Ubiquitin-dependent sorting of integral membrane proteins for degradation in lysosomes

Introduction

One important signal for sending post-Golgi membrane proteins to the lysosome for degradation is ubiquitin (Ub). Ub attachment to membrane proteins correlates strongly with their degradation. Ub attachment not only helps destroy damaged proteins but, in mammalian cells, acutely regulates a variety of cell surface receptors including EGFR (epidermal growth factor receptor), PDGFR (platelet derived growth factor receptor), IL1-R (interleukin 1 receptor) and TrkA (nerve growth factor receptor) which undergo “downregulation” and lysosomal degradation to attenuate their signaling potential. Ub may act as a sorting signal for membrane proteins at the TGN, the plasma membrane and at endosomes. Although it has been recognized for many years that late endosomes are morphologically identical to MVBs (multivesicular bodies)[1], only recently have we have started to understand the mechanisms by which endocytosed, ubiquitinated membrane proteins are sorted into endosomal lumenal vesicles. This sorting occurs prior to the fusion of late endosomes with lysosomes that results in protein degradation[2,3]. Sorting into the lumenal vesicles of MVBs is mediated in part by cytosolic proteins functioning as ESCRT (endosomal sorting complexes required for transport) complexes together with ESCRT-associated proteins [4–7]. In this article we review how cells use Ub as a sorting signal for membrane proteins, the function of ESCRT complexes in linking MVB formation to endosome movement along microtubules and the relationship between MVB formation and fusion of late endosomes with lysosomes.

Ubiquitin as a sorting signal

The correlation between the degradation of membrane proteins and their ubiquitination has been established for sometime [8]. Now we understand that this correlation represents a conglomerate of ubiquitin-dependent trafficking steps that eventually send proteins to the interior of the lysosome for degradation. The first Ub-dependent sorting step that was discovered was internalization from the cell surface using the yeast G-protein coupled receptor Ste2p [9]. Removing the ubiquitinated lysines from the C-terminus of Ste2p blocked its internalization while in-frame fusion of Ub restored its internalization. Ub can also direct proteins to endosomes from the TGN, in some cases serving as a quality control device to divert damaged or functionally inappropriate proteins to lysosomes and prevent them from appearing on the cell surface. In yeast, this has been shown for nutrient transporters such as Fur4p and Gap1p, which are downregulated by an excess of their substrates as the direct result of ubiquitination [10–12]. Ub can also cause proteins to sort into the interior of endosomes by incorporating them into lumenal vesicles that accumulate to form MVBs [13,14]. These three Ub-dependent pathways, namely endocytosis, delivery from TGN to endosomes and sorting into lumenal vesicles were first described in yeast but also operate in animal cells (Figure 1). Studies in both systems are beginning to resolve the general mechanisms for how Ub is recognized as a sorting signal and how ubiquitinated proteins are incorporated into various transport intermediates.

How efficient a single Ub signal is for these three sorting pathways to lysosomes has been unclear. Membrane proteins can carry multiple single Ubs attached directly to substrate protein (multiple monoubiquitination) or short polyubiquitin chains. Thus proteins such as MHC-I [15] and TrkA [16] have a single lysine that may be modified by a K63-linked polyUb chain but EGFR carries multiple monoUb [17] and/or polyUb [18]. The consensus view is that it is the presence of many Ubs rather than the presence of a particular polyubiquitin linkage that affords efficient Ub-dependent sorting to lysosomes. The physical basis of this lies in the relatively low affinity of the candidate Ub-sorting receptors for a single Ub (often in the range 10⁻⁴->10⁻³M). For all three of the sorting steps, monoubiquitination appears to be sufficient in yeast. Thus, fusion of Ub onto reporter proteins causes internalization of Ste2p [19], redirection of the SNARE Snc1p to endosomes, and entry of stable vacuolar membrane proteins into MVBs [20]. The problem with these studies however, is that the presence of Ub itself can promote ubiquitination of the fusion protein which could theoretically occur on any number of exposed lysines or cysteines left in the fusion protein or the glycine tail of ubiquitin at the C-terminus. Other experiments have demonstrated the inefficiency of a single in-frame fusion of Ub as either an internalization signal, a TGN-to-endosome signal, or an MVB sorting signal. One potential problem, however, with these latter experiments is that they use mutant forms of Ub in which all lysines (including lysine 27 which protrudes into the core of Ub) are altered to arginine and which contain alterations in the C-terminal tail which could negatively impact the ability of known and unknown Ub-binding protein machinery to properly recognize these Ub-fusion reporter proteins (for instance, altering the surface lysines of Ub reduces its binding to the GGA GAT domain, RCP unpublished data).

Discovering exactly what protein complexes recognize Ub as a lysosomal sorting signal has been difficult for two main reasons. The first is that many proteins that play a role in protein sorting at the plasma membrane, TGN and endosomes have been found to contain UBDs (Ub-binding domains). These include proteins such as Eps15, Epsin, CIN85, Hrs, Tsg101, Tom1, Tollip Vps36/Eap45, STAM and GGAs 1–3, which bind Ub via UIM, SH3, GAT, CUE, NZF, and GLUE domains (reviewed in [21]). This not only presents a large number of candidate Ub-sorting proteins to analyze but also implies that there may be redundant Ub-recognition complexes. The second reason is that while it is tantalizing to suppose the primary purpose of these Ub-binding motifs is to contact ubiquitinated cargo, ubiquitination can serve many other cellular roles [22]. Indeed, most of the candidate Ub-sorting proteins are themselves ubiquitinated in animal cells upon overexpression in a process termed coupled monoubiquitination [23]. Ubiquitination of the machinery is strictly by virtue of the ability of these proteins to bind ubiquitinated proteins non-covalently and likely works by positioning the machinery on or near Ub ligase activity [23,24]. Such ubiquitination could drive the assembly of UBDs into large networks. Alternatively, it could drive an intramolecular interaction between the UBD and covalently attached Ub, which causes a conformational change that prevents other intermolecular interactions [25]. For instance, ubiquitinated GGA3 and Eps15 fail to bind Ub non-covalently and ubiquitinated Epsin fails to bind clathrin [26,27]. Furthermore, overexpression of Ub ligases that ubiquitinate Eps15 or TSG101 attenuate internalization and degradation of the EGFR [24,28].

Ubiquitin-mediated sorting at the cell surface, TGN and endosomes

At the plasma membrane, Epsin, Eps15 and their yeast homogues mediate the endocytic internalization of a variety of ubiquitinated membrane proteins and has been well described (reviewed in [29]). Interestingly, the low affinity of Epsin for a single Ub moiety correlates with the need for multiple Ubs to constitute an efficient internalization signal, thus implicating these factors in the direct recognition of Ub-cargo [30,31]. Intriguingly, Eps15 has also been found complexed with STAM and Hrs, two endosomal proteins associated with ESCRT complexes and implicated in sorting into the MVB lumen [32–34]. Thus, Eps15 and Epsin could act as switch hitters, lending their Ub-binding/regulatory capacity to a variety of sorting events.

Ub-dependent sorting of proteins from the TGN to endosomes in yeast requires GGA proteins, which bind Ub via motifs within the GAT domain. Loss of GGA or loss of Ub binding by GGAs disrupts this sorting process [35]. Whether GGAs can similarly mediate sorting of ubiquitinated proteins from the TGN to endosomes in animal cells has yet to be convincingly demonstrated. In mammalian cell, the transport of the polytopic lysosomal membrane protein LAPTM5 requires both GGA3 and ubiquitination but this process shows some interesting twists. LAPTM5 moves from the TGN to endosomes but does not require ubiquitination [36]. Rather, LAPTM5 requires both a UIM domain that binds Ub and a PY motif, which can bind the Nedd4 ubiquitin ligase. These data imply a model whereby LAPTM5 recruits Nedd4 to ubiquitinate an associated protein to which LAPTM5 can then bind via its UIM domain. Interestingly, one of the candidate ubiquitinated partners may be GGA itself, which upon ubiquitination can associate with LAPTM5, possibly pulling the membrane protein into transport vesicles that depart the TGN.

At the endosome, a host of proteins have been found which might bind ubiquitinated cargo and sort it to the MVB interior. These include the Hrs/STAM complex, GGA3, the TOM1/Tollip complex, as well as subunits of the ESCRT-I and ESCRT-II complexes thought to participate in MVB formation (Figure 2). These studies are further complicated since these protein complexes appear to support multiple cellular functions, each of which might potentially integrate with the Ub system [6]. The idea that the Hrs/STAM complex directly binds and sorts ubiquitinated cargo is supported by the observation that mutations that inactivate Ub binding by this complex result in specific defects in sorting ubiquitinated cargo into the MVB interior [37]. If indeed the Hrs/STAM complex does serve as an endosomal Ub-cargo sorting complex, it is likely not the only one—especially since homologues cannot be found in plants and Dictyostelium [38]. Other candidate Ub-sorting receptors include GGA3, which party localizes to endosomes, binds Ub, and can influence sorting of ubiquitinated EGFR [39]. Blocking Ub binding by GGA3 with mutations in the GAT domain greatly influences the trafficking of EGFR. However, those mutations also block association with other GAT-associated factors. Also, members of the Tom1 family of proteins (Tom1, Tom1L1, Tom1L2) are likely candidates as they bind Ub, can associate with other Ub binding proteins, localize to endosomes, and are represented in both plant and Dictyostelium genomes [40–42]. Depletion of Tom1 family proteins potentiates signaling from PDGFR and retards degradation of ubiquitinated IL-1R [43,44]. Tom1 also associates with another endosomal Ub-binding protein Tollip, which associates with ubiquitinated IL-1R and which requires its Ub-binding CUE domain to effect IL-1R transport into lysosomes [44]. All of these complexes can bind to clathrin, which may allow them to localize to clathrin-rich subdomains on endosomes which are thought to be the staging ground for protein sorting into lumenal vesicles [40,45–47]. All of these complexes also interact with TSG101, a Ub binding subunit of the ESCRT-I complex. The ESCRT-I and ESCRT-II complexes are localized to endosomes and in yeast are required for MVB biogenesis. At least in yeast, ESCRT-I can associate with ESCRT-II to form a proposed supercomplex. An attractive model is that Ub-binding complexes such as Hrs/STAM or Tollip/Tom1 could transfer their Ub-cargo to the ESCRT-I/II complexes for further sorting to lumenal vesicles of MVBs. However, clear evidence that Vps23p/TSG101 (yeast and animal ESCRT-I subunits subunits that bind Ub) and Vps36p/Eap45 (yeast and animal ESCRT-II subunits that bind Ub) act as specific Ub-sorting receptors is lacking. Now that the interfaces for ESCRT-I and ESCRT-II complexed with Ub have been solved [21], better testing of whether these proteins recognize Ub-cargo during cargo sorting can be done. Similarly, the role of Ub binding by mammalian Vps36/Eap45 can be addressed now that recent structural studies provide a model for how its GLUE domain binds Ub [48,49].

Studies in yeast defined a set of ESCRT protein complexes comprised of class E vps (vacuolar protein sorting) proteins that are in some way required for the biogenesis of MVBs (Figure 2). Loss of TSG101, an ESCRT-I component, from animal cells causes many of the same phenotypes that arise in yeast—accumulation of multicisternal endosomes, a block in degradation of ubiquitinated cargo proteins such as EGFR and a block in gag-dependent virus budding, a process functionally related to the budding of intralumenal vesicles into the endosomal lumen [50,51]. However, loss-of-function studies of other ESCRT components in animal cells have yielded a more complex picture. A dramatic example is depletion of ESCRT-II which does not result in gross morphological changes to the endocytic system, does not affect viral budding, the transport of ligands such as EGF to lysosomes, or the degradation of ubiquitinated membrane proteins such as MHC-I [52,53]. While there is a modest effect of ESCRT-II loss on EGFR degradation, it appears that the process of ushering ubiquitinated proteins into properly forming MVBs is largely intact. As a potential Ub-sorting receptor, ESCRT-II does not appear to be an obligate part of the sorting process, indicating that it may be only used for a subset of ubiquitinated cargo

Endosome sorting, movement and delivery to lysosomes

Microtubules play an important role in both cargo sorting and delivery in the endocytic pathway. Receptor sorting in early endosomes is retarded in the absence of dynein[54]. Movement along microtubules plays a key role in delivering cargo to lysosomes and lysosome biogenesis. The maturation of Rab5-positive early endosomes to Rab7-positive endosomes does not require microtubules [55]. However, endosomes move bidirectionally along microtubules using both dynein and kinesin motors, and efficient transport of endocytosed material from early endosomes to lysosomes is dependent on an intact microtubule cytoskeleton. One of the key proteins required for microtubule movement of endosomes is RILP (Rab7 interacting lysosomal protein) [56,57]. RILP appears to coordinate several events with regard to endosome movement since it binds the dynein-dynactin complex [56]. Intriguingly, RILP has recently been found to also bind to ESCRT-II, expanding the repertoire of possible functions with which it is involved [58,59]. The RILP Rab7 complex also binds the oxysterol-binding protein homologue ORP1L and overexpression of either RILP or ORP1L clusters lysosomes around the MTOC (microtubule organizing centre) [60]. Recently, a detailed biochemical mechanism has been proposed to explain the recruitment dynein-dynactin to late endosomes and its activation to drive endosome movement toward the minus-end of microtubules. In this model Rab7 binds RILP to recruit dynein-dynactin via p150 glued and localizes this complex in endosomal patches via an association between Rab7, ORP1L, and beta-III spectrin [61]. Beta-III spectrin can then associate with Arp1 of dynactin to lock and load the dynein-dyactin complex onto endosomes. Given the ability of mammalian ESCRT-II subunits Eap30/Vps22p and Eap45/Vps36p to directly interact with different regions of RILP, it will be interesting to determine how the molecular events that RILP orchestrates are regulated at the molecular level of ESCRT-II association. Further evidence for an interaction of ESCRT II with the microtubule cytoskeleton has come from experiments showing that it is required for proper localization of Drosophila bicoid mRNA, which encodes a homeo-domain transcription factor [62]. Bicoid mRNA is transported along microtubules by dynein and accumulates at the anterior end of the oocyte, coincident with the minus ends of the oriented microtubules. Bicoid binds directly to the GLUE domain of Eap45/Vps36p of ESCRT-II, and fails to localize properly upon loss of any of the ESCRT-II subunits [62]. ESCRT-II has also been implicated in other microtubule-dependent processes such as proper formation of the MTOC during meiosis, which is altered in fission yeast lacking the ESCRT-II Vps22p protein [63].

Movement of endosomes along microtubules may simply place organelles near subsequent fusion partners, thus fostering the next step. For instance, overexpressing the Rab5-dependent early endosomal kinesin KIF16B shunts early endosomes towards the cell periphery and accelerates recycling while depleting KIF16B clusters endosomes around the MTOC, retards recycling and accelerates delivery to lysosomes [64]. Alternatively, microtubules may help define morphological features of endosomes that promote sorting and fusion events. For instance, phagosomes form RILP-positive tubular structures along microtubules that then fuse with lysosomes [65]. Similar microtubule-dependent tubular structures also mediate some kissing and fusion events between late endosomes and lysosomes [3]. Perhaps these tubules are enriched in fusogenic proteins such as SNAREs and various tethering factors. The connection between the microtubule system and the MVB biogenesis pathway afforded by ESCRT-II may provide a potential mechanism to ensure proper timing of lysosomal fusion such that it occurs when MVB biogenesis is complete. Further evidence suggesting that MVB formation must be complete before fusion with lysosomes comes from the observation that depletion of the mammalian ESCRT-III protein Vps24p results in accumulation of EGFR in MVBs [66].

Conclusions

The protein machinery responsible for sorting ubiquitinated membrane proteins is rapidly being identified and its molecular structure solved. However, it is not at all clear exactly which complexes move ubiquitinated membrane proteins and how many observations concerning ubiquitination of Ub sorting proteins themselves reflect bonafide physiological regulatory events. Thus, there is some way to go before it is possible to decipher which Ub-binding proteins actually recognize ubiquitinated cargo for sorting and which recognize Ub to do something else. The links between sorting machinery for ubiquitinated proteins and other machinery such as that for moving organelles along microtubules and the fusion of endosomes with lysosomes are similarly at an early stage of understanding.