The Mon1–Ccz1 GEF activates the Rab7 GTPase Ypt7 via a longin-fold–Rab interface and association with PI3P-positive membranes

Within the endomembrane system, vesicles transport cargo between organelles. These organelles need to maintain their identity despite the fact that they continuously fuse with incoming vesicles and lose membrane as a result of</emph> vesicle generation. Rab GTPases, tethering factors and the SNARE machinery localize to specific organelles and cooperate to promote fusion of membranes. Rab proteins are small GTP/GDP-binding proteins with C-terminal prenyl anchors, which function as molecular switches. They bind in their GTP form to tethering factors to bring membranes into proximity before SNAREs present in each membrane form four-helix bundles and thus drive bilayer mixing.

As a result of the slow rate of dissociation of GDP, Rabs require guanine nucleotide exchange factors (GEFs) for their conversion into the active GTP-bound form, which can interact with effectors on membranes such as tethering factors or lipid kinases. In addition, Rabs depend on a GTPase activating protein (GAP) to hydrolyze the GTP to GDP, and then become substrates of the GDI chaperone, which can extract the Rab-GDP and keep it soluble in the cytoplasm (Lachmann et al., 2011; Hutagalung and Novick 2011; Itzen and Goody 2011; Barr, 2013).

Among the Rab interactors, GEFs are most crucial because they provide the activated Rab at a specific location in the cell. In this context, structures of Rab–GEF complexes and the subsequent analyses have been very informative. For the exocytic Sec4 Rab, the GEF Sec2 promotes nucleotide exchange via its α-helical domain (Dong et al., 2007; Sato et al., 2007). Rab5-like proteins that operate in the endosomal pathway depend on Vps9-domain-containing GEFs (Delprato et al., 2004; Delprato and Lambright 2007). DENN domains were identified as a signature domain for multiple GEFs (Yoshimura et al., 2010; Allaire et al., 2010; Marat et al., 2011) and molecular details of their function were revealed for DENND1-BS and Rab35 (Wu et al., 2011).

GEFs might also act as part of a multiprotein complex. The best-characterized example is the heptameric TRAPPI, which interacts with the Rab1/Ypt1 protein through a surface consisting of four subunits: Bet5, Trs23 and the two flanking copies of Bet3 (Kümmel et al., 2006; Kim et al., 2006; Cai et al., 2008). GEF dimers are required to activate Rab6/Ypt6 (Ric1–Rgp1) (Pusapati et al., 2012; Siniossoglou et al., 2000) and Rab7/Ypt7 (Mon1–Ccz1) (Nordmann et al., 2010; Gerondopoulos et al., 2012), although structural information on these complexes is lacking. In agreement with the function of the Mon1–Ccz1 dimer in Ypt7 activation, endosomal biogenesis is strongly impaired if either of the two proteins is defective (Wang et al., 2002; Kinchen and Ravichandran 2010; Poteryaev et al., 2010; Nordmann et al., 2010; Yousefian et al., 2013).

TRAPP subunits Bet5 and Trs23 interact both with each other and with Ypt1 through their longin domains (Kim et al., 2006; Cai et al., 2008). These are arranged into a continuous ten-strand β-sheet supporting two parallel α-helices, which provide a significant portion of the interacting residues. Interestingly, the formation of the Mon1–Ccz1 heterodimer also requires the predicted longin domains of both subunits (Nordmann et al., 2010). This suggests that the architecture and mode of GEF action of Mon1 and Ccz1 on Ypt7 might resemble those of the TRAPPI complex. Here, we present evidence that Mon1 and Ccz1, which are linked to Rab7/Ypt7 both in mammals and yeast, cooperate via their longin-domain surface to activate the Rab GTPase. This activity is strongly promoted upon membrane recruitment of the GEF, suggesting that previously determined activities might have been significantly underestimated.

Secondary structure prediction using PredictProtein (http://predictprotein.org) and PsiPred (http://bioinf.cs.ucl.ac.uk/psipred/) identified possible longin domains in Ccz1 (aa 1–156) and Mon1 (aa 180–287), which are required for the assembly of the Mon1–Ccz1 heterodimer (Nordmann et al., 2010). This suggests a similar structure–function relationship for Mon1–Ccz1 and the TRAPPI GEF complex, because the two TRAPP subunits Trs23 and Bet5 also contain longin domains that contact the Rab GTPase Ypt1 directly (Kim et al., 2006; Cai et al., 2008). We therefore built a homology model for the predicted longin domains of Mon1–Ccz1 using Trs23 and Bet5 as templates.

Longin domains consist of a central antiparallel β-sheet (S1–S5) with one α-helix (H1) on one side and two α-helices (H2, H3) on the opposing side. Trs23 and Bet5 interact edge to edge through the central β-sheets of their longin domains and contact Ypt1 primarily through residues residing in the H1 helices of both subunits (Fig. 1A). Consequently, mutants in each H1 helix result in loss of GEF activity towards Ypt1 (Cai et al., 2008) (Fig. 1B). In addition, Ypt1 is recognized by loop S1–S2 and strand S2 of Trs23 (Fig. 1B) and some residues provided by the two copies of Bet3 (Cai et al., 2008). We used the Genesilico Metaserver (Kurowski et al., 2003) to assign Ccz1 (aa 1–156) to Bet5 and Mon1 (aa 180–287) to Trs23, and built a homology model for Ypt7/Mon1 and Ccz1 longin domains based on the Ypt1–TRAPP complex (see Materials and Methods for details). The resulting model showed only few clashes with satisfactory stereochemistry (Fig. 1C).

On the basis of this model and the previously characterized TRAPP mutants, we introduced mutations in Mon1 and Ccz1 that should interfere with GEF activity (Fig. 1D). Notably, for TRAPPI and DENND1B-S, double and triple mutants had to be generated in order to abolish GEF activity completely (Cai et al., 2008; Wu et al., 2011). We therefore used a similar approach here, and first followed vacuole morphology of Mon1 and Ccz1 mutants. Deletion of Mon1 or Ccz1 resulted in multiple small vacuoles (Fig. 2A–C). This defect was rescued by expression of Mon1 and Ccz1, either from a normal promoter or when overproduced (Fig. 2A,C). For Mon1, we identified mutations in helix H1 (G209W, T213K) and the neighboring S1–S2 region (G191P, K192D) that resulted in vacuolar defects (Fig. 2A). Defects in the single G191P mutant were only observed if expression of Mon1 was lowered (Fig. 2C). Ccz1 mutants displayed similar defects if mutated in the H1 region (G47W, G51M) (Fig. 2B). These defects were not due to mislocalization, because GFP-tagged Mon1 and Ccz1 mutants were still found in dot-like structures, which probably represent endosomes, similar to the wild-type proteins (Fig. 2D,E). To test whether any of the mutations interfered with the Rab interaction, we relied on a yeast three-hybrid assay, where both Mon1 and Ccz1 were co-expressed with Ypt7. This assay clearly demonstrated interaction of Mon1 and Ccz1 with the GDP-locked version of Ypt7 (T22N), whereas none of the mutants of Mon1 or Ccz1 did (Fig. 2F). In agreement with this, the exclusive vacuolar localization of Ypt7 was lost in both mutants, as observed for the GEF deletion strains (Fig. 3A) (Cabrera and Ungermann, 2013). We next asked whether the observed defect was due to complex assembly. However, neither mutation affected the formation of the dimer and we could purify Mon1–Ccz1 complexes with similar stoichiometries as observed for the wild-type complex using the established tandem-affinity purification protocol (Fig. 3B) (Nordmann et al., 2010). We then asked whether the purified mutant complexes were still functional and used the vacuole fusion assay to measure their GEF activity (Fig. 3C). In this assay, fusion of vacuoles from two different tester strains allows the protease present in one vacuole population to process immature alkaline phosphatase, which is found in the other vacuole population (Wickner, 2002). To test for GEF activity, we preincubated vacuoles with a non-specific GAP to generate Ypt7-GDP and then used Mon1–Ccz1 to reactivate Ypt7 (Nordmann et al., 2010). Whereas the wild-type Mon1–Ccz1 complex readily restored fusion, neither mutant was able to counteract GAP activity (Fig. 3C). In sum, mutations in either Mon1 or Ccz1, which had been introduced based on our homology model, maintained the GEF complex and its localization, but abolished both Ypt7 interaction and GEF activity.

Having established that Mon1 and Ccz1 cooperate in activation of Rab7/Ypt7, we asked how activity of the complex might be regulated. Previous analyses of the C. elegans Mon1 protein SAND-1 indicated that it binds phosphoinositide-3-phosphate (PI3P) (Poteryaev et al., 2010), even though depletion of PI3P did not affect Mon1–Ccz1 localization in Drosophila (Yousefian et al., 2013). We used single Mon1 and Ccz1 proteins isolated from yeast deleted for the other binding partner, and tested their binding specificity to nitrocellulose strips coated with different lipid species (Fig. 4A). When blots were probed for Mon1 or Ccz1, we observed strong interactions with PI3P, PI5P and phosphatidylserine (PS). Interestingly, the human Mon1a protein had the same preference for PI3P and PS, indicating conserved membrane preferences of Mon1–Ccz1 across species. To confirm binding of yeast Mon1–Ccz1 to membranes with negatively charged lipids, we generated liposomes with vacuolar lipid composition with or without PI3P, incubated them with Mon1–Ccz1 complex and separated the lipid-bound fraction from the unbound by flotation (Cabrera et al., 2010). Surprisingly, Mon1–Ccz1 was mainly recovered in the liposome fraction (Fig. 4B), regardless of the presence of PI3P in the bilayer. This indicates that either Mon1–Ccz1 does not show a lipid preference or that our methodology is not sensitive enough to reveal possible differences. We thus used reflectance interference (Rif), which is a label-free detection method that determines protein binding to solid-supported membranes (Fig. 4C) (Gavutis et al., 2005). Using this experimental strategy, we observed more efficient binding of Mon1–Ccz1 to membranes containing PS and PI3P than to PC membranes (Fig. 4D, table on right). This indicates that Mon1–Ccz1 has an inherent affinity to membranes that is strengthened by negatively charged phospholipids.

Although GEFs operate on membranes to promote both activation and localization of Rab, their activities are usually determined in solution. We wondered whether the activity of Mon1–Ccz1 might have been underestimated in our previous measurements (Nordmann et al., 2010). To mimic at least part of the Rab activation cascade, we immobilized Ypt7 tagged with C-terminal hexahistidine on vesicles containing DOGS-NTA lipids (1,2-dioleoyl-sn-glycero-3-{[N-(5-amino-1-carboxypentyl) iminodiacetic acid]succinyl}) (Sot et al., 2013), and then measured GEF activity. Unlike results obtained in solution, where modest Mon1–Ccz1 activity was observed (Nordmann et al., 2010), binding of Ypt7 to vesicles resulted in dramatically enhanced activity (Fig. 5A). We then reduced the concentration of Mon1–Ccz1, and estimated an ∼1600-fold increase in GEF activity (Fig. 5B,I). Of note, the determined GEF activity of Mon1–Ccz1 without membrane-anchored Ypt7 is very similar to our previous approximation (Nordmann et al., 2010). This increase in GEF activity was dependent on the membrane localization of Ypt7 because vesicles without DOGS-NTA had no effect (Fig. 5C). As expected, the Mon1 longin mutant (G209W, T213K) and Rab5-GEF Vps9 were inactive in the GEF assay (Fig. 5D). We then asked whether the increase in membrane binding as observed before (Fig. 4D), correlated with GEF activity. We therefore generated DOGS-NTA-containing vesicles without or with PS and PI3P in different concentrations and determined GEF activity. Whereas activity was clearly observed if Ypt7 was immobilized on PC-containing membranes, it was increased twofold by the addition of negatively charged phospholipids (Fig. 5E,I). To discriminate whether PS and PI3P only contribute to the membrane recruitment of the GEF complex, we used a His-tagged version of the complex that can associate with membranes independently of these lipids. Interestingly, the presence of PS and PI3P resulted in additional stimulation of the GEF activity, which might reflect an additional role of these lipids in Mon1–Ccz1 activation (Fig. 5F).

The HOPS complex binds activated Ypt7 on the late endosome and vacuole, and might act in a positive-feedback loop to promote Mon1–Ccz1 activity. Indeed, the Vps39 subunit of the HOPS complex interacts with Mon1 (Wang et al., 2003; Nordmann et al., 2010), and additional interactions of metazoan Mon1 with HOPS subunits were reported (Poteryaev et al., 2010). We therefore localized GFP-tagged Mon1 or Ccz1 in cells overexpressing the entire HOPS complex. Under these conditions, Mon1 and Ccz1 accumulated on vacuolar membranes, presumably as a result of the reported interactions of the GEF complex with HOPS (Fig. 5G). We then asked if the interaction also affects activity of the GEF complex, and therefore added purified HOPS (approximately twofold excess) to the GEF assay (Fig. 5H). Even though we observed a slight increase in activity, we consider the effect rather minimal. In sum, the GEF activity of Mon1–Ccz1 is strongly increased on membranes, but is not affected by the HOPS complex.

In this study, we have addressed the molecular function of the Mon1–Ccz1 complex. Mon1–Ccz1 activity requires residues within the predicted longin domains of both subunits, which we could identify based on a homology model with the TRAPPI complex. Consistent with this, Mon1 and Ccz1 longin mutants still form a heterodimer and localize like the wild-type complex, but cannot promote nucleotide exchange in Ypt7. Using a modified GEF assay, we further show that Mon1–Ccz1 is strongly active once it is recruited to the membranes, suggesting that GEF activities might have been underestimated previously.

The presence of longin domains in Mon1–Ccz1 and TRAPPI, and the similar phenotype of the mutants suggest that both GEFs might share a related Rab-binding surface and nucleotide exchange mechanism. Also, the recently resolved DENND1 GEF domain contains a longin fold, although it binds Rab35 via a different surface compared with the interaction between TRAPPI and Ypt1 (Wu et al., 2011). We are aware that within the TRAPPI complex, the two Bet3 subunits contribute to the binding of Ypt1 and the C-terminal residues of Bet3A have a crucial role to promote GEF activity (Cai et al., 2008). We thus consider it likely that additional segments proximal to the longin domains of Mon1–Ccz1 will be required for full activity. We also note that our model is limited by the structural differences between Ypt1 and Ypt7, especially in their N-terminal regions. However, at this point, our working model allows us to predict essential residues for GEF activity (Figs 1,2).

In this context, we realized that roadblock domains that are found in subunits of the EGO/ragulator complex have been predicted as GEFs for the Rag GTPases (BarPeled et al., 2013). Even though the order of secondary structure elements differs between roadblock and longin domains, their overall structure is quite similar (Kurzbauer et al., 2004; Qian et al., 2005; Zhang et al., 2012). These domains have probably evolved as general interactors of small GTPases (Levine et al., 2013), and a subfamily might act as dimeric GEFs. Importantly, the dimeric BLOC-3 complex also harbors a longin domain in each of its two subunits Hps1 and Hps4 and might function in a very similar manner to Mon1–Ccz1 (Gerondopoulos et al., 2012). It thus is conceivable that these domains have evolved as general interactors of small GTPases. Intriguingly, the dimeric Ric1–Rgp1 complex, which activates Ypt6 (Siniossoglou et al., 2000), also depends on both subunits for activity, and might also use a shared surface to interact with its Rab Ypt6/Rab6.

Even though we have shown GEF activity of Mon1–Ccz1 previously in vitro, we did not expect a 1600-fold increase in activity (k_cat/K_M) upon immobilization of Ypt7 on membranes. This might be due in part to the membrane recruitment of both proteins. Although the assay does not recapitulate the entire cycle, it revealed that Mon1–Ccz1 activity is far higher than previously anticipated. GEFs such as the Legionella DrrA protein have strong activity in vitro, which is sufficient to counteract Rab extraction by GDI (Schöbel et al., 2009; Zhu et al., 2010; Suh et al., 2010). It is thus likely that other GEFs use similar mechanisms and will explain how GEFs that are mistargeted to the mitochondria can drive the recruitment of Rabs to this organelle (Gerondopoulos et al., 2012; Blümer et al., 2013), even though some targeting information might reside in additional proteins and the Rab itself (Cabrera and Ungermann, 2013). Our adjustment in the GEF assay now provides insight into the activity of Mon1–Ccz1. We demonstrated that acidic phospholipids such as PS and PI3P clearly enhance the inherent ability of the yeast and human complex to bind membranes, in agreement with earlier findings on C. elegans Mon1 (Poteryaev et al., 2010). Interestingly, in Drosophila, the Mon1–Ccz1 complex localizes even in the absence of PI3P to membranes (Yousefian et al., 2013). We consider it likely that acidic phospholipids also facilitate the Rab-GEF interaction by optimally positioning the GEF complex relative to the Rab. We believe that similar assays with other GEFs will also uncover stronger GEF activities. A more extreme case was observed for Rasal, a RasGAP that displays activity only when it is recruited with its Ras substrate to the same membrane (Sot et al., 2013). Because GEFs function only on membranes in vivo, we consider our approach more suitable to mimic the in vivo situation on organelles.

One aspect that we could not clarify was the timing of Mon1–Ccz1 recruitment. It is possible that the complex is always active on endosomal membranes or it might be activated only once endosomes have matured, and thus can contribute to the exchange of Rab5 for Rab7 (Rink et al., 2005; Poteryaev et al., 2010). With respect to yeast, it is puzzling that the amounts of Ypt7 on endosomes are rather low, even though we observe Mon1–Ccz1 mostly on endosomes (Nordmann et al., 2010). This suggests that the complex might be further regulated in vivo. At this point, our data indicate that Mon1–Ccz1 acts primarily at endosomes and its interaction with HOPS might create an additional GEF microdomain on vacuoles. Within our experimental set-up, Mon1–Ccz1 activity was not strongly affected by HOPS, and Ypt7 localization was not disturbed in the absence of Vps39, which binds to Mon1 in vitro (Nordmann et al., 2010). We expect that the further dissection of the role of other endosomal factors that might function in a Rab cascade will help us to reveal how Mon1–Ccz1 function is controlled in vivo. In summary, our data reveal that the Mon1–Ccz1 GEF complex activates Ypt7 via a common interface, has highest activity on membranes and acts independently of the HOPS complex.

Strains and plasmids used in this study are listed in supplementary material Tables S1 and S2, respectively. Deletions and tagging of genes were done using homologous recombination of PCR fragments (Janke et al., 2004; Puig et al., 1998). mCherry-tagged Ypt7 was expressed from plasmid pRS414-TPIpr-mCherry-V5, which was generated from plasmid pRS415-TPIpr-mCherry-V5-ATG8 (a gift from Fulvio Reggiori, University Medical Center Utrecht, The Netherlands). Mon1 and Ccz1 mutants were generated using the QuikChange site-directed mutagenesis kit from Stratagene (La Jolla, CA). The catalytic domain of Gyp1 (249–637) was purified from a BL21 strain containing plasmid pET22-Gyp1-46 (Nordmann et al., 2010). Protein expression was induced by addition of 0.5 mM isopropyl-d-thiogalactoside overnight at 20°C. Purification was performed using the nickel nitrilotriacetic acid resin (Qiagen, Hilden, Germany) and elution with 300 mM imidazole.

YEP352-Mon1, pACT2-Ccz1 and pFBT9-Ypt7 plasmids containing wild-type and mutant versions of Mon1, Ccz1 and Ypt7 were co-transformed into PJ69-4A strain and plated on minimal medium lacking leucine, tryptophan and uracil. Transformants were patched first on plates lacking leucine, tryptophan and uracil and afterwards on plates lacking leucine, tryptophan, uracil, histidine and adenine. Four clones were analyzed for each combination and one is shown.

The Genesilico metaserver (https://genesilico.pl/meta2) (Kuroski and Bujnicki 2003) was used to identify remotely related proteins of known structure (by allowing for much more freedom in sequence alignments). Secondary structure predictions were carried out with either PredictProtein or PsiPred (http://predictprotein.org, http://bioinf.cs.ucl.ac.uk/psipred/). Both published structures of the TRAPP complex were identified (PDB IDs 3CUE 2J3T; protein data bank codes from www.pdb.org). 3CUEc and 2J3Tc (chain c of yeast and mouse Bet5 respectively) were identified for the first 150 residues of Ccz1. The server also reports Bet5 structures for the Mon1 fragment (180–287) but with a lower score. The template for modeling consisted of Ypt7-GDP (PDB ID 1KY3) (Constantinescu et al., 2002) and mammalian Bet5 and Trs23 (PDB ID 2J3T) and was obtained by 3D alignment with the respective chains of Ypt1/TRAPP structure (PDB ID 3CUE). A PDZL domain, which is absent in yeast Trs23 was removed from the template. The complex formed by Ypt7 GDP and the longin domains of Mon1 and Ccz1 was modeled using YASARA 13.4.21 (Krieger et al., 2002). The resulting model resembled the Ypt1–Trs23–Bet5 structure and showed only few clashes in the contacting residues. The model scored as optimal or satisfactory considering the correctness of backbone (Ramachandran plot) and side-chain dihedrals, as well as packing interactions. Note that the model does not include Ypt7 residues (38–40 and 67–76), which are missing in the reported structure.

Yeast cells were grown to mid-log phase in YPD, YPG or synthetic complete (SDC) medium lacking selected amino acids or nucleotides, collected by centrifugation, washed once with SDC or SGC medium supplemented with all amino acids, and immediately analyzed by fluorescence microscopy. For FM4-64 staining of vacuoles, cells were incubated with 30 µM FM4-64 for 30 minutes, washed twice with YPD medium, and incubated in the same medium without dye for 1 hour. Images were acquired with a Leica DM5500 B microscope equipped with a SPOT Pursuit camera equipped with an internal filter wheel (D460sp, BP460-515 and D580lp; Leica Microsystems GmbH), fluorescence filters [49002 ET-GFP (FITC/Cy2): Exc. ET470/40x, Em. ET525/50m; Wide Green: Exc. D535/50, Em. E590lp; 49008 ET-mCherry, Texas Red: Exc. ET560/40x, Em. ET630/75m Chroma Technology Corp.], and Metamorph 7 software (Visitron Systems, Munich, Germany). Images were processed using ImageJ 1.42 (National Institutes of Health) and Autoquant x v1.3.3 (Media Cybernetics, Inc.).

Tandem affinity purification was performed as described (Nordmann et al., 2010; Ostrowicz et al., 2010; Bröcker et al., 2012). Three liters of culture were grown at 30°C to OD₆₀₀ ∼4 and cells were harvested by centrifugation. Cells were lysed in buffer containing 50 mM HEPES-NaOH, pH 7.4, 300 mM NaCl, 1.5 mM MgCl₂, 1× FY protease inhibitor mix (Serva, Heidelberg, Germany), 0.5 mM PMSF and 1 mM DTT. Lysates were centrifuged for 90 minutes at 100,000 g, and supernatants were incubated with IgG Sepharose beads for 2 hours at 4°C. Beads were isolated by centrifugation at 800 g for 5 minutes, and washed with 15 ml lysis buffer containing 0.5 mM DTT. Bound proteins were eluted by TEV cleavage, and analyzed on SDS-PAGE. For Mon1–Ccz1, buffer exchange via NAP5 columns to the fusion reaction buffer was performed. For HOPS, buffer exchange was omitted.

Vacuoles were isolated from yeast strains BJ3505 (pep4Δ) and DKY6281 (pho8Δ) via ficoll gradient centrifugation as described (Cabrera and Ungermann 2008). Fusion reactions containing 3 µg of each vacuole type were incubated in fusion reaction buffer (1 mM PIPES-KOH, pH 6.8, 20 mM sorbitol, 5 mM MgCl₂, 125 mM KCl), 10 µM CoA, 0.01 µg His-Sec18, 1 mM GTP, with or without ATP-regenerating system (0.5 mM ATP, 40 mM creatine phosphate, 0.1 mg/ml creatine kinase) for 90 minutes at 27°C and developed for 5 minutes. Where indicated, Mon1–Ccz1 or recombinant Gyp1-46 was added. Fusion values correspond to the average of 2–4 samples.

Single Mon1 and Ccz1 molecules were purified from strains lacking the respective binding partner. Proteins were incubated with the lipid strips (Invitrogen) that had been blocked with TBS-Tween and 3% fatty acid free BSA for 1 hour at room temperature and protein binding detected by immunoblotting. Purification of human Mon1a was done as described (Gerondopoulos et al., 2012). Blots were blocked with TBS-Tween and 3% fatty acid free BSA as described by the manufacturer (Invitrogen). Incubation with human Mon1–Ccz1 was carried out at 4°C overnight.

Lipids were purchased from Avanti Polar Lipids, except ergosterol (Sigma) and phosphatidylinositol-3-phosphate (PI3P) (Mobitech/Echelon). A vacuolar lipid mixture (Zinser et al., 1991; Mima et al., 2008) containing (mol %) di-oleoyl-phosphatidylcholine (DOPC; 47%), di-oleoyl-phosphatidyl-ethanolamine (DOPE; 18%), soy phosphatidyl-inositol (PI; 18%), di-oleoyl-phosphatidyl-serine (DOPS; 4.4%), di-oleoyl-phosphatidic acid (DOPA; 2%), cardiolipin (1.6%), ergosterol (8%), diacylglycerol (DAG; 1%) and NBD-PE (1%) was prepared by evaporation, and resuspended in HEPES-KOAc (HK) buffer (50 mM HEPES-KOH, pH 7.2 and 120 mM KOAc). Where indicated, phosphatidylinositol-3-phosphate (PI3P, 1%) was added. After five steps of thawing and freezing in liquid nitrogen, the liposome suspension (2 mM) was extruded through polycarbonate filters of 400, 200, 100 nm pore size using a hand extruder (Avanti). The liposome size was determined by dynamic light scattering (DynaPro Titan, Wyatt Technology Europe GmbH, Dernbach, Germany). To obtain the vesicles used in the GEF assays, the lipid mixture was prepared by evaporation of the desired lipids, resuspended in HEPES-NaCl buffer containing 20 mM HEPES-NaOH, pH 7.4, 150 mM NaCl, and 1 mM MgCl₂ to a final concentration of 3 mM, and subjected to nine steps of thawing and freezing in liquid nitrogen.

The assay was performed as described (Cabrera et al., 2010). In brief, purified Mon1–Ccz1 complex (200 nM) was incubated with liposomes (0.75 mM) in 150 µl HKM buffer (HK buffer with 1 mM MgCl₂) at room temperature for 5 minutes. 100 µl of a 75% sucrose solution in HKM buffer was added to adjust the sucrose concentration to 30%. The suspension was overlaid with two layers (200 µl of 25% sucrose solution in HKM and 50 µl HKM buffer). Gradients were centrifuged at 220,000 g for 1 hour at 20°C. Proteins from the bottom (250 µl) and top (50 µl) fractions were trichloroacetic acid (TCA)-precipitated, analyzed by SDS-PAGE and Sypro Orange staining (Invitrogen). Protein binding was detected using VersaDoc imaging system (Bio-Rad GmbH, Munich).

Binding of Mon1–Ccz1 to artificial membranes was monitored in real time by reflectance interference (RIf) detection in a flow system as in principle described previously (Gavutis et al., 2005). Briefly, RIf allows label-free detection of protein binding to the surface of a glass substrate coated with a thin silica layer. All experiments were carried with HEPES-NaCl buffer (20 mM HEPES, pH 7.5, 150 mM NaCl) as running buffer. Continuous membranes were obtained by fusing small unilamellar lipid vesicles composed of POPC only, 90% POPC and 10% POPS or 90% POPC and 10% PI3P, onto clean transducer slides. Prior to each measurement the system and membrane were washed by successive injections of imidazole (500 mM, 250 µl, 86 seconds) and EDTA (250 mM, 250 µl, 86 seconds). Subsequently, 100 nM of the Mon1–Ccz1 complex was injected in a volume of 250 µl for 185 seconds, and its dissociation was monitored for 150 seconds by rinsing with HBS. A final washing step with imidazole removed all the bound protein from the membrane to recover it for the next measurement and served as proof of an intact bilayer.

GEF assays were performed as described (Nordmann et al., 2010). Briefly, 500 pmoles of the Rab Ypt7-His were preloaded with MANT-GDP, and incubated with 160 µl vesicles for 10 minutes at 25°C before incubation with different amounts of Mon1–Ccz1 complex. MANT fluorescence was recorded for the indicated time in a fluorimeter with a temperature-controlled cuvette and a stirring device (Jasco, Gross-Umstadt, Germany). Samples were excited at 366 nm and fluorescence was detected at 443 nm. After 200 seconds, GTP was added to final concentrations of 0.1 mM to trigger the exchange reaction. The decrease of MANT-GDP fluorescence is used as read-out of nucleotide release.

We would like to thank Daniel Kümmel for feedback on the manuscript, and all members of the Ungermann lab for critical suggestions.