitz Diaplan microscope. The system is associated with an informatic unit (ELTEC, Motorola microprocessor 68020 (32 Bits, 20 MHz), memory size 22 MO including 2 MO of RAM video). It is linked (Ethernet) to a Silicon Graphics 4D35GT and the sections are visualized on a 32.5 cm Mitsubishi screen (800 holes per 560 lines). An argon-krypton laser (Omnichrome) adjusted at 488, 567 and 647 nm was used. Photographs were made from the hard disk memory using a freeze frame system from Polaroid.

For this study, the second antibodies were from Amersham and conjugated to either Texas Red or fluorescein isothiocyanate. In control preparations, a background staining of neuron perikarya was observed with the Texas Red-labeled sheep anti-mouse IgG, but not with Texas Red-labeled donkey anti-rabbit IgG. Thus, the confocal analysis was performed with Texas Red (TR) as a marker for Rab6p, or for TGN 38, and fluorescein (FITC) as a marker for Sy.

For Rab6p and Sy, double fluorescence images were acquired simultaneously according to the following procedure. The two signals were simultaneously excited at wavelengths of 488 nm (for FITC) and 567 nm (for TR), respectively, and detected on the two channels using appropriate filters. First, a double dichroic mirror gave an optimal separation and a 90% transmission of the emitted lights in the wavelength ranges of 535 nm (FITC) and 610 nm (TR), respectively. A second dichroic splitter (Leica -TK 580) separated green and red signals. Then, two barrier filters were placed in front of each one of the two photomultipliers: a narrow band filter (Schott; 535 ± 8 nm) for FITC and a long wave pass filter (Schott; 610 nm) for TR. Upon visualization of the optical sections, the intensity of the green signal was generally higher than that of the red one, suggesting a possible difference in the sensitivity of the two photomultipliers. Thus, we verified the absence of bleeding of the green channel to the red one by exciting the preparation at a wavelength of 567 nm only. This gave the same intensity of the red signal. Focal series of horizontal sections were performed, with a ×100 objective (plan fluotar, op. 1.30) (pinhole closed, 0.3 µm sections). Depending on neurons, 12-14 paired sections were collected from the bottom to the top of the culture.

For TGN38 and Sy double fluorescence, images were acquired successively, first red, then green, using the same conditions and control as above, to avoid any bleeding from the green channel to the red one. Focal series of horizontal sections (0.8 µm thickness) were performed using a ×63 objective (op. 1.30).

Monolayers were fixed in 4% paraformaldehyde and 0.01% glutaraldehyde in Sorensen phosphate buffer for 90 min at room temperature. They were then processed, without permeablization, as previously described for Sy localization (Tixier-Vidal et al., 1988).

Cells cultured for increasing times were scraped in phosphate buffer saline containing phenylmethylsulfonyl fluoride (1 mM) and a mixture of protease inhibitors (leupeptin, chymostatin, pep-statin, antipain and aprotinin; 0.1 µg/ml). After addition of SDS (2%), the lysates were heated for 5-10 min at 95°C, sonicated, and centrifuged at 10,000 g for 5 min. Samples of the supernatant were used for protein determination according to the method of Peterson (1977). Laemmli buffer (Laemmli, 1970) was then added and the samples were stored at-20°C until use. Western blotting and quantification of the amount of Rab6p in the different fractions, using affinity-purified anti-Rab6p antibodies and ¹²⁵I-labeled Protein A (Amersham), were as described previously (Goud et al., 1990).

To follow the distribution of Rab6p between cytosol and membranes, cultures were homogeneized with a Dounce homogenizer in buffer H (Leube et al., 1989) (10 mM triethanolamine-acetic acid, pH 7.4, 1 mM EGTA, 1 mM EDTA, 0.1 mM DTT, 0.2 mM phenylmethylsulfonyl fluoride). The postnuclear supernatant was subjected to ultracentrifugation (140,000 g for 1.5 h). The resulting pellet was resuspended in buffer H. Samples of the supernatant and of the solubilized membrane pellet were then boiled in Laemmli buffer and stored at-20°C before assays. Western blotting and Rab6p quantification were performed as described above, in parallel for the 7 stages examined.

Hypothalamic cultures taken at two developmental stages, 6 and 15 days in vitro (DIV), were subjected, in parallel, to subcellular fractionation on sucrose gradients followed by immunoisolation (fractions 2-6 from the gradient) with A-Sy coupled to Dynabeads as previously described (Leube et al., 1989). The only modification was that postnuclear supernatants were not concentrated prior to gradient centifugation. At each stage, the same amount of proteins was loaded onto the sucrose gradient. The immunoisolated membranes were subjected to SDS-PAGE followed by western blotting using crude A-Rab6p anti-serum, polyclonal A-Sy antiserum and A-protein SV2.

We have shown previously that the polyclonal antibody we have generated against purified human recombinant Rab6p did not cross-react with Rab1A, Rab2, Rab4 and Rab5 proteins (Goud et al., 1990). We checked in the present study whether this antibody could recognize Rab3Ap, also called smg p25A (Matsui et al., 1988), a rab protein only expressed in neurons and neuro-endocrine cells (Burstein and Macara, 1989; Olofsson et al., 1988; Sano et al., 1989) and found to be associated with the synaptic vesicle membrane (Fischer von Mollard et al., 1990). As shown in Fig. 1, no cross-reaction with Rab3Ap was detected by western blotting even when a large amount of purified Rab3Ap was loaded onto the gel. In addition, affinity-purified anti-rab6 antibody lights up only one spot on 2-D gels, in the 20-25 kDa region of various cell extracts, suggesting that this antibody is truly Rab6p specific (B. Goud, unpublished results).

At the earliest stage examined, 2 days in vitro (DIV), as well as at all later culture stages, Rab6p was conspicuously localized in a juxta-nuclear region reminiscent of the Golgi zone, in all presumptive neurons (Fig. 2). Rab6p was also expressed in neurites but the intensity and distribution of the immunostaining varied with time in culture. At early stages (4 DIV; Fig. 2a), it was found as an homogeneous staining in short neurites, which reacted positively with a monoclonal antibody against microtubule associated protein 2 (MAP2; not shown). In the following days, while neurite length increased, the Rab6p immunosignal became concentrated in varicosities distributed along thin neurites (Fig. 2c). Most of these thin neurites were immunonegative for MAP2, suggesting that they represent growing axons. At later stages (12 DIV onward), while the complexity of the neuritic network and the density of varicosities increased, Rab6p expression became restricted to a small proportion of varicosities (Fig. 2e). In contrast, the intensity of the immunosignal remained the same in perikarya.

The comparison of Rab6p and Sy distributions by double immunofluoresence microscopy revealed apparent co-localizations. However, differences in the respective intensity and distribution of the two signals, depending on developmental stages and neuron compartments, were observed. In perikarya, the Sy immunosignal was distributed in the same juxta-nuclear region as Rab6p at the earliest stages examined and then became progressively less conspicuous than that of Rab6p, with increasing time in culture. In emerging neurites, Sy co-localized with Rab6p and, in thin neurites (presumably axons), both became simultaneously restricted to varicosities (Fig. 2a,b,c,d). However, at the time of synapse formation, which begins around 10 DIV in our culture conditions (Tixier-Vidal et al., 1986, 1988), a dramatic decrease in the proportion of double-labeled varicosities occurred (Fig. 2e,f). The density of Sy-containing varicosities and boutons increased considerably from day 4 to day 21, whereas that of Rab6p-containing varicosities decreased after approximately 10 DIV. Thus, Rab6p seems to co-localize with Sy-containing membrane during the first phase of axonal growth up to synapse formation.

The respective subcellular distributions of Rab6p and Sy were further compared using CLSM on cultures taken at 6 DIV, a stage where the two antigens were often found apparently co-localized in neurons (Fig. 2c,d). As previously assumed, the limit of the resolution of the optical microscope (0.2 µm) should theoretically permit the observation of two distinct or mixed staining patterns of immunolabeling between Golgi subcompartments (Antony et al., 1992). When the distribution of Rab6p and Sy were compared on successive horizontal sections of a same neuronal perikaryon, both co-localizations and distinct pat-terns could be observed depending on the neurons and the sections (Figs 3, 4). The extent of co-localization was further analyzed by merging the images of both channels (Fig. 5). Examples of two characteristic situations are depicted in Figs 3, 4 and 5. In the neuron perikaryon analysis shown in Fig. 3D-J, a good overlapping of both markers is observed in a juxta-nuclear Golgi reticulum, at every successive paired section depicted. However, on merged images of both channels (Fig. 5A), an heterogeneity of the distribution of the colors from green to yellow and red, suggests the possibility of quantitative differences in the distribution of both antigens along Golgi membranes; however, this is observed at the limit of the resolution of the light microscope. Another situation is shown in the neuron perikaryon depicted in Figs 4 and 5B. In the second optical paired sections (Fig. 4A), Sy is already strongly expressed in a crescent area of the cytoplasm (Fig. 4Ag) where Rab6p is not detected (Fig. 4Ar). Then, in the following sections, the Rab6p signal became progressively expressed and strictly co-localized with Sy, as illustrated in Fig. 4B, which represents the 5th paired section. This situation was maintained in the three following paired sections. Then, the Sy signal progressively decreases in intensity in the Golgi area, whereas the Rab6p signal remains conspicuous, but in a more juxta-nuclear crescent (Fig. 4C, 11th section). On merged images of both channels (Fig. 5B), one can see, moreover, a sort of polarization of the respective distribution of both markers within Golgi subcompartments: the co-localization pattern is observed in the tra n s area, whereas the Rab6p signal becomes progressively restricted to a more juxta-nuclear Golgi subcompartment. This preferential localization of Sy in a tra n s Golgi position is consistent with the good co-localization pattern observed with TGN 38, at every section examined and on merged images of both channels (Fig. 5C).

In the other neuron compartments, clearly Sy and Rab6p coexist in large emerging neurites (Fig. 5B). How-ever, the heterogeneous distribution of the staining and the limit of the resolution of the light microscope (200 nm) does not permit us to discriminate a cytoplasmic from a membrane staining (Fig. 5B). In varicosities, a strict co-localization of Sy and Rab6p was often observed. By comparing successive paired optical sections (see Fig. 3A-E), one can follow very well the superposition of both signals over the 2nd to the 6th section (from bottom to top), indicating that varicosities are lying very close to the culture substratum and that their mean diameter is around 1.5 µm. On merged images of both channels (Fig. 5A), various degrees of dual staining are observed from yellow to green, sometimes red, suggesting that the relative proportion of both markers varies among varicosities. In some cases, the comparison of successive sections (Fig. 3A-E) revealed a preferential localization of Rab6p near the substratum (Fig. 3Ar) with a progressive shift to the Sy green labeling toward the top of the culture medium (Fig. 3Eg).

In conclusion, the confocal analysis confirms that, within the limits of the resolution of the optical microscope, Rab6p and Sy are co-localized in the tra ns-most Golgi compartments, but are also sometimes dissociated in other Golgi subcompartments. The degree of apparent spatial dissociation of the two antigens also varies from one neuron to another. A similar diversity in the respective intensity of the Rab6p and Sy signals was also observed in varicosities. Taken together, such observations are compatible with a transient association of Rab6p to Sy-containing membranes during their postGolgi transport into neurites and varicosities.

Rab6p could be visualized in the Golgi region of almost all neurons, regardless of the developmental stage (5 DIV to 23 DIV). The peroxidase reaction product was localized to the cytoplasmic face of the membrane of a subcompartment of the Golgi cisternae as well as to numerous buds and vesicles in their vicinity (Fig. 6). The immunostaining was always polarized, as previously observed in NRK cells (Goud et al., 1990). Although it is difficult to assign the polarity of the labeled subcompartment, its topographical localization and morphology suggest that Rab6p is predominantly associated with membranes of a t ra n s-Golgi compartment, including cisternae and vesicles. Such a distribution is reminiscent of that previously reported for Sy (Tixier-Vidal et al., 1988). The membrane of small dense core vesicles was not significantly labeled (Fig. 6).

In cultures taken before synapse formation (5-6 DIV), Rab6p immunostaining was found in the cytoplasm as well as associated with the membrane of small vesicles, which were present in varicosities and some growth cones (not shown). These vesicular structures have been previously immunostained with a monoclonal antibody against Sy (Tixier-Vidal et al., 1988).

In cultures containing differentiated synapses (12-23 DIV), Rab6p immunostaining was clearly associated with the membrane of SSVs, but only in a portion of synaptic boutons (Fig. 7). Moreover, the intensity of the vesicle labeling and the proportion of labeled vesicles varied greatly among terminals, and, in numerous axon terminals, SSVs were not immunostained (Fig. 7). Neurites were not stained, except at the plasma membrane where discontinuous patches were sometimes labeled (Fig. 7). The plasma membrane of synaptic boutons was also discontinuously labeled, particularly at the contact with neurites (Fig. 7). However, this plasma membrane labeling may represent a diffusion artefact of the peroxidase reaction.

Immunoblotting of total cell extracts using affinity-purified anti-Rab6p antibodies revealed a single band, which migrated with the same electrophoretic mobility as reference Rab6p (Fig. 8). In two independent culture series, Rab6p concentration was determined as a function of culture times. As shown in Fig. 8, Rab6p concentration increased 3-to 4-fold during the first 10-12 days, which corresponds to the period of active axonal growth and migration of Sy to varicosities and axon terminals. Afterwards, the level of Rab6p tended to decrease. By contrast, the level of Sy increased, under identical conditions of cultures, continuously for up to 21 days, at least (Tixier-Vidal et al., 1992).

Rab proteins, including Rab6p, exist at a steady state in cytoplasmic (soluble) and membrane-bound pools (Goud et al., 1990). In order to determine whether the ratio between the two pools might vary in developing neurons, cells were fractionated into high speed supernatants (cytosol) and high speed pellets (crude membranes) after various days in culture. Interestingly, Rab6p was found to be mostly soluble at the very beginning of culture (days 0 and 2; Fig. 9). Then, at 6 DIV, the proportion of soluble Rab6p stabilized at around 30-40% of the total protein, a value comparable to that found in most cell types (Goud et al. 1990 and unpublished results). In our culture system, the period of increased association of Rab6p with membranes corresponds to neurite initiation and elongation and to the onset of Sy immunostaining of varicosities (Tixier-Vidal et al., 1988, 1992).

To establish further the association of Rab6p with Sycontaining membranes, these membranes were immunoisolated from 6- and 15-day cultures after fractionation on sucrose gradients (fractions 2-6) as previously described (Leube et al., 1989; see also Materials and Methods). By western blotting (Fig. 10), these membranes were found to contain, as expected, two proteins characteristic of the membrane of SSV proteins SV2 and Sy. These proteins were more abundant at day 15 than at day 6. Only a faint immunosignal for Rab6p was detected at day 6 (Fig. 10a). To ascertain the presence of Rab6p on immunoisolated Sy membranes, an experiment was repeated with a 6-day-old culture using a large number of dishes. In this case, a significant immunosignal for Rab6p was obtained on immunoisolated membranes (Fig. 11).

Electron microscopic examination of immunobeads obtained from 6- and 15-day-old cultures revealed the presence at their surface of numerous small vesicles of regular diameter, which were intermingled with elongated cisternae of similar diameter (not shown). These cisternae may represent Sy-containing Golgi-derived membranes. In younger cultures, the proportion of small vesicles as compared to other membranes was smaller than at the oldest stage.

Rab6p, a ubiquitous small GTP-binding protein is found in the present study to be associated with membranes of two neuron compartments: (1) in perikarya, on cisternae and vesicles of a Golgi subcompartment, presumably trans, regardless of the developmental stage; and (2) in varicosities and axon terminals, on vesicles that possess morphological and biochemical features of SSVs, at certain developmental stages.

The Golgi-membrane association of Rab6p is consistent with previous observations performed in NRK cells (Goud et al., 1990). However, in neurons, Rab6p-labeled vesicles and buds in the Golgi region are much more abundant than in NRK cells (Goud et al., 1990).

The association of Rab6p with synaptophysin-containing membranes is firstly supported by our morphological findings. By double immunostaining and confocal laser scanning microscopy, we have found that, in the limit of the resolution of the light microscope, Rab6p and Sy were colocalized in a Golgi subcompartment that can be assigned to the TGN in view of the co-localization of the immunosignals for TGN38 and Sy that we observed separately. In addition, a co-localization of Rab6p and Sy was often observed in large emerging neurites and in axonal varicosities at young stages of neuron development. By electron microscopy, Rab6p immunoperoxidase reaction product was observed on membranes (cisternae and vesicles) of the Golgi zone and of SSVs that were previously found to be immunostained with A-Sy (Tixier-Vidal et al., 1988). Attempts to localize simultaneously Rab6p and Sy at the electron microscope level, using a combination of immunoperoxidase and immunogold methods that previously permitted the simultaneous localization of Sy and Secretogranin II (Kagotani et al., 1991), were unsuccessful with Rab6p: binding of one antibody seemed to inhibit the binding of the other. Secondly, Rab6p could also be detected by western blotting on immunoisolated Sy-containing membranes obtained at an early stage of synaptogenesis (6 DIV).

Our data indicate that Rab6p expression in neurons is developmentally regulated as shown morphologically and biochemically. Rab6p is expressed in the hypothalamus at an early post-mitotic neuroblast stage (15th fetal day) and is concentrated in the Golgi area of neurons examined at the onset of neuritogenesis. Its migration into growing neurites and axons and its concentration in varicosities correlate with its association with membranes from 0 to 6 DIV, and its increased concentration from 4 to 12 DIV. Then, the restriction of Rab6p expression to a minor population of varicosities and boutons correlates with the steady state of its membrane association and cell concentration. These variations are likely to reflect neuronal developmental changes. Indeed, our cultures contain no more than 20% of flat, non-neuronal cells (Faivre-Bauman et al., 1984), which exhibit a perinuclear reticulum also immunostained with A-Rab6p. In addition, in our medium conditions, the proportion of non-neuronal cells remains constant and the distribution of Rab6p immunostaining of those cells did not vary with time in culture (data not shown).

Our findings suggest that Rab6p association with Sy bearing vesicles is transient: (1) its concentration in culture stops increasing after the onset of synapse formation, which itself lasts for several days as attested by the continuous increase of the number of synaptophysin-immunostained varicosities and boutons and of Sy concentration (Tixier-Vidal et al., 1992). (2) In the Golgi zone, depending on the neurons, the co-localization of Rab6p and Sy is sometimes restricted to a Golgi trans subcompartment and even to a post-Golgi compartment (emerging neurites), whereas Rab6p only is detected in a presumably medial compartment. (3) In varicosities taken at developmental stages where Rab6p and Sy are co-localized, the relative intensity of the two signals greatly varies; moreover, at increasing ages of cultures, after the onset of synaptogenesis, Rab6p is found in a decreasing proportion of varicosities and synaptic boutons.

Based on data obtained from studies on Sec4p and Ypt1p, a model for the function of small GTP-binding proteins in membrane traffiking has been presented (Bourne, 1988; Walworth et al., 1989). This model provides an explanation for how small GTP-binding proteins are used in transport vesicle targeting. It predicts that the vectorial transport of vesicles between successive intracellular compartments is coupled to the cycle of a small GTP-binding protein specific for a given transport step between an inactive (GDP-bound), cytoplasmic form and an active (GTP-bound) form associated with membranes. Evidence now exists to show that some of the rab proteins are indeed controlling various targeting/fusion events (for reviews, see Goud and McCaffrey, 1991; Pfeffer, 1992). However, recent data indicate that small GTP-binding proteins might also control the formation of transport vesicles. Such a role has been attributed to ARF (ADP-rybosylation factor) proteins in mammalian cells (Serafini et al., 1991) and to Sar1p in yeast (D’Enfert et al., 1991).

In embryonic neuronal cells, Rab6p may play a critical role in sorting and post-Golgi transport of Sy-containing membranes. The transientness of the association of Rab6p with membranes during active synaptogenesis supports this hypothesis. Moreover, it was recently reported that GTPγs inhibits fast organelle transport along axonal microtubules of the squid giant axon and that this effect might involve small GTP-binding proteins (Bloom et al., 1993). A role for other proteins of the rab family has been proposed in neurons. Rab2p has been shown to increase neural adhesion and neurite initiation (Ayala et al., 1990). Rab3Ap, a rab protein found on the membrane of SSVs, could be involved in targeting/fusion of these vesicles with synaptic membranes (Fischer von Mollard et al., 1991). Interestingly, Rab3Ap appears to be absent from the Golgi area in adult as well as in developing neurons, suggesting that it associates with vesicles distal to the Golgi complex (Matteoli et al., 1991). Altogether, these data indicate that during the cascade of events underlying neuron development, different Rab proteins could be involved, each at a precise step.

We thank Dr A. Zahraoui for providing us with purified recombinant Rab3Ap. We thank Prof. Gérard Buttin for his constant support, and Drs Mary McCaffrey and Michèle Roa for critical reading. We acknowledge the help of Mr R. Hellio for the CLSM analysis and of Mr C. Pennarun for the preparation of photographs. This work was supported by grants from the Centre National de la Recherche Scientifique and from the Fondation pour la Recherche Médicale to Dr A. Tixier-Vidal and to Dr B. Goud, by grants from the Université Pierre et Marie Curie and from the Ligue Nationals contre le Cancer to Dr B. Goud and by grants from the Deutsche Forschungsgemenshoft (Wi 617/5-2) and Forschungsrat Rauchen und Gesundheiz to Dr B. Wiedenmann. The co-operation between B. Wiedenmann and A. Tixier-Vidal was supported by the German Academic Exchange Service and the french Ministère des Affaires Etrangères (PROCOPE).