The role of the hypervariable C-terminal domain in RabGTPases membrane targetingFu Lia,b, Long Yia, Lei Zhaoa,b, Aymelt Itzena,c, Roger S. Goodya, and Yao-Wen Wua,b,1

aDepartment of Physical Biochemistry, Max Planck Institute of Molecular Physiology, 44227 Dortmund, Germany; bChemical Genomics Centre of the MaxPlanck Society, 44227 Dortmund, Germany; and cCenter for Integrated Protein Science Munich, Department Chemie, Technische Universität München, 85748Garching, Germany

Edited by Gregory A. Petsko, Weill Cornell Medical College, New York, NY, and approved January 15, 2014 (received for review July 18, 2013)

Intracellular membrane trafficking requires correct and specificlocalization of Rab GTPases. The hypervariable C-terminal domain(HVD) of Rabs is posttranslationally modified by isoprenyl moietiesthat enable membrane association. A model asserting HVD-directed targeting has been contested in previous studies, but therole of the Rab HVD and the mechanism of Rab membranetargeting remain elusive. To elucidate the function of the HVD, wehave substituted this region with an unnatural polyethylenglycol(PEG) linker by using oxime ligation. The PEGylated Rab proteinsundergo normal prenylation, underlining the unique ability of theRab prenylation machinery to process the Rab family with diverseC-terminal sequences. Through localization studies and functionalanalyses of semisynthetic PEGylated Rab1, Rab5, Rab7, and Rab35proteins, we demonstrate that the role of the HVD of Rabs in mem-brane targeting is more complex than previously understood. TheHVD of Rab1 and Rab5 is dispensable for membrane targeting andappears to function simply as a linker between the GTPase domainand the membrane. The N-terminal residues of the Rab7 HVD areimportant for late endosomal/lysosomal localization, apparently dueto their involvement in interaction with the Rab7 effector Rab-interacting lysosomal protein. The C-terminal polybasic cluster ofthe Rab35 HVD is essential for plasma membrane (PM) targeting,presumably because of the electrostatic interaction with negativelycharged lipids on the PM. Our findings suggest that Rab membranetargeting is dictated by a complex mechanism involving GEFs, GAPs,effectors, and C-terminal interaction with membranes to varyingextents, and possibly other binding partners.

chemical protein modification | synthetic protein probe | RILP

Rab proteins are key regulators of intracellular vesicle trans-port in eukaryotic cells (1, 2). They comprise the largest

subgroup of the Ras superfamily of small GTPases, with morethan 60 members in humans and 11 members in yeast (3).Interacting with a complex network of Rab regulators andeffectors, Rab GTPases regulate these processes through a spa-tiotemporally controlled GTPase cycle and their distribution incells. The GTPase cycle is strictly regulated by guanine nucleo-tide exchange factors (GEFs) that mediate GDP/GTP exchangeand by GTPase-activating proteins (GAPs) that accelerate thehydrolysis of GTP. Active (GTP-bound) Rab proteins associatewith distinct intracellular compartments and direct vesiculartransport by recruiting a multitude of Rab-specific effectors,including tethering complexes and motor proteins.Rab proteins are posttranslationally modified at the C termi-

nus with prenyl groups that function as membrane anchors. Rabprenylation involves covalent attachment of the geranylgeranyl(C-20 isoprenyl) moiety to one or two C-terminal cysteine resi-dues of the protein substrate via a stable thioether linkage (4).Unlike other protein prenyltransferases that recognize the C-terminal CaaX motif of protein substrates (e.g., Ras and Rho),Rab geranylgeranyl transferase (RabGGTase) does not recog-nize its protein substrates (Rab proteins) directly but requiresthe adaptor Rab escort protein (REP). Rab prenylation requiresthe formation of a ternary catalytic Rab:REP:RabGGTase com-plex (5–7). It remains elusive how the single Rab prenylation

machinery can process the whole Rab family with diverseC termini.Cycling between the cytosol and membranes is an essential

feature of the mode of action of Rabs and is made possible byreversible interaction with GDP dissociation inhibitor (GDI),which can solubilize the otherwise water-insoluble geranylger-anylated Rab molecules (8). Membrane-bound GDI displace-ment factors were proposed to disrupt GDI:Rab complexes,leading to insertion of the prenylated Rab into the membrane inthe GDP form and release of GDI into the cytosol (9, 10). Oneof the most perplexing questions is how Rab proteins are spe-cifically targeted to their cognate membranes. The hypervariableC-terminal domain (HVD) was proposed to function as a signalfor targeting Rab proteins to specific subcellular membranes(11). However, later studies suggested that several features ofRab molecules, Rab effectors, and GEFs are involved in thetargeting process (12–15). The role of the Rab HVD in mem-brane targeting is controversial, because contradictory resultswere obtained by swapping the hypervariable domains of Rabproteins (11–13). Thus, the function of the Rab HVD anda complete model for Rab membrane targeting remain to beestablished. To further understand the significance of the HVDin Rab membrane targeting and prenylation, unique methods areneeded to manipulate the structure of Rab C terminus.In this study, we replaced the Rab C-terminal sequence with

the polyethylenglycol (PEG) linker as a nonpeptidic chain. ThePEG chain containing two cysteine residues or a simple thiolgroup at one end was coupled to truncated Rab proteins by ox-ime ligation. We found that the PEGylated Rab proteins undergonormal prenylation in vitro, confirming that the Rab prenylationmachinery does not require a specific C-terminal sequence but

Significance

Rab GTPases are key regulators of intracellular vesicular trans-port in eukaryotic cells, for example in neurotransmission andendocytosis. Each of more than 60 Rabs in human cells regulatesthese processes at a specific subcellular membrane. The hyper-variable C-terminal domain (HVD) of Rab has been proposed tocontain a targeting signal, but recent studies have argued thismodel. However, the role of the Rab HVD and the mechanism ofRab membrane targeting remain elusive. In this study, usinga combination of unique synthetic protein probes and cell bi-ological tools, we dissect the role of the HVD for subcellular Rabtargeting. We find that the HVD can act as a structurally flexiblespacer, a subcellular localization determinant, and/or an en-hancer of membrane affinity.

Author contributions: Y.-W.W. designed research; F.L. and L.Y. performed research; L.Y.and L.Z. contributed new reagents/analytic tools; F.L. and Y.-W.W. analyzed data; andA.I., R.S.G., and Y.-W.W. wrote the paper.

The authors declare no conflict of interest.

This article is a PNAS Direct Submission.1To whom correspondence should be addressed. E-mail: yaowen.wu@mpi-dortmund.mpg.de.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1313655111/-/DCSupplemental.

2572–2577 | PNAS | February 18, 2014 | vol. 111 | no. 7 www.pnas.org/cgi/doi/10.1073/pnas.1313655111

rather outsources the specificity to the REP molecule. By com-bining this semisynthetic strategy with cell imaging, we elucidatethe role of the hypervariable C-terminal domain for subcellular Rabtargeting. In some instances, the HVD is dispensable for correctsubcellular localization (Rab1, Rab5), but is essential in other casesbecause of specific interactions with effectors (Rab7) or electro-static interactions with membranes (Rab35). The results furtherelaborate the model for Rab prenylation and membrane targeting.

ResultsConstruction of PEGylated Rab Protein Probes. To further un-derstand the function of the Rab HVD in membrane targetingand prenylation, other methods are needed to manipulate thestructure of the Rab C terminus. Advances in protein chemicalmethods allow for site-specific modification of proteins (16, 17).Recently, we reported a facile method for protein C-terminalmodification through intein-mediated incorporation of a (bis)oxyamine moiety into the C terminus of proteins, which is thenavailable for subsequent conjugation under mild conditions witha fluorophore containing a ketone functionality (18). We choseto equip Rab1, Rab5, Rab7, and Rab35, which localize on theGolgi apparatus, early endosome, late endosome/lysosome, andplasma membrane, respectively, with synthetic nonnative C ter-mini and analyze the effect on subcellular localization. Based onsequence alignments, a dibasic repeat (KK) that is conserved inRab1, Rab5, and Rab35 sequences was defined as the hyper-variable domain junction (11, 12, 19).Initially, we investigated to what extent the Rab prenylation

machinery could tolerate changes in the Rab C-terminal structure.The hypervariable domain is largely unstructured. Only N-termi-nal residues of the hypervariable domain are visible in the crystalstructure of Rab:GDI and Rab:REP complexes, where a hy-drophobic interaction of the C-terminal interacting motif (CIM)within the hypervariable domain with REP and GDI is observed(6, 20). Because the CIM is important for binding to REP andGDI, we made two kinds of structural changes in the HVDflanking the CIM (7). The fragment C-terminal to the CIM wasreplaced with a PEG linker containing prenylatable cysteines ora thiol group using the oxime ligation strategy. The upstreamfragment was replaced by flexible GGS repeats (Fig. 1). The(GGS)n-peptide linker has a random-coil conformation and hasbeen used as a flexible peptide linker in proteins (21). In theconstruct referred to as Rab-PEG-CC, the C-terminal sequencedownstream of the CIM was replaced by a PEG-CC moiety. InRab-(GGS)n-PEG-CC, the whole hypervariable domain was sub-stituted by a GGS repeat and a PEG-CC moiety (Fig. 1).To this end, truncated Rab7-thioester proteins were produced

via intein-mediated protein splicing by using 2-mercaptoetha-nesulfonate (MESNA) as the thiol reagent. Subsequently, Rab7-thioester proteins were treated with (bis)oxyamine at pH 7.5 onice for 4 h, converting them to oxyamine conjugates, Rab7Δ-ONH2, which are competent for oxime ligation with a compoundcontaining a ketone moiety (Fig. 1 and SI Appendix, Fig. S3A).Keto-PEG-C(StBu)C(StBu) was synthesized by stepwise solid-

phase (the synthetic procedure is described in SI Appendix). In-cubation of Keto-PEG-C(StBu)C(StBu) and Rab7Δ15-ONH2 onice overnight led to ca. 90% conversion of the oxyamine to anoxime adduct, as shown by electrospray ionization (ESI)-MS (SIAppendix, Fig. S3B). Subsequently, the thiol protecting groupsStBu were removed by treatment with 500 mM MESNA to yieldRab7-PEG-CC. A dimeric (Keto-PEG-SH)2 is connected bya disulfide bond that can be easily cleaved by using 5 mMDithioerythritol (DTE) during protein conjugation. Ligation ofKeto-PEG-SH with Rab7Δ15-ONH2 yielded Rab7-PEG-SH (SIAppendix, Fig. S3C).

PEGylated Rab Proteins Undergo Prenylation in Vitro. The obtainedPEGylated Rab7 proteins were tested in an in vitro prenylationassay by using NBD-FPP, a fluorescent analog of the lipid sub-strate GGPP (22). Rab7-PEG-CC and Rab7-(GGS)n-PEG-CCprotein conjugates display identical prenylation efficiency to that

of the wild-type Rab7 protein (Fig. 2). In keeping with NBD-FPPprenylation results, prenylation of Rab7-PEG-CC and Rab7-(GGS)n-PEG-CC using GGPP leads to the doubly geranylger-anylated protein, as shown by mass spectrometry (SI Appendix,Fig. S4). Next, we further examined whether intact cysteineresidues are indispensable for prenylation. A Rab7-PEG-SHprotein conjugate with two cysteines being replaced by a simplethiol group was subjected to in vitro prenylation (Fig. 1). Rab7-PEG-SH undergoes prenylation with an observed rate that is ca.twofold higher than that of wild-type protein (Fig. 2 and SIAppendix, Fig. S4). These findings suggest that Rabs do not re-quire a specific sequence in the HVD for prenylation, except forthe CIM and an SH group as an isoprenyl acceptor. Instead, theamino acids downstream of the CIM can be replaced by a non-peptidic PEG linker containing cysteines or thiol groups withoutperturbing Rab prenylation in vitro.To gain insights into the role of the Rab C terminus in mem-

brane targeting, we prepared PEGylated EGFP-Rab1, EGFP-Rab5, EGFP-Rab7, and EGFP-Rab35 proteins by ligating Keto-PEG-CC to the truncated EGFP-Rab and EGFP-Rab-(GGS)nproteins, as done for the PEGylated Rab7 conjugates (Fig. 1).These protein conjugates were subjected to in vitro prenylation.As shown in SI Appendix, Fig. S5, all PEGylated Rab proteinsundergo prenylation in vitro.

The HVD Sequence of Rab1 and Rab5 Is Not Required for CorrectSubcellular Localization and Function. Next, we microinjected thePEGylated fluorescent Rab proteins into HeLa or MDCK cellsexpressing Cherry-Rab wild-type proteins to examine their sub-cellular localizations. All of the PEGylated Rab proteins witha dicysteine motif localized to membranes, suggesting that

Fig. 1. Modification of Rab HVD. (A) Semisynthesis of Rab-PEG-CC and Rab-PEG-SH and comparison of the C-terminal structures with that of wild-typeprotein. (B) Substitution of the Rab HVD with different C-terminal structures.The CIM and prenylatable cysteines are highlighted.

Li et al. PNAS | February 18, 2014 | vol. 111 | no. 7 | 2573

CELL

BIOLO

GY

PEGylated Rab proteins are prenylated in the cell (Fig. 3 and SIAppendix, Fig. S6). Both Rab1-PEG-CC and Rab1-(GGS)n-PEG-CC colocalize with the wild-type Rab1 on the Golgi ap-paratus (Fig. 3 A and B). To demonstrate that the PEGylatedRab proteins are functional, we prepared Rab5Q79L mutantsthat are deficient in GTPase activity and are constitutively in theactive GTP-bound state. Constitutively active Rab5 promotesthe formation of enlarged endosomes, in line with its functionin homotypic endosome fusion (23). Likewise, microinjectionof Rab5Q79L-PEG-CC or Rab5Q79L-(GGS)n-PEG-CC proteininto the cell leads to the formation of enlarged vesicles (Fig. 3H).Moreover, both of them colocalized with the Rab5Q79L mutanton the enlarged early endosomes when they were introduced intocells expressing Cherry-Rab5Q79L (Fig. 3 C and D). Theseresults suggest that PEGylated Rab1 and Rab5 proteins arecorrectly targeted and functional, indicating that the hyper-variable domain of Rab1 and Rab5 is not essential for theirmembrane targeting and function.To further confirm our conclusions, we substituted the hyper-

variable domain of Rab1 and Rab5 with a (GGS)n-VKL-(GGS)n-CSC or (GGS)n-VKL-(GGS)n-CC fragment, referred to as RabΔ-CSC or RabΔ-CC, respectively (SI Appendix, Fig. S7). In thesechimeric proteins only the CIM and the prenylation motif wereretained while the rest of the HVD was replaced with flexibleGGS repeats. To examine the effects of carboxyl methylation onRab protein localization, two kinds of prenylation motifs, CXCand CC, were included. After prenylation, Rab proteins ending inCXC undergo carboxyl methylation by isoprenylcysteine carboxy-methyltransferase (Icmt) at the ER membrane, whereas thosehaving CC and CCXX sequences in their C termini are not car-boxymethylated (24–26). Analogous scenarios were observed inthe cells expressing Rab1 and Rab5 chimeric proteins, i.e., correctmembrane localization and induction of the enlarged earlyendosomes by Rab5Q79LΔ35-CSC or Rab5Q79LΔ35-CC (SIAppendix, Fig. S7).Interestingly, Rab protein conjugates (Rab1, Rab7, Rab5, and

Rab5Q79L) with the replacement of the C-terminal residuesdownstream from the CIM by a PEG-SH moiety are largely cy-tosolic and not functional in the cell (Fig. 1 and SI Appendix, Fig.S8), although they can be prenylated in vitro (SI Appendix, Fig.S5C). These results suggest that digeranylgeranylation rather thanthe sequence of the prenylation motif is essential for the correctsubcellular localization and function of Rab proteins, in keepingwith a previous report on Rab5 membrane localization (27).

Binding to Rab-Interacting Lysosomal Protein Is Essential for Rab7Membrane Targeting. Rab7-PEG-CC with the C-terminal se-quence downstream from the CIM replaced by a PEG-CCmoiety colocalizes with wild-type Rab7 at late endosomes orlysosomes in the perinuclear region (Fig. 3 E and K). How-ever, Rab7-(GGS)n-PEG-CC with the substitution of the whole

hypervariable domain is located differently from wild-type Rab7protein (Fig. 3 F and K). We found that Rab7-(GGS)n-PEG-CCcolocalizes with the Golgi marker, Giantin (Fig. 3 I and K). Toconfirm its Golgi localization, cells were treated with nocoda-zole, which disrupts the microtubule network, leading to theformation of Golgi fragments that are distributed throughout thecell (28). Rab7-(GGS)n-PEG-CC was found to colocalize ex-tensively with Giantin on the Golgi fragments in the nocodazole-treated cells (SI Appendix, Fig. S9 A–C). These results suggest

Fig. 2. Reaction kinetics of prenylation from the SDS/PAGE assay using thefluorescent analog of GGPP, NBD-FPP. The solid lines show single exponen-tial fits for Rab7 wild type (kobs = 0.021 min−1), Rab7-PEG-CC (kobs = 0.031min−1), Rab7-(GGS)n-PEG-CC (kobs = 0.018 min−1), and Rab7-PEG-SH (kobs =0.050 min−1).

Fig. 3. Subcellular localization of EGFP-Rab-PEG-CC and EGFP-Rab-(GGS)n-PEG-CC proteins. EGFP-Rab1-PEG-CC (A) and EGFP-Rab1-(GGS)n-PEG-CC (B)colocalize with mCherry-Rab1 wild-type protein at the Golgi apparatus.EGFP-Rab5Q79L-PEG-CC (C) and EGFP-Rab5Q79L-(GGS)n-PEG-CC (D) colo-calize with mCherry-Rab5Q79L on enlarged endosomes. EGFP-Rab7-PEG-CC(E) colocalize with mCherry-Rab7 wild-type protein at the late endosome/lysosome. EGFP-Rab7-(GGS)n-PEG-CC does not colocalize with mCherry-Rab7wild-type protein (F) but colocalizes with mKate2-Giantin at the Golgi ap-paratus (I). (G) EGFP-Rab35-PEG-CC localizes in the perinuclear region butnot the plasma membrane as shown for the ECFP-Rab35 wild-type protein inMDCK cells. (J) EGFP-Rab35-PEG-CC colocalizes with mKate2-Giantin at theGolgi apparatus. (H) EGFP-Rab5Q79L-PEG-CC and EGFP-Rab5Q79L-(GGS)n-PEG-CC induce formation of enlarged vesicles. (K) Pearson’s colocalizationcoefficient analysis of chimeric Rab proteins with wild-type Rab proteins orthe Golgi marker. Measurements were performed in HeLa cells except whereotherwise indicated. (Scale bars: 10 μm.) In the measurements shown in A, B,I, and J, cells were fixed and subject to extraction of cytosolic proteins bysaponin.

2574 | www.pnas.org/cgi/doi/10.1073/pnas.1313655111 Li et al.

that the HVD of Rab7 upstream from the CIM is essential formembrane targeting, and mutation in this region leads to mis-targeting to the Golgi apparatus.To map the region that is essential for Rab7 targeting, we

substituted the Rab7 HVD partially or completely with (GGS)n-VKL-(GGS)n-CSC or (GGS)n-VKL-(GGS)n-CC fragments. Thechimeric Rab7 proteins were expressed in HeLa cells, and theirsubcellular localization was examined. Rab7Δ24-CSC colocalizeswith the wild-type protein, whereas partial colocalization and nosignificant colocalization but rather mislocalization to the Golgiapparatus were observed in Rab7Δ27-CSC and Rab7Δ34-CSC/CC proteins, respectively (Fig. 4 A–C and SI Appendix, Fig. S10),suggesting that residues (174–183) of the HVD are involved inRab7 membrane targeting. These residues are known to be in-volved in binding to the Rab7 effector, Rab-interacting lyso-somal protein (RILP). In particular, V180, L182, and Y183 areinvolved in hydrophobic interactions with RILP (29). Binding ofRab7 full-length, Rab7Δ24-CSC, Rab7Δ27-CSC, and Rab7Δ34-

CSC to RILP declines sequentially as shown in the pull-downassay (Fig. 4 K–L). No detectable binding is observed betweenRab7Δ34-CSC and RILP. To exclude the possibility that C-terminal truncation might affect nucleotide exchange, we examinedthe GTP/GDP ratio of Rab7 and Rab7Q67L (GTPase deficient)proteins in HeLa cells. We found that the truncated Rab7 chi-mera proteins undergo normal nucleotide exchange in the cell.(SI Appendix, Fig. S11). Taken together, these results suggestthat residues of Rab7 HVD, particularly amino acids 174–183,which are indispensable for the Rab7–RILP interaction, are alsoessential for correct targeting. Therefore, the interaction withRILP appears to be essential for Rab7 membrane targeting.Indeed, colocalization of RILP with Rab7 wild-type, Rab7Δ24-CSC, Rab7Δ27-CSC, and Rab7Δ34-CSC/CC proteins in theclustered late endosomes/lysosomes decreases sequentially (Fig.4 E–G and J), in line with the Rab7WT colocalization and theRILP pull-down results (Fig. 4I).To further investigate the role of RILP in Rab7 membrane

targeting, we knocked down RILP in HeLa cells by siRNA (Fig.5H). Rab7 wild type and Rab7-PEG-CC are largely cytosolic inRILP knock-down cells (Fig. 5 D, E, and I and SI Appendix, Fig.S12). The subcellular localization of these proteins is normal incontrol siRNA cells (Fig. 5 A, B, and I). In contrast, the mem-brane localization of Rab7-(GGS)n-PEG-CC and Rab7Δ34-CSC, which do not bind to RILP and are mistargeted to theGolgi apparatus, is not affected by RILP knock-down (Fig. 5 Fand G). These results demonstrate that RILP plays an essentialrole in Rab7 membrane targeting.

The C-Terminal Polybasic Cluster Is Important for Rab35 Localizationto the Plasma Membrane. Rab35-PEG-CC, with its C-terminalpolybasic cluster substituted by the PEG linker, does not localizeto the plasma membrane (PM) but mistargets to the Golgi ap-paratus (Fig. 3 G and J and SI Appendix, Fig. S6). Replacementof the Rab35 HVD with a (GGS)n-VKL-(GGS)n-CSC or (GGS)n-VKL-(GGS)n-CC fragment also leads to mistargeting to the Golgiapparatus (SI Appendix, Fig. S7 E and F). To confirm its Golgilocalization, cells were treated with nocodazole and Rab35-PEG-CC was found to colocalize extensively with Giantin on the Golgifragments in the nocodazole-treated cells (SI Appendix, Fig. S7 D–F). These results demonstrate that the polybasic region is essentialfor the PM targeting and/or PM membrane affinity, presumablybecause of interaction with negatively charged phosphatidylino-sitol phosphate lipids (30). It is not clear whether Rab35 initiallytargets to the Golgi apparatus and subsequently redistributes tothe PM.

DiscussionIn the present study, we have combined synthetic chemistry,bioorthogonal chemistry, and protein engineering to introduceunnatural C-terminal fragments into Rab proteins. This methodresolves inherent problems of the traditional HVD swappingapproach in the analysis of Rab membrane targeting. Becausethe HVD may be only a partial determinant for membrane tar-geting in some Rabs, investigation of chimeric Rab proteins maylead to ambiguous results (11–13).In vitro prenylation analysis of the PEGylated Rab proteins

probes provides insights into the mechanism of Rab proteinprenylation. This study together with previous reports suggeststhat the specificity of the Rab prenylation machinery is conferredby three binding interfaces, involving the GTPase domain, theCIM, REP, and RabGGTase (5, 6). The HVD of Rabs, with theexception of the CIM, does not contribute specificity to the as-sembly of the ternary protein complex. Once the ternary proteincomplex is established, the Rab C terminus is concentratedwithin the microenvironment of the complex, thus enhancing theprobability of C-terminal cysteines reaching the active site ofRabGGTase. As a consequence, the protein substrate specificityof RabGGTase does not need to be encoded in the Rab C ter-minus, in contrast to that of CaaX protein prenyltransferases.Rab proteins therefore outsource the specificity to REP. The

Fig. 4. Subcellular localization of EGFP-Rab7Δ-CSC proteins. The HVD ofRab7 is partly or completely substituted. The HVD is highlighted in cyan, theCIM is highlighted in yellow, and prenylatable cysteines are in red. EGFP-Rab7 wild-type protein and EGFP-Rab7Δ-CSC proteins are coexpressed withmCherry-Rab7 (A–D) or mCherry-RILP (E–H) in HeLa cells. The arrowheadsindicate Rab7 vesicles that do not contain Rab7Δ27-CSC, whereas the arrowsindicate Rab7Δ27-CSC vesicles that do not contain Rab7. (Scale bars: 10 μm.)Pearson’s correlation coefficients (PCC) for colocalization with Rab7 (I) andRILP (J). (K) Binding of Rab7Δ-CSC proteins with RILP. HeLa cells weretransfected with full-length and truncated (C-terminally replaced) HA-Rab7Q67L (constitutively active) proteins. Proteins were pulled down fromcell lysate with MBP-RILP or MBP (control)-coated beads. Beads were sub-jected to SDS/PAGE and Western blotting with anti-HA antibody (Upper).The protein level of cell lysate, MBP-RILP and MBP on the beads was ex-amined by Coomassie blue staining (Lower). C, control; L, cell lysate; P, pull-down. (L) Quantitation of Rab7 proteins bound to RILP as shown in K.Measurements were performed in triplicate.

Li et al. PNAS | February 18, 2014 | vol. 111 | no. 7 | 2575

CELL

BIOLO

GY

model is consistent with the fact that RabGGTase has essentiallyno sequence preference for the context of the prenylatable cys-teines, and the C-terminal sequences occurring in Rab GTPasesinclude CC, CXC, CCX, CCXX, CCXXX, and CXXX. Hence,any cysteine- or thiol-containing fragment that can be properlypresented to the active site of RabGGTase is able to undergoprenylation. This property allows RabGGTase to process allmembers of the family of more than 60 Rab proteins withhypervariable C-terminal sequences, a feature that is uncommonin protein-modifying enzymes. This unique property of the Rabprenylation machinery also enables it to process Rab proteinswith unnatural C-terminal moieties.Analysis of subcellular localization and function of the

PEGylated Rab proteins probes facilitates elucidation of Rabmembrane targeting and the role of the hypervariable domain inthis process. Our findings suggest that the HVDs of individualRab proteins play distinct roles in membrane targeting. In thecase of Rab1 and Rab5, the HVD is not required for membranetargeting, probably because it is not involved in binding withpotential targeting factors, including effectors and GEFs (12, 31–33). It serves rather as an anchoring chain that physically con-nects the functional GTPase domain with the membrane. TheHVD can be substituted by a nonnative PEG linker of sufficientlength. The digeranylgeranyl lipid anchor is important for mem-brane association, whereas the sequence of the prenylation motifis not essential for correct membrane targeting. However, nativelymonogeranylgeranylated Rabs (e.g., Rab8, Rab13, Rab18, Rab23,and Rab28) obviously do not require diprenylation for appropriatetargeting to subcellular membranes.In contrast, some of the N-terminal residues of the Rab7 HVD

are involved in binding with the Rab7 effector RILP, and theresults presented here suggest that this interaction stabilizesRab7 association with late endosomes/lysosomes. Because resi-dues of the Rab7 HVD are one of the elements involved in theinteraction with RILP (binding to the GTPase domain is alsorequired), a Rab chimeric protein with the Rab7 HVD wouldbind to RILP less efficiently (29). In keeping with this result,replacing the hypervariable domain of Rab5 with that of Rab7only led to partial mislocalization from early endosomes (13).

Moreover, knockdown of RILP renders Rab7 largely cytosolic incells. Therefore, RILP appears to be a targeting factor for Rab7and determines the steady-state distribution of Rab7 on sub-cellular compartments. Evidence has been presented that theRab9 effector TIP47 might interact with the Rab9 HVD, and thisinteraction has been implicated in Rab9 localization (12).The C-terminal polybasic cluster of Rab35 HVD appears to

play a key role for PM targeting, probably due to the electrostaticinteraction with negatively charged lipids on the PM (30). How-ever, we cannot rule out the possibility that other unidentifiedRab35 effectors that associate with the C terminus of Rab35 mayplay a role in enhancing its interaction with the PM.It is not clear how and why those Rabs lacking in affinity to

effectors or lipids are mistargeted to the Golgi apparatus. Theremight be two hypotheses for the role of the Golgi membranes asthe initial site for Rab attachment or as the default localization ofmislocalized Rabs. In the former case, newly synthesized preny-lated Rabs would be delivered to the Golgi membranes and besubsequently sorted to the designated compartments. Loss of tar-geting elements would lead to accumulation of Rabs in the Golgiapparatus. In the latter case, Rabs would initially target to theirsubcellular compartments and be transported to the Golgi appa-ratus via GDI recycling and/or vesicular transport, because Rabscould not stably associate with their cognate membranes. However,the question then arises as to the origin of the specificity of Golgilocalization. Further work is required to elaborate this question.From the results presented here together with evidence from

earlier studies, it appears that Rab membrane targeting is gov-erned by complex mechanisms, with the involvement of Rabregulators and binding partners including GEFs, GAPs, andeffectors. GEF-mediated nucleotide exchange provides the ther-modynamic driving force for Rab membrane insertion, which isindispensable for the stable attachment of Rabs to membranes(15, 34, 35). GEFs and GAPs appear to regulate Rab localiza-tion, because the nucleotide-bound state defines Rab membraneassociation (15). Indeed, emerging evidence suggests that Rabsare relayed on the trafficking pathway through GEF and GAPcascades, which determine the boundary between Rab proteins(36–40). In such cascades, Rab A recruits the GEF that targetsRab B along the pathway to the membrane, and Rab A is sub-sequently inactivated by a GAP recruited through Rab B and,hence, is detached from the membrane. As a consequence,conversion of a Rab A to a Rab B membrane is achieved. OnceRab proteins are activated and stabilized on the membrane(GTP-bound), the steady-state Rab localization might be dic-tated by binding partners, including effectors and lipids (Fig. 6).Such a stabilization of Rabs on their cognate membranes byeffectors appears to play an essential role in Rab membranetargeting, because depletion of effectors (RILP knockdown) orloss of binding to effectors (Rab HVD replacement) leads tomislocalization of Rab proteins. Some effectors associate witha specific membrane independently of Rabs (41). For thoseeffectors that are recruited to membranes in a Rab-dependent

Fig. 5. Effect of RILP RNAi on Rab7 localization. (A–C) Subcellular locali-zation of EGFP-Rab7 WT (A), EGFP-Rab7-PEG-CC (B), and EGFP-Rab7-(GGS)n-PEG-CC (C) in control siRNA cells. (D–G) Subcellular localization of EGFP-Rab7WT (D), EGFP-Rab7-PEG-CC (E), EGFP-Rab7Δ34-CSC (F), and EGFP-Rab7-(GGS)n-PEG-CC (G) in RILP siRNA cells. Endogens RILP was detected by im-munofluorescence. (H) Knockdown of endogenous RILP in HeLa cells,detected by Western blot using anti-RILP antibody. Scrambled siRNA (scr)cells were used as a control. (I) Quantification of membrane localization ofRab7 WT and Rab7-PEG-CC in siRNA or control cells by counting Rab7-posi-tive vesicles in the cell. ***P < 0.001. (Scale bars: 10 μm.)

Fig. 6. Model for Rab membrane targeting. Initial insertion of Rab proteinsinto membranes is driven by GEF-mediated nucleotide exchange. Binding ofactivated Rab proteins (GTP-bound) with effectors or other binding partnersdetermines the steady-state distribution of Rab proteins on membranes.GAPs deactivate Rab proteins at specific sites and trigger the recycling of Rabproteins from membranes. The GTPase cycle controlled by GEFs and GAPsdictates the thermodynamic equilibrium of Rab membrane localization anddefines the boundary of a Rab realm.

2576 | www.pnas.org/cgi/doi/10.1073/pnas.1313655111 Li et al.

manner (2), there must be a synergy between Rabs and effectorsin membrane localization. For example, Rabenosyn-5 has twodistinct binding sites for Rab4 and Rab5, suggesting its co-ordinative role in Rab4 and Rab5 localization and function (42).

MethodsPreparation of PEGylated Rab Proteins. Keto-PEG-C(StBu)C(StBu) and (Keto-PEG-SH)2 were synthesized on solid phase and liquid phase, respectively.Purification of protein thioesters, subsequent oxime ligation with PEGmoieties, quality analysis by LC-ESI-MS, and in vitro prenylation assay wereperformed as described (18, 22) (for details, see SI Appendix).

Cell Culture and Transfection. HeLa or MDCK cells were maintained at 37 °Cand 5% CO2 in high glucose minimum essential medium (MEM) (21969-035;Invitrogen) supplemented with 10% (vol/vol) FBS, 1% sodium pyruvate, 1%non-essential amino acids (NEAA), and 1% L-Glutamine, and MDCK cells inMEM were supplemented with 10% FBS. For confocal microscopy, 2.0 × 105

cells were cultured on 35-mm glass bottom dishes (MatTek) for 12 h beforetransfection. Transient plasmid expression was achieved by overnighttransfection with X-treme GENE HP DNA transfection reagent (06366244001;Roche). For nocodazole treatment, cells were treated with 20 μM nocoda-zole for 1.5 h. For the details of cell fixation and immunofluorescence, RNAi,pull down assay, and determination of GTP/GDP ratio, see SI Appendix.

Quantitative Analysis of Colocalization. Confocal micrographs of various pro-teins and microinjected proteins were selected randomly. After the backgroundsubtractionusing theBGSubtraction fromROIplugin of ImageJ software (version1.44; National Institutes of Health), Pearson’s correlation coefficient (PCC) valuesfor the relation between the EGFP signals and Cherry signals were then calcu-lated with the Intensity Correlation Analysis plug-in of ImageJ software.

ACKNOWLEDGMENTS. This work was supported in part by DeutscheForschungsgemeinschaft Grants SPP 1623 (to Y.-W.W.) and SFB 642 (to R.S.G.and Y.-W.W.).

1. Hutagalung AH, Novick PJ (2011) Role of Rab GTPases in membrane traffic and cellphysiology. Physiol Rev 91(1):119–149.

2. Stenmark H (2009) Rab GTPases as coordinators of vesicle traffic. Nat Rev Mol Cell Biol10(8):513–525.

3. Pereira-Leal JB, Seabra MC (2001) Evolution of the Rab family of small GTP-bindingproteins. J Mol Biol 313(4):889–901.

4. Casey PJ, Seabra MC (1996) Protein prenyltransferases. J Biol Chem 271(10):5289–5292.5. Pylypenko O, et al. (2003) Structure of Rab escort protein-1 in complex with Rab

geranylgeranyltransferase. Mol Cell 11(2):483–494.6. Rak A, et al. (2004) Structure of the Rab7:REP-1 complex: Insights into the mechanism

of Rab prenylation and choroideremia disease. Cell 117(6):749–760.7. Wu YW, Goody RS, Abagyan R, Alexandrov K (2009) Structure of the disordered C

terminus of Rab7 GTPase induced by binding to the Rab geranylgeranyl transferasecatalytic complex reveals the mechanism of Rab prenylation. J Biol Chem 284(19):13185–13192.

8. Wu YW, Tan KT, Waldmann H, Goody RS, Alexandrov K (2007) Interaction analysis ofprenylated Rab GTPase with Rab escort protein and GDP dissociation inhibitor ex-plains the need for both regulators. Proc Natl Acad Sci USA 104(30):12294–12299.

9. Dirac-Svejstrup AB, Sumizawa T, Pfeffer SR (1997) Identification of a GDI displacementfactor that releases endosomal Rab GTPases from Rab-GDI. EMBO J 16(3):465–472.

10. Sivars U, Aivazian D, Pfeffer SR (2003) Yip3 catalyses the dissociation of endosomalRab-GDI complexes. Nature 425(6960):856–859.

11. Chavrier P, et al. (1991) Hypervariable C-terminal domain of rab proteins acts asa targeting signal. Nature 353(6346):769–772.

12. Aivazian D, Serrano RL, Pfeffer S (2006) TIP47 is a key effector for Rab9 localization.J Cell Biol 173(6):917–926.

13. Ali BR, Wasmeier C, Lamoreux L, Strom M, Seabra MC (2004) Multiple regions con-tribute to membrane targeting of Rab GTPases. J Cell Sci 117(Pt 26):6401–6412.

14. Beranger F, Paterson H, Powers S, de Gunzburg J, Hancock JF (1994) The effectordomain of Rab6, plus a highly hydrophobic C terminus, is required for Golgi appa-ratus localization. Mol Cell Biol 14(1):744–758.

15. Wu YW, et al. (2010) Membrane targeting mechanism of Rab GTPases elucidated bysemisynthetic protein probes. Nat Chem Biol 6(7):534–540.

16. Hackenberger CP, Schwarzer D (2008) Chemoselective ligation and modificationstrategies for peptides and proteins. Angew Chem Int Ed Engl 47(52):10030–10074.

17. Vila-Perelló M, Muir TW (2010) Biological applications of protein splicing. Cell 143(2):191–200.

18. Yi L, et al. (2011) One-pot dual-labeling of a protein by two chemoselective reactions.Angew Chem Int Ed Engl 50(36):8287–8290.

19. Pereira-Leal JB, Seabra MC (2000) The mammalian Rab family of small GTPases:Definition of family and subfamily sequence motifs suggests a mechanism for func-tional specificity in the Ras superfamily. J Mol Biol 301(4):1077–1087.

20. Rak A, et al. (2003) Structure of Rab GDP-dissociation inhibitor in complex withprenylated YPT1 GTPase. Science 302(5645):646–650.

21. Evers TH, van Dongen EM, Faesen AC, Meijer EW, Merkx M (2006) Quantitative un-derstanding of the energy transfer between fluorescent proteins connected viaflexible peptide linkers. Biochemistry 45(44):13183–13192.

22. Wu YW, et al. (2006) A protein fluorescence amplifier: Continuous fluorometric assay

for rab geranylgeranyltransferase. ChemBioChem 7(12):1859–1861.23. Stenmark H, et al. (1994) Inhibition of rab5 GTPase activity stimulates membrane

fusion in endocytosis. EMBO J 13(6):1287–1296.24. Bergo MO, et al. (2001) Isoprenylcysteine carboxyl methyltransferase deficiency in

mice. J Biol Chem 276(8):5841–5845.25. Li G, Stahl PD (1993) Post-translational processing and membrane association of the

two early endosome-associated rab GTP-binding proteins (rab4 and rab5). Arch Bio-

chem Biophys 304(2):471–478.26. Smeland TE, Seabra MC, Goldstein JL, Brown MS (1994) Geranylgeranylated Rab

proteins terminating in Cys-Ala-Cys, but not Cys-Cys, are carboxyl-methylated by

bovine brain membranes in vitro. Proc Natl Acad Sci USA 91(22):10712–10716.27. Gomes AQ, et al. (2003) Membrane targeting of Rab GTPases is influenced by the

prenylation motif. Mol Biol Cell 14(5):1882–1899.28. Lippincott-Schwartz J (1998) Cytoskeletal proteins and Golgi dynamics. Curr Opin Cell

Biol 10(1):52–59.29. Wu M, Wang T, Loh E, Hong W, Song H (2005) Structural basis for recruitment of RILP

by small GTPase Rab7. EMBO J 24(8):1491–1501.30. Heo WD, et al. (2006) PI(3,4,5)P3 and PI(4,5)P2 lipids target proteins with polybasic

clusters to the plasma membrane. Science 314(5804):1458–1461.31. Cai Y, et al. (2008) The structural basis for activation of the Rab Ypt1p by the TRAPP

membrane-tethering complexes. Cell 133(7):1202–1213.32. Delprato A, Lambright DG (2007) Structural basis for Rab GTPase activation by VPS9

domain exchange factors. Nat Struct Mol Biol 14(5):406–412.33. Mishra A, Eathiraj S, Corvera S, Lambright DG (2010) Structural basis for Rab GTPase

recognition and endosome tethering by the C2H2 zinc finger of Early Endosomal

Autoantigen 1 (EEA1). Proc Natl Acad Sci USA 107(24):10866–10871.34. Tarafder AK, et al. (2011) Rab27a targeting to melanosomes requires nucleotide ex-

change but not effector binding. Traffic 12(8):1056–1066.35. Blümer J, et al. (2013) RabGEFs are a major determinant for specific Rab membrane

targeting. J Cell Biol 200(3):287–300.36. Kinchen JM, Ravichandran KS (2010) Identification of two evolutionarily conserved

genes regulating processing of engulfed apoptotic cells. Nature 464(7289):778–782.37. Nordmann M, et al. (2010) The Mon1-Ccz1 complex is the GEF of the late endosomal

Rab7 homolog Ypt7. Curr Biol 20(18):1654–1659.38. Poteryaev D, Datta S, Ackema K, Zerial M, Spang A (2010) Identification of the switch

in early-to-late endosome transition. Cell 141(3):497–508.39. Rivera-Molina FE, Novick PJ (2009) A Rab GAP cascade defines the boundary between two

Rab GTPases on the secretory pathway. Proc Natl Acad Sci USA 106(34):14408–14413.40. Zhu H, Liang Z, Li G (2009) Rabex-5 is a Rab22 effector and mediates a Rab22-Rab5

signaling cascade in endocytosis. Mol Biol Cell 20(22):4720–4729.41. Houghton FJ, Chew PL, Lodeho S, Goud B, Gleeson PA (2009) The localization of the

Golgin GCC185 is independent of Rab6A/A’ and Arl1. Cell 138(4):787–794.42. Eathiraj S, Pan X, Ritacco C, Lambright DG (2005) Structural basis of family-wide Rab

GTPase recognition by rabenosyn-5. Nature 436(7049):415–419.

Li et al. PNAS | February 18, 2014 | vol. 111 | no. 7 | 2577

CELL

BIOLO

GY