|Home | About | Journals | Submit | Contact Us | Français|
Mammalian arrestins have a major role in the intracellular trafficking of seven-transmembrane (7TM) receptors. The fungal ambient pH signaling pathway involves an arrestin-related protein, PalF/Rim8, and the ESCRT (endosomal sorting complex required for transport) machinery. We found that in Saccharomyces cerevisiae, Rim8 binds to both the putative 7TM pH sensor Rim21 and the ESCRT-I subunit Vps23. We show that an SXP motif in Rim8 mediates binding to the Vps23 ubiquitin E2 variant (UEV) domain and that a monoubiquitinated residue near the SXP motif contributes to this interaction. We present evidence that Rim8 ubiquitination is dependent on the Rsp5 E3 ubiquitin ligase and triggered upon binding of Vps23 UEV to both the SXP motif and ubiquitin, thus suggesting a two-step binding mechanism. We further show that Rim8 coimmunoprecipitates with ESCRT-I subunits Vps23 and Vps28, supporting the idea that binding of Rim8 to Vps23 mediates the association of Rim8 with the ESCRT-I complex. Fluorescence microscopic analyses indicate that overexpressed Rim8 and Vps23 colocalize at cortical punctate structures, providing additional evidence of the interaction between these two proteins. Strikingly, our findings indicate that evolutionary conserved mechanisms control the recruitment of the ESCRT machinery to Pal/Rim proteins in fungi and retroviral Gag proteins in animal cells.
Arrestins play an essential role in the regulation of seven-transmembrane (7TM) receptors (31). Although initially identified on the basis of their ability to uncouple 7TM receptors from heterotrimeric G proteins in a process known as desensitization, arrestins were later found to serve as endocytic adaptors that recruit clathrin and the clathrin adaptor protein AP-2 and facilitate 7TM receptor internalization via clathrin-coated vesicles (17, 29, 30).
Until recently, arrestins were thought to be restricted to the animal kingdom. However, the identification of an arrestin-related protein in Aspergillus nidulans, PalF, demonstrated that members of this protein family are also present in fungi (22). PalF and its yeast homolog Rim8 are involved in the Pal/Rim signaling pathway, which mediates the ambient pH response and is activated in neutral-alkaline environments (41). As mammalian beta-arrestins, PalF, which contains arrestin N-terminal and C-terminal domains, binds to the cytoplasmic domain of a 7TM protein, the putative pH sensor PalH, and is ubiquitinated in a signal- and 7TM receptor-dependent manner (22).
The PalF/Rim8 signaling function does not appear to be related to receptor desensitization, as there is no evidence of G protein involvement in the Pal/Rim pathway. However, evidence strongly suggests that these proteins, as mammalian β-arrestins, play a role as endocytic adaptors. Most of the components of the ESCRT (endosomal sorting complex required for transport) machinery involved in the MVB (multivesicular body) sorting pathway (24, 48), play an essential role in the Pal/Rim signaling pathway (4, 7, 12, 15, 20, 28, 43, 45, 62). According to the current model for the Rim pathway in Saccharomyces cerevisiae, the ESCRT-I and ESCRT-II complexes, together with the Snf7 and Vps20 subunits of ESCRT-III, act downstream of a plasma membrane complex containing the two 7TM proteins Rim21 and Dfg16, Rim8, and probably the 3TM protein Rim9 (41). The prevailing view is that pH signaling requires Rim8-mediated endocytic internalization of Rim21 and/or Dfg16, thus making the connection between the plasma membrane complex and the endosomal ESCRT machinery. Snf7, which oligomerizes upon membrane binding (1, 47, 53), would serve as a docking platform for recruitment of a protein complex that triggers the pH-dependent proteolytic activation of the Rim101 transcription factor by the Rim13 cysteine protease. Rim20, a Bro1 domain-containing protein, mediates the recruitment of Rim101 to Snf7 (61), and evidence suggests that this interaction is a regulated step in the pathway (8).
An important question concerns the molecular mechanisms by which the pH signal is transmitted from the 7TM protein-arrestin complex to the ESCRT-associated Rim101-processing machinery. Intriguingly, Vps27, which mediates the recruitment of ESCRT-I to endocytic cargos, is not required for pH signal transduction, suggesting that a Rim pathway-specific protein may substitute Vps27 for ESCRT-I recruitment (45, 62).
Here, we present genetic and biochemical evidence that in S. cerevisiae, the arrestin-related protein Rim8 mediates the recruitment of ESCRT-I to the 7TM protein Rim21. Rim8 binds to both Rim21 and the ESCRT-I subunit Vps23, and we present evidence that an SXP motif and a monoubiquitinated residue in Rim8 contribute to binding to the Vps23 UEV domain. We show that Rim8 ubiquitination is dependent on the E3 ubiquitin (Ub) ligase Rsp5, which binds a PXY motif in Rim8. We further show that Rim8 associates with both ESCRT-I subunits Vps23 and Vps28 in yeast extracts and that overexpressed Rim8 and Vps23 colocalize at cortical punctate structures. Our results, therefore, identify a protein of the arrestin family as an adaptor for the recruitment of the ESCRT machinery to a 7TM protein.
The Saccharomyces cerevisiae strains were MHY501 (MATα DOA4 his3- Δ200 leu2-3,112 ura3-52 lys2-801 trp1-1) and the isogenic doa4::LEU2 derivative MHY623 (40), 23346c (MATa ura3), and its isogenic derivative 27038a (MATa ura3 npi1) (21). The strain Y00000 (BY4741; MATa his3- Δ1 leu2 Δ0 met15 Δ0 ura3 Δ0) and isogenic derivatives Y04414 (rim8 Δ), Y01150 (rim21 Δ), Y11806 (dfg16 Δ), Y06196 (rim9 Δ), Y03416 (vps23 Δ), Y07333 (rim20 Δ), Y04752 (ygr122w Δ), Y00936 (rim101 Δ), and Y06643 (rim13 Δ) were obtained from Euroscarf (Frankfurt, Germany). The OVY25 strain expressing hemagglutinin (HA)-tagged RIM101 and the OVY121 strain expressing Flag-tagged VPS28 were constructed by transforming Y04414 (rim8 Δ) with a fragment of plasmid pKR41 and a PCR fragment generated from plasmid pFA6a-6xGLY-3xFLAG-HIS3MX6 (Addgene plasmid 20753), respectively, as previously described (14, 45). CTY10.5d (MATa ade2-101 his3- Δ200 leu2- Δ1 trp1- Δ901gal4 gal80 URA3::lexAop-lacZ) (3) and W303-1A (MATa trp1-1 leu2-3,112 his3-11,15 ura3-1 ade2-1 can1-100) were used for two-hybrid and fluorescence microscopy experiments, respectively. Standard genetic methods were followed, and yeast cultures were grown in synthetic dextrose (SD) medium lacking appropriate amino acids to maintain selection for plasmids (44). Growth tests on yeast extract-peptone-dextrose (YPD) plates containing 200 mM LiCl have been described elsewhere (15).
Plasmids encoding Gal4 activation domain (GAD)-Vps23 and GAD-Rim21Cterm were constructed by cloning the Vps23 open reading frame (ORF) or the Rim21 C-terminal cytoplasmic tail (codons 327 to 534) in the polylinker of pACT2 (Clontech). Plasmids encoding glutathione S-transferase (GST)-Vps23 and GST-Rsp5WW are derivatives of pGEX-5X-1 (Pharmacia) containing the Vps23 ORF and the Rsp5 WW domains (codons 228 to 430), respectively. pADH1-HA-Vps23UEV is a derivative of pWS93 (50) containing the Vps23 UEV domain (codons 1 to 161). Vps23 mutations resulting in a Ser162Stop substitution in GAD-Vps23 and GST-Vps23 and in a Ser55Ala Asp56Ala Gly57Ala triple substitution in GAD-Vps23 and HA-Vps23UEV were obtained by using mutagenic PCR. pVps23-Flag and pVps23-RFP, encoding phenotypically wild-type (WT) (data not shown) Flag- or red fluorescent protein (RFP)-tagged Vps23 under the control of its native promoter, are derivatives of the centromeric plasmid pRS316 and the 2μm high-copy-number plasmid pRS424 (49), respectively. A NotI site was engineered immediately 5′ of the stop codon of the Vps23 coding sequence, with a 367-bp 5′ sequence and a 128-bp 3′ sequence. This site was used to introduce a double-stranded oligonucleotide containing a triple Flag tag or a NotI fragment containing the mCherry RFP sequence. Plasmids encoding LexA-Rim8 and LexA-Rim8(528-542) or 3MYC-Rim8 and 3MYC-Rim8(461-542) were constructed by cloning the corresponding Rim8 coding sequence in pLexA(1-202)+PL (46) or pGBKT7 (Clontech). pFlag-Rim8, pHA-Rim8, and pRim8-HA, encoding phenotypically WT (data not shown) Flag- or HA-tagged Rim8 under the control of its native promoter are derivatives of the centromeric plasmid pRS313 (49) containing the Rim8 coding sequence with a 1,000-bp 5′ sequence and a 427-bp 3′ sequence. For pHA-Rim8 and pFlag-Rim8, a NotI site was engineered after Rim8 codon 178 and was used to introduce a NotI fragment from pGTEP1 (56) containing a triple HA tag or a double-stranded oligonucleotide containing a triple Flag tag in a region of Rim8 corresponding to the loop connecting β strands 9 and 10 in bovine arrestin (22). The same procedure was used for pRim8-HA, except that the NotI site was created immediately 5′ of the stop codon. pRim8-GFP, encoding phenotypically WT (data not shown) green fluorescent protein (GFP)-tagged Rim8, is a derivative of the 2μm high-copy-number plasmid pRS423 containing the same Rim8 fragment as pRim8-HA, except that the NotI fragment containing the triple HA tag was replaced by a NotI fragment from pSFGP1 (27) containing the GFP sequence. pHA-Rim8(U) and pRim8-HA(U) were obtained by cloning the HA-Rim8- or Rim8-HA-containing fragments from pHA-Rim8 or pRim8-HA in pRS316. Rim8 mutations resulting in Pro506Ala, Lys521Arg, Lys527Arg, Pro536Leu, and Val505Stop substitutions or in a Glu533Ala Ser534Ala Asp535Ala Pro536Ala quadruple substitution were obtained by using mutagenic PCR. To construct the plasmid encoding the Flag-Rim8(1-504)-Vps23 chimeric protein, a NcoI site was engineered after Rim8 codon 504 in pFlag-Rim8 and used to introduce the Vps23 coding sequence. YEp96 (2μm TRP1 Ub) and YEp112 (2μm TRP1 HA-Ub) contain synthetic yeast Ub and HA-tagged Ub under the control of the copper-inducible CUP1 promoter, respectively (23). pCUP1-Ub0K was constructed by replacing the BamHI-KpnI Ub fragment in Yep96 with a fragment from LHP306 containing Ub-no Lys (55). pLGn+3xNRE22D carries a CYC1-NRE-lacZ reporter (45).
For two-hybrid analysis, transformants were grown to mid-log phase in a selective SD medium. β-Galactosidase was quantitatively assayed in permeabilized yeast cells grown to mid-log phase in a selective SD medium and expressed in Miller units (35). For analysis of the CYC1-NRE-lacZ reporter, transformants were grown in a selective SD medium containing 0.1 M HEPES titrated to pH 7.0 and β-galactosidase activity was assayed with protein extracts prepared by cell disruption with glass beads and normalized to the protein concentration of the extract (18).
Yeast protein extracts were prepared from transformants grown to mid-log phase in a selective SD medium (final pH value, 3.5) and shifted to pH 7 for 30 min by adding 10 mM KOH to activate the Rim pathway. When indicated, protein extracts were prepared before the shift. Protein extracts for Rim101-HA and Flag-Rim8 immunoblot analysis were prepared by a rapid boiling method (57). Protein extracts for HA-tagged Rim8 and Vps23 immunoblot analysis were prepared with immunoprecipitation (IP) buffer (50 mM HEPES, pH 7.5, 150 mM NaCl, 0.1% Triton X-100, 1 mM dithiothreitol [DTT], 10% glycerol, 5 mM N-ethylmaleimide, Complete protease inhibitor mixture [Roche]), as described previously for immunoprecipitation assays (57), or by alkaline lysis (59) when indicated. Immunoprecipitation procedures were essentially as described previously (57). Protein lysates (300 to 400 μg) were incubated with 20 μl of anti-HA (Roche) or anti-Flag (Sigma) affinity matrix for 1 h on a rotating wheel. Total (50 μg, or 1 × 107 cell equivalents, for rapid boiling and alkaline lysis) and immunoprecipitated extracts were analyzed by 7.5% SDS-PAGE and immunoblotting with monoclonal anti-HA (3F10; Roche) or anti-Flag (M2; Sigma) antibodies (Abs). Antibodies were detected by enhanced chemiluminescence with ECL Plus reagents (Amersham).
Recombinant GST-Vps23, GST-Vps23UEV, and GST-Rsp5WW were purified from Escherichia coli BL21 (Novagen) as described previously (58). [35S]3MYC-tagged Rim8, Rim8Cterm (codons 461 to 542), and Rim8 mutant derivatives were synthesized in vitro by using the Promega TNT system in the presence of [35S]methionine [1,000 Ci/mmol (1 Ci = 37 GBq)]. Sepharose beads loaded with GST fusion proteins were incubated with 5 μl of labeling reaction mixture for 1 h at 4°C in 500 μl of STE buffer (10 mM Tris-HCl, pH 8.0, 1 mM EDTA, 150 mM NaCl) with 1% (vol/vol) Triton X-100, washed extensively with the same buffer, and boiled in the sample buffer. Bound proteins were separated by 10% SDS-PAGE and detected by autoradiography (labeled preys) or Coomassie staining (baits). For pulldown assays with yeast protein extracts, Sepharose beads loaded with GST fusion proteins were incubated in IP buffer with 250 μg of protein lysates. Bound proteins were separated by 10% SDS-PAGE and analyzed by immunoblotting. For quantitation of percent binding, several autoradiogram exposures of a gel were scanned as TIFF images and analyzed with ImageJ software.
Yeast strains expressing GFP and RFP fusion proteins were grown to mid-log phase in a selective SD medium. Cells from 1-ml cultures were harvested by centrifugation and resuspended in ~100 μl of the residual medium, and 2.5 μl of the suspension was placed on a microscope slide. GFP and RFP localization in live cultures was monitored by direct fluorescence within 5 min. Cells were viewed using a Nikon Eclipse 90i fluorescence microscope. Images were captured with a DS-Qi1Mc digital camera (Nikon) using Nis elements BR 3.0 software and processed with ImageJ.
Signal transduction through the Pal/Rim pathway involves a 7TM receptor-arrestin complex and most of the components of the ESCRT machinery but not Vps27, which mediates the recruitment of ESCRT-I to endocytic cargos. We then tested whether the arrestin-related protein Rim8 could substitute Vps27 for recruitment of the ESCRT-I subunit Vps23. Two-hybrid assays revealed a strong interaction between a LexA-Rim8 fusion protein and a fusion of Vps23 to the Gal4 activation domain (GAD) (Fig. (Fig.1A).1A). Our previous work showed that the Rim8 homolog in A. nidulans, PalF, binds directly to the C-terminal cytoplasmic tail of the 7TM protein PalH (22). This interaction is conserved in yeast, as we detected a strong two-hybrid binding between Rim8 and the C-terminal cytoplasmic tail of Rim21 (codons 327 to 534), one of the two PalH homologs in S. cerevisiae (Fig. (Fig.1A).1A). Thus, Rim8 interacts with both the 7TM protein Rim21 and the ESCRT-I subunit Vps23. Rim8 binds directly to Vps23, as shown by GST pulldown assays with purified, bacterially expressed GST-Vps23 as bait and in vitro-synthesized 35S-labeled Rim8 (Fig. (Fig.1B1B).
Multiple sequence alignment of Rim8 family members did not reveal regions of sequence conservation outside the arrestin domains, with the exception of a short sequence (box 1) at the C terminus of Rim8 in S. cerevisiae (Fig. (Fig.1C).1C). A related sequence (box 2) is found in Candida albicans, whereas both boxes 1 and 2 are present in A. nidulans and Yarrowia lipolytica. The C-terminal region of Rim8, containing box 1, is necessary (Fig. (Fig.1A)1A) and sufficient (Fig. (Fig.1B)1B) for interaction with Vps23. In contrast, the C terminus of Rim8 is not required for binding to the 7TM protein Rim21 (Fig. (Fig.1A),1A), as expected from the direct involvement of arrestin domains in the interaction between Rim8 and Rim21 homologs in A. nidulans (22). Furthermore, the Vps23 UEV domain (codons 1 to 161) is sufficient for interaction with Rim8 (Fig. (Fig.1A1A).
Previous studies showed that Vps23 UEV interacts with PSDP and PTVP motifs in Vps27 (6, 26). Rim8 box 1 contains a highly conserved SXP motif (SDP; codons 534 to 536) corresponding to a PSXP motif in box 2 (Fig. (Fig.1C).1C). We showed that the 15 C-terminal residues of Rim8 (codons 528 to 542), containing the SXP motif, are sufficient for interaction with Vps23 (Fig. (Fig.1A)1A) and that substitution of four alanine residues in place of the ESDP sequence (codons 533 to 536) containing the SXP motif abolished both two-hybrid interaction and in vitro binding between Rim8 and Vps23 (Fig. 1A and B). As expected, the ESDP/AAAA substitution did not impair the interaction between Rim8 and the 7TM protein Rim21 (Fig. (Fig.1A).1A). A previous study showed that binding of the mammalian Vps23 homolog, Tsg101, to the related P(T/S)AP motif in the human immunodeficiency virus type 1 (HIV-1) Gag protein is strongly dependent on the second Pro residue in this motif, as its replacement by a Leu residue significantly reduced Tsg101 UEV binding (16). We found that Rim8-Vps23 two-hybrid interaction was also disrupted upon replacement of the Pro536 residue with Leu in the SXP motif (Fig. (Fig.1A).1A). However, in contrast to the ESDP(533-536)/AAAA mutation, the Pro536Leu substitution allowed some residual binding to Vps23 in pulldown assays (Fig. (Fig.1B),1B), thus suggesting that this mutation does not fully inactivate the SXP motif. Together, these findings indicate that the conserved SXP motif in Rim8 box 1 is a Vps23-binding site.
Rim8 box 1 contains a highly conserved Lys residue (K521) located 12 residues upstream of the SXP motif (Fig. (Fig.1C).1C). Because the Rim8 homolog in A. nidulans is ubiquitinated (22), we wondered whether K521 might be a ubiquitination site. To test whether Rim8 is ubiquitinated, we examined by immunoblot analysis either Flag- or HA-tagged Rim8, expressed from its own promoter, on a centromeric plasmid. Anti-Flag immunoblot analysis of Flag-tagged Rim8 immunoprecipitates detected a major band migrating at 82 kDa and a minor band of lower-level mobility (Fig. (Fig.2A,2A, upper panel, lane 1). We showed that the band of lower-level mobility markedly increases and is up-shifted upon overexpression of HA-tagged Ub (Fig. (Fig.2A,2A, upper panel, lane 2). This band corresponds to an HA-Ub conjugate as it yields detectable HA immunoreactivity (Fig. (Fig.2A,2A, lower panel). Thus, Rim8, as PalF of A. nidulans, is ubiquitinated. Unexpectedly, we found that Rim8, in contrast to PalF, is ubiquitinated in a pH signal-independent manner (Fig. (Fig.2B),2B), although increased ambient pH activates the Rim pathway leading to Rim101 processing (Fig. (Fig.2C).2C). In addition, ubiquitinated Rim8 levels remained unchanged over 3 h following a shift from either pH 3.5 to pH 7 or pH 7 to pH 3.5 (data not shown).
A previous study showed that the pool of free Ub in the cell, which is depleted in a doa4 Δ mutant, can be restored to wild-type levels by expressing Ub under the CUP1 promoter without adding extra copper to the medium (38). In agreement with this study, immunoblot analysis of HA-Rim8 showed that ubiquitinated Rim8 is undetectable in a doa4Δ mutant and restored upon expression of Ub from the CUP1 promoter in the absence of extra copper (Fig. (Fig.2D,2D, upper panel, lanes 2 and 3). The same result was obtained when Ub lacking all of its Lys residues (Ub0K), which prevents the formation of polyubiquitin chains, was expressed (Fig. (Fig.2D,2D, upper panel, lane 4), indicating that the band of lower-level mobility detected in the wild-type strain represents monoubiquitinated Rim8. Control experiments showed that overexpression of Ub upon addition of extra copper to the medium results in the formation of an additional band, which is not detected when Ub0K is overexpressed (Fig. (Fig.2D,2D, lower panel, lanes 3 and 4). This band is presumed, owing to its size, to be diubiquitinated Rim8, which is consistent with the previously reported accumulation of di-Ub species upon Ub overexpression in a doa4Δ mutant (38). We next examined whether the conserved K521 residue in Rim8 box 1 is involved in Rim8 ubiquitination. Immunoblot analysis of HA-Rim8 showed that a Lys521Arg substitution abolishes Rim8 ubiquitination (Fig. (Fig.2E,2E, lane 2). As a control, substitution of a second Lys residue in Rim8 box 1, Lys527, had no effect (Fig. (Fig.2E,2E, lane 3). Thus, Rim8 is monoubiquitinated at residue K521 within its C terminus.
Previous work showed that mammalian β-arrestin-1 interacts with AIP4/Itch, an E3 Ub ligase of the Nedd4 family, to regulate CXCR4 endosomal sorting (5). As Rim8 interacts with a component of the endosome-associated ESCRT machinery, we tested whether Rsp5, the only member of the Nedd4 family in yeast, is involved in Rim8 ubiquitination. Immunoblot analysis of HA-Rim8 shows that Rim8 ubiquitination is undetectable in the npi mutant (Fig. (Fig.3A),3A), which produces very low levels of Rsp5 (51), suggesting that Rsp5 is responsible for Rim8 ubiquitination.
Rsp5 recognizes a large number of substrates through its WW domains (19). We showed that in vitro-synthesized, 35S-labeled Rim8 binds directly to a purified, bacterially expressed fusion of GST and the WW domains of Rsp5 (codons 228 to 430) in pulldown assays (Fig. (Fig.3B).3B). Previous studies showed that Rsp5 WW domains bind to PPXY or PXY motifs (11). A PXY motif (PKY; codons 506 to 508), located 12 residues upstream of the ubiquitinated Lys residue at the C terminus of Rim8 (Fig. (Fig.1A),1A), is highly conserved in Rim8 homologs in other Saccharomyces species (data not shown). Binding of in vitro-synthesized Rim8 to GST-Rsp5WW was almost abolished by a Pro506Ala substitution in the PXY motif (Fig. (Fig.3B),3B), suggesting that this motif mediates the direct interaction detected between these two proteins. To assess the physiological role of the PXY-mediated interaction between the Rim8 and Rsp5 WW domains, we tested the effect of the Pro506Ala substitution on Rim8 ubiquitination in vivo. Immunoblot analysis of HA-Rim8 shows that Rim8 ubiquitination was markedly reduced upon replacement of the Pro506 residue with Ala (Fig. (Fig.2E,2E, lane 4), which is consistent with the idea that the PXY motif in Rim8 mediates a functional interaction with Rsp5 in vivo.
To address the physiological role of the monoubiquitinated residue and the SXP motif in Rim8, we examined the effect on Rim signaling of mutations inactivating one or both of these elements. In agreement with previous work (32, 45, 61), abrogation of Rim signaling in a rim8 Δ mutant prevents processing of the Rim101 transcription factor (Fig. (Fig.4A,4A, lane 3), repression of the Rim101-dependent reporter CYC1-NRE-lacZ (Fig. (Fig.4B,4B, lane 3), and growth in the presence of lithium (Fig. (Fig.4C,4C, lane 3). Truncation of the C terminus of Rim8 (codons 505 to 542), which contains the ubiquitination site and the SXP motif, resulted in a stringent loss of function phenotype in vivo (Fig. (Fig.4A,4A, B, and C, lanes 2). The ESDP(533-536)/AAAA substitution, which eliminates the SXP motif and abolishes Vps23 binding, also produced a loss-of-function phenotype (Fig. (Fig.4A,4A, B, and C, lanes 8), thus indicating that binding of Vps23 to the Rim8 SXP motif plays an essential role in Rim signal transduction. In contrast, the Pro536Leu substitution, which allows some residual binding to Vps23 (see above), or the Lys521Arg substitution, which inactivates the ubiquitination site, had no detectable effect on Rim8 function (Fig. (Fig.4A,4A, B, and C, lanes 4 and 5). In agreement with these findings, inactivation of the PXY motif, which mediates the interaction with the Rsp5 Ub ligase, did not impair Rim signaling (data not shown). Strikingly, a combination of both Pro536Leu and Lys521Arg substitutions produced a strong loss-of-function phenotype (Fig. (Fig.4A,4A, B, and C, lanes 6). Together, these results suggest that partial inactivation of the SXP motif upon replacement of the Pro536 residue with Leu is not sufficient to block Rim signal transduction unless the ubiquitination site is mutated. In light of the fact that the Vps23 UEV domain binds Ub (25), these results are consistent with a contribution of the ubiquitinated residue to the interaction with Vps23 in vivo.
To assess the role of Rim8 ubiquitination in Vps23 binding, we used two different approaches. As Rim8 ubiquitination may be a rate-limiting step for the detection of two-hybrid interactions, we first examined whether Ub overexpression restores the two-hybrid binding of Vps23 to Rim8-P536L. Two-hybrid interaction between Vps23 and Rim8-P536L was restored to about 5% of the wild-type levels upon Ub overexpression, and this effect depends on the integrity of the ubiquitinated K521 residue since no two-hybrid interaction was detected between Rim8-K521R-P536L and Vps23 (Fig. (Fig.1D).1D). In contrast to Rim8-P536L, Rim8-ESDP(533-536)/AAAA did not interact with Vps23 when Ub was overexpressed (Fig. (Fig.1D),1D), further supporting the idea that the Pro536Leu substitution does not fully inactivate the SXP motif. We then carried out pulldown assays with purified, bacterially expressed GST-Vps23UEV as bait and Rim8 from yeast protein extracts. Ub was overexpressed to improve the detection of ubiquitinated Rim8. In agreement with a role for ubiquitin in Vps23-Rim8 interaction, the Vps23 UEV domain preferentially bound the ubiquitinated form of Rim8 (Fig. (Fig.5A,5A, lane 3), although binding to nonubiquitinated Rim8 was also detected under less stringent conditions (data not shown). In contrast, Rim8 ubiquitination did not improve its interaction with Rsp5 WW domains (Fig. (Fig.5A,5A, lane 4). Together, these findings indicate that Rim8 ubiquitination contributes to Vps23 binding.
Previous work showed that the translational fusion of the mammalian Vps23 homolog, Tsg101, to the HIV-1 Gag protein bypasses the requirement for a functional Tsg101-binding Gag PTAP motif in retrovirus budding (34). Then, we tested whether the fusion of Vps23 to C-terminally truncated Rim8 (codons 1 to 504), which lacks both the SXP motif and the ubiquitination site, bypasses the requirement for the C terminus of Rim8 in Rim signaling. Chimeric fusion of Vps23 to Rim8(1-504) fully restored Rim8 function (Fig. (Fig.4A,4A, B, and C, compare lanes 2 and 7) and pH-dependent regulation of Rim101 processing (data not shown). The same result was obtained by fusing a truncated form of Vps23, which lacks the UEV domain (data not shown). Thus, the essential role of the Rim8 C terminus, as the HIV-1 Gag PTAP motif, is to recruit the Vps23/Tsg101 UEV domain.
To confirm the interaction of Vps23 and Rim8 in vivo, we tested whether these proteins coimmunoprecipitate from cell extracts. A rim8 Δ strain was cotransformed with plasmids expressing functional Rim8-HA and Vps23-Flag from their own promoter on a centromeric vector, and protein extracts were prepared before and after a shift from pH 3.5 to pH 7 to activate the Rim pathway. Anti-HA immunoblot analysis of Flag-tagged Vps23 immunoprecipitates shows that a fraction of Rim8 coimmunoprecipitated with Flag-Vps23 in an ambient pH-independent manner (Fig. (Fig.5B).5B). In agreement with in vitro binding assays, Vps23 is preferentially associated with ubiquitinated Rim8 although a small fraction of the nonubiquitinated protein is detected in the precipitates. In control experiments, no Rim8-HA was precipitated by anti-Flag in the absence of Vps23-Flag (Fig. (Fig.5B).5B). These results further support the role of Rim8 ubiquitination in Vps23 binding and show that the Rim8-Vps23 interaction is independent of the pH signal. In agreement with these findings, coimmunoprecipitation of Rim8 and Vps23 was also observed in the absence of the putative 7TM pH sensor Rim21 (data not shown). A previous work showed that other components of the ESCRT-I complex are required for Rim signaling (62). We then tested the association of Rim8 with the ESCRT-I subunit Vps28. A rim8 Δ strain expressing chromosomal Vps28-Flag was transformed with a centromeric plasmid expressing Rim8-HA. Anti-HA immunoblot analysis of Flag-tagged Vps28 immunoprecipitates shows that ubiquitinated Rim8 coimmunoprecipitated with Flag-Vps28 (Fig. (Fig.5C).5C). Thus, Rim8 coimmunoprecipitates with both Vps23 and Vps28, strongly suggesting that binding of Rim8 to Vps23 mediates the association of Rim8 with the ESCRT-I complex.
Ubiquitination of PalF, the Rim8 homolog in A. nidulans, is fully dependent on the pH signal and the 7TM protein PalH, partially dependent on the 3TM protein PalI, and independent of the other specific components of the ambient pH signaling pathway (22). In contrast, and consistent with its lack of regulation by ambient pH, Rim8 ubiquitination does not require any specific component of the Rim pathway, including Rim21 and Dfg16, the two yeast homologs of the putative 7TM pH sensor PalH (Fig. (Fig.6A,6A, lanes 1 to 7). In addition, Rim8 ubiquitination was also detected in a strain carrying deletions of both RIM21 and DFG16 (data not shown). Unexpectedly, we found that Rim8 ubiquitination is undetectable in a vps23 Δ mutant strain (Fig. (Fig.6A,6A, lane 8). Thus, ubiquitination of Rim8 is independent of the activation of the Rim pathway but strongly dependent on the presence of its binding partner, Vps23. Overexpression of the Vps23 UEV domain is sufficient to restore Rim8 K521 ubiquitination in a vps23 Δ mutant (Fig. (Fig.6B,6B, lanes 2 and 3), although its does not restore Rim signaling (data not shown). Strikingly, the level of monoubiquitinated Rim8 increases up to 50% of the total protein, suggesting that binding of Vps23 UEV to Rim8 promotes the ubiquitination of Rim8. As the UEV domain binds both Ub and the SXP motif, we examined the contribution of each interaction to Rim8 ubiquitination under these conditions. We first generated a Vps23 UEV mutant derivative defective in Ub binding but still able to interact with the Rim8 SXP motif. Structural analysis of the Vps23 UEV domain in a complex with Ub showed that a patch of residues (codons 52 to 57) is specifically involved in Ub recognition (54). In agreement with the structural data, substitution of the last three residues of this patch to Ala [SDG(55-57)AAA] abolished the Ub-mediated two-hybrid interaction of Vps23 with Rim8 (Fig. (Fig.1D)1D) without affecting binding of Vps23 to the Rim8 SXP motif (Fig. (Fig.1A).1A). Overexpression of the SDG(55-57)AAA mutant derivative of Vps23 UEV did not restore Rim8 ubiquitination in a vps23 Δ strain (Fig. (Fig.6B,6B, lane 5), thus indicating that Rim8 ubiquitination is strictly dependent on the ability of the UEV domain to bind Ub. Binding of Vps23 UEV to the SXP motif was also required for Rim8 ubiquitination since ubiquitinated Rim8 was undetectable upon inactivation of the SXP motif in Rim8-ESDP(533-536)/AAAA, either in the presence or in the absence of Vps23 UEV overexpression (Fig. (Fig.6C,6C, lane 4, and D). In contrast, residual Vps23 UEV-induced ubiquitination of the K521 residue in Rim8-P536L was still detectable (Fig. (Fig.6C,6C, compare lanes 2 and 3), which is consistent with the residual Vps23-binding activity of the SXP motif in this mutant. Altogether, these findings indicate that binding of Vps23 UEV to Rim8 promotes its ubiquitination through a mechanism that involves binding of the UEV domain to both Ub and the SXP motif in Rim8.
Previous studies showed that Vps23 localizes to endosomal compartments and that this localization is lost in the absence of Vps27 (25, 26). As Rim8 appears to substitute Vps27 for recruitment of Vps23, we tested whether overexpression of Rim8 redirects Vps23 to another compartment. Therefore, we examined the localization of RFP-tagged Vps23 either expressed alone or expressed in the presence of GFP-tagged Rim8. Each fusion protein was expressed from its endogenous promoter on a 2μm high-copy-number plasmid. As shown previously (25), and consistent with its endosomal distribution, Vps23-RFP localizes to the cytoplasm and internal punctate structures in the absence of Rim8-GFP (Fig. (Fig.7).7). Cooverexpression with Rim8-GFP, however, resulted in recruitment of Vps23-RFP to cortical punctate structures, where it colocalized with Rim8-GFP (Fig. (Fig.7).7). Notably, some internal Vps23-RFP foci that do not coincide with Rim8-GFP foci were still visible under these conditions (Fig. (Fig.7,7, white arrowhead). Relocalization of Vps23-RFP was not observed when Vps23 was cooverexpressed with the Rim8-ESDP(533-536)/AAAA mutant protein, which fails to bind to Vps23 (Fig. (Fig.7).7). Under these conditions, as well as in the absence of overexpressed Vps23-RFP, Rim8-GFP localization was diffuse throughout the cytoplasm, with some nuclear enrichment. Together, these results indicate that colocalization of Rim8 and Vps23 in cortical punctate structures is strictly dependent on the interaction between these two proteins.
Previous genetic studies suggested that a Rim pathway-specific protein substitutes Vps27 for specialized recruitment of the ESCRT machinery. Here, we present evidence that the arrestin-related protein Rim8 mediates the recruitment of the ESCRT-I complex to the 7TM protein Rim21. Rim8 interacts with the C-terminal cytoplasmic tail of Rim21 and, as Vps27, binds directly to the Vps23 subunit of ESCRT-I. Bindings of Vps23 to Vps27 and Rim8 share common features, as they involve the Vps23 UEV domain and related P(S/T)XP and SXP motifs in Vps27 and Rim8, respectively. These findings support a model in which the specific recruitment of the ESCRT machinery to the 7TM protein Rim21 is mediated by Rim8 arrestin domains whereas recruitment to a large number of ubiquitinated cargos for MVB sorting is mediated by Vps27 UIM domains. In agreement with this model, staining with the dye FM4-64 indicates that the rim8 Δ mutant does not have an MVB defect (data not shown), which is consistent with previous work showing that Rim signaling components such as Dfg16, Rim21, and Rim101 are not required for MVB pathway function (62).
Both Rim8 and mammalian β-arrestins serve as adaptors linking 7TM proteins to the endocytic machinery. However, whereas β-arrestins bind to clathrin and the clathrin adaptor AP-2, Rim8 interacts with Vps23, a component of the ESCRT machinery. The Vps23-binding site is located at the C terminus of Rim8 in a position similar to that of clathrin and AP-2 binding sites in β-arrestins. Interaction of clathrin with β-arrestin-2 is regulated as accessibility of the clathrin binding site increases upon binding of arrestin to the activated 7TM receptor (60). In contrast, the Rim8-Vps23 interaction does not appear to be regulated by the pH signal in coimmunoprecipitation assays. In addition, pH sensing does not require the dissociation of Vps23 from Rim8, as shown by the ability of the Rim8(1-504)-Vps23 chimera to trigger the pH response. These findings support a model in which, in the absence of pH signal transduction, Rim8 is bound to Vps23. Coimmunoprecipitation of Rim8 with both Vps23 and Vps28 suggests that binding of Rim8 to Vps23 mediates the association of Rim8 with ESCRT-I. We hypothesize that, by analogy with the regulation of the receptor-arrestin complex in mammals, binding of Rim8 to the 7TM protein Rim21 mediates the recruitment of the Rim8-bound ESCRT-I complex in response to the pH signal. Vps23 appears to exist exclusively as a component of ESCRT-I (25), and our results indicate that a small fraction of Rim8, which is ubiquitinated, associates with Vps23 and Vps28 in coimmunoprecipitation assays. Ubiquitinated Rim8 is present in a much smaller amount than Vps23 in the cell, suggesting that Rim8 associates with a very small fraction of ESCRT-I. This finding could explain why Rim8 has not been identified as an ESCRT-I associated protein in previous studies.
According to the current model of the Rim pathway, Rim8-mediated endocytic internalization of Rim21 and/or Dfg16 would make the connection between the plasma membrane pH-sensing complex and the endosomal ESCRT-associated Rim101-processing machinery. Our findings that Rim8 binds to both Rim21 and Vps23 would provide the missing link between these two complexes. The hypothetical existence of two spatially segregated complexes in the Rim pathway was proposed on the basis that Rim20 colocalizes with the endosomal marker Snf7-RFP (8). However, it has been shown that C-terminal GFP fusion of Snf7 produces a class E mutant phenotype and accumulates on endosomes (53). Additionally, class E mutations that block ESCRT-III disassembly have been found to bypass the ambient-pH regulatory system, permitting Rim101 processing in the absence of pH-sensing components (20). Thus, we cannot exclude the possibility that location of Rim20 upon Snf7-RFP expression does not represent the Rim20 population that mediates the pH signal under physiological conditions. Then, alternative models, which do not involve spatially segregated complexes, can also be considered. Rim8 appears to associate with ESCRT-I in an ambient pH-independent manner, which is consistent with the idea that binding of Rim8 to the putative ambient pH sensor Rim21 could also mediate recruitment of ESCRTs to the plasma membrane. Such localization of ESCRTs would not be without precedent, since these protein complexes are involved in retroviral budding at the plasma membrane in mammalian cells (36). Additionally, previous studies of A. nidulans showed that components of the pH-sensing complex localize to the plasma membrane (9) and that the pH signaling component PalC, which binds ESCRT-III subunit Snf7, is recruited to the plasma membrane in an ambient pH-dependent manner (15). Interestingly, we found that overexpressed Rim8 and Vps23 colocalize at cortical punctate structures. These structures are not observed when the two proteins fail to interact or when either protein is overexpressed alone. This result, which provides additional evidence of the Rim8-Vps23 interaction, is consistent with the localization of Rim8-GFP expressed from the RIM8 genomic locus to peripheral puncta (33). However, these data should be considered with caution since Rim8 overexpression may cause mislocalization of Vps23 to the plasma membrane. Additionally, the fluorescence signal is not representative of ESCRT-I, since Vps23 is overexpressed and the other ESCRT-I subunits are not present stoichiometrically. Thus, further studies will be necessary to determine the location of Rim8-mediated recruitment of the ESCRT machinery.
Our results indicate that the SXP motif in Rim8 is the primary determinant of the interaction with Vps23. Inactivation of the SXP motif in the Rim8-ESDP(533-536)/AAAA mutant abolished Vps23 binding and Rim signaling. In contrast, we found that replacement of the Pro536 residue with Leu does not fully inactivate the SXP motif. Although interaction of Vps23 with Rim8-P536L was not detectable in two-hybrid assays, the residual binding detected between these two proteins in pulldown experiments appears to be sufficient to support Rim8 function in vivo. Several lines of evidence support the idea that the SXP motif in Rim8 is related to the previously described P(S/T)XP motif which binds both Vps23 and its mammalian homolog Tsg101 (2, 6, 16, 26). First, both SXP and P(S/T)XP motifs bind to the Vps23/Tsg101 UEV domain. Second, the SXP motif in Rim8 box 1 corresponds to a PSXP motif in the related box 2 sequence found in Rim8 homologs in other fungi. These proteins contain either box 1, box 2, or both, suggesting that SXP and PSXP motifs play the same function. Third, the last Pro residue in the HIV-1 Gag PTAP motif and in the Rim8 SXP motif plays an important role, as its replacement with a Leu residue markedly reduced in vitro binding to Tsg101 and Vps23, respectively. In contrast, the replacement of the first Pro residue of the Gag PTAP motif with Ala only moderately reduced Tsg101 binding (16), which is consistent with the ability of its yeast homolog Vps23 to bind both an SDP motif in Rim8 and PSDP motifs in Vps27. It is noteworthy that a recent study showed that the first Pro residue in the feline immunodeficiency virus (FIV) Gag PSAP motif is not required for its function in retrovirus budding (10).
We present genetic and biochemical evidence that a monoubiquitinated Lys residue located 12 residues upstream of the SXP motif contributes to Vps23 binding in vivo. Ub overexpression partially restores the two-hybrid interaction between Vps23 and the Rim8-P536L mutant protein, and this effect is dependent on the integrity of the ubiquitinated residue. In addition, pulldown and coimmunoprecipitation assays indicate that Vps23 preferentially binds to the ubiquitinated form of Rim8. In agreement with these results, partial inactivation of the SXP motif upon replacement of the Pro536 residue with Leu is not sufficient to block Rim signal transduction unless the ubiquitination site is mutated. The contribution of these two elements to Vps23 UEV binding is consistent with evidence that the UEV domain of Tsg101 can bind both Ub and the related PTAP motif simultaneously (52). Structural studies indicate that the P(S/T)XP and Ub binding sites in Tsg101 and Vps23 UEV could create a continuous binding surface upon monoubiquitination of a Lys residue upstream of the P(S/T)XP motif (42, 54). This model is supported by our findings and is consistent with the position of the monoubiquitinated residue in Rim8 and with the high level of conservation of the distance (12 residues) between this residue and the SXP motif in Rim8 homologs.
We further show that Rim8 ubiquitination is dependent on Rsp5, an E3 Ub ligase of the Nedd4 family. Ubiquitinated Rim8 is undetectable in an npi mutant, which has very low levels of Rsp5. In addition, a PXY motif located 12 residues upstream of the ubiquitinated residue in Rim8 mediates a direct interaction with Rsp5 WW domains and is required for Rim8 ubiquitination. This result is contrary to previously published data showing that Rim8 from yeast cell extracts, in contrast to other yeast arrestin-related proteins, does not bind Rsp5 WW domains in vitro (33). We found that Rim8 is ubiquitinated in a pH-independent manner, suggesting that the interaction with Rsp5 is not regulated by the pH signal. The extraction procedure used in this earlier study may not disrupt the association of Rim8 with endogenous Rsp5, which would prevent the binding of GST-WW to the PXY motif. Thus, our data provide further support to the role of yeast arrestin-related proteins as ubiquitin ligase adaptors (33, 37). Furthermore, our results indicate that the Rsp5-dependent monoubiquitination of Rim8 contributes to the recruitment of the ESCRT machinery. This function of Rim8 may be shared by other yeast arrestin-related proteins (ARTs) that regulate the internalization by endocytosis of specific plasma membrane proteins (33, 37). As mammalian β-arrestin-1 appears to play additional roles in later postinternalization events (5), it is possible that other yeast ARTs mediate the recruitment of the ESCRT machinery for sorting in the MVB pathway.
Our results indicate that binding of Vps23 UEV to Rim8 promotes the ubiquitination of Rim8. Rim8 ubiquitination is undetectable in a vps23 Δ mutant and is restored and even strongly increased upon overexpression of Vps23 UEV. In addition, this effect is dependent on the interaction of the UEV domain with both Ub and the SXP motif. A possible explanation for this unexpected result is that binding of Vps23 to Rim8 occurs in two steps. First, Vps23 would bind to the SXP motif in Rim8 and to Ub linked to the active site of the E3 Ub ligase Rsp5. These two interactions could be required for the correct positioning of Ub and its subsequent linkage to the Rim8 K521 residue. Rim8 ubiquitination would then stabilize the interaction of Rim8 with Vps23. Residual ubiquitination of Rim8-P536L would explain why this mutant protein remains functional unless the K521 residue is mutated. The hypothesis that binding of Ub to Vps23 UEV precedes the linkage of Ub to Rim8 is supported by previous work showing that the Mms2 UEV domain is required for the correct positioning of Ub for linkage-specific polyubiquitin chain formation (13). Although Rim8 ubiquitination appears to contribute to the interaction with Vps23 UEV, it does not play an essential role in Rim signaling. Inactivation of the ubiquitination site does not impair Rim signaling, and we observed only a slight defect in the kinetics of Rim101 processing in response to the pH signal (data not shown). Thus, further studies will be necessary to determine whether Rim8 ubiquitination plays a specific role in the fine-tuning of the Rim pathway.
Strikingly, our findings indicate that the same or related motifs mediate the recruitment of the ESCRT machinery to Pal/Rim proteins in fungi and retroviral Gag proteins in animal cells (Fig. (Fig.8).8). Recruitment of Tsg101, AIP1/Alix, and members of the Nedd4 family of Ub ligases to viral proteins is mediated by P(S/T)AP, YPXL, and PPXY motifs, respectively (36). Here, we show that SXP and PXY motifs located at the C terminus of Rim8 interact with the Tsg101 and Nedd4 fungal homologs Vps23 and Rsp5, and we previously found that YPX(L/I) motifs in PacC interact with the AIP1/Alix homolog PalA (58). The analogy between Pal/Rim signaling and retrovirus budding is further supported by other evidence. While the translational fusion of Tsg101 to HIV-1 Gag restores retrovirus budding in the absence of a functional Tsg101-binding PTAP motif in the Gag protein (34), fusion of Vps23 to C-terminally truncated Rim8, which lacks the Vps23-binding site, restores Rim8 function and Rim signaling. In addition, it has been suggested that the PTAP-mediated binding of HIV-1 Gag to Tsg101 is stabilized by Gag monoubiquitination (16). We show here that such a molecular mechanism actually occurs in the Rim signaling pathway, where both a monoubiquitinated residue and an SXP motif contribute to the interaction with Vps23 (Fig. (Fig.8).8). Finally, we found that overexpression of Rim8 redirects Vps23 to the plasma membrane in a manner reminiscent of the HIV-1 Gag-dependent relocalization of Tsg101 to the plasma membrane of mammalian cells (39). Thus, on the basis of these new findings and our previous identification of the YPXL motif as an AIP1/Alix binding site (58), we believe that the analysis of the Pal/Rim signaling pathway in fungi will provide new clues to the molecular mechanisms involved in retrovirus budding.
We thank Jacqueline Segall, Mark Hochstrasser, Linda Hicke, and Francisco Portillo for plasmids and strains.
This work was supported by grants from the Spanish CICYT (BFU2005-01970 and BFU2008-02005), the Comunidad de Madrid and CSIC (CCG07-CSIC/SAL-2145), and Intramural-CSIC (200620l094). A.H. is recipient of a CSIC-JAE predoctoral fellowship, and S.H. held a Plan de Formación de Personal Investigador studentship.
Published ahead of print on 22 December 2009.