Site-specific monoubiquitination downregulates Rab5 by disrupting effector binding and guanine nucleotide conversion
Abstract
Rab GTPases, which are involved in intracellular trafficking pathways, have recently been reported to be ubiquitinated. However, the functions of ubiquitinated Rab proteins remain unexplored. Here we show that Rab5 is monoubiquitinated on K116, K140, and K165. Upon co-transfection with ubiquitin, Rab5 exhibited abnormalities in endosomal localization and EGF-induced EGF receptor degradation. Rab5 K140R and K165R mutants restored these abnormalities, whereas K116R did not. We derived structural models of individual monoubiquitinated Rab5 proteins (mUbRab5s) by solution scattering and observed different conformational flexibilities in a site-specific manner. Structural analysis combined with biochemical data revealed that interactions with downstream effectors were impeded in mUbRab5K140, whereas GDP release and GTP loading activities were altered in mUbRab5K165. By contrast, mUbRab5K116 apparently had no effect. We propose a regulatory mechanism of Rab5 where monoubiquitination downregulates effector recruitment and GDP/GTP conversion in a site-specific manner.
https://doi.org/10.7554/eLife.29154.001Introduction
Rab GTPases, the largest family of small GTPases, regulate various vesicular trafficking processes such as endocytosis, exocytosis, and ER to Golgi transport (Hutagalung and Novick, 2011). Rab GTPases are activated by GTP binding and inactivated by the hydrolysis of GTP to GDP (Pfeffer, 2005). This regulatory cycle of Rab is facilitated by other regulators. A GDP dissociation inhibitor (GDI) stabilizes GDP-bound Rab in the inactive state. Binding of a GDI displacement factor (GDF) to the GDI:GDP-Rab complex releases GDI, freeing GDP:Rab for subsequent activation. A guanidine nucleotide exchange factor (GEF) binds to GDP:Rab5 and catalyzes GDP release and GTP loading for Rab. The activated GTP-bound Rab exerts its effects by binding to various effector proteins. Inactivation of GTP:Rab is aided by a GTPase activating protein (GAP) (Cherfils and Zeghouf, 2013), rendering the Rab GTPase ready for another cycle.
Ubiquitination, a major post translational modification, is associated with various cellular processes, including vesicular trafficking (Hochstrasser, 2009; Komander and Rape, 2012). Both the number of ubiquitin moieties and their chain linkage types, such as polyubiquitination and monoubiquitination, can provide the molecular bases for the regulation of diverse cellular activities. Two representative polyubiquitination linkages have been studied extensively. K48-linked polyubiquitination guides the substrate to the proteasome, resulting in degradation of the substrate (Komander and Rape, 2012; Hochstrasser, 2009; Akutsu et al., 2016). By contrast, K63-linked polyubiquitination regulates cellular signaling pathways (Bhoj and Chen, 2009). Monoubiquitination modulates histone modification, virus budding, and endocytosis (Hicke, 2001). Although ubiquitination in endocytosis is well-documented (Piper et al., 2014), little is known about ubiquitination of Rab GTPases, the master regulators of endocytic trafficking (Hutagalung and Novick, 2011; Mizuno-Yamasaki et al., 2012).
Evidence is mounting that ubiquitination can alter the functional consequences of small GTPases, which has been well documented for Ras GTPases. Site-specific monoubiquitination of K147 of Ras leads to activation of signaling by impairing its interaction with GAP (Baker et al., 2013a; Baker et al., 2013b; Sasaki et al., 2011). In contrast, monoubiquitination of K117 of Ras facilitates GDP release from Ras, independent of a GEF (Baker et al., 2013b). Ubiquitination of Rab has also been recently reported. HACE1, a HECT-type E3 ubiquitin ligase, ubiquitinates Rab6A, Rab8A, and Rab11A in conjunction with the β2-adrenergic receptor (β2AR) (Lachance et al., 2014). Ubiquitination of K145 of Rab11A activates Rab11A. Atypical serine ubiquitination through the formation of a phosphodiester bond independent of E1 and E2 enzymes has been reported for Rab (Qiu et al., 2016; Bhogaraju et al., 2016; Bhogaraju and Dikic, 2016). Rab33b is monoubiquitinated through this atypical serine ubiquitination by SdeA, a pathogenic effector protein of Legionella pneumophila. However, the molecular mechanisms by which ubiquitination of Rab regulates Rab function remain largely unexplored.
Rab5 regulates formation of early endosomes and maturation to late endosomes (Hutagalung and Novick, 2011). Three isoforms of mammalian Rab5 - Rab5a, Rab5b and Rab5c regulate endocytosis in a co-operative manner (Bucci et al., 1995). However, recent studies suggest that these three isoforms may have differential roles. Rab5a is required for EGF-induced EGFR degradation while Rab5b and Rab5c are not (Barbieri et al., 2000; Chen et al., 2009). Rab5c is involved in cell migration whereas Rab5a and Rab5b are not (Chen et al., 2014a). Proteomic studies have suggested that the Rab5 isoforms can be differentially ubiquitinated in cells (Chen et al., 2014b; Wagner et al., 2011; Wagner et al., 2012). Since Rab5a has been extensively studied for its involvement in endosomal fusion (Bucci et al., 1992; Gorvel et al., 1991; Barbieri et al., 2000; Hoffenberg et al., 1995; Rybin et al., 1996; Stenmark et al., 1994), we decided to focus on Rab5a for our studies (hereafter referred to as ‘Rab5’). Here, we identified monoubiquitination sites of Rab5 in cultured cells. Moreover, a modified chemical conjugation method enabled us to produce pure monoubiquitinated Rab5 proteins (mUbRab5s) in large quantities. Structural information about the mUbRab5s obtained from small-angle X-ray scattering and biochemical studies revealed that mUbRabs have different functions according to the site of ubiquitination. This research sheds light on how ubiquitination can regulate the function of Rab GTPases.
Results
Rab5 undergoes monoubiquitination at multiple sites
To determine whether Rab5 is ubiquitinated in cells as proteomics studies have suggested, we performed ubiquitination assays using HEK 293T cells. Rab5 was predominantly monoubiquitinated when FLAG-Rab5 was overexpressed with HA-ubiquitin in HEK 293T cells (Figure 1a). Because Rab5 is not known to be ubiquitinated by certain E3 ubiquitin ligases that can ubiquitinate other Rab proteins, such as the β2AR-HACE complex (Lachance et al., 2014) and the bacterial effector protein SdeA (Qiu et al., 2016; Bhogaraju et al., 2016; Bhogaraju and Dikic, 2016), monoubiquitination of Rab5 is likely to be performed by an as yet unidentified E3 ligase(s). Monoubiquitinated Rab5 migrated in a gel as doublet bands, implying that monoubiquitination of Rab5 can occur at multiple sites. Appearance of doublet bands from monoubiquitination was previously noted for caveolin-1 (Ritz et al., 2011; Kirchner et al., 2013; Hayer et al., 2010). To explore the possibility that double bands might have been caused by prenylation of Rab5 (Xu and Nagy, 2016; Wojtkowiak et al., 2011), we treated cells with a geranylgeranyl transferase inhibitor (GGTI-198) and then conducted ubiquitination assays (Figure 1b). Double bands from the monoubiquitinated Rab5 remained unchanged in cells treated with GGTI-198, demonstrating that the double bands are the result of monoubiquitination of Rab5 at multiple sites. These double bands were not observed when a high percentage polyacrylamide gel (12%) was used (Figure 1c,e–f). Next, we examined whether ubiquitination is dependent on the nucleotide binding state of Rab5. Ubiquitination assays with GDP- and GTP-bound forms of Rab5 mutants (S34N and Q79L, respectively) revealed that the monoubiquitination of Rab5 is not dependent on the nucleotide binding state of Rab5 (Figure 1c).

Rab5 is monoubiquitinated in the cellular environment.
(a) Monoubiquitination of FLAG-Rab5 in HEK 293T cells. FLAG-Rab5 was co-transfected with HA-ubiquitin or HA vector and immunoprecipitated from total cell lysates (TCL). Immunoblotting was performed with the indicated antibodies. (b) Cells were treated with a geranylgeranyl transferase inhibitor (GGTI-298) or DMSO. After 24 hr of treatment, cells were harvested and subjected to ubiquitination assay. Short and long refer to exposure time during immunoblotting (c) GDP/GTP-bound mutants of Rab5 (S34N, Q79L) were examined by the ubiquitination assay (d) Each of 13 lysine (K) residues was mutated to arginine (R), and all single mutants were subjected to ubiquitination assay. (e) A 13KR (no lysine) mutant of Rab5 was generated and subjected to the ubiquitination assay. (f) FLAG-Rab5 was co-transfected with HA-ubiquitin or HA vector and subjected to ubiquitination assay in HeLa cells. (g) Immunofluorescence assay of 13KR, 12KR (13KR with R33K or R134K), and 11KR (13KR with R33/134K) mutants. FLAG-Rab5 WT and mutants were transfected into HeLa cells that were then stained with FLAG-mouse-IgG/rhodamine (red). Scale bar, 10 μm. (h) Ubiquitination assay of single lysine mutants in HEK 293T cells. The 11KR mutant was used as a negative control. Each of 11 arginine (R) residues was mutated back to lysine residues as indicated.
Proteomic studies reported the following ubiquitination sites for Rab5: K116 in human Rab5 (Wagner et al., 2011; Chen et al., 2014b) and K134 and K140 in mouse Rab5 (Wagner et al., 2012; Chen et al., 2014b) (Table 1). To verify whether those ubiquitination sites are monoubiquitinated and to examine if other lysine residues of Rab5 can be ubiquitinated, we mutated each of the 13 lysine (K) residues of Rab5 to arginine (R) and repeated the ubiquitination assay for each K-to-R mutant in HEK 293T cells. We found that the upper band of the doublet was completely abolished in the K140R mutant, while the monoubiquitination bands of Rab5 were not altered in the other K-to-R mutants (Figure 1d). This result suggests that K140 is a major monoubiquitination site of Rab5, and that at least two or more sites are responsible for the lower band.. To identify the ubiquitination sites on Rab5 comprehensively, we prepared a lysine zero mutant in which all 13 lysine residues of Rab5 were mutated to arginine (13KR). However, the 13KR mutant was not expressed in HEK 293T cells (Figure 1e).We reasoned that this was due to mutation of both K33 and K134 based on previous structural studies of Ras and EF-Tu GTPases (Pai et al., 1989; Brünger et al., 1990; Jurnak, 1985; Berchtold et al., 1993). Two lysine residues, the P-loop lysine in the GxxxGKS/T motif and the invariant lysine in the NKxD motif, are crucial because they form direct contacts with the guanine nucleotide. Similarly, for Rab5, K33 and K134 form direct contacts with both GDP and GTP molecules (Figure 1—figure supplement 1b and c, PDB ID: 1TU4 and 3MJH, respectively). K33, which sits in the nucleotide-binding site, performs two polar interactions with GDP/GTP: the Nζ atom and amide nitrogen interact with the Oβ and Oγ atoms of guanine nucleotides (2.7–3.4 Å for GDP; 2.8–3.5 Å for GTP; Figure 1—figure supplement 1b and c). K134 also contacts GDP/GTP through two interactions: Nζ atom interacts with O4’ of the ribose ring (3.3 Å for GDP and 3.0 Å for GTP), and the amide nitrogen interacts with O6 of the guanine base (3.4 Å for both cases). Subsequently, we generated an 11KR mutant with intact K33 and K134 and examined its function by observing Rab5-positive endosome puncta in HeLa cells to visualize Rab5-positive endosomes more efficiently (Miaczynska et al., 2004; Borg et al., 2014; Kajiho et al., 2003). First, we checked whether Rab5 is monoubiquitinated in HeLa cells (Figure 1f). Rab5 was monoubiquitinated in HeLa cells, confirming that ubiquitination of Rab5 is conserved in the two cell lines. Whereas Rab5 puncta were not observed for the 13KR and the two 12KR mutants, puncta were formed as clearly in 11KR mutant cells as in WT cells (Figure 1g), suggesting that the functionality of the 11KR mutant is preserved. No monoubiquitinated Rab5 band was detected for the 11KR mutant in the ubiquitination assay (Figure 1h), suggesting that K33 and K134 are not responsible for monoubiquitination of Rab5. Considering that K134 was previously reported to be ubiquitinated by proteomic studies (Chen et al., 2014b; Wagner et al., 2011; Wagner et al., 2012), it is plausible that K134 may undergo polyubiquitination. Given the crucial role of K134 in the structural integrity of Rab5 (Figure 1—figure supplement 1b and c), we hypothesized that ubiquitination of K134 might be related to disruption/degradation of Rab5, thereby preventing us from observing ubiquitination of K134 in the current experimental conditions.
Putative ubiquitination sites of Rab5A based on published studies.
https://doi.org/10.7554/eLife.29154.004Lysine site | Peptide from mass spectrometry | Predicted protein |
---|---|---|
116 | RAKNWVKELQRQA | hRab5a (K116/215)(Chen et al., 2014b; Wagner et al., 2011) |
134 | IALSGNKADLANK | mRab5a (K134/215)(Chen et al., 2014b; Wagner et al., 2012) |
140 | KADLANKRAVDFQ | mRab5a (K140/215)(Chen et al., 2014b; Wagner et al., 2012) |
Using the 11KR mutant as a negative control, we introduced single R-to-K back-mutations in the background of the 11KR mutant and performed ubiquitination assays. Monoubiquitination bands were present for R116K, R140K, and R165K mutants in contrast to the 11KR mutant (Figure 1h). These results verified the previously known ubiquitination sites (K116 and K140) and revealed a novel ubiquitination site (K165). Interestingly, K165 corresponds to K147 of a Ras GTPase whose monoubiquitination activates Ras signaling (Baker et al., 2013a). Notably, R22K, R112K, and R183K mutants exhibited polyubiquitination patterns. Because no such polyubiquitination patterns were present in the WT (Figure 1a–f), these polyubiquitination patterns are likely artifacts from using the 11KR mutant as the background.
We analyzed these putative ubiquitination sites using the crystal structures of Rab5:GDP and Rab5:GTP (PDB ID: 1TU4 and 3MJH, respectively). None of the three lysine residues are located in the Rab5 switch regions (SW I and SW II; Figure 1—figure supplement 1a). In addition, all of the lysine residues are surface-exposed and accessible to ubiquitin (Figure 1—figure supplement 1a). Taken together, these findings indicate that, among the three lysine residues known to be ubiquitination sites of Rab5, K116 and K140 are monoubiquitinated, while K134 is not. We also identified K165 as a novel monoubiquitination site of Rab5.
Monoubiquitination of K140 and K165 of Rab5 plays a negative role in the endocytic pathway
To elucidate the consequences of monoubiquitinated Rab5, we examined the effects of monoubiquitination on the formation of Rab5-positive puncta, because Rab5 itself is widely used as an early endosome marker protein (Zhang et al., 2015; Liu et al., 2015; Alexopoulou et al., 2016). We performed immunofluorescence assays in HeLa cells to examine the formation of endogenous Rab5-positivie puncta upon overexpression of ubiquitin (Figure 2a–b). Upon transfection of ubiquitin into HeLa cells, endogenous Rab5-positive puncta diffused to the cytosol and failed to co-localize with early endosome antigen-1 (EEA1) protein, another early endosome marker protein (Figure 2a). These results raise two possibilities: (i) monoubiquitinated Rab5 failed to be recruited to endosomal membranes or (ii) monoubiquitinated Rab5 disrupt Rab5-positive endosomal formation while it localized on the endosomal membranes. To distinguish between these two possibilities, we conducted a fractionation assay followed by immunoprecipitation to enrich for mUbRab5 in cytosolic and membrane fractions (Figure 2c). We observed that the monoubiquitinated Rab5 was predominantly localized in the membrane fraction. These results prompted us to hypothesize that monoubiquitinated Rab5 could disrupt Rab5-positive endosomal formation while it was localized on the membrane. Together with the immunofluorescence assay results (Figure 2a–b), these findings strongly suggest that ubiquitination of Rab5 has negative effects on the regulatory cycle of Rab5.

Endosome delocalization of Rab5 through overexpression of ubiquitin.
(a) HeLa cells were transfected with HA-ubiquitin or HA vector. Endogenous Rab5 and EEA1 were immunostained with Rab5-mouse-IgG/rhodamine (red) and EEA1-rabbit-IgG/Alexa-fluor-488 (green). Images were obtained by confocal microscopy. Scale bar, 10 μm. (b) Quantification of immunofluorescence results in b. Number of Rab5 positive puncta co-localized with EEA1 is counted per cell (n = 45). p-value is calculated by t-test. (c) Fractionation assay. Cytosol and membrane fractions were subjected to immunoprecipitation and analyzed by immunoblotting as indicated. (d) EGF-induced EGF receptor (EGFR) degradation was monitored in HeLa cells. Cells were treated with EGF, harvested at the indicated time points, and analyzed by immunoblotting. (e) FLAG-Rab5 WT and K-to-R mutants were transfected with or without HA-ubiquitin into HeLa cells as indicated. (f) FLAG-Rab5 K11R and indicated 11KR/single R-to-K mutants were transfected with or without HA-ubiquitin into HeLa cells. Cells were immunostained with FLAG-mouse-IgG/Rhodamine (red) and HA-rabbit-IgG/Alexa-fluor-488 (green). Images were obtained by confocal microscopy. Scale bar, 10 μm. (g, h) Quantification of immunofluorescence results in f and g, respectively. Number of cells with Rab5-positive endosomes were counted from 100 cells transfected with ubiquitin. Data are presented as mean ± S.D (n = 3, *p < 0.1, ***p < 0.001, by one-way ANOVA).
Next, we investigated whether the three putative ubiquitination sites (K116, K140, and K165) of Rab5 are involved in the endocytic pathway. We conducted EGF-induced EGF receptor (EGFR) degradation assays to examine the effects of monoubiquitinated Rab5 on the endocytic trafficking pathway (Hoeller et al., 2006; Balaji et al., 2012; Smith et al., 2013). EGFR was gradually degraded in EGF-treated HeLa cells, and complete degradation was observed after 90 min of EGF treatment (Figure 2d upper-left panel). Overexpression of Rab5 apparently facilitated EGF-induced EGFR degradation, as judged by earlier complete degradation after 60 min of EGF treatment (Figure 2d upper-right panel). Interestingly, overexpression of ubiquitin and Rab5 attenuated the degradation of EGFR, supported by the retention of some EGFR after 90 min of EGF treatment (Figure 2d middle-left panel). These results are consistent with the disruption of endosomal localization of Rab5 under ubiquitin overexpression (Figure 2a–b). Thus, ubiquitination of Rab5 is likely to play a negative role in the Rab5-mediated endocytic pathway. When the K140R mutant was overexpressed with ubiquitin EGFR degradation was completely restored and no attenuation was observed and almost all of the EGFR were degraded after 90 min of EGF treatment (Figure 2d lower-left panel), while the other mutants (K116R and K165R) still possesses such attenuation. This result suggests that ubiquitination of K140 of Rab5 has a crucial role in downregulating the Rab5-mediated endocytic pathway. To examine the effects of the three putative ubiquitination sites in the endosomal localization of Rab5, we conducted immunofluorescence assays. Consistent with the results obtained using endogenous Rab5 (Figure 2a–b), Rab5-positive endosomal puncta were abolished when Flag-Rab5 WT was co-transfected with HA-ubiquitin (Figure 2e,g). Similarly, no puncta were observed for the K116R mutant upon co-transfection with HA-ubiquitin (Figure 2e,g), indicating that monoubiquitination of K116 does not affect Rab5-positive endosomal formation. By contrast, Rab5-positive endosomal puncta were detectable when either K140R or K165R mutants were co-transfected with HA-ubiquitin (Figure 2e,g). These observations suggest that monoubiquitination of K140 or K165 inhibits the Rab5-positive endosomal formation. To further investigated, we performed immunofluorescent assays using Rab5-11KR and three single-K-mutants (11KR-R116K, 11KR-R140K, and 11KR-R165K). Rab5-11KR mutant still displayed Rab5-positive endosomal puncta when it was co-transfected with HA-ubiquitin (Figure 2f). 11KR-R116K also showed Rab5-positive endosomal formation consistent with the previous finding that ubiquitination of K116 does not have an adverse effect on Rab5 function (Figure 2e–h). However, no endosomal puncta were observed for either 11KR-R140K or 11KR-R165K mutants upon co-transfection with HA-ubiquitin (Figure 2f,h). These results confirm the inhibitory role of monoubiquitination of either K140 or K165. These observations indicate that Rab5 monoubiquitination has functional consequences opposite those of Ras monoubiquitination (Baker et al., 2013a; Baker et al., 2013b): monoubiquitination of Rab5 on either K140 or K165 disrupts its function, whereas that of Ras on either K117 or K147 activates it. Taken together, our data demonstrate that monoubiquitination of K140 and K165 downregulates Rab5-mediated endocytic pathway, establishing monoubiquitination at specific sites of Rab5 as inhibitory signals, while monoubiquitination of K116 has no apparent effect.
Modified chemical ubiquitination with iterative ubiquitin addition to obtain fully monoubiquitinated Rab5 protein
To understand how monoubiquitinated Rab5s negatively affect the Rab5 regulatory cycle at molecular level, we undertook biochemical studies of monoubiquitinated Rab5s. Structural, biophysical, and biochemical characterization of monoubiquitinated Rab5 proteins requires a large quantity of sample of the highest purity. Among the previously reported methods for generating ubiquitin-protein covalent linkages (Merkley et al., 2005; McGinty et al., 2008; Freudenthal et al., 2010; Virdee et al., 2011; Rösner et al., 2015), we adopted a chemical conjugation method where a disulfide bond is formed in place of the isopeptide bond, using K-to-C mutants of Rab5 (K116C, K140C and K165C) and the C-terminal G-to-C mutant of ubiquitin (G76C). A known problem with existing chemical conjugation methods is the incompleteness of conjugation. Because Rab5 and monoubiquitinated Rab5s have similar molecular masses and isoelectric points (25 kDa and pI 7.81 for Rab5; 34 kDa and pI 7.72 for monoubiquitinated Rab5), the presence of unmodified Rab5 would prevent separation of the two species by either size exclusion chromatography or ion exchange chromatography. Therefore, we modified the published methods (Merkley et al., 2005; Baker et al., 2013a) to overcome incomplete monoubiquitination of Rab5 by introducing iterative ubiquitinG76C addition (Figure 3a). First, we altered four native cysteine residues in Rab5 to prevent non-specific disulfide bond formation: two cysteine residues on the surface of Rab5 were mutated to serine (C19S and C63S), and four C-terminal residues, including two cysteine residues (CCSN; residues 213–216), were removed, inspired by a previous study (Baker et al., 2013a). GEF assays confirmed that the engineered Rab5 Del_Cys (C19S/C63S/Δ213–216) had the similar functionality as WT Rab5 (Figure 3—figure supplement 1a,b). The production of chemically conjugated monoubiquitinated Rab5 (hereafter referred to as ‘mUbRab5’) at a specific lysine residue by iterative addition of ubiquitinG76C at regular intervals was monitored by SDS-AGE under reducing (R) and non-reducing (NR) conditions. Increments in mUbRab5 and decrements in Rab5 band intensities were observed as more cycles of ubiquitinG76C addition were completed (Figure 3b and Figure 3—figure supplement 1c,d). Size exclusion chromatography successfully separated fractions containing only mUbRab5 (Figure 3c and Figure 3—figure supplement 1c,d). Finally, we obtained chemically synthesized mUbRab5K116, mUbRab5K140, and mUbRab5K165 (Figure 3d). This simple modification resulted in significant improvement of mUbRab5 production and enabled us to perform structural and biochemical studies.

Chemical synthesis of mUbRab5s by iterative addition of ubiquitinG76C.
(a) Schematic diagram of the chemical ubiquitination of Rab5 by iterative addition of ubiquitinG76C. (b) Production of mUbRab5K140 was monitored by SDS-PAGE under reducing (R) or non-reducing (NR) conditions. Intervals between addition of ubiquitinG76C (Ub) are indicated. O/N, overnight. Data for the production of mUbRab5K116 and mUbRab5K165 are shown in Figure 3—figure supplement 1c,d. (c) A representative size exclusion chromatography chromatogram of the final reaction mixture for mUbRab5K140 synthesis. (Inset) Fractions were analyzed by SDS-PAGE. White arrowhead indicates a fraction containing pure mUbRab5K140. (d) Final products of chemically synthesized mUbRab5K116, mUbRab5K140, and mUbRab5K165.
Structural models of mUbRab5s derived by solution scattering
Despite the great interest in ubiquitinated proteins, very few structures of ubiquitinated proteins have been reported (Merkley et al., 2005; Zhang et al., 2012; Freudenthal et al., 2010). This might be due to the difficulties in purifying ubiquitinated proteins and the dynamics of the ubiquitin moiety; the ubiquitin moiety in ubiquitinated proteins has been reported to be able to adopt multiple conformations (Baker et al., 2013a; Ye et al., 2012). Consistently, structural simulations of ubiquitin moieties of mUbRas have predicted flexibility of these moieties (Baker et al., 2013a). Because of the inherent flexibility of the ubiquitin moiety of ubiquitinated proteins, we opted to employ small-angle X-ray scattering (SAXS). Three mUbRab5s (mUbRab5K116, mUbRab5K140, and mUbRab5K165) were subjected to dynamic light scattering to evaluate whether the chemically synthesized mUbRab5s were stable and monodisperse in solution. All three mUbRab5s were monodisperse in solution (polydispersity <12%; Figure 4—figure supplement 1a). To conduct SAXS experiment mUbRab5K116, mUbRab5K140, and mUbRab5K165 were concentrated up to 2.9, 7.3, and 2.6 mg/ml, respectively, and subjected to SAXS measurements (Table 2). All three mUbRab5s showed linear Guinier regions at low q, leading to determination of reliable values for a radius of gyration (Rg) (Svergun, 1992). To check whether the ubiquitin moieties of Rab5 are flexible, we performed flexibility analysis using the Porod-Debye method (Rambo and Tainer, 2011). SAXS curves from the three mUbRab5s did not have a plateau in the Porod-Debye plot, implying that all of the mUbRab5s were flexible in solution (Figure 4—figure supplement 1b). To understand the conformational dynamics of the mUbRab5s, we analyzed SAXS data using an ensemble-optimized method (EOM) (Tria et al., 2015). EOM was originally designed for multi-domain proteins with a flexible linker. Input files for EOM are a single linear protein sequence covering entire protein, atomic coordinates for each domain, and SAXS data (Tria et al., 2015). Because ubiquitinated proteins are not single polypeptides but branched ones, we generated the sequences and atomic coordinates of mUbRab5s as described in the Materials and methods. A pool of 10,000 independent models was generated based on the sequence and structural information from SAXS curve by RANCH (embedded in EOM, ATSAS package). Then, a genetic algorithm for the selection of an ensemble was performed by GAJOE with 100 times (embedded in EOM, ATSAS package). Finally, the best ensemble matched with the SAXS curve with lowest χ2 was selected. Distribution of conformational ensembles of mUbRab5K140 was clearly distinct from those of mUbRab5K116 and mUbRab5K165. Two distinct populations of conformational ensembles were evident for mUbRab5K140, with a radius-of-gyration (Rg) of 23 and 27 Å, respectively, and maximum distance (Dmax) of 75 and 90 Å (Figure 4b,e). By contrast, mUbRab5K116 and mUbRab5K165 comprised a single population with an Rg of 25 Å and Dmax of 80 Å (Figure 4a,d and and c,f). These results strongly suggest that monoubiquitination at these three lysine sites has different functional consequences. EOM-derived structural models of mUbRab5s revealed specific orientations of the ubiquitin moiety. Three major conformations were derived for mUbRab5K140 (Figure 4h) and two conformations for each of mUbRab5K116 and mUbRab5K165 (Figure 4g,i). This was in contrast to K147-monoubiquitinated Ras (mUbRasK147) that exhibited different orientations of the ubiquitin moiety in the top 10 lowest-energy models from simulations (Baker et al., 2013a). The number of relative orientations of the ubiquitin moiety appears to be less than that of the major conformations derived: essentially, two orientations for mUbRab5K140 and a single orientation each for mUbRab5K116 and mUbRab5K165. This indicates that the ubiquitin moiety on K140 is more flexible than those on K116 and K165. Our results suggest that the ubiquitin moieties on multiple lysine residues of Rab5 can assume different conformations, resulting in different modes of regulation. Similar to the molecular dynamics simulation of mUbRas (Baker et al., 2013a), none of the ubiquitin molecules from the EOM models were located proximal to Rab5. This raises the possibility that monoubiquitination of Rab5 can regulate its function by altering its interactions with partner proteins such as GEF, GAP, and downstream effectors.

Ensemble models of mUbRab5s from small angle X-ray scattering (SAXS) obtained using the ensemble optimized method (EOM).
(a, b, c) Radius of gyration (Rg) and (d, e, f) maximum inter-atomic distance (Dmax) plots of the randomly generated pool (solid black) and EOM-selected model (orange) are shown for mUbRab5K116, mUbRab5K140, and mUbRab5K165, respectively. (g, h, i) (Upper) Final ensemble models of mUbRab5K116, mUbRab5K140, and mUbRab5K165. Each EOM-selected ensemble and switch I and II regions of Rab5 (SW I and SW II) are labeled. Rab5 is colored in cyan and ubiquitin is colored in series of red. (Lower) Raw data with the calculated SAXS curve for each model are plotted.
-
Figure 4—source data 1
SAXS raw data and EOM results of mUbRab5K116.
- https://doi.org/10.7554/eLife.29154.013
-
Figure 4—source data 2
SAXS raw data and EOM results of mUbRab5K140.
- https://doi.org/10.7554/eLife.29154.014
-
Figure 4—source data 3
SAXS raw data and EOM results of mUbRab5K165.
- https://doi.org/10.7554/eLife.29154.015
SAXS data collection and analysis statistics.
https://doi.org/10.7554/eLife.29154.016mUbRab5K116 | mUbRab5K140 | mUbRab5K165 | |
---|---|---|---|
Data-collection parameters | |||
Synchrotron beamlines | PAL-4C | ||
Beam geometry | Capillary | ||
Wavelength (Å) | 1.24 | ||
Exposure time (sec) | 10 | ||
Concentration range (mg/ml) | 0.4–2.9 | 0.9–7.3 | 1.0–2.6 |
Sample parameters | |||
Polydispersity (%, by DLS) | 10.9 | 11.6 | 11.8 |
Structural parameters | |||
I(0) (cm−1) [from Guinier] | 0.73 ± 0.01 | 4.29 ± 0.03 | 10.69 ± 0.05 |
Rg (Å) [from Guinier] | 25.54 ± 0.24 | 23.38 ± 0.34 | 27.23 ± 0.34 |
I(0) (cm−1) [from P(r)] | 0.73 | 4.35 | 10.74 |
Rg (Å) [from P(r)] | 26.55 | 24.61 | 27.81 |
Dmax (Å) | 86.56 | 83.51 | 93.69 |
Porod volume estimate (Å3) | 44428.4 | 32040.2 | 46685.0 |
Software employed | |||
Primary data reduction | RAW | ||
Dara processing | PRIMUS | ||
Ensemble analysis | EOM and FoXS MES | ||
Three-dimensional representations | PyMOL |
To determine if monoubiquitination affected the interaction of Rab5 with its partners, we superimposed our mUbRab5 models with known complex structures of Rab5:Rabex-5 (a GEF), RabGAP-5 (a GAP), and Rabaptin5 and EEA1 (effector proteins). In the mUbRab5:Rabex-5 complex models obtained by superpositioning of mUBRab5 and the Rab5:Rabex-5 structure (PDB ID: 4DQU) (Zhang et al., 2014), none of the ubiquitin moieties of mUbRab5s appeared to clash with Rabex-5 (Figure 4—figure supplement 2). We speculated that the ubiquitin moieties might not interfere with activities related to a GEF, such as GDP release and GTP loading. Next, we generated structural models of mUbRab5:Rabaptin5 and mUbRab5:EEA1 by superposition on Rab5:Rabaptin5 and Rab5:EEA1 structures (PDB IDs: 1TU3 and 3MJH, respectively). The ubiquitin moieties of mUbRab5K116 and mUbRab5K165 were far from both Rabaptin5 and EEA1 (Figure 4—figure supplement 3). However, the switch I (SWI) region of Rab5 was juxtaposed with the flexible ubiquitin moiety of mUbRab5K140, raising the possibility that a flexible ubiquitin moiety could cause a conformational change in the SWI. Because this conformational change of SWI is required for activation of Rab5 through interactions with downstream effector proteins, potential interruption of the SWI by mUbRab5K140 could have a severe impact on the Rab5 regulatory cycle. Finally, we generated a structural model for mUbRab5:RabGAP-5 by superposition to a known Rab:GAP structure (PDB ID: 4HLQ). The ubiquitin moiety of mUbRab5K116 did not clash sterically with RabGAP-5 (Figure 4—figure supplement 4a). Both mUbRab5K140 and mUbRab5K165 seemingly bring the ubiquitin moieties close to RabGAP-5 (Figure 4—figure supplement 4b,c), suggesting that neither mUbRab5K140 nor mUbRab5K165 would have enhanced interactions with RabGAP-5. Taken together, our structural analyses of mUbRab5s with partner proteins such as GEF, GAP, and effector proteins strongly suggest that the inhibitory effects of some mUbRab5s are highly likely to be caused by impedance of interactions with effector proteins.
mUbRab5s retain GEF-mediated guanine nucleotide conversion
To dissect the inhibitory role of mUbRab5s at the molecular level, we examined whether monoubiquitination of Rab5 could interfere with guanine nucleotide conversion, because the major regulatory mechanism of Rab GTPase is the conversion between a GDP-bound state (inactive) and GTP-bound state (active) (Novick, 2016; Mizuno-Yamasaki et al., 2012; Cherfils and Zeghouf, 2013). Rab5 guanidine nucleotide exchange factor (Rabex-5) mediates the conversion from GDP to GTP for Rab5 (Zhang et al., 2014; Thomas and Strutt, 2014; Aikawa et al., 2012). We measured GDP dissociation and GTP association rates of the mUbRab5s in the presence and absence of a GEF (Rabex-5132-393). Guanine nucleotide analogues 2’-(or-3’)-O-(N-methyl-anthraniloyl) GDP (MANT-GDP) and MANT-GTP were used to detect the interaction between guanine nucleotides and Rab5. GDP release from either Rab5 or mUbRab5 was measured by monitoring the decrease in the amount of fluorescence from MANT-GDP when non-fluorescent GDP was added in the presence and absence of Rabex-5, a GEF (Figure 5a,b and Table 3). Both mUbRab5K116 and mUbRab5K140 showed similar GDP-release activities to WT Rab5 in the presence of Rabex-5, indicating that monoubiquitination of K116 and K140 does not interfere with the GEF function (Figure 5a,b). Notably, mUbRab5K165 exhibited reduced GDP release in the presence of GEF, and the curve could not be fitted by a one-phase decay linear regression model. Next, GTP loading was monitored using MANT-GTP and Rabex-5. Again, mUbRab5K116 and mUbRab5K140 showed similar levels of GTP loading to Rab5 (Figure 5c,d). However, the fluorescence signal from mUbRab5K165 decreased continuously after it reached a maximum. This observation indicates that monoubiquitination of K165 does not block GTP loading of Rab5, but can weaken the interaction between GTP and Rab5. Our data support the structural analyses that predicted that monoubiquitination of Rab5s would have no effect on the Rab5-GEF interaction (Figure 4—figure supplement 2a–c). Taken together, these findings suggest that monoubiquitination of Rab5 on K116 and K140 does not affect GEF activity, while monoubiquitination of K165 interferes with the GEF-mediated conversion cycle, as indicated by this protein’s reduced response to GDP release and leakage during GTP loading.

Intrinsic and GEF-mediated GDP release and GTP loading activities of mUbRab5s.
(a) MANT-GDP release assay with (green) or without a GEF (orange; Rabex-5132-393). Raw data and fitted one-phase exponential dissociation curves are shown. (b) Dissociation rates of MANT-GDP with or without the GEF are shown as mean ± standard deviation (S.D.; n = 3). (c) MANT-GTP loading was monitored by fluorescence emission at different time points. GEF was added to the reaction mixture to facilitate the GTP loading at the indicated time (gray arrow). (d) Association rates of MANT-GTP with the GEF are shown as mean ± S.D. (n = 3). Raw data and fitted one-phase exponential association curves are shown. (b, d) Dissociation and association rates of mUbRab5K165 with GEF were not determined (labeled ‘N.D.”) because the data could not be fitted to one-phase exponential dissociation and association curves.
-
Figure 5—source data 1
Raw data of GDP release assay for Rab5 and mUbRab5s.
- https://doi.org/10.7554/eLife.29154.018
-
Figure 5—source data 2
Raw data of GTP loading assay for Rab5 and mUbRab5s.
- https://doi.org/10.7554/eLife.29154.019
GDP dissociation and GTP association rates for Rab5 and mUbRab5s.
https://doi.org/10.7554/eLife.29154.020GDP dissociation rate ± SD* (10−4 s−1) | GTP association rate ± SD (10−4 s−1) | |||
---|---|---|---|---|
Without GEF | With GEF | |||
Rab5 | 5.8 ± 0.2 | 20.2 ± 0.3 | 6.2 ± 0.2 | |
mUbRab5K116 | 8.6 ± 0.2 | 18.8 ± 0.1 | 8.2 ± 0.2 | |
mUbRab5K140 | 5.3 ± 0.2 | 13.6 ± 0.2 | 9.8 ± 0.2 | |
mUbRab5K165 | 1.7 ± 0.2 | ND† | ND† |
-
* SD, standard deviation
† ND, not determined.
Monoubiquitination of K140 of Rab5 decreases its binding affinity to effector proteins
Our data demonstrate that endosome localization of Rab5 is downregulated by monoubiquitination of either K140 or K165, while monoubiquitination on K116 is not important for localization. Notably, we found that mUbRab5K165 showed reduced GDP release activity and weakened GTP loading. However, the GEF activity of mUbRab5K140 remained unchanged relative to WT. We then conducted downstream effector protein binding assays with Rab5 and mURab5s. Several studies have reported that monoubiquitination of a target protein can alter its interactions with other proteins (Lin et al., 2016; Lau et al., 2015; Duan et al., 2016). Previous studies of Ras monoubiquitination have revealed that monoubiquitinated Ras has different binding affinities to its effector proteins compared to WT Ras (Sasaki et al., 2011). This prompted us to examine the interactions between mUbRabs and effector proteins. We determined the dissociation constants (Kd) of Rab5 and mUbRab5s for two Rab5 effector proteins, Rabaptin5 and EEA1 (Mishra et al., 2010; Zhu et al., 2004), using bio-layer interferometry (Figure 6). Each of the Rab5 binding domains of Rabaptin5 and EEA1 (Rabaptin5551-862 and EEA136-91) was highly purified as a GST fusion protein (Figure 6a). Rabaptin5, a well-characterized Rab5 effector protein, has been widely used to study biochemical and biophysical mechanisms involved in Rab5-mediated endocytosis (Zhu et al., 2007; Zhu et al., 2004; Zhang et al., 2014). Surprisingly, the Kd value of mUbRab5K140 for Rabaptin5 was five-fold higher than that of unmodified Rab5 (37.7 ± 4.9 μM vs. 8.0 ± 3.4 μM; Figure 6b,c and Table 4), indicating the reduced affinity of mUbRab5K140 for Rabaptin5. However, mUbRab5K116 and mUbRab5K165 showed little change in the Kd values: 8.0 ± 0.6 μM and 4.4 ± 0.2 μM, respectively. These results strongly suggest that monoubiquitination of K140 of Rab5 inhibits its binding to downstream effector proteins. Because effector proteins bind to Rab5 via similar binding modes (Mishra et al., 2010), it is plausible that mUbRab5K140 can inhibit interactions with other Rab5 effector proteins. To evaluate this possibility, we determined the Kd values of Rab5 and mUbRab5s for early endosome antigen-1 (EEA1), another well-known Rab5 effector protein. EEA1 recognizes phosphatidylinositol 3-phosphate-positive membranes and recruits Rab5 (Christoforidis et al., 1999; Rubino et al., 2000; Murray et al., 2016). mUbRabK140 exhibited an approximately 20-fold increase in Kd for EEA1 compared to unmodified Rab5 (108.0 ± 22.0 μM vs. 5.8 ± 0.3 μM, respectively; Figure 6d,e and Table 4). The Kd value of unmodified Rab5 for EEA1 was similar to that reported previously (5.8 ± 0.3 μM vs. 2.4 ± 0.23 μM) (Mishra et al., 2010). mUbRab5K116 showed a modest increase in Kd value for EEA1 (18.4 ± 0.8 μM), while mUbRab5K165 showed virtually no change (4.7 ± 0.5 μM) in Kd value relative to that of unmodified Rab5. These data are consistent with the structural models, which suggested a possible clash of mUbRabK140 with effector proteins but no such clash for mUbRabK116 and mUbRabK165 (Figure 4—figure supplement 2d–i). Our results collectively establish that monoubiquitination of K140 of Rab5 disrupts its interactions with down-stream effector proteins, while monoubiquitination of K116 only has a minor effect on these interactions.

Impairment of the interaction of mUbRab5K140 with effector proteins (Rabaptin5 and EEA1).
(a) Highly purified GST-tagged Rabaptin5551-862, GST-EEA136-91, and GST proteins were prepared. (b, d) Bio-layer interferometry assay was performed to determine dissociation constants (Kd) for mUbRab5K140 and either Rabpatin-5 or EEA1. Several concentrations of Rab5 or mUbRab5s were used as indicated. (c, e) Dissociation constants were calculated by global binding fitting. Kd values are shown as mean ± S.E.M.
-
Figure 6—source data 1
Raw data of BLI assays of Rab5 and mUbRab5s with Rabaptin5.
- https://doi.org/10.7554/eLife.29154.022
-
Figure 6—source data 2
Raw data of BLI assays of Rab5 and mUbRab5s with EEA1.
- https://doi.org/10.7554/eLife.29154.023
Binding kinetics of Rab5 and mUbRab5s to Rabaptin5551-862 and EEA136-91.
https://doi.org/10.7554/eLife.29154.024Rabaptin5 | EEA1 | |||||
---|---|---|---|---|---|---|
kon ± SEM* (102 M−1s−1) | koff ± SEM (10 −3s−1) | Kd ±SEM (μM) | kon ±SEM (102 M−1s−1) | koff ± SEM (10 −3s−1) | Kd ±SEM (μM) | |
Rab5 | 6.6 ± 2.8 | 2.2 ± 0.2 | 8.0 ± 3.4 | 4.5 ± 0.2 | 2.6 ± 0.1 | 5.8 ± 0.3 |
mUbRab5K116 | 2.4 ± 0.1 | 1.9 ± 0.1 | 8.0 ± 0.6 | 1.9 ± 0.2 | 5.0 ± 0.2 | 18.4 ± 0.8 |
mUbRab5K140 | 13.9 ± 1.8 | 52.3 ± 1.4 | 37.7 ± 4.9 | 1.8 ± 0.4 | 19.7 ± 0.8 | 108.0 ± 22.0 |
mUbRab5K165 | 3.2 ± 0.1 | 1.4 ± 0.1 | 4.4 ± 0.2 | 1.9 ± 0.8 | 0.9 ± 0.1 | 4.7 ± 0.5 |
-
* SEM, standard error of the mean
Discussion
Rab GTPases have recently been revealed to be ubiquitinated (Lachance et al., 2014; Qiu et al., 2016). However, the functional roles of ubiquitinated Rab GTPases have not been explored yet. Here, we demonstrated that Rab5 is monoubiquitinated at K116, K140, and K165 in two different cell lines (HEK 293T and HeLa). We also obtained structural ensemble models of the mUbRab5s by solution scattering. While the ubiquitin molecule on mUbRab5K140 was flexible, it was not flexible from the other two mUbRab5s. mUbRab5K140 shows decreased binding affinity to downstream effectors such as Rabaptin5 and EEA1, but its GEF-mediated GDP release and GTP loading activities were not changed. Through GDP release and association assays, we found that mUbRab5K165 had altered binding to GDP/GTP. Based on our data, we propose a molecular mechanism for how site-specific monoubiquitination of Rab5 acts as an inhibitory signal (Figure 7). The canonical regulatory mechanism of Rab5 involves a GEF- and GAP-mediated GDP/GTP conversion cycle and interaction between activated Rab5 and downstream effector proteins (Figure 7a). Monoubiquitination of K116 of Rab5 does not interfere with any part of the functional cycle (Figure 7b). By contrast, monoubiquitination of either K140 or K165 of Rab5 negatively regulates the functional cycle by either reducing the interaction of Rab5 with downstream effector proteins (K140 monoubiquitination; Figure 7c) or altering the intrinsic GDP/GTP conversion cycle (Figure 7d).

Proposed model of Rab5 regulation by site-specific monoubiquitination.
(a) Canonical Rab5 regulatory cycle. Inactive GDP-bound Rab5 is activated by a guanine nucleotide exchange factor (GEF such as Rabex-5) through the exchange of GDP to GTP. The activated GTP-bound Rab5 interacts with various effector proteins to form early endosomes. Activated Rab5 is switched-off by a GTPase activating protein (GAP) through the hydrolysis of GTP to GDP. (b) Monoubiquitination on K116 has no effect on the regulatory cycle. Thus, mUbRab5K116 is regulated in the same way as unmodified Rab5. (c) Monoubiquitination of K140 causes the ubiquitin moiety to impair the interaction of Rab5 with its effector proteins. Therefore, mUbRab5K140 cannot form early endosomes. (d) Monoubiquitination of K165 alters the GDP/GTP conversion cycle by interfering with the activities of GEF, thus lowering affinity to GTP. mUbRab5K165 cannot form early endosomes.
Our proposed model for the role of monoubiquitination in the functional cycle of Rab5 is clearly distinct from the case of Ras (Baker et al., 2013a), where monoubiquitination interferes with GAP activity, leading to activation of Ras signaling. Therefore, monoubiquitination seems to play different roles in Rab and Ras GTPases: monoubiquitination in Rab5 serves as an inhibitory signal, whereas it serves as a stimulatory or activating signal in Ras. It will be interesting to determine the ubiquitination sites of other Rab GTPases and to investigate how ubiquitination alters the intrinsic functions of these proteins. Proteomics studies to screen for ubiquitination sites of Rab5 have also indicated that other Rab GTPases are ubiquitinated. As Rab5 monoubiquitination is different from Ras monoubiquitination, other Rab GTPases might be regulated differently by site-specific ubiquitination. For these studies, the optimized chemical ubiquitination method described here will be useful. Recently, atypical serine ubiquitination was reported for Rab33b (Qiu et al., 2016; Bhogaraju et al., 2016; Bhogaraju and Dikic, 2016), but the underlying molecular mechanism has not been clarified. Our chemical ubiquitination method can be used to determine the functional role of serine ubiquitination in Rab GTPases. Further investigation is required to evaluate the diversity of roles of monoubiquitination among different small GTPase super families and/or among members of the same superfamily.
Advances in proteomic techniques have revealed the ubiquitination sites of a small number of Rab proteins. Analysis of published proteomics data revealed that 28 of 63 human Rab GTPases (including isoforms) are ubiquitinated in cells. When we compiled ubiquitination sites of Rab proteins reported by proteomic studies, we found that ubiquitination sites of Rab5 found in the current study - K116 and K140 - are located in non-conserved regions, while K165 is in the conserved G5 motif (xxxETSx’K’) (Figure 8a). It is plausible that ubiquitination of non-conserved lysine residues in Rab proteins and possibly small GTPases in general exhibits different functionalities. We noticed ubiquitination of K33 and K134 (numbering based on the Rab5 sequence) in some Rab proteins; both of these residues are critical for structural integrity (Figure 1e,g, Figure 1—figure supplement 1). Analysis of Rab5 proteins from eukaryotic species revealed different degrees of conservation of the three ubiquitination sites discovered in this study: both K116 and K165 are strictly conserved from yeast to human, whereas K140 is less conserved (Figure 8b). K140 is replaced by isoleucine in Drosophila and glutamate in yeast, implying that K140 might be a point of functional differentiation in these organisms. In contrast to eukaryotic Rab5 proteins, the three ubiquitination sites were not conserved in Rab5 homologue from Acanthamoeba polyphaga mimivirus (APMV). Recent crystal structure of APMV Rab5 revealed that K165 of hRab5a is substituted to F152 in APMV Rab5 and plays a major role for binding to GTP molecule (Ku et al., 2017). Therefore, it seems that monoubiquitination of Rab5 is a subsidiary signal for its regulation in eukaryotic cells. It is also intriguing that ubiquitination of a conserved site can have different functional consequences in Rab and Ras GTPase proteins. Monoubiquitination of K165 of Rab5 has an inhibitory effect, whereas ubiquitination of the corresponding K147 of Ras has a stimulatory effect (Baker et al., 2013a; Baker et al., 2013b). Therefore, monoubiquitination appears to be a versatile modification tool.

Multiple sequence alignment of human and mouse Rab GTPases reported to be ubiquitinated.
Rab GTPases in the ubiquitination database (mUbSiDa [Chen et al., 2014b]) were obtained, and their sequences were aligned to human Rab5A (this study) using Clustal X2 (Larkin et al., 2007). The graphical representation of the aligned sequences was prepared using BoxShade (ExPASy server). Ubiquitination sites from proteomic studies are indicated with the letter ‘K’ in the red background (K). Ubiquitination sites and structural key lysine residues revealed from this study are marked as inverse triangles in black () and green (
), respectively. ‘h’ and ‘m’ refer to human and mouse, respectively. (b) Rab5 GTPases from different species were aligned to human Rab5A. Ubiquitination sites from this study are indicated with the letter ‘K’ on a red background (K).
Oxidative conjugation of a ubiquitin moiety to a cysteine residue of a target protein has been widely used for chemical synthesis of ubiquitinated proteins. Despite the relative simplicity of this method in comparison with enzymatic conjugation, oxidative conjugation often suffers from incompleteness of the process, rendering subsequent separation of the ubiquitinated protein from unreacted protein very challenging (Baker et al., 2013b; Merkley et al., 2005). Such an inseparable mixture can obscure correct interpretation of biochemical and structural data. Here, we modified a pre-existing method by iterative addition of ubiquitinG76C to drive the chemical synthesis to completion. Our optimized method for chemical synthesis of ubiquitinated proteins will facilitate studies on the effects of protein ubiquitination at the molecular level when no identified E3 ubiquitin ligase is available or a large quantity of ubiquitinated protein is desired.
In this study, we found that monoubiquitination of Rab5 had an inhibitory effect on the regulatory cycle of Rab5. However, we did not find a stimulus signal or E3 ligase(s) responsible for monoubiquitination of Rab5. As result, only a small portion of Rab5 could be ubiquitinated when ubiquitin was overexpressed. These results suggest that ubiquitination of Rab5 occurs rarely in cells under basal conditions and may explain why the ubiquitination of Rab GTPases has not been studied until recently (de la Vega et al., 2011). Despite the small population of monoubiquitinated Rab5, the effects of monoubiquitination of Rab5 were strong. We observed dramatic failure of localization of Rab5 to the endosome and EGFR degradation pathway. This suggests that monoubiquitination of Rab5 is tightly regulated in the cell through specific E3 ligase or DUB enzymes. It will be of great interest to determine how monoubiquitination of Rab5 is regulated in cells.
Materials and methods
Plasmids and cell culture
Request a detailed protocolThe pCDNA-GFP-Rab5 clone was provided by Dr. Yoshikatsu Aikawa at Doshisha University, Japan. The pGEX-EEA136-91 clone was obtained from Dr. Bonsu Ku at the Korea Research Institute of Bioscience and Biotechnology (KRIBB). cDNA clones of Rabaptin5 and Rabex-5 were purchased from Open Biosystems. The Rab7 gene was chemically synthesized (Bioneer). Each gene was amplified and cloned into bacterial expression vectors (parallel-His, parallel-GST (Sheffield et al., 1999), and pQE30) or mammalian expression vectors (pCDNA-FLAG and pCDNA-HA). Human embryonic kidney 293T (HEK 293T) and HeLa cells were purchased from American Type Culture Collection (RRID: CVCL_0063 and CVCL_0030, respectively) and maintained in Dulbecco’s modified Eagle’s medium (DMEM) supplemented with 10% fetal bovine serum (heat-inactivated). Plasmids were transfected into cells using Lipofectamine 2000 (Invitrogen) for HeLa cells and PEI 25000 (23966, Polysciences) for HEK 293T cells. Mycoplasma detection kit (LT07-318, Lonza, Switzerland) was regularly used to check absence of mycoplasma.
Protein expression and purification
Request a detailed protocolGST-ubiquitin G76C, His-Rab5 (WT and mutants), and His-Rabex-5132-393 were expressed in E. coli BL21(DE3) cells. GST-Rabaptin5551-862 and GST-EEA136-91 were expressed in E. coli Rosetta2 (DE3) cells. Cells were inoculated and grown to an OD600 of 0.6–0.8 at 37 ˚C. To induce protein expression, 0.5 mM isopropyl d-thiogalactoside (IPTG) was added to the cultures, and they were incubated at 20 ˚C overnight. Induced cells were harvested by centrifugation at 4,000 rpm. Cells were resuspended in buffer A (50 mM Tris-HCl pH 7.5 and 150 mM NaCl). For GST-Ub G76C and His-Rab5 K-to-C mutants, 2 mM TCEP was added during the purification procedure. Resuspended cells were lysed by sonication and centrifuged at 13,000 rpm. The supernatant was applied to glutathione-S-sepharose resin (GE HealthCare) for GST fusion proteins and Ni-NTA-agarose resin (Qiagen) for His-tagged proteins. Non-specific proteins were washed out with buffer B (50 mM Tris-HCl pH 7.5 and 500 mM NaCl) for GST fusion proteins and buffer C (50 mM Tris-HCl pH 7.5, 500 M NaCl, and 20 mM imidazole) for His-tagged proteins. To elute target proteins, buffer D (50 mM Tris-HCl pH 8.0 and 50 mM NaCl, supplemented with 15 mM reduced glutathione) was used for GST-fusion proteins, and buffer E (50 mM Tris-HCl pH 7.5, 300 mM imidazole, and 500 mM NaCl) was used for His-tagged proteins. GST and the His-tag were cleaved from the fusion proteins using GFP-tagged tobacco etch virus protease in cleavage buffer (25 mM Tris-HCl pH 7.5, 75 mM NaCl, and 0.5 mM EDTA) (Wu et al., 2009). Cleaved samples were applied to glutathione-S-sepharose or Ni-NTA-agarose to remove the released His- or GST- tags. Proteins were further purified by size exclusion chromatography on a Superdex 75 or 200 prep grade 16/60 column (GE HealthCare). Fractions were analyzed by SDS-PAGE, and samples were pooled and concentrated by centrifugal concentrators (Amicon Ultra 3, 10, or 30 kDa, Millipore). Final concentration was determined by either absorbance at 280 nm or Bradford assay.
Ubiquitination assay
Request a detailed protocolUbiquitinated Rab5 was detected based on previously described methods (Choo and Zhang, 2009; Jung et al., 2013). HA-ubiquitin and FLAG-Rab5 were co-transfected into HEK 293T cells. After 24 hr, cells were washed with ice-cold phosphate-buffered saline (PBS) twice and lysed with lysis buffer LA (2% SDS, 150 mM NaCl, 10 mM Tris-HCl, pH 8.0, 2 mM sodium ortho-vanadate, 50 mM sodium fluoride, and a protease inhibitor cocktail from Roche). Lysed cells were transferred into a tube and boiled for 10 min at 95 ˚C. The boiled sample was then sonicated (microtip 40% amplitude, pulse 1 s on/1 s off, 20 s, VC 750) and diluted with dilution buffer (150 mM NaCl, 10 mM Tris-HCl, pH 8.0, 2 mM EDTA, and 1% Triton X-100). After centrifugation at 16,200 g for 10 min, the resulting supernatant was pre-cleared with mouse IgG and protein G-beads (GeneDepot). Pre-cleared lysates were immuno-precipitated with FLAG antibody (F1804, Sigma Aldrich, RRID:AB_262044) and immunoblotted with HA-HRP (3F10, Roche, RRID:AB_2314622) or FLAG-IgG-rabbit (F7425, Sigma Aldrich, RRID:AB_439687) followed by Rabbit-IgG-HRP (ADI-SAB-300-J, Enzo Life Sciences, RRID:AB_11179983). For the geranylgeranyl transferase inhibitor assay, 20 μM GGTI-298 (G5169, Sigma) was added for 6 hr after transfection. Twenty-four hours after GGTI treatment, cells were harvested and subjected to ubiquitination assay. These experiments were repeated three times.
Cytosol/membrane fractionation assay
Request a detailed protocolCytosol/membrane fractionation assay was performed according to a previously described method (Baghirova et al., 2015). Briefly, HEK 293T cells were harvested by trypsin treatment and washed with ice-cold PBS. Fractionation buffer FA (150 mM NaCl, 50 mM HEPES pH 7.4, 100 μg/ml digitonin (D141, Sigma), and 1 M hexylene glycol) supplemented with 1x protease inhibitor cocktail (P8340, Sigma, 100x stock solution) was added to the harvested cells to release cytosolic proteins. After incubation for 15 min on an end-to-end rotator at 4 ˚C, samples were centrifuged at 2,000 g for 10 min. The resulting supernatant was transferred and marked as cytosolic proteins. Fractionation buffer FB (150 mM NaCl, 50 mM HEPES pH 7.4, 1% (v/v) Nonidet-P40, and 1 M hexylene glycol) supplemented with 1x protease inhibitor cocktail (P8430, Sigma, 100x stock solution) was added to the resulting cell pellets to release membrane proteins. After incubation for 30 min on ice, samples were centrifuged at 7,000 g for 10 min. The resulting supernatant was transferred and marked as membrane proteins. Both cytosolic and membrane fractions were further immunoprecipitated using a FLAG-M2-affinity gel (A2220, Sigma, RRID:AB_10063035) and analyzed by immunoblotting. These experiments were repeated twice.
EGF-induced EGFR degradation assay
Request a detailed protocolEGFR degradation was assessed based on previously described methods (Hoeller et al., 2006; Balaji et al., 2012; Smith et al., 2013). Briefly, HeLa cells were transfected with FLAG-Rab5 WT, K116R, K140R, or K165R with HA-ubiquitin. After 24 hr of transfection, cells were treated with EGF (20 ng/ml, final concentration) and harvested at the indicated time points after treatment. Harvested cells were lysed with lysis buffer LB (50 mM Tris-HCl pH 7.5, 1% (v/v) NP-40, 150 mM NaCl, and 10% (v/v) glycerol) supplemented with a protease inhibitor cocktail (05-056-489-001, Roche) and subjected to immunoblotting with anti-EGFR antibody (SC-03-rabbit, Santa Cruz, RRID:AB_631420). These experiments were repeated three times.
Chemical ubiquitination of Rab5
Request a detailed protocolChemical ubiquitination of Rab5 by formation of a disulfide bond between ubiquitin and a specific lysine residue of Rab5 was based on previously published studies (Baker et al., 2013a; Baker et al., 2013b; Merkley et al., 2005). Purified ubiquitin G76C and a Rab5 K-to-C mutant (K116C, K140C, or K165C) were stored in reducing buffer (2 mM TCEP, 50 mM Tris-HCl pH 7.5, and 150 mM NaCl). The proteins were then mixed at a 1:5 molar ratio (Rab5: ubiquitin) and dialyzed with oxidizing buffer (50 mM Tris-HCl pH 7.5, 150 mM NaCl, 5 mM MgCl2, and 20 μM CuCl2). During dialysis, ubiquitin G76C was added to the dialysis bag at regular intervals. Iterative additions of ubiquitin resulted in 95% yield of monoubiquitinated Rab5 (mUbRab5). Production of mUbRab5 was monitored by SDS-PAGE under non-reducing conditions. Finally, mUbRab5 was purified by size-exclusion chromatography on a Superdex 75 16/60 prep grade column (GE HealthCare) pre-equilibrated with buffer F (50 mM Tris-HCl pH 7.5, 150 mM NaCl, and 10 mM MgCl2).
Small angle X-ray scattering (SAXS) measurement and data processing
Request a detailed protocolProtein samples were concentrated in Amicon ultra centrifugal devices (3 kDa cut-off). Final concentrations were measured by absorbance at 280 nm. Buffer F was used to record the reference buffer scattering profile. Each sample was measured six times and monitored for radiation damage. Samples were also diluted (two-, four-, and eight-fold) with buffer F to check for concentration dependency. Scattering profile was generated using in-house software of the Pohang Accelerator Laboratory. Rg (radius of gyration) and I(0) from the Guinier plot were calculated using AUTORG (Svergun, 1992). GNOM was used to calculate the pair distribution function and Porod volume (Svergun, 1992). Molecular ensembles were generated by EOM (Tria et al., 2015). Final ensemble models were cross-validated with the FoXS server minimal ensemble search MES (Pelikan et al., 2009). Since EOM accepted only single polypeptide chain with multiple domains, we generated the sequences and atomic coordinates of the mUbRab5s in the following four steps. Firstly, the input sequence started with the ubiquitin sequence followed by the partial Rab5 sequence which covered from the target lysine of Rab5 to the C-terminus and then the remaining Rab5 sequence ranging from the N-terminus to the residue before the target lysine of Rab5. For example, the input sequence for mUbRab5K140 was as follows: (N-[Ubiquitin residues 1–76]-[K140-C-term of Rab5]-[N-term to residue 139 of Rab5]-C). Using this sequence, we were able to use EOM to analyze the ubiquitin moiety of mUbRab5 as a domain of a single polypeptide chain. Secondly, the input coordinates were prepared following the same way as in the preparation of the input sequences. Crystal structures of ubiquitin and Rab5 were used as the coordinates for domains 1 and 2, respectively (PDB ID: 1UBQ, 3MJH). The residues of Rab5 were re-numbered so that the new numbering was the same as the numbering of the prepared sequence. For example, K140 of Rab5 was re-numbered K77. Thirdly, we determined how many residues in the C-terminal region of ubiquitin should be allowed to be flexible. To define flexible residues in ubiquitin, we analyzed crystal/NMR structures of ubiquitin (PDB IDs: 1UBQ and 1D3Z). In both cases, the C-terminal five residues of ubiquitin (residues 72–76) after the β3 strand were not assigned to secondary structures, which imply that those residues are flexible. Since Arg 72 of ubiquitin is juxtaposed at the end of the β3 strand, Arg 72 was retained in the input coordinate file. Subsequently, four residues of ubiquitin (residues 73–76) were allowed to be flexible and removed in the input coordinate file. These C-terminal four residues of ubiquitin were used for flexibility analysis through EOM in our previous study on linear ubiquitin chains (Thach et al., 2016). Finally, we used the native isopeptide linkage for ensemble model calculations in EOM since a disulfide bond linkage is known to resemble the native ubiquitination linkage (Baker et al., 2013a).
GEF-mediated GDP release and GTP loading assays
Request a detailed protocolGDP release from Rab5 or mUbRab5 was measured using 2’-(or-3’)-O-(N-methylanthraniloyl) (MANT)-GDP (Molecular Probes) (Delprato and Lambright, 2007; Baker et al., 2013a). One micromolar Rab5 or mUbRab5 was loaded with 1 μM MANT-GDP for 30 min at room temperature in buffer F in the presence or absence of 1 μM Rabex-5132-391. To initiate MANT-GDP release, 1 mM GDP was added to the solution. Fluorescence intensities were measured (excitation at 360 nm and emission at 450 nm) on a TECAN M200 multi-plate reader, analyzed, and fitted to a one-phase exponential decay curve with GraphPad Prism software. GTP loading was measured using MANT-GTP. Two micromolar Rab5 or mUbRab5 was mixed with 1 μM of MANT-GTP in buffer F and stabilized for 10 min. To facilitate GTP loading by GEF, 1 μM Rabex-5132-391 was added to the solution. Fluorescence intensities were measured (excitation at 360 nm and emission at 450 nm) on a TECAN M200 multi-plate reader, analyzed, and fitted to a one-phase exponential association curve using GraphPad Prism software. These experiments were repeated three times.
Effector protein binding assay by biolayer interferometry (BLI)
Request a detailed protocolBLI experiments were performed on a BLitz system (ForteBio). For this, 0.5 µM GST-Rabaptin5551-862 or GST-EEA136-91 was immobilized on anti-GST sensors. After baseline stabilization with reaction buffer (50 mM Tris-HCl, 150 mM NaCl, 10 mM MgCl2, and 1 mM 5'-guanylyl imidodiphosphate, a non-hydrolysable GTP analog (GppNHP, Jena Bioscience, Germany)), GppNHP-loaded Rab5 or mUbRab5 at different concentrations was loaded until signal saturation, and this experiment was repeated twice. To initiate dissociation, the sensor was then placed into in reaction buffer again. Association rate constant (kon), dissociation rate constant (koff), and SEM (standard error of mean) values were calculated by BLItz Pro (ForteBio) based on sensorgram results. Dissociation constants (Kd) and SEMs of dissociation constants were calculated by the following equation:
where Δ refers to the SEMs of Kd, kon, and koff, respectively.
Immunofluorescence
Request a detailed protocolAfter transfection, cells were rinsed in PBS and fixed for 10 min at room temperature with 3% paraformaldehyde. Cells were then permeabilized with 0.1% Triton X-100 in PBS for 5 min and blocked with 2% bovine serum in PBS for 1 hr at room temperature. For immunostaining, cells were incubated with antibodies diluted in 2% bovine serum at room temperature for 1 hr, then rinsed with PBS and incubated with either anti-rabbit-Rhodamine or anti-mouse-Alexa-Fluor-633-conjugated secondary antibodies (R6394;RRID:AB_1500693, A11001;RRID:AB_2534069, respectively, Life Technologies,) for 30 min at room temperature. Cells were washed three times for 5 min with PBS. Images were obtained using a LSM-700 (Zeiss).
References
-
Spatiotemporal regulation of the ubiquitinated cargo-binding activity of Rabex-5 in the endocytic pathwayJournal of Biological Chemistry 287:40586–40597.https://doi.org/10.1074/jbc.M112.411793
-
Ubiquitin chain diversity at a glanceJournal of Cell Science 129:875–880.https://doi.org/10.1242/jcs.183954
-
Site-specific monoubiquitination activates Ras by impeding GTPase-activating protein functionNature Structural & Molecular Biology 20:46–52.https://doi.org/10.1038/nsmb.2430
-
Differences in the regulation of K-Ras and H-Ras isoforms by monoubiquitinationJournal of Biological Chemistry 288:36856–36862.https://doi.org/10.1074/jbc.C113.525691
-
RIN1 orchestrates the activation of RAB5 GTPases and ABL tyrosine kinases to determine the fate of EGFRJournal of Cell Science 125:5887–5896.https://doi.org/10.1242/jcs.113688
-
Epidermal growth factor and membrane trafficking. EGF receptor activation of endocytosis requires Rab5aThe Journal of Cell Biology 151:539–550.https://doi.org/10.1083/jcb.151.3.539
-
A novel interaction between Rab7b and actomyosin reveals a dual role in intracellular transport and cell migrationJournal of Cell Science 127:4927–4939.https://doi.org/10.1242/jcs.155861
-
Rab5 isoforms differentially regulate the trafficking and degradation of epidermal growth factor receptorsJournal of Biological Chemistry 284:30328–30338.https://doi.org/10.1074/jbc.M109.034546
-
Regulation of small GTPases by GEFs, GAPs, and GDIsPhysiological Reviews 93:269–309.https://doi.org/10.1152/physrev.00003.2012
-
Detection of Protein UbiquitinationJournal of Visualized Experiments e1293.https://doi.org/10.3791/1293
-
Structural basis for Rab GTPase activation by VPS9 domain exchange factorsNature Structural & Molecular Biology 14:406–412.https://doi.org/10.1038/nsmb1232
-
Structure of monoubiquitinated PCNA and implications for translesion synthesis and DNA polymerase exchangeNature Structural & Molecular Biology 17:479–484.https://doi.org/10.1038/nsmb.1776
-
Caveolin-1 is ubiquitinated and targeted to intralumenal vesicles in endolysosomes for degradationThe Journal of Cell Biology 191:615–629.https://doi.org/10.1083/jcb.201003086
-
Protein regulation by monoubiquitinNature Reviews. Molecular Cell Biology 2:195–201.https://doi.org/10.1038/35056583
-
Regulation of ubiquitin-binding proteins by monoubiquitinationNature Cell Biology 8:163–169.https://doi.org/10.1038/ncb1354
-
Functional and structural interactions of the Rab5 D136N mutant with xanthine nucleotidesBiochemical and Biophysical Research Communications 215:241–249.https://doi.org/10.1006/bbrc.1995.2459
-
Role of Rab GTPases in membrane traffic and cell physiologyPhysiological Reviews 91:119–149.https://doi.org/10.1152/physrev.00059.2009
-
RIN3: a novel Rab5 GEF interacting with amphiphysin II involved in the early endocytic pathwayJournal of Cell Science 116:4159–4168.https://doi.org/10.1242/jcs.00718
-
Ubiquitination of the N-terminal region of caveolin-1 regulates endosomal sorting by the VCP/p97 AAA-ATPaseJournal of Biological Chemistry 288:7363–7372.https://doi.org/10.1074/jbc.M112.429076
-
The ubiquitin codeAnnual Review of Biochemistry 81:203–229.https://doi.org/10.1146/annurev-biochem-060310-170328
-
Ubiquitylation and activation of a Rab GTPase is promoted by a β₂AR-HACE1 complexJournal of Cell Science 127:111–123.https://doi.org/10.1242/jcs.132944
-
Clustal w and clustal x version 2.0Bioinformatics 23:2947–2948.https://doi.org/10.1093/bioinformatics/btm404
-
Molecular architecture of the Ub-PCNA/Pol η complex bound to DNAScientific Reports 5:15759.https://doi.org/10.1038/srep15759
-
Monoubiquitination Inhibits the Actin Bundling Activity of FascinJournal of Biological Chemistry 291:27323–27333.https://doi.org/10.1074/jbc.M116.767640
-
Ubiquitin manipulation by an E2 conjugating enzyme using a novel covalent intermediateJournal of Biological Chemistry 280:31732–31738.https://doi.org/10.1074/jbc.M505205200
-
GTPase networks in membrane trafficAnnual Review of Biochemistry 81:637–659.https://doi.org/10.1146/annurev-biochem-052810-093700
-
Structure and flexibility within proteins as identified through small angle X-ray scatteringGeneral Physiology and Biophysics 28:174–189.https://doi.org/10.4149/gpb_2009_02_174
-
Structural clues to Rab GTPase functional diversityJournal of Biological Chemistry 280:15485–15488.https://doi.org/10.1074/jbc.R500003200
-
Ubiquitin-dependent sorting in endocytosisCold Spring Harbor Perspectives in Biology 6:a016808–016821.https://doi.org/10.1101/cshperspect.a016808
-
Selective membrane recruitment of EEA1 suggests a role in directional transport of clathrin-coated vesicles to early endosomesJournal of Biological Chemistry 275:3745–3748.https://doi.org/10.1074/jbc.275.6.3745
-
Overcoming expression and purification problems of RhoGDI using a family of "parallel" expression vectorsProtein Expression and Purification 15:34–39.https://doi.org/10.1006/prep.1998.1003
-
The E3 ubiquitin ligases RNF126 and Rabring7 regulate endosomal sorting of the epidermal growth factor receptorJournal of Cell Science 126:1366–1380.https://doi.org/10.1242/jcs.116129
-
Inhibition of rab5 GTPase activity stimulates membrane fusion in endocytosis.The EMBO Journal 13:1287–1296.
-
Determination of the regularization parameter in indirect-transform methods using perceptual criteriaJournal of Applied Crystallography 25:495–503.https://doi.org/10.1107/S0021889892001663
-
New conformations of linear polyubiquitin chains from crystallographic and solution-scattering studies expand the conformational space of polyubiquitinActa Crystallographica Section D Structural Biology 72:524–535.https://doi.org/10.1107/S2059798316001510
-
Traceless and site-specific ubiquitination of recombinant proteinsJournal of the American Chemical Society 133:10708–10711.https://doi.org/10.1021/ja202799r
-
A proteome-wide, quantitative survey of in vivo ubiquitylation sites reveals widespread regulatory rolesMolecular & Cellular Proteomics 10:M111.013284.https://doi.org/10.1074/mcp.M111.013284
-
Proteomic analyses reveal divergent ubiquitylation site patterns in murine tissuesMolecular & Cellular Proteomics 11:1578–1585.https://doi.org/10.1074/mcp.M112.017905
-
Aborted autophagy and nonapoptotic death induced by farnesyl transferase inhibitor and lovastatinJournal of Pharmacology and Experimental Therapeutics 337:65–74.https://doi.org/10.1124/jpet.110.174573
-
A Novel Method for High-Level Production of TEV Protease by Superfolder GFP TagJournal of Biomedicine & Biotechnology 2009:591923.https://doi.org/10.1155/2009/591923
-
Structural basis of Rab5-Rabaptin5 interaction in endocytosisNature Structural & Molecular Biology 11:975–983.https://doi.org/10.1038/nsmb832
-
Rabaptin-5-independent membrane targeting and Rab5 activation by Rabex-5 in the cellMolecular Biology of the Cell 18:4119–4128.https://doi.org/10.1091/mbc.E07-02-0100
Decision letter
-
Suzanne R PfefferReviewing Editor; Stanford University School of Medicine, United States
In the interests of transparency, eLife includes the editorial decision letter and accompanying author responses. A lightly edited version of the letter sent to the authors after peer review is shown, indicating the most substantive concerns; minor comments are not usually included.
Thank you for submitting your article "Site–specific monoubiquitination downregulates Rab5 by disturbing effector binding and guanine nucleotide conversion" for consideration by eLife. Your article has been reviewed by three peer reviewers, one of whom is a member of our Board of Reviewing Editors and the evaluation has been overseen by Ivan Dikic as the Senior Editor. The following individuals involved in review of your submission have agreed to reveal their identity: Marino Zerial (Reviewer #2); David Lambright (Reviewer #3).
You will see that the reviewers believe that the work is potentially suitable for presentation in eLife. However, they describe a number of aspects that need to be fully addressed and they noted that the manuscript needs rewriting, perhaps with help from a native English speaker. I enclose the full reviews to help in recrafting this story for submission and re–evaluation for presentation in eLife. We hope you find the comments constructive; note that the ubiquitination of Rab5 is already reported in PhosphoSite Plus mass spec database (when you respond to reviewer 3); the Rab7 story can be omitted.
Reviewer #1
This is an interesting paper reporting the consequences of Rab5 and Rab7 monoubiquitynation on Rab protein function. The authors try to combine cell biological assays and biochemical analysis with SAXS structural analysis, the kind of broad mix of approaches that eLife would like to present. They present a method for efficient in vitro generation of pure mono–ubiquitin conjugates. It is likely that the authors have shown that mono–ubiquitination decreases Rab activity when driven in cells by artificial expression of extra HA–ubiquitin. This is shown with direct binding assays for Rab5 effectors and some light microscopy.
Unfortunately, it appears that the figures were photographs of drawings and are of too poor resolution for publication and in some cases even for reviewer evaluation. This reviewer cannot see any EEA1 staining for example, as the images are out of focus or off too low a resolution to see even blown up on a computer screen. Also, the language is difficult to understand and the manuscript will need significant editing and rewriting for clarity. The EGFR degradation needs to be presented graphically so that half times for degradation relative to time zero can be compared.
I will await the other referee comments and hope that this story can be presented more clearly so that its findings can be appreciated by the scientific community and brought up to the standard required for presentation in eLife.
(Please use photoshop, high resolution and a drawing program to label figures such as Adobe Illustrator)
Quantitation of the cell biology data is missing
Reviewer #2
In this study, the authors employ a range of techniques to investigate Rab5 monoubiquitination both functionally and structurally, culminating in a model of how differential monoubiquitination of Rab5 has different effects on the downstream Rab5 machinery and dynamics at the early endosome. While the premise of this manuscript is highly interesting and is in the scope of eLife, I do not believe that the manuscript in the current form meets the standards of this journal. With sufficient revisions, however, this could be a conceptually novel and compelling paper.
In general, the story presented here is not very coherent. The authors have performed all the right experiments to present a good case for their model. However, the experiments are often poorly described and presented in an order that is only logical if you already know the outcome of later experiments. The authors should take care to present their story in a stepwise and logical manner and to explain in detail both the experimental approaches and underlying logic.
General points:
1) It is essential to be clear about which cell lines are used and also provide the rationale for using two different cell lines for the biochemistry and microscopy experiments. This approach is not coherent. Similarly, the change from cell work to recombinant proteins is not made explicit in the text.
2) The authors should go through the manuscript carefully to fix grammatical mistakes. Especially in the case of subsection “Monoubiquitination on K140 of Rab5 decreases its binding affinity to effector prot”, grammatical errors may confuse the message the authors are trying to convey.
There are several Rab5 isoforms and it must be made clear from the start which one is being used and provide a sound rationale for using that specific isoform alone, as the three isoforms show differential ubiquitination patterns.
Next, the authors use GGTI–198, a GGTase inhibitor, to rule out prenylation as the source of the double bands upon ubiquitination. However, this experiment is not sufficiently described in the figure legend or in the text. What are the two different IP FLAG plots in Figure 1B, what do the labels short and long mean?
Importantly, it is not clear to the reader why out of all lysines in Rab5, K33 and K165 were chosen in addition to the known K116, K134 and K140 sites from protEOMics. There is some explanation for K165 in sequence homology to K147 in Ras, but otherwise the authors make no effort to explain their rationale. This section should be reordered to make it more logical. I would suggest reordering the first part of the manuscript: the story would flow better if the validation of protEOMics hits preceded the investigation of the now proven ubiquitination sites. For example, instead of starting with candidate ubiquitination sites and an analysis of crystal structures in subsection “Rab5 undergoes monoubiquitination at multiple sites”, it would be more logical to begin the experiments presented in Figure 1D as a validation of published protEOMics data. From there the story can then focus on the relevant monoubiquitination sites and their location on the Rab5 structure (Figure 1—figure supplement 1A).
Furthermore, the experiments done with K to E mutations and protein expression are not essential and should be moved to the supplement or left out entirely so as not to distract from the more sophisticated K to R mutation experiments.
In Figure 2C, the authors show that monoubiquitinated Rab5 (mUbRab5) localises to membranes. This raises many questions: mUbRab5 is a very small percentage of total membrane Rab5, why do the authors observe such a dramatic effect on recycling/degradation? This would seem to be a highly significant result but is not addressed at all. In addition to this, Figure 2B shows a dramatic removal of Rab5 from EEA1 positive endosomes upon ubiquitination and indeed in the text the authors refer to this as 'diffusing to cytosol' which would not agree with the results presented in Figure 2C.
If K134 mutations are not viable, how has this been identified as a ubiquitination site in vivo? The authors should comment on this.
The authors introduce the immunofluorescence of Rab5 endosomes and justify the use of HeLa cells in subsection “Monoubiquitination on K140 and K165 of Rab5 plays negative roles in endocytic pathway”, but results from these experiments had already been shown in subsection “Rab5 undergoes monoubiquitination at multiple sites” without explanation of what Rab5 puncta are or what cell line was used.
In Figure 2A and C, UbRab5 does not appear as a double band as in other blots, why is this?
Subsection “Monoubiquitination on K140 and K165 of Rab5 plays negative roles in endocytic pathway” is a bit of a reach of interpretation at this point in the manuscript and are only logical in hindsight, knowing the results that are still to come.
From the SAXS analysis the authors derive that mUbRab5K140 features two conformations while mUbRab5K116 and mUbRab5K165 have only one conformation. However, the authors also state in the text that mUbRab5 is structurally very flexible, contradicting the existence of distinct conformations. To better evaluate the data, the authors need to explain in detail how distinct conformations were extracted from the data using ensemble–optimized methods (EOM) and how the final models were derived (Figure 4G, H and I). Furthermore, the authors need to show the shape of the particle derived from the SAXS data, not only the model. In general, the unprocessed SAXS data are missing and the reader is presented immediately with something highly processed. For a more coherent story, the intermediate stages of deriving these structural models must be presented.
From structural models the authors derive that the close proximity of the ubiquitin moieties in mUbRab5K140 to the switch I region of Rab5 might disturb Rabaptin–5 and EEA1 binding, while interaction of Rab5 with GEF (Rabex–5) and GAP (RabGAP–5) is unchanged. This especially needs clarification, because the switch I region is involved in nucleotide binding as well, and hence prone to be influenced by changes in nucleotide state. Knowing the outcome of the binding experiments described later, this conclusion makes sense, but when reading this manuscript for this first time this appears a somewhat illogical conclusion.
Reviewer #3
This manuscript describes experiments examining monoubiquitination of Rab5 when overexpressed with ubiquitin and affects of monoubiquitination of the specific–lysine residues on Rab5 structure, localization, EGFR degradation, intrinsic and Rabex–5 catalyzed nucleotide exchange, and binding of effectors (Rabaptin5 and EEA1). Similar cell–based experiments were performed to examine Rab7 ubiquitination. The identification of putative monoubiquitination sites, the characterization of homogenously monoubiquitinated Rab5 by SAXS, and the effects of monoubiquitination on GEF activity, effector binding affinity, localization, and EGFR degradation suggest that monoubiquitination of at least two lysines could potentially roles regulate Rab5 and Rab7 functions.
All of the experiments involve either purified proteins or transfections of cell lines. There are no experiments assessing whether the endogenous Rab proteins are monoubiquinated (i.e. without ubiquitin overexpression) and under what conditions. This limitation should be discussed, and conclusions about the potential biological significance appropriately qualified.
Subsection “Rab5 undergoes monoubiquitination at multiple sites”, K33 and K134 are indeed crucial but this is already known, indeed from the very first structures of EF–Tu and Ras. K33 is the P–loop lysine in the GxxxxGKS/T motif; K134 is the invariant lysine in the NKxD G4 motif. The role and interactions of these residues in GTPases in general has been extensively documented. Its not clear why these invariant residues were chosen for mutational analysis. Expression in inclusion bodies in bacteria isn't particularly informative or unexpected. The rationale for generating lysine–to–glutamate mutations is also unclear as these mutations don't mimic ubiquitination.
How can anything be concluded about inhibition of Rab5 activation (subsection “Monoubiquitination on K140 and K165 of Rab5 plays negative roles in endocytic pathway”) from the IF and fractionation experiments in Figure 2B and 2C? These experiments do not indicate whether ubiquitination affects the nucleotide status of Rab5. Perhaps the authors meant inhibition of a Rab5 function rather than activation of Rab5 per se. Needs clarification.
Overexpression of ubiquitin presumably affects the function of many targets including proteins other than Rab5 that are also important for endocytic trafficking. Does overexpression of ubiquitin alone have any effects on (for example) EGFR degration?
The cell-based experiments lack quantification and analysis of statistical significance.
The resolution of the images in the PDF is low, making it difficult to evaluate the quality of these experiments.
The buffer–subtracted SAXS curves should be shown somewhere in the main or supplemental figures. Accurate modeling of SAXS data, in particular by ensemble approaches, requires data with high signal to noise that typically extends to a high resolution cutoff near the first Porod minimum. The data and calculated curves in Figure 4 are truncated in the Kratky region and the signal–to–noise does not appear to be very high. Are the data of sufficient quality for ensemble modeling?
No details are provided regarding the EOM modeling. For example, how many residues we're allowed to be flexible and what criteria where to determine which residues show be flexible? Do the results change if additional residues are allowed to be flexible? A narrow selected distribution near the high Rg and Dmax tails of the pools in Figure 4 suggest that the number of flexible residues used to generate the pool might not be adequate to represent the flexibility in solution. Also unclear is how the linkage between Rab5 and ubiquitin is modeled. Does the EOM model disulfide linkages? If not, how was the linkage modeled? The details of the modeling are important and how they affect the results needs to be evaluated and discussed.
"Interestingly, none of the EOM–derived structural models for mUbRab5s implicates any direct interaction between ubiquitin and Rab5" (subsection “Structural models of mUbRab5s derived by solution scattering”). Is EOM capable of producing models with direct interactions? Even if it could, what would it mean at the low resolution of SAXS?
Figure 4—figure supplement 1: Kratky–Debye and Porod–Debye plots indicate how extended a molecule is. Flexibility is inferred. Presumably the same results would be obtained from the calculated EOM curves that fit the data. The question is whether the SAXS data can distinguish flexible vs. extended. The interpretation should be more nuanced.
https://doi.org/10.7554/eLife.29154.028Author response
Reviewer #1
[…] Unfortunately, it appears that the figures were photographs of drawings and are of too poor resolution for publication and in some cases even for reviewer evaluation.
We are submitting our revised manuscript with high resolution image files.
This reviewer cannot see any EEA1 staining for example, as the images are out of focus or off too low a resolution to see even blown up on a computer screen. Also, the language is difficult to understand and the manuscript will need significant editing and rewriting for clarity.
The revised manuscript was re–written and has been edited by native English speaker. We would be delighted to provide you a certificate by the editing company upon request.
The EGFR degradation needs to be presented graphically so that half times for degradation relative to time zero can be compared.
The graphical representation of EGFR degradation has been included in the revised manuscript (Figure 2D).
I will await the other referee comments and hope that this story can be presented more clearly so that its findings can be appreciated by the scientific community and brought up to the standard required for presentation in eLife.
(Please use photoshop, high resolution and a drawing program to label figures such as Adobe Illustrator)
We are now providing high resolution images prepared by Adobe Photoshop and Illustrator as the reviewer requested.
Quantitation of the cell biology data is missing
We have added statistics and the significance for cell-based experiments in our revised manuscript (Figure 2B, G, H).
Reviewer #2
[…] 1) It is essential to be clear about which cell lines are used and also provide the rationale for using two different cell lines for the biochemistry and microscopy experiments. This approach is not coherent. Similarly, the change from cell work to recombinant proteins is not made explicit in the text.
The rationale for the transition from HEK293T cells to HeLa cells is described in the revised manuscript as follows:
(Subsection “Rab5 undergoes monoubiquitination at multiple sites”) “Subsequently, we generated an 11KR mutant with intact K33 and K134 and examined its function by observing Rab5–positive endosome puncta in HeLa cells to visualize Rab5–positive endosomes more efficiently (Miaczynska et al., 2004, Borg et al., 2014, Kajiho et al., 2003).”
The rationale for the transition from cell work to recombinant proteins is described in the revised manuscript as follows:
(Subsection “Modified chemical ubiquitination with iterative ubiquitin addition to obtain fully monoubiquitinated Rab5 protein”) “To understand how monoubiquitinated Rab5s negatively affect the Rab5 regulatory cycle at molecular level, we undertook biochemical studies of monoubiquitinated Rab5s.”
2) The authors should go through the manuscript carefully to fix grammatical mistakes. Especially in the case of subsection “Monoubiquitination on K140 of Rab5 decreases its binding affinity to effector prot”, grammatical errors may confuse the message the authors are trying to convey.
We thank the reviewer for careful reading of the manuscript. The revised manuscript has been edited by a native English speaker through a professional editing service company.
There are several Rab5 isoforms and it must be made clear from the start which one is being used and provide a sound rationale for using that specific isoform alone, as the three isoforms show differential ubiquitination patterns.
We used Rab5a in this study. Since Rab5a is extensively studied compared to Rab5b and Rab5c, we decided to use the Rab5a isoform. We revised the manuscript as follows:
(Introduction) “Three isoforms of mammalian Rab5 – Rab5a, Rab5b and Rab5c regulate endocytosis in a co–operative manner (Bucci et al., 1995). However, recent studies suggest that these three isoforms may have differential roles. Rab5a is required for EGF–induced EGFR degradation while Rab5b and Rab5c are not (Barbieri et al., 2000a, Chen et al., 2009). Rab5c is involved in cell migration whereas Rab5a and Rab5b are not (Chen, Schauer, et al., 2014). ProtEOMic studies have suggested that the Rab5 isoforms can be differentially ubiquitinated in cells (Chen, Zhou, et al., 2014, Wagner et al., 2011, Wagner et al., 2012). Since Rab5a has been extensively studied for its involvement in endosomal fusion (Bucci et al., 1992, Gorvel et al., 1991, Barbieri et al., 2000b, Hoffenberg et al., 1995, Rybin et al., 1996, Stenmark et al., 1994), we decided to focus on Rab5a for our studies (hereafter referred to as ‘Rab5’). Here, we identified monoubiquitination sites of Rab5 in cultured cells.”
Next, the authors use GGTI–198, a GGTase inhibitor, to rule out prenylation as the source of the double bands upon ubiquitination. However, this experiment is not sufficiently described in the figure legend or in the text. What are the two different IP FLAG plots in Figure 1B, what do the labels short and long mean?
The treatment of GGTI–198 is described in the Materials and methods section as follows:
(Subsection “Ubiquitination assay”) “For the geranylgeranyl transferase inhibitor assay, 20 μM GGTI–298 (G5169, Sigma) was added for 6 hours after transfection. Twenty–four hours after GGTI treatment, cells were harvested and subjected to ubiquitination assay. These experiments were repeated three times.”
We described that label in Figure 1 legend as follows:
(Figure 1 legend) “Short and long refer to relative exposure time during immunoblotting.”
Importantly, it is not clear to the reader why out of all lysines in Rab5, K33 and K165 were chosen in addition to the known K116, K134 and K140 sites from protEOMics. There is some explanation for K165 in sequence homology to K147 in Ras, but otherwise the authors make no effort to explain their rationale. This section should be reordered to make it more logical. I would suggest reordering the first part of the manuscript: the story would flow better if the validation of protEOMics hits preceded the investigation of the now proven ubiquitination sites. For example, instead of starting with candidate ubiquitination sites and an analysis of crystal structures in subsection “Rab5 undergoes monoubiquitination at multiple sites”, it would be more logical to begin the experiments presented in Figure 1D as a validation of published protEOMics data. From there the story can then focus on the relevant monoubiquitination sites and their location on the Rab5 structure (Figure 1—figure supplement 1A).
We completely revised the manuscript according to the logical flow the reviewer suggested. Briefly, ubiquitination sites from protEOMics studies (K116, K134 and K140) were validated with ubiquitination assay. We found that K116 and K140 were monoubiquitinated. By contrast, K134 was not responsible for monoubiquitination on Rab5, implicating that K134 might undergo polyubiquitination. Another monoubiquitination site K165, homologous to K147 of Ras, was identified. Finally the three putative monoubiquitination sites (K116, K140 and K165) were chosen for further studies. These changes have been implemented in the lines 78–152 of the revised manuscript.
Furthermore, the experiments done with K to E mutations and protein expression are not essential and should be moved to the supplement or left out entirely so as not to distract from the more sophisticated K to R mutation experiments.
Pursuant to the reviewer’s suggestion, the results on K to E mutation and their expression are now omitted from the revised manuscript.
In Figure 2C, the authors show that monoubiquitinated Rab5 (mUbRab5) localises to membranes. This raises many questions: mUbRab5 is a very small percentage of total membrane Rab5, why do the authors observe such a dramatic effect on recycling/degradation? This would seem to be a highly significant result but is not addressed at all. In addition to this, Figure 2B shows a dramatic removal of Rab5 from EEA1 positive endosomes upon ubiquitination and indeed in the text the authors refer to this as 'diffusing to cytosol' which would not agree with the results presented in Figure 2C.
The dramatic effects of a small percentage of monoubiquitinated Rab5 on recycling/degradation is now discussed in the revised manuscript as follows:
(Discussion) “In this study, we found that monoubiquitination of Rab5 had an inhibitory effect on the regulatory cycle of Rab5. However, we did not find a stimulus signal or E3 ligase(s) responsible for monoubiquitination of Rab5. As result, only a small portion of Rab5 could be ubiquitinated when ubiquitin was overexpressed. These results suggest that ubiquitination of Rab5 occurs rarely in cells under basal conditions and might the reason why the ubiquitination of Rab GTPases has not been studied until recently (de la Vega et al., 2011). Despite the small population of monoubiquitinated Rab5, the effects of monoubiquitination of Rab5 were strong. We observed dramatic failure of localization of Rab5 to the endosome and EGFR degradation pathway. This suggests that monoubiquitination of Rab5 is tightly regulated in the cell through specific E3 ligase or DUB enzymes. It will be of great interest to determine how monoubiquitination of Rab5 is regulated in cells.”
The confusions for describing localization of Rab5 upon ubiquitin transfection is now clarified as follows:
(Subsection “Monoubiquitination of K140 and K165 of Rab5 plays a negative role in the endocytic pathway”) “We observed that the monoubiquitinated Rab5 was predominantly localized in the membrane fraction. These results prompted us to hypothesize that monoubiquitinated Rab5 could disrupt Rab5–positive endosomal formation while it was localized on the membrane. Together with the immunofluorescence assay results (Figure 2A–B), these findings strongly suggest that ubiquitination of Rab5 has negative effects on the regulatory cycle of Rab5.”
If K134 mutations are not viable, how has this been identified as a ubiquitination site in vivo? The authors should comment on this.
Our ubiquitination assay data suggests that K134 is not monoubiquitinated. Since K134 was previously reported as an ubiquitination site by protEOMic studies, it is likely that K134 may be polyubiquitinated, although we have no direct evidence for it. Polyubiquitination of K134 would destabilize the structural integrity of Rab5, ultimately leading to its degradation. We have commented about this concern as follows:
(Subsection “Rab5 undergoes monoubiquitination at multiple sites”) “No monoubiquitinated Rab5 band was detected for the 11KR mutant in the ubiquitination assay (Figure 1H), suggesting that K33 and K134 are not responsible for monoubiquitination of Rab5. Considering that K134 was previously reported to be ubiquitinated by protEOMic studies (Chen, Zhou, et al., 2014, Wagner et al., 2011, Wagner et al., 2012), it is plausible that K134 may undergo polyubiquitination. Given the crucial role of K134 in the structural integrity of Rab5 (Figure 1—figure supplement 1B and C), we hypothesized that ubiquitination of K134 might be related to disruption/degradation of Rab5, thereby preventing us from observing ubiquitination of K134 in the current experimental conditions.”
The authors introduce the immunofluorescence of Rab5 endosomes and justify the use of HeLa cells in subsection “Monoubiquitination on K140 and K165 of Rab5 plays negative roles in endocytic pathway”, but results from these experiments had already been shown in subsection “Rab5 undergoes monoubiquitination at multiple sites” without explanation of what Rab5 puncta are or what cell line was used.
We revised our manuscript guided by the reviewer’s suggestion as follows:
(Subsection “Rab5 undergoes monoubiquitination at multiple sites”) “Subsequently, we generated an 11KR mutant with intact K33 and K134 and examined its function by observing Rab5–positive endosome puncta in HeLa cells to visualize Rab5–positive endosomes more efficiently (Miaczynska et al., 2004, Borg et al., 2014, Kajiho et al., 2003).”
In Figure 2A and C, UbRab5 does not appear as a double band as in other blots, why is this?
Different percentage of gel was used for Figure 2A, C (10% gel for Figure 1, 12% gel for Figure 2A, C) since we already confirmed that both double bands are responsible for monoubiquitinated Rab5. This point is explicitly stated in the revised manuscript as follows:
(Subsection “Rab5 undergoes monoubiquitination at multiple sites”) “These double bands were not observed when a high percentage polyacrylamide gel (12%) was used (Figures 1C, E–F).”
Subsection “Monoubiquitination on K140 and K165 of Rab5 plays negative roles in endocytic pathway” are a bit of a reach of interpretation at this point in the manuscript and are only logical in hindsight, knowing the results that are still to come.
We removed these sentences in the revised manuscript, and the preceding paragraph was entirely re–written as the reviewer suggested.
From the SAXS analysis the authors derive that mUbRab5K140 features two conformations while mUbRab5K116 and mUbRab5K165 have only one conformation. However, the authors also state in the text that mUbRab5 is structurally very flexible, contradicting the existence of distinct conformations. To better evaluate the data, the authors need to explain in detail how distinct conformations were extracted from the data using ensemble–optimized methods (EOM) and how the final models were derived (Figure 4G, H and I). Furthermore, the authors need to show the shape of the particle derived from the SAXS data, not only the model. In general, the unprocessed SAXS data are missing and the reader is presented immediately with something highly processed. For a more coherent story, the intermediate stages of deriving these structural models must be presented.
To determine whether we can apply ab initio modeling of mUbRab5s, we carefully analyzed our SAXS curves of mUbRab5s (described in subsection “Structural models of mUbRab5s derived by solution scattering”). As shown in Figure 4—figure supplement 1B, all the mUbRab5s feature a loose plateau in the Porod–Debye plot, supporting that all the mUbRab5s are flexible in solution. Because mUbRab5 shows clear feature of flexibility, we decided to analyze the SAXS curve with EOM. The details of how we extracted distinct conformations of mUbRab5s through EOM have been described as follows:
(Subsection “Structural models of mUbRab5s derived by solution scattering”) “EOM was originally designed for multi–domain proteins with a flexible linker. Input files for EOM are a single linear protein sequence covering entire protein, atomic coordinates for each domain, and SAXS data (Tria et al., 2015). Because ubiquitinated proteins are not single polypeptides but branched ones, we generated the sequences and atomic coordinates of mUbRab5s as described in the Materials and methods. A pool of 10,000 independent models was generated based on the sequence and structural information from SAXS curve by RANCH (embedded in EOM, ATSAS package). Then, a genetic algorithm for the selection of an ensemble was performed by GAJOE with 100 times (embedded in EOM, ATSAS package). Finally, the best ensemble matched with the SAXS curve with lowest χ2 was selected."
(Subsection “Small angle X-ray scattering (SAXS) measurement and data processing”) “Since EOM accepted only single polypeptide chain with multiple domains, we generated the sequences and atomic coordinates of the mUbRab5s in the following four steps. [...] Finally, we used the native isopeptide linkage for ensemble model calculations in EOM since a disulfide bond linkage is known to resemble the native ubiquitination linkage (Baker, Lewis, et al., 2013).”
Based on the ab initio model from initial SAXS curve one can obtain an idea for the shape of the particle. Therefore, we have generated ab initio models for mUbRab5s and the results were deposited on SASBDB. (Since the data will be released after acceptance on publication, you can only access them through the link below.)
mUbRab5K116: https://www.sasbdb.org/data/SASDCM6/tkmwsdjl7l/
mUbRab5K140: https://www.sasbdb.org/data/SASDCN6/0ec25yspdo/
mUbRab5K165: https://www.sasbdb.org/data/SASDCP6/7wnrjfzgyv
We question whether we should include the ab initio models in the revised manuscript. Technically ab initio modelling should be performed with monodisperse sample, which is not flexible and does not have oligomeric state. However, mUbRab5s revealed signs of high flexibility (Figure 4—figure supplement 1B). If one generates ab initio model with a flexible sample, the overall size of the molecule will be much larger than expected, and the χ2 value with an atomic model will be poorer than the value generated by the EOM analysis. In addition, analyzing and showing a SAXS curve with two different systems (in this case, flexible and ab initio) is technically improper. One should choose the best method that describes their SAXS curve. With this regards, we suggest to include our SASBDB deposition ID numbers along with the EOM analysis in the revised manuscript.
Since the resolution of our figures was not sufficient in the initially submitted manuscript, it might be hard to find the unprocessed SAXS data which is shown as the black dots in figure 4G–I (labeled as “Experiment”). We hope that the high resolution Figure 4 in the revised manuscript will resolve this issue.
From structural models the authors derive that the close proximity of the ubiquitin moieties in mUbRab5K140 to the switch I region of Rab5 might disturb Rabaptin–5 and EEA1 binding, while interaction of Rab5 with GEF (Rabex–5) and GAP (RabGAP–5) is unchanged. This especially needs clarification, because the switch I region is involved in nucleotide binding as well, and hence prone to be influenced by changes in nucleotide state. Knowing the outcome of the binding experiments described later, this conclusion makes sense, but when reading this manuscript for this first time this appears a somewhat illogical conclusion.
As the reviewer suggested, we clarified the paragraph with logical flow as below. Briefly, the conformational change of SWI is interrupted by ubiquitin moiety, which would cause the activation of Rab5 through interaction with downstream effector proteins. We revised the manuscript as follows:
(Subsection “Structural models of mUbRab5s derived by solution scattering”) “However, the switch I (SWI) region of Rab5 was juxtaposed with the flexible ubiquitin moiety of mUbRab5K140, raising the possibility that a flexible ubiquitin moiety could cause a conformational change in the SWI. Because this conformational change of SWI is required for activation of Rab5 through interactions with downstream effector proteins, potential interruption of the SWI by mUbRab5K140 could have a severe impact on the Rab5 regulatory cycle.
Reviewer #3
[…]All of the experiments involve either purified proteins or transfections of cell lines. There are no experiments assessing whether the endogenous Rab proteins are monoubiquinated (i.e. without ubiquitin overexpression) and under what conditions. This limitation should be discussed, and conclusions about the potential biological significance appropriately qualified.
We appreciate the reviewer’s careful reading of our manuscript and suggestion to state the limitation of our study. Pursuant to the reviewer’s suggestion, we discussed the limitation that no experiment on the monoubiquitination of endogenous Rab without ubiquitin overexpression and the biological significance in the revised manuscript as follows:
(Discussion) “In this study, we found that monoubiquitination of Rab5 had an inhibitory effect on the regulatory cycle of Rab5. However, we did not find a stimulus signal or E3 ligase(s) responsible for monoubiquitination of Rab5. As result, only a small portion of Rab5 could be ubiquitinated when ubiquitin was overexpressed. These results suggest that ubiquitination of Rab5 occurs rarely in cells under basal conditions and might the reason why the ubiquitination of Rab GTPases has not been studied until recently (de la Vega et al., 2011). Despite the small population of monoubiquitinated Rab5, the effects of monoubiquitination of Rab5 were strong. We observed dramatic failure of localization of Rab5 to the endosome and EGFR degradation pathway. This suggests that monoubiquitination of Rab5 is tightly regulated in the cell through specific E3 ligase or DUB enzymes. It will be of great interest to determine how monoubiquitination of Rab5 is regulated in cells.”
Subsection “Rab5 undergoes monoubiquitination at multiple sites”, K33 and K134 are indeed crucial but this is already known, indeed from the very first structures of EF–Tu and Ras. K33 is the P–loop lysine in the GxxxxGKS/T motif; K134 is the invariant lysine in the NKxD G4 motif. The role and interactions of these residues in GTPases in general has been extensively documented. Its not clear why these invariant residues were chosen for mutational analysis. Expression in inclusion bodies in bacteria isn't particularly informative or unexpected. The rationale for generating lysine–to–glutamate mutations is also unclear as these mutations don't mimic ubiquitination.
We have cited the references as the reviewer suggested and revised our manuscript as below. In addition, the lysine–to–glutamate mutation studies are removed in the revised manuscript.
(Subsection “Rab5 undergoes monoubiquitination at multiple sites”) “However, the 13KR mutant was not expressed in HEK 293T cells (Figure 1e). We reasoned that this was due to mutation of both K33 and K134 based on previous structural studies of Ras and EF–Tu GTPases (Pai et al., 1989, Brunger et al., 1990, Jurnak, 1985, Berchtold et al., 1993). Two lysine residues, the P–loop lysine in the GxxxGKS/T motif and the invariant lysine in the NKxD motif, are crucial because they form direct contacts with the guanine nucleotide. Similarly, for Rab5, K33 and K134 form direct contacts with both GDP and GTP molecules (Figure 1—figure supplement 1B and C, PDB ID: 1TU4 and 3MJH, respectively).”
How can anything be concluded about inhibition of Rab5 activation (subsection “Monoubiquitination on K140 and K165 of Rab5 plays negative roles in endocytic pathway”) from the IF and fractionation experiments in Figure 2B and 2C? These experiments do not indicate whether ubiquitination affects the nucleotide status of Rab5. Perhaps the authors meant inhibition of a Rab5 function rather than activation of Rab5 per se. Needs clarification.
We performed ubiquitination assay with Rab5 GDP–/GTP– bound mutants (S34N and Q79L, respectively) and the results were presented in Figure 1. Since we found clear monoubiquitination band from both S34N and Q79L mutants, we concluded that the monoubiquitination is independent of the nucleotide binding status of Rab5. However, as the reviewer commented, we cannot definitively indicate whether ubiquitination on Rab5 affects the nucleotide status of Rab5. Therefore, we revised the expression from “activation” to “inhibition” of Rab5, and it is described in the revised manuscript as follows:
(Subsection “Monoubiquitination of K140 and K165 of Rab5 plays a negative role in the endocytic pathway”) “Together with the immunofluorescence assay results (Figure 2A–B), these findings strongly suggest that ubiquitination of Rab5 has negative effects on the regulatory cycle of Rab5.”
Overexpression of ubiquitin presumably affects the function of many targets including proteins other than Rab5 that are also important for endocytic trafficking. Does overexpression of ubiquitin alone have any effects on (for example) EGFR degration?
We agree with the reviewer that ubiquitin overexpression could affect other proteins in the endocytic trafficking. To explore such possibility, we examined whether overexpression of ubiquitin alone affects the EGFR degradation (Figure 2). We observed no significant difference in the EGFR degradation by ubiquitin overexpression itself. While we cannot exclude the possibility that ubiquitin overexpression may affect proteins in the endocytic pathways, our results with K–to–R or single K mutants together with biochemical assay strongly support that these observations are tightly linked with ubiquitination of Rab5 (Figure 5, 6).
The cell-based experiments lack quantification and analysis of statistical significance.
We added statistics and the significance for cell-based experiments in our revised manuscript (Figure 2B, G, H).
The resolution of the images in the PDF is low, making it difficult to evaluate the quality of these experiments.
We are submitting our revised manuscript with high resolution image files.
The buffer–subtracted SAXS curves should be shown somewhere in the main or supplemental figures. Accurate modeling of SAXS data, in particular by ensemble approaches, requires data with high signal to noise that typically extends to a high resolution cutoff near the first Porod minimum. The data and calculated curves in Figure 4 are truncated in the Kratky region and the signal–to–noise does not appear to be very high. Are the data of sufficient quality for ensemble modeling?
Since the resolution of our figures was not sufficient in the initially submitted manuscript, it might be hard to find the unprocessed SAXS data which is shown as the black dots in figure 4H–I (labeled as “Experiment”). We hope that the high resolution Figure 4 in the revised manuscript will resolve this issue.
The experimental setup at the synchrotron beamline where we collected SAXS data prevented us from obtaining high resolution data. However, technically EOM does not work better or worse with limited angular range. The main output of EOM is a histogram of the Rg values (and sometimes a histogram of Dmax values). EOM just characterizes the protein as compact or extended and provides a way of comparing the degree of compactness. The Rg and Dmax information is contained in the lower angles. The high angle information is smeared due to polydispersity of the sample in case of a flexible protein. Nevertheless, we fully agree with your concerns about the q range and we did validation process through the SAXS data deposition on SASBDB. Through the review process in SASBDB, our data for mUbRab5K116, mUbRab5K140 and mUbRab5K165 have been successfully deposited to the SASBDB (ID: SASDCM6, SASDCN6 and SASDCP6, respectively). Therefore, we believe that the SAXS data we used are of sufficient quality for EOM modelling. Since the data will be released after acceptance on publication, you can only access them through the link below.
mUbRab5K116: https://www.sasbdb.org/data/SASDCM6/tkmwsdjl7l/
mUbRab5K140: https://www.sasbdb.org/data/SASDCN6/0ec25yspdo/
mUbRab5K165: https://www.sasbdb.org/data/SASDCP6/7wnrjfzgyv
No details are provided regarding the EOM modeling. For example, how many residues we're allowed to be flexible and what criteria where to determine which residues show be flexible? Do the results change if additional residues are allowed to be flexible? A narrow selected distribution near the high Rg and Dmax tails of the pools in Figure 4 suggest that the number of flexible residues used to generate the pool might not be adequate to represent the flexibility in solution. Also unclear is how the linkage between Rab5 and ubiquitin is modeled. Does the EOM model disulfide linkages? If not, how was the linkage modeled? The details of the modeling are important and how they affect the results needs to be evaluated and discussed.
We added details for EOM analysis in the revised manuscript as described below. We believe that the preparation steps of input files for ubiquitinated proteins in EOM analysis are key steps, and it will be very useful to other scientists who wish to apply EOM to similar cases. The detailed description for EOM analysis is now described as follows:
(Subsection “Structural models of mUbRab5s derived by solution scattering”) “EOM was originally designed for multi–domain proteins with a flexible linker. Input files for EOM are a single linear protein sequence covering entire protein, atomic coordinates for each domain, and SAXS data (Tria et al., 2015). Because ubiquitinated proteins are not single polypeptides but branched ones, we generated the sequences and atomic coordinates of mUbRab5s as described in the Materials and methods. A pool of 10,000 independent models was generated based on the sequence and structural information from SAXS curve by RANCH (embedded in EOM, ATSAS package). Then, a genetic algorithm for the selection of an ensemble was performed by GAJOE with 100 times (embedded in EOM, ATSAS package). Finally, the best ensemble matched with the SAXS curve with lowest χ2 was selected."
(Subsection “Small angle X-ray scattering (SAXS) measurement and data processing”) “Since EOM accepted only single polypeptide chain with multiple domains, we generated the sequences and atomic coordinates of the mUbRab5s in the following four steps. […] Finally, we used the native isopeptide linkage for ensemble model calculations in EOM since a disulfide bond linkage is known to resemble the native ubiquitination linkage (Baker, Lewis, et al., 2013).”
"Interestingly, none of the EOM–derived structural models for mUbRab5s implicates any direct interaction between ubiquitin and Rab5" (subsection “Structural models of mUbRab5s derived by solution scattering”). Is EOM capable of producing models with direct interactions? Even if it could, what would it mean at the low resolution of SAXS?
We revised the sentence as follows:
(Subsection”Structural models of mUbRab5s derived by solution scattering”) “Similar to the molecular dynamics simulation of mUbRas (Baker, Lewis, et al., 2013), none of the ubiquitin molecules from the EOM models were located proximal to Rab5.”
Figure 4—figure supplement 1: Kratky–Debye and Porod–Debye plots indicate how extended a molecule is. Flexibility is inferred. Presumably the same results would be obtained from the calculated EOM curves that fit the data. The question is whether the SAXS data can distinguish flexible vs. extended. The interpretation should be more nuanced.
As the reviewer commented, it is very important to characterize a sample when performing SAXS experiments. Since any kinds of SAXS curve could be used for obtaining ab initio or ensemble model, one should be extremely careful in preparing the sample and initial data assessment. If the sample is homogenous and rigid, SAXS curve can be used for ab initio modeling. In contrast, if the sample is homogenous and flexible, ensemble analysis should be chosen for the method of choice. Judgement of flexibility of the sample is really important in the initial data assessment. From a SAXS curve, we can analyze whether the protein is folded or unfolded (extended) through traditional Kratky analysis (s vs. s2*I(s)). However, it was challenging to analysis of flexible versus unfolded protein using traditional Kratky analysis. Rambo and Tainer described a new analysis, called Porod–Debye law (s4 vs. s4*I(s)), for obtaining information of flexibility of a protein from a SAXS curve (Rambo & Tainer, 2011). Briefly, Porod–Debye plot can distinguish discrete conformational change (apo vs holo form of a protein with a substrate) and flexibility of the protein, while traditional Kratky analysis cannot. To obtain this information, one should take a look whether there is a plateau in Porod–Debye plot. If there is a plateau, it implies that the protein is compact or rigid, while loss of plateau reflects flexibility of the protein. The graphical explanation for this analysis is also well–presented in a website (http://www.bioisis.net/tutorial/12). In addition the Porod–Debye law has another advantage over Kratky analysis. While Kratky analysis requires relatively high angle data (0.01 < s < 0.3 Å–1), the Porod–Debye analysis can be done with similar low angular range as Guinier region. To make it easier for a reader to understand, we inserted the interpretation of flexibility analysis in our revised manuscript as follows:
(Subsection “Structural models of mUbRab5s derived by solution scattering”) “To check whether the ubiquitin moieties of Rab5 are flexible, we performed flexibility analysis using the Porod–Debye method (Rambo & Tainer, 2011). SAXS curves from the three mUbRab5s did not have a plateau in the Porod–Debye plot, implying that all of the mUbRab5s were flexible in solution (Figure 4—figure supplement 1B).
https://doi.org/10.7554/eLife.29154.029Article and author information
Author details
Funding
National Research Foundation of Korea (NRF-2015R1A2A1A15055951)
- Donghyuk Shin
- Gyuhee Kim
- Jiseok Baek
- Sangho Lee
Rural Development Administration (PJ009106)
- Donghyuk Shin
- Gyuhee Kim
- Jiseok Baek
- Sangho Lee
National Research Foundation of Korea (2012-0009597)
- Donghyuk Shin
- Gyuhee Kim
- Jiseok Baek
- Sangho Lee
National Research Foundation of Korea (SRC-2017R1A5A1014560)
- Donghyuk Shin
- Wooju Na
- Ji-Hyung Lee
- Gyuhee Kim
- Jiseok Baek
- Seok Hee Park
- Cheol Yong Choi
- Sangho Lee
The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.
Acknowledgements
We thank the staff members at beamline 4C of Pohang Accelerator Laboratory for technical assistance in SAXS data collection. This work was supported by the Basic Science Research Program (NRF-2015R1A2A1A15055951), the Science Research Center Program (SRC-2017R1A5A1014560), and the Pioneer Research Center Program (2012–0009597) through National Research Foundation of Korea (NRF) grants funded by the Korea Ministry of Science and ICT and by the Woo Jang Chun program (PJ009106) through the Rural Development Agency.
Reviewing Editor
- Suzanne R Pfeffer, Stanford University School of Medicine, United States
Publication history
- Received: May 31, 2017
- Accepted: September 5, 2017
- Version of Record published: October 2, 2017 (version 1)
Copyright
© 2017, Shin et al.
This article is distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use and redistribution provided that the original author and source are credited.
Metrics
-
- 2,137
- Page views
-
- 344
- Downloads
-
- 20
- Citations
Article citation count generated by polling the highest count across the following sources: Crossref, PubMed Central, Scopus.
Download links
Downloads (link to download the article as PDF)
Open citations (links to open the citations from this article in various online reference manager services)
Cite this article (links to download the citations from this article in formats compatible with various reference manager tools)
Further reading
-
- Biochemistry and Chemical Biology
Ciliopathies manifest from sensory abnormalities to syndromic disorders with multi-organ pathologies, with retinal degeneration a highly penetrant phenotype. Photoreceptor cell death is a major cause of incurable blindness in retinal ciliopathies. To identify drug candidates to maintain photoreceptor survival, we performed an unbiased, high-throughput screening of over 6,000 bioactive small molecules using retinal organoids differentiated from induced pluripotent stem cells (iPSC) of rd16 mouse, which is a model of Leber congenital amaurosis (LCA) type 10 caused by mutations in the cilia-centrosomal gene CEP290. We identified five non-toxic positive hits, including the lead molecule reserpine, which maintained photoreceptor development and survival in rd16 organoids. Reserpine also improved photoreceptors in retinal organoids derived from induced pluripotent stem cells of LCA10 patients and in rd16 mouse retina in vivo. Reserpine-treated patient organoids revealed modulation of signaling pathways related to cell survival/death, metabolism, and proteostasis. Further investigation uncovered dysregulation of autophagy associated with compromised primary cilium biogenesis in patient organoids and rd16 mouse retina. Reserpine partially restored the balance between autophagy and the ubiquitin-proteasome system at least in part by increasing the cargo adaptor p62, resulting in improved primary cilium assembly. Our study identifies effective drug candidates in preclinical studies of CEP290 retinal ciliopathies through cross-species drug discovery using iPSC-derived organoids, highlights the impact of proteostasis in the pathogenesis of ciliopathies, and provides new insights for treatments of retinal neurodegeneration.
-
- Biochemistry and Chemical Biology
Mutations within Ras proteins represent major drivers in human cancer. In this study, we report the structure-based design, synthesis, as well as biochemical and cellular evaluation of nucleotide-based covalent inhibitors for KRasG13C, an important oncogenic mutant of Ras that has not been successfully addressed in the past. Mass spectrometry experiments and kinetic studies reveal promising molecular properties of these covalent inhibitors, and X-ray crystallographic analysis has yielded the first reported crystal structures of KRasG13C covalently locked with these GDP analogues. Importantly, KRasG13C covalently modified with these inhibitors can no longer undergo SOS-catalysed nucleotide exchange. As a final proof-of-concept, we show that in contrast to KRasG13C, the covalently locked protein is unable to induce oncogenic signalling in cells, further highlighting the possibility of using nucleotide-based inhibitors with covalent warheads in KRasG13C-driven cancer.