Tethering complexes in the endocytic pathway: CORVET and HOPS



A. Spang, Biozentrum, Universität Basel, Klingelbergstraße 50/70, CH-4056 Basel, Switzerland

Fax: +41 61 267 2148

Tel: +41 61 267 2380

E-mail: anne.spang@unibas.ch


Endocytosis describes the processes by which proteins, peptides and solutes, and also pathogens, enter the cell. Endocytosed material progresses to endosomes. Genetic studies in yeast, worms, flies and mammals have identified a set of universally conserved proteins that are essential for early-to-late endosome transition and lysosome biogenesis, and for endolysosomal trafficking pathways, including autophagy. The two Vps-C complexes CORVET (class C core vacuole/endosome tethering) and HOPS (homotypic fusion and vacuole protein sorting) perform diverse biochemical functions in endocytosis: they tether membranes, interact with Rab GTPases, activate and proof-read SNARE assembly to drive membrane fusion, and possibly attach endosomes to the cytoskeleton. In addition, several of the CORVET and HOPS subunits have diversified in metazoans, and probably form additional specialized complexes to accomodate the higher complexity of trafficking pathways in these cells. Recent studies offer new insights into the complex relationships between CORVET and HOPS complexes and other factors of the endolysosomal pathway. Interactions with V-ATPase, the ESCRT machinery, phosphoinositides, the cytoskeleton and the Rab switch suggest an intricate cooperative network for endosome maturation. Accumulating evidence supports the view that endosomal tethering complexes implement a regulatory logic that governs endomembrane identity and dynamics.


alkaline phosphatase


class C core vacuole/endosome tethering


endosomal sorting complex required for transport


guanine nucleotide exchange factor


homotypic fusion and vacuole protein sorting


intralumenal vesicles


Rab-interacting lysosomal protein


Sec1/Munc18-like SNARE master


soluble NSF (N-ethylmaleimide-sensitive factor) attachment protein receptor


transforming growth factor β receptor-associated protein


vacuolar protein sorting


Endosomal pathways play a central role in membrane traffic in the cell. Imagining these processes as a simple ‘highway to the lysosome’ would be a grave over-simplification. Indeed, endosomes have been shown to be infinitely more complex than ever imagined when they were first discovered [1-4]. Probably the image of a busy city at the cross-roads of several important trafficking pathways is more accurate. Endocytic pathways (clathrin-dependent and -independent pathways, macropinocytosis and others) internalize extracellular components and fluids as well as membrane-bound factors, including lipids [5, 6]. The vesicles formed through the endocytic process eventually fuse with early endosomes. Recycling back to the plasma membrane occurs mostly via a sorting process in early endosomes, but recycling from late endosomes has also been reported [7, 8]. Early endosomes are also part of the biosynthetic pathway, receiving proteins and lipids from the trans-Golgi network. Cargoes from the anterograde pathway may be trafficked to the plasma membrane, the lysosome or retrieved back to the trans-Golgi network; the latter process is performed by the retromer pathway [9, 10]. Alternatively, proteins may reach the lysosomes from the trans-Golgi network through late endosomes. Constant exchange and selective sequestration of proteins and lipids eventually lead to maturation from early to late endosomes. Because early endosomes may be derived from various endocytic pathways, fulfil several distinct functions and exist in different states of maturation on before becoming late endosomes, they represent a rather heterogeneous population.

A hallmark of early-to-late endosome maturation is Rab conversion, in which Rab5 present on early endosomes is replaced by the late endosomal Rab7. This event is regulated by the SAND1/MON1–CCZ1 complex [11, 12]. Rab conversion most likely represents only the end-point of a highly dynamic process: formation of intra-lumenal vesicles (ILVs) is performed by the endosomal sorting complex required for transport (ESCRT) machinery, which sequesters ubiquitinated cargo by inward budding into the endosomal lumen [13]. Progressive acidification of endosomal compartments is accomplished by V-ATPase, a highly complex, multi-subunit proton pump [14]. The phosphoinositide content of the endosomal membrane changes from phosphoinositol-3-phosphate (PI(3)P) to phosphoinositol-3,5-bis phosphate (PI(3,5)P2) [15, 16]. Endosomes change in size and morphology, losing tubular extensions and recycling capacity, and move towards the perinuclear region of the cell. In addition, membrane fusion specificity is changed by replacing the class C core vacuole/endosome tethering factor (CORVET) by the late endosomal/lysosomal homotypic fusion and vacuole protein sorting (HOPS) complex [17]. This change is accompanied by substitution of early-endosomal by late-endosomal soluble NSF (N-ethylmaleimide-sensitive factor) attachment protein receptors (SNAREs) [18, 19]. Finally, late endosomes fuse with lysosomes, from which all degraded materials as well as endosomal components are retrieved to allow subsequent rounds of fusion (Fig. 1 and Table 1) [4].

Table 1. ‘To-do list’ for endosomal tethering complexes.
1Recognize and interact with Rab GTPases [22-24]
2Collaborate with Rab-switch SAND1/MON1–CCZ1 [11]
3Recognize and interact with phosphoinositides [52, 92]
4Recognize correct SNAREs and bind them [19, 20]
5Tether and dock membranes (together with SAND1/MON1–CCZ1) [47]
6Form trans-SNARE complex and perform fusion activity (in collaboration with V-ATPase) [20, 48-50]
7Proof-read trans-SNARE complexes and inhibit non-specific fusion [54]
8Reactivate cis-SNARE complexes (in collaboration with NSF/SNAP) [20]
9Interact with the cytoskeleton (microtubules and actin) [34]
10Coordinate endosome motility through interaction with RILP (in collaboration with ESCRT?) [33, 96, 97]
11Manage endosome size (in collaboration with ESCRT?) [84, 86, 95]
12Consume vesicles from the biosynthetic ALP pathway [60-63]
Figure 1.

Endosome maturation requires an intricate series of events performed by multi-protein complexes. The endosomal tethering complexes CORVET and HOPS perform their function concomitantly and in parallel to ILV formation (by ESCRT machinery), acidification (by V-ATPase), Rab conversion (by SAND1/MON1–CCZ1), phosphatidylinositol conversion (production of PI(3)P and PI(3,5)P2 by phosphatidylinositol 3-kinases and phosphatases), endosome motility (by microtubule motor proteins) and other processes [4]. Acidification is indicated by increasing shades of red. Described interactions between endosome maturation modules are indicated by red arrows. Images are shown according to previously published models: CORVET/HOPS complexes are drawn according to Plemel et al. [22], the ESCRT machinery corresponds to the concentric circle model by Nickerson et al. [13], the V-ATPase is based on Lee et al. [14], the late endosome dynein–dynactin motor is shown as in Huotari and Helenius [4], and the overall model and organelle fusion are from Pryor and Luzio [21].

The importance of the endolysosomal pathways is highlighted by the universal evolutionary conservation of their components (Fig. 2 and Table 2), the severity of the phenotypes caused by mutations in the genes involved in these pathways (almost all are lethal in metazoans) and the various human diseases associated with these mutations.

Table 2. Homologues of CORVET/HOPS complex subunits in metazoans
Organism Saccharomyces cerevisiae Caenorhabditis elegans Drosophila melanogaster a Mus musculus Homo sapiens
  1. a dor, deep orange; car, carnation; lt, light; fob, full-of-bacteria. b No homologous gene found. c Nearest homologous genes to VPS3 are VPS39 and TRAP1.

Core VPS11 vps-11 Vps11 VPS11 VPS11
VPS16 vps-16 Vps16 VPS16 VPS16
VPS18 vps-18dor (Vps18)a VPS18 VPS18
VPS33 vps-33.1car (Vps33a)a VPS33A VPS33A
vps-33.2 Vps33b VPS33B VPS33B
VPS8 vps-8 Vps8 VPS8 VPS8
HOPS VPS39 vps-39 Vps39 VPS39 VPS39
VPS41 vps-41lt (Vps41)a VPS41 VPS41
Othersbspe-39fob (Vps16b)a SPE39 SPE39
VPS45 vps-45 Vps45 VPS45 VPS45
Figure 2.

CORVET/HOPS subunits are evolutionarily conserved. Phylogenetic trees of class C core subunits: (A) VPS11, (B) VPS16 and (C) VPS18. VPS16b/SPE39 homologs have been added to (B). (D) VPS8 and VPS41 subunits cluster into two distinct groups, both showing homology to yeast VPS41. (E) VPS39 homologs show homology to human and mouse TRAP1 as well as to yeast VPS3. (F) SM proteins VPS33A, VPS33B and VPS45 share sequence homology. Organism abbreviations: sc, Saccharomyces cerevisiae; ce, Caenorhabditis elegans; dm, Drosophila melanogaster; mm, Mus musculus; hs, Homo sapiens. The phylogenetic trees were created using ClustalW2 Multiple Sequence Alignment (protein weight matrix Gonnet; gap open 10; gap extension 0.2; gap distances 5; no end gaps; no iteration; Numiter 1; NJ clustering) at http://www.ebi.ac.uk/Tools/msa/clustalw2/, followed by ClustalW2 Phylogeny at http://www.ebi.ac.uk/Tools/phylogeny/clustalw2_phylogeny/, with default settings and the neighbor-joining method.

Endosomal tethering complexes, which were initially thought to bring together two membrane-bound endocytic structures, may have additional functions (Fig. 1 and Table 1). Indeed, it appears that CORVET/HOPS-like complexes are involved in almost all aspects of endosome maturation, and may have a crucial role in coordinating and controlling endosomal traffic to the lysosome. The multi-subunit nature of the complexes, coupled with the apparent ability to combine different subunits to fulfil specific tasks, renders the analysis very challenging. The best-characterized complex is the HOPS complex in yeast (Saccharomyces cerevisiae), and most analyses have concentrated on this prototypical endolysosomal tether [17, 20-23]. In addition, the CORVET complex, which plays a role on early endosomes, and some intermediates between CORVET and HOPS have been described, but their functions are not yet understood [24]. In higher eukaryotes, only the existence of the HOPS complex has been reported [25-34], but almost certainly there will be also a CORVET complex and several complexes with shared subunits, some of them with CORVET-like functions and some with more specialized roles. In this review, we summarize some of the known properties of yeast HOPS complex and then concentrate on the not so well-characterized functions of endosomal tethers in metazoans and their interactions with other factors of the endosomal maturation machinery.

Identification of the HOPS and CORVET genes in yeast

Vacuolar protein sorting (vps) mutants were first isolated in yeast, using defective transport of vacuolar proteins as a readout [35, 36]. This large collection of complementation groups contained more than 40 genes. In addition to transport of caboxypeptidase Y and other vacuolar enzymes, morphological analyses of the mutants have been used to group the vps genes into several classes [37-39]. Interestingly, mutations of the CORVET and HOPS subunits were found to fall into distinct classes: mutations of the core subunits VPS11, VPS16, VPS18 and VPS33 showed the most severe phenotype (class C, lack of coherent vacuoles), mutations of HOPS subunits (VPS39 and VPS41) showed fragmented vacuoles (class B), and mutations of CORVET subunits (VPS8 and VPS3) were the least affected, showing vacuoles similar to wild-type (VPS8: class A, vacuoles similar to wild-type; VPS3: class D, vacuole morphology similar to wild-type, but vacuole inheritance and acidification are defective) [39]. Because mutations in the core subunits (VPS11, VPS16, VPS18 and VPS33) eliminate both CORVET and HOPS complexes, they have the most severe effects on vacuolar biogenesis. Deletion of the HOPS-specific subunits VPS39 and VPS41 results in deficient vacuolar fusion, leading to smaller, multiple vacuoles. It is interesting to note that loss of the CORVET subunits has very little effect on vacuole morphology, probably because they function upstream of vacuole formation. Yeast, which does not depend on endocytosis for survival, is able to survive without the HOPS and CORVET complexes. However, loss-of-function mutations in almost all of these genes in higher eukaryotes cause lethality.

The structure of the HOPS complex

As knowledge of the arrangement and overall 3D structure of the HOPS tethering complex would allow insights into its function, efforts have been made to elucidate its molecular architecture (Fig. 3) [22, 23]. Exhaustive analyses of protein–protein interactions between single subunits as well as with Rab proteins have provided a structural basis for the genetic observations. The core subunits Vps11, Vps16 and Vps18 present the SNARE-interacting Vps33 subunit on one side and bind the Rab GTPase interaction module through Vps3/Vps8 (in CORVET) or Vps39/Vps41 (in HOPS) on the other side (Fig. 3A) [22]. These observations are in agreement with biochemical analyses of purified partial HOPS complexes [40]. Electron microscopy combined with single particle analysis and tomography were used to obtain 3D structural information on the HOPS complex (Fig. 3B) [23]. These analyses showed a different arrangement of subunits, in which the Rab GTPase-interacting subunits were found at opposite ends of an elongated complex [23]. This structure offers an elegant explanation for the ‘tethering’ function of the HOPS complex. Two organelles bearing Rab GTPases may be bridged and joined, allowing SNARE assembly and fusion (Fig. 3B) [23]. On the other hand, this 3D structure does not appear to be in agreement with earlier analyses that showed direct binding between the two Rab-interacting subunits [22]. Perhaps the HOPS complex exists in stretched and closed conformations, whereby Vps39 and Vps41 first contact different membranes and then come together and bind to each other [23].

Figure 3.

Subunit arrangement and functional distinctions between subunits of CORVET and HOPS complexes. (A) Interactions of CORVET and HOPS subunits with one another and with cognate Rabs according to Plemel et al. [22]. (B) Organization of the HOPS complex obtained by electron microscopy and single-particle analysis according to Bröcker et al. [23]. Functional interactions with Rab5 and Rab7 GTPases, SAND1/MON1–CCZ1 and SNAREs are indicated by dashed red lines.

It is interesting to note that different tethering complexes have apparently evolved independently. Members of the CATCHR (complex associated with tethering containing helical rods) family of tethering complexes (including conserved oligomeric complex (COG), DSL1, exocyst and Golgi-associated retrograde protein (GARP) are distinct in sequence and 3D structure from the HOPS complex [41, 42]. While the DSL1 complex has a stiff, rod-like conformation [43], COG appears to be more flexible, with three ‘legs’ [44]. The exocyst complex has a Y-shaped structure, with a central bundle of rods forming a scaffold and arms bridging different membranes with GTPases [45]. Transport protein particle (TRAPP) tethering complexes are different again, having a rather compact structure close to the membrane and relying on long coiled-coil proteins to capture vesicles [46]. Even though the basic function of all these tethering factors appears to be the same, they rely on different assemblies and combinations of proteins to perform these functions.

The HOPS complex is essential for vacuolar fusion

A large contribution to the understanding of HOPS function involved use of in vitro fusion assays with purified yeast vacuoles or liposomes [20]. Yeast offers many advantages to study vacuolar fusion. In addition to powerful genetic tools, vacuoles may be readily purified and analyzed for fusion. Fusion between purified vacuoles from different strains may be used to determine directionality and the order of events of the reaction. Freshly isolated vacuoles are dispersed and only undergo tethering and docking upon ATP addition. Fusion requires several distinguishable steps that are all catalyzed by specific protein factors. As mentioned above, many of these factors were identified by genetic screens [37-39]. The three major ‘ingredients’ are Rab GTPases, SNARE proteins and the HOPS tethering complex. Also of particular importance are lipids, including ergosterol, diacylglycerol, phosphoinositides, phosphatidylethanolamine and phosphatidic acid. Additional factors involved in vacuolar fusion are SAND1/MON1–CCZ1 [47] and the V0 sector of the V-ATPase [48-50] (discussed below). The vacuolar Ypt7 GTPase (the yeast homolog of Rab7) has been studied in detail, and its main role in the fusion process appears to be recruitment of the HOPS complex. Although Vps39 was initially believed to be a guanine nucleotide exchange factor (GEF) for Ypt7, it is now widely accepted that the Mon1-Ccz1 complex plays this role [22, 23]. However, Vps39 has the unusual ability to bind Ypt7 in both GDP- and GTP-bound conformations, but this may be more important for either GEF recruitment, compartment tethering, or both. The vacuolar SNAREs comprise the R-SNARE Nyv1 and the three Q-SNAREs Vam3, Vti1 and Vam7, which are required for homotypic fusion. Vam7 lacks an integral membrane anchor and is recruited to the vacuole by binding to PI(3)P [51], HOPS and the other SNAREs [52]. The four SNAREs associate to form a characterisic tetrameric coil of α-helices. Three highly conserved glutamines (Q) and one arginine (R) are found at the center of this structure (the 0-layer), and categorize the SNAREs as Q- or R-SNAREs. This fusion-proficient trans-SNARE complex usually exists in parallel with inactive cis-complexes. Sec18 (the yeast ortholog of mammalian NSF), Sec17 (SNAP, soluble NSF attachment protein) and ATP disassemble cis-SNARE complexes and make them available for trans-SNARE complex assembly. Probably a similar set of factors are used for all fusion reactions in the cell, including Rab family GTPases that recruit effectors that are responsible for the tethering and synthesis of phosphoinositides, and SNARE proteins and NSF/SNAP [20].

Proof-reading of SNARE complexes by HOPS

As mentioned above, the HOPS complex is an essential component in vacuolar fusion, binding and stabilizing trans-SNARE complexes, which are crucial for membrane fusion. In this context, ‘proof-reading’ means that correct pairing of SNAREs is favored to obtain fusion specificity [53], while errors in SNARE complex formation lead to inhibition of fusion [54]. The HOPS complex contains the Sec1/Munc18-like SNARE master (SM) protein VPS33. SM proteins bind to trans-SNARE complexes to direct their fusogenic activity. Individual fusion reactions are performed by distinct combinations of SNARE and SM proteins to ensure specificity, and are controlled by regulators that embed the SM–SNARE fusion machinery into a physiological context [55-57]. Consequently, HOPS does not promote non-specific formation of trans-SNARE complexes, but ensures correct arrangement of SNAREs in the canonical 3Q:1R conformation [20, 53]. Purified HOPS complex from yeast was found to inhibit in vitro vacuole fusion when mismatches in the SNARE 0-layer were present [54]. Sensing of these slight changes in the buried 0-layer depends on the HOPS complex as a whole, and is not mediated exclusively by the SM subunit Vps33 [54].

Membrane fusion experiments with reconstituted proteoliposomes and purified yeast SNAREs showed that non-cognate R-SNAREs still allow fusion, but the presence of the wrong Qc-SNARE in the Qabc-SNARE bundle abolished fusion activity [58]. Thus trans-SNARE complex formation is the critical step for defining fusion specificity, and this specificity is not provided just by the SNAREs [58]. Experiments in mammals supported conclusions regarding the need for proof-reading abilities of CORVET and HOPS complexes. Using post-nuclear supernatants from rat neurons, SNAREs in themselves were not sufficient for fusion specificity in vitro. Although specific sets of SNAREs were found on early and late endosomes, their ability to form complexes and fuse proteoliposomes was rather non-specific [19]. The proof-reading abilities in CORVET and HOPS complexes may be very important to ensure correct fusion and to maintain organelle integrity.

The HOPS complex is required for AP-3-dependent transport from the Golgi to the vacuole

Biosynthetic transport of proteins to the yeast vacuole proceeds through two separate pathways, the caboxypeptidase Y pathway and the alkaline phosphatase (ALP) pathway [59]. The ALP pathway also transports the vacuolar SNARE protein Vam3 and requires the AP-3 adaptor protein complex. Purification of AP-3 and ALP-containing vesicles revealed the presence of Vps41 on these membranes [60]. Further analyses showed that Vps41 is required for ALP pathway vesicle formation at the Golgi, and that Vps41 physically associates with the Apl5 subunit of AP-3 [60]. Genetic screening for factors involved in the ALP pathway uncovered several vps41 and AP-3 alleles, but did not reveal any additional candidates [61]. As the soluble fraction of Vps41 was found to exist as a homo-oligomer, it was hypothesized that Vps41 may form a coat around AP-3 vesicles [61]. Indeed, Vps41 is predicted to have structural similarities to clathrin, COPI and COPII coat proteins [17]. However, subsequent reports revealed that the function of Vps41 in AP-3 transport is performed as part of the HOPS complex, and consisted mainly of consumption of AP-3 vesicles at the vacuole [62]. Specific phosphorylation of Vps41 by the yeast casein kinase Yck3 exposes the AP-3 binding site in Vps41 and allows fusion of AP-3 vesicles only with the vacuole and not with endosomes, on which Vps41 is also present [63]. Even though the involvement of Vps41 in AP-3 trafficking appears to be HOPS-dependent, this does not preclude the possibility that members of the HOPS complex have functions in addition to being part of a tethering complex.

The CORVET complex may tether endosomal compartments

The function of the CORVET complex is less well understood. Initially the CORVET complex was identified as a tethering complex that interacts with Vps21, a yeast homolog of Rab5, that is necessary for endolysosomal biogenesis [24]. Interestingly, except for Vps3 and Vps8, the CORVET complex is identical to the HOPS complex. Thus Vps8 provides the interaction site for Vps21, similar to Vps39 being the Ypt7 interactor. Vps8 cooperates with Vps21 to tether late endosomal compartments [64]. Strikingly, however, not all CORVET subunits are required for this tethering process, so the function of the full CORVET complex remains elusive. It has been speculated that the CORVET complex tethers endosomal membranes upstream of the HOPS complex, mainly because of its interaction with Vps21 [22, 65].

Homologs of CORVET/HOPS subunits

All CORVET and HOPS proteins are evolutionarily conserved (Fig. 2 and Table 2) [31, 32]. The core subunits VPS11 and VPS18 each have a clear homolog in all metazoans (Fig. 2A,C) [26, 28, 31, 34, 66-68]. Homologs of VPS16 have been found and characterized (Fig. 2B) [26, 31, 68]. An additional homolog (VPS16B/SPE39) should probably also be counted as a CORVET/HOPS complex subunit (Fig. 2B) [30, 69, 70]. SPE-39 was initially described in Caenorhabditis elegans, where it was found to associate with the HOPS complex [70, 71]. A homolog of SPE-39 called VPS16b or fob (full-of-bacteria) is present in Drosophila melanogaster [30, 69]. VPS33, the SM subunit, has two clear homologs in all metazoans (Fig. 2F and Table 2) [29, 31, 32, 70]. The expression of VPS33B appears to be ubiquitous, but no expression data are currently available for VPS33A [31]. However, the genes appear to have differential effects on endosomal and lysosomal compartments [32, 70]. The cluster of VPS33 homologs (Fig. 2F) is interrupted by another SM protein, VPS45, which also acts in the endolysosomal transport [34, 72, 73]. In the initial morphological classification of vps mutants in yeast, VPS45 was found to be a member of class D, together with the CORVET subunit VPS3 [39]. Moreover, the phenotypes of vps45 mutants are indistinguishable from those of vps3∆ strains [39]. Vps45 has been characterized as an SM protein in S. cerevisiae. Finally, Vps45 is required for transport from the Golgi to vacuoles [74]. The binding of Vps45 to the SNARE Tlg2, which localizes to the trans-Golgi network and endosomes, has been analyzed in detail [75]. Interestingly, VPS45 also displayed genetic interaction with MON1–CCZ1, resulting in synthetic growth phenotypes [72]. In C. elegans, VPS-45 is involved in early steps of the endosomal pathway, accumulating small vesicles that are indicative of a fusion defect [73]. In phagosome maturation, RNAi of VPS-45 had similar effects to knockdown of CORVET/HOPS members, but was postulated to work in a step distinct from the HOPS complex member VPS-41 [26]. Interactions between VPS45 and other members of the CORVET/HOPS complex have also been observed in mammals, indicating a function on early endosomes [34]. Taken together, these results strongly suggest that VPS45 is an alternative SM protein interacting with CORVET/HOPS, and that it may play a role in endosome maturation. How the two VPS33 isoforms interact with HOPS and CORVET has not yet been established, and whether VPS45 may substitute for either or both VPS33 proteins remains to be seen.

VPS8 and VPS41, which are homologous to each other, have one clear corresponding gene in higher eukaryotes (Fig. 2D) [25, 26, 76]. VPS3 appears to be missing in C. elegans and D. melanogaster, which contain only one VPS39 homolog (Fig. 2E and Table 2) [26]. Two VPS39 homologs are found in mammals [VPS39 and transforming growth factor β receptor-associated protein (TRAP1)] [33, 77, 78]. Currently, it is unclear whether TRAP1 is a Vps3 homolog or whether it evolved later. The question regarding the ‘missing’ VPS3 in C. elegans and D. melanogaster has two possible answers: either VPS39 takes over all functions performed by two proteins in yeast, or there is an as yet to be identified factor with low sequence homology that performs the function of VPS3/39. In the latter case, this unknown protein may have homologs in mammals, providing further possible combinations for tethering complexes.

After the description of all homologs of CORVET/HOPS subunits in metazoans, a small note on nomenclature may be useful. In this review, we use the yeast designations for CORVET and HOPS, assuming that CORVET is active on early endosomes (because of the interaction with Rab5/Vps21) and HOPS on late endosomes and lysosomes (the Rab7-positive compartments). The intermediate complexes i-CORVET and i-HOPS exist in yeast, but their exact roles have not yet been established [24]. The proliferation of subunits in worms (C. elegans) and flies (D. melanogaster) (adding a second VPS33 subunit, and SPE39) and in mammals (TRAP1), and the possible addition of VPS45, adds many theoretically possible complexes. Not all of these probable complexes may exist in the cell at any given time, but certainly there will be more than just CORVET and HOPS. The available data appear to indicate that six-subunit complexes may be the rule, but other assemblies with specialized functions may occur. For example, VPS18 may be part of several complexes, and awareness of this fact should guide the interpretation of data on one individual single subunit.

Characterization of CORVET/HOPS genes in metazoans

The core subunits VPS18, VPS11, VPS16 and VPS33 of CORVET/HOPS have been described in humans, where they are ubiquitously expressed [31]. VPS11, VPS16 and VPS18 are found on early endosomes and are associated with a transferrin receptor-positive compartment. VPS18 co-immunoprecipitated with the early endosomal SNAREs SYN13 and SYN6. Although CORVET is associated with late endosomes in yeast [64], this is the best evidence to date for the existence of CORVET complexes with a probable function in early endosome fusion in metazoans [34]. Most probably there is more than one CORVET-type complex, as early endosomes are heterogeneous, showing differences in pH [79] or localization of recycling early endosomes in specialized cells [80]. This heterogeneity of endosomes may be one reason for the presence of multiple homologs of CORVET/HOPS subunits in metazoans.

VPS18 binds both SM proteins VPS45 [34] and VPS33A [81], suggesting the existence of a HOPS-like assembly in mammals. Mouse VPS39 and TRAP1 were found to be ubiquitously expressed and caused embryonic lethality when knocked out [77]. Human Vps39/Vam6 were shown to cluster and fuse late endosomes and lysosomes when over-expressed [78]. Similar functions were reported for mouse VPS18 and VPS39, causing co-localization with VPS11 and VPS33 (the antibodies could not differentiate between VPS33A and B), as well as other factors such as actin and actin-associated proteins (ezrin and myosin) [33].

Vps41 and the HOPS complex were found to be important in a variety of biological processes, ranging from antigen presentation and microbial killing in immune cells [82] to protection of neurons from the neurotoxicity of α-synuclein and progression of Parkinson's disease [76]. In addition, they may also play a role in AP-3-dependent transport, defects of which cause Hermansky–Pudlack syndrome, due to transport deficiency to lysosomes and lysosome-related organelles [83]. Taken together, the mammalian data are in agreement with the existence of CORVET and HOPS complexes similar to those found in yeast.

VPS18 (deep orange) and VPS33a (carnation) have also been characterized in Drosophila. VPS18 has been implicated in programmed autophagy [67], biogenesis of pigment granules and normal delivery of proteins to lysosomes [84]. It interacts with VPS33a in a large protein complex [84]. VPS18 and VPS33a collaborate in the maturation of late endosomal compartments and their fusion to tubular lysosomes in primary cultures of Drosophila hemocytes [29]. VPS16 appears to be part of the VPS18/VPS33a complex and to have similar phenotypes to VPS18 in terms of pigment granule biogenesis and lysosomal delivery [30]. These results indicate that VPS33a may be part of the HOPS complex in Drosophila. An additional complex containing VPS16b/SPE39 and VPS33b was also identified [30], potentially representing an ortholog of the CORVET complex. At least in Drosophila, the CORVET/HOPS homologs share a similar organization to those described in yeast. In zebrafish (Danio rerio), Vps18 and Vps39 were found to affect melanosome maturation and to accumulate vesicles, presumably playing a function in correct trafficking of endosomes [66, 85].

All CORVET/HOPS subunits have been implicated in phagosome maturation in C. elegans [26]. The VPS-18 core subunit has been analyzed in detail and was found to have strong effects on the biogenesis of endosomes and lysososmes [28]. The HOPS subunits VPS-39 and VPS-41 have specific effects on the size of late endosomes/lysosomes, indicating a function in fusion of these organelles [86].

Overall, it appears that the CORVET and HOPS complexes may be conserved throughout the animal kingdom, and that they all perform functions in the endolysosomal pathway.

The CORVET-to-HOPS switch

It has been established that, during endosome maturation, Rab5 on early endosomes is replaced by Rab7 on late endosomes/multivesicular bodies (MVBs) through a process called Rab conversion [11, 87, 88]. The SAND1/MON1–CCZ1 complex plays a dual role in this process. First, it interrupts the positive feedback loop of Rab5 activation by displacing Rabex5 (the Rab5 GEF) from endosomal membranes [11]. Second, as a Rab7 GEF, it recruits and activates Rab7 [12, 89]. A switch from CORVET to HOPS should happen concomitantly. As expected, the HOPS subunits (VPS16, VPS18, VPS33A and VPS41) interact with SAND1/MON1 in vitro [11]. A proposed mechanism for the CORVET-to-HOPS switch is inter-conversion from one complex to the other via intermediate complexes (i-CORVET and i-HOPS) [24]. Currently it is unclear whether this inter-conversion is coupled to the Rab conversion or instead indicates the existence of specialized CORVET/HOPS complexes for specialized fusion events (e.g. retrieval of material from early and late endosomes). Given the expansion of homologous factors in metazoans, specialized functions of individual complexes appear to be more likely.

In addition to the interaction of the HOPS complex with SAND1/MON1–CCZ1 during the conversion phase from Rab5-positive to Rab7-positive endosomes [11, 89], these two factors also function during homotypic vacuolar fusion [47]. The Mon1–Ccz1 complex in yeast is part of the vacuolar fusion machinery, and plays a role in the docking/tethering stage of vacuole fusion [47]. Loading of Mon1–Ccz1 onto vacuoles is dependent on the core subunits Vps11, Vps16 and Vps18, but Mon1–Ccz1 accumulates on vacuoles in strains lacking Vps39 or Vps41, [47]. Apparently, there are different HOPS sub-complexes required for stabilization and down-regulation of Mon1–Ccz1 on the vacuole [47]. The SAND1/MON1–CCZ1 complex combines two functions: Rab conversion and tethering/docking during vacuolar fusion. Both functions are probably performed together with the HOPS complex. SAND1/MON1–CCZ1 may be initially recruited through Rab5, the core subunits of CORVET (VPS16 and VPS18) and PI(3)P [11, 89]. After displacement of Rab5, this will eventually start a positive feedback loop, whereby SAND1/MON1–CCZ1 recruits and activates Rab7, which in turn allows the HOPS complex to displace CORVET and subsequently bind more SAND1/MON1–CCZ1 through its VPS41 subunit [11]. After the Rab conversion, a separate recruiting step may be necessary for the function of SAND1/MON1–CCZ1 during vacuolar fusion. It may be that involvement of the switch complex coordinates the fusion of late endosomes and lysosomes/vacuoles with a new round of early-to-late endosome transition.

Interaction of CORVET and HOPS complexes with other endosome maturation machinery

The CORVET-to-HOPS switch may be coupled to Rab conversion. However, as endosome maturation involves a number of very complex events that may or may not be coordinated, it is useful to investigate and review the interaction of CORVET and HOPS complex members with other components of the endosome maturation machinery.

Interactions with phosphoinositides

Phosphoinositides exist in various phosphorylated forms that are recognized by a plethora of proteins through specific interaction domains (e.g. PX and FYVE domains). The phosphorylations are reversible, and levels of phosphoinositides are controlled by the action of specific kinases and phosphatases (myotubularins) [15]. In the endolysosomal system, there are two main phosphoinositides: PI(3)P on early endosomes and PI(3,5)P2 on late endosomes and lysosomes. In C. elegans, the two kinases VPS-34 and PIKI-1 (which produce PI(3)P) were shown to regulate phagosome maturation [90]. The myotubularin MTM-1 was found to down-regulate PI(3)P levels and thereby drive phagosome maturation [90]. Similar effects of phosphatidylinositol 3-kinase on phagosomes were found in mouse macrophages and CHO cells [91]. In humans, the UVRAG and Beclin-1 proteins play a crucial role in regulating PI(3)P levels on autophagosomes, but a role for UVRAG together with CORVET/HOPS in the endosomal pathway was also postulated [68]. UVRAG was found to co-localize and co-immunoprecipitate with all core subunits (VPS11, VPS16, VPS18 and VPS33) and the HOPS subunit VPS39. This interaction led to enhanced autophagosome fusion in late stages of the autophagic pathway, but also to accelerated protein degradation in the endocytic pathway [68]. It is possible that UVRAG plays a role in up-regulating PI(3)P levels on endosomes, but this activity is independent of Beclin-1 [68]. Clearly, the presence of PI(3)P is a crucial factor for the recruitment of many proteins with functions in endosome (and phagosome) maturation. It may even play a central role in coordinating the various machineries and ensure their deployment at the correct place and time during endosome maturation. The HOPS complex was shown to have affinity for phosphoinositides [52], binding PI(3)P, PI(3,5)P2 and PI(4,5)P2 in in vitro assays [52]. Which of these phosphoinositides is the main interactor in vivo remains unclear, and may be different for CORVET and HOPS. However, interaction of the HOPS complex with PI(3)P and PI(4,5)P2 plays a role during vacuole fusion [92], where both phosphoinositides are enriched to promote optimal fusion [93]. On the other hand, interaction with PI(3)P on early endosomes may be important for CORVET function. Intriguingly, SAND1/MON1–CCZ1 was found to bind PI(3)P and was suggested to be recruited by a high concentration of PI(3)P to promote the Rab conversion event [11]. Concomitant with the Rab switch, the HOPS complex may be recruited through its interaction with SAND1/MON1–CCZ1 [11], thus also driving the switch from CORVET to HOPS. PI(3,5)P2 is found on late endosomes and lysosomes, and regulates the size and shape of endo/lysosomal membranes [16]. PI(3,5)P2 is also required for vacuole acidification, as lack of PI(3,5)P2 causes defects in vacuolar acidification in yeast despite the presence of the V-ATPase in the vacuolar membrane [16]. The HOPS complex also functions in these processes. In addition, PI(3,5)P2 was shown to promote vacuolar fission [94]. Such a function opposes the fusion activities of HOPS and the V-ATPase [16]. To ensure correct vacuole size and morphology, cross-talk between the fission and fusion activites is crucial. The binding of HOPS to PI(3,5)P2 may be one way to coordinate fission and fusion events.

Interactions with the ESCRT machinery

The ESCRT machinery consists of several sub-complexes (ESCRT 0, I, II and III) that form circular assemblies on endosomal membranes and actively induce the formation of ILVs. ESCRT 0, I and II are mainly involved in recognizing ubiquitinated membrane proteins and clustering them together, while ESCRT III is responsible for invagination of the ILVs and turnover of the whole machinery for subsequent rounds of ILV formation [13]. The coordination between the ESCRT machinery and CORVET/HOPS complexes has not been investigated to a great extent, but, in order to achieve efficient maturation of endosomes, the processes of ILV formation and homotypic fusion need to be balanced in some way. Fusion leads to expansion of the membrane and introduces new membrane-bound proteins that need to be sorted and recycled, and ceases after a certain degree of maturation has been achieved. On the other hand, ILV formation consumes membrane and removes proteins from the endosomal surface, and this process must be regulated so that proper sorting may occur and endosomes may keep their surface available for other maturation processes [4]. Indeed, antagonistic roles of CORVET/HOPS complexes and ESCRT have been observed in recycling of membrane proteins in yeast [95]. Whether direct interactions between the two complexes are required for these effects is not clear. ESCRT II interacts directly with Rab-interacting lysosomal protein (RILP) through its VPS22 and VPS36 subunits [96, 97]. The consequences of this interaction are not well understood: either RILP coordinates ILV biogenesis with endosome motility or it has additional Rab7-independent functions [4]. If ESCRT function and endosomal transport along microtubules are indeed coupled to prevent endosomal clustering [97], then potential attachments of endosomes to the cytoskeleton through CORVET or HOPS subunits should be relinquished and reformed at the new location. Again, ESCRT and HOPS may have antagonistic roles that need to be balanced for a smooth maturation process.

Interactions with the V-ATPase

In the initial morphological classification of VPS mutants in yeast, class E vps mutations lead to sequestration of V-ATPase subunit B (part of the cytoplasmic V1 rotor) into a separate compartment [39]. Class E comprises mostly ESCRT subunits. In addition, vps3 (CORVET) and vps41 (HOPS) mutants failed to properly assemble V-ATPase onto the cytoplasmic surface of the vacuole [39]. These findings demonstrate the inter-dependence of the various endosome maturation factors (in this case, the ESCRT machinery, the CORVET/HOPS tethering factors and the V-ATPase). A direct involvement of V-ATPase in yeast vacuole fusion has been shown for the V0 transmembrane part of the complex [48-50]. In this case, the V-ATPase was involved directly downstream of the HOPS complex in promoting the final steps in vacuolar fusion. V0 transcomplexes may connect proteolipid cylinders from opposing membranes, creating a condition that resembles the situation postulated in pore models of fusion [98]. A function of V-ATPase in the release of Vps41 from vacuoles has also been observed [99]. In a screen for mutations that caused dot-like accumulations of Vps41 proximal to the vacuole, mutants of subunits of the vacuolar V1/V0 ATPase were found. Surprisingly, these V-ATPase mutants showed a strong vacuolar fusion defect. In addition to the vacuolar fusion defect of vma16 (V0 subunit c' of the V-ATPase), mutations in V1 subunits and the presence of the V-ATPase pumping inhibitor concanamycin A also affected Vps41 localization [99]. V-ATPase-driven vacuolar fusion appears to be necessary to release HOPS from the vacuole for another round of tethering and fusion. However, whether there is a more direct interaction or whether a potential feedback mechanism also exists remains to be demonstrated.

Interactions with actin and microtubules

Mouse VPS18 and VPS39 were shown to have a function in clustering of endocytic organelles and recruiting other HOPS subunits and additional factors such as actin and actin-associated proteins (ezrin and myosin) to these clusters [33]. In addition, VPS18 and VPS16 were found to co-localize with actin [34]. The actin cytoskeleton is important for endosome maturation [4, 100-104]. Actin-driven processes are crucial for membrane fission in early endosomes and membrane fusion between late endosomes and lysosomes/vacuoles [100-104]. The interaction of the HOPS complex with the actin cytoskeleton indicates that the complex fusion machinery at the lysosome probably also contains actin and associated myosin motors [33, 104]. Thus, HOPS may also have a function in ‘tethering’ endosomes to the actin cytoskeleton to keep the compartment at a defined position in the cell [34].

VPS18 and VPS39 were shown to recruit the RAB7 effector RILP to clustered endocytic organelles [33]. RILP is known to interact with dynein–dynactin motors, and controls the transport of late endocytic organelles [105, 106]. By recruiting RILP to late endosomes, the HOPS complex may directly or indirectly influence the transport of endosomal compartments. Another connection of CORVET/HOPS complexes with the cytoskeleton is probably through the Hook1 microtubule-interacting protein, which interacts with VPS18 [34]. Moreover, the HOPS subunits VPS39 and VPS41 co-localized with microtubules [34]. These differential interactions with the cytoskeleton were proposed to be important to keep endosomal compartments fixed at specific locations in the cell, suggesting that cytoskeletal functions are coordinated with the SNARE-based fusion apparatus [34].

Conclusions and perspectives

The CORVET and HOPS complexes are large, multi-protein machines that are capable of interacting with multiple other factors of the endolysosomal pathway and display an impressive array of molecular activities (Fig. 1 and Table 1). As many of these interactions and activities probably only occur in the context of the complexes, easy biochemical analysis may not be possible. CORVET and HOPS complexes proved to be quite stable and purifiable in yeast, which provides hope regarding isolation of similar complexes from higher eukaryotes as well. This would allow a definition of existing complexes in metazoans, which contain more CORVET/HOPS subunits than yeast. A genetic approach may also be envisaged, but may be hampered by the very importance of the endolysosomal pathway in multicellular organisms, whereby most loss-of-function mutations are lethal. A comprehensive parallel analysis of all CORVET/HOPS subunits in the same system may allow allocation of individual proteins to defined complexes that may be involved in specific tethering/fusion reactions. If CORVET/HOPS complexes truly are central regulators of the endolysosomal pathway (as the available data suggest), their function will only become apparent in the context of other key players (e.g. V-ATPase, ESCRT, SAND1/MON1–CCZ1, phosphatidylinositol 3-kinases and myotubularins, actin and microtubules). Analyzing the cross-talk between all these factors appears to be a daunting task, but must be attempted to understand the regulation of endocytic pathways.