The contractile vacuole network of Dictyostelium as a distinct organelle: its dynamics visualized by a GFP marker protein

In addition to the endoplasmic reticulum and the Golgi apparatus, Dictyostelium cells contain two major vesicle systems with distinct functions: the endosomal system and the contractile vacuole (CV) network. Both these systems are connected to the plasma membrane. The endosomal pathway of nutrition begins with the uptake of particles by phagocytosis or, in mutated strains, by the ingestion of fluid by macropinocytosis (Hacker et al., 1997). This pathway ends with exocytosis after a trafficking and membrane processing period of about 90 minutes, which means that plasma membrane and endosomal membranes are interconvertible.

The CV system consists of ducts, cisternae and bladders (Heuser et al., 1993), and is transiently connected with the plasma membrane by pores through which the bladders expel water. The CV system acts primarily as an osmoregulatory organelle, but is also involved in Ca²⁺ regulation (Moniakis et al., 1999). Endosomes and contractile vacuoles share two proteins within or at their membranes, the vacuolar H⁺-ATPase (Temesvari et al., 1996) and rabD, a rab4-like small GTPase (Bush et al., 1994). These common markers suggested that the two compartments are connected to each other by fusion of their membranes (Bush et al., 1996).

In order to distinguish the endosomal and CV compartments in living cells, specific markers are required. Endosomes can be loaded with fluorescent dextran (Cardelli et al., 1989; Aubry et al., 1993; Hacker et al., 1997), and late endosomes are selectively decorated by GFP-tagged vacuolin B (Jenne et al., 1998). For the CV system, two markers applicable in vivo have been described, the styryl dye FM4-64 and GFP-tagged drainin. The fluorescent dye transiently highlights the bladder but does not resolve the ducts, and it gradually occupies also other internal membranes (Heuser et al., 1993). Drainin is a peripheral membrane protein that is specifically involved in discharge of the bladder (Becker et al., 1999). In accord with this function, GFP-tagged drainin decorates the bladder of the CV system.

Since drainin-GFP does not significantly label the ducts of the CV system, no specific and stable marker is available for the labeling of the entire CV network in vivo. We fortuitously discovered an appropriate marker in an attempt to supply a cell-adhesion molecule, the contact site A (csA) glycoprotein, with a GFP tag. CsA has been the prototype of proteins modified by a ceramide-based phospholipid anchor (Stadler et al., 1989). A previous study showed that this anchor is not essential for the protein’s function in cell adhesion, but is responsible for excluding the protein from a pathway of internalization and subsequent degradation (Barth et al., 1994). Therefore, without impairing the activity of csA as a cell-adhesion molecule, the lipid anchor could be replaced by the transmembrane domain and cytoplasmic tail of another protein in order to tag the C terminus of the protein with GFP. The other protein used to construct a transmembrane chimera has provisionally been designated as P29F8, the number of a cDNA clone comprising its full-length coding region (Barth et al., 1994). We refer here to P29F8 as dajumin, combining the initials of those who contributed to its discovery and analysis.

When the C terminus of dajumin was tagged with GFP, the fusion protein unexpectedly localized to the membranes of the CV system rather than to the cell surface. With this observation as a starting point, we studied the CV network during its activity in vivo. The CV system is known for rapid changes in its organization that accompany each cycle of activity. In the absence of any label, reflection interference contrast microscopy visualized elements of the system that are in close proximity to the plasma membrane (Gingell et al., 1982). This technique provided evidence that ducts and bladders are interconvertible (Heuser et al., 1993).Dajumin-GFP constructs are stably integrated into the CV membranes. They proved to be highly specific not only to the network of ducts and bladders in interphase cells, but also to the fragmented network of cells in mitosis (Zhu et al., 1993). Using these constructs as markers of the CV complex, we address the following questions. (1) Is there a connection between the CV network and the endosomal system; (2) is there an exchange of membrane constituents between the bladder and the plasma membrane during discharge; (3) which are the steps in differentiation of a bladder out of the network of ducts; (4) how is the fragmented CV network reorganized after mitosis?

To construct the GFP-tagged csA/dajumin chimera (Fig. 1A), the csA/dajumin construct was isolated from the pDEV/CP fusion vector (Barth et al., 1994). For the full-length dajumin sequence, a genomic clone was amplified by PCR. These fragments were cloned into the EcoRI site of pDEX RH (Faix et al., 1992) in frame with the GFP S65T sequence (Heim and Tsien, 1996), which was inserted into the HindIII site of the vector. The hexapeptide linker EFKKLK between dajumin and GFP resulted from the cloning procedure. The amplified dajumin sequence and the correct fusion of the csA, dajumin, and GFP coding sequences were confirmed by custom sequencing (Toplab, Martinsried).

Cells transformed by electroporation were selected for G418 resistance using 20 µg/ml of Geneticin (Sigma) and subsequently cloned by spreading onto SM agar plates with Klebsiella aerogenes. G418-resistant clones were screened for the expression of vector-encoded proteins by microscopic assessment of GFP fluorescence in growth-phase cells.

The D. discoideum parent strain of transformants was AX2 clone 214. For dajumin-GFP as a CV marker, two independent transformant clones, HG1752 and HG1753, were used with indistinguishable results. In each clone, the dajumin-GFP construct was brilliantly expressed in almost all of the cells. For the expression of csA/dajumin-GFP, clone HG1764 was used. All clones tested expressed this chimera only in a small percentage of cells. The GFP-α-tubulin expressing strain was HG1668 (Neujahr et al., 1998), the control strain producing free GFP was HG1694. For Fig. 3, the csA-null mutant HG1287 was used as a control, HT-C1 as a csA expressing and HT-CP8 as a csA/dajumin expressing transformant of this mutant (Barth et al., 1994).

Cells were cultivated in plastic Petri dishes in liquid nutrient medium, washed in 17 mM K/Na-phosphate buffer, pH 6.0, to remove fluorescent compounds from the medium, and subjected to confocal microscopy.

For antibody labeling, cells were washed twice in 17 mM Na/K-phosphate buffer, pH 6.0, and allowed to adhere to glass coverslips for 20 minutes. The cells were then fixed with picric acid/paraformaldehyde (Humbel and Biegelmann, 1992), postfixed with 70% ethanol, and incubated overnight with mAb 24-210-2 against O-linked oligosaccharides or mAb 41-71-21 against the csA protein (Bertholdt et al., 1985), followed by 2 hours of incubation with TRITC-conjugated goat anti-mouse IgG (Jackson ImmunoResearch). Images were taken using an Axiophot 2 microscope (Zeiss) with a ×100/1.3 Neofluar objective and a cooled SensiCam CCD camera (PCO Computer Optics).For immunoblotting, proteins of total cell homogenate from 1×10⁶ cells per lane were resolved by SDS-PAGE in 10% gels, blotted and immunolabeled either with mAb 264-449-2 against GFP (a gift from Markus Maniak), mAb 24-210-2 against O-linked carbohydrate epitopes or mAb 33-294-17 recognizing the csA protein moiety (Bertholdt et al., 1985). For detection of the first antibodies, goat anti-mouse IgG conjugated with alkaline phosphatase was used (Jackson ImmunoResearch).

For covalent labeling of plasma membrane proteins in vivo, HG1752 cells were washed twice in 17 mM Na/K-phosphate buffer, pH 8.0, and resuspended in a solution of 50 µM of the monofunctional NHS ester of the fluorescent dye Cy 3.5 (Amersham Pharmacia Biotech) in the same buffer. The cells were rotated slowly in the dark at 4°C for 30 minutes, washed twice in 17 mM K/Na-phosphate buffer, pH 8.0, and once in the same buffer adjusted to pH 6.0. Cells were warmed up to room temperature, allowed to adhere to glass coverslips, and subjected to confocal imaging with or without agar overlay (Yumura et al., 1984).

For the labeling of endocytic vesicles with a fluid-phase marker, adherent cells were incubated with 17 mM K/Na-phosphate buffer, pH 6.0, containing 2 mg/ml of TRITC-dextran (Sigma). To label phagocytic vesicles, cells were incubated with heat-killed yeast particles (Sigma) in liquid nutrient medium for 1 hour. Thereafter, the medium was replaced by 17 mM K/Na-phosphate buffer, pH 6.0, and the cells were subjected to agar overlay. Confocal imaging was performed with an LSM 410 laser scanning microscope (Zeiss) equipped with a ×100/1.3 Plan-Neofluar objective. 3-D images were constructed using AVS software (Advanced Visual Systems) as described by Neujahr et al. (1997).

Cells were suspended in 5 mM Na-glycinate buffer, pH 8.5, containing 100 mM sucrose (Padh et al., 1991), and lysed by passing through two layers of Nucleopore track etch membranes, pore size 8 µm (Corning). The lysate was layered on top of a gradient of 28, 40, and 60% sucrose in the Na-glycinate buffer, and centrifuged for 1 hour at 40,000 rpm in a Sorvall SW 41 rotor.

With the intention of connecting the csA cell adhesion protein through a transmembrane domain with GFP, a csA/dajumin-GFP chimera was constructed which turned out to label the bladders as well as the ducts of the CV system (Fig. 1A). The series of confocal fluorescence images shown in Fig. 1B illustrates the periodic changes in the pattern of the network that are associated with cycles of vacuole activity.

To explore the possibility that the dajumin portion is responsible for the unpredicted localization of the GFP-tagged csA/dajumin chimera to the CV complex, we tagged full-length dajumin with GFP (Fig. 1A). In a plane close to the substrate surface, the fusion protein visualized the characteristic network of ducts and cisternae of the CV complex, in which larger cisternae act as bladders (Fig. 1C). An optical cross-section through the cisternae slightly beyond the network of ducts showed not only discharge of a bladder but also the merging of cisternae along connecting ducts (Fig. 1D).

Almost all the cells in a population expressed the dajumin-GFP, and its fluorescence was brilliant. This construct could be used, therefore, to visualize the three-dimensional organization of the contractile vacuole complex. Serial sectioning of an entire cell by a set of confocal planes revealed extension of the CV network around the entire cell body, although it is most densely knit at the substrate-attached area of the cell surface (Fig. 2). Similarly, bladders are most often formed in the bottom half, but also close to the upper surface of the cell.

The restriction of the csA/dajumin-GFP fusion protein to the CV network was unpredicted since csA/dajumin acts as a cell-adhesion protein, which means that the protein is exposed on the cell surface (Barth et al., 1994). To confirm this observation, we immunolabeled cells expressing either normal, lipid-anchored csA or csA/dajumin without the GFP tag (Fig. 3). Using an anti-csA antibody, the chimera was not only localized to intracellular vesicles, some of them recognizable as contractile vacuoles (arrowhead in Fig. 3D), but also to the cell surface (Fig. 3C-F). Similar to the phospholipid-anchored csA protein in Fig. 3A, the chimera accumulated at areas of intercellular adhesion (Fig. 3E) and remained there on tethers connecting two separated cells (Fig. 3F). In accord with the location of the csA/dajumin chimera on the cell surface, cells expressing this protein strongly tended to agglutinate. One has to conclude from these data that the fusion of GFP to the C terminus of csA/dajumin supports retention of the protein within the CV system.

To provide evidence that dajumin enters the CV complex after being passaged through the Golgi apparatus, modification of the protein by O-glycosylation was examined. Among a series of monoclonal antibodies that recognize exclusively epitopes on O-glycosylated proteins (Bertholdt et al., 1985; Hohmann et al., 1987), mAb 210 was selected because the epitopes for this antibody were restricted to two O-glycosylated proteins, csA and dajumin. Both these developmentally regulated proteins were undetectable in growth-phase cells, and both were clearly recognized after 6 hours of starvation in aggregating wild-type cells (Fig. 4A and B). The dajumin-GFP fusion protein, which was constitutively expressed under control of the actin15 promoter, was recognized by mAb 210 in growth-phase as well as in aggregating cells, and was also detected by an anti-GFP antibody (Fig. 4A and C). These results indicate that the GFP-fusion protein, like normal dajumin, is passaged through the Golgi apparatus where it is O-glycosylated before being sorted to the CV system.

In Fig. 4C a protein fragment with an apparent molecular mass of 36 kDa is labeled with anti-GFP antibody. This fragment is about 8 kDa larger than free GFP (Fig. 4D), and thus appears to be a product of proteolytic cleavage at a lumenal site of dajumin.

The dajumin-GFP clusters left after discharge of a bladder are often flat and spread beneath the cell boundary, so one might falsely assume them to be incorporated into the plasma membrane. However, live observation revealed that the clusters can be refilled, which means they represent folded membranes underlying the cell surface rather than protein that has diffused from the vacuole into the plasma membrane.

Since the GFP-tag may prevent not only the csA/dajumin chimera but also full-length dajumin from being incorporated into the plasma membrane, GFP fusion proteins may not be representative for the exchange of membrane proteins between the cell surface and the CV system. To explore the possibility of an exchange during the period of discharge of the bladder, when the plasma membrane is connected to the CV system through a pore, the fluorescent dye Cy 3.5 was covalently and unselectively conjugated to constituents of the cell surface.

As a control, Fig. 5A shows the internalization of a cell-surface area during macropinocytosis. This entry into the endosomal pathway is characterized by the budding of a fluid-filled vesicle from the plasma membrane. The Cy 3.5 label reliably identified the internalized membrane during and after the uptake process and, at a later stage, decorated also membranes folded within the lumen of the vesicle (410 second frame of Fig. 5A).

In Fig. 5B, the cell-surface label is superimposed to the dajumin-GFP fluorescence. Several vacuoles are refilled during the period of recording. The sequence of shape changes is best represented by the vacuole on the left (arrowheads in the 0 and 280 second frames of Fig. 5B). The filling bladder spreads initially beneath the cell surface, most obviously in the 130 second frame of Fig. 5B. Subsequently, the area of contact with the cell cortex is reduced, so that the vacuole rounds up before discharge of its contents (160 and 330 second frames of Fig. 5B). No overlap in the vacuole and cell-surface label has been recognizable during the periods of discharge and refilling of the bladders, indicating that translocation of proteins from the plasma membrane to the vacuolar membrane during a cycle of CV activity is negligible.

After a period of more than 10 minutes of internalization into endosomes, the membrane label of Cy 3.5 fades out either by bleaching or proteolytic degradation. Therefore, the experiments shown in Fig. 5 do not rule out an exchange between the endosomal and the CV system at a later stage of endosomal processing. In order to probe for an exchange, we have used two non-digestible markers of the endosomal pathway, boiled yeast particles and TRITC-dextran (Rauchenberger et al., 1997). Fig. 6A depicts phase-contrast images and confocal sections through two cells loaded with yeast. The numerous ingested yeast particles visualized by phase contrast (in blue) occupied much of the intracellular space in these cells. The CV network (in green) was well developed in a particle-free plane close to the substrate surface, and otherwise was squeezed in form of ducts and vesicles into the space between phagosomes filled with the yeast particles. In cells not fed with yeast particles, endosomes are recognized by phase contrast as fluid-filled vesicles of variable sizes, which are not labeled with dajumin-GFP. In the example of Fig. 6B, a tube of the CV complex surrounds in a U-turn one of the larger unlabeled vesicles, suggesting close contact but no fusion of the two vesicle systems. For unequivocal distinction of endosomes from structures of the CV complex, cells were loaded with the fluid-phase marker TRITC-dextran. Since the average residence time between endocytosis and exocytosis of a fluid phase marker is about 60 minutes in Dictyostelium cells (Jenne et al., 1998), bathing the cells for at least 90 minutes in a solution of TRITC-dextran visualizes the lumen of endosomes at all stages of their pathway. To provide optimal conditions for optical inspection, the cells were compressed between a glass and agar surface. In the double-labeled cells we never observed merging of the two labels: the endosomal fluid-phase marker did not leak into the CV network, nor did we recognize an enrichment of dajumin-GFP at the site of endosomal membranes. Analysis of double-labeled cells by series of confocal sections revealed that even ducts of the CV system and endosomes residing in close contact to each other did not merge into one hybrid vesicle filled with TRITC-dextran and decorated on its surface with GFP-dajumin (Fig. 6C).

Since no TRITC-dextran was detectable in vesicles labeled with dajumin-GFP, it appeared unlikely that external fluid enters the CV system during discharge or refilling of a bladder. The lack of significant fluid uptake is explicitly demonstrated in Fig. 6D for the two bladders marked by arrowheads. These data indicate that during discharge of a bladder the outside-directed flow surpasses the rate of diffusion of macromolecular dextran into the opposite direction, and they confirm that the bladder is closed against the environment during the filling period. As a bonus, the sequence of Fig. 6D shows in the 490 and 500 second frames the fusion of an expanded bladder with a second one that is just starting to refill.

To establish that constituents of the CV complex are separable from endosomes, double-labeled cells were lysed and particles fractionated in a sucrose gradient according to the method of Padh et al. (1991). The endosomes loaded with TRITC-dextran were recovered on the bottom of the gradient in the 60% sucrose fraction, in coincidence with fractionation of the lysosomal marker acid phosphatase (Barth et al., 1994). GFP labeled structures microscopically identifiable as ducts and bladders were separated from the endosomes. They peaked in the 28% sucrose fraction in accord with acidosomes (Nolta et al., 1991), which are thought to be fragments of the CV complex (Heuser et al., 1993; Bush et al., 1994).

By the labeling of fixed cells with calmodulin antibodies, Zhu and Clarke (1992) have shown that the CV network is fragmented during mitosis. In accord, many small contracting vacuoles are seen in time-lapse movies to be dispersed throughover the periphery of dividing cells (Gerisch, 1964). The disconnected vesicles in mitotic cells make it possible to study the periodic activity of a minimal unit of the CV system apart from its integration into the circumferential network of interphase cells (Fig. 7A). A typical sequence of bladder-duct interconversion in a fragment of the CV complex is shown in Fig. 7B. In the filling phase the bladder is connected with short ducts that have blind ends. These pieces of ducts disappear or are disconnected when the bladder starts to contract. The membranes of the bladder collapse during contraction into a fluorescent cluster unresolvable by optical means. The high local fluorescence intensity in this cluster indicates that the membranes are strongly folded, enveloping a minimum of lumenal space. Refilling is initiated by the radial outgrowth of tubules from the cluster. Subsequently, the lumen of the complex dilates locally into ventricles, at the site of the previous bladder and also within the elongating tubules. Finally these ventricles merge, giving rise to an expanding vacuole, thus reconstituting after a period of 75 seconds the minute complex of a bladder with short connected ducts.

At the end of mitosis, the CV fragments start to re-assemble around each of the daughter nuclei (Fig. 7C). The separate CV vesicles consecutively fuse with each other in a variable pattern. In comparing all records of mitotic cells we have collected, we could not find a pre-fixed site of post-mitotic CV assembly. This is clearly different from re-assembly of the Golgi apparatus. Golgi vesicles assort along microtubules and are transported towards their minus end, where the Golgi apparatus is reconstituted close to the centrosome (Schneider et al., 1999). In contrast, CV vesicles cluster and fuse at any side of the nuclei, including the zone opposite to centrosome position. These data exclude the centrosome as the center of CV assembly.

In Fig. 7D the microtubule system is visualized by GFP-α- tubulin in order to relate CV reassembly to the state of mitosis. The spindle is already disassembled when cleavage of the cell progresses: the 100 second frame of Fig. 7D represents the same stage of cytokinesis as the 10 second frame in Fig. 7C, where CV reassembly is just commencing. The red label in Fig. 7D shows endosomes loaded with TRITC-dextran, demonstrating the absence of any pattern of endosome assembly in the course of mitotic cell division.

To establish that contractile vacuoles stay separate from endosomes during mitosis, we labeled cells expressing dajumin-GFP with TRITC-dextran and searched for dividing cells. Through all stages of mitosis, the GFP-labeled vacuoles and TRITC-dextran loaded endosomes existed as two distinct classes of vesicles (Fig. 7E). This example has been chosen because it demonstrates in the upper daughter cell how a CV ring around the nucleus is generated, giving rise to the arrangement of ducts and bladders established in the interphase cell of Fig. 1B. When dividing cells were strongly compressed between a glass and agar surface, the endosomes were slightly concentrated in the mid-zone, while the CV vesicles assembled in the polar regions (Fig. 7F).

In this paper we have introduced a specific marker of the contractile vacuole system of Dictyostelium cells. As a transmembrane protein, dajumin-GFP provides a reliable label to study the dynamics of the entire CV complex in interphase and mitotic cells. The CV compartment illuminated by dajumin-GFP shows the organization and variability of unlabeled CV structures visualized by reflection interference contrast microscopy (Heuser et al., 1993). The fluorescent network is also coincident with electron micrographs of freeze-dried broken cells showing clusters and arrays of ducts and cisternae (Heuser et al., 1993; Clarke and Heuser, 1997). In particular, the proposed interconversion of ducts into bladders during cycles of activity in the CV complex (Heuser et al., 1993) is reflected in the fluorescence images obtained with dajumin-GFP as a marker (Figs 1B and 7B).

Since the vacuolar H⁺-ATPase is present not only on the CV complex but also on the endosomes, dajumin-GFP is the only integral membrane protein available to unequivocally identify constituents of the contractile vacuole system in cell fractions. Potential applications of the dajumin-GFP marker include the analysis of mutants with impaired structure and function of the CV complex. Examples are mutants deficient in clathrin heavy chains or drainin, and cells overexpressing dominant negative rabD. In the clathrin mutant, bladders are rudimentary (Ruscetti et al., 1994), in the drainin mutant they are excessively expanded and arrested at a stage preceding discharge (Becker et al., 1999), and in the rabD mutant the CV complex appears to be collapsed into one patch (Bush et al., 1996).

It has not been within the scope of this paper to investigate the function of dajumin used here as a marker. Dajumin knock-out mutants have shown that the protein is not essential for survival of the cells (Müller-Taubenberger, 1989). Endogenous dajumin expressed under control of its own promoter is a strictly developmentally regulated protein (Fig. 4A). In the growth-phase cells, which have been used throughout the present study, no dajumin mRNA has been detected by Northern blotting (Gerisch et al., 1985; in this reference dajumin has been referred to as P29F8). Dajumin mRNA accumulates during development to the aggregation stage in response to periodic cyclic AMP signals, in parallel with a number of other membrane proteins that serve different functions in aggregating cells. The role of dajumin at this stage remains to be elucidated.

Our data underline previous results indicating that the CV network is an intracellular compartment unique in its capacity to coordinate activities in space and time by membrane fusion, extension and vesiculation of tubules, pore formation, and contraction (Heuser et al., 1993; Clarke and Heuser, 1997). This is illustrated by the steps involved in discharge of the bladder. The bladder is ready for discharge when it reaches a certain size. Mediated through a sensing and signal transduction mechanism that probably involves drainin, the state of filling appears to regulate the interaction of the bladder with the plasma membrane (Becker et al., 1999). In a first stage, the expanding bladder is attached to the cell cortex over a large area of its surface, as depicted in the 20 and 130 second frames of Fig. 5B. In drainin-null mutants, largely oversized bladders are arrested in that particular state. Actin filaments and palisade-like arrays of spacers separate the vacuolar membrane from the plasma membrane in drainin-deficient cells. In wild-type cells, this state is transient and followed by rounding up of the bladders in preparation of their discharge (Fig. 5B). When the pore between bladder and plasma membrane is opened, ducts have to be closed or detached from the bladder in order to avoid discharge of fluid into the ducts rather than the environment (John Heuser, personal communication). This separation of the bladder from surrounding ducts is not always perfect; sometimes we have seen the blowing up of adjacent tubular regions during the contraction of a bladder.

A functional differentiation between bladders and ducts is illustrated by the preferential association of drainin with the bladders at all stages of a contraction cycle (Becker et al., 1999). On the other hand, our results substantiate previous observations that ducts and bladders are readily interconverted during filling and discharge (Heuser et al., 1993). Locally regulated contractility is one of the mechanisms involved in bladder to duct conversion. The transfer of fluid from one vacuole into another, and the differentiation of membrane clusters into ducts and bladders during the filling period (Figs 1D, 5B and 6D) suggests the presence of a delicately controlled system for contractility in the CV complex. The availability of dajumin-GFP as a specific marker makes it possible to localize unconventional myosins or other motor proteins to the CV arrays and to identify constituents of the cytoskeleton that anchor the network at the cell cortex.

Dajumin has an N-terminal hydrophobic leader sequence and a transmembrane domain in its C-terminal half, indicating that the protein is integrated into the membrane upon its synthesis in the ER. O-glycosylation is a post-translational modification in Dictyostelium as in other cells (Hohmann et al., 1987). The presence of O-linked oligosaccharides on dajumin and its GFP fusion product suggests that the protein is passaged through the Golgi apparatus on its way to the CV complex. The question is how dajumin-GFP is distinguished from proteins sorted to the plasma membrane or to the endosomal pathway. In the csA/dajumin-GFP chimera the N-terminal csA portion is localized on the lumenal side of the vacuolar membrane. Since csA is a phospholipid anchored cell-adhesion molecule normally kept on the cell surface (Barth et al., 1994), it is unlikely that the csA portion targets the chimera to the CV complex. Since free GFP is uniformly distributed in the cytoplasm of Dictyostelium cells and slightly enriched in the nuclear matrix, the C-terminal GFP sequence cannot be responsible for targeting the chimera to the CV system. Therefore, localization to the CV complex is attributed to the dajumin fragment in the middle of the csA/dajumin-GFP chimera. This dajumin fragment consists of the C-terminal region of dajumin including a stretch of 18 amino-acid residues on the lumenal side of the membrane, a trans-membrane domain of 23 hydrophobic residues and a cytoplasmic tail of 34 residues (Fig. 1A). It will be necessary to dissect the protein further in order to analyse the sequence requirements for protein targeting to the CV complex.

Although the GFP moiety has no targeting capacity by itself, it assists in restricting the fusion protein to the CV system. Untagged csA/dajumin is detectable by antibody on the cell surface (Fig. 3), and is active there as a cell adhesion protein (Barth et al., 1994). Only the GFP-tagged csA/dajumin is efficiently entrapped within the CV system (Fig. 1B). The presence of the GFP moiety is therefore crucial for the use of dajumin constructs as CV markers.

The results obtained with the dajumin-GFP marker indicate that membrane flow and the exchange of contents between the CV complex and other compartments is strictly controlled. During discharge of the bladder, dajumin-GFP accumulates beneath the plasma membrane in patches that remain part of the contractile vacuole system (Figs. 5B and 6D). This retention of the marker argues against an unselective transit of proteins from the vacuolar membrane to the cell surface. Only exceptionally we have found dajumin-GFP at the plasma membrane, perhaps resulting from a slight injury of the cells. (The label at the upper border of the cell in the 440 to 500 second frames of Fig. 6D could be due to such an effect.) Evidence for regulated localization of a vacuolar protein has been provided by Moniakis et al. (1999), who reported that the Ca²⁺-ATPase PAT1 populates the plasma membrane in response to high extracellular Ca²⁺ concentrations.

Our data provide no evidence for a collateral communication between the endosomal compartment and the CV complex. These results are in accord with observations indicating that bladders decorated on their cytoplasmic surface with GFP-drainin were clearly distinguishable from endosomes filled with the fluid-phase marker TRITC-dextran (Becker et al., 1999). In the present study we show that the entire network of ducts and vesicles of the CV system is separated from endosomes, and that the same is true for the disassembled vesicles in mitotic cells. Neither did dajumin-GFP detectably decorate endosomes at any stage of their pathway, nor did

TRITC-dextran enter the ducts or bladders of the CV network. We conclude that under normal conditions the CV compartment is closed to vesicles of the endosomal pathway, prohibiting lumenal transit from endosomes to the CV network.

The sequence of the C-terminal region of dajumin, responsible for targeting to the CV system, reveals no obvious sequence relationships to other proteins in the database. However, another Dictyostelium protein, drainin, which is specifically associated with the contractile vacuole and involved in its discharge, is the prototype of a protein family represented in Caenorhabditis elegans and man (Becker et al., 1999). The question is whether the drainin homologues of higher eukaryotes serve similar functions as drainin in Dictyostelium, and whether they are localized to a compartment as distinct from endosomes as the CV complex. A compartment of apparent similarity is the sub-plasmalemmal tubulocisternal system characterized in neuroendocrine cells by Schmidt et al. (1997). This membrane system is connected with the cell surface by narrow channels and gives rise to synaptic-like microvesicles, the counterpart of neuronal synaptic vesicles. It will be intriguing to find out whether the CV complex, previously considered to be a peculiarity of protozoa living under conditions of low osmolarity, is indeed a specialized version of a compartment in eukaryotic cells, connected to the plasma membrane and functioning in the conversion of ducts into vesicles.

We thank John Heuser for intense discussions and Gerard Marriott for recommending the Cy 3.5 membrane labeling procedure to us. The work has been supported by the Deutsche Forschungsgemeinschaft (SFB266/C6) and the Fonds der Chemischen Industrie.