A Rab GTPase Is Required for Homotypic Assembly of the Endoplasmic Reticulum*

To define the requirements for the homotypic fusion of mammalian endoplasmic reticulum (ER) membranes, we have developed a quantitative in vitro enzyme-linked immunosorbent assay. This assay measures the formation of IgG (H 2 L 2 ) following the fusion of ER microsomes containing either the heavy or light chain subunits. Guanine nucleotide dissociation inhibitor (GDI), a pro- tein that extracts Rab GTPases in the GDP-bound form from membranes, potently inhibits fusion. Inhibition was not observed using GDI mutants defective in Rab binding. Kinetic analysis of the inhibitory effects of GDI revealed that Rab activation is required immediately preceding or coincident with fusion and that this step is preceded by a priming event requiring a member of the AAA ATPase family. Our results suggest that homotypic fusion of ER membranes requires Rab and that Rab activation is a transient event necessary for the forma- tion of a fusion pore leading to the mixing of luminal contents of ER microsomes.


From the Departments of Cell and Molecular Biology, The Scripps Research Institute, La Jolla, California 92037
To define the requirements for the homotypic fusion of mammalian endoplasmic reticulum (ER) membranes, we have developed a quantitative in vitro enzyme-linked immunosorbent assay. This assay measures the formation of IgG (H 2 L 2 ) following the fusion of ER microsomes containing either the heavy or light chain subunits.

Guanine nucleotide dissociation inhibitor (GDI), a protein that extracts Rab GTPases in the GDP-bound form from membranes, potently inhibits fusion. Inhibition was not observed using GDI mutants defective in Rab binding. Kinetic analysis of the inhibitory effects of GDI revealed that Rab activation is required immediately preceding or coincident with fusion and that this step is preceded by a priming event requiring a member of the AAA ATPase family. Our results suggest that homotypic fusion of ER membranes requires Rab and that Rab activation is a transient event necessary for the formation of a fusion pore leading to the mixing of luminal contents of ER microsomes.
Regulated fusion is a critical feature of heterotypic membrane interactions involved in vesicular transport of cargo through the exocytic and endocytic pathways (reviewed in Ref. 1) and homotypic events leading to the reassembly of intracellular organelles following their disassembly during mitosis (reviewed in Ref. 2). A number of advances have been made in recent years in recognition of components comprising the targeting/fusion machinery used by vesicles to deliver cargo to subcellular organelles (reviewed in Refs. 3 and 4). In contrast, less is known about the mechanism of homotypic fusion that controls the assembly of these compartments.
To study homotypic fusion, cell-free assays have been developed that measure the fusion of endosomes (5), mitotic Golgi fragments (6,7), yeast vacuoles (8 -10), and ER 1 microsomes (11)(12)(13)(14)(15)(16). In yeast, homotypic fusion of ER microsomes has been suggested to require the luminal molecular chaperone KAR2 (the yeast homolog of mammalian BiP) (11) and Cdc48p (17), the latter being a member of a larger gene family of N-ethylmaleimide (NEM)-sensitive AAA ATPases, which includes the intra-Golgi targeting/fusion factor NSF (reviewed in Refs. 3, 18, and 19). Similarly, liver ER microsomes inactivated by NEM lack fusion activity (14). Moreover, the mammalian homolog to Cdc48p, p97, in conjunction with NSF has been shown to be required for homotypic fusion of vesiculated Golgi membranes (6,7). The potential role of p97 in the fusion of mammalian ER membranes has not been tested.
In addition to a role for members of the AAA ATPase gene family in fusion, Rab GTPases have also been shown to be essential for the targeting and/or fusion of membranes throughout the exocytic and endocytic pathways (reviewed in Refs. 20 and 21). Morphological, genetic, and biochemical approaches have revealed an essential role for Rab1 in ER to Golgi transport in mammalian cells (22)(23)(24)(25). Rab5 is involved in the homotypic fusion of endosomes (26), whereas Ypt7p is involved in the homotypic fusion of yeast vacuoles (8,9). Given the participation of different Rab GTPases in homotypic fusion of at least two intracellular compartments, a novel Rab protein may also mediate homotypic fusion of ER membranes. To test this hypothesis, in the present study we describe the development of a simple ELISA-based assay, which allows us to efficiently quantitate the fusion of mammalian ER microsomes. We now report that fusion is mediated by a small monomeric GTPase(s) belonging to the Rab family.
Membrane Fusion/Assembly Assay-Post-nuclear supernatants from P3U.1 and Ag8(8) cells were incubated at 32°C for the time indicated under "Results" (generally 15 min) in a total volume of 200 l containing 25/125 supplemented with 2.5 mM MgOAc, 1 mM ATP and an ATP-regenerating system (4.8 mM creatine phosphate and 5 IU/ml creatine phosphate kinase). The assay was terminated by transfer to ice and supplemented in order with 25 mM iodoacetate (to alkylate free sulfhydryl groups) and 20 l of ice-cold lysis buffer (1% Triton X-100 in 200 mM Tris-HCl, 1.5 M NaCl, 25 mM EDTA, pH 7.5). After vortexing the mix for 10 s, the insoluble material was pelleted at 16,000 ϫ g for 10 * The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. min at 4°C. The supernatant (90 l) was added to wells of ELISA strips that had been precoated for 2 h with 2 g/ml protein G in 100 mM Na 2 CO 3 , washed with 25 mM Tris-HCl, 0.8% NaCl, 0.2% KCl, pH 8 (TBS), then blocked for 1 h with TBS containing 0.2% Tween 20, and washed twice with TBS. Purifed IgG (0 -10 ng/well) was used for a standard curve for quantitation in each experiment. The ELISA strips were incubated in the dark at 4°C for 3 h, after which they were washed four times with TBS before 100 l of chicken anti-mouse IgG horseradish peroxidase (diluted 1 in 1000 in TBS, 0.2% Tween 20) was added to each well, and samples were incubated at 4°C overnight. Unreacted antibody was then removed by washing four times with TBS, and 100 l of assay mixture (10 mg of o-phenylenediamine/25 ml of 27 mM citric acid, 51 mM Na 2 HPO 4 , pH 5 plus 1 l of H 2 O 2 ) was added to each well. The absorbance at 490 nm was determined using a Bio-Rad 3550 microplate reader, and absolute values were determined by comparing individual mean sample absorbances to those from the IgG standard curve. All incubations were carried out in duplicate with a standard error of Ϯ 10 -15% as indicated under "Results."

RESULTS
The ER is an organelle that specializes in the folding and oligomerization of proteins for export. We previously described a cell-free assay based on the ability of radiolabeled ([ 35 S]) heavy (H) and light (L) chain containing ER microsomes prepared from two different cell lines to promote the fusion-dependent oligomerization of mature IgG molecules (12). Using this assay, we demonstrated that H 2 L 2 assembly is efficient (ϳ50% of the total H and L chain pool is assembled) and that oligomerization is not rate-limiting. Therefore, the rate assembly of mature IgG is a direct measure of membrane fusion.
We now report the development of an ELISA assay to more rapidly quantitate ER fusion. This assay encompasses the general principles of the previous fusion reaction (12) in that microsomes are prepared from Ag8(8) and P3U.1 cells expressing the IgG H and L chains, respectively. Incubation at 32°C in the presence of ATP and cytosol leads to luminal continuity between the two membrane populations and oligomerization of the H and L chains to form mature IgG. To quantitate the appearance of H 2 L 2 , membranes are solubilized by detergent and centrifuged to remove insoluble material, and the supernatant is added to protein G-protein-coated ELISA plates, which bind the H chain found only in mature H 2 L 2 . Unbound material is removed by washing, followed by incubation with a horseradish peroxidase-conjugated anti-mouse antibody. The amount of H 2 L 2 in each well is quantitated by spectroscopy. Fig. 1 shows the kinetics of the reaction over a 30-min time course in the absence or the presence of 0.1% Triton X-100. The addition of detergent allows us to distinguish between the appearance of H 2 L 2 from bonafide fusion-dependent assembly of sealed membranes and fusion-independent oligomerization, which as shown previously (12) can occur if the membranes are lysed by addition of detergent at the beginning of the incubation. Using intact membranes, there is a rapid onset of H 2 L 2 assembly that proceeds at a linear rate following a brief lag (Ͻ1-2 min) and reaches a plateau after 15 min of incubation (Fig. 1). In contrast, in the presence of 0.1% Triton X-100, the kinetics of H 2 L 2 assembly has a prolonged lag period (10 min) and a reduced rate. The reduced kinetics of H 2 L 2 assembly in the presence of detergent undoubtedly reflects the loss of the highly specialized folding/oligmerization environment of the ER (12). Given the kinetic differences between assembly observed in the absence or the presence of detergent, incubations are limited to the 15-min time period where nonluminal H 2 L 2 assembly in response to any potential membrane lysis would contribute only minimally to the signal derived from fusionrelated assembly. A detergent-treated sample is always included as an internal control in each experiment, and this value, which measures the maximal contribution of luminal independent assembly, is subtracted from all reported values. All of the basic properties of the ELISA based assay were found to be identical to those reported for ER fusion detected by appearance of radiolabeled H 2 L 2 (12) (not shown).
Previous studies by our group and others using GTP or nonhydrolyzable GTP analogs such as GTP␥S have suggested a potential role for GTPases in the fusion of mammalian ER membranes (12,15,16). To define the GTP requirement for ER assembly, we first examined whether the GTP-dependent step required a membrane-associated or cytosolic component. Pretreatment of cytosol with 50 M GTP␥S for 15 min at 32°C in the presence of ATP and an ATP-regenerating system, followed by the addition of 10 mM GTP to neutralize the inhibitory effect of GTP␥S, had no effect on the subsequent ability of cytosol stimulate H 2 L 2 assembly compared with untreated cytosol (not shown). In contrast, the addition of GTP␥S to the assay potently inhibited ER fusion (Fig. 2A, lane g) (12). Identical results were observed with GDP␤S ( Fig. 2A, lane h), suggesting that ER fusion requires a complete GTPase cycle. No effect of either analog was observed on the assembly of H 2 L 2 in the presence of detergent (not shown).
The ability of both GTP␥S and GDP␤S to inhibit fusion is diagnostic of the activity of small GTPases belonging to the Ras superfamily. Two guanine nucleotide binding proteins associated with the ER and compartments of the early secretory pathway are the ARF1 and Sar1 GTPases. Mutants that restrict these GTPases to the GDP-or GTP-bound forms have potent trans-dominant effects on ER to Golgi transport in vivo (25) and in vitro (22)(23)(24) by inhibiting the assembly/disassembly of COPII and COPI coat components, respectively (29 -32).
To determine if either of these two GTPases affect ER assembly, we incubated microsomes with the GDP-(inactive) or GTPrestricted (active) forms. Fig. 2A (lanes b-e) demonstrates that the mutants had little effect at concentrations that potently inhibit ER to Golgi vesicular transport (30 -32). We conclude that the Sar1 and ARF1 GTPases are not involved in ER assembly.
Rab family GTPases are believed to play an unknown but critical role in vesicle targeting and fusion (reviewed in Ref. 21). To test if members of the Rab GTPase family are involved in ER assembly, we treated membranes with Rab GDP dissociation inhibitor (GDI), a protein essential for the cycling of Rab between GDP-and GTP-bound forms. Previous studies have demonstrated that GDI binds exclusively to the GDP-bound form of Rab proteins and that the addition of GDI to a variety of Rab-dependent in vitro fusion assays leads to potent inhibition (33-36), presumably due to the ability of GDI to efficiently extract the GDP-bound form of Rab proteins from the mem- brane (reviewed in Refs. 37 and 38). As shown in Fig. 2A, preincubation of microsomes with GDI on ice prior to incubation at 32°C leads to a complete, dose-dependent inhibition of fusion with an IC 50 of ϳ0.5 M. Pretreatment of either membrane alone was sufficient to inactivate fusion (not shown), emphasizing the need for Rab on each fusion partner. No inhibition of H 2 L 2 assembly was observed in the presence of detergent (not shown), demonstrating that GDI blocks the fusion of intact membranes. As additional controls, we examined the effect of selected GDI mutants on ER fusion. Residues involved in Rab binding have been recently shown to occur in sequence conserved regions, which form a compact structure at the apex of GDI (39). Mutation of the surface residues Tyr 39 , Tyr 249 , or Met 250 found in sequence conserved regions 1 and 3B, respectively, potently block the ability of GDI to bind Rab in vitro and to extract Rab from membranes (39) and prevent the ability of GDI to inhibit ER to Golgi transport in vitro (34). 2 Incubation of ER microsomes with these mutants at a 5-fold excess over the concentration of wild-type GDI necessary to elicit complete inhibition of ER assembly ( Fig. 2A) had at most a modest effect on ER fusion (Fig. 2B). In addition, Rho-GDI, which extracts Rho GTPases and inactivates Rho/Rac-dependent events (reviewed in Ref. 40), had no effect on ER assembly at concentrations up to 50 M (not shown). Assembly does not require the Rab1 isoform involved in ER to Golgi and intra-Golgi transport because addition of a trans-dominant mutant (Rab1A[N124I]), which fails to bind guanine nucleotide and which potently inhibits the fusion of ER-derived vesicles to Golgi compartments (22,24,25,41), had no effect on homotypic fusion (Fig.  2A, lane f). These results demonstrate that a novel Rab protein is required for the homotypic fusion of ER membranes.
To assess whether the requirement for Rab in ER assembly occurs in conjunction with the activity of a NEM-sensitive factor(s), we first examined whether our assay is sensitive to sulfhydryl alkylating reagents. Although NEM has been widely used in the past to inactivate AAA ATPase family members and found to inhibit the GTP-dependent assembly of liver microsomes (14), it is membrane permeant and would be expected to inactivate sulhydryl groups required for the assembly of H and L chains in the lumen of the ER. We therefore examined the effects of a number of membrane-impermeant analogs of NEM including sulfo-MBS, sulfo-SMPB, and sulfo-SMCC on the ability of membranes or cytosol to promote fusion (Fig. 3A). Following treatment for 15 min on ice, the reagents were inactivated by the addition of excess glutathione, and the treated membranes or cytosol were subsequently incubated in the presence of ATP for 15 min at 32°C. Whereas treatment of cytosol had little effect on ER assembly (Fig. 3A, lanes b and c), pretreatment of microsomes with each of the reagents completely inhibited the appearance of assembled H 2 L 2 (Fig. 3A, lanes  d-g). As expected, a similar effect was observed in the presence of detergent (not shown) due to alkylation of the sensitive sulfhydryl groups required for H and L chain oligmerization. To avoid the possibility that ER microsomes were potentially leaky to these membrane impermeant regents, we also analyzed the effect of a large bulky thiol-blocking reagent, BSA-PDP synthesized by conjugating the bifunctional reagent Nsuccinimidyl 3-(2-pyridyldithio)propionate to BSA (14,42). Previous studies using fluorescent quenching as a measured of fusion of rat liver ER microsomes have shown that this reagent inhibits lipid bilayer mixing (14). H and/or L chain-containing microsomes were pretreated with BSA-PDP for 15 min on ice. Following neutralization of unreacted PDP groups with excess glutathione, membranes were incubated for 15 min at 32°C. Treatment of either the H or L chain containing membranes alone was sufficient to inactivate fusion of ER membranes when mixed with the untreated partner (Fig. 3A, lanes h-j). No effect was observed with unmodified BSA (not shown). Attempts to reactivate the fusion of ER membranes pretreated with either the membrane impermeant NEM analogs or BSA-PDP using purified NSF or p97, two AAA ATPases previously implicated in yeast ER and mammalian Golgi reassembly (6,7,17), were unsuccessful (not shown).
To assess the temporal sensitivity of the assay to Rab activation or sulfhydryl-blocking reagents, membranes were incubated for increasing time at 32°C. At the indicated time (Fig.  3B, ⌬t), membranes were transferred to ice and either retained on ice (Fig. 3B, closed squares) or treated with GDI (Fig. 3B, open squares), GTP␥S (Fig. 3B, open circles), or BSA-PDP (Fig.  3B, closed circles) and reincubated at 32°C for a total time of 15 min. The addition of BSA-PDP (Fig. 3B, closed circles) inhibited the assembly of H 2 L 2 only when added within the first 2-5 min of incubation at 32°C, confirming that H 2 L 2 assembly is inaccessible to the bulky thiol-containing reagent. Similar results were observed with membrane impermeant NEM analogs (not 2 W. E. Balch, unpublished results.

FIG. 2. ER fusion is mediated by a Rab GTPase.
A, microsomes were preincubated on ice for 30 min with the indicated concentration of Rab-GDI and subsequently incubated for 15 min at 32°C in the presence of ATP and cytosol. Inset, microsomes were incubated in the presence of ATP and cytosol at 32°C for 15 min with the indicated addition. a, no addition; b, Sar1-GDP (5 g); c, Sar1-GTP (5 g); d, ARF1-GDP (10 g); e, ARF1-GTP (10 g); f, Rab1A N124I (7.5 g); g, 50 M final GTP␥S; h, 1 mM final GDP␤S. Recombinant mutant GTPases were prepared as described (24,29,32). H 2 L 2 assembly was quantitated as described under "Experimental Procedures." B, mutant GDIs defective in Rab binding do not inhibit ER assembly. The indicated mutants were prepared as described (39) and added to the assay at a final concentration of 5 M. Membranes were incubated for 15 min at 32°C in the presence of cytosol and ATP, and H 2 L 2 assembly was quantitated as described under "Experimental Procedures." shown). Although the temporal sensitivity to GTP␥S yielded a similar result to that of sulfhydryl blocking reagents (Fig. 3B, open circles), H 2 L 2 assembly remained sensitive to GDI throughout the entire time course (Fig. 3B, open squares)). The addition of GDI at any time point abruptly blocked ER fusion, similar to the effect of transferring cells to ice (Fig. 3B, closed  squares). Thus, the requirement for Rab is a transient event, occurring immediately prior to membrane fusion. DISCUSSION We have developed a convenient ELISA assay to measure homotypic ER fusion based on the unique protein folding environment of the ER (12). Fusion of H and L chain containing microsomes requires a factor sensitive to sulfhydryl blocking reagents, as has been observed previously in other assays that measure ER assembly using fluorescent lipid probes (11,13,14). Consistent with the observation that NSF cannot reverse the GTP-dependent fusion of microsomes prepared from rat liver (14), we have been unable to reverse NEM-inhibited ER fusion by either purified NSF or p97, proteins that are required for the reassembly of NEM-treated Golgi membranes (6,7). The latter reagent (p97/yeast Cdc48p) is involved in the assembly of yeast ER fragments (17). It would appear that the fusion of mammalian ER microsomes may be mediated by a novel member of the AAA ATPase family. Alternatively, the inactivated factor(s) may remain associated with a docking/fusion complex(es), functioning as a dominant inhibitor.
The principle focus of our study was to examine the hypothesis that a Rab GTPase may mediate ER fusion. Previous observations using GTP and/or GTP analogs have implicated the involvement of a GTPase(s) in the homotypic assembly of the mammalian ER membranes (12,15,16). We eliminated the possibility that Sar1, ARF1, and Rab1 are involved in ER fusion as trans-dominant inhibitory forms of these proteins, which inhibit ER to Glogi transport, had no effect on ER fusion in vitro. The inability of the GTP-restricted forms of either the Sar1 or ARF1 GTPases to inhibit fusion eliminates the possibility that the inhibition observed with GTP␥S is somehow related to the activation of endogenous Sar1 or ARF1 leading to the stable coating of ER membranes with either COPII or COPI coats, respectively (30,32,43). However, we did find that Rab-GDI, but not Rho-GDI, had potent effects on homotypic fusion and that this inhibition was specific, because GDI mutants defective in Rab binding were not inhibitory. These results demonstrate for the first time the involvement of a member of the rab gene family in ER assembly. A requirement for Rab in ER fusion is paralleled by the need for Ypt7p in the homotypic fusion of vacuolar membranes in yeast (8) and Rab5 in the homotypic fusion of early endosomes in vitro (26,44). Curiously, the homotypic assembly of Golgi cisternae has been reported to be insensitive to both Rab-GDI and GTP␥S (6,7). Likewise, the homotypic fusion of yeast ER membranes is GTP␥S-insensitive (11). These results are at odds with a large body of data that suggests that most if not all cellular fusion events involve Rab (reviewed in Ref. 21). We suggest that the Rab GTPases likely to be involved in each case remain to be detected.
We found that the timing of Rab activation is coupled to the formation of a pore that provides luminal continuity between H and L chain-containing membranes. This is because H 2 L 2 assembly occurs immediately upon fusion (12), and GDI, which only recognizes the GDP-bound form of Rab (reviewed in Refs. 34 and 35), was able to inactivate fusion at early and late time points. Our results are similar to the late requirement for Ypt7p in the homotypic fusion of vacuolar membranes (8). Although GDI blocked a late step coincident with fusion, the temporal effects of GTP␥S mimicked that of sulhydryl reagents, which only blocked an early membrane priming step. If the endogenous Rab responsible for ER fusion is also a target for GTP␥S, these results suggest that although Rab can become activated during priming, premature stable activation by the nonhydrolyzable analog is inhibitory. Consistent with this conclusion, the sensitivity to GDI throughout the time course of incubation suggests that functional activation from the GDPto the GTP-bound form occurs immediately prior to or coincident with fusion. These results are, in part, similar to the observation that Rab5 continuously cycles between GDP-and GTP-bound forms prior to membrane interaction (45). This cycling has been proposed to serve as a timer to kinetically proofread membrane fusion events (45,46). In ER assembly, a similar timer function may be in effect. Therefore, an ordered reaction involving a protein belonging to the AAA ATPase gene family during the priming of membranes for fusion followed by a transient activation of Rab at the fusion site may be a common feature of the homotypic fusion of both endocytic and exocytic compartments.  (14)). Following the addition of 2 mM glutathione to neutralize excess reagent, treated and untreated membranes/cytosol were combined and incubated for 15 min at 32°C in the presence of ATP, and H 2 L 2 assembly was quantitated by ELISA. B, microsomes incubated in the presence of ATP and cytosol at 32°C for the indicated time (⌬t) were transferred to ice and either held on ice (f) or supplemented with 2 M GDI (Ⅺ), 50 M GTP␥S (E), or BSA-PDP (1 mM PDP) (q) (final concentrations). After 30 min on ice, GDI treated samples were transferred to 32°C and incubated for a total time of 15 min. After 15 min on ice, BSA-PDP of GTP␥S treated samples were supplemented with either 2 mM glutathione (BSA-PDP-treated samples) or 10 mM GTP (GTP␥S-treated samples) to neutralize excess reagent, transferred to 32°C, and incubated for a total time of 15 min.