Monoubiquitination of EEA1 in cells
To investigate whether EEA1 is ubiquitinated in cells, we expressed full-length FLAG-tagged wild type EEA1 together with hemagglutinin (HA)-tagged ubiquitin in HEK293 cells. EEA1-FLAG appeared functional as a fraction of it was localized to the endosomes [15]. Cell extracts were subject to immunoprecipitation with the FLAG antibody. Immunoblotting showed that when EEA1-FLAG was expressed, the EEA1 antibody stained not only the ectopically expressed EEA1, but also a few EEA1-containing bands migrating more slowly. These additional EEA1-containing species corresponded to ubiquitinated EEA1 because they were also recognized by anti-HA antibodies, as demonstrated by two-color immunoblotting (Figure 1A). As expected, no ubiquitinated EEA1 could be detected when FLAG-EEA1 plasmid was omitted during transfection. Ubiquitinated EEA1 was also detected with endogenous EEA1 using a similar method (Figure 1B). The migration pattern of ubiquitinated EEA1 remained similar when a ubiquitin mutant lacking lysine residues was expressed (Figure 1C). Since this lysine-less ubiquitin mutant did not support the formation of ubiquitin chain, our results suggest that a fraction of EEA1 is ubiquitinated in cells at several sites, each with a single ubiquitin moiety.
EEA1 has an intrinsic affinity to ubiquitin conjugating enzymes
Protein ubiquitination usually requires three types of enzymes, an E1 activating enzyme, an E2 conjugating enzyme and an E3 ligase [18]. A conjugation reaction also depends on ubiquitin and ATP. In most cases, these enzymes would form polymerized chains on substrates and the type of ubiquitin chains can dictate the functional consequence of a ubiquitination reaction [19].
To dissect the mechanism by which EEA1 preferentially undergoes monoubiquitination, we developed an in vitro ubiquitination assay using post-nuclear extract. To this end, cells expressing EEA1-FLAG were treated with a low salt buffer to disrupt the plasma membrane. We reasoned that the enzymes involved in EEA1 ubiquitination likely reside in either cytosol or on the endosome membranes. We therefore obtained a post-nuclear extract fraction containing both endosomes and cytosol and incubated it with HA-tagged ubiquitin and an ATP regenerating system. Immunoblotting showed that EEA1 was rapidly ubiquitinated in the presence of ATP and ubiquitin (Figure 2A). In the absence of exogenously added ATP, only a small amount of ubiquitinated EEA1 was generated, likely due to the residual endogenous ATP, which was present at millimolar concentrations in cells. By contrast, no ubiquitinated EEA1 was detected in the absence of ubiquitin. This was not surprising given that endogenous ubiquitin was significantly diluted during cell lysis.
We next wished to reconstitute ubiquitination of EEA1 using purified proteins. We incubated purified EEA1 (Figure 2B), ubiquitin activating enzyme (E1), ubiquitin, and ATP together with a panel of E2 conjugating enzymes. After incubation, immunoblotting showed that several E2 enzymes were able to support EEA1 ubiquitination even in the absence of any E3 ubiquitin ligases. These included Ube2A(HR6a), Ube2H(UbcH2), Ube2D1(UbcH5a) and Ube2E1(UbcH6). Ube2A appeared to be the most efficient one in mediating EEA1 ubiquitination (Figure 2C).
As mentioned above, ubiquitination often involves an E3 ligase that serves as a matchmaker to bring a substrate in proximity to a ubiquitin-charged E2 enzyme [20]. One method that allows bypass of an E3 ligase in ubiquitination is to utilize a ubiquitin recognition motif (UIM) in a substrate, which binds directly to the ubiquitin moiety on an activated E2 [21, 22]. This mechanism requires a hydrophobic surface composed of Ile44 and Leu8 in ubiquitin, which serves as the docking site for UIM. EEA1 does not contain any identifiable UIMs, but it might have a cryptic ubiquitin binding site that mediates its own ubiquitination. To exclude this possibility, we performed in vitro EEA1 ubiquitination using ubiquitin mutants defective in UIM binding [23]. The result showed that these ubiquitin mutants were as effective as wild type ubiquitin in ubiquitination of EEA1 (Figure 2D). This observation raised the possibility that EEA1 may directly communicate with an E2 enzyme(s) to receive ubiquitin from it. This conclusion was indeed confirmed by Surface Plasmon Resonance (SPR) experiments using purified EEA1 and an E2. These in vitro binding experiments demonstrated that EEA1 had an intrinsic affinity to Ube2A with Kd of ~9.7 ± 0.54 μM (Figure 2E, F). By contrast, Ube2G2 bound EEA1 with a significantly reduced affinity (data not shown), consistent with the fact that Ube2G2 could not ubiquitinate EEA1 in vitro.
Expression of a ubiquitin-EEA1 chimera generates giant vacuole-like endosomes
To understand the functional consequence of EEA1 monoubiquitination, we took advantage of the observation that in-frame fusion of ubiquitin to a protein often mimics endogenously generated mono-ubiquitinated protein [24, 25]. We substituted the C-terminal glycine in ubiquitin to valine to avoid cleavage of ubiquitin by cellular deubiquitinases and then fused the coding sequence of UbG76V to the 5′-end of the EEA1 gene. This construct allowed the expression of a constitutively 'ubiquitinated' EEA1 variant (Ub-EEA1), carrying just a single ubiquitin moiety at the N-terminus. We expressed EEA1 or Ub-EEA1 in COS7 cells and stained the cells with an EEA1 antibody to visualize the endosomes. In untransfected cells, the EEA1 antibody stains clusters of small vesicles in a perinuclear region. Since EEA1 normally functions as a tether that ‘primes’ endosomes for fusion, overexpression of wild type EEA1 induced endosome clustering and fusion, resulting in enlarged endosome vesicles (Figure 3A). Intriguingly, approximately 60% of cells expressing Ub-EEA1 contained giant vacuole-like EEA1-positive membrane structures in proximity to the nucleus (Figures 3A, B). Co-expressing the early endosome marker Rab5-GFP showed that Ub-EEA1 and Rab5 were co-localized in this structure (Additional file 1: Figure S1), confirming that this structure originated from early endosome vesicles, likely due to uncontrolled fusion of early endosomes. Under electron microscopy, Ub-EEA1-expressing cells often contained large vacuole-like structures resembling late endosome/lysosome (Figure 3C). Since immunoblotting showed that Ub-EEA1 was expressed at a much lower level than EEA1 (Figure 3D), Ub-EEA1 appeared much more active than unmodified EEA1 in inducing endosome fusion, causing enlarged and seemingly over-matured endosomes.
To see if the enlarged endosome phenotype depends on the interplays between Ub-EEA1 and an unknown ubiquitin receptor in cells, we substituted Ile44 in Ub-EEA1 with alanine. This isoleucine residue is required for ubiquitin recognition by all forms of ubiquitin binding domains in cells. Thus, if an ubiquitin binding effector is required for Ub-EEA1 to induce enlarged endosomes, the mutation should abolish the phenotype. Indeed, even though UbI44A-EEA1 was expressed at a higher level than Ub-EEA1, as demonstrated by immunoblotting (Figure 3D), cells expressing this mutant only displayed a small increase in size of endosomes, comparable to cells expressing EEA1 (Figure 3A). Similar observations were made using a set of EEA1 constructs that carried a FLAG tag in addition to the ubiquitin fusion in COS7 cells and in another mammalian cell line, the HEK293 cells (Additional file 1: Figure S2).
We further characterized the giant endosome structures generated by Ub-EEA1 using antibodies that label a variety of subcellular organelles. The results showed that the Ub-EEA1 positive endosomes did not co-localize with the Golgi marker β-COP. Interestingly, the morphology of the Golgi in cells expressing Ub-EEA1 was dramatically altered compared to control cells. Instead of forming a stack in a perinuclear region, β-COP positive Golgi vesicles are completely dispersed throughout the cell, suggesting a connection between the endocytic pathway and Golgi morphology (Figure 4A-C). By contrast, the ER and mitochondrial morphology were largely unaffected, and Ub-EEA1 did not colocalize with either ER or mitochondrial markers (Figure 4D-I). Intriguingly, immunostaining with a LAMP1 antibody that labels lysosomes showed that many lysosome vesicles were docked on the large endocytic structures formed upon Ub-EEA1 expression (Figure 4J-L). These results suggest that Ub-EEA1 disrupts endocytic homeostasis at least in part by inducing uncontrolled fusion of endosomes that can fuse further with the lysosomes. The results also suggest that endocytic membrane homeostasis is critical for maintaining normal Golgi stack in cells,
EEA1-Ubiquitin expression blocks internalization of transferrin
It was known from previous studies that constitutive activation of Rab5 induces uncontrolled fusion of early endosomes, resulting in giant endosomes [26]. However, endosomes generated by Rab5 activation (e.g. expression of the constitutive activated Rab5) are morphologically distinct from those produced by Ub-EEA1. The former generates enlarged endosomes that are mostly spherical whereas as the endosome sheets formed by Ub-EEA1 are often irregular in shapes. In addition, the endosome structures generated by Ub-EEA1 are also generally larger than those by Rab5 Q79L expression. These differences suggest that uncontrolled fusion may not be the only contributor that forms the abnormal endocytic structures in Ub-EEA1 expressing cells.
We examined whether Ub-EEA1-expressing cells had other endocytic trafficking defects. We first analyzed the internalization of Texas red-labeled transferrin in cells expressing either EEA1 or Ub-EEA1. Following incubation of the cells with labeled-transferrin on ice, we removed unbound transferrin by washing and then incubated the cells at 37°C in a transferrin free medium for different time points. In EEA1-expressing cells, transferrin was internalized and accumulated in EEA1 positive endosomes over time (Figure 5). In general, cells internalizing transferrin gradually lost the transferrin signal as a result of an endocytic recycling process. Interestingly, cells expressing Ub-EEA1 failed to take up transferrin (Figure 5). We questioned whether enlargement of endosomes in general can block transferrin uptake. We therefore performed the transferrin uptake experiment using cells expressing Rab5 Q79L. Despite the presence of a large number of enlarged endosomes, these cells effectively internalized Texas red-labeled transferrin (Additional file 1: Figure S3), consistent with a previous report [26]. These results further underscore the phenotypic difference between cells containing activated Rab5 and those expressing Ub-EEA1.
Ubiquitin-EEA1 traps a cell surface receptor at endosomes
One possible explanation for the enlarged endosome phenotype associated with Ub-EEA1 is that activation of EEA1 by ubiquitination may not only induce endosome fusion, but also block an endosome fission process required for recycling of certain membrane receptors such as the transferrin receptor to the plasma membrane. This could lead to a depletion of transferrin receptor from the cell surface, resulting in abnormal transferrin uptake. This model also explains why the endosomes in Ub-EEA1 cells are even larger than those in Rab5 Q79L expressing cells, as the endosome enlargement induced by Rab5 activation may be solely due to increased vesicle fusion.
To test this hypothesis, we stained cells with an antibody that recognizes the transferrin receptor. In wild type cells, transferrin receptor was mostly localized in endosomal vesicles concentrated in a perinuclear region, but small vesicles localized to peripheral regions of the cells were also frequently observed. These vesicles represented recycling endosomes, which are responsible for re-distributing internalized transferrin receptor to the plasma member. Only a few transferrin receptor molecules were detected on the plasma membrane, suggesting that it was transiently localized to the cell surface. In Ub-EEA1 expressing cells, transferrin receptor was almost entirely trapped in the enlarged endocytic structures. Almost no transferrin receptor positive recycling endosome vesicles could be detected (Figure 6). This observation supports our model that overactivation of EEA1 via ubiquitin conjugation not only induces endosome fusion, but also blocks endosome recycling.