Disruption of Golgi structure and function in mammalian cells expressing a mutant dynamin

The formation of nascent secretory vesicles from the trans-Golgi network (TGN) involves the coordinated participation of multiple enzymes, coat proteins, and cytoskeletal elements. Currently, the mechanisms that liberate nascent budding vesicles from the TGN are undefined. The large GTPase dynamin is an attractive candidate to participate in this function for several reasons. It is a mechanoenzyme that has been shown to bind and hydrolyze nucleotide while undergoing a conformational change to distort and sever membrane tubules (Sweitzer and Hinshaw, 1998) and has been referred to as a ‘molecular pinchase’ (McNiven, 1998; McNiven et al., 2000). Further, a role for dynamin in the liberation of both clathrin-coated pits and caveolae from the plasma membrane during endocytosis has been reported by multiple laboratories (Henley et al., 1998; McNiven et al., 2000).

The plasma membrane and Golgi apparatus utilize many of the same protein families to produce nascent vesicles during endocytosis and secretion. It seems likely that dynamin could function at both cellular locations. In support of this premise, several reports have implicated dynamin in the formation of secretory vesicles from the TGN (Henley and McNiven, 1996; Jones et al., 1998; Maier et al., 1996). These initial findings included the immunofluorescense and immunoelectron microscopic localization of dynamin to the Golgi apparatus using multiple dynamin antibodies (Henley and McNiven, 1996; Maier et al., 1996) and the immunoisolation of rat liver Golgi using dynamin antibody-coated beads (Henley and McNiven, 1996). In a recent study, a specific form of dynamin 2 coupled to green fluorescent protein, [Dyn 2(aa)-GFP] was observed to target to clathrin-coated pits at both the plasma membrane and the Golgi in cultured cells (Cao et al., 1998). Most recently, dynamin function has been shown to be required for the formation of TGN-derived secretory vesicles in vitro using a well characterized, cell-free assay (Jones et al., 1998).

The combined observations described above strongly suggest that Dyn 2 associates with the Golgi apparatus in mammalian cells where it may release newly forming secretory vesicles from the TGN. In this present study we have closely monitored the morphology and function of the Golgi apparatus in cultured cells expressing a mutant dynamin protein. Specifically, we have constructed a point mutation in the first GTP-binding element (K44A) of Dyn 2(aa), the dynamin form that is normally expressed in epithelial cells and has been shown to associate with the Golgi apparatus (Cao et al., 1998). We find that epithelial cells expressing this mutant dynamin protein display either compacted, vesiculated, or tubulated Golgi structures that are defective in the transport of GFP-tagged vesicular stomatitus viral glycoprotein (VSVG) to the plasma membrane.

These findings are in direct support of previous studies implicating dynamin in Golgi function (Henley and McNiven, 1996; Jones et al., 1998; Maier et al., 1996) and provide the first evidence for dynamin function at the TGN in living cells. Futher, these findings are consistent with the previously proposed concept suggesting that dynamin participates in distinct vesicle budding events in both the endocytic and secretory pathways (Cao et al., 1998; Urrutia et al., 1997; McNiven et al., 2000).

MiniPrep Express™ Matrix and Luria-Bertani medium were from BIO 101, Inc. (Vista, CA). Restriction enzymes were from Promega (Madison, WI) and Life Technologies, Inc. (Gaitherburg, MD). 1 kb Plus DNA Ladder was from Gibco BRL (Grand Island, NY). An anti-clathrin monoclonal antibody (X22) was collected from the supernatant of the X22 hybridoma cell line (ATCC, Rockville, MD). Monoclonal anti-rat-mannosidase II (Mann II) was purchased from BabCO (Richmond, CA). The following antibodies were obtained as gifts: Golgi-specific monoclonal anti-TGN38 (Dr K. E. Howell, University of Colorado School of Medicine, Denver, CO), HUDY 1 monoclonal anti-dynamin (Dr S. L. Schmid, The Scripps Research Institute, La Jolla, CA). The Dyn 2 and Pan-MC63 antibodies have been previously described (Henley et al., 1998; Henley and McNiven, 1996). A pAb to the rat TGN38 was made to a peptide having the sequence ALEGKRSKVTRRPKASDYQRLNLK, which represents amino acids from 316 to 339 C-terminal of TGN38. All other chemicals and reagents unless otherwise stated were from Sigma (St Louis, MO).

Dyn 2(aa)wt was generated as described previously (Cao et al., 1998). The point mutant Dyn 2(aa)K44A was generated using PCR-based mutagenesis methods. The TGN38 open reading frame (ORF) was amplified by RT-PCR using rat brain cDNA based on the TGN38 sequence from GenBank (accession number X53565). The Dyn 2(aa)K44A PCR products and TGN38 ORF were TA-ligated with the pCR3.1 vector (Invitrogen, Carlsbad, CA) then sub-cloned into the pEGFP-N¹ vector (Clontech, Palo Alto, CA). Vector containing the insert encoding for VSVG was a kind gift from Dr Jennifer Lippincott-Schwartz (NIH, Bethesda, MD). This insert was sub-cloned into the pEGFP-N¹ vector (Clontech, Palo Alto, CA). All DNA constructs were verified by restriction enzyme and sequence analysis (The Mayo Molecular Biology Core (ABI PRISM 377 DNA sequencer, Perkin Elmer-Cetus)). Sequences were analyzed using DNA* analysis software (DNA star, Madison, WI).

BHK-21 cells, (ATCC CCL-10, Rockville, MD) were maintained in EMEM (minimum essential medium Eagle) containing 2 mM L-glutamine and Eagle’s BSS supplemented with 10% fetal bovine serum (Gibco BRL), 100 U/ml penicillin, and 100 μg/ml streptomycin. Clone 9 cells, (ATCC CRL-1439, Rockville, MD) were maintained in Ham’s F-12K medium supplemented with 10% fetal bovine serum (Gibco BRL) and antibiotics. Cells were cultured in T-75 flasks (Fisher Scientific, Pittsburgh, PA) and on 22-mm coverslips for transfections and immunocytochemistry, respectively. All cells were kept in a 5% CO²/95% air incubator at 37°C unless otherwise stated. Plasmid DNA containing dynamin inserts were purified by equilibrium centrifugation in CsCl-ethidium bromide gradients (Sambrook et al., 1989). Cells were transiently transfected using the Lipofectamine Plus™ Reagent. (Gibco BRL). Confluent cells were trypsinized (1â trypsin-EDTA, 0.25% trypsin, 1 mM EDTA-4Na, Gibco BRL) and cultured on coverslips overnight. Transfections were carried out the following day according to the manufacturer’s instructions using 1-2 μg of plasmid DNA per transfection.

Cells were transiently co-transfected with VSVG-GFP and either Dyn 2(aa)wt or Dyn 2(aa)K44A using the Lipofectamine Plus™ reagent (Gibco BRL). After transfection, cells were cultured in 5% CO₂/95% air at 40°C for 16 hours. Thirty minutes prior to a temperature shift from 40°C to 32°C, 100 μg/ml cycloheximide was added to each dish (see time line below) following the methods of Simons and co-workers (Toomre et al., 1999). Cells were subsequently fixed at 15 minutes, 60 minutes, or 120 minutes after being shifted to 32°C, then processed for fluorescence microscopy (Cao et al., 1998; Henley and McNiven, 1996) and quantitation.

Images for quantitation were acquired using a cooled CCD camera (Photometrics SenSys, Tuscon, AZ) driven by the image acquisition program Metamorph (Universal Imaging, West Chester, PA). Images of VSVG-GFP in Dyn 2(aa)K44A cells were taken at full resolution (1400â1000) using the same acquisition settings (exposure time = 3 seconds, 12 bit greyscale). Using the fluorescence quantitation feature of Metamorph, images were thresholded to an average pixel value of 90 (arbitrary units), effectively removing the background, and giving an area with pixel values greater than or equal to 90. Over the perinuclear area was laid a circle of defined size (22109 pixels) and an average pixel intensity value was calculated for the thresholded pixels within the circle. This value multiplied by the thresholded area gave a numerical value for the image, which was directly related to the fluorescence intensity of the VSVG-GFP. We then calculated the fluorescence intensity of 18 and 28 cells at each time point, for both the wild-type dynamin cells, and the Dyn 2(aa)K44A cells. This fluorescence intensity data was then moved to Excel 98 (Microsoft Corp., Seattle, WA) where average intensities were calculated for each data set, and graphs were generated from the average intensity values.

For studies using electron microscopy, all cells were transfected by microinjection of plasmid DNA. This approach insured that most, if not all, cells examined within a given area (1 mm²) of a gridded coverslip were transfected. Injected DNA was diluted to 50 ng/μl in microinjection buffer (10 mM KH₂PO₄, pH 7.2, and 75 mM KCl) and Texas Red-dextran (400 μM) was included to serve as an indicator for successful microinjections. Uninjected or damaged cells were removed from the gridded area with the injection pipette. Cells were pressure injected at 37°C and prepared for electron microscopy as previously described (Henley et al., 1998).

To further emphasize the localization of dynamin with the Golgi using morphological methods, we stained a cultured hepatocyte cell line (Clone 9) with three distinct antibodies to dynamin, or transfected these cells with Dyn 2(aa)-GFP (Fig. 1). The dynamin antibodies utilized here include a polyclonal peptide antibody to a highly conserved region adjacent to the GTP-binding domain (MC63; Henley and McNiven, 1996), a polyclonal peptide antibody to the tail domain specific for Dyn 2 (Henley et al., 1998), and a widely used monoclonal antibody to the dynamin tail (HUDY 1; Damke et al., 1994; Warnock et al., 1995). Clone 9 cells were double labeled with antibodies to dynamin and the trans-Golgi network (TGN38). As demonstrated in Fig. 1a-c, all three antibodies produced a punctate labeling at both the plasma membrane and the Golgi apparatus, as confirmed by double staining with the TGN38 antibody (Fig. 1a′-c′). To support these immunofluorescence observations, cells were also transfected with the plasmid encoding Dyn 2(aa)-GFP. As previously published (Cao et al., 1998), the distribution of Dyn 2(aa)-GFP appeared very similar to that of endogenous dynamin. The tagged Dyn 2(aa) displayed a prominent localization at both the plasma membrane and the Golgi (Fig. 1d,d′). These observations, obtained using distinct dynamin reagents generated by us and others, demonstrate a consistent association of dynamin with the TGN and support the functional studies in living cells described in the following section.

Based on the ability of dynamin to tubulate and constrict lipids both in vivo and in vitro (Stowell et al., 1999; Sweitzer and Hinshaw, 1998; Takei et al., 1995) it is likely that disrupting the function of a Golgi-associated dynamin could have marked effects on the morphology of this organelle. To test this prediction we expressed a mutant Dyn 2 construct in cultured Clone 9 cells expressing GFP-tagged trans-Golgi Network 38 (TGN38-GFP), a widely used marker for the TGN (Banting and Ponnambalam, 1997). The mutant dynamin protein utilized here was a specific spliced variant of Dyn 2 normally expressed in epithelial cells and previously shown to associate with clathrin-coated pits at the plasma membrane and the Golgi (Fig. 1d,d′; Cao et al., 1998). A standard point mutation (K44A) in the first GTP-binding element of Dyn 2(aa) was selected based on previous studies showing a potent inhibition of clathrin-mediated endocytosis in cells expressing this altered protein (Damke et al., 1994; Herskovits et al., 1993; van der Bliek et al., 1993). Cultured Clone 9 cells were co-transfected with TGN38-GFP and Dyn 2(aa)K44A constructs and allowed to recover for 18 hours. Subsequently, cells were fixed, permeabilized, stained with a Pan-dynamin antibody (MC63) and observed by fluorescence microscopy. Cells transfected with the Dyn 2(aa)K44A construct showed a large increase in the intensity of dynamin staining compared to untransfected cells. Control cells expressing TGN38-GFP and endogenous Dyn 2 showed a prominent localization of the TGN38-GFP protein to the Golgi. The distribution and morphology of the Golgi in these cells (Fig. 2a) mimicked that of untransfected cells stained with TGN38 antibodies (not shown). In contrast, cells expressing the mutant Dyn 2(aa)K44A displayed three different types of altered Golgi morphologies. These Golgi phenotypes included a compacted perinuclear Golgi mass (Fig. 2b), an enlarged vesiculated Golgi cluster comprised of numerous budded Golgi cisternae (Fig. 2c), and a highly tubulated Golgi in which numerous Golgi tubules extended outward about the nucleus, seemingly along cytoskeletal filaments (Fig. 2d,e).

To more closely examine the morphology of those altered Golgi structures in cells expressing mutant Dyn 2(aa)K44A, electron microscopy was performed. As shown in Fig. 3, control Clone 9 cells displayed normal Golgi structures with associated vesicle buds that were distributed sparsely about the perinuclear region. The number of Golgi stacks and buds visualized were modest in number but typical for this cell type. In marked contrast, cells expressing Dyn 2(aa)K44A possessed many closely situated, highly tubulated Golgi stacks with an extraordinary number of associated vesicle buds. Higher magnification images of these altered Golgi clusters (Fig. 3c-f) clearly showed the relationship between the Golgi cisternae, tubules, and numerous vesicles. These accumulated vesicle buds were particularly striking, not only due to their excessive numbers, but also because they had extended necks. Further, most of these buds were delineated by a dark, electron dense coat that did not appear to be clathrin, although some spiked coats could be seen. These electron micrographs are remarkably consistent with the fluorescence images in Fig. 2 displaying compacted, vesiculated, or tubulated Golgi morphologies in the dynamin mutant cells expressing TGN38-GFP.

To extend the qualitative observations described above (Figs 2, 3), we tested if secretion of nascent protein from mutant cells was impaired. Specifically, we tested if transport of GFP-tagged VSVG from the Golgi to the cell surface was delayed or reduced. VSVG has proven to be a useful tool to study the secretory pathway in mammalian cells and has been utilized previously in many classic studies. For review see (Bergmann, 1989). Further, a temperature sensitive mutation in this protein (VSVG-ts045-GFP) has been shown to prevent its transport out of the ER to the Golgi at the restrictive temperature (40°C) thereby providing a pulse of nascent protein that can be tracked with the fluorescence microscope (Cole et al., 1998; Presley et al., 1997; Toomre et al., 1999). We have followed the fate of VSVG-ts045-GFP in another normal non-transformed cultured epithelial cell line (BHK) transfected with the Dyn 2(aa)K44A mutant construct. BHK cells were utilized for this portion of the study because they are easily co-transfected while, in our hands, Clone 9 cells did not readily express VSVG-ts045-GFP.

BHK cells were co-transfected with VSVG-ts045-GFP and either Dyn 2(aa)K44A or Dyn 2(aa)wt constructs as described and illustrated in Materials and Methods. By combining cycloheximide treatment with a temperature shift a defined pulse of VSVG-ts045-GFP could be followed as it was transported to and then from the Golgi (Toomre et al., 1999). Thus, control cells were expected to transport VSVG out of the Golgi to the plasma membrane within 30-60 minutes while mutant cells might accumulate VSVG-ts045-GFP in a specific secretory compartment. This method has been utilized by others to study vesicle-based protein trafficking in the secretory pathway (Cole et al., 1998; Hirschberg et al., 1998; Presley et al., 1997; Toomre et al., 1999). As demonstrated in Fig. 4, the cytoplasmic distribution of VSVG-ts045-GFP at the restrictive temperature is generally diffuse with some perinuclear localization due to cell thickness, a large number of ER cisternae extending outward from the area, and some potential leakage of nascent protein from the ER to the Golgi compartment. This distribution is similar to that observed by others (Cole et al., 1998; Hirschberg et al., 1998; Presley et al., 1997; Toomre et al., 1999). Cells expressing Dyn 2(aa)wt (a-d) transported VSVG-ts045-GFP out of the ER to the Golgi in 15-30 minutes. By 60 minutes most, if not all, of the VSVG-ts045-GFP was transported out of the Golgi and inserted in the plasma membrane. In sharp contrast, cells expressing Dyn 2(aa)K44A transported the viral protein to the Golgi with normal kinetics although transport out of the Golgi was markedly reduced (Fig. 4e-h). Indeed, by 60 minutes at the permissive temperature these cells showed substantial accumulations of VSVG-ts045-GFP in Golgi-like structures at the perinuclear area. Further, this accumulation was not reduced even after 120 minutes at the permissive temperature. Fluorescence quantitation of over 350 cells showed a consistent and significant retention of nascent viral protein in the perinuclear region of cells expressing mutant dynamin as compared to wild-type dynamin (Table 1).

To confirm that the perinuclear accumulation of nascent viral protein in the mutant cells represented a retention in the Golgi, mutant cells expressing VSVG-ts045-GFP were fixed and stained with antibodies to the Golgi resident protein Mann II. As shown in Fig. 5, cells expressing wild-type Dyn 2(aa) accumulate VSVG-ts045-GFP in a Mann II positive compartment prior to transport out of the cell. Mutant cells incubated at the permissive temperature for 15-30 minutes also showed a near exact co-localization between the perinuclear VSVG-ts045-GFP and Mann II. Interestingly, as the time of incubation at the permissive temperature was increased to 60-120 minutes, the amount of viral protein accumulated at the Golgi region in the mutant cells increased well beyond the boundaries of the Mann II localization. When these cells were stained with antibodies to TGN38, we observed a near exact co-localization between nascent VSVG and the TGN38 protein. Little, if any, VSVG-ts045-GFP was observed to co-localize with marker antibodies for the pre-Golgi compartment (ERGIC53, β-COP, and Rab1; data not shown).

In this study we have tested the role of the large GTPase dynamin in regulating the structure and function of the Golgi apparatus in mammalian epithelial cells. Specifically, we have manipulated one form of this protein, Dyn 2(aa), that is expressed in epithelial cells (Cook et al., 1994; Sontag et al., 1994) and has been shown to preferentially associate with the TGN compared to the other three identified spliced variants of Dyn 2 (Cao et al., 1998). First, we have utilized three distinct poly- or monoclonal dynamin antibodies and GFP-tagged Dyn 2(aa) to demonstrate an association of Dyn 2 with the TGN (Fig. 1). Second, we have expressed, in the same cultured epithelial cells, a mutant form of Dyn 2(aa) that induces profound alterations in Golgi structure. Fluorescence microscopy of mutant cells expressing a GFP-tagged trans-Golgi network protein (TGN38-GFP) or electron microscopy of mutant cells showed consistent alterations in Golgi structure. These changes included a hypertrophied cluster of multiple perinuclear Golgi stacks exhibiting various levels of compaction, extention, tubulation, and vesiculation (Figs 2, 3). Consistent with these morphological changes is a substantial disruption of Golgi function in mutant cells as displayed in Fig. 4. In contrast to cells expressing Dyn 2(aa)wt where VSVG-GFP reached the plasma membrane, cells expressing Dyn 2(aa)K44A showed a marked perinuclear accumulation of VSVG-ts045-GFP (Fig. 4). Double labeling of cells with Golgi marker antibodies (Fig. 5) revealed that this protein accumulation occurred within a Mann II positive compartment at early time periods after removal of the temperature block (15-30 minutes) while filling and distending the TGN at later times (60-120 minutes).

The association of dynamin with the Golgi apparatus in cultured epithelial cells described here supports and extends similar observations made by us (Cao et al., 1998; Henley and McNiven, 1996) and others (Maier et al., 1996). Henley and McNiven (1996) have demonstrated that two different dynamin antibodies label the Golgi and are capable of immunoisolating intact Golgi stacks from liver cell homogenates. To avoid the use of antibodies, Cao et al. (1998) expressed different spliced variants of Dyn 2 tagged with GFP in cultured cells and observed a striking association of Dyn 2(aa) with the TGN. Interestingly, another form of Dyn 2, differing only by a four amino acid deletion, showed a very modest Golgi association. Finally, Maier et al. (1996) demonstrated significant labeling of the TGN at the ultrastructural level using immunogold electron microscopy. Despite this consistent series of observations using distinct methodologies, Altschuler et al. (1998) have not observed a Golgi-dynamin association by either antibody cell staining and/or expression of tagged dynamin in cultured cells. While it is not surprising that distinct laboratories using different reagents may obtain contrasting results, we believe that technical approaches used could also explain the negative results obtained by this group. In addition, different dynamin antibodies were used in these studies. To address this difference, we obtained one of these reagents, HUDY 1, which is a well characterized and widely used monoclonal antibody to the dynamin tail domain that provides striking images of punctate, clathrin-like localization in cultured cells (Damke et al., 1994; Warnock et al., 1995). In our hands, this antibody provided the best and most consistent Golgi labeling (Fig. 1c,c′) of all reagents that we have tested regardless of cell type used.

Consistent with the concept of dynamin residing at the Golgi, we observed gross morphological changes in this organelle in mutant cells expressing TGN38-GFP. These structural alterations can be grouped into three distinct phenotypes and include a tightly compacted, highly vesiculated, or extensively tubulated cluster of Golgi stacks (Fig. 2). TGN38-GFP was utilized as a Golgi marker based on observations by Banting et al. (1998) and colleagues indicating that the steady state of this protein is not significantly altered in dynamin mutant cells defective in clathrin-mediated endocytosis. Thus, it is likely that the changes we have observed in Golgi structure are a direct consequence of defective dynamin action at the Golgi as opposed to perturbing TGN38 trafficking from the plasma membrane. Indeed, the ultrastructural morphology of the Golgi in Dyn 2 mutant cells (Fig. 3) is consistent with a predicted block in vesicle formation at the Golgi that, in turn, leads to the formation of highly tubulated and vesiculated cisternae. These vesicle buds have elongated necks and are densely coated either with or without clathrin. This observation correlates with our previous study (Jones et al., 1998) demonstrating that dynamin participates in the formation of both clathrin and non-clathrin-coated secretory vesicles from rat liver Golgi in vitro. Further, these structural alterations are also consistent with the accumulation of clathrin-coated and uncoated membrane invaginations observed at the plasma membrane in cells of the shibire^ts flies that express a temperature sensitive dynamin mutant (Koenig and Ikeda, 1989; Kosaka and Ikeda, 1983).

The marked 8- to 11-fold accumulation of nascent GFP-VSVG-ts045 in the Golgi (Figs 4, 5, and Table 1) of cells expressing Dyn 2(aa)K44A suggests that this mutant dynamin induces a block in the exit of proteins from the TGN. This functional impairment is consistent with the morphological defects observed by fluorescence and electron microscopies described in Figs 2 and 3. Interestingly, mutant cells that are fixed and examined at early times (15-30 minutes) after being shifted to the permissive temperature (32°C) accumulate VSVG-ts045-GFP almost exclusively in a Mann II positive compartment (Fig. 5). Cells examined after longer times at the permissive temperature (60-120 minutes) generally display much larger perinuclear accumulations of nascent protein that exceed the borders of the Mann II compartment and appear to engorge and distend the trans-Golgi compartment. This hypertrophy of the TGN is particularly graphic when compared to adjacent untransfected cells. We had predicted that much of the nascent protein may accumulate and ‘spill back’ into a pre-Golgi compartment, however, staining of these cells with marker antibodies to this compartment (β COP, ERIC38, and RabI) showed little co-localization (data not shown).

The observations described above are consistent with a similar study conducted by another group (Kreitzer and Rodriguez-Boulan, 2000) that has tested the role of both Dyn 2 and kinesin function in the exit of a GFP-tagged membrane protein (neurotrophin receptor (GFP-p75)) from the Golgi. Cells expressing mutants in either of these mechanoenzymes showed enlarged Golgi structures filled with nascent GFP-p75. This study was published during the review of our manuscript, however, our study has extended the observations of this group in several ways. First, the above study examined the effects of mutant dynamin on the exit of a an apically targeted protein (p75) from the TGN following release from a 20°C block while we have followed the transit of a basolateral viral protein released from the ER. Second, we have provided graphic, high resolution electron microscopic images of tubulated and vesiculated Golgi cisternae in Dyn 2 mutant cells. Third, detailed computer-aided quantitation of VSVG-ts045-GFP accumulation in the Golgi region has been provided. Fourth, we have tracked the transit of nascent protein through an early Golgi, mannosidase positive compartment to show accumulation in the TGN proper. Despite the wealth of observations supporting a role for Dyn 2 in Golgi function in both of these studies, a recent report (Altschuler et al., 1998) has suggested that cells expressing a mutant Dyn 2 or Dyn 1 exhibit normal trafficking of nascent proteins from the Golgi to either the cell surface (transferrin receptor (TfnR), polymeric IgA receptor (pIgAR)) or the lysosome (cathepsin D). We believe that these negative findings should be interpreted with caution for several reasons. First, experiments following the pIgAR were conducted in MDCK cells in which mutant Dyn 2 had only a limited inhibitory effect (40-80%) on classic receptor mediated endocytosis. Second, transport of TfnR to the cell surface or cathepsin D to the lysosome was indeed retarded or reduced by 20-30% in Dyn 2 mutant cells. This reduction was downplayed by the authors as being insignificant. Third, this study did not conduct a block or pre-pulse of nascent protein into a secretory compartment as is done in many studies measuring protein transit out of the Golgi. Fourth, and perhaps most importantly, this study utilized biochemical methods to measure the insertion of nascent receptor proteins into the plasma membrane and did not follow protein trafficking directly by microscopic methods to look for retention in the Golgi. In our view, it would not be surprising if non-polarized or neoplastic cells cultured in vitro, when induced to accumulate large quantities of nascent protein in the Golgi, might traffic proteins to the lysosome or surface by alternative mechanisms. Indeed, an interesting study by Stoorvogel et al. (1996) convincingly demonstrated a differential effect of the fungal metabolite brefeldin A (BFA) on exit of nascent proteins out of the Golgi. Although BFA was observed to potently inhibit the transport of α1-antitrypsin from the Golgi to the plasma membrane, transport and processing of cathepsin D to the lysosome was unperturbed. The authors suggest that a ‘continuous network’ of the TGN and endosomes may function in targeting cathepsin to the lysosomal compartment. Based on these findings, a tubulated TGN, induced by an inhibition of dynamin function, may continue to support Golgi to lysosome transport. The modest but significant reduction of cathepsin D processing observed by Altschuler et al. (1998) in the mutant dynamin expressing cells is consistent with this premise. In another study, Balch and colleagues (Nishimura et al., 1999) have shown that nascent VSVG-ts045-GFP lacking a DXE motif is not concentrated in COPII vesicles and subsequently accumulated in the ER. Surprisingly, despite this accumulation, equal levels of wild-type (DXE) or mutant (AXA) viral protein were found to reach the cell surface after a two hour release from the restrictive temperature. These findings are again consistent with the concept that cultured cells may continue to transport substantial levels of nascent secretory proteins despite the fact that this pathway has been impaired. Such alternative transport pathways underline the importance of morphological viewing of secretion in manipulated cells.

Our findings reported here and those of others implicate the action of Dyn 2 at the TGN and are consistent with the prediction that the dynamin family participates in a significant number of vesicle budding events within cells. These distinct processes include non-clathrin mediated endocytosis (Henley et al., 1998), phagocytosis (Gold et al., 1999), and endosome to Golgi transport (Llorente et al., 1998). It will now prove interesting and worthwhile to define proteins that target dynamin to the TGN while identifying specific dynamin-mediated Golgi processes in both ortho- and retrograde trafficking pathways.

Special thanks to Dr Sophie Dahan for her help with the fluorescence quantitation of VSVG-ts045-GFP and to Dr Yisang Yoon for help in plasmid construction of VSVG-ts045-GFP. This study was supported by NIH grant DK44650 to M.A.M.