Direct Role for Proliferating Cell Nuclear Antigen in Substrate Recognition by the E3 Ubiquitin Ligase CRL4Cdt2*

Background: CRL4Cdt2 requires that a substrate bind to proliferating cell nuclear antigen (PCNA) on DNA prior to ligase recruitment, but the precise role of PCNA is unclear. Results: A specific PCNA residue is required for destruction of CRL4Cdt2 substrates. Conclusion: CRL4Cdt2 recognizes a composite surface composed of PCNA and substrate residues. Significance: This is the first ubiquitin ligase whose substrate recognition requires creation of a bipartite substrate surface. The E3 ubiquitin ligase Cullin-ring ligase 4-Cdt2 (CRL4Cdt2) is emerging as an important cell cycle regulator that targets numerous proteins for destruction in S phase and after DNA damage, including Cdt1, p21, and Set8. CRL4Cdt2 substrates contain a “PIP degron,” which consists of a canonical proliferating cell nuclear antigen (PCNA) interaction motif (PIP box) and an adjacent basic amino acid. Substrates use their PIP box to form a binary complex with PCNA on chromatin and the basic residue to recruit CRL4Cdt2 for substrate ubiquitylation. Using Xenopus egg extracts, we identify an acidic residue in PCNA that is essential to support destruction of all CRL4Cdt2 substrates. This PCNA residue, which adjoins the basic amino acid of the bound PIP degron, is dispensable for substrate binding to PCNA but essential for CRL4Cdt2 recruitment to chromatin. Our data show that the interaction of CRL4Cdt2 with substrates requires molecular determinants not only in the substrate degron but also on PCNA. The results illustrate a potentially general mechanism by which E3 ligases can couple ubiquitylation to the formation of protein-protein interactions.

The E3 ubiquitin ligase Cullin-ring ligase 4-Cdt2 (CRL4 Cdt2 ) is emerging as an important cell cycle regulator that targets numerous proteins for destruction in S phase and after DNA damage, including Cdt1, p21, and Set8. CRL4 Cdt2 substrates contain a "PIP degron," which consists of a canonical proliferating cell nuclear antigen (PCNA) interaction motif (PIP box) and an adjacent basic amino acid. Substrates use their PIP box to form a binary complex with PCNA on chromatin and the basic residue to recruit CRL4 Cdt2 for substrate ubiquitylation. Using Xenopus egg extracts, we identify an acidic residue in PCNA that is essential to support destruction of all CRL4 Cdt2 substrates. This PCNA residue, which adjoins the basic amino acid of the bound PIP degron, is dispensable for substrate binding to PCNA but essential for CRL4 Cdt2 recruitment to chromatin. Our data show that the interaction of CRL4 Cdt2 with substrates requires molecular determinants not only in the substrate degron but also on PCNA. The results illustrate a potentially general mechanism by which E3 ligases can couple ubiquitylation to the formation of protein-protein interactions.
Eukaryotic cells contain hundreds of E3 ubiquitin ligases that each ubiquitylate one or more target proteins, modulating their activity or marking them for destruction by the 26 S proteasome (1). Several of the best-studied E3 ligases regulate cell cycle progression. For example, the Skp2-containing Cullin ring ligase 1 (CRL1 Skp2 also known as SCF Skp2 ) 2 is composed of a Cul1 scaffold, a Skp1 adaptor, the Skp2 substrate receptor, and Rbx1, which interacts with a ubiquitin-conjugating enzyme. Skp2 binds directly to substrates via a "phosphodegron," a short peptide motif on the substrate whose phosphorylation by cyclin-dependent kinases promotes its interaction with the leucine-rich repeat motif of Skp2 (2,3). In late G 1 phase, when CDK activity rises, CRL1 Skp2 targets the CDK inhibitor p27 for destruction, promoting S phase entry. Another cell cycle-regulated ubiquitin ligase is the anaphasepromoting complex (APC Cdc20 ). This multisubunit enzyme targets a number of factors for destruction, including Cyclin B and securin (4). In this case, the ligase itself is phosphorylated by mitotic CDKs, leading to substrate-ligase interactions.
Recently, an unusual ubiquitin ligase called CRL4 Cdt2 has been characterized, which promotes the ubiquitylation of several proteins in S phase and after DNA damage (5,6). CRL4 Cdt2 is composed of a Cul4 scaffold, a Ddb1 adaptor, and Cdt2, the putative substrate receptor. In vertebrates, CRL4 Cdt2 targets the licensing factor Cdt1 (6 -9), the CDK inhibitor p21 (Xic1 in frogs) (10 -13), and the histone methyltransferase Set8 (14 -18) for proteolysis in S phase. In all three cases, destruction appears to contribute to the block to re-replication. Set8 destruction also promotes transcription and prevents premature chromatin compaction (14,15,18). CRL4 Cdt2 targets the transcription factor E2F in flies (to shut off G 1 transcription in S phase and to regulate endocycles) (19,20), the translesion DNA polymerase in worms (perhaps to restrict access of this mutagenic polymerase to undamaged DNA) (21), and the ribonucleotide reductase inhibitor Spd1 in fission yeast (to activate nucleotide synthesis in S phase and after DNA damage) (22). Other substrates of CRL4 Cdt2 are likely to emerge.
The activity of CRL4 Cdt2 is coupled to DNA replication and damage via PCNA, a homotrimeric, ring-shaped molecule that encircles DNA. Given its topological embrace of DNA, PCNA tethers to DNA any proteins with which it interacts, including DNA polymerases, DNA ligases, chromatin-remodeling factors, and numerous other proteins involved in DNA replication and repair (23). Most proteins that bind PCNA do so via an eight-amino acid motif called a PIP box (see Fig. 1A, green amino acids). The aromatic and hydrophobic residues in the PIP box interact with a hydrophobic pocket underlying the interdomain connector loop (IDCL) (23)(24)(25)(26). CRL4 Cdt2 substrates contain a PIP box, through which they bind to chromatin-bound PCNA (PCNA chromatin ) at sites of DNA damage or at the replication fork. Most CRL4 Cdt2 substrates also contain a TD motif at positions 5 and 6 of the PIP box (Fig. 1A, blue amino acids), which confers especially high affinity binding to PCNA (27,28). However, a PIP box and TD motif are not sufficient for CRL4 Cdt2 activity. All substrates also contain a basic residue four amino acids downstream of the PIP box (the "Bϩ4" residue). When this residue is mutated to alanine in Xenopus Cdt1, the resulting protein binds normally to PCNA, but CRL4 Cdt2 is not recruited to the Cdt1⅐PCNA complex (27). These data explain why most PIP box proteins, such as DNA polymerases, which lack the Bϩ4 residue, are not destroyed. In summary, the above data indicate that, during CRL4 Cdt2 -mediated proteolysis, a substrate docks onto PCNA chromatin , leading to recruitment of CRL4 Cdt2 , followed by ubiquitin transfer (see Fig. 1B).
Notably, an alternative model has recently been proposed in which CRL4 Cdt2 binds PCNA independently of a substrate, and in which the PCNA and Cdt2 binding regions within substrates can be separated (11). Thus, there is significant disagreement over the role of PCNA in promoting substrate recognition by CRL4 Cdt2 .
In this report, we investigate the mechanism by which CRL4 Cdt2 interacts with its substrates to trigger proteolytic degradation. First, we provide evidence against the recent proposal that CRL4 Cdt2 and its substrates dock onto PCNA independently of one another (11), thereby affirming that CRL4 Cdt2 is initially recruited to a PIP degron⅐PCNA complex. We then addressed whether the sole function of PCNA is to position the substrate's PIP degron for binding to CRL4 Cdt2 (indirect role) or if a specific surface of PCNA is required, together with the PIP degron, to recruit CRL4 Cdt2 (direct role). In support of the latter model, we identify an acidic residue (Asp-122) on the surface of PCNA that is essential for CRL4 Cdt2 activity. Importantly, this residue is not necessary for PIP box binding to PCNA but is essential for CRL4 Cdt2 recruitment to the PCNA⅐PIP degron complex on chromatin. Asp-122 is also essential for CRL4 Cdt2 activity in fission yeast. Our findings support the idea that CRL4 Cdt2 recognizes the PCNA⅐PIP degron interface by making direct contacts with residues in both polypeptides. This mechanism suggests new possibilities for the regulation of proteolysis.

EXPERIMENTAL PROCEDURES
Xic1 in Vitro Transcription and Translation-pCS2ϩ/Xic1 (11) was in vitro translated according to the manufacturer's protocol using TNT SP6 Quick Coupled Transcription Translation kit from Promega (Madison, WI). In each reaction 125 ng of DNA was used per 10 l of TNT Quick Master Mix (Promega), and reactions were scaled according to the amount of Xic1 needed.
Depletion of Xenopus PCNA-The polyclonal PCNA antibody used for PCNA depletion was generated as described previously (34). To deplete PCNA from HSS, three rounds of depletion were performed using 3 l of PCNA antibody per 1 l of rProtein A-Sepharose FastFlow resin (Amersham Biosciences). To deplete PCNA from LSS, two rounds of depletion were performed. The antibody and resin were pre-bound, and 0.2 volume of resin was used per microliter of egg extract.
MMS DNA Preparation and Bead Spin-down Assay-Methylated DNA was generated as described previously (35). A biotinylated 1-kb PCR product was generated, treated with methyl methane sulfonate (MMS), and coupled to M-280 streptavidin Dynabeads (Invitrogen) as previously described (27). To spin down MMS-DNA beads, we used a modification of our standard chromatin spin-down protocol (33), in which the egg lysis buffer (10 mM HEPES (pH 7.7), 250 mM sucrose, 50 mM KCl, 2.5 mM MgCl 2 ) wash step was supplemented with 0.6% Triton X-100.
Fission Yeast General Methods-Strains used in this study are listed in supplemental Table S1. Standard genetic methods and flow cytometry were used as described previously (36,37).
PCNA Mutagenesis in Fission Yeast-To construct the pcn1-D122A mutant strain 2640, two partially complementary fragments for the mutation (NsiI-fragment and BamHI-fragment) were amplified using primers containing the mutation and an NsiI or BamHI restriction site; subsequently, the fragments were annealed and amplified. After DNA purification, the fragment was cloned into NsiI-and BamHI-digested pSMH and integrated at the pcn1 ϩ locus after linearizing with XhoI (in the strain 2069). Hygromycin-resistant transformants were selected, and the correct integration of the plasmid was checked with the primers 619 and 979. The plasmid and all the strains constructed were checked by sequencing (primer 978). Primers used are shown in supplemental Table S2.
Cell Cycle Synchronization and UV Irradiation Experiments-Cells were arrested in G 1 phase in Edinburgh minimal medium (65) lacking NH 4 Cl for 16 h at 25°C (38), then released into the cell cycle in rich yeast extract and supplements medium (65) at 32°C. nda3-311 strains were grown at 32°C in rich medium and arrested in M phase by incubation for 4 h at 20°C. For UV exposure, cells were resuspended in water and irradiated with 100 J/m 2 of 254 nm UV light in a 6-mm deep stirred suspension at 20°C.
Protein Analysis of Fission Yeast Samples-Protein extracts were made by TCA extraction and analyzed by Western blotting as described previously (39). Tandam affinity purification (TAP)-tagged Cdt1 was detected with peroxidase⅐anti-peroxidase-soluble complex (P1291, Sigma), and ␣-tubulin was used as loading control and detected with antibody T5168 (Sigma).

RESULTS
To study the mechanism of CRL4 Cdt2 -mediated proteolysis, we used two approaches that involve different types of Xenopus egg extracts (summarized in supplemental Fig. S1). First, DNA that had been treated with MMS to induce methylation damage was added to an HSS of egg cytoplasm. Nucleotide excision repair of the methylation damage involves a PCNA-dependent gap-filling step that promotes destruction of endogenous Cdt1 as well as other substrates by CRL4 Cdt2 (7,15). Second, sperm chromatin was added to an LSS of egg cytoplasm. Upon nuclear-envelope assembly around the sperm, chromosomal DNA replication initiates, leading to PCNA-dependent, CRL4 Cdt2mediated proteolysis (7,33). Most experiments were performed using DNA damage-induced destruction in HSS, but key conclusions were confirmed using the S phase pathway in LSS.
The Bϩ4 Residue Can Be Partially Compensated for in Xic1 by Residues Upstream of the PIP Box-The Bϩ4 residue is conserved in all CRL4 Cdt2 substrates (Fig. 1A), and it is known to be essential for Cdt1 destruction (27,28). We wanted to know whether Bϩ4 is also important for destruction of other substrates. We therefore examined Xic1 and Set8. When added to Xenopus egg extracts, Xic1, a Xenopus CDK inhibitor, is destroyed in a manner that requires its PIP box, chromatinloaded PCNA, and Cdt2 (11, 40 -42). However, in contrast to Cdt1 Bϩ4A , which was completely stable (27), Xic1 Bϩ4A was still destroyed, albeit more slowly than Xic1 WT ( Fig. 2A, bottom, light blue trace). This residual destruction of Xic1 Bϩ4A was PCNA-dependent (data not shown). We next tested the histone H4 lysine 20 methyltransferase Set8, which is also destroyed during S phase and after DNA damage in human cells or when added to Xenopus egg extracts (14 -18). Like Xic1 Bϩ4A , Set8 Bϩ3/4A was destroyed slowly (Fig. 2B, lanes 7-9; note that in Set8 Bϩ3/4A , the Bϩ3 residue was also mutated to alanine in case of charge redundancy), and the residual destruction was still PCNA-dependent (see below). Similar to Cdt1 Bϩ4A (27), Set8 Bϩ3/4A bound normally to PCNA chromatin , and its slow destruction was due to reduced recruitment of CRL4 Cdt2 to chromatin, resulting in reduced ubiquitylation of Set8 Bϩ3/4A (Fig. 2C). In summary, as seen in Cdt1, the Bϩ4 residue in Xic1 and Set8 is important for efficient destruction due to a role in CRL4 Cdt2 recruitment, although it is less crucial for destruction of the latter substrates.
Given the different effects of the Bϩ4A mutation on different CRL4 Cdt2 substrates, we re-examined their sequences. Notably, the PIP box of Cdt1 is located at the extreme amino terminus, whereas the PIP boxes of p21, Xic1, and Set8 are internal or C-terminal to these proteins. Additionally, the latter substrates contain one or more basic amino acids immediately upstream of the PIP box (Fig. 1A, pink) (27). We speculated that these basic residues might contribute to destruction in the absence of the Bϩ4 residue. To test this hypothesis, we mutated the basic residues N-terminal of the PIP box to alanines in Xic1, yielding Xic1 RRKR/AAAA ( Fig. 2A, top). Mutation of these residues alone had little or no effect on Xic1 destruction ( Fig. 2A, yellow trace). However, when combined with the Bϩ4 mutant, these mutations rendered the protein completely stable, similar to Xic1 ⌬PIP ( Fig. 2A, green trace). Together, these data show that, in a CRL4 Cdt2 substrate where the PIP degron is not located at the extreme N terminus, additional basic residues located just upstream of the PIP box can contribute to CRL4 Cdt2 -mediated destruction, but they are only essential in the absence of the Bϩ4 mutant. The upstream basic residues likely promote substrate destruction in a manner that is distinct from the downstream basic residues, including Bϩ4.
The PCNA and CRL4 Cdt2 Binding Motifs of Substrates Cannot Be Separated-Recently, Yew and colleagues proposed a new model for substrate recognition by CRL4 Cdt2 in which the substrate and CRL4 Cdt2 initially bind independently to PCNA before coming together in a complex (Fig. 3A). This model was based on two considerations. First, they showed that the C terminus of Cdt2 can bind to PCNA independently of substrate (11). However, our previous data established that Cdt2 does not bind to PCNA on chromatin in the absence of substrate (7,27), demonstrating that a Cdt2-PCNA interaction is insufficient to support chromatin recruitment of the ligase. Second, Yew and colleagues proposed that, within the primary amino acid sequence of Xic1, PIP box and Cdt2 binding motifs can be widely separated and still promote destruction (11). Their conclusion, which would further indicate that CRL4 Cdt2 does not bind PCNA in the context of substrate, was based on the following experiment. The p21 PIP degron, in which a proposed Cdt2-binding region had been deleted, was fused onto the N terminus of Xic1, in which the endogenous PIP box had been compromised through mutation of Ile-174 to alanine, yielding p21 PIP -Xic1 I174A (Fig. 3, B and C; originally named NPIP2-Xic1 I174A in Ref. 11). This construct was destroyed normally in egg extracts (11) even though it was thought to contain well separated PCNA and Cdt2 interaction motifs, as illustrated in Fig. 3B. We repeated this experiment and obtained the same result (Fig. 3D, yellow trace). However, this experiment has two caveats. First, the Bϩ4 residue of the added p21 PIP degron was not mutated, likely still allowing binding to Cdt2. Second, only one residue (Ile-174) in the endogenous Xic1 PIP box was mutated. Thus, the mutated Xic1 PIP box and the added p21 PIP box might bind cooperatively via interactions with two PCNA subunits (Fig. 3E). In this case, each PIP box might be able to recruit some CRL4 Cdt2 and thus promote destruction via the mechanism we previously proposed (Fig. 3E) (27).
To test whether the destruction of p21 PIP -Xic1 I174A involved residual binding of the Xic1 PIP box to PCNA, we mutated all three Xic1 PIP box residues, creating p21 PIP -Xic1 I174A,Y177A,F178A ( Fig. 3C; abbreviated as p21 PIP -Xic1 ⌬PIP ). Unlike p21 PIP -Xic1 I174A , p21 PIP -Xic1 ⌬PIP was destroyed much less efficiently than Xic1 WT (Fig. 3D, green line). From these data we conclude that ϳ40% of p21 PIP -Xic1 I174A destruction can be attributed to residual binding of the Xic1 PIP box to PCNA. When the Bϩ4 residue of the added p21 PIP box was also mutated to alanine, the resulting protein, p21 PIP/Bϩ4A -Xic1 ⌬PIP (Fig. 3C), was destroyed at background levels (Fig. 3D, purple line), indicating that the p21 PIP box retained Cdt2 binding capacity. Thus, the original p21 PIP -Xic1 I174A construct did not contain adequately separated PCNA and Cdt2-binding functions. When these domains are effectively separated, the resulting fusion protein is not destroyed. Therefore, there is no evidence that these domains can function separately.
Identification of a PCNA Residue That Is Essential for CRL4 Cdt2 -mediated Destruction-We next wanted to further characterize what role PCNA plays in the recognition of the PIP degron by CRL4 Cdt2 . Specifically, we wished to distinguish between direct and indirect roles for PCNA. In the "indirect" model, CRL4 Cdt2 only contacts residues in the substrate's PIP degron, and the sole function of PCNA is to position the degron for recognition by CRL4 Cdt2 . In the "direct" model, CRL4 Cdt2 recognizes a composite surface created by the two proteins. The second model predicts that substrate recognition by CRL4 Cdt2 requires residues in PCNA that do not influence PIP degron binding. To look for such residues, we examined the co-crystal structure of PCNA with the PIP degron of p21 Reactions were stopped at the indicated time points, and the amount of Xic1 remaining was quantified by autoradiography. Results from three independent experiments were averaged and graphed. Bars represent the standard error of the mean. In B: Top, sequence alignment of Set8 PIP degron with upstream residues and the various mutants examined. Bottom, HSS was supplemented with 5 ng/l MMS plasmid and 50 nM human Set8 WT , Set8 ⌬PIP , or Set8 Bϩ3/4A . At the indicated times, samples were blotted for endogenous Cdt1 or Set8. C, HSS was supplemented with immobilized 1-kb MMS DNA and 2 mg/ml methyl ubiquitin. Buffer and 50 nM Set8 WT , Set8 ⌬PIP , or Set8 Bϩ3/4A was also added, and after 10 min, chromatin was recovered from the extract and washed, and the indicated proteins were visualized by Western blotting.
( Fig. 4A) (24). Interestingly, the structure shows that two conserved acidic residues, Asp-122 and Glu-124 on the IDCL of PCNA, cradle the Bϩ4 residue of p21 (Fig. 4, A and B,  purple). Because the Bϩ4 residue is not required for binding to PCNA (27), we postulated that Asp-122 and/or Glu-124 contact CRL4 Cdt2 and help recruit it to the PCNA⅐PIP degron complex.
We first tested whether Asp-122 and Glu-124, as well as another nearby acidic residue, Asp-120, are required for destruction of CRL4 Cdt2 substrates. Asp-120, Asp-122, and Glu-124, or Asp-122 and Glu-124, were mutated to alanines, and the resulting proteins (PCNA D120A , PCNA D122A , PCNA E124A , and PCNA DE/AA ) were purified (supplemental Fig.  S2A). PCNA-depleted HSS was supplemented with the differ- 11. B, model for destruction of p21 PIP -Xic1 I174A using separable Cdt2 and PCNA binding motifs. Yew and colleagues fused residues 135-164 of p21 containing the PIP box to the N terminus of Xic1 to create a p21 PIP box-Xic1 fusion protein (11). In this construct one canonical PIP box residue in Xic1 (Ile-174) was mutated to an alanine to inhibit PCNA binding, and six residues (156 -161) just past the Bϩ4 residue of p21 (residues ϩ5 to ϩ10) were deleted to inhibit the Cdt2-binding region proposed by Yew and colleges. C, scheme of Xic1 and p21 PIP -Xic1 mutations used in our experiments. D, graph showing the percentage of 35 S-labeled Xic1 mutants remaining when incubated with HSS and 5 ng/l MMS-damaged plasmid at the indicated times. Results from three independent experiments were averaged and graphed. Bars represent the standard error of the mean. E, our model for CRL4 Cdt2 -mediated destruction of p21 PIP -Xic1 I174A binding to two PCNA monomers. ent recombinant PCNA proteins, damaged DNA, as well as Set8 and Cdt1 1-243 , an N-terminal fragment of Cdt1 that is destroyed by the same mechanism as Cdt1 WT (27). Although PCNA D120A behaved essentially like PCNA WT , PCNA E124A displayed a noticeable, but minor defect, especially in Set8 destruction (Fig. 4C, lanes 17-20). Strikingly, PCNA D122A was completely inactive for destruction of Cdt1, Cdt1 1-243 , and Set8 (Fig. 4C, lanes 9 -12). PCNA D122A was also unable to support efficient Xic1 destruction (supplemental Fig. S2C). As expected, PCNA DE/AA was unable to support CRL4 Cdt2 activity in HSS extracts ( Fig. 4C and supplemental Fig. S2B). In addition, PCNA DE/AA did not support replication-dependent Cdt1 destruction in the context of sperm chromatin replication carried out in LSS extracts (supplemental Fig. S3A).
We examined the role of other PCNA residues near the IDCL. Alanine substitution of His-44, which contacts the 5 and 6 positions of bound PIP boxes, caused a slight defect in destruction of Cdt1, due to deficient substrate binding to PCNA, whereas S42A and S230A, which reside on either side of the IDCL, had no effect. 3 It was important to rule out the possibility that the D122A and E124A mutations in PCNA inhibited CRL4 Cdt2 activity due to indirect effects on DNA replication. As shown in supplemental Fig. S2D, PCNA DE/AA supported normal levels of M13 DNA replication in HSS, a model for leading strand synthesis in Xenopus egg extracts, implying that this mutant is also loaded normally on chromatin. PCNA DE/AA was also fully competent for sperm chromatin replication in LSS (supplemental Fig.  S3B). Together, our results show that the D122A and E124A residues of PCNA play a specific role in potentiating CRL4 Cdt2 function.

Asp-122 Is Required for CRL4 Cdt2 Docking, but Not Substrate
Binding to PCNA-To determine why the Asp-122 residue of PCNA is required for CRL4 Cdt2 function, we first examined whether it mediates binding of substrates to PCNA. We took advantage of the fact that Cdt1 1-243 only binds to chromatin via PCNA in a PIP box-dependent manner (27). MMS-treated DNA coupled to magnetic beads was added to PCNA-depleted HSS supplemented with Cdt1 1-243 and either buffer, PCNA WT , PCNA DE/AA , or PCNA DE/KR , a charge-reversal mutant that also failed to support CRL4 Cdt2 function and behaved identically to PCNA DE/AA (supplemental Figs. S2B and S3B). We also included PCNA LI/AA , a PCNA mutant previously studied in Saccharomyces cerevisiae (25), which carries mutations in residues Leu-126 and Ile-128 in the IDCL of PCNA and does not bind well to PIP box proteins (24 -26, 43). Each of the recombinant PCNAs bound to chromatin at similar levels (Fig. 5A, PCNA), suggesting that none of the mutations affected replication factor C-mediated loading of PCNA onto DNA. Indeed, it was previously reported that even PCNA LI/AA is efficiently loaded onto DNA by replication factor C (25). Importantly, equal amounts of Cdt1 1-243 bound to DNA in the presence of PCNA WT , PCNA DE/AA , and PCNA DE/KR (Fig. 5A, lanes 4, 6,  and 7). In contrast, PCNA IL/AA did not support Cdt1 loading above background levels, as expected, based on previous studies using DNA polymerases (Fig. 5A, compare lanes 2 and 5) (25,26). In conclusion, the defect in CRL4 Cdt2 function observed with mutations in Asp-122 and Glu-124 was not due to a failure to recruit the substrate to chromatin-bound PCNA.
We next examined CRL4 Cdt2 recruitment to chromatin in the presence of the PCNA mutants. MMS-treated 1-kb DNA conjugated to magnetic beads was added to PCNA-depleted extract supplemented with recombinant PCNAs and  Recombinant Set8 WT , Cdt1 1-243 , and 5 ng/l MMS plasmid were also added to the extract. Reactions were stopped at the indicated times and blotted for Cdt1 (both endogenous and Cdt1 1-243 ) or Set8. Samples were run on two separate gels that were processed under the same conditions. Additionally, PCNA WT samples were run on both gels, exposures were matched, and similar exposures were used.
were inactive for Cdt2 recruitment and Cdt1 1-243 ubiquitylation (Fig. 5B). PCNA D120A was unaffected, whereas PCNA D122A behaved like PCNA DE/AA with respect to Cdt1 binding, CRL4 Cdt2 recruitment, and ubiquitylation (Fig. 5C, lanes 3 and  4). Consistent with its intermediate effects on Cdt1 destruction, PCNA E124A supported intermediate levels of Cdt2 recruitment (Fig. 5C, lane 5). Similar results were observed for Set8 (Fig.  5D). PCNA DE/AA and PCNA DE/KR also did not support Cdt2 recruitment or Cdt1 ubiquitylation in the context of sperm chromatin replication (Fig. 5E). Together, these data show that Asp-122 is essential and Glu-124 is important for substrate-dependent CRL4 Cdt2 recruitment to chromatin, whereas these residues have no role in mediating the binding of PIP box proteins to PCNA.
PCNA Residues Asp-122 and Glu-124 Are Required for CRL4 Cdt2 -mediated Destruction Independently of the Bϩ4 Residue-Based on the above results, we postulated that CRL4 Cdt2 binds directly to residues in the PIP degron (Bϩ4), as well as in PCNA (Asp-122 and Glu-124). However, one alternative explanation was that Asp-122 and Glu-124 do not directly contact CRL4 Cdt2 (or an unknown ligase cofactor) but rather function to position Bϩ4 for interaction with the ligase. To address this possibility, we examined whether mutations in Asp-122 and Glu-124 still affect destruction of a CRL4 Cdt2 sub- . Samples were run on two separate gels that were processed under the same conditions. Additionally, one sample set was run on both gels, exposures were matched, and similar exposures were used. strate lacking the Bϩ4 residue. This experiment was not possible in the context of Cdt1, where mutation of Bϩ4 alone completely eliminates destruction (27). Therefore, we examined the destruction of Set8 Bϩ3/4A in PCNA-depleted extracts supplemented with PCNA WT or PCNA DE/KR . As shown in Fig. 5F, the intermediate level of Set8 Bϩ3/4A destruction was completely abolished in the presence of PCNA DE/KR (compare lanes 10 -12 with 16 -18). Thus, PCNA residues 122 and 124 are critical for CRL4 Cdt2 -mediated destruction even in the absence of the Bϩ4 residue, indicating that residues on PCNA are directly involved in ligase recruitment.
PCNA Residue Asp-122 Is Required for CRL4 Cdt2 -mediated Destruction in Fission Yeast-In Schizosaccharomyces pombe, CRL4 Cdt2 targets Cdt1 for destruction in a PCNA-dependent manner during S phase and after DNA damage (39,44). To determine whether the function of aspartic acid 122 in PCNA is conserved and whether it affects endogenous CRL4 Cdt2 substrates in an in vivo model, we made the D122A mutation in the S. pombe PCNA gene, pcn1. Upon UV irradiation, cells express-ing wild-type PCNA destroyed Cdt1, while cells expressing PCNA D122A did not (Fig. 6A, compare lanes 5 and 10). In addition, analysis of Cdt1 levels in cells released from a G 1 arrest showed that the S phase destruction of Cdt1 is inhibited in the mutant (Fig. 6B, compare lanes 4 and 5 to lanes 11 and 12). Stabilization of TAP-tagged Cdt1 by PCNA D122A appeared to cause a minor delay in either S phase entry or progression through S phase (Fig. 6C, compare 3-h time points), but this could not explain the effect on destruction. Similar results for Cdt1 stabilization were seen in cells released from a mitotic block (supplemental Fig. S4). Therefore, as in Xenopus egg extracts, Asp-122 in S. pombe PCNA is required to support CRL4 Cdt2 activity during replication and after DNA damage.

DISCUSSION
In this study we explore the molecular mechanism by which CRL4 Cdt2 recognizes its substrates in the context of PCNA. We provide evidence that recognition by CRL4 Cdt2 requires amino acids not only in the substrate's PIP degron, but also in PCNA itself. The simplest interpretation is that CRL4 Cdt2 makes direct contacts with both polypeptides during substrate recognition. This mechanism appears to be conserved from humans to fission yeast.
The activity of most ubiquitin ligases is regulated either by the assembly of ligase subunits, or post-translational modification of the ligase or substrate. To our knowledge, CRL4 Cdt2 is the first example of a ubiquitin ligase whose activity is regulated by the creation of a bipartite surface when a substrate interacts with another polypeptide. Specifically, the event that triggers destruction of CRL4 Cdt2 substrates is the creation of a composite surface composed of PCNA and the substrate. Conversely, CRL1 Tir -auxin (SCF Tir1 -auxin) (45) and CRL1 Skp2 -Cks1 (SCF Skp2 -Cks1) (46) require ligase cofactor interactions to recognize their substrates. In the case of CRL1 Tir1 -auxin, binding of auxin to the substrate receptor Tir1 creates a surface on the ligase that mediates binding to and ubiquitylation of its substrates (45). Although ubiquitylation of the CRL1 Skp2 substrate p27 requires prior formation of a complex between the substrate receptor Skp2 and its cofactor Cks1, this interaction is not the initiating event to trigger p27 destruction (46,47). Rather, CRL1 Skp2 -Cks1-mediated proteolysis is promoted by a phosphorylation event on threonine 187 of the substrate p27, which mediates the p27-CRL1 Skp2 -Cks1 interaction (3,46,49).
Mechanism of Substrate Recognition by CRL4 Cdt2 -Our data suggest the following model for the assembly of the ternary PCNA DNA ⅐PIP degron⅐CRL4 Cdt2 complex. First, substrates bind PCNA DNA via their PIP degron, an event that does not require CRL4 Cdt2 (27) and therefore almost certainly precedes binding of the ligase. Next, CRL4 Cdt2 docks onto the PCNA⅐PIP degron complex, likely using the WD40-repeat-containing ␤-propeller of Cdt2. A model for the structure of CRL4 Ddb2 , which ubiquitylates xeroderma pigmentosum, complementation group C (XPC), in the context of nucleotide excision repair (50 -52), provides a framework for the possible structure of CRL4 Cdt2 . Thus, like the ␤-propeller protein Ddb2, Cdt2 likely contacts the adaptor protein Ddb1 via a helix located near the bottom surface of its propeller (50,53). Accordingly, the top surface of the propeller of Cdt2 would interact with the PCNA⅐PIP degron complex. We have shown that the Bϩ4 residue within the PIP degron and at least one residue on PCNA that cradles Bϩ4 are essential for stably recruiting CRL4 Cdt2 . It is presently unclear whether residues in the PIP degron other than Bϩ4 or residues in PCNA other than Asp-122 and Glu-124 make contact with CRL4 Cdt2 . Our data suggest that Cdt2 contains a surface with an appropriate arrangement of positive and negative charges that binds the PCNA⅐PIP degron complex (Fig. 7). Importantly, because substrate recognition by CRL4 Cdt2 has not been reconstituted with purified components, we cannot rule out the possibility that the binding of CRL4 Cdt2 to the PCNA⅐PIP degron complex is indirect. However, given the direct binding of several other ␤-propeller WD40 proteins to substrate (50, 54 -58), this appears unlikely.
An important question is whether PCNA functions primarily as a match-maker that promotes interactions between CRL4 Cdt2 and its substrates (as illustrated in Fig. 7), or whether it also regulates ubiquitin transfer allosterically, by inducing conformational changes in the substrate or ligase. Our identification of residues on PCNA that are specifically required to recruit CRL4 Cdt2 to the PCNA⅐substrate complex provides strong evidence for the former view, although it leaves open the possibility that PCNA could play additional roles in ubiquitin transfer.
Recently, Yew and colleagues proposed a two-step recognition model in which Xic1 and CRL4 Cdt2 bind independently to two different subunits of PCNA and only later come together for Xic1 ubiquitylation (Fig. 3A) (11). This conclusion was based in part on an experiment in which a hybrid Xic1 substrate was constructed that was thought to have well separated PCNA and Cdt2 recognition motifs (Fig. 3B). However, we show here that the motifs were in fact not well separated (Fig. 3, C and D). Together with the observation that short peptides derived from Cdt1 (59) and p21 (27) are sufficient to support CRL4 Cdt2 recruitment and activity, the data strongly favor a model in which the PCNA-and CRL4 Cdt2 -binding activities in the substrate are closely linked.
Contribution of Positively Charged Residues Upstream of the PIP Box-In Cdt1, mutation of the Bϩ4 residue completely abolished destruction (27,28), whereas in Xic1 and Set8, destruction was slowed but not eliminated (this report). In all cases, the mutation dramatically reduced the recruitment of CRL4 Cdt2 to the PCNA⅐PIP degron complex. Importantly, the FIGURE 7. Putative model of CRL4 Cdt2 recruitment to the PCNA⅐substrate complex. Substrate-dependent recruitment of CRL4 Cdt2 to PCNA DNA requires residues in the PIP degron (Bϩ4), as well as PCNA Asp-122. PCNA Glu-124 also contributes to CRL4 Cdt2 recruitment but is not essential. We speculate charged residues in Cdt2 make direct contacts with Bϩ4, Asp-122, and Glu-124. residual destruction of Set8 Bϩ3/4A (Fig. 5F) and Xic1 Bϩ4A (data not shown) still required Asp-122, arguing that this residue does not merely function to position the Bϩ4 residue for recognition by CRL4 Cdt2 . Notably, the PIP degron of Cdt1 is located at the extreme N terminus of the protein, whereas in Set8, Xic1, and p21, this is not the case. The latter class of substrates also contains a cluster of basic residues immediately upstream of the PIP box. Mutation of these residues alone didn't interfere with destruction of Xic1. However, when they were mutated in combination with Bϩ4, Xic1 was no longer degraded. The upstream basic residues are likely to contribute to the free energy of ternary complex formation, but this contribution is only readily detectable when the PIP degron is otherwise compromised through mutation of Bϩ4. The discovery that Set8 and Xic1 lacking the Bϩ4 residue can be inefficiently destroyed at reduced rates suggests that CRL4 Cdt2 could also modify proteins that lack the Bϩ4. However, the absence of this residue would have to be compensated for by other features to enhance CRL4 Cdt2 recruitment, as illustrated by the basic residues upstream of the Xic1 PIP box. In addition, we know that the ϩ5 downstream basic residue near the Bϩ4 enhances PCNA binding, which contributes to efficient CRL4 Cdt2 -mediated destruction (13,24,27,28). In fact, if the dimensions of the Cdt2 ␤-propeller are similar to those of Ddb2, separate areas of the "top" surface of Cdt2 could simultaneously contact the Bϩ4 and upstream basic residues in the p21 PIP degron⅐PCNA complex.
Recently two distinct E2 ubiquitin-conjugating enzymes were identified that mediate ubiquitylation of substrates with CRL4 Cdt2 (60). UBCH8 cooperates with CRL4 Cdt2 to target p21 and Set8, whereas UBE2G1 and UBE2G2 collaborate with CRL4 Cdt2 to ubiquitylate Cdt1. During polyubiquitin chain formation, E2s can recognize the surface of ubiquitin or the substrate near the lysine to be ubiquitylated (48,(61)(62)(63)(64). Therefore, given the different PIP box locations and additional contributions of upstream residues of CRL4 Cdt2 substrates, it is tempting to speculate that these different recognition determinants could contribute to the use of distinct E2s for Cdt1 versus p21 and Set8.
New Perspectives on the Degron-Our work raises interesting questions about the nature of degrons and the regulation of proteolysis. Because substrate recognition by CRL4 Cdt2 requires residues in the PIP degron and in PCNA, PCNA could be considered part of the degron. However, to conform to the field's implicit understanding of the term, we propose that "degron" be reserved for recognition elements within the protein that gets destroyed. Eukaryotic cells contain hundreds of distinct ubiquitin ligases, most of which are completely uncharacterized, and we speculate that some of these might be regulated similarly to CRL4 Cdt2 . Thus, we propose that, among the thousands of transient protein-protein interactions that form during cellular growth and metabolism, some create a composite recognition surface that attracts a specific E3 ubiquitin ligase. Degron recognition could also involve the binding of protein substrates to other macromolecules such as nucleic acids, sugars, or lipids. The utility of this strategy is that it couples proteolysis to the final outcome of signaling events, which is usually the assembly of macromolecular complexes.