Composition of SCFPop1p-Pop2p
Based on the composition of known SCF complexes, we identified in the S. pombe genome database psh1 (pombe skp1 homologue) and pip1 (pop interacting protein 1), two genes encoding proteins with strong similarity to human SKP1 and HRT1/RBX1/ROC1, respectively (data not shown). Consistent with these proteins being components of the putative SCFPop1p-Pop2p complex, they all co-purified with Pop1p, Pop2p, and Pcu1p when overexpressed pairwise (data not shown).
Co-immunoprecipitation experiments using affinity purified rabbit antisera confirmed these interactions at the level of the endogenous proteins. While each of the five antisera co-precipitated at least one of the other subunits, Pip1p, Pop1p, and Pop2p antisera co-precipitated all five proteins from wild-type cell lysate (Fig. 1A). Size fractionation of total cell lysates prior to immunoprecipitation revealed co-elution of Pip1p with Pop1p, Pop2p, Pcu1p, and Psh1p in a high molecular weight complex of approximately 500 kDa, which we refer to as SCFPop1p-Pop2p (Fig. 1B). The composition of the core complex Pip1p/Pcu1p/Psh1p did not undergo major variations during the cell cycle (Fig. 1C). We have carefully reexamined potential cell cycle variations of Pcu1p neddylation apparent in the IP/immunoblotting experiment in Fig. 1C. These variations were not seen when samples were denatured in SDS immediately following extract preparation (data not shown), suggesting that they arise from varying degrees of deneddylation presumably occuring during the immunoprecipitation step. In addition, in a separate experiment, Pop1p-Pop2p heterooligomerization was largely constant during the cell cycle (Fig. 1C). These findings indicated that cell cycle-dependent substrate degradation is unlikely to be controlled at the level of SCFPop1p-Pop2p complex formation.
SCFPop1p-Pop2p mediates polyubiquitylation of Rum1p in vitro
Genetic experiments suggested that degradation of Rum1p depends on the Pop1p and Pop2p F-box proteins [18, 20], but also requires phosphorylation of Rum1p on serine 58 and threonine 62 by cyclin-dependent kinase (CDK) [21]. Thus, phosphorylated Rum1p may be a substrate for SCFPop1p-Pop2p-mediated polyubiquitylation. To test this, we first confirmed that Rum1p is an in vitro substrate for the Cdc2p kinase in association with the cyclin Cig1p [21] as judged by a mobility shift on SDS gels (Fig. 2A). Bacterially expressed Rum1p purified to apparent homogeneity was also efficiently phosphorylated by Cdc2p/Cig1p complexes (Fig. 2A).
To determine whether phosphorylated Rum1p interacts with Pop1p and Pop2p, protein lysate was prepared from cells co-overexpressing epitope-tagged combinations of Pop1p and Pop2p. Upon affinity purification on Ni-NTA resin, HA-Pop1p/His-Myc-Pop2p complexes were incubated with bacterially expressed, phosphorylated Rum1p. In this reaction, Pop1p-Pop2p complexes specifically bound phosphorylated Rum1p (Fig. 2B, lane 1). Consistent with the genetic data [20], His-Myc-Pop1p and His-Myc-Pop2p individually purified upon overexpression in pop1 pop2 double mutants exhibited no Rum1p binding above background (Fig. 2B, lanes 2 and 3).
Given our ability to prepare immunopurified SCFPop1p-Pop2p that bound phosphorylated Rum1p, we sought to reconstitute Rum1p polyubiquitylation in vitro. In addition, we required an ubiquitin activating enzyme (E1) and an ubiquitin conjugating enzyme (UBC). While human E1 is highly similar to its fission yeast counterpart (data not shown), inspection of the S. pombe genome revealed fourteen potential UBCs, none of which is an obvious homologue of human UBC3 or budding yeast Cdc34p, since all lack the characteristic C-terminal extension (data not shown). We therefore purified recombinant human E1 and UBC3 (CDC34) upon expression in bacteria (Fig. 2C).
In the presence of human E1, UBC3, ubiquitin, and ATP, SCFPop1p-Pop2p complexes immunopurified with Pip1p antibodies converted a small portion of phosphorylated Rum1p into high molecular weight species (Fig. 2D). This conversion was dependent on the addition of E1, ATP (data not shown), and wild-type UBC3 (Fig. 2D, lanes 2,4,6). The activity of SCFPop1p/Pop2p was augmented when Pip1p complexes where purified from csn5 mutants (Fig. 2D, lane 5). This mutant accumulates Pcu1p exclusively in a form carrying the stimulatory Nedd8p modification, due to a defect in COP9/signalosome-mediated cullin deneddylation [24, 25]. Replacing wild-type ubiquitin by a mutant lacking all lysine residues prevented the formation of high molecular weight products (Fig. 2D, lane 7), indicating that they represent polyubiquitylated Rum1p species generated in the reaction. Similar polyubiquitylated reaction products were detected upon incubation of phosphorylated Rum1p with Pcu1p immunocomplexes, further suggesting that the activity is mediated by SCFPop1p-Pop2p (data not shown). Moreover, Rum1p ubiquitylation was not obtained with Pip1p complexes prepared from cell lysate of pop2 deletion strains, proving the F-box protein dependency of this reaction (Fig. 2D, lane 3). In addition, the reaction was specific for human UBC3, as no ubiquitylation was obtained with fission yeast Ubc1p, Ubc7p, Ubc11p, or Ubc13p (Fig. 2, lanes 10,11). Taken together these results strongly suggest that SCFPop1p-Pop2p mediates the polyubiquitylation of CDK phosphorylated Rum1p in vitro.
Differential subcellular localization of SCFPop1p-Pop2p subunits
The co-purification of the five identified SCFPop1p-Pop2p subunits and their in vitro activity toward Rum1p suggested that they coexist in a common subcellular compartment. The nuclear localization of the only known substrates, Cdc18p [26] and Rum1p (D.A.W., unpublished), indicated that a substantial portion of SCFPop1p-Pop2p is enriched in the nucleus. To test this assumption, all five SCFPop1p-Pop2p subunits were expressed as fusion proteins with green fluorescent protein (GFP) at low levels from an inducible pRep81 plasmid. While Pip1p, Psh1p, Pcu1p, and Pop2p were present in both the cytoplasm and the nucleus, surprisingly, GFP-Pop1p was largely restricted to the nucleus (Fig. 3A). These localization patterns were consistently observed in each single cell of an asynchronous population, excluding major variations during the cell cycle.
To rule out the possibility that overexpression or N-terminal fusion to GFP affects their localization, Pop1p and Pop2p were modified with 13 C-terminal Myc epitope-tags at the endogenous genomic locus. Immunoblotting proved the expression of correctly sized proteins and, in addition, showed that endogenous Pop1p is approximately twofold more abundant in S. pombe cells than Pop2p (Fig. 3B). Indirect immunofluorescence staining with Myc antibodies confirmed that Pop1p is predominantly localized to cell nuclei, whereas Pop2p is expressed in both the cytoplasm and the nucleus (Fig. 3C).
To confirm these localization patterns, cells derived from the epitope-tagged strains were biochemically fractionated into cytoplasmic and nuclear components. The efficiency of enrichment of nuclear and cytoplasmic components was estimated by analyzing fractions with antibodies recognizing the nuclear marker PCNA and cytoplasmic tubulin (Fig. 3D). Although both fractions showed some contamination, Pop1p was detected mostly in nuclear fractions, while Pop2p was apparent in both nuclear and cytoplasmic fractions (Fig. 3D). Thus, all five SCFPop1p-Pop2p subunits appear to coexist in the nucleus, although all but Pop1p are also present in the cytoplasm.
Since Pop1p and Pop2p interact with each other [20], we asked whether their localization patterns depended on the presence of the respective interaction partner. For this, we created a pop2 deletion strain carrying Pop1p modified with 13 Myc epitope tags at the endogenous genomic locus (pop1-13myc Δpop2 strain). In addition, we created the reciprocal pop2-13myc Δpop1 strain containing epitope-tagged endogenous Pop2p in a pop1 deletion background. The distinct localization patterns of Pop1p-13Myc and Pop2p-13Myc were fully maintained in these strains (Fig. 3E). This observation was confirmed by overexpressing GFP-tagged Pop1p or Pop2p in pop1 pop2 double deletion mutants (data not shown). These data indicate that Pop1p and Pop2p assume their subcellular localization pattern independent of each other, indicating the possibility of distinct nuclear and cytoplasmic homooligomeric SCFPop1p and SCFPop2p complexes.
Differential F-box requirements of Pop1p and Pop2p
As shown above and in previous genetic work [18, 20], SCFPop1p-Pop2p-dependent Rum1p degradation requires two different proteins with highly conserved F-boxes (Fig. 4A). To better understand the apparent dual F-box requirement for SCFPop1p-Pop2p function, we generated mutants of Pop1p and Pop2p lacking F-boxes (Pop1p-ΔF, Pop2p-ΔF; Fig. 4B). In addition, we prepared a set of mutants, in which the F-boxes of Pop1p and Pop2p were swapped (Pop1p-2F, Pop2p-1F; Fig. 4B). The mutants were tested for their ability to suppress polyploidy and Rum1p accumulation in the respective pop mutant strains.
As described previously [20], wild-type Pop1p mildly overexpressed from a pRep81 plasmid fully complemented the polyploidization phenotype of pop1 mutants as determined by flow cytometric measurement of the cellular DNA content (Fig 4C). In addition, Rum1p accumulation in pop1 mutants was efficiently reversed by wild-type Pop1p (Fig. 4D). In contrast, Pop1p lacking its F-box (Pop1p-ΔF) or Pop1p, in which the F-box was replaced by the F-box of Pop2p (Pop1p-2F) were largely inactive in both assays (Fig. 4C,4D). Thus, as with many other F-box proteins, the F-box of Pop1p is essential for its in vivo functions.
In contrast, wild-type Pop2p, the corresponding F-box mutant, and Pop2p containing the Pop1p F-box were equally effective in preventing Rum1p accumulation (Fig. 4D). The same wild-type and mutant proteins also reversed the mild polyploidy phenotype of pop2 disruptants (Fig. 4C). Thus, in contrast to Pop1p, Pop2p does not seem, to require its F-box to mediate Rum1p degradation in vivo.
To further substantiate this conclusion, we examined Rum1p protein stability in wild-type and pop mutant strains expressing F-box-deleted Pop proteins from the weak pRep81 promoter. Since the sensitivity of our Rum1p sera was insufficient to detect the low levels present in wild-type cells (see Fig. 4D, lane 9), these experiments were conducted in a background where endogenous Rum1p was modified with 13 C-terminal c-Myc epitope tags. Rum1p half-life was increased from ~20 minutes in wild-type to greater than 100 minutes in pop1 or pop2 mutants (Fig. 5). While F-box deleted Pop2p expressed from plasmids reduced Rum1p half-life to ~20 minutes in pop2 mutants, F-box-deleted Pop1p was completely defective in rescuing the Rum1p proteolysis defect of pop1 mutants (Fig. 5). Instead, expression of Pop1p-ΔF in pop1 mutants led to even greater stabilization of Rum1p, potentially due to dominant negative interference with the residual activity of Pop2p and/or other F-box proteins.
F-box independent interaction of Pop1p and Pop2p
The failure of the Pop2p F-box to replace the F-box of Pop1p as well as the finding that it is not essential for Rum1p degradation could be explained, if it was not critically involved in SCFPop1p-Pop2p protein interactions. We therefore tested the possibility that Pop2p can be tethered to the SCF core complex independently of its F-box via an interaction with Pop1p. Consistent with this hypothesis, co-immunoprecipitation experiments of overexpressed proteins revealed that the Pop1p-Pop2p interaction occurs independently of the F-boxes of both Pop1p and Pop2p (Fig. 6A).
We had previously mapped the domain of Pop2p that interacts with Pop1p to an N-terminal fragment consisting of the first 241 amino acids and lacking the F-box ([20], Fig. 6B, lane 8). In co-immunoprecipitation experiments with overexpressed proteins, this fragment also bound to an N-terminal piece containing the first 402 residues of Pop1p (Fig. 6B, lane 12). Thus, the Pop1p-Pop2p interaction is mediated by their N-terminal domains. A further truncation mutant mapped the Pop2p binding domain to a region between residues 228 and 402 of Pop1p (Fig. 6B, lanes 9,10).
Individual SCFPop1p and SCFPop2p complexes bearing ubiquitin ligase activity
The apparent dispensibility of the F-box of Pop2p for Rum1p degradation and binding to Pop1p raised the question of why Pop2p does contain an F-box. Based on the subcellular localization data, we considered the possibility that the F-box of Pop2p may mediate the assembly of a cytoplasmically localized SCFPop2p complex, independent of Pop1p. To demonstrate this, we again used the strain in which endogenous Pop2p was modified by 13 Myc epitope tags in a pop1 deletion background. The same experiments were carried out with the reverse stain, which contained Pop1p-13Myc in a pop2 background. As a reference for SCFPop complex formation, we used strains carrying 13Myc epitope-tagged Pop1p or Pop2p integrated into the genome of wild-type cells.
Pip1p immunoprecipitates were prepared from lysates of these four strains and appropriate controls, and co-purification of SCF components was determined by immunoblotting. These experiments showed that both F-box proteins, in the absence of their respective heterooligomerization partner, could individually bind to Pip1p in complexes that also contained Psh1p and Pcu1p (Fig. 7A). These findings indicate the existence of distinct SCFPop1p and SCFPop2p complexes in vivo.
To further substantiate this conclusion, we used gel filtration to compare the elution profiles of Pop1p and Pop2p in the presence or absence of their respective dimerization partners. If recruitment of Pop2p into a high molecular weight SCF complex required heterooligomerization with Pop1p, its elution profile would be expected to shift to a smaller size in the absence of Pop1p. Consistent with the results presented in Fig. 1C, Pop2p, together with SCF core subunits, eluted in fractions corresponding to 400 – 600 kDa, irrespective of whether Pop1p was present or not (Fig. 7B). In the reverse experiment, the elution profile of Pop1p was found to be independent of the presence of Pop2p (Fig. 7B).
While these data support the existence of distinct SCFPop1p and SCFPop2p complexes, individual binding of Pop1p and Pop2p to SCF core components in the absence of their heterodimerizing F-box protein partners does not rule out the possibility that these complexes represent inactive intermediates formed during the normal assembly of functional SCFPop1p-Pop2p complexes. To exclude this possibility, we asked whether distinct SCFPop1p and SCFPop2p complexes bear ubiquitin ligase activity in vitro. To this end, we performed in vitro ubiquitylation assays. Since the substrates of putative SCFPop1p and SCFPop2p ubiquitin ligases are unknown, we adopted a substrate-independent assay originally described by Lyapina et al. [27]. For this experiment, we again used the strains harboring genomically integrated Myc epitope-tagged Pop1p or Pop2p in a background deficient in the respective heterooligomerization partner (Δpop2 pop1-13myc, Δpop1 pop2-13myc strains). Pop1p and Pop2p complexes were immunopurified with Myc antibodies and employed in in vitro ubiquitylation assays upon addition of E1, UBC3, ubiquitin, and ATP. High molecular weight products generated in the reaction were detected by immunoblotting with ubiquitin antibodies. As references, we used strains expressing Myc-tagged Pop1p or Pop2p in a wild-type background. The experiment demonstrated that Pop1p and Pop2p each associate with polyubiquitylation activity even in the absence of their respective heterooligomerizing F-box proteins (Fig. 7C). Thus Pop1p and Pop2p appear to assemble into distinct SCF complexes bearing ubiquitin ligase activity in vitro.