Proliferating Cell Nuclear Antigen-dependent Coordination of the Biological Functions of Human DNA Polymerase ι*

Y-family DNA polymerases are believed to facilitate the replicative bypass of damaged DNA in a process commonly referred to as translesion synthesis. With the exception of DNA polymerase η (polη), which is defective in humans with the Xeroderma pigmentosum variant (XP-V) phenotype, little is known about the cellular function(s) of the remaining human Y-family DNA polymerases. We report here that an interaction between human DNA polymerase ι (polι) and the proliferating cell nuclear antigen (PCNA) stimulates the processivity of polι in a template-dependent manner in vitro. Mutations in one of the putative PCNA-binding motifs (PIP box) of polι or the interdomain connector loop of PCNA diminish the binding between polι and PCNA and concomitantly reduce PCNA-dependent stimulation of polι activity. Furthermore, although retaining its capacity to interact with polη in vivo, the polι-PIP box mutant fails to accumulate in replication foci. Thus, PCNA, acting as both a scaffold and a modulator of the different activities involved in replication, appears to recruit and coordinate replicative and translesion DNA synthesis polymerases to ensure genome integrity.

Y-family DNA polymerases are believed to facilitate the replicative bypass of damaged DNA in a process commonly referred to as translesion synthesis. With the exception of DNA polymerase (pol), which is defective in humans with the Xeroderma pigmentosum variant (XP-V) phenotype, little is known about the cellular function(s) of the remaining human Y-family DNA polymerases. We report here that an interaction between human DNA polymerase (pol) and the proliferating cell nuclear antigen (PCNA) stimulates the processivity of pol in a template-dependent manner in vitro. Mutations in one of the putative PCNA-binding motifs (PIP box) of pol or the interdomain connector loop of PCNA diminish the binding between pol and PCNA and concomitantly reduce PCNA-dependent stimulation of pol activity. Furthermore, although retaining its capacity to interact with pol in vivo, the pol-PIP box mutant fails to accumulate in replication foci. Thus, PCNA, acting as both a scaffold and a modulator of the different activities involved in replication, appears to recruit and coordinate replicative and translesion DNA synthesis polymerases to ensure genome integrity.
Y-family DNA polymerases are widely distributed among the three kingdoms of life. Human cells contain at least four; Rev1, pol, 1 and two RAD30 paralogs, pol and pol (1,2). The best characterized is pol, which has been shown to be defective in humans with the sun-sensitive, cancer-prone Xeroderma pigmentosum variant (XP-V) syndrome (3,4). The biological functions of pol, pol, and Rev1 polymerases remain, however, largely unknown. Indeed, the cellular function of human pol is particularly enigmatic given its unique enzymatic properties in vitro. Detailed biochemical analysis of the purified enzyme indicates that it exhibits up to 100,000-fold differences in the frequency of nucleotide misincorporation depending upon the template base replicated. At template T, the wobble base G is incorporated 3-11 times more often than the correct Watson-Crick base, A. However, at template A, misincorporations occur with a frequency of ϳ10 Ϫ4 (5)(6)(7).
Human pol also exhibits a unique ability to incorporate bases opposite certain DNA lesions. Unlike the related pol, which bypasses cis-syn cyclobutane thymine dimers efficiently and relatively accurately (8,9), pol frequently misincorporates nucleotides opposite the 3ЈT of the dimer (10), and the efficiency of bypass is sequence context-dependent (11). In contrast, pol very efficiently inserts bases opposite the more structurally distorting 6 -4 pyrimidine-pyrimidone lesion (6,10,11), a benzo[a]pyrene adducted deoxyadenosine (12), and a noncoding abasic site (6,13). In the latter three cases, the enzyme is unable to facilitate lesion bypass but instead requires assistance from another translesion DNA synthesis enzyme, such as pol (6) or pol (12).
Given the unique and generally error-prone properties of pol in vitro, it makes teleological sense that these activities would be tightly regulated in vivo. One appealing mechanism would be to strictly control access of the polymerase to a growing primer terminus. For the replicative polymerases, such activity is coordinated by the replicative processivity factors (prokaryotic ␤-clamp and eukaryotic PCNA). The process is mediated by a clamp loader (which in eukaryotes is replication factor C (RFC)), which recognizes the DNA primer terminus and opens and assembles a PCNA ring around the nascent DNA (reviewed in Ref. 14). At least three human Y-family polymerases, including pol, are believed to physically interact with the homotrimeric PCNA clamp (15)(16)(17), and such interactions have been proposed to play a central role in enabling access of the translesion DNA synthesis polymerases to a blocked replication fork (18 -20).
In an attempt to understand how these interactions may specifically influence the enzymatic and biological properties of pol, we show that the processivity of pol is stimulated in the presence of PCNA in vitro, have identified regions in both pol and PCNA that are important for a physical and functional interaction between the two proteins, and demonstrate that a PCNA-pol interaction is required for the normal recruitment of pol into cellular replication factories in vivo. C-3Ј), from pET39hPCNA (21) digested with NdeI-BamHI and cloned into the NdeI-BamHI sites of vectors pGBKT7 and pGADT7 (Clontech) producing plasmids pGBKT7-PCNA and pGADT7-PCNA . To generate  I126A and L128A mutations in the IDCL of PCNA, the last 423 nucleotides of hPCNA cDNA were amplified with mutagenic primer PCNA3  (5Ј-GGA ATT CGA CGT CGA ACA AGC TGG AGC TCC AGA ACA GGA  GTA CAG CTG TG-3Ј) and PCNA2, digested with EcoRI-BamHI and  cloned into the EcoRI-BamHI sites of vector pGBKT7, generating plasmid pGBKT7-PCNA[364 -786]. Next, the first 363 nucleotides of hPCNA were amplified by PCR with PCNA1 and PCNA4 (5Ј-GGA ATT CGA CGT CTA AAT CCA TCA ACT TCA TTT C-3Ј), digested with NdeI-AatII and cloned into the NdeI-AatII-digested pGBKT7-PCNA[364 -786] vector to produce pGBKT7-PCNA-IDCL. pACT2-pol, pACT2-pol , pACT2-pol , pACT2-pol[492-715], pGBKT7pol, and pGBKT7-pol were constructed as previously described (22). To produce pACT2-pol PIP1 (Y426A,Y427A), pACT2-pol PIP2 (F546A,F547A), and pGBKT7-pol PIP1, the EcoRI-BamHI 674-nucleotide fragment from peYFP-pol PIP1 or pEYFP-pol PIP2 was subcloned into the similarly digested pACT2-pol, pGBKT7-pol plasmids, respectively. The pol PIP3 (F710A,H711A) mutant was constructed by PCR amplification of the POLI gene with primers: 5Ј-GTC GGG TCA TGT ATA CAA TAA TCA G-3Ј and 5Ј-GCG GAT CCT TAT TTA TGT CCA ATG GCG GCA TCT GAT CCT G-3Ј, and ligation of the AatII-BamHIdigested PCR products to the similarly digested pACT2-pol. pGBKT7pol[484 -713] was produced by PCR amplification with primers: 5Ј-GCA TAT GGA ATT CCA AAA AGC TGC AGA AAG-3Ј and 5Ј-GCG GTC GAC TAA TGT GTT AAT GGC TTA AAA AAT G-3Ј using pG-BKT7-pol as a template. The PCR product was then digested with EcoRI-SalI and cloned into the EcoRI-SalI sites of pGBKT7.
Proteins-Glutathione S-transferase-tagged human pol was purified from baculovirus-infected insect cells as previously described (5). Histidine-tagged human PCNA and IDCL-PCNA proteins were purified on Ni 2ϩ -charged nickel-nitrilotriacetic acid His-Bind Resin (Novagen) as recommended by the manufacturer. The His-PCNA containing eluates were then dialyzed against H Buffer (20 mM potassium phosphate, pH 7.5, 10 mM 2-mercaptoethanol, 0.01% Nonidet P-40, 10% glycerol) and applied to a hydroxylapatite (Bio-Rad) column previously equilibrated with H Buffer. The column was subsequently washed with H Buffer, and the His-PCNA fusion proteins were eluted with 10 column volumes of a 10 -300 mM linear gradient of potassium phosphate in H Buffer. His-PCNA containing fractions were aliquoted and stored at 80°C.
In Vitro Transcription/Translation of Proteins-In vitro transcription/translation of full-length pol, pol PIP1, and PCNA was performed using a TNT-coupled reticulocyte lysate system (Promega) according to the manufacturer's instructions. The expression vectors encoding the full-length pol (pGBKT7-pol), pol PIP1 (pGBKT7-pol PIP1), and PCNA (pGBKT7-PCNA) were added separately to the reaction mixtures and then incubated for 90 min at 30°C in the presence of [ 35 S]methionine. The reaction products were analyzed by SDS-PAGE and used in replication reactions and GST and Ni 2ϩ pull-down assays.
Replication Reactions-The DNA template used for the study was a circular single-stranded M13mp18 DNA (New England Biolabs) primed with a 17-mer 5Ј-32 P-labeled oligonucleotide. 5Ј-CGA GAA CAA GCA AGC CG-3Ј (Lofstrand Laboratories) that is complementary to base pairs 3440 -3456 of M13mp18.
Replication Assays in the Presence of a DNA Trap-Reaction mixtures containing pol (5 nM) with or without PCNA, RFC, and RPA were preincubated with the circular M13 DNA template-primer in the standard reaction buffer in the absence of dNTPs at 37°C for 5 min to preform the DNA-polymerase complex. Enzymatic reactions were started by the addition of dNTPs (100 M) or dNTPs and an excess of competitor DNA (1 mg/ml of nonradiolabeled herring sperm DNA) to capture pol molecules that dissociated from the radiolabeled template and incubated at 37°C for 10 min. As a control, a reaction mixture containing pol, PCNA, RFC, and RPA was preincubated with an excess of DNA competitor and the DNA template prior to addition of the dNTPs.
Two-hybrid Assay-The interaction between human pol and PCNA was analyzed in vivo using the Saccharomyces cerevisiae two-hybrid Matchmaker III system (Clontech). Strain AH109 was co-transformed with the GAL4 binding domain (pGBKT7-X) and GAL4 activation domain (pACT2-X, pGADT7-X) fusion constructs as indicated in the figures. Co-transformants were selected on DOBA-Trp-Leu (Bio 101) plates. Colonies were subsequently replica-plated on DOBA-Trp-Leu-His-Ade (Bio 101) plates. An interaction between the two fusion proteins resulted in the ability of the transformed yeast cells to grow in the selective medium.
Ni 2ϩ Pull-down Assay-Equal amounts of nickel-nitrilotriacetic acid beads alone or coupled to His-tagged PCNA or His-IDCL-PCNA proteins (20 g) were mixed with 10 l of a standard TNT-lysate reaction containing 35 S-labeled pol, pol PIP1, or PCNA in 500 l of binding buffer (50 mM Tris⅐HCl, pH7.5, 200 mM NaCl, 50 mM imidazole, 1 mM EDTA, 0.1% Nonidet P-40, 1 mM dithiothreitol) in the presence of a "Complete EDTA-free" mixture of protease inhibitors (Roche Applied Science). After incubation for 16 h at 4°C, the beads were washed four times with binding buffer, and the bound proteins were separated on a 4 -20% SDS-PAGE. The dried gels were scanned using a FujiFilm FLA-3000 Phosphor Imager.
GST Pull-down Assay-Equal amounts of GST or GST-pol[484 -715] (20 g) coupled to glutathione-agarose beads were mixed with 35 S-labeled pol or pol-PIP1 (10 l of a standard TNT lysate system reaction) in 500 l of binding buffer (50 mM Tris⅐HCl, pH7.5, 200 mM NaCl, 1 mM EDTA, 0.1% Nonidet P-40, and 1 mM dithiothreitol), containing Complete protease inhibitors. After incubation for 15 min at 37°C, the beads were washed four times with binding buffer, and the bound proteins were separated on a 4 -20% SDS-polyacrylamide gel.
Plasmids-peYFP-pol and peCFP-pol were produced in a similar way to peCFP-pol and peYFP-pol described in Ref. 22. The pol PIP1 mutant was constructed by using the QuikChange mutagenesis kit (Stratagene) with the primers 5Ј-GAA AGG GCT TAT CGA TGC TGC TTT AAT GCC ATC ATT ATC-3Ј and 5Ј-GAT AAT GAT GGC ATT AAA GCA GCA TCG ATA AGC CCT TTC-3Ј on the plasmid template peYFPpol (22). For the convenience of subsequent identification, these primers also contain a silent change that results in a novel ClaI restriction enzyme site (underlined). The pol PIP1 mutant was recloned into peYFP-C1 to produce peYFP-pol PIP1. To generate peYFP-pol PIP2, the pol gene was amplified by PCR from peYFP-pol with the following primers: 5Ј-CAT GCC TCT AGA GGA GTA TTA TCT GCC GCT TCT AAA AAA C-3Ј and 5Ј-TAT GGC TGA TTA TGA TCA GTT ATC TAG ATC CGG TGG ATC C-3Ј and subcloned into peYFP-pol.
Transfection and Immunofluorescence-peCFP-pol and peYFP-pol were co-transfected into transformed fibroblasts (at a ratio of 1:1) using FuGENE TM 6, according to the manufacturer's protocol (Roche Applied Science). Twenty hours after transfection, the cells were irradiated at 7 J/m2 and incubated for a further 8 -12 h. Fixation of cells was carried out as previously described (22). Fluorescence images of cell nuclei were acquired on a Zeiss Axiophot2 microscope (Carl Zeiss) equipped with an Orca ER CCD camera (Hamamatsu) using Simple PCI software. The images were captured by alternating the excitation at 440 ϩ 12.5 and 500 ϩ 10 nm and detection of CFP and YFP emission at 465 ϩ 12 and 530 ϩ 15 nm, respectively (CFP/YFP filter set; Omega). A minimum of 200 nuclei were analyzed for each cell line and treatment.
Model Building Procedure-The pol peptide suspected to bind PCNA was aligned with the p21 peptide co-crystallized with PCNA (Protein Data Bank code 1AXC). To make the model complex of pol and PCNA, the graphic program ONO (25) was used to substitute residues of p21 in 1AXC with those of pol based on the sequence alignment. Interestingly, residues 423-427 of pol are predicted to form a short helix corresponding to the 3 10 helix formed by residues 147-151 of p21. Most of the rotamers of mutated side chains were kept the same as in the original crystal structure. Side chains of Lys 420 and Lys 421 , which replaced Gln 144 and Thr 145 , were adjusted choosing one of the favored rotamers and manually fitted to avoid clashes.

RESULTS
Stimulation of pol-dependent Polymerase Activity in the Presence of PCNA, RFC, and RPA-Previous studies using steady-state kinetic analysis revealed that the catalytic activity of pol is stimulated up to 150-fold in in vitro reactions containing RFC, PCNA, and RPA, yet processivity was unaltered, and pol continued to behave in a highly distributive manner (15).
Studies on pol suggest, however, that the properties of the enzyme are especially sensitive to both the template base replicated and the sequence context in which it is located. At templates G, C, and especially T, the enzyme misincorporates nucleotides at high frequency and has difficulty in extending from the mispaired primer terminus (26). In contrast, at template A, pol inserts the correct nucleotide, T, 25-100-fold more efficiently than it makes any other correct base pair (5). We therefore monitored DNA polymerase activity of pol by using a 7-kb circular single-stranded M13mp18 template in which the first five template bases replicated are dAs, the sixth base is dT, and the following four bases are dAs. Based upon previous biochemical analysis of pol, we expected that synthesis on the template dAs to be robust, but upon encountering template dT, the enzyme would misincorporate dGMP and, because of its reduced ability to extend the G:T mispair, would pause or might even terminate synthesis (Fig. 1). Indeed, pol alone or in the presence of various combinations of PCNA, RFC, or RPA readily extended the primer by six bases, with a strong pause opposite template dT (Fig. 1A, lanes 3-6). Quite strikingly, however, in the presence of PCNA, RFC, and RPA, replication products were up to 14 nucleotides longer (Fig. 1A, lane 7). The stimulation in activity results from more efficient elongation of the replication products, because the amount of the initial radiolabeled primer utilized remained practically identical in the presence or absence of PCNA, RFC, and RPA (63% versus 71% respectively; Fig. 1A, lane 7 versus lane 3). The effect of PCNA, RFC, and RPA on the enzymatic activity of pol is obvious when analyzing incorporation at bases past the presumed pol-generated G:T mispair at n ϭ 6. Indeed, the percentage of n Ն 7 elongation products increases ϳ13-fold from ϳ2% in the absence of PCNA, RFC, and RPA, to ϳ26% in their presence. Our data therefore suggest either that pol rebinds preferentially to those substrates encircled by PCNA or that it remains on the same DNA molecule and performs more processive DNA synthesis.
To examine this further, we studied the effect of PCNA on pol processivity under conditions where the polymerase might only be expected to encounter a primer-template once. To achieve this goal, the reaction mixtures containing pol alone (Fig. 1B, lanes 1 and 4) or pol, PCNA, RFC, and RPA (Fig. 1B,  lanes 2 and 5) were preincubated at 37°C in the absence of dNTPs to preform a polymerase-DNA complex. The reactions were initiated by the addition of dNTPs (Fig. 1B, lanes 1 and 2) or dNTPs and an excess of competitor DNA (nonradiolabeled herring sperm DNA) (Fig. 1B, lanes 4 and 5), which serves as a trap to capture pol molecules as they dissociate from the template. Fig. 1B (lane 1 versus lane 2) shows the stimulation of pol activity by PCNA, RFC, and RPA under our standard assay conditions. Preincubation in the presence of an excess of DNA competitor resulted in the trapping of all pol molecules by the cold substrate, and no DNA polymerase activity was detected (Fig. 1B, lane 3). When the competitor DNA was added after preincubation, pol alone incorporated up to six nucleotides (Fig. 1B, lane 4). In contrast, in the presence of PCNA, RFC, and RPA products up to 20 nucleotides in length were observed. Our data therefore clearly demonstrate that under the appropriate assay conditions, PCNA, RFC, and RPA substantially increase the processivity of pol.
Identification of the PCNA-binding Site in pol-In the pol paralog pol, a conserved PCNA-binding motif (PIP box) (27,28) is located at the very C terminus of the enzyme and is required for the interaction with PCNA (16). However, the corresponding sequence in pol is absent, and the precise location of the pol PIP box remains to be elucidated. Analysis of the primary amino acid sequence of pol reveals three potential PIP boxes (identified as PIP1-3), although none of them has a perfect match to the PIP box consensus sequence, QXX(I/L/ M)XXFF (27,28) (Fig. 2, A and B).
To identify which of the potential PIP boxes is involved in the pol-PCNA interaction, we first used a yeast two-hybrid analysis. In these experiments, a series of pol truncations (Fig. 2B) fused to the GAL4 activation domain (GAL4-AD) and human PCNA in fusion with the GAL4 binding domain (GAL4-BD) were co-expressed in the yeast reporter strain, AH109. An interaction between the two fusion domains resulted in the ability of the transformed yeast cells to grow in selective medium in the absence of tryptophan, leucine, histidine, and adenine. As shown in Fig. 2C, only cells expressing full-length pol or the truncated form comprising the first 492 amino acids were able to grow on selective medium. In the latter case,  1 and 2), or dNTPs and an excess of competitor DNA (1 mg/ml), to capture pol molecules after they dissociated from the radiolabeled primer-template (lanes 4 and 5). The reaction mixtures were then incubated at 37°C for 10 min. In a control reaction pol, PCNA, RFC, and RPA were preincubated with an excess of DNA competitor and the radiolabeled primer-template prior to the addition of the dNTPs (lane 3). The locations of pol-dependent pause sites corresponding to misincorporation at template T are indicated by arrows.
however, the interaction with PCNA was clearly not as robust as that observed between full-length pol and PCNA. The twohybrid results, although not definitive, tentatively mapped the PCNA-binding site between amino acids 279 and 492 (Fig. 2C). In this region, we identified a potential PIP box, PIP1 (KKGLIDYY), between pol amino acids 420 and 427. However, because the interaction between the truncated pol-  and PCNA was suboptimal, we felt compelled to further investigate the possibility that the two other potential PIP boxes of pol may nevertheless facilitate an interaction between pol and PCNA, especially because PIP2 (SRGVLSFF) had previously been proposed as a potential PCNA-binding site (15), and PIP3 (KRTGSDFH), which although far from the consensus PIP sequence, aligns with the pol PIP box.
We therefore generated base substitutions in full-length POLI that resulted in amino acid substitutions in each poten-tial PIP box and assayed the mutants via the two-hybrid assay. As shown in Fig. 2D, both the PIP2 and PIP3 mutants retained the capacity to interact with PCNA, whereas amino acid substitutions of Y426A and Y427A in PIP1 completely eliminated the interaction with PCNA (Fig. 2D). The diminished ability of the PIP1 mutant to interact with PCNA is specific, because like wild-type pol (22), the mutant (along with the PIP2 and PIP3 mutants) retains an ability to interact with both full-length pol and a C-terminal fragment of pol.
To confirm that the PIP1 mutant disrupts pol-PCNA interactions, we used purified His-tagged PCNA protein in a pulldown assay (Fig. 3). pol and the pol PIP1 mutant under the control of the T7 promoter on the two-hybrid vector pGBKT7 were used as a template for in vitro transcription and translation in the presence of 35 S-labeled methionine. As shown in Fig.  3A (lanes 5 and 6), SDS gel electrophoresis of the labeled proteins showed bands of the expected size for human pol. In vitro translated 35 S-labeled pol or pol PIP1 were incubated with either Ni 2ϩ -charged beads alone or with beads coupled with His-PCNA. After extensive washing, the bound proteins were resolved by SDS-PAGE. Consistent with the reported interaction between the two proteins, pol was bound to the His-tagged PCNA but not to the Ni 2ϩ -charged beads alone (Fig.  3A, cf. lanes 1 and 2). In contrast, the amount of the pol PIP1 mutant retained on either the Ni 2ϩ -charged beads or His-PCNA was dramatically reduced and similar to the nonspecific binding of wild-type pol to the Ni 2ϩ -charged beads (Fig. 3A, cf.  lanes 1, 3, and 4). Our in vitro findings are therefore in good agreement with the two-hybrid assay identifying PIP1, but not PIP2 or PIP3, as a bona fide pol-PCNA-binding site. As a control, we carried out a parallel pull-down experiment using GST protein and a GST-pol[484 -713] construct. Both, pol and the pol PIP1 mutant protein retained an ability to interact with pol and were bound to the beads coupled with GSTpol[484 -713] but not to GST alone (Fig. 3B and Ref. 22). pol PIP1 Is Catalytically Active but Is Not Stimulated by PCNA, RFC, and RPA-To determine the effect of the PIP1 mutation on the catalytic activity of pol, we compared the polymerase activity of the wild-type and PIP1 mutant pol proteins synthesized in vitro, in our standard replication assay with circular ssM13mp18 as substrate (Fig. 4). As a control, the pGBKT7 cloning vector was used as template for protein expression reaction, and equivalent amounts of the three in vitro translation reaction mixtures were used in the replication assays. Although no labeled protein products were observed in the control pGBKT7 reaction (data not shown), the preparation exhibited weak polymerase activity (Fig. 4, lanes 1-3). However, this unidentified polymerase activity forms a very even ladder of products without any pausing and is therefore clearly distinguishable from the signature of pol, with its characteristic pausing at template Ts. Indeed, as observed in the earlier experiments with highly purified GST-pol, the pol synthesized in vitro extended the primer by five to six bases in the absence of PCNA and exhibited increased processivity in the presence of PCNA, RFC and RPA (Fig. 4, lanes 4 -6). The in vitro synthesized pol PIP1 mutant exhibited activity comparable with the wild-type enzyme in the absence of PCNA, RFC, and RPA. In the presence of these accessory factors, there was a modest stimulation of polymerase activity (Fig. 4, cf. lanes  7-9), but this appears to be largely independent of RFC (and presumably PCNA) because the level of stimulation was similar in the presence and absence of RFC (Fig. 4, cf. lanes 8 and   FIG. 4. Effect of PCNA 1-3), pGBKT7-pol (lanes 4 -6), or pGBKT7-pol-PIP1 (lanes 7-9) were incubated for 20 min at 37°C with the circular M13 primer-template substrate (5 nM) and dNTPs (100 M) in the absence or presence of PCNA (50 nM), RFC (2 nM), and RPA (475 nM) as indicated.  9). Taken together, our results indicate that the pol PIP1 mutant has a greatly reduced ability to participate in RFC-PCNA-dependent replication reactions (Fig. 4, cf. lanes 5 versus  8).
PCNA Targets pol to the Replication Machinery in Human Cells-The precise mechanisms by which Y-family polymerases are recruited to the replication machinery and how the cell chooses a specific polymerase to bypass any particular lesion remain to be determined. In an earlier study, we demonstrated that wild-type pol localizes in replication factories during S phase and that these replication foci accumulate after UV irradiation. Furthermore, pol foci formation was coordinated with the appearance of pol foci (22). To investigate the role that PCNA might play in targeting pol into replication factories, we analyzed the nuclear localization pattern of the pol PIP1 mutant (Y426A,Y427A) fused to enhanced yellow fluorescent protein (eYFP) and compared it with the pattern of foci observed for similar constructs expressing wild-type pol or the pol PIP2 mutant (F546A, F547A), both of which interact with PCNA (Fig. 2). The three eYFP-pol constructs were individually co-transfected with eCFP-pol into normal MRC5 fibroblasts, cells were irradiated with 7 J/m 2 , and 12 h post-irradiation pol and pol were visualized by autofluorescence of eCFP (Fig. 5, left column) and eYFP, respectively (Fig. 5, middle  column). In agreement with our earlier studies, the number of cells containing wild-type pol and pol foci increases significantly from 10 -15% before irradiation to Ͼ60% of the cells after UV, suggesting that both proteins accumulate at replication forks stalled at sites of UV damage. A similar pattern was also observed with the pol PIP2 mutant, which also co-localizes with pol (Fig. 5, middle row). In dramatic contrast, only ϳ3% of the cells expressing the pol PIP1 mutant formed foci before (not shown) or after UV irradiation, even though pol foci formation occurred normally (Fig. 5, bottom row). These observations strongly suggest that constitutive localization of pol in the cellular replication machinery and its accumulation at sites of UV damage are largely dependent on a fully functional interaction with PCNA.
The Physical and Functional Interaction between pol and PCNA Occurs at the Interdomain Connector Loop of PCNA-Human PCNA exists as a homotrimer of three identical subunits, each consisting of two structurally similar domains (29,30). The two domains within each monomer are connected by a central loop, which has been referred to as the IDCL. Previous studies have shown that the IDCL of PCNA plays a key role in promoting the interaction with DNA-polymerase ␦, so as to stimulate its processivity (31)(32)(33). The IDCL also binds the cell cycle regulatory protein p21 (30,34). In vitro, p21 binds PCNA through the IDCL region and inhibits its interaction with DNA-polymerase ␦ (21,33,35,36) by competition for the site normally occupied by pol␦. To study the role of the well characterized IDCL in the interaction and stimulation of pol, we generated a mutant PCNA with amino acid substitutions I126A and L128A in the IDCL. The IDCL-PCNA mutant was cloned in-frame to the GAL4-BD, and the resulting two-hybrid construct was used to co-transform AH109 yeast cells. No interaction between pol and the IDCL-PCNA mutant (and control vectors without an insert) was observed in a two-hybrid assay (Fig. 6A). In contrast, the IDCL-PCNA mutant retained its ability to interact with wild-type PCNA, and such observations are consistent with previous reports showing that similar mutations in S. cerevisiae PCNA do not affect the ability to form homotrimeric structures (31).
The reduced ability of the IDCL-PCNA mutant to interact with pol was confirmed by incubating in vitro translated 35 Slabeled pol with purified His-PCNA or His-IDCL mutant PCNA (Fig. 6B, upper panel). The pull-down assay using Ni 2ϩcharged affinity beads showed that the binding of pol to the IDCL mutant was greatly reduced in comparison with the wild-type PCNA protein. In our control experiment (Fig. 6B, bottom panel), in vitro translated PCNA co-purified with both mutant and wild-type His-PCNA, supporting our previous twohybrid results. No pol or PCNA was pulled-down with Ni 2ϩcharged beads alone.
Next, we assayed pol-dependent replication on a primed circular template (M13mp18) to assess the effect of the IDCL-PCNA mutant on the processivity of pol (Fig. 6C). As reported above, wild-type PCNA in the presence of RFC and RPA greatly increased the processivity of pol (cf. lane 4 versus lane 5). In contrast, the ability of the IDCL mutant to stimulate the processivity of pol was reduced considerably (cf. lane 4 versus lanes 6 -8). Taken together, it seems reasonable to hypothesize that the IDCL region of PCNA is most likely the primary docking location for the PIP1 box of pol. However, it should also be noted that recent studies suggest that in addition to this docking site, the "little finger" domain of Y-family polymerases are likely to make considerable protein-protein interactions with the replicative sliding clamps (20). Such interactions therefore most likely explain why our ICDL mutants exhibited a greatly reduced ability to interact with pol, but the PCNApol interactions were not entirely eliminated.
Modeling of the pol PIP1 Residues onto PCNA-Having identified the contacting residues between pol and PCNA necessary for binding and stimulation of the processivity of pol, we were interested in gaining structural insights into these protein-protein interactions. Although the homology of the PIP box of pol to the PIP consensus is low outside the (I/L/M)XX(F/ Y)(F/Y) five-residue region (Fig. 2), secondary structure prediction of pol by PsiPred (bioinf.cs.ucl.ac.uk/psipred/) indicated that the residues immediately flanking Tyr 426 and Tyr 427 of pol fold into a short helix (residues 423-427) preceded and followed by extended regions just like the residues of p21 that intimately interact with PCNA (Fig. 7A). Moreover, the most prominently conserved aromatic side chains, Tyr 426 and Tyr 427 , are readily modeled onto the crystal structure of p21 peptide and PCNA complex replacing the Phe 150 and Tyr 151 of p21 (30) ( Fig. 7B). The additional hydroxyl group of Tyr 426 in pol is easily accommodated by slight adjustment of torsion angles between C␤ and C␥. Like Tyr 151 of p21, Tyr 427 of pol fits snugly into a hydrophobic pocket on the surface of PCNA encompassed by Leu 126 and Ile 128 in the IDCL (Fig. 7). The less conserved residues, such as the replacement of Gln 144 of p21 by Lys 420 of pol, is also readily achieved because the extended Lys side chain is flexible and can replace solvent molecules surrounding Gln 144 .
This model of the pol-PCNA complex provides a clear explanation for the reduced interaction between the two proteins carrying the mutations in pol and IDCL. Substitution of the two aromatic residues in pol with alanines obviously removes the most extended hydrophobic interactions between the two proteins (Fig. 7B). Vice versa, replacement of the bulky hydrophobic Leu 126 and Ile 128 with alanines enlarge the hydrophobic pocket that snugly accommodates Tyr 427 of pol. We therefore predict that pol and PCNA interactions are similar to those occurring between p21 and PCNA, particularly surrounding the two consecutive aromatic residues. DISCUSSION PCNA plays a pivotal role in coordinating the ability of the cell to duplicate its genome accurately and efficiently. This critical function occurs largely through protein-protein interactions between PCNA and key proteins, such as the main replicases of the cell, pols ␦ and ⑀, or with various repair proteins such as Fen1, Msh3, and the XPG protein (28). In the present study, we have investigated how PCNA also coordinates the biochemical and biological activities of a Y-family polymerase, human pol.
By using a template that is optimal for pol synthesis in vitro, we demonstrated that PCNA in the presence of RFC and RPA significantly increased the processivity of the enzyme. Importantly, these factors also help pol traverse a kinetic pause at template T, presumably caused by the "signature" misinsertion by pol of dGMP opposite template T. As a consequence it is possible that an interaction with PCNA in vivo may lead to an increase in pol-dependent mutagenesis. The stimulation in the processivity of pol appears to occur through direct proteinprotein interactions between PCNA and pol. By using a combination of yeast two-hybrid assays, site-directed mutagenesis, and pull-down assays, we were able to identify a noncanonical PIP box (KKGLIDYY) (Fig. 2A). In contrast to the PIP box in pol that is located at its very C terminus, the pol PIP box is located some 290 amino acids from its C terminus. Interestingly, however, this would place the PIP box immediately downstream of the little finger domain (37) of pol and as such would place it in the same approximate structural location as the analogous ␤-clamp-binding site in the prokaryotic and archaeal Y-family polymerases (20, 38 -40).
The cognate-binding site of PCNA was identified as the IDCL, and despite the fact that the pol PIP box differs from the consensus at its N terminus, the middle and C-terminal residues were modeled and shown to fit snugly into a hydrophobic pocket in the IDCL. In a previous study, we reported that pol and pol co-localize constitutively and in a coordinate manner with PCNA in nuclear foci (22). The region required for foci formation was mapped to a region encompassing residues 490 -715 of pol that is also required for the physical interaction between pols and , suggesting that such an interaction is a prerequisite for foci formation. Indeed, in XPV cells lacking pol, the number of UV-induced pol foci dropped 3-6-fold (from ϳ60% in a wild-type cell to 10 -20% in an XP-V cell), indicating that pol foci formation is at least partially dependent upon pol (22). In our present study, we show that foci formation is essentially abolished in a full-length pol PIP mutant (Ͻ3% cells with foci formation), whereas pol foci formation remains unaltered. Our data suggest, therefore, that a functional interaction with PCNA is the key determinant for the accumulation of pol into replication factories but that an interaction between pols and probably also contributes to stabilizing these structures.
Because of the homotrimeric structure of PCNA, up to three DNA polymerases could potentially bind to the clamp simultaneously, in a manner similar to the 'tool belt' model proposed by Pages and Fuchs (19). Support for such an idea comes from the recent structural analysis of Bunting et al. (20), who co-crystallized the little finger domain of Escherichia coli polIV with the ␤-clamp and demonstrated that in principle, the clamp could accommodate multiple polymerases without any steric hindrance of the polymerase engaged at the primer terminus. Because different regions of pol are involved in the interactions with PCNA (Fig. 4) and pol (22), one can conceive of a possible scenario in which pols ␦, , and could potentially form FIG. 7. Model of the pol and PCNA complex. A, the pol peptide complexed with one of the three equivalent PCNA subunits is shown in a ribbon diagram (44). Like residues 147-150 of p21, pol is predicted to have a 3 10 helix between residues 423 and 427. The backbone of the pol peptide is shown in purple and that of the PCNA subunit is shown in pink. Tyr 426 and Tyr 427 of pol and Leu 126 and Ile 128 of PCNA, which anchor the interactions between the two proteins, are highlighted in ball-and-stick presentations. B, close-up of the hydrophobic pocket in PCNA that holds Tyr 427 of pol. PCNA is presented in molecular surface format using graphic program PyMol (www.pymol.org). The subsurface backbone trace and side chains of PCNA are shown with Leu 126 and Ile 128 highlighted in orange. The pol peptide trace is shown as a worm in purple, and the side chains of Tyr 426 and Tyr 427 are highlighted as green sticks. a constitutive translesion-synthesis complex, whose stability would depend ultimately on pairwise interactions between their components. Further regulation may occur through posttranslational modifications of PCNA by mono-and polyubiquination and sumoylation (reviewed in Ref. 41) or acetylation (42), which change the respective affinities of the various polymerases for PCNA and help facilitate switching between replicative and translesion DNA synthesis polymerases. Indeed, in support for such an idea, human pol has recently been shown to bind much more tightly to monoubiquitinated PCNA than to unmodified PCNA (43).