Ulp1 localization to the nuclear envelope and the septin ring
As part of a larger study to identify how Ulp1 is targeted to its mitotic desumoylation substrates, we analyzed the localization of GFP-tagged versions of both the full-length wild-type (WT) Ulp1 and a catalytically inactive mutant of Ulp1 (Ulp1C580S) in G2/M-arrested yeast cells (see Methods and Additional file 3). The C580S mutation replaces the catalytic cysteine with a serine residue, rendering the Ulp1 SUMO protease catalytically inactive [17]. Both fusion proteins were expressed under the control of the Ulp1 promoter on low-copy plasmids, and images were collected using a fluorescence microscope. Consistent with its localization to NPCs, WT Ulp1 stained only the NE of arrested yeast cells (Figure 1A, left). Unexpectedly, however, full-length Ulp1C580S was enriched at both the bud-neck and the NE of G2/M-arrested cells (Figure 1A, right). This bud-neck localization of Ulp1C580S is reminiscent of the localization of the septin ring. Several sumoylated septins have been shown to be Ulp1 substrates, and we show in this study that the septin Cdc3 is highly sumoylated during G2/M arrest (Figure 1B). Furthermore, a catalytically inactive Ulp1 mutant colocalizes with the septin Cdc11 in G2/M-arrested (noc) cells (Figure 1C). Therefore, Ulp1C580S resides at the bud-neck localized septin ring.
Our data suggest that introducing the C580S mutation into the catalytic domain of Ulp1 somehow alters the subcellular distribution of this SUMO protease, causing it to localize with a bud-neck-associated substrate, possibly a sumoylated septin protein. Localization changes have also been reported for catalytically inactive, substrate-trapping mutants of phosphatases that form stable complexes with their substrates in vivo [43].
SUMO conjugation is required for Ulp1 localization to the septin ring
We tested whether the C580S mutation that visually increased the ability of Ulp1 to associate with the septin ring in vivo was, in fact, SUMO-dependent. For this purpose, the Ulp1C580S construct was expressed in two Smt3 mutants (smt3-331 and smt3-R11, 15, 19) or two SUMO pathway mutants (ubc9-1, siz1Δ siz2Δ) [13, 39, 41, 44–46]. Logarithmically growing cells of each mutant were arrested in G2/M, and images were collected to assess the septin ring localization of Ulp1C580S in comparison to an SMT3 WT strain. In our analyses, we found that in the absence of both SUMO chains (in the R11, 15, 19 mutants) and a mutant SUMO protein (in the smt3-331 mutant), the localization of Ulp1C580S to the septin ring was reduced (22% in smt3-331 and 36% in smt3-R11, 15, 19) in frequency and intensity but not abolished (Figure 2). We obtained different results in the ubc9-1 strain, a mutant of the SUMO E2-conjugating enzyme which impairs SUMO conjugation, and in the siz1Δ siz2Δ strain, a SUMO E3 ligase double-mutant that lacks sumoylation of septins and many other proteins [40, 41, 44]. Consistent with a role for Smt3 in the localization of Ulp1C580S, we were unable to detect septin ring localization of Ulp1C580S in ubc9-1 and siz1Δ siz2Δ strains. However, Ulp1C580S was retained at the NE (Figure 2A). As an additional control, the septin ring localization of GFP-tagged Smt3 was undetectable in both ubc9-1 and siz1Δ siz2Δ strains (Figure 2B).
Smt3 conjugation is required for Ulp1 localization to the septin ring. Therefore, Ulp1 is targeted to the septin ring of dividing cells in a SUMO-dependent fashion. Our data also suggest that the formation of SUMO chains on substrates may enhance this targeting of Ulp1.
Distinct and separate Ulp1 domains are required for localization to the septin ring
Our finding that a single point mutation in Ulp1, C580S, dramatically enhanced the localization of full-length Ulp1 to the septin ring in a SUMO-dependent fashion warranted a more detailed analysis of the targeting domains in Ulp1. Therefore, we generated a collection of GFP-tagged Ulp1 truncations and domains that were expressed under control of the Ulp1 promoter. We reasoned that the truncations and domains of Ulp1 that retained substrate targeting information would also localize to the septin ring in G2/M-arrested cells. In all, we assessed the localization of ten GFP-tagged constructs in comparison to full-length WT Ulp1 and full-length Ulp1C580S (C580S). Our choice of individual constructs was guided by previous findings that Ulp1 consists of functionally separate domains. These domains include a Kap121-binding domain with a role in septin localization (region 1), a Kap95- and Kap60-binding domain with a role in NPC anchoring (region 2), a cc domain harboring a nuclear export signal (CC) and the catalytic UD (region 3) [25–27]. Representative images of these domains and their subcellular localization are shown in Figures 3A and 3B. As previously reported, we found that the Ulp1 protein lacking region 2 (Δ2) localized to the septin ring in the majority of large-budded, arrested cells [27]. Therefore, region 2 of Ulp1 normally antagonizes localization and/or retention at the septin ring. This result is complemented by our novel finding that the full-length Ulp1C580S localized to the septin ring in 33% of all arrested, large-budded cells (Figures 1A and 3A).
Next we investigated other residues of Ulp1 that could affect the septin ring localization of the Ulp1C580S mutant, possibly by interfering with its targeting to sumoylated substrates. Aspartate 451 (D451) in Ulp1 is required to form an essential salt bridge with arginine 64 of Smt3 [31, 47]. Therefore, we introduced a D451N mutation into Ulp1C580S and found that it abolished the accumulation of the full-length Ulp1 double-mutant (D451N and C580S) at the septin ring (Figure 3A). This finding underscores the importance of Smt3 in targeting full-length Ulp1 to the septin ring shown in Figure 2. Additionally, it may indicate that D451 is required for targeting of sumoylated proteins and the C580S mutation is required for retention of Ulp1 at the septin ring.
Most intriguingly, we found that a truncation consisting only of region 3 with the C580S mutation (Ulp1(3)(C580S)) displayed robust septin ring localization in 59% of cells (Figures 1C, right panel, and Figure 3B). In stark contrast, regions 1 and 2 and WT region 3, lacking the C580S mutation, failed to localize to the septin ring (Figures 3A and 3B). However, in strains with diffuse Ulp1 truncations, the septin ring stays intact. Therefore, necessary and sufficient SUMO-dependent targeting information is contained in region 3 of Ulp1, but not in regions 1 and 2. The latter conclusion is confirmed by two-hybrid assays with Smt3.
The previously published cocrystal structure of Ulp1 with Smt3 (MMDB database 13315) reveals that amino acids 418 to 447 of region 3 make extensive contact with Smt3 and constitute an exposed SBS [31] (see also Additional files 1 and 2). The SBS is situated next to, but does not include, the critical D451 residue that contacts Smt3 [31, 47]. Additionally, deletion of this SBS in region 3 of Ulp1 abolishes the complementation of a ulp1Δ deletion mutant [26]. In an attempt to identify critical residues in the evolutionary conserved SBS domain, we used psi-blast to compare the protein sequence of the yeast Ulp1 catalytic domain with all nonredundant protein sequences in the National Center for Biotechnology Information database for seven iterations and limited the output to the top 250 matches. Our results contained 81 different species. Among these species, 61% of the sequences were identified as verified or predicted sentrin/SUMO protease/Ulp1 genes, 24% were identified as unnamed protein products or hypothetical genes and 15% were classified as "other" (crystal structures, unanalyzed sequences and so on). The alignment of these sequences allowed us to identify areas of strong conservation (see Additional file 1). Using this approach, we identified several highly conserved residues in the SBS. However, these amino acids did not contact Smt3 in the published cocrystal structure and likely play structural roles in Ulp1 folding [31].
We investigated the effect of deleting the entire SBS domain on the localization of Ulp1(3)(C580S). A Ulp1(3)(C580S)SBSΔ construct did not localize to the septin ring in the majority of cells (96%). These results match those obtained by Li and Hochstrasser [26] using a WT Ulp1(3)ΔSBS construct (C173). We confirmed that SBSΔ and other Ulp1(3) constructs are expressed as soluble proteins, suggesting that they are not grossly misfolded. We also cloned and expressed the SBS domain as a fusion with GFP (SBS-GFP). This construct was distributed diffusely throughout the cell and failed to localize to the septin ring (Figure 3, middle). These data suggest that the SBS domain of region 3 may be required for the initial interaction with sumoylated substrates, but additional features of Ulp1 are required for targeting (D451) and retention (C580S) of this SUMO protease at the septin ring.
Next we directed our attention to the temperature-sensitive ulp1ts-333 allele. This mutant allele causes cells to arrest in mitosis and accumulates unprocessed SUMO precursor and sumoylated proteins [17]. Our ulp1ts construct of region 3, Ulp1(3)ts, contains three mutations (I435V, N450S and I504T), and introduction of C580S into Ulp1(3)ts showed a greatly reduced incidence and intensity of septin ring localization (compare panels in Figures 3B and 3C). We noted that the (N450S) mutation in the ts construct was located next to the salt bridge-forming residue D451 described above and that both residues were highly conserved in the consensus sequence of Ulp1-like molecules (Additional file 1). This suggests that residues altered in ulp1ts-333, specifically N450, may contribute to Smt3 interaction and possibly substrate targeting. It is possible that N450S perturbs the salt-bridge interaction formed between D451 of Ulp1 and R64 of Smt3, thus reducing the interaction with Smt3 and contributing to the temperature-sensitive phenotype. In support of this hypothesis, correction of the N450S mutation in Ulp1(3)ts (S450N) partially rescued the slow growth defect of a ulp1Δ strain at 30°C and 37°C (data not shown). The effect of the ulp1ts mutation on Ulp1's ability to interact with Smt3 is explored in more detail below.
We tested which domains of Ulp1 are required for targeting and retention at the septin ring in vivo. Using our region 1 and region 2 GFP-tagged constructs (see Figure 3A), we show that septin-targeting information is not contained in the domains that are known to interact with karyopherins. The Δ2 construct recapitulates the previous finding that Kap121-binding to region 1 regulates access of Ulp1 to the septin ring. The full-length Ulp1C580S mutant reveals that a single substrate-trapping mutation in Ulp1 suffices to enrich Ulp1 at the septin ring. To show that an Smt3 interaction is required for the septin localization of Ulp1C580S, we created the double-mutant (D451N and C580S). The D451N mutation is known to destroy an essential salt bridge formed between Smt3 and Ulp1. Next, using the Ulp1(3)C580S construct, we show that the septin-targeting information is limited to region 3 of Ulp1 (Figure 3B). Further truncating Ulp1(3)C580S revealed that a previously identified SBS domain in Ulp1(3)C580S is also involved in septin targeting and retention. To test whether mutations found in region 3 of the ulp1ts mutant play a role in septin localization, we introduced three additional mutations, I435V, N450S and I504T, into Ulp1(3)C580S. This Ulp1(3)ts C580S construct showed a reduced ability to enrich at the septin ring (Figure 3C), suggesting that its ability to interact with sumoylated septins may be reduced but not abolished.
Kap121-independent SUMO-targeting information resides in the catalytic domain of Ulp1
In the preceding sections, we described our identification of necessary and sufficient substrate-targeting information in the catalytic domain (region 3) of Ulp1. However, region 3 of Ulp1 may not be the only domain involved in targeting to the septin ring. Region 1 of Ulp1, the Kap121-binding domain, has previously been implicated in septin targeting. Specifically, it has been reported that Kap121 is required for targeting Ulp1 to the septin ring during mitosis [27]. Therefore, we decided to assess the role of Kap121 in the substrate-targeting of Ulp1(3)(C580S). Specifically, we used a kap121ts mutant [48] to assess the septin ring-targeting of WT Ulp1, full-length Ulp1C580S and Ulp1(3)(C580S). In our analysis, we found that full-length Ulp1C580S required Kap121 function for targeting to the septin ring. At the permissive temperature (30°C), Ulp1C580S demarcated the NE and septin ring of G2/M-arrested cells. After a shift to the nonpermissive temperature, however, Ulp1C580S could no longer be detected at the septin ring (Figure 4, middle). Surprisingly, the Ulp1(3)(C580S) truncation was localized to the septin ring at the permissive and nonpermissive temperatures in a kap121ts strain. As shown herein, Ulp1(3)(C580S) resided both inside the nucleus and at the septin ring at 30°C and 37°C (Figure 4, right).
Our data suggest that Ulp1 contains both Kap121-dependent and Kap121-independent septin ring-targeting information. The only requirement to detect full-length Ulp1 at the septin ring is the C580S mutation and functional Kap121 (Figures 1, 2 and 4). In contrast Ulp1(3)(C580S), which lacks all domains required for NPC interaction through Kap121, Kap60 and Kap95, localizes to the septin ring and inside the nucleus. This finding provides strong evidence that Kap121-independent septin ring-targeting information resides in the catalytic domain (region 3) of Ulp1.
Multiple features in the catalytic domain of Ulp1 affect SUMO interactions
Our finding that a single amino acid change in the catalytic domain of Ulp1 results in greatly enhanced, SUMO-dependent localization to the septin ring also prompted us to investigate the two-hybrid interactions of Ulp1 with budding yeast SUMO (Smt3-BD; Smt3 fused to the Gal4 DNA-binding domain). Full-length WT Ulp1, the full-length catalytically inactive Ulp1C580S mutant, the Ulp1 Kap121-interacting domain (region 1), the Ulp1 Kap60/Kap95-interacting domain (region 2) and the catalytic domain (region 3) failed to interact with Smt3-BD (data not shown). However, the catalytically inactive Ulp1(3)(C580S) truncation interacted reproducibly and above background with Smt3 (see Figure 5, C580S).
Ulp1(3)(C580S) appears to interact only weakly with our Smt3 two-hybrid bait construct, as indicated by fewer colonies on the reporter media (Figure 5, Ade). However, such an interpretation assumes an equally available pool of both bait and prey. One possible explanation for this result is that the Ulp1(3)(C580S) two-hybrid prey construct interacts with a number of available substrates in the cell (for example, free Smt3 and other sumoylated proteins), and, as a result, this sequestration is no longer available to the Smt3 two-hybrid bait construct, thus creating the appearance of a weak interaction. We reasoned that introduction of the ulp1ts mutations could weaken the potential substrate-trapping phenotype of Ulp1(3)(C580S), making more of the pool available to engage the Smt3 bait construct. Consistent with this model, when we introduced the ulp1ts mutations into the substrate-trapping Ulp1(3)(C580S) prey construct, we observed a more robust two-hybrid interaction with the Smt3 bait (Figure 5, compare C580S and ts(C580S). It must also be noted, however, that currently we cannot fully explain the variations in our in vivo and in vitro assays used to assess ability of Ulp1(3)(C580S) to interact with Smt3.
Next we focused on the D451N mutant of Ulp1 that prevents the interaction of Ulp1 with SUMO [31, 47]. As shown above, D451N, when introduced into Ulp1(C580S), prevents localization of this construct to the septin ring (Figure 3A, C580S/D451N). Correspondingly, we found that introduction of the D451N mutation into Ulp1(3)(C580S) completely abolished the two-hybrid interaction with Smt3 (Figure 5, compare C580S, D451N/C580S and D451N). These observations provide evidence that the targeting of Ulp1 to sumoylated substrates is a closely balanced act involving both Smt3 targeting and retention.
Ulp1(3)(C580S)truncation binds SUMO and SUMO-modified proteins
We hypothesized that if Ulp1(3)(C580S) were to interact avidly with Smt3, this mutated moiety of Ulp1 could efficiently interact with SUMO adducts in vitro. Therefore, to test the direct interaction of Ulp1(3)(C580S) with SUMO, we fused this domain to the carboxy terminus of maltose-binding protein (MBP) and expressed the recombinant fusion protein in bacteria. Subsequently, the MBP-Ulp1(3)(C580S) fusion protein was purified from bacterial extracts and bound to amylose resin (see Methods). As a control to assess the ability of MBP-Ulp1(3)(C580S) to interact with sumoylated proteins, we also purified a second MBP-fused Ulp1(3)(C580S) construct lacking the SBS domain (3(C580S)ΔSBS).
First we determined the ability of MBP-Ulp1(3)(C580S) to affinity-purify sumoylated proteins from crude yeast cell extracts. ulp1ts-333 cells expressing FLAG-tagged SMT3 were grown to log phase prior to preparation of yeast cell extracts (see Methods). These extracts were then incubated with resin-bound MBP-Ulp1(3)(C580S), MBP-Ulp1(3)(C580S)-ΔSBS or unbound amylose resin. After washing, bound yeast proteins were eluted, separated on SDS-PAGE gels and examined by Western blot analysis with an anti-FLAG antibody. Flag-SMT3-modified proteins present in the whole-cell extracts (WCEs) (Figure 6, lane 2) could clearly be detected bound to MBP-Ulp1(3)(C580S) (lane 5) but not to the MBP-Ulp1(3)(C580S)-ΔSBS control protein (lane 4). We identified both unconjugated Flag-Smt3 proteins as well as several higher-molecular-weight adducts. These data suggest that Ulp1(3)(C580S) can efficiently bind and enrich sumoylated proteins from crude yeast cell extracts. To demonstrate the versatility of Ulp1(3)(C580S)-aided Smt3 purification, we also purified monomeric and conjugated GFP-Smt3 from yeast cells (Figure 6B). Additionally, we probed the extracts and eluted proteins shown in Figure 6B with an anti-Cdc11 antibody, which revealed the specific copurification of Cdc11 with immobilized Ulp1(3)(C580S) (Figure 6C).
In the reciprocal experiment, we tested whether a GFP-tagged Ulp1(3)(C580S) construct expressed in yeast cells could bind immobilized SUMO2, which is highly conserved to yeast Smt3. In this experiment, yeast cells expressing CEN-plasmid levels of GFP-tagged Ulp1(3), Ulp1(3)(C580S) or the Ulp1(3)(C580S)-ΔSBS (see Figure 3) were grown to log phase prior to preparation of yeast cell extracts. Individual extracts were then incubated with SUMO2 immobilized on agarose beads (see Methods). After washing, bound yeast proteins were eluted, separated on SDS-PAGE gels and examined by Western blot analysis with an anti-GFP antibody. This time the GFP-tagged Ulp1(3)(C580S) could be detected in the WCEs and bound to the SUMO2 agarose (Figure 6D). In contrast, neither the WT catalytic domain of Ulp1 (Ulp1(3)) nor Ulp1(3)(C580S)(SBSΔ) was bound to SUMO2 agarose. Similarly, the Ulp1(3)(C580S) could also be purified on SUMO1 agarose (data not shown).
We also tested whether immobilized Ulp1(3)(C580S) could be used to purify SUMO chains. In this experiment, we incubated purified SUMO2 chains with our immobilized Ulp1(3)(C580S) or the unbound amylose resin. After washing, bound SUMO2 chains were eluted, separated on SDS-PAGE gels and examined by Western blot analysis with an anti-SUMO2 antibody. SUMO2 chains could clearly be detected in the input (Figure 6E, lane 2) and bound the MBP-Ulp1(3)(C580S) (lane 4), but not the resin-only control (Figure 6E, lane 3). Both lower- and higher-molecular-weight adducts of SUMO2 were purified with a preference for higher-molecular-weight chains (5-mer, 6-mer or 7-mer). These data suggest that Ulp1(3)(C580S) can efficiently bind and enrich SUMO2 chains in vitro and that the MBP fusion of Ulp1(3)(C580S) may also be useful for the purification of sumoylated proteins from mammalian cells.
A SUMO2-binding platform for substrate ubiquitination
STUbLs such as the yeast Slx5/Slx8 heterodimer and the human RNF4 protein efficiently ubiquitinated proteins modified with SUMO chains [49, 50]. These proteins interact with their respective sumoylated ubiquitinated targets through SIMs. STUbL reactions have been reconstituted in vitro, but the purification of target proteins modified with SUMO chains has been technically difficult or prohibitively expensive. The ability of Ulp1(3)(C580S) to interact with SUMO may therefore provide a simple way to purify a SUMO chain-modified STUbL target of choice.
To test whether Ulp1(3)(C580S) can serve as a platform to modify a purified protein with SUMO2 chains, we incubated the immobilized MBP-Ulp1(3)(C580S) with SUMO2 chains. Unbound SUMO2 chains were removed by washing. The MBP-Ulp1(3)(C580S) SUMO2 chain complex was then eluted and added into a STUbL in vitro ubiquitinated reaction containing recombinant RNF4 (K A Fryrear and O Kerscher, unpublished reagents). Proteins in the STUbL-mediated ubiquitination assay were separated on SDS-PAGE gels and examined by Western blot analysis with an anti-SUMO2 antibody (Figure 7A). Consistent with previous observations, we were able to detect ubiquitinated SUMO2 chains after the STUbL reaction. This ubiquitination was dependent on RNF4 and SUMO2 chains. On the basis of these results, we propose that Ulp1(3)(C580S) may provide a useful, widely applicable tool for the study of sumoylated proteins and STUbL targets (Figure 7B).