The ESCRT-III Adaptor Protein Bro1 Controls Functions of Regulator for Free Ubiquitin Chains 1 (Rfu1) in Ubiquitin Homeostasis*

Background: Rfu1 localizes to endosomes and has a role in ubiquitin homeostasis. Results: Bro1 recruits Rfu1 to endosomes by binding through the Bro1 V domain and the YPEL motif of Rfu1, and Bro1 overexpression prevents Rfu1 degradation in heat-shocked cells. Conclusion: Bro1 regulates both deuibiquitinating enzymes Doa4 and Rfu1, which inhibits Doa4. Significance: The system that maintains ubiquitin homeostasis is elaborately regulated. Yeast Rfu1 (regulator for free ubiquitin chain 1) localizes to endosomes and plays a role in ubiquitin homeostasis by inhibiting the activity of Doa4. We show that Bro1, a member of the class E vacuolar protein sorting proteins that recruits Doa4 to endosomes and stimulates Doa4 deubiquitinating activity, also recruits Rfu1 to endosomes for involvement in ubiquitin homeostasis. This recruitment was mediated by the direct interaction between a region containing the YPEL motif in Rfu1 and the V domain in Bro1, which could be analogous to the interaction between the mammalian Alix V domain and YPXnL motifs of viral and cellular proteins. Furthermore, overexpression of Bro1, particularly the V domain, prevented Rfu1 degradation in response to heat shock. Thus, Bro1, a Doa4 positive regulator, regulated Rfu1, a negative regulator of Doa4. Rfu1 degradation partly involved the proteasome and a ubiquitin ligase Rsp5, suggesting that Rfu1 stability was regulated by ubiquitin-proteasome pathways.

Yeast Rfu1 (regulator for free ubiquitin chain 1) localizes to endosomes and plays a role in ubiquitin homeostasis by inhibiting the activity of Doa4. We show that Bro1, a member of the class E vacuolar protein sorting proteins that recruits Doa4 to endosomes and stimulates Doa4 deubiquitinating activity, also recruits Rfu1 to endosomes for involvement in ubiquitin homeostasis. This recruitment was mediated by the direct interaction between a region containing the YPEL motif in Rfu1 and the V domain in Bro1, which could be analogous to the interaction between the mammalian Alix V domain and YPX n L motifs of viral and cellular proteins. Furthermore, overexpression of Bro1, particularly the V domain, prevented Rfu1 degradation in response to heat shock. Thus, Bro1, a Doa4 positive regulator, regulated Rfu1, a negative regulator of Doa4. Rfu1 degradation partly involved the proteasome and a ubiquitin ligase Rsp5, suggesting that Rfu1 stability was regulated by ubiquitin-proteasome pathways.
Ubiquitin is a highly conserved 76-amino acid protein that is covalently attached to lysine residues within target proteins via a carboxyl-terminal glycine (1). The ubiquitin modification plays a critical role in regulating various cellular processes, such as protein degradation, protein trafficking, chromosome surveillance, apoptosis, and signal transduction. To perform such a remarkable array of cellular tasks, many ubiquitins are required in a cell. Ubiquitins are expressed abundantly from several ubiquitin-encoding genes, and ubiquitin expression appears to be regulated because either an excess or shortage of ubiquitin is detrimental to cells (2). Indeed, a cell has several different mechanisms to control and maintain ubiquitin homeostasis. For example, in response to heat stress, which induces protein ubiquitination, the transcription of polyubiquitin genes is upregulated to produce more ubiquitins (3). In addition, some deubiquitinating enzymes (Dubs) 2 play important roles in ubiquitin homeostasis. For example, mice with a mutation in the gene encoding UCH-L1, an abundant brain-specific Dub that causes gracile axonal dystrophy, exhibit reduced ubiquitin levels (4). Similarly, the ataxia (ax J ) mutation caused by reduced expression of Usp14, a proteasome-associated mammalian Dub, causes reduced ubiquitin levels (5). In yeast, mutations in Ubp6, an Usp14 homolog, cause reduced levels of monomer ubiquitin (6).
Endosomes facilitate the transport of membrane proteins between the plasma membrane and lysosomes/vacuoles (7). Many ubiquitins are used to regulate trafficking processes at endosomes (8). In yeast, Doa4 is an endosome-associated Dub that recovers ubiquitin from ubiquitinated cargos when they are incorporated into multivesicular bodies (9 -13). In addition, Doa4 contributes to ubiquitin recycling and is important for maintaining the monomeric ubiquitin pool within a cell. In the absence of Doa4, the level of monomer ubiquitin decreases and small ubiquitin species or free ubiquitin chains increase (12)(13)(14)(15). Previously, we identified Rfu1 (regulator for free ubiquitin chains 1) as an inhibitor of Doa4 (16). In the absence of Rfu1, the level of free ubiquitin chains decreased, and ubiquitin monomers increased. Moreover, Rfu1 inhibited the Doa4-mediated deubiquitination of a ubiquitinated cargo in vivo, and the deubiquitinating activity of Doa4 for ubiquitin chains in vitro. During heat shock, Rfu1 was degraded presumably to allow Doa4 to produce ubiquitin monomers to meet the need for more ubiquitin within the cell.
Interestingly, Doa4 is also regulated by another factor, Bro1. Bro1 is an ESCRT-III adaptor protein that binds to Snf7, one of the ESCRT-III subunits, to enhance the stability of ESCRT-III (17,18). Bro1 recruits Doa4 to endosomes and activates Doa4 deubiquitinating activity (19 -21) reported to have a similar ubiquitin profile as the doa4 mutation (22). Notably, recent studies showed that that the V domains of mammalian Alix, a homolog of Bro1, and of various yeast Bro1 were bound to K63-linked ubiquitin chains (23)(24)(25).
In this study, we examined the localization and degradation mechanisms of Rfu1 and revealed that both mechanisms are largely dependent on Bro1.
Yeast Strains-A list of the yeast strains used in this study is provided in supplemental Table S1. To delete BRO1 with HIS3, a PCR-generated EcoRV fragment carrying HIS3 was inserted into the blunted HindIII sites (ϩ446, ϩ2339) of BRO1 (nucleotide Ϫ434 to ϩ2585) in BSII to create E766. Using the E766 plasmid, a fragment covering Ϫ150 to ϩ2535 of BRO1 with HIS3 inserted into BRO1, was generated by PCR and then transformed.
Plasmids-A list of the plasmids used in this study is provided in supplemental Table S2. The Rfu1-GFP and Rfu1 mutant-GFP plasmids (pRfu1(1-200)-GFP, pRfu1(1-124)-GFP, and pRfu1(61-200)-GFP), in which the fusion proteins are expressed under the RFU1 promoter, were created as follows. SpeI and EcoRI fragments of RFU1 with the RFU1 promoter (Ϫ740 to Ϫ1) were PCR amplified using genomic DNA as a template. These PCR fragments were cut with SpeI, and EcoRI was inserted into the SpeI and EcoRI sites of pGCU10 (26) to create pRfu1(1-200)-GFP and pRfu1(1-124)-GFP. For pRfu1(60 -200)-GFP, two PCR fragments were obtained. The two fragments were cut with SpeI-BamHI and BamHI-EcoRI, respectively, and inserted into the SpeI and EcoRI sites of pGCU10.
Plasmids expressing HA-tagged Bro1-N, Bro1-C, and Bro1-V under the GPD promoter (E710, E711, and E772, respectively), were created as follows. PCR fragments were generated using a genomic library, cut with KpnI-SalI, and ligated with the EcoRI-SalI fragment of pRS426 and KpnI-EcoRI fragment of the 3HA-GPD promoter from E276.
Immunoblotting-Preparation of whole cell extracts and immunoblot analysis were performed as previously described (27) except cells were harvested in the early log phase. To analyze the overall ubiquitin profiles, total cell proteins were separated by 10 -20% gradient gels (Biocraft Inc.) using Tricinebased buffer, followed by transfer to Immobilon-P membranes (Millipore). Blots were incubated with mouse anti-ubiquitin monoclonal antibody (P4D1-HRP, Santa Cruz). Alternatively, the blots were incubated with a mouse anti-GFP monoclonal antibody (Roche Applied Science), anti-HA antibody (HA.11, COVANCE), or anti-yeast PGK antibody (Molecular Probes), followed by HRP-conjugated anti-mouse IgG (NA931V, Amersham Biosciences), and then visualized using ECL-plus reagent (Amersham Biosciences). To detect GST, an HRP-conjugated anti-GST antibody (Wako Chemicals) was used. A rabbit antiyeast Bro1 antibody was generated by immunizing with purified GST-Bro1 V.
In Vitro Binding between Various MBP-fused Rfu1 and GSTfused Bro1 Proteins-MBP or various MBP-fused Rfu1 derivatives (3 g each) were mixed with 18 g of GST or various GST-Bro1 fusion proteins in buffer B (50 mM Tris-HCl, pH 7.5, 100 mM NaCl, 10% glycerol) in a 200-l reaction volume for 1 h at 25°C, after which amylose resin was added. After 30 min, the resin was washed with buffer B and eluted with buffer B containing 10 mM maltose. The eluted samples were analyzed by immunoblot analysis using anti-MBP and anti-GST antibodies.
Microscopy-FM4-64 (Molecular Probes, Inc.) staining was performed as described (28). Cells were imaged at room temperature using a confocal microscope (LSM780; Carl Zeiss) equipped with an ␣Plan-Apochromat ϫ100 oil objective lens. Images were processed using the LSM image browser, and the brightness and contrast were adjusted using Adobe Photoshop CS4.

Bro1 Mediates the Endosomal Localization of Rfu1-Both
Rfu1 and Doa4 localize to endosomes (9,29). Based on coimmunoprecipitation analyses, Rfu1 was noted to bind to Bro1 as well as Doa4 (16). Because the endosomal localization of Doa4 is largely dependent upon Bro1 (19 -21), we tested whether Rfu1 localization was also affected by the presence of Bro1. The localization of exogenous Rfu1-GFP expressed under the RFU1 promoter was examined in ⌬rfu1⌬vps4 cells. In ⌬vps4 cells, endosome-localized ESCRT factors are concentrated in the class E compartment, an aberrant late endosomal structure that is adjacent to the vacuole (30), and FM4-64 specifically labels class E compartments and vacuolar membranes. As expected, Rfu1-GFP fluorescence was present at foci that overlapped with FM4-64-stained class E compartments in ⌬vps4⌬rfu1 cells (Fig. 1, A and B). In BRO1-deleted ⌬bro1⌬vps4 ⌬rfu1 cells, however, the localization of Rfu1-GFP at class E compartments was mainly lost. Many Rfu1-GFPexpressing cells showed only highly diffuse cytosolic GFP fluorescence. Some Rfu1-GFP-expressing cells showed much lower levels of punctuate GFP fluorescence at class E compartments against highly diffuse cytosolic GFP fluorescence. An anti-GFP immunoblot analysis showed that Rfu1-GFP was expressed at 10 -40% lower levels in ⌬bro1⌬vps4⌬rfu1 cells than in ⌬vps4⌬rfu1 cells, suggesting that Bro1 might also affect the stability of Rfu1 (Fig. 1C). We confirmed that the localization of Rfu1-GFP was unaffected in ⌬doa4⌬vps4⌬rfu1cells, suggesting that Doa4 was not involved in Rfu1 localization (Fig. 1, A and B). These results indicated that the endosomal localization of Rfu1 was mainly dependent on Bro1.
Bro1 and Rfu1 Directly Interact-The association between Bro1 and Rfu1 was detected even in the ⌬doa4 mutant (16). These results suggested that Doa4 was not required for the interaction between Bro1 and Rfu1, but rather that Bro1 and Rfu1 likely formed a direct interaction. We tested this possibility using purified recombinant proteins. MBP-Rfu1, but not MBP, specifically bound to GST-Bro1, but not GST (Fig. 2B,  lanes 8 and 9). We observed that purified recombinant GST-Bro1 was always in a doublet form, and partial degradation might occur during the expression and purification of GST-Bro1. We subsequently examined which region of Bro1 bound to Rfu1 ( Fig. 2A). We first divided Bro1 into the following two domains: the N-terminal Bro1 domain (Bro1-N, aa 1-388), which binds to Snf7 and stabilizes ESCRT-III (18,31) and the C-terminal region (Bro1-C, aa 388 -844). We found that MBP-Rfu1 specifically bound to GST-Bro1-C, but not GST-Bro1-N (Fig. 2B, lanes 10 and 11). The Bro1-C region could be divided into three regions. A search of the Pfam database revealed a domain called the V domain from aa 475-704 (32,33). Secondary structure alignments of Bro1 and Alix suggested that a region within Bro1 from approximately aa 370 -706 contains a complete V domain (33). It was previously reported that the Bro1 region from aa 793-844 was sufficient to mediate the interaction with Doa4 and that the PSVF sequence within aa 831-834 of Bro1 was also important for binding (20). Another study reported that Bro1 with a frameshift at aa 732 had impaired Doa4 binding (19). Thus, Bro1-C contains the V domain, a proline-rich region, and a binding region for Doa4 (20). An interaction was observed between MBP-Rfu1 and GST-Bro1-C⌬ (aa 388 -820, Fig. 2B, lane 12), which was defective in the Doa4-binding region. In contrast, MBP-Rfu1 did not interact with the Doa4-binding region of Bro1, GST-Bro1-D (aa 793-844; Fig. 2B, lane 13). We also observed an interaction between MBP-Rfu1 and GST-Bro1-V (aa 388 -720) in which the proline-rich region was mostly deleted (Fig. 1B, lane 14).
We determined that GST-Bro1-V (aa 388 -720) bound to MBP-Rfu1 as efficiently as GST fusion of a region containing a predicted complete V domain, Bro1 (aa 361-720; Fig. 2C). These results indicated that a region containing the Bro1 V domain is a binding site for Rfu1.
We subsequently generated MBP fusions for Rfu1 deletion mutants to determine which region of Rfu1 was responsible for Bro1-V binding (Fig. 3A). Rfu1 is a 200-amino acid protein, and aside from an N-terminal region with weak homology to the MIT (microtubule interacting and trafficking) domains, Rfu1 does not have distinct domains. Therefore, we generated various deletion mutants and tested their ability to bind to GST-Bro1-V (Fig. 3, A and B). We determined that Bro1-V bound to MBP-Rfu1(61-200), but exhibited significantly reduced binding to MBP-Rfu1(1-172) and MBP- Rfu1(1-140). These results indicated that the C-terminal region of Rfu1 encompassing aa 61-200, particularly the most C-terminal part of Rfu1, was responsible for Bro1-V binding.
In mammalian cells, the V domain of ALIX, a homolog of Bro1, binds to YPX n L motifs (where X varies in sequence and length), which are found in the viral Gag proteins and cellular proteins (32)(33)(34)(35). We searched for the YPX n L motif within Rfu1. A YPEL sequence was detected at aa 190 -193, which is the most carboxyl-terminal region of Rfu1, and was within the region for Bro1 V binding (Fig. 3C). We therefore tested whether the YPEL sequence might be a binding motif for the Bro1 V domain. Indeed, the binding activity for GST-Bro1-V was reduced in a mutant in which YPEL was changed to AAEL in MBP-Rfu1(aa 61-200), indicating that the YPEL motif was important for binding to the Bro1 V domain (Fig. 3D).
Importance of the YPEL Motif of Rfu1 for the Localization of Endosome and Ubiquitin Homeostasis-The importance of the C-terminal region and the YPEL motif in Rfu1 for Bro1 binding was verified by examining the localization of various Rfu1-GFP fusions expressed in yeast (Fig. 4). When an N-terminal-truncated form of Rfu1(60 -200)-GFP was expressed in ⌬vps4⌬rfu1 cells, GFP foci co-localized with the class E compartments and exhibited a similar fluorescence pattern as Rfu1(1-200)-GFP, although its expression was lower than that of Rfu1(1-200)-GFP. In contrast, Rfu1(1-124)-GFP fluorescence localization was different. A significant decrease in specific localization on class E compartments were observed, and many Rfu1(1-124)-GFP expressing cells showed higher diffusive fluorescence; these fluorescence patterns were similar to those of Rfu1(1-200)-GFP expressed in BRO1-deleted cells (Fig. 1A). These results suggested that the carboxyl region was important for the endosomal localization of Rfu1-GFP.
We also examined the localization of a YPEL-mutated form of Rfu1(Y190A,P191A)-GFP. Its endosomal localization was mainly lost and there were considerably smaller and fewer foci at class E compartments with a more highly diffuse background (Fig. 4A). We observed that the expression level of Rfu1(Y190A,P191A)-GFP was slightly lower than that of Rfu1(1-200)-GFP, but much higher than that of Rfu1(60 -200)-GFP, which showed foci at the class E compartments in ⌬vps4⌬rfu1 cells (Fig. 4B). These results suggested that the YPEL motif was important for the endosomal localization of Rfu1.

Bro1-dependent Functions of Rfu1
We subsequently determined whether the YPEL motif of Rfu1 was important for ubiquitin homeostasis by examining cellular ubiquitin profiles. Rfu1-GFP or Rfu1(Y190A,P191A)-GFP was expressed in ⌬rfu1 cells, and cellular ubiquitin profiles were examined (Fig. 5A). Wild type cells grown in early growth phase showed a bulk ubiquitin profile with monomeric ubiqui-  tin, free ubiquitin chains, and slowly migrated high-molecular weight forms (16, 36 -38). As reported previously, the amount of free ubiquitin chains was decreased, whereas the level of monomeric ubiquitin was increased in ⌬rfu1 cells (16). We found that the expression of Rfu1-GFP, but not Rfu1-(Y190A,P191A)-GFP restored a normal ubiquitin profile, indicating that the YPEL motif of Ruf1 was critical for maintaining ubiquitin homeostasis by Rfu1. We investigated whether the YPEL motif of Rfu1 functioned in the interaction with Bro1 in vivo (Fig. 5B). Immunoprecipi-tation analysis using lysates from ⌬rfu1 cells expressing Rfu1-GFP or Rfu1(Y190A,P191A)-GFP was performed. Endogenous Bro1 was specifically immunoprecipitated by Rfu1-GFP, but not Rfu1(Y190A,P191A)-GFP. These results indicated that the YPEL motif of Rfu1 was important for ubiquitin homeostasis and Bro1 binding in vivo.
Rsp5 Contributes to Rfu1-3F Degradation-Rfu1-3xFLAG (Rfu1-3F) is degraded upon heat shock (16), possibly allowing more Rfu1-free Doa4 deubiquitination to produce ubiquitin monomers. Concurrently with the interaction analysis of Bro1 and Rfu1, we investigated the mechanism of Rfu1 degradation during heat shock. We found that degradation of exogenously expressed Rfu1-3F was partially inhibited in the rpt1 mutant, which has a temperature-sensitive mutation in one of the 26 S proteasome subunits (Fig. 6, A and B), suggesting that proteasomal degradation was involved. In contrast, Rfu1-3F degradation was not impaired in the ⌬pep4⌬pbr1 mutant in which many vacuolar degradation events were inhibited (Fig. 6C). In addition, we found that Rfu1-3F degradation was not impaired by the doa4 mutation (Fig. 6D).
Bro1 Overexpression Prevents Rfu1-3xFLAG Degradation Upon Heat Shock-Bro1 recruits Doa4 to endosomes and activates the deubiquitinating activity of Doa4 (20). Therefore, it is possible that Bro1 has other effects on Rfu1 in addition to 3 Y. Kimura, J. Kawawaki, and K. Tanaka, unpublished observation.

Bro1-dependent Functions of Rfu1
AUGUST 1, 2014 • VOLUME 289 • NUMBER 31 recruiting Rfu1 to endosomes. We therefore tested the effects of Bro1 overexpression on Rfu1-3F degradation. Cycloheximide was added when the temperature was shifted to 39°C to prevent further production of Rfu1-3F. We observed that overexpression of full-length Bro1 inhibited Rfu1-3F degradation (Fig. 7A). We then examined the effects of overexpressing HAtagged Bro1-N, Bro1-C, and Bro1-V. We observed that Bro1-C as well as Bro1-V inhibited Rfu1-3F degradation (Fig. 7, C and  D). Bro1-N expression had no effect, although it was expressed at a significant level (Fig. 7B). These results indicate that the Bro1 V domain expression inhibited Rfu1 degradation after heat shock.

DISCUSSION
In addition to trafficking and sorting various cargos, endosomes seem to have an important role in ubiquitin homeostasis. In this study, we showed that the endosome localization of Rfu1 was important for ubiquitin homeostasis, and this was achieved by an Rfu1 interaction with Bro1. It was reasonable to assume that Rfu1 must be present on endosomes to elicit its function, because Doa4, a target of Rfu1, is localized at endosomes. Many ubiquitins are utilized on endosomes, and Bro1 recruits both Doa4 and Rfu1 for them to function. Thus, Bro1, a positive regulator of Doa4, also regulates a negative regulator of Doa4, Rfu1 (Fig. 8). Although the detailed regulation of Bro1 on Doa4 and Rfu1 is still unclear, from previous and present analysis, Bro1 was observed to bind to both Rfu1 and Doa4, but different regions within Bro1 mediate the binding to Doa4 and Rfu1; the C-terminal region is required for the interaction with Doa4, and the V domain mediate binding to Rfu1. Therefore, one Bro1 molecule may simultaneously interact with Doa4 and Rfu1, and regulate them. Certainly, it will be important to determine how the three proteins interact with each other.
Mammalian Alix V binds to the YPX n L motif of viral and cellular proteins, and these interactions are important for virus budding, multivesicular body sorting, or generation of exosomes (32)(33)(34)(35). Our analysis revealed for the first time that the yeast Bro1 V domain also binds the YPX n L motif (YPEL in Rfu1) and has important functions as well, notably recruiting Rfu1 through the YPEL motif and preventing Rfu1 degradation when

Bro1-dependent Functions of Rfu1
AUGUST 1, 2014 • VOLUME 289 • NUMBER 31 it was overexpressed. Therefore, although amino acid sequences are not entirely highly conserved between the V domains of Alix and Bro1, the specific binding mode for a YPX n L motif may be similar between them.
Recently, it was reported that the Bro1 V domain binds to K63-linked ubiquitin chains (25). Thus, it is tempting to speculate that the Bro1 V domain binds to both Rfu1 and K63linked ubiquitin chains. In a cell, binding of the V domain to Rfu1 may affect the efficiency of the V domain binding to ubiquitin chains, or vice versa, and the regulated binding of Bro1 V domain to Rfu1 and ubiquitin chains may result in the regulation of ubiquitin homeostasis. Further detailed analysis on the V domain is required to clarify this point.
We also addressed the mechanism of Rfu1 degradation in response to heat shock. Our results showed that Rfu1 degradation was partially inhibited in the rpt1 mutant and that degradation was not inhibited by ⌬pep4⌬prb1, suggesting that at least part, if not all, of Rfu1 is degraded by the proteasome. Furthermore, we found that Rsp5, a ubiquitin ligase, was required for Rfu1 degradation. Thus, a regulator of ubiquitin homeostasis may be regulated by a ubiquitin-related system; however, the detailed mechanisms for Rfu1 degradation along with the role of Bro1 in Rfu1 degradation remains to be investigated. Moreover, further analysis is required to clarify whether Rsp5 is directly or indirectly involved in Rfu1 degradation. Although we have not been able to detect the ubiquitinated form of Rfu1 by immunoprecipitation analysis in MG132-treated heat-shocked cells, 3 Rfu1 may be ubiquitinated by Rsp5 during heat shock, and Bro1 may prevent Rfu1 ubiquitination. It is noteworthy that Rsp5 is also related to ubiquitin homeostasis (41).
In conclusion, we showed that Bro1 regulates the localization and the activity of Rfu1. Our data suggests that the system that maintains ubiquitin homeostasis would be elaborately regulated.