Ubiquitin recognition in endocytic trafficking – with or without ESCRT-0 Free

Plasma membrane-localised proteins, including various transporters and receptors, are essential for the regulation of growth and development as well as adaptation and the correct response to biotic and abiotic stress factors (reviewed in He et al., 2018). For the organism to coordinate its growth and development with the environment, it is pivotal for it to regulate the abundance and activity of many of these plasma membrane proteins. The abundance of proteins can be regulated at multiple levels including transcription, translation, protein targeting, post-translational modifications and protein degradation (reviewed in Harper and Bennett, 2016; Noack et al., 2014). For rapid removal from the cell surface and subsequent degradation, plasma membrane proteins are first endocytosed and then transported to the lysosome or vacuole via the endosomal trafficking pathway.

Endocytosis of plasma membrane proteins is mediated mainly by clathrin, which is conserved among eukaryotes (reviewed in Reynolds et al., 2018; Robinson, 2015). In concert with several adaptor proteins, clathrin cages – composed of clathrin trimers that form symmetrical three-legged structures called triskelia – are assembled at the plasma membrane allowing the internalisation of the proteins in clathrin-coated vesicles (CCVs). Upon disassembly of the clathrin cage, cargoes can be delivered to the endosomes. The endosomal transport of cargoes to the lysosome or vacuole depends on post-translational modification with ubiquitin (Ub), or ubiquitylation. Cargoes are passed on to and between conserved membrane-bound protein complexes that recognise and bind the Ub chain on the cargo protein. In the end, cargo proteins are typically sorted into intraluminal vesicles (ILVs) of multivesicular endosomes (MVEs) (reviewed in Paez Valencia et al., 2016). After fusion of the MVEs to the lysosomes or vacuoles, ILVs are digested by processing enzymes that reside in these compartments.

Ub is a small modifier protein that can be conjugated to target proteins in different manners (reviewed in Komander and Rape, 2012). In endocytic protein degradation, Ub chains linked through its lysine residue 63 (K63) are reported to act as degradation signals for the targeted protein substrate (Lauwers et al., 2009). All endosomal cargoes reported in Arabidopsis to date are K63-ubiquitylated (Dubeaux et al., 2018; Kasai et al., 2011; Leitner et al., 2012; Lu et al., 2011; Martins et al., 2015). In addition, a K63-Ub sensor (Vx3K0–GFP) localises to the plasma membrane, to endosomes and on the tonoplast in addition to at nuclear foci (Johnson and Vert, 2016), showing that the K63-linked Ub modification is involved in endosomal degradation also in plants.

The transport of ubiquitylated endosomal cargo is mediated by the endosomal sorting complexes required for transport (ESCRTs), ESCRT-0, ESCRT-I, ESCRT-II and ESCRT-III (reviewed in Schöneberg et al., 2017). Essential in multicellular organisms, defects in ESCRTs cause severe developmental and growth phenotypes (reviewed in Hurley, 2015). ESCRT-0, ESCRT-I and ESCRT-II each contain at least one Ub-binding- and one membrane-binding subunit that enables them to recognise and retain ubiquitylated cargoes at endosomal membranes (Fig. 1). Mechanisms of ESCRT-dependent endosomal degradation in plants are summarised in recent review articles (Gao et al., 2017; Isono and Kalinowska, 2017; Paez Valencia et al., 2016; Romero-Barrios and Vert, 2018).

Besides their essential function in membrane trafficking, ESCRTs are also implicated in other cellular processes that depend on ESCRT-III activity (reviewed in Hurley, 2015; Schöneberg et al., 2017). ESCRT-III assembles at the internal face of membrane constriction sites and mediates membrane scission. This process is essential for cytokinesis, mitotic nuclear envelope reformation and repair (Carlton and Martin-Serrano, 2007; Olmos et al., 2015; Vietri et al., 2015; Raab et al., 2016; Denais et al., 2016). Furthermore, the ability of ESCRT-III to act as a cellular scissor is also used by HIV and other viruses that recruit the ESCRT machinery to promote their budding from the host cells (Garrus et al., 2001; Martin-Serrano et al., 2003).

Studies in yeast have established that ESCRT-I and ESCRT-II are required for the assembly and the organisation of ESCRT-III (Henne et al., 2012; Teis et al., 2010). In plants, ESCRT-I, ESCRT-II and ESCRT-III have been shown to be distributed along the endosomal pathway, with ESCRT-I and ESCRT-II mainly on early endosomes, and ESCRT-III mainly on late endosomes (Scheuring et al., 2011). In contrast, it was recently reported that all ESCRTs are instead recruited, in a coordinated manner, to early endosomes in mammalian cells (Wenzel et al., 2018). Another study in yeast has revealed that ESCRT-III assembly can be triggered also by ESCRT-0 and the BCK1-like resistance to osmotic shock (Bro1) protein, thereby bypassing ESCRT-I and ESCRT-II (Tang et al., 2016). Whether the distribution and recruiting mechanisms of ESCRTs are different in plants remains to be investigated.

Once the cargoes are internalised in ILVs, ESCRT-III disassembly is ensured by the activity of the AAA-ATPase vacuolar protein sorting 4 (Vps4) (Babst et al., 1998). Vps4 exists in the cytosol as inactive monomers or dimers. Upon interaction with ESCRT-III subunits and binding of its activator Vps twenty associated 1 (Vta1), Vps4 assembles as hexamer on MVEs to promote ESCRT-III disassembly (reviewed in McCullough et al., 2018). The plant orthologue of Vta1, LYST-INTERACTING PROTEIN 5 (LIP5), is a positive regulator of the Vps4 orthologue SUPPRESSOR OF K⁺ TRANSPORT GROWTH DEFECT 1 (SKD1) and is involved in ILV formation and the constitutive degradation of the plasma membrane-localised auxin efflux carriers PIN-FORMED 2 (PIN2) and PIN3 (Buono et al., 2016).

Although all eukaryotic organisms possess genes coding for ESCRT subunits, the subunit composition and copy number of genes for each subunit are surprisingly diverse (Leung et al., 2008; Wideman et al., 2014). Among the four ESCRTs, the two-subunit ESCRT-0 is only found in opisthokonta and is absent in all other organisms, including plants (Leung et al., 2008; Wideman et al., 2014; Winter and Hauser, 2006). Furthermore, two subunits of ESCRT-I, multi-vesicular body sorting factor of 12 kDa (MVB12) and Ub-associated protein 1 (UBAP1) (Morita et al., 2007; Stefani et al., 2011), are also absent in plants. The different composition of ESCRTs and ESCRT-interacting protein complexes in metazoan, fungi and plants suggest that the molecular mechanisms of endocytosis and endosomal transport could differ in these organisms.

The opisthokonta-specific ESCRT-0 binds ubiquitylated cargoes through its Vps27–Hrs–STAM (VHS) domain, the Ub-interacting motif (UIM) and the Src-homology 3 (SH3) domain and interacts with ESCRT-I, which then recruits further ESCRTs to sort the cargoes into the ILVs of the MVEs (Katzmann et al., 2003; Lange et al., 2012; Wenzel et al., 2018) (Table 1). The VHS domain binds preferentially K63-linked Ub, which is a signal for ESCRT-dependent endosomal transport (Ren and Hurley, 2010). The ESCRT-0 subunit Vps27 (Hrs or HGS in metazoans) contains a FYVE domain through which it binds endosomal membranes.

ESCRT-0 knockout mutants in yeast show a typical class E phenotype, in that they are defective in proper vacuolar sorting and accumulate so-called class E compartments that are multilamellar pre-vacuolar structures (Babst et al., 1997; Raymond et al., 1992). Knockout of ESCRT-0 in metazoan causes lethality in early developmental stages and affects endosomal structures (Komada and Soriano, 1999; Lloyd et al., 2002; Roudier et al., 2005).

Given the indispensable role of ESCRT-0 in opisthokonta, we here outline how other eukaryotes, especially multicellular organisms that do not have this heterodimeric complex, recognise ubiquitylated endocytic cargoes. With a focus on the model plant Arabidopsis (Box 1), we will discuss Ub adaptors with conserved Ub-binding entities that function in the endosomal pathway and that have similar characteristics to ESCRT-0 of metazoan and fungi. To conclude, we speculate that SH3 domain-containing proteins are good candidates for endophilin and amphiphysin orthologues that mediate membrane tubulation in Arabidopsis.

Target of Myb protein 1 (Tom1) proteins are widely conserved in eukaryotes and are thought to have appeared early in the evolution as proteins that function in capturing ubiquitylated cargoes (Herman et al., 2011; Wideman et al., 2014). A Tom1 orthologue is absent in Saccharomyces cerevisiae, but is present in the amoeba Dictyostelium, where it acts to recruit the ESCRT-I protein tumour susceptibility gene 101 (DmTsg101, the yeast VPS23 homologue) to sort ubiquitylated cargo proteins to MVEs (Blanc et al., 2009). In mammals, Tom1 and its two homologues, Tom1L1 and Tom1L2, represent a subfamily of the VHS domain-containing protein family that includes the ESCRT-0 subunits, Hrs and STAM, and the Golgi-localised, γ-ear-containing, ARF-binding proteins (GGA) homologues, which act as Ub receptors in metazoan and fungi but are absent in plants (reviewed in Shields and Piper, 2011) (Table 1).

Mammalian Tom1 proteins bind and recruit clathrin through their C-terminal region (Katoh et al., 2006), and Ub through the Ub-binding VHS and GAT (GGA and TOM1) domains (Wang et al., 2010). Tom1 and Tom1L2 bind both K48- and K63-linked Ub chains with a preference for the K63 linkage (Nathan et al., 2013). Based on these findings, mammalian Tom1 proteins are proposed to be mediating the sorting of ubiquitylated cargoes. Moreover, Tom1 proteins interact with Toll-interacting protein (Tollip) that binds Ub and endosomal membranes and is involved in protein sorting and degradation (Yamakami et al., 2003). Tom1L1 can also be recruited on endosomes interacting with Tollip, Hrs or TSG101 (Puertollano, 2005).

The TOM1 subfamily has expanded in plants, as there are nine Arabidopsis TOM1-LIKE PROTEINs (TOLs) (Winter and Hauser, 2006) (Table 1). TOLs, like Tom1, have VHS and GAT domains (Fig. 2A–C) that are also present in ESCRT-0 and GGAs and are important domains for capturing ubiquitylated cargoes. TOLs bind Ub in vitro and localise to the plasma membrane and on early endosomal structures (Korbei et al., 2013).

Arabidopsis TOLs function redundantly, as only higher-order mutant combinations show defects in endosomal cargo trafficking and severe defects in plant development (Korbei et al., 2013). Although it has not been investigated yet, spatio-temporal regulation of the nine TOLs, including the possible differential expression patterns in different cell types, organs or during different developmental stages might fine-tune the function of TOLs as a Ub adaptor in trafficking processes. To date, it has not been investigated whether TOLs, like their orthologues in other organisms, interact with the ESCRT machinery and other factors involved in the endosomal transport pathway. It is also yet to be established where TOLs recognise ubiquitylated targets and whether they contribute to clathrin recruitment.

Mammalian ALG-2-interacting protein X (ALIX; also known as PDCD6IP) and its yeast counterpart Bro1 or Vps31 are ESCRT-I-and ESCRT-III-binding proteins that bind Ub and function in the endosomal trafficking pathway (reviewed in Bissig and Gruenberg, 2014). The yeast Bro1 is important for the recruitment of the deubiquitylating enzyme (DUB) degradation of alpha 4 (Doa4) to endosomes, a protein that is important for recycling of Ub and endosomal protein degradation (Dupre and Haguenauer-Tsapis, 2001; Luhtala and Odorizzi, 2004; Nikko and André, 2007). Furthermore, Bro1 has been proposed to function in parallel to ESCRT-0 as a Ub receptor and to mediate ESCRT-III assembly by activating the ESCRT-III subunit Snf7 (Pashkova et al., 2013; Tang et al., 2016).

Arabidopsis has one ALIX homologue, which has the same domain structure as the mammalian ALIX and yeast Bro1 (Table 1). The structure of these proteins comprises an N-terminal Bro1 domain, a V-shaped domain and a C-terminal proline-rich domain (PRD) (Cardona-Lopez et al., 2015; Kalinowska et al., 2015). Although the sequence identity between orthologues is only ∼20%, the organisation of their protein domains is identical. In all organisms tested so far, including plants, the Bro1 domain is essential for the interaction with ESCRT-III (Cardona-Lopez et al., 2015; Fisher et al., 2007; Kim et al., 2005). Although ALIX proteins do not have a typical Ub-binding domain, the V domain binds mono-Ub, K63-linked di-Ub and K63-linked tetra-Ub (Dowlatshahi et al., 2012; Kalinowska et al., 2015; Keren-Kaplan et al., 2013; Pashkova et al., 2013) (Fig. 2A,D). The C-terminal PRD of ALIX and Bro1 interacts with E3 Ub ligases and is necessary to bind and recruit Doa4 in yeast (Nikko and André, 2007). The PRD of ALIX is the binding site for ESCRT-I, whereas the PRD is dispensable for Bro1–ESCRT-I interaction (Nikko and André, 2007; von Schwedler et al., 2003).

In mammals, an additional Bro1 domain-containing protein, His domain phosphotyrosine phosphatase (HD-PTP, also known as PTPN23), has been reported to be involved in ESCRT-dependent intracellular trafficking events (Ali et al., 2013; Gahloth et al., 2016). Besides ALIX, four Bro1 domain-containing proteins can be found in the Arabidopsis genome (Shen et al., 2018). Although a second Bro1 domain-containing protein named BRO1-DOMAIN PROTEIN AS FYVE1/FREE1 SUPPRESSOR (BRAF) has been shown to be involved in plant endosomal protein degradation, there is no obvious homolog of HD-PTP in the Arabidopsis genome (Shen et al., 2018).

Arabidopsis ALIX localises to endosomes and is essential for plant growth and development (Cardona-Lopez et al., 2015; Kalinowska et al., 2015). In addition to binding K63-linked Ub, the ESCRT-I subunit VPS23 and the ESCRT-III subunit SNF7, Arabidopsis ALIX interacts and recruits the DUB-associated molecule with the SH3 domain of signal transduction adaptor molecule (STAM) 3 (AMSH3), a metalloproteinase DUB that is unrelated to Doa4 (Kalinowska et al., 2015). This is particularly interesting, since the ESCRT-0 subunit STAM is known to interact with and activate AMSH (also known as STAMBP) in mammals (McCullough et al., 2006; Tanaka et al., 1999). Whether, in addition to STAM, mammalian ALIX is also involved in AMSH regulation has yet to be established.

One of the biochemical properties of ESCRT-0 is to bind phospholipids. Mammalian ALIX has been implicated in viral infection processes in that it binds the MVE membrane in a Ca²⁺-dependent manner (Bissig et al., 2013). This lipid-binding module of ALIX recognises and binds an atypical membrane lipid lysobisphosphatidic acid (LBPA) that is specific to late endosomes in mammals. The LBPA insertion loop in human ALIX is not present in yeast, nor in plants, which is in accordance with the fact that they are not known to produce LBPA as endomembrane lipids (Bohdanowicz and Grinstein, 2013). Whether plant ALIX functions as an alternative to ESCRT-0 and whether and how it binds membrane lipids will be an interesting topic for future studies.

Altogether, by binding Ub and ESCRT-I, ALIX proteins, like Tom1 orthologues, could serve as functional equivalents of ESCRT-0 in capturing ubiquitylated cargoes at the endosomes and transfer them to the ESCRT machinery.

To recruit ubiquitylated cargoes to the endosomal membranes, phospholipid-binding domains, such as pleckstrin homology (PH), phox homology (PX) and Fab1, YOTB, Vac1 and EEA1 (FYVE) domains are essential (reviewed in Lemmon, 2008). The Zn²⁺-binding FYVE domain present in the ESCRT-0 subunit Hrs (Vps27 in yeast) specifically interacts with PI3P, a phospholipid present in endosomal membranes (Burd and Emr, 1998; Gaullier et al., 1998). The FYVE domain of both Hrs and Vps27 is necessary for the recruitment of ESCRT-0 to the endosomal membrane (Katzmann et al., 2003; Raiborg et al., 2001). Interestingly, the Arabidopsis genome encodes 16 FYVE domain-containing proteins (van Leeuwen et al., 2004; Wywial and Singh, 2010). Among them, FYVE1 [also known as FYVE DOMAIN PROTEIN REQUIRED FOR ENDOSOMAL SORTING (FREE)1] localises on endosomes and is involved in the control of vacuole biogenesis and membrane protein localisation as well as endosomal and autophagic protein degradation (Barberon et al., 2014; Belda-Palazon et al., 2016; Gao et al., 2014, 2015; Kolb et al., 2015). Recent studies have shown that, by binding to transcription factors involved in abscisic acid signalling, FYVE1 inhibits their DNA-binding abilities, thus suggesting a dual function of FYVE1 on endosomes and in the nucleus (Li et al., 2019).

FYVE1 binds PI3P through its FYVE domain and colocalises with late endosome markers (Barberon et al., 2014; Gao et al., 2014; Kolb et al., 2015). This is in accordance with the fact that although PI3P is present in early endosomes in animals (Gaullier et al., 2000), it is a constituent of late endosomes in plants (Simon et al., 2014). In contrast to Vps27, none of the 16 FYVE domain-containing proteins in Arabidopsis, including FYVE1, have a VHS domain. Although a canonical Ub-binding motif cannot be detected, FYVE1 binds Ub and interacts with the ESCRT-I subunit VPS23 (Table 1) (Gao et al., 2014). Whether and how FYVE1 coordinates the transfer of ubiquitylated cargoes with the ESCRT machinery on endosomes has still to be elucidated.

As FYVE1 interacts with Ub, binds PI3P and localises to endosomes and interacts with ESCRT-I, it possesses major binding characteristics of ESCRT-0 and thus could fulfil ESCRT-0 functions in plants.

The ESCRT-0 heterodimer contains STAM (Hse in yeast), a protein with multiple Ub-binding domains; an N-terminal VHS domain, a UIM and an SH3 domain. A subset of SH3 domain-containing proteins including STAM binds UBs (Lange et al., 2012; Stamenova et al., 2007). The SH3 domain also acts as an interaction surface with other proteins, mostly with PRDs. In contrast to mammals, which have up to 300 SH3 domain-containing proteins (Zarrinpar et al., 2003), our in silico search identified only five SH3 domain-containing proteins in Arabidopsis. All five are plant specific, and none has a VHS domain or a UIM similar to Hse or STAM. Four of these proteins – in addition to the SH3 domain – have a Bin/amphiphysin/Rvs (BAR) domain at the N-terminus (Gadeyne et al., 2014; Zhuang et al., 2013). One is TPLATE-ASSOCIATED SH3 DOMAIN-CONTAINING PROTEIN (TASH), which is a plant-specific subunit of the TPLATE complex, which acts in concert with adaptor protein 2 (AP2) in clathrin-mediated endocytosis (CME) (Gadeyne et al., 2014; Hirst et al., 2014).

The other three proteins, SH3P1, SH3P2 and SH3P3 (together SH3Ps), are homologous to each other and are proposed to be involved in clathrin-mediated processes (Lam et al., 2001). The SH3 domain of SH3P2 (Fig. 2A,E) preferentially binds K63-linked tetra-Ub, and interacts with the ESCRT-I subunit VPS23, the endosome-associated DUB AMSH3 in vitro and also with the autophagsome component ATG8 (Nagel et al., 2017; Zhuang et al., 2013). Interestingly, human AMSH was first identified as an interactor of the SH3 domain of the ESCRT-0 subunit STAM and was later shown to be an ESCRT-associated DUB (McCullough et al., 2004; Tanaka et al., 1999). In addition to an SH3 domain, STAM has a UIM and activates AMSH in a UIM-dependent manner (Davies et al., 2013; McCullough et al., 2006). However, the addition of SH3P2 did not impact the in vitro DUB activity of Arabidopsis AMSH (Nagel et al., 2017), which could be attributed to the fact that, unlike STAM, SH3P2 does not possess a UIM. Whether and how SH3P2 is regulating AMSH has still to be investigated.

Both ESCRT-0 subunits contain clathrin-binding domains and are essential for clathrin recruitment to the endosomal membrane (McCullough et al., 2006; Raiborg et al., 2001). ESCRT-0 in C. elegans is recruited to the plasma membrane, where it binds ubiquitylated cargoes by interacting with plasma membrane-localised adaptor complexes (Mayers et al., 2013). Plant SH3P2 localises to the plasma membrane, associates with CCVs, colocalises with clathrin light chain-labelled structures and co-immunoprecipitated with clathrin heavy chains (Nagel et al., 2017). In contrast to SH3P2, it has yet to be established whether Arabidopsis ALIX, FYVE1 and TOLs function on clathrin-positive membranes, although in silico analysis has revealed that six of the nine TOLs contain putative clathrin-binding motifs (Korbei et al., 2013).

In addition to the plasma membrane, SH3P2 is also localised to the growing cell plate and is involved in cell plate formation (Ahn et al., 2017). During cell plate formation, active trafficking events, including CME, occur at the developing cell plate (reviewed in Van Damme et al., 2008). Whether trafficking events at the cell plate require Ub adaptors and whether SH3P2 serves as such an adaptor is not yet known. As described in the following section, to date, SH3P2 has only been implicated in membrane tubulation at the cell plate.

Taken together, SH3P2 is an plasma membrane- and endosome-localised Ub-binding protein that is associated with CCVs, ESCRT-I and with the DUB AMSH3, thereby sharing interactors with ESCRT-0 and thus, similar to ESCRT-0, could function as a Ub adaptor in the ESCRT pathway.

Although SH3P2 shows biochemical and cell biological characteristics that are typical of ESCRT-0, recent studies suggest that it might function as an orthologue of yet another protein family.

Human amphiphysins and endophilins belong to the BAR domain protein family. Amphiphysins and endophilins are implicated in CME and share a similar domain organisation to SH3P2: a BAR domain at the N-terminus and an SH3 domain at the C-terminus (Fig. 3A). BAR domains bind membranes upon dimerisation and can induce membrane curvature and tubulation (Carman and Dominguez, 2018). The BAR domain of SH3P2 binds phospholipids, tubulates membranes in vitro and localises to the constricted regions of an expanding cell plate (Ahn et al., 2017).

Arabidopsis SH3P1, SH3P2 and SH3P3 share only 17%, 19% and 16% amino acid identity with human endophilin A1, respectively (Fig. 3B). Similar to endophilin, SH3P2 and SH3P3 have a short linker region between the BAR- and SH3 domains [SH3P1, 108 amino acids (aa); SH3P2, 38 aa; SH3P3, 17 aa] in contrast to mammalian amphiphysin, which has a linker length of up to 400 aa (Fig. 3A). Human amphiphysin-1 contains a binding motif for clathrin (sequence LLDLD) and a clathrin–AP-2 binding domain (CLAP) (sequence PWDLW) in the linker region (Zhang and Zelhof, 2002) (Fig. 3A). Our sequence analysis did not reveal clear clathrin-binding motifs in Arabidopsis SH3Ps, although they associated with clathrin-containing structures (Lam et al., 2001; Nagel et al., 2017). Endophilin has also been implicated in clathrin-independent endocytosis (CIE) (Boucrot et al., 2015; Renard et al., 2015). Despite indications for the existence of CIE in plants (Bandmann and Homann, 2012), the molecular mechanisms driving CIE have yet to be investigated.

Both endophilin and amphiphysin interact with the PRD of the GTPase dynamin through their SH3 domain and recruit it to the budding vesicle neck, which is an essential step during CCV budding (Sundborger et al., 2011; Takeda et al., 2018; Wigge et al., 1997). Arabidopsis DYNAMIN RELATED PROTEIN 2b (DRP2B) is recruited to clathrin foci at the plasma membrane (Fujimoto et al., 2010). Arabidopsis SH3Ps interact with DRP2A and with DRP1A (Ahn et al., 2017; Lam et al., 2002), and could potentially be involved in the recruitment of dynamins. Although further in-depth studies are required, SH3Ps are probably not strictly required for dynamin function, as, in contrast to DRP2-encoding genes that are essential in Arabidopsis (Backues et al., 2010), the sh3p triple mutant does not show apparent developmental phenotypes nor is it impaired in the uptake of the endocytic tracer dye FM4-64 (Nagel et al., 2017). Whether SH3Ps fulfil dual functions at the neck of the budding CCV and also on endosomes as Ub adaptors for the ESCRT pathway, or whether they are more specialised in one of the tasks has to be further investigated.

Although plants do not possess the ESCRT-0 heterodimer, domains that are present in opisthokont ESCRT-0 subunits can be found in plant-specific proteins that function together with the ESCRT machinery. As discussed above, the Ub- and ESCRT-binding proteins TOLs, ALIX, FYVE1 and SH3Ps function in similar protein networks to ESCRT-0 in opisthokonta and thus present an overall similar ESCRT pathway in plants, fungi and metazoan. The differences in individual proteins between organisms might serve to fine-tune the ESCRT pathway specific to the need of the organisms, thereby enabling rapid and flexible adaptation to environmental changes of plants.

Although endomembrane trafficking has diversified in the course of evolution, the protein or lipid interaction domains that are involved in this pathway are conserved and appear in different combinations. Similar to ESCRT-0 components Vps27 and Hse, TOLs, ALIX, FYVE1 and SH3Ps all localise to endomembrane compartments, interact with Ub and either interact with or have conserved motifs for the interaction with ESCRT-I (Fig. 4). To date, there is no evidence of a protein heterodimer similar to ESCRT-0 known to function in plant endocytic trafficking. Almost nothing is known about the transcriptional, translational and post-translational regulation of plant Ub adaptors. Whether these proteins are expressed in the same cell types, developmental stage or environmental conditions awaits future in-depth analyses.

Another yet poorly explored process in plant endocytosis is CIE. In mammals, among others, endophilins, caveolins, and flotillins are important in clathrin-independent endocytosis events [reviewed in (Ferreira and Boucrot, 2018)]. There are three homologues of flotillin in Arabidopsis that localise on the plasma membrane; however, none of them has been well characterised to date (Junkova et al., 2018). Homologues of caveolin do not exist in plants and, as discussed before, a true functional orthologue of endophilins has not been identified yet, though SH3Ps could represent possible candidates.

Even highly conserved components such as ESCRT-III have functional divergence in different organisms. For example, in contrast to what is found in yeast, mutants defective in the ESCRT-pathway in Arabidopsis do not typically accumulate stacked membranes called class-E compartments but rather show clustered or enlarged MVEs (Haas et al., 2007; Kalinowska et al., 2015). Plant MVEs, upon closer ultrastructural examination, do not show typical ILVs, but often generate concatenated vesicles (Buono et al., 2017). Thus, studying the regulation of endocytosis and membrane trafficking in plants not only gives insights into the evolutional flexibility of this pathway, but could also reveal insights into the molecular mechanisms of endocytic trafficking processes in other species. In recent years, more and more proteomics, transcriptomics and protein interaction data were generated and sophisticated light and electron microscopy techniques have become available for research. By using these tools, future research will unravel the molecular network supporting endocytosis and endosome-mediated transport pathways and elucidate the intriguing evolutional history of organisms to use and adapt protein domains to serve complex cellular processes.