Differential Inhibition of Multiple Vesicular Transport Steps between

THE JOURNAL OF BIOLOGICAL CHEMISTRY 0 1993 by The American Society for Biochemistry and Molecular Biology, Inc.

Val. 268, No. 6, Issue of February 25, pp. 4216-4226,1933 Printed in U. S A .

Differential Inhibition of Multiple Vesicular Transport Steps between the Endoplasmic Reticulum and Trans Golgi Network*

(Received for publication, September 21, 1992)

Howard W. Davidsonz and William E. Balch From the Departments of Cell and Molecular Biology, The Scripps Research Znstitute, La Jolla, California 92037

Using the glycoprotein of the ts045 mutant of vesicular stomatitis virus (VSV-G) as a marker, we have developed a system capable of measuring vesicular transport from the endoplasmic reticulum (ER) to the trans Golgi network (TGN) in vitro. Movement from the ER to the cis Golgi compartment was assessed by the conversion of VSV-G from a totally endoglycosidase D (endo D)-resistant form to a species containing one endo D-resistant and one endo D-sensitive oligosaccharide (GD1). Similarly, delivery to the medial cisternae was measured by the appearance of the completely endo D-sensitive form of VSV-G (GDz) or by the acquisition of complete resistance to endoglycosidase H (endo H) (GH~) and delivery to the TGN by the appearance of an endo H-resistant form of VSV-G which was sensitive to digestion with neuraminidase and subsequently @-galactosidase (GH,).

Movement between each sequential compartment required ATP and soluble proteins (cytosol) and was inhibited by nonhydrolyzable analogues of GTP and by an antibody toward the N-ethylmaleimide-sensitive factor NSF. In contrast, fractionation of the cytosol by ammonium sulfate precipitation indicated that distinct proteins were required for movement between succes- sive compartments. Similarly, inclusion of a mutant form of the small molecular weight GTP-binding protein rablA inhibited movement between the ER and cis Golgi, and between the cis and medial cisternae, but did not affect transport from the medial Golgi to the TGN. Conversely, the protein kinase inhibitor staurosporine prevented movement between the medial Golgi and the TGN but did not influence transport between the ER and early Golgi compartments. This study pro- vides the first demonstration that vesicular transport between the ER and TGN can be reconstituted in a cytosol-dependent fashion in vitro, allowing a direct analysis of the roles of individual components in multiple transport events.

The eukaryotic secretory apparatus is composed of a series of discrete membrane-limited compartments (reviewed in Pa- lade, 1975; Rothman and Orci, 1992; Mellman and Simons, 1992). Movement of proteins between sequential elements is generally vectorial and is mediated by transport vesicles. However, the biochemical machinery responsible for vesicular

GM33301, GM42336, and CA27489 (to W. E. B.). This is manuscript * This work was supported by National Institutes of Health Grants

No. 7637-CB from The Scripps Research Institute. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore he hereby marked “aduer- tisement” in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

$Supported by a Science and Engineering Research Council of Great Britain/NATO research fellowship and by the G. Harold and Leila Y. Mathers Charitable Foundation.

transport remains to be elucidated. Progress toward the char- acterization of components involved in this process has been made following two advances: the isolation of a series of secretory mutants from yeast (Novick et al., 1980) and the development of systems from a variety of eukaryotic cells which reconstitute protein trafficking in vitro (reviewed in Balch, 1989; Rothman and Orci, 1992; Pryer et al., 1992). For example, genetic approaches have suggested the involvement of various SEC gene products and several members of the ras superfamily of 20-25-kDa GTP-binding proteins (including the YPT/SEC4/rab, SAR, and ARF families) (reviewed in Balch, 1990; Goud and McCaffrey, 1991; Pryer et al., 1992). Similarly, in vitro reconstitution studies have also provided evidence for the involvement of GTP-binding proteins, based on the effects of the inhibitor GTP@ (e.g. Melancon et at., 1987; Baker et at., 1988; Ruohola et al., 1988; Schwaninger et al., 1992) and have identified the N-ethylmaleimide-sensitive factor NSF (Block et al., 1988) and soluble NSF attachment proteins (SNAPS; Clary and Rothman, 1990). Together these and related studies have indicated that there is a remarkable degree of conservation between the exocytic machineries of widely divergent species. For example, the mouse yptl (rabl) gene product can rescue a yeast YPTl mutant (Haubruck et al., 1989), whereas the SEC18 protein isolated from Saccha- romyces cerevisiae can substitute for NSF in a cell-free assay using Chinese hamster ovary (CHO) cells (Wilson etal., 1989).

To date most of the proteins which have been isolated appear to be involved in multiple vesicular transport events. Thus, analysis of the effects of a monoclonal antibody which inhibits NSF function suggest that this protein is involved both in transport from the ER to the cis Golgi (Beckers et al., 1989) and between Golgi cisternae (Wilson et ai., 1989). Similarly, a-SNAP, originally identified as an essential component of transport between the cis and medial Golgi (Clary and Rothman, 1990), is homologous to SEC17, a protein required for movement between the ER and Golgi in yeast (Griff et al., 1992). However, the discontinuous nature of the exocytic pathway suggests that some transport components should only be required a t distinct steps, both to maintain the fidelity of the various compartments and to ensure vectorial flow. Evidence to support this hypothesis has been provided by the different phenotypes in temperature shift experiments of yeast carrying thermosensitive mutations in the SEC18 and SEC23 genes (Graham and Emr, 1991). However, not all thermosensitive mutant proteins are rapidly inactivated upon shift to the nonpermissive temperature, which prevents analysis of many of the early SEC genes by this protocol.

‘The abbreviations used are: GTP-yS, guanosine 5’-3-0- (thi0)triphosphate; BHK, baby hamster kidney; CHO, Chinese hamster ovary; NRK, normal rat kidney; ER, endoplasmic reticulum; endo D, endoglycosidase D; endo H, endoglycosidase H; PAGE, polyacrylamide gel electrophoresis; VSV, vesicular stomatitis virus; VSV-G, vesicular stomatitis virus glycoprotein; Sia, sialic acid TGN, trans Golgi network.

4216

ER to TGN Transport in Vitro 4217

To determine whether a transport component is required at a particular step by analysis in vitro generally requires a comparison between several different assay systems, each of which focuses upon a single step (e.g. Balch et al., 1984; Rothman, 1987; Beckers e t al., 1987). To overcome this re- striction we have developed an assay which is capable of measuring movement of secretory proteins from the ER through sequential Golgi cisternae to the trans Golgi/trans Golgi network (TGN). Using this system we confirm that NSF is required for each transport step between the ER and TGN. We further demonstrate that GTPyS-sensitive proteins are also involved a t each step but show that a specific GTP- binding protein, RablA, is only required for movement between the ER and medial Golgi.

EXPERIMENTAL PROCEDURES

Materials-Staurosporine, endoglycosidase D (endo D), and Vibrio cholera neuraminidase were each obtained from Calbiochem; GTPTS, Diplococcuspneumoniae 8-N-acetylglucosaminidase, bovine testicular 8-galactosidase, and endoglycosidase H (endo H) were from Boehrin- ger Mannheim; TranRsS-label was from ICN Biomedicals Inc, Costa Mesa, CA. Hybridoma clone 4A6 was kindly provided by J. E. Roth- man (Sloan-Kettering Institute, New York), and anti-NSF IgM was purified by published procedures (Block et al., 1988). RablA proteins were expressed in Escherichia coli and purified from bacterial lysates by procedures described elsewhere.' All other reagents except where indicated were obtained from Sigma.

Assay of ER-to-Golgi Transport in Vitro-Normal rat kidney (NRK) cells maintained as described previously (Davidson et al., 1992) were infected a t a multiplicity of 10-20 plaque-forming units with the ts045 strain of vesicular stomatitis virus (VSV) (Flamand, 1970; Beckers et al., 1987). Four hours post-infection the cells were transferred to the restrictive temperature (40 "C), radiolabeled with TranRsS-label for 10 min (100 pCi/lO-cm dish), and perforated by the swelling and scraping procedure (Beckers et al., 1987). Washed cells were resuspended in 50 mM Hepes-KOH, pH 7.2, 90 mM potassium acetate (approximately 120 pl/lO-cm dish) and 5-pl aliquots incubated a t 30 "C in assay buffer (Davidson et al., 1992) supplemented with sugar nucleotides as indicated. At the conclusion of the assay, cell membranes were collected by centrifugation, digested with exo- and/ or endoglycosidases as described below, and analyzed by sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) and. autoradiography (Beckers et al., 1987). Autoradiographs were quantitated by scanning densitometry (Hoeffer GS300 transmission scanning densitometer, Hoeffer Scientific Instruments, San Francisco).

Digestion of Membrane Glycoproteins with Endo- and Eroglycosi- dases-For digestion with endo H, cell pellets were resuspended in 20 p1 of 100 mM sodium acetate, pH 5.6, containing 0.3% SDS and 0.3 M 2-mercaptoethanol and boiled for 5 min. The samples were then diluted with 40 p1 of 100 mM sodium acetate, pH 5.6, and allowed to cool to room temperature. After addition of 3 milliunits of endo H, they were incubated a t 37 "C for 16 h, at which time the reactions were terminated by the addition of concentrated SDS-PAGE sample buffer (Beckers et al., 1987). To those samples to be digested with neuraminidase, 5 milliunits of this enzyme was added concurrent with the endo H. Those samples to be digested with @-galactosidase were first digested with endo H in the presence or absence of neuraminidase as appropriate. Triton X-100 (Surfact-Amps'", Pierce Chemical Co.) was then added to a final concentration of 1%. followed by 5 milliunits of &galactosidase. The resulting solutions were incubated for a further 24 h a t 37 "C and then terminated as described above.

For digestion with endo D, cell pellets were resuspended in 50 pl of 0.1 M sodium citrate/sodium phosphate, pH 6.0, containing 0.2% Triton X-100, 5 mM EDTA, 100 pg/ml leupeptin, 0.3 milliunit of endo D and 2 milliunits of @-N-acetylglucosaminidase, incubated a t 37 "C for 16 h, and terminated by the addition of SDS-PAGE sample buffer. Where indicated, the digestion mixture also contained 5 milliunits of @-galactosidase.

Ammonium Sulfate Fractionation of Rat Liver Cytosol-Rat liver cytosol (10 ml) prepared as described previously (Davidson et al., 1992) was made 30% saturated in ammonium sulfate by the stepwise addition of solid a t 4 "C with constant mixing. The insoluble material was collected (10 min a t 10,000 rpm, Beckman JA 20 rotor; Beckman

S. Pind, C. Nuoffer, T. Compton, H. W. Davidson, and W. E. Balch, manuscript in preparation.

Instruments Inc.), resuspended in 1 ml of 25 mM Hepes-KOH, pH 7.2,125 mM potassium acetate, 1 mM sodium mercaptoethanesulfonic acid (buffer A) containing 100 p~ GDP and 100 p~ magnesium acetate (final volume 1.5 ml), and dialyzed extensively against buffer A. The supernatant was made 80% saturated in ammonium sulfate, collected, resuspended in buffer A, and dialyzed as described above.

The two retentates were each collected, centrifuged (14,000 X g for 10 min a t 4 "C), and aliquots of the supernatants stored frozen a t -80 "C prior to use. Typically the protein concentration of the 0- 30% cut was 2 mg/ml and that of the 30-8076 cut 10 mg/ml.

RESULTS

Terminal Glycosylation of VSV-G in Semi-intact NRK Cells-In mammalian cells most asparagine-linked oligosaccharides are extensively modified during the passage of glycoproteins through the ER and Golgi apparatus (reviewed in Kornfeld and Kornfeld, 1985). These changes include the removal of Glc and Man residues from the common core oligosaccharide and the subsequent addition of various combinations of GlcNAc, Gal, fucose, and sialic acid (Sia) residues. The enzymes catalyzing these reactions are confined to distinct locations within the secretory pathway and show limited substrate specificities, allowing their products to be used as markers of delivery to particular subcellular compartments. In addition the action of individual enzymes renders their products differentially sensitive or resistant to digestion with various combinations of exo- and/or endoglycosidases, providing a method of measuring their activity.

The Indiana serotype of VSV expresses a single glycoprotein (VSV-G), which in the mature form contains two complex (sialylated) N-linked oligosaccharide chains (Etchison et al., 1977). A strain of virus with a temperature-sensitive mutation in this protein (ts045) has been isolated (Flamand, 1970) and used extensively to define the compartmentalization of oligosaccharide processing and as a marker protein in the study of trafficking within the exocytic pathway (e.g. Beckers et al., 1987; for review see Bergmann, 1989). Since the sialyltransferases responsible for terminally glycosylating VSV-G are confined to the TGN (Roth et al., 1985; Chege and Pfeffer, 1990), we decided to use the acquisition of Sia residues by VSV-G as a marker for transport between the ER and TGN in vitro.

As shown in Fig. 1, incubation of perforated NRK cells at 30 "C under conditions shown previously to support vesicular transport in semi-intact CHO cells (Davidson et al., 1992), supplemented with UTP to stabilize endogenous sugar nucleotides, produced an endo H-resistant form of VSV-G which migrated slower on SDS-PAGE than the undigested high mannose form of the glycoprotein (Fig. 1, trucks 1-4; compare tracks 1 and 3 (undigested) with tracks 2 and 4 (digested)). This species (GHt) was sensitive to digestion with neuraminidase (Fig. 1, track 5) and subsequently @-galactosidase (Fig.

1 2 3 4 5 6 7 0 9

Endo H - + - + + + + + + Neuraminidase - - - - + + - - - p-Galactosidase - - - - - + + "

FIG. 1. Terminal glycosylation of VSV-G in vitro. Semi- intact NRK cells were incubated on ice (tracks 1 and 2 ) or a t 30 'C (tracks 3-9) in a transport cocktail containing rat liver cytosol supplemented with UTP (100 p ~ ) (tracks 1-7), UTP and UDP-GlcNAc (500 p ~ ) (track 8) , or UTP, UDP-GlcNAc, and UDP-Gal (1 mM) (track 9) for 150 min. Cell membranes were collected, either mock- digested (tracks I and 3 ) or digested with enzymes as indicated (tracks 2 and 4-9) and analyzed by SDS-PAGE and autoradiography. GH*, G H r , and G H t indicate the endo H-sensitive, endo H-resistant, and terminally glycosylated forms of VSV-G, respectively.

4218 ER to TGN Transport in Vitro

1, track 6; note the slight increase in mobility between tracks 5 and 6 ) but was insensitive to digestion with &galactosidase in the absence of neuraminidase (Fig. 1, track 7). Thus GHt represents the complex form of the protein containing both Gal and Sia residues. Interestingly GHt was not formed when incubations were conducted in the presence of UDP-GlcNAc (500 p ~ ) (Fig. 1, truck 8 ) which instead produced an endo H- resistant species migrating identically to the neuraminidase and P-galactosidase-treated form (GHr; Fig. 1, track 6 ) . This inhibition could be antagonized by also including 1 mM UDP- Gal (Fig. 1, truck 9 ) , suggesting that it results from interactions between the various sugar nucleotide pools, and that the ratios of these substrates may be critical in determining the extent of sialylation observed in vitro.

The G H ~ form of VSV-G was also generated in incubations containing semi-intact cells prepared from several other cell lines, including baby hamster kidney (BHK), NIH 3T3, and HeLa, although not always with the same efficiency as that routinely obtained with NRK cells (data not shown).

Glycosylation in Vitro Can Be Modulated by Manipulating Sugar Nucleotide Pools-Our ability to detect terminally glycosylated VSV-G after incubation in assay mixtures not supplemented with exogenous sugar nucleotides (Fig. 1) must indicate that these substrates are provided either by the cell membranes, cytosol (a 100,000 X g supernatant of rat liver), or a combination of these components. Since sugar nucleotides are not synthesized within the lumen of the Golgi and so must be transported across the Golgi membranes (reviewed in Hirschberg and Snider, 1987), we examined the effect of removing low molecular weight components from the cytosol by gel filtration using Sephadex" G-25.

Incubation in assay mixtures containing gel-filtered ("de- sal ted) cytosol, but lacking additional sugar nucleotides, resulted in the majority of VSV-G remaining in the endo H- sensitive form (GHJ (Fig. 2, track 3) . The remaining 20-30% was present as material migrating between GHr and GHs (prob- ably comprised of a mixture of complete and partially modified VSV-G). Inclusion of UTP (Fig. 2, track 4 ) , but not CTP (data not shown), stimulated the formation of both GHr and GHt, whereas UDP-GlcNAc addition resulted in nearly quan- titative production of GHr (Fig. 2, track 5 ) . Surprisingly, addition of either UDP-Gal or CMP-Sia alone also stimulated the formation of endo H-resistant species (Fig. 2, tracks 6 and 7; compare with truck 3) , suggesting that these compounds may stabilize existing membrane associated pools of UDP- GlcNAc.

In the absence of UTP, formation of GHt required the addition of UDP-Gal and either UDP-GlcNAc or CMP-Sia, (Fig. 2, tracks 8-10) with the most efficient processing being obtained when all three substrates were added (Fig. 2, track 11 ). These results suggest that UDP-Gal is the least stable component. Consistent with this conclusion the formation of

1 2 3 4 5 6 7 8 9 1 0 1 1

- Z I a . , r r " ' = ~ 4-Gm

UDP-GlcNAc - - - - + - - + + - + UDP-Gal - - - - - + - + - + + CMp-Sia - - - - - - + - + + +

FIG. 2. Evidence for cytosolic and membrane-associated sugar nucleotide pools. Semi-intact NRK cells were incubated on ice (track 1 ) or at 30 "C (tracks 2-11) for 150 min in a transport mixture containing untreated (trucks 1 and 2 ) or desalted (tracks 3- 11 ) rat liver cytosol and supplemented with UTP (100 p ~ ) (track 4 ) or UDP-GlcNAc (200 pM), UDP-Gal (500 p ~ ) , and CMP-Sia (200 pM) as indicated. Cell membranes were collected by centrifugation, digested with endo H, and analyzed by SDS-PAGE and autoradiography.

GHt in incubations containing cytosol which had not been gel- filtered could often be stimulated by the addition of this sugar nucleotide (not shown). Optimization experiments demonstrated that maximal conversion to the GHr form required at least 100 p~ UDP-GlcNAc and that the ratio of UDP-Gal to UDP-GlcNAc be at least 2:l to prevent inhibition of GHt formation. Accordingly, to ensure consistency between different membrane preparations, except where indicated, all subsequent experiments were conducted with desalted cytosol and assay mixtures containing 100 pM UDP-GlcNAc, 250 p M UDP-Gal, and 100 p~ CMP-Sia.

Sialylation of VSV-G in Vitro Requires Multiple Vesicular Transport Steps-At the restrictive temperature (40 "C) ts045 VSV-G cannot exit from the ER (Bergmann, 1989). Consequently the protocol used in this study ensures that before the start of the incubation at the permissive temperature (30 "C) the pulse of radiolabeled VSV-G is retained within the ER. The data in Fig. 1 demonstrate that under the incubation conditions used VSV-G apparently reaches compartments enriched in galactosyl- and sialyltransferases; presumably the trans Golgi and TGN (Kornfeld and Kornfeld, 1985; Rothman and Orci, 1992). Moreover, given the substrate specificity of the various sugar transferases, they also indicate that the protein must have been exposed to enzymes located in the cis and medial cisternae. Although the most likely explanation for these observations is that the system we are using reconstitutes multiple transport events, an alternative explanation is that incubation in vitro results in the formation of a single compartment containing all of the required enzymes. In vivo a similar situation can be artificially induced using the drug brefeldin A, which causes the collapse of Golgi membranes into the ER, although the TGN (where sialyltransferase is located) does not appear to behave like the rest of the Golgi complex (Doms et al., 1989; Chege and Pfeffer, 1990). To examine the possibility that VSV-G moved to a single compartment we analyzed the kinetics of the formation of GHt. Movement through the Golgi stack in vivo is rapid and so intermediates between the high mannose (ER form) and sialylated (TGN form) forms of VSV-G do not accumulate. In contrast, transport in uitro is generally 3-4-fold slower (see Beckers et al., 1987). Therefore, we reasoned that if multiple transport events occurred, we should be able to detect the various intermediate forms of VSV-G.

Analysis of the products formed at various times during a typical incubation revealed that this was indeed the case. The initial product formed was G H ~ (Fig. 3, open squares) which could be detected within 20-30 min of shifting to 30 "C. This

1

I% 20 0 0 30 60 90 120 150 180

Time (mins)

FIG. 3. Kinetics of VSV-G transport in vitro. Semi-intact cells were incubated at 30 "C for times as indicated in a transport mixture containing desalted rat liver cytosol and supplemented with UDP-GlcNAc (100 p ~ ) , UDP-Gal(250 p ~ ) , and CMP-Sia (100 pM). Results were quantitated by scanning densitometry and are expressed as the percentage of the total radiolabeled VSV-G present in the (0) and G H ~ (W) forms, respectively.

ER to TGN Tra

0 30 60 90 120 150 Time (min)

B

- e

0 30 60 90 120 150 Time (min)

FIG. 4. Inhibition of vesicular transport by GTPyS. Semi- intact cells were incubated at 30 "C under conditions described in the legend to Fig. 3. At the times indicated GTPyS was added to a final concentration of 10 p~ and the incubations either shifted to ice (0) or continued at 30 "C for the remainder of a total of 150 min (U). Cell membranes were then collected, digested with endo H, and radiolabeled VSV-G analyzed by SDS-PAGE, autoradiography, and scanning densitometry.

product reached a maximum after 60-90 min and thereafter declined, concomitant with the appearance of the sialylated form (GHt) which reached a maximum after approximately 150 min. Typically 70-90% of the total VSV-G was converted to endo H-resistant species, of which 60-80% was terminally glycosylated.

The significant delay between the appearance of G H ~ and that of GHt strongly suggests that they are formed in distinct subcellular compartments. To further test the need for multiple transport steps to achieve terminal glycosylation in. vitro, we examined the effect of adding various inhibitors to the reaction mixture. A variety of previous studies have demonstrated that vesicular transport in vitro can be inhibited by GTPyS (reviewed in Balch, 1990; Goud and McCaffrey, 1991), presumably reflecting the involvement of GTP-binding proteins in these processes. This inhibitor does not affect the incorporation of exogenous radiolabeled UDP-GlcNAc, UDP- Gal, or CMP-Sia into endogenous acceptor proteins within sealed rat liver Golgi vesicles3 Thus any decrease in processing which we observe reflects a direct inhibition of vesicular transport per se rather than inhibition of any of the various sugar nucleotide transporters or transferases.

To determine the temporal involvement of GTP-binding proteins in transport between the ER and TGN, GTPyS was added to semi-intact NRK cells at various times during the incubation and the cells either shifted directly to ice (Fig. 4, open squares) or incubated at 30 "C for the remainder of the

B. K. Hayes, A. Varki, and W. E. Balch, unpublished data.

Insport in Vitro 4219

150 min time course to allow VSV-G which had progressed beyond the stage sensitive to the inhibitor to be fully processed (Fig. 4, closed squares). As expected, addition of the inhibitor at time 0 completely inhibited the formation of all endo H- resistant species (Fig. 4A). In contrast, addition at later times led to a progressive decrease in the efficiency of inhibition (e.g. after 30 min, a time at which less than 10% of the total VSV-G had become endo H-resistant, more than 45% of the total signal (GH, + GHt) could not be inhibited by GTPyS). At early times G H ~ was the sole product formed in the presence of the inhibitor. GHt could not be detecteduntil approximately 50% of the total GH, signal had become resistant to GTPyS (compare Fig. 4, A and B , 30-min time point). These results indicate that formation of GHt involves at least two distinct GTPyS steps. This is consistent with the hypothesis that in semi-intact NRK cells the G H ~ and GHt forms of VSV are markers for delivery to the cis/medial Golgi and trans Golgi/ TGN, respectively, and that terminal glycosylation requires at least two (ER to cis/medial Golgi and cis/medial Golgi to trans Golgi/TGN) and most likely at least three (ER to cis Golgi, cis to medial Golgi, and medial Golgi to trans Golgi/ TGN) distinct budding and fusion steps, each one of which involves GTP hydrolysis.

Essentially identical results to those obtained with GTPyS were also obtained in experiments conducted with an inhibi- tory monoclonal antibody toward NSF (Wilson et al., 1989) (Fig. 5). Again, there was an approximate 5-min lag before any resistant signal could be detected (Fig. 5A) and a second lag of 10-20 min between the appearance of anti-NSF-resist-

A

B Time (min)

0 30 60 90 120 150 180 Time (min)

FIG. 5. Inhibition of vesicular transport by anti-NSF. Semi- intact cells were incubated at 30 "C under conditions described in the legend to Fig. 3. At the times indicated, 4 pg of monoclonal antibody 4A6 was added and the reaction mixtures shifted to ice for 45 min to allow antibody binding to occur. Reactions were then either maintained on ice (0) or shifted to 30 "C and incubated for the remainder of a total of 180 min (U). Cell membranes were then collected, digested with endo H, and radiolabeled VSV-G analyzed by SDS-PAGE, autoradiography, and scanning densitometry.


ant material in the GHr form and that in the GHt form (Fig. 5, A and B, closed squares). These results suggest that NSF is also required for transport both to and from the cis/medial Golgi and are consistent with other studies which have implicated NSF/Secl8 in multiple vesicular transport events within the exocytic pathway (Beckers et al., 1989; Wilson et al., 1989; Graham and Emr, 1991; Rothman and Orci, 1992).

At Least Three Vesicular Transport Steps Precede the Ap- pearance Terminally Glycosylated VSV-G in Vitro-No lag in the appearance of GTPyS or anti-(NSF)-resistant signals is detected when transport between the ER and cis Golgi is analyzed in semi-intact CHO clone 15B cells (Beckers and Balch, 1989; Beckers et al., 1989). In contrast, a 5-10-min lag preceded the appearance of inhibitor resistant G H ~ in semi- intact NRK cells (Figs. 4 and 5), suggesting that this form of VSV-G may not accurately measure delivery to the cis Golgi under the conditions used. Transport between the ER and cis Golgi in 15B cells is assessed by the acquisition of sensitivity to endo D. This enzyme recognizes a distinct subset of oligosaccharides (Mizuochi et al., 1984) which are products of a- 1,2 mannosidase I, the first Golgi processing enzyme (Korn- feld and Kornfeld, 1985). In contrast, acquisition of endo H resistance also requires the action of both GlcNAc transferase I and a-1,2 mannosidase I1 (Tarentino and Maley, 1974) and in intact cells is generally considered a marker of the medial Golgi (Kornfeld and Kornfeld, 1985).

To examine whether endo H resistance was a measure of delivery to the cis or medial compartments in uitro, we investigated the kinetics of the appearance of endo D-sensitive forms of VSV-G. In intact NRK cells glycosyltransferases acting after a-1,2 mannosidase I recreate an endo D-resistant structure (structures not formed in clone 15B cells (Gottlieb et al., 1975)) which can complicate subsequent analysis using endo D. However, our ability to manipulate the sugar nucleotide pools (Fig. 2) allowed us to circumvent this problem. To ensure that no terminal glycosylation occurred, and to allow parallel analysis of endo H resistance under identical conditions, the incubations contained only UDP-GlcNAc (250 p ~ ) . Since the addition of GlcNAc residues catalyzed by GlcNAc- transferase I is the key step in re-establishing endo D resistance, the post-assay digestions contained @-N-acetylglucosaminidase to remove these residues, in addition to endo D. Under these circumstances two distinct endo D-sensitive species could be detected (Fig. 6). The first of these (GDI), which could be detected within 15-20 min of shifting to 30 "C, migrated more slowly than fully deglycosylated VSV-G (Fig. 6, tracks 1-5, open squares). The formation of GD1 was completely sensitive to kifuensine (an inhibitor of a-1,2 mannosidase I (Elbein et al., 1990)). Similarly, in the presence of swainsonine (an inhibitor of a-1,2 mannosidase I1 (Tulsiani et al., 1982)), GD, was fully sensitive to digestion with endo H (data not shown), indicating that i t is a form of VSV-G having a single processed oligosaccharide. The second endo D-sensitive species (GDz), which co-migrated with GHs (see Fig. 9), appeared after a lag of 30-40 min and thereafter accumulated (Fig. 6, tracks 5-12, closed squares). These kinetics strongly suggest that the two endo D-sensitive forms of VSV-G are the products of distinct compartments, presumably the cis and medial Golgi, respectively.

Analysis of the effect of GTPyS upon the formation of endo D-sensitive species also suggested that GD1 was formed in a distinct compartment to GDz (Fig. 7). Approximately 90% of VSV-G remained resistant to endo D when the inhibitor was added at the start of the incubation, with the small resistant signal being converted exclusively to the GDl form (Fig. 7A, closed squares). This form remained the only endo D-sensitive species detected if GTPyS was added to reaction mixtures which had been incubated at 30 "C for up to 20 min,

1 2 3 4 5 6 7 8 9 1 0 1 1 1 2 =.

a= 2; 0 5 10 x) 30 40 50 M) 75 90 ?x) 120 Time(min)

"1

z20k 0 0 30 Time 60 (min) 90 120

FIG. 6. Kinetics of transport measured by the acquisition of endo D sensitivity. Semi-intact cells were incubated a t 30 "C for the times indicated in a transport mixture supplemented with UDP- GlcNAc (250 p ~ ) and containing desalted rat liver cytosol. Cell membranes were collected by centrifugation, digested with endo D + 8-N-acetylglucosaminidase (tracks 1-11 ) or endo D alone (track 12), and analyzed by SDS-PAGE and autoradiography. Results were quantitated by scanning densitometry and are expressed as the percentage of the total radiolabeled VSV-G present in the GD1 (0) and Go2 (m) forms, respectively.

A

B Time (rnin)

ow-, > 0 30 60 90 120 s?

Time (rnin) FIG. 7. Differential inhibition of transport by GTPrS. Semi-

intact cells were incubated a t 30 "C under conditions described in the legend to Fig. 6. At the times indicated GTPyS was added to a final concentration of 10 p ~ , and the incubations either shifted to ice (0) or continued a t 30 "C for the remainder of a total of 120 min (m). Cell membranes were then collected, digested with a mixture of endo D and 8-N-acetylglucosaminidase, and radiolabeled VSV-G analyzed by SDS-PAGE, autoradiography, and scanning densitometry.


and the reactions continued for a total of 120 min to allow VSV-G which had moved beyond the site of inhibition to be fully processed. At this time (20 min) approximately 10% of the VSV-G had been converted to the G D ~ form (Fig. 7A, open squares), but a total of 35% had reached compartments insensitive to the inhibitor (closed squares). Throughout the course of the experiment the amount of GTPyS-resistant VSV-G continued to increase with time, with the GDz form predomi- nating when the inhibitor was added at times after the first 20-30 min of incubation (Fig. 7 B ) .

Comparisons between the kinetic profiles of the formation of the two endo D-sensitive forms and that of the formation of G H r indicated that the production of G H r more closely paralleled the formation of GDz than that of G D ~ (compare Fig. 6, closed squares, with Fig. 3, open squares). Similarly, comparison between the formation of GTPyS-resistant G H ~ and the appearance of inhibitor-resistant GD1 and GDz also suggested that the majority of the GHr form of VSV-G was produced in the same compartment as GD2 (compare Fig. 7 with Fig. 4A). Together these results are consistent with the hypothesis that endo H resistance is primarily, but not exclusively, a measure of delivery to the medial rather than the cis Golgi, whereas GD1 and GD2 are selective markers of the cis and medial Golgi, respectively.

Terminal Glycosylation of VSV-G Involves Multiple Cytosol- dependent Steps-Intercompartmental movement of proteins in vitro requires both a source of energy (ATP) and soluble proteins (reviewed in Balch, 1989; Rothman and Orci, 1992; Pryer et al., 1992). To further analyze the requirements for the transport of VSV-G from the ER through sequential Golgi compartments to the TGN, we investigated the effect of removing the soluble components at various times during the overall incubation. Aliquots of semi-intact NRK cells were incubated at 30 "C in a complete mixture for various times, the membranes collected by centrifugation and the supernatant discarded. The semi-intact cells were then resuspended, added to fresh reaction mixes containing cytosol (Fig. 8, open squares and closed triangles) or an equivalent volume of buffer (Fig. 8, closed squares), and either incubated at 30 "C (Fig. 8, closed symbols) or maintained on ice (Fig. 8, open symbols) for the remainder of a total of 150 min.

As shown in Fig. 8, removal of cytosolic components at any point during the first 90 min of incubation significantly inhibited subsequent transport to both the cis/medial Golgi (Fig. 8 A ) and trans Golgi/TGN (Fig. 8B) as measured by the formation of the GHr and GHt forms of VSV-G, respectively. In contrast, there was almost no loss in total transport from the ER ( G H ~ + GHt) when the second 30 "C incubation was conducted in the presence of fresh cytosol (Fig. 8A, triangles), indicating that the pelleting and resuspension procedure was not itself responsible for the loss of signal observed in the absence of cytosol. The cytosol-independent signals seen under these conditions were significantly less than the GTPyS and anti-NSF-resistant signals observed at equivalent time points (compare Figs. 4, 5, and 8). This suggests that soluble components are required later than the point sensitive to the GTP analog or the antibody within each transport step. It also confirms that transport rather than processing of VSV- G oligosaccharides is the rate-limiting step of the assay under the conditions used, since processing occurs within the lumen of the Golgi and should not be influenced by the presence or absence of soluble proteins outside this compartment if suf- ficient concentrations of sugar nucleotides are provided.

Distinct Cytosolic Components Are Required at Different Stages of Vesicular Transport-The results presented above strongly suggest that the assay described reconstitutes multiple vesicular transport steps between the ER and TGN and that the GD1, GD2, and GH, forms of VSV-G are markers for

' 0 0 30 60 90

Time (rnin) B

8o 1

0 30 Time (mm) 60 90

FIG. 8. Vesicular transport in vitro involves multiple cytosol dependent steps. Semi-intact cells were incubated at 30 "C in a transport mixture containing desalted rat liver cytosol and supplemented with UDP-GlcNAc (100 p ~ ) , UDP-Gal(250 j", and CMP- Sia (100 PM). At the times indicated the semi-intact cells were collected by centrifugation (3 min at 500 X g, 4 "C), the supernatant discarded, and the membranes resuspended in 50 mM Hepes-KOH, pH 7.2, 90 mM potassium acetate. Aliquots were added to fresh transport mixtures either containing (A, 0) or lacking (W) rat liver cytosol and maintained on ice (0) or incubated at 30 "C (W, A) for the remainder of a total of 150 min. Membranes were again collected, digested with endo H, and radiolabeled VSV-G analyzed as described previously.

the cis, medial, and trans Golgi/TGN, respectively. To investigate whether this system can be used to distinguish those components required for individual transport steps from those acting at multiple steps, we examined the effect of subfrac- tionating the cytosol. As shown in Fig. 9A, separation of those cytosolic proteins insoluble in 30% saturated ammonium sulfate (fraction A) from those soluble in 30%, but insoluble in 80% (fraction B ) , suggested that this was indeed the case. Incubation with fraction B as the sole source of soluble components resulted in the conversion of greater than 80% of VSV-G to the GD, form without any significant formation of G D ~ (Fig. 9A, truck 6). In contrast, fraction A supported movement to compartments in which both oligosaccharides were processed (Fig. 9A, track 51, although transport was stimulated by the addition of the fraction B components (Fig. 9A, truck 7). Consistent with our hypothesis that the two forms of endo D-sensitive VSV-G are markers of distinct compartments, the small cytosol-independent signal was exclusively in the GD1 form (Fig. 9A, truck 3 ) .

The difference between the properties of the two cytosolic fractions was also apparent when transport was assayed by the acquisition of endo H resistance (Fig. 9, B and C). In this case, the apparent level of transport supported by fraction B was significantly lower than that measured using endo D (compare Fig. 9A, track 5, with 9B, truck 4 ) , consistent with our suggestion that ts045 VSV-G is not efficiently converted to endo H-resistant forms in the cis Golgi under the conditions used.

Neither of the two cytosolic fractions were able to support


A 1 2 3 4 5 6 7

OH 00 .C +C A B A+B

8 1 2 3 4 5 r@ ..-

C C G H l h~ & &. Z G H r

0 tC A R A t 0 GH*

@ m c C G H l (I i & - - Z G H r

0 .C tC A E A+B GHS

FIG. 9. Differential cytosolic requirements for multiple vesicular transport events. A, semi-intact cells were incubated on ice (OH, OD; trucks I and 2 ) or at 30 "C (trucks 3-7) in transport mixtures supplemented with UDP-GlcNAc (250 p ~ ) and containing desalted rat liver cytosol (OH, OD, and +C), fraction A ( A ) , fraction B (R), or in the absence of soluble components ( 4 ) for 120 min. Membranes were collected and digested with endo H (truck I ) or a mixture of endo D and 8-N-acetylglucosaminidase (tracks 2-7) and analyzed by SDS-PAGE and autoradiography. B, semi-intact cells were maintained on ice (truck I ) or incubated a t 30 "C for 150 min in a transport mixture supplemented with UDP-GlcNAc (100 pM) , UDP-Gal (250 p ~ ) , and CMP-Sia (100 p ~ ) and containing soluble components as indicated. Endo H-resistant products were analyzed as described previously. C, semi-intact cells were incubated a t 30 "C in a transport mixture containing desalted rat liver cytosol and supplemented with UDP-GlcNAc (100 p ~ ) , UDP-Gal (250 pM) , and CMP-Sia (100 p ~ ) . After 30 min the semi-intact cells were collected by centrifugation (3 min at 500 X g, 4 "C), the supernatant discarded, and the membranes resuspended in 50 mM Hepes-KOH, pH 7.2, 90 mM potassium acetate. Aliquots were added to fresh transport mixtures containing soluble components as indicated and incubated for a further 120 min at 0 "C (truck I ) or 30 "C (tracks 2-6). Endo H- resistant forms of VSV-G were analyzed as described above.

the terminal glycosylation of VSV-G when they were the sole source of soluble components (Fig. 9B, tracks 3 and 4 ) , but did support transport to the TGN when combined (Fig. 9B, track 5). In the case of fraction B, this result is consistent with transport being blocked at the cis Golgi. To further examine the involvement of components in this fraction at later stages, we followed a pelleting and resuspension protocol similar to that described for Fig. 8. Semi-intact NRK cells were first incubated a t 30 "C in a complete mixture containing rat liver cytosol for 30 min to allow a pulse of VSV-G to be transported beyond the cis Golgi. The membranes were then collected by centrifugation, the supernatant discarded, and the semi-intact cells resuspended in buffer. Aliquots were added to fresh reaction mixes containing the various soluble components and incubated at 30 "C for a further 120 min (Fig. 9C). Under these conditions, fraction B was able to support delivery of the fraction of VSV-G which had exited from the cis Golgi during the first incubation to the trans Golgi (Fig. 9C, track 5; compare with track 2). In contrast, fraction A alone could not support delivery to compartments in which GHt was formed, although it did stimulate movement to the medial Golgi (Fig. 9C, track 4 ) and augmented transport supported by fraction B (Fig. 9C, track 6). These results are consistent with the hypothesis that fraction A contains factors essential for transport from the cis to medial Golgi, whereas fraction B lacks factors necessary for budding from the cis Golgi but contains components required for fusion with the trans Golgi/TGN.

A Mutant Form of rablA Inhibits Vesicular Transport from the ER to Cis Golgi and from the Cis to Medial Golgi, but Not Subsequent Steps-Recently we demonstrated that selected members of the rab gene family of low molecular weight GTP-

binding proteins are involved in vesicular transport between the ER and Golgi, both in vitro (Plutner et al., 1991, 1992) and in intact cells (Tisdale et al., 1992). In particular, we showed that a mutation in the GTP binding domain of rablA, AS^'*^ + Ile, (which is equivalent to the oncogenic mutation Asn'I6 4 Ile in ~21'"" (Walter et al., 1986)) created a trans dominant inhibitor of transport, causing VSV-G to accumulate in a post-ER pre-cis Golgi compartment (Tisdale et al., 1992). However, these experiments could not determine whether more distal transport steps were also inhibited by the mutant protein.

T o investigate the potential role of rablA in transport within the Golgi stack we examined the effects of adding bacterially expressed rab 1A proteins' to the transport assay a t various times during the course of an incubation. The results of this experiment are shown in Fig. 10. Consistent with our observations in semi-intact CHO clone 15B cells? addition of 2 pg of the wild-type rablA had no effect on transport (Fig. 10, triangles), whereas addition of an equivalent amount of the mutant rablA12411e prior to the start of the 30 "C incubation inhibited transport to the cis/medial Golgi by over 75% (Fig. lOA, closed squares). However, in contrast to the results obtained with GTPyS and anti-NSF (Figs. 4 and 5), over half of the signal resistant to the mutant rablA a t time 0 was in the GHt form (Fig. 10B), indicating that only transport between the ER and cis/medial Golgi was inhibited.

A

p 100,

!i? 0- ' 0 30 60 90 120 150

B Time (min)

40

20

0 0 30 60 90 120 150

Time (min)

FIG. 10. Differential inhibition of vesicular transport by a mutant form of rablA. Semi-intact cells were incubated a t 30 "C in a transport mixture containing desalted rat liver cytosol and supplemented with UDP-GlcNAc (100 p ~ ) , UDP-Gal (250 p ~ ) , and CMP-Sia (100 p ~ ) . At the times indicated the semi-intact cells were collected by centrifugation (3 min at 500 X g, 4 "C), the supernatant discarded, and the membranes resuspended in 50 mM Hepes-KOH, pH 7.2, 90 mM potassium acetate. Aliquots were added to fresh transport mixtures containing rat liver cytosol and either 2 pg of rablA wild-type (A) or 2 pg of rablA,241re (M, 0) and maintained on ice (0) or incubated a t 30 "C (M, A) for the remainder of a total of 150 min. Membranes were again collected, digested with endo H, and radiolabeled VSV-G analyzed as described previously.


Consistent with this conclusion, incubation of semi-intact NRK cells at 30 "C prior to the addition of the mutant led to a rapid time-dependent loss of the inhibition, with almost complete resistance to the inhibitor being observed after the first 60 min. At all times GHt was the major endo H-resistant species detected and there was no accumulation of G H ~ (compare Fig. 10, A and B ) .

The rapid loss of sensitivity to the mutant rablA might suggest that only the intial step in transport, ER to cis Golgi, requires rablA. However, as the results presented above suggest that G H ~ is not a good marker for delivery of ts045 VSV- G to the cis Golgi compartment, we investigated the effects of the mutant protein on transport as assessed by the acquisition of endo D sensitivity. Preliminary analysis revealed that the GDz form of VSV-G was almost undetectable following incubation in the presence of although, in contrast, almost 50% of the marker protein was converted to the GD1 form (data not shown). This observation suggested that transport both from the ER to the cis Golgi and between the cis and medial Golgi compartments required functional rablA but that the mutant rab protein is a kinetically slow inhibitor.

To more accurately define the temporal aspects of inhibition by the mutant rablA, we attempted to take advantage of the fact that vesicular transport between exocytic compartments can be reversibly inhibited by the chelation of magnesium ions. We reasoned that, by incubating the mutant rablA with the semi-intact cell membranes and cytosol at 30 "C under circumstances in which transport was arrested, we would allow the inhibitor more time to interact with other transport components and that this might increase the efficiency of the inhibition of transport between the ER and cis Golgi compartment. Accordingly, semi-intact cells were incubated in complete reaction mixtures containing 250 p M UDP-GlcNAc for various times and the cell membranes collected by centrifugation and resuspended in buffer. Aliquots were added to fresh incubation mixes without magnesium acetate or ATP but containing 2.5 mM EDTA and mutant or wild-type rablA as appropriate. After incubation at 30 "C for 15 min, 5 mM magnesium acetate and the ATP regenerating system were added and aliquots either transferred to ice (Fig. 11, open squares) or the incubations continued for a total of 135 min (Fig. 11, closed symbols). Under these conditions approximately 75% of VSV-G remained resistant to digestion with endo D when the mutant protein was added to perforated cells directly from ice (Fig. l lA), with the residual endo D- sensitive signal almost exclusively in the G D ~ form. Transport to the cis Golgi (assessed as the sum of GD1 and GDz) rapidly became insensitive to the mutant rab protein, with greater than 80% of the maximal signal being resistant to the inhibitor within 15 min (Fig. 1lA). At this time (15 min) approximately one-third of the maximal G D ~ signal was resistant to the inhibitor (Fig. 11B), with GDz becoming completely resistant after 45-60 min, consistent with the results obtained when transport was assessed by endo H resistance (Fig. 10). Together these data indicate that the mutant rablA inhibits transport from the ER to cis Golgi and between the cis and medial Golgi, but not beyond the medial compartment.

Intercompartmental Transport Is Differentially Sensitive to Protein Kinase Inhibitors-Previously, we demonstrated that export of VSV-G from the ER was inhibited by protein phosphorylation and that movement from the ER to the medial Golgi in semi-intact NRK cells did not require protein kinase activity based upon its insensitivity to the inhibitors staurosporine and H-8 (Davidson et al., 1992). However, when we examined the effects of staurosporine on terminal glycosylation, we were surprised to observe that transport beyond the medial Golgi was completely inhibited (Fig. 12, closed

loo 1

0 10 20 30 40 50 Time (mins)

B

0 10 20 30 40 50 Time (mins)

FIG. 11. Inhibition of vesicular transport between the ER and medial Golgi by a mutant form of rablA. Semi-intact cells were incubated at 30 "C in a transport. mixture containing desalted rat liver cytosol supplemented with UDP-GlcNAc (250 pM). At the times indicated the semi-intact cells were collected by centrifugation (3 min at 500 X g, 4 "C), the supernatant discarded, and the membranes resuspended in 50 mM Hepes-KOH, pH 7.2,90 mM potassium acetate. Aliquots were added to fresh transport mixtures lacking magnesium acetate and ATP but containing 2.5 mM EDTA, rat liver cytosol, and either 2 pg of rablA wild-type (A) or 2 pg of rablAlz41r. (U, 0) and incubated at 30 "C for 15 min. 5 mM magnesium acetate and the ATP-regenerating system were then added to each aliquot which were either maintained on ice (0) or incubated at 30 'C (U, A) for the remainder of a total of 135 min. Membranes were again collected, digested with endo D + P-N-acetylglucosaminidase, and radiolabeled VSV-G analyzed as described previously.

squares, 0-min time points). This result could not be attributed to the inhibition of oligosaccharide processing, rather than vesicular transport, since staurosporine did not reduce the incorporation of radiolabeled UDP-GlcNAc, UDP-Gal, or CMP-Sia into endogenous acceptor proteins within sealed rat liver Golgi vesicles?

Staurosporine interacts with the ATP-binding site of protein kinases and so is competitive with respect to this nucleotide (Ruegg and Burgess, 1989). Consequently, prior to further investigations of the inhibition, we first defined the lowest concentration of ATP required to support vesicular transport to the TGN in semi-intact NRK cells. These studies showed that transport to compartments containing sialyltransferase required the concentration of ATP to be maintained in excess of 200-250 pM (data not shown). In the presence of a 250 pM ATP regenerating system, 0.25 p~ staurosporine caused half- maximal inhibition of transport to the TGN based on the formation of GH~. As expected, the inhibition could be prevented by increasing the concentration of ATP to 2.5 mM, but it was not antagonized by the addition of 2.5 mM GTP.


A

E 100 m ln I

‘ji 80 I r 8 60 - ,g 40 - 2

c 0

L

4 2 0 - > (0 ’ o * - , , , , , , I , , s? 0 30 60 90 120 150

Time (min)

B

1

m ln ._

> v) > s?

20 40p 0 0 30 60 90 120 150

Time (rnin)

FIG. 12. Inhibition of vesicular transport by staurosporine. Semi-intact cells were incubated at 30 “C in a transport mixture containing 250 @I ATP, 5 mM creatine phosphate, 5 milliunits of creatine phosphokinase, desalted rat liver cytosol, and sugar nucleotides as described in the legend to Fig. 3. At the times indicated staurosporine was added to a final concentration of 2.5 pM and the incubations either shifted to ice (0) or continued at 30 ”C for the remainder of a total of 150 min (m). Radiolabeled VSV-G was analyzed as described in the legend to Fig. 4.

Temporal analysis revealed that an early step in transport from the medial cisternae was affected (Fig. 12B). For example, at 45 min, a time a t which less than 5% of VSV-G had become sialylated, 28% (approximately 50% of the maximal sialylation observed) was insensitive to the addition of staurosporine (2.5 PM) (Fig. 12B). At this time approximately 40% of the VSV-G had been rendered endo H-resistant (50% of the maximal signal) (Fig. 12A).

Staurosporine did not affect transport to the medial Golgi as assessed by the formation of G D ~ , although in the presence of the inhibitor inclusion of UDP-Gal and CMP-Sia did not generate species resistant to P-N-acetylglucosaminidase (data not shown), indicating that both galactosylation and sialylation were prevented. Identical results to those reported above were also obtained with the protein kinase inhibitor H-8, except that half-maximal inhibition required approximately 250 WM H-8 (data not shown). Together these results suggest that either budding from the medial Golgi or targeting to the trans Golgi requires the action of one or more protein kinases.

DISCUSSION

The primary objective of this study was to develop an assay system capable of measuring movement of newly synthesized membrane proteins throughout the secretory pathway. This system could then be used to distinguish those components required at a single vesicular transport step from those involved at multiple steps, in a single assay. Previously we demonstrated that CHO cells perforated by the swelling and scraping technique can reconstitute movement between the

ER and medial Golgi in an ATP- and cytosol-dependent fashion, but our attempts to detect in uitro galactosylation and sialylation of VSV-G (an indication of transport to the trans Golgi and TGN) were unsuccessful (Schwaninger et al., 1991). Madin-Darby canine kidney cells perforated by the “nitrocellulose stripping” procedure sialylated VSV-G in vitro, although transport was independent of added cytosol and proceeded to a limited extent even in the absence of added ATP (Simons and Virta, 1987). This suggested that it might be possible to develop an ATP- and cytosol-dependent system to follow transport of marker proteins throughout the secretory pathway i n uitro. We therefore analyzed the character- istics of semi-intact cells from several different cell lines to see if differences existed between them.

The data described above demonstrate that a combination of semi-intact NRK cells and rat liver cytosol can reconstitute movement from the ER to compartments capable of sialylat- ing VSV-G. This was also possible using several other cell lines, and at present we are uncertain of the reason why we cannot reconstitute this step using CHO cells. One possible explanation concerns the behavior of the Golgi apparatus in perforated cells. Recently, we developed a morphological assay for transport based upon the co-localization of immunoreactive ts045 VSV-G and a-1,2 mannosidase I1 (a marker of the medial Golgi (Farquahar, 1985)) (Plutner et al., 1992). In that study we also showed that the Golgi fragments during incubations i n uitro, concomitant with changes in the organization of the cytoskeleton. In perforated CHO cells transport between the ER and medial Golgi (Schwaninger et al., 1991) appears to proceed at a slower rate than is observed in NRK cells (Davidson et al., 1992). Thus, a possible reason for our inability to detect sialylation in semi-intact CHO cells is that the more distal Golgi cisternae become dispersed within the assay mix before efficient targeting and fusion to these compartments can occur.

Our ability to manipulate the concentrations of soluble components in the assay mixture allowed us to examine the dependence of in vitro sialylation on individual sugar nucleotides. These studies indicated that both soluble and membrane associated pools were involved and revealed unantici- pated interactions between them. UDP-GlcNAc and UDP- Gal are synthesized in the cytoplasm and are subsequently transported across Golgi membranes by facilitated diffusion using distinct UMP-coupled antiports, whereas CMP-Sia is synthesized in the nucleus and transported from the cytoplasm into the Golgi lumen in exchange for CMP (reviewed in Hirschberg and Snider, 1987). We were therefore surprised to observe that high concentrations of UDP-GlcNAc inhibited the formation of GHt. The effect could be antagonized by UDP-Gal, suggesting that it resulted from a direct interaction between these sugar nucleotides and arguing against the inhibition being due to a block in transport following the spurious addition of GlcNAc to serine or threonine residues of an essential cytosolic component by soluble GlcNAc transferases (reviewed in Hart et al., 1989). Enzymatic digests indicated that in the presence of 500 pM UDP-GlcNAc, VSV- G oligosaccharides were not modified beyond the Glc- NAc2Man3G1cNac2 form,4 suggesting that the loss of sialylation could not be attributed to the formation of oligosaccharide chains which were not substrates of galactosyltransferase. Thus, one possible explanation for the inhibition we observe is that, at elevated concentrations, UDP-GlcNAc can compete with UDP-Gal for the galactosyl transporter and that unla- beled acceptors (either VSV-G or endogenous proteins) present throughout the Golgi stack deplete any luminal pool of UDP-Gal prior to the arrival of radiolabeled VSV-G in the

H. W. Davidson and W. E. Balch, unpublished results.


trans Golgi. Alternatively, UDP-GlcNAc might directly deplete the available UDP-Gal by providing an alternative acceptor molecule. Although UDP-GlcNAc is not itself an acceptor for galactosyltransferase, GlcNAc and GlcNAc poly- mers are, with (GlcNAc)z having a submillimolar K,,, value (Schanbacher and Ebner, 1970). Thus, conversion of luminal UDP-GlcNAc to these compounds could result in the deple- tion of UDP-Gal pools and the consequent inhibition of VSV- G sialylation. Whatever the cause, this observation confirms the need to establish and maintain the correct balance between individual sugar nucleotides when using oligosaccharide modifications to measure vesicular transport through sequential compartments of the exocytic pathway.

The data described above strongly suggest that the sialylation of VSV-G which we observe is a consequence of multiple vesicular transport reactions. Both the kinetic and inhibitor profiles argue that it occurs by an ordered series of discontinuous steps which are characteristic of intercompartmental transfer.The distribution of Golgi-associated oligosaccharide- processing enzymes has been well documented (Kornfeld and Kornfeld, 1985), and together with the data from studies using the drug brefeldin A (Doms et al., 1989; Chege and Pfeffer, 1991) suggest that a-1,2 mannosidase 11, galactosyltransferase, and sialyltransferase are likely to be markers of the medial Golgi, trans Golgi, and TGN, respectively. On this basis we believe that our assay reconstitutes at least four distinct budding and fusion reactions (ER to cis Golgi, cis to medial Golgi, medial to trans Golgi, and trans Golgi to TGN) and that we can measure the first, second, and fourth steps by the appearance of t h e G D ~ , G D ~ , a n d G H ~ forms of ts045, respectively.

The enzymatic digestion protocols which we used in this study to analyze transport between the ER and TGN did not allow us to specificly measure galactosylation. Thus, we could not readily resolve the third step, delivery to the trans Golgi (the likely site of this modification) from the presumed fourth step, delivery to the TGN (the expected site of sialylation). Addition of galactose residues causes only a small decrease in the electrophoretic mobility of the GHr form which, although detectable, cannot be separately quantitated. Similarly, measuring galactosylation on the basis of endo D sensitivity requires a comparison between digests conducted with endo D and P-N-acetylglucosaminidase in the presence or absence of P-galactosidase. This analysis is complicated by the fact that one of the two VSV-G oligosaccharides is processed primarily to a form which is resistant to digestion with D. pneumoniae P-N-acetylglucosaminidase after terminal Gal and Sia residues have been removed: suggesting that the complex oligosaccharides formed in perforated NRK cells are a mixture of bi-antennary and tri-antennary structures, similar or identical to those previously isolated from infected BHK cells (Hunt et al., 1983; Turco and Pickard, 1982). We are therefore currently investigating alternative methods of quantitating galactosylation, including the use of immobilized lectins such as Ricinus communis agglutinin, which will enable us to determine whether galactose and sialic acid residues are added in distinct, or the same, compartments in vitro.

Several pieces of evidence support the conclusion that under the conditions used only one of the two ts045 VSV-G oligosaccharides is processed in the cis Golgi. 1) There was a 10- 15 min lag between the initial appearance of G D ~ and that of the G D ~ form, 2) GTP-yS and anti-NSF resistant signals at early time points are exclusively as GD1, and 3) incubations in the absence of cytosol or the presence of fraction B produced only the G D ~ form. At present the mechanisms govern- ing whether the oligosaccharide chain at a particular site in a glycoprotein is a substrate of a processing enzyme are not fully understood, although the protein structure adjacent to

the site is likely to be a key determinant (Hubbard, 1988). This suggests that movement of VSV-G between the cis and medial cisternae is accompanied by a conformational change within the glycoprotein which allows processing of the second chain to proceed and which presumably reflects differences in the prevailing ionic and enzymatic concentrations within the lumen of these elements of the Golgi stack. A similar distinc- tion between cis and medial processing of VSV-G has also been observed in CHO and NRK cells infected with wild-type VSV (Schwaninger et al., 1991; Davidson et al., 1992). How- ever, in contrast to the results presently described, in CHO cells both VSV-G oligosaccharides were processed to endo D- sensitive forms in the cis Golgi. This suggests that equivalent Golgi compartments may vary between cell types, as has been discussed recently by Mellman and Simons (1992).

As expected from the results of previous studies which examined vesicular transport in vitro (reviewed in Balch, 1989; Rothman and Orci, 1992), movement of VSV-G between each adjacent compartment of the exocytic pathway in perforated NRK cells required ATP, soluble proteins, and functional NSF and could be inhibited by nonhydrolyzable analogues of GTP such as GTPyS. However, in contrast to these common properties, we also demonstrated that individual steps had distinct requirements. For example, at least one cytosolic component required for movement between the cis and medial Golgi was identified among the proteins insoluble in 30% ammonium sulfate (fraction A). This component(s) was not essential for movement between the ER and cis Golgi or transport beyond the medial cisternae, although we cannot exclude the possibility that there is also a membrane associated pool which can support these reactions. Similarly, both fraction A and fraction B could support transport from the ER to the cis Golgi, but the proteins required for delivery to the trans Golgi and TGN were only present in the soluble fraction (fraction B). Such results are consistent with those obtained previously in yeast by Graham and Emr (1991) who demonstrated that although SEC18/NSF was required for movement between multiple exocytic compartments, SEC23 (and by implication SEC12, SEC24, and SARl (Pryer et al., 1992)) was only required for transport between the ER and cis Golgi compartment (Graham and Emr, 1991).

Another difference between transport from the ER to the medial Golgi and that from the medial to the trans Golgi was a differential sensitivity to protein kinase inhibitors. In a previous study we showed that export from the ER could be inhibited by protein phosphorylation (Davidson et al., 1992), which may be related to the arrest of vesicular transport during mitosis. The inhibition could be prevented by staurosporine and H-8 showing that ER to medial Golgi transport normally does not involve protein kinase activity. In contrast, movement from medial to trans Golgi compartments was completely inhibited by staurosporine and H-8, indicating that later steps of the pathway require protein phosphorylation. At present the identity of the kinase(s) required for movement between the medial and trans cisternae, it's substrates, and the possible relationship to the enzyme(s) responsible for inhibiting export from the ER remain to be estab- lished. Preliminary results suggest that both are membrane- associated and unrelated to cyclic AMP or calcium-dependent enzyme^.^ Protein kinases have also been implicated in the regulation of endocytic transport (Tuomikowski et al., 1989; Woodman et al., 1992), transport to the yeast vacuole (Her- man et al., 1991), fragmentation of the Golgi during mitosis (Lucocq et al., 1991), and delivery of endocytosed asialogly- coproteins to lysosomes (Ohashi and Ohnishi, 1991), although their precise roles are presently unknown.

It is now generally accepted that vesicular transport between compartments of the exocytic pathway requires multi-


ple GTP-binding proteins including members of the rab gene family (reviewed in Balch, 1990; Goud and McCaffrey, 1991; Pryer et al., 1992). rablA is 76% identical to yptlp (Touchot et al., 1987; Chavrier et al., 1990 ), a protein required for vesicular transport between the ER and Golgi in yeast (Schmitt et al., 1988; Segev et al., 1988). It can functionally substitute for yptlp in YPTl null cells and thus appears to be the mammalian homologue (Haubruck et al., 1989). Re- cently, we have shown that trans dominant mutants of rabl can inhibit transport from the ER to cis Golgi both in uiuo (Tisdale et al., 1992) and in uitro.’ However, as immunocyto- chemical studies have shown that immunoreactive rabl/yptlp co-localizes not only with markers for the ER and cis Golgi, but also with the medial Golgi marker mannosidase I1 (Plut- ner et al., 1991), and to a lesser extent with wheat germ agglutinin staining (Segev et al., 1988) (which is primarily a marker of the more distal Golgi cisternae (Kornfeld and Kornfeld, 1985)), we thought it likely that rabl might also function in transport within the Golgi stack. This hypothesis could not be tested in the previous studies which focussed upon a single vesicle transport step. The biochemical data presently described demonstrate that rablA is indeed required for movement between the ER and cis Golgi, and the cis and medial Golgi, but is not involved in transport from the medial Golgi to the TGN. At present the precise function of rab proteins is unknown, although it has been suggested that they may be involved in ensuring that transport vesicles fuse to the appropriate acceptor membranes (Bourne, 1988). A key prediction of this “targeting” hypothesis is that individual rab proteins should only regulate a single vesicle fusion step. Our results instead demonstrate that rablA is involved in two sequential fusion reactions and thus suggest that at least some rab proteins may not be directly responsible for vesicle targeting.

In conclusion, we believe that the system which we have developed, which allows us to analyze the effects of reagents (including antibodies, subfractions of cytosol, and bacterially expressed mutant proteins) on the movement of membrane proteins between multiple subcellular compartments in uitro, will enable us to more accurately define the complex biochemical interactions involved in vesicular transport.

Acknowledgments-We thank J. E, Rothman (Sloan-Kettering Institute, New York) for supplying anti-NSF clone 4A6 and C. Nuoffer (Scripps Research Institute, San Diego, CA) for the wild- type and mutant rablA proteins, We are also grateful to B. K. Hayes and A. Varki (University of California San Diego, San Diego, CA) for helpful criticism and advice and J. D. Watkins (Scripps Research Institute) for critical reading of the manuscript.

REFERENCES Baker, D., Hicke, L., Rexach, M., Schleyer, M., and Schekman, R. (1988) Cell

Balch, W. E. (1989) J. Biol. Chem. 264,16965-16968 Balch, W. E. (1990) Trends Biochem. Sci. 16,473-477 Balch, W. E., Dunphy, W. G., Braell, W. A., and Rothman, J. E. (1984) Cell

Beckers, C. J. M., and Balch, W. E. (1989) J. Cell Biol. 108 , 1245-1256 Beckers, C. J. M., Keller, D. S., and Balch, W. E. (1987) Cell 60,523-534

64,335-344

39,405-416

Beckers, C. J. M., Block, M. R., Glick, B. S., Rothman, J. E., and Balch, W. E,

Bergmann, J. E. (1989) Methods Cell Biol. 3 2 , 85-110 Block, M. R., Glick, B. S., Wilcox, C. A., Wieland, F. T., and Rothman, J. E. Bourne, H. (1988) Cell 63,669-671

(1988) Proc. Natl. Acad. Scr. U. S. A. 86,7852-7856

Chavrier, P., Vingron, M., Sander, C., Simons, K., and Zerial, M. (1990) Mol.

Chege, N. W., and Pfeffer, S. R. (1990) J. Cell Biol. 111,893-899

Davidson, H. W., McGowan, C. H., and Balch, W. E. (1992) J. Cell Bwl. 116 , Clary, D. O., and Rothman, J. E. (1990) J. Bwl. Chem. 266,10109-10117

(1989) Nature 339,397-398

Cell Biol. 10 , 6578-6585

1.143-1 355 ”” _~.. Doms, R. W., Russ, G., and Yewdell, J. W. (1989) J. Cell Biol. 109,61-72 Elbein, A. D., Tropea, J. E., Mitchell, M., and Kaushal, G. P. (1990) J. Biol.

Etchison, J. R., Robertson, J. S., and Summers, D. F. (1977) Virology 78,375-

Farquahar, M. G. (1985) Annu. Rev. Cell Biol. 1,447-488 Flamand, A. (1970) J. Gen. Virol. 8,187-195 Gottlieb, C., Baenziger, J., and Kornfeld, S. (1985) J. Bioi. Chem. 2 6 0 , 3303-

Goud, B., and McCaffrey, M. (1991) Curr. Opin. Cell Biol. 3,626-633 Goud, B., Salminen, A., Walworth, N. C., and Novick, P. J. (1988) Cell 6 3 ,

Graham, T. R., and Emr, S. D. (1991) J. Cell Biol. 114,207-218 Griff, I. C., Schekman, R., Rothman, J. E., and Kaiser, C. A. (1992) J. Biol.

Hart, G. W., Haltiwanger, R. S., Holt, G. D., and Kelly, W. G. (1989) Annu.

Haubruck, H., Prange, R., Vorgias, C., and Gallwitz, D. (1989) EMBO J. 8,

Herman, P. K., Stack, J. H., DeModena, J. A., and Emr, S. D. (1991) Cell 6 4 ,

Hirschberg, C. B., and Snider, M. D. (1987) Annu. Rev. Biochem. 66, 63-88 Hubbard, S. C. (1988) J. Bwl. Chem. 263.19303-19317 Hunt, L. A., Davidson, S. K., and Golemboski, D. B. (1983) Arch. Biochem.

Kornfeld, R., and Kornfeld, S. (1985) Annu. Rev. Biochem. 64,631-664 Lucocq, J., Warren, G., and Pryde J. (1991) J. CeU Sci. 100,753-759 Melancon, P., Glick, B. S., Malhoira, V., Weidman, P. J., Serafini, T., Gleason,

Mellman, I., and’simons, K. (1992) Cell 68, 829-840 Mhyyhi , T., Amano, J., and Kobata, A. (1984) J. Biochm. (Tokyo) 96,1209-

Chem. 266,15599-15605

392

3309

753-768

Chem. 267,12106-12115

Rev. Biochem. 68,841-874

1427-1432

425-437

Bwphys. 226,347-356

M. L., Orci, L. and Rothman, J. E. (1987) Cell 61,1053-1062

Pryer, N. K., Wuestehube, L. J., and Schekman, R. (1992) At

Roth, J., Taatjes, D. J., Lucocq, J. M., Weinstein, J., and Paulson, J. e. (1985)

Rothman, J. E. (1987) J. Biol. Chem. 2 6 2 , 12502-12510

R u e p U T;. and Bur ess, G M. (1989) Trends Pharmacol. Sci. 10,218-220 Rothman, J. E., and Orci, L. (1992) Nature 356,409-415

Ruo ola H Kabcenefl, A. K., and Ferro-Novick, S. (1988) J. Cell Bzol. 107 ,

Schanbacher, F. L., and Ebner, K. E. (1970) J. Biol. Chem. 246,5057-5061 Schmitt, H. D., Puzicha, M., and Gallwitz, D. (1988) Cell 6 3 , 635-647 Schwaninger, R., Beckers, C. J. M., and Balch, W. E. (1991) J. Biol. Chem.

Schwaninger, R., Plutner, H., Bokoch, G. M., and Balch, W. E. (1992) J. Cell

Segev, N., Mulholland, J., and Botatein, D. (1988) Cell 62,915-924 Simons, K., and Virta, H. (1987) EMBO J. 6,2241-2247 Tarentino, A. L., and Maley, F. (1974) J. Biol. Chem. 249,811-817 Tisdale, E. J., Bourne J. R., Khosravi-Far, R., Der, C. J., and Balch, W. E. Touchot, N., Chardin, P., and Tavitian, A. (1987) Proc. Natl. Acad. Sei. U. S. A.

Tulsiani, D. R. P., Harris, T., and Touster, 0. (1982) J. Biol. Chem. 267,7936-

Tuomikowski, T., Felix, M.-A., Doree, M., and Gruenberg, J. (1989) Nature

61,471-516

Cell 4 3 , 287-295

1465-1476

2 6 6 , 13055-13063

Biol. 119,1077-1096

(1992) J. Cell B ~ I . i m , 749-761

84,8210-8214

7939

342,942-945 Turco, S. J., and Pickard. J. L. (1982) J. Biol. Chem. 267,8674-8679 Walter, M., Clark, S. G., and Levinson, A. D. (1986) Science 2 3 3 , 649-652 Wilson D. W. Wilcox C. A,, Flynn G. C., Chen, E., Kuang, W.-J., Henzel, W.

J., Biock, M . R., Ulirich, A., and’Rothman, J. E. (1989) Nature 3 3 9 , 355-

Woodman, P. G., Mundy, D. I., Cohen, P., and Warren, G. (1992) J. CeU Biol. 116,331-338

Documents

Differential Inhibition of Multiple Vesicular Transport Steps between