Assembly of functional replication factor C expressed using recombinant baculoviruses.

Replication factor C (RF-C), a complex of five subunits, is an essential eukaryotic protein involved in both DNA replication and DNA repair. To generate an easily accessible source of human RF-C for biochemical and genetic studies, we cloned the cDNAs of all five subunits into baculoviruses so that each subunit could be expressed both as a non-fused polypeptide and as an N-terminal His-tagged fusion (-his). Co-infection of insect cells with five baculoviruses encoding individual RF-C subunits (p140, p40, p38, p37, and p36) yielded a protein preparation active in two assays characteristic for authentic RF-C: stimulation of DNA polymerase delta DNA synthesis on singly primed single-stranded DNA template and formation of a complex of proliferating cell nuclear antigen with circular double-stranded DNA. Functional recombinant RF-C containing p40-his, p37-his, or p36-his was isolated using affinity resin. Active RF-C was reconstituted only by co-expression of all five subunits. A model for assembly of RF-C from individual subunits was derived from co-purification experiments performed with various combinations of His-tagged and non-fused subunits expressed by co-infection of insect cells with recombinant baculoviruses. p37 and p36 are proposed to form the first intermediate, which, upon addition of either p40 or p38, generates stable tertiary complexes: p40.p37.p36 and p38.p37.p36. The remaining fourth small subunit binds to the tertiary complex to form a quaternary complex p40.p38. p37.p36. Large subunit p140 binds last to form the five-subunit protein.

ssDNA in the presence of PCNA, RP-A (or Escherichia coli SSB), and ATP. Although discovered as a protein essential for DNA replication, RF-C itself had little or no effect on DNA synthesis by polymerases ␣, ␦, and ⑀. In contrast, the combination of RF-C with PCNA, RP-A, and ATP strongly inhibited pol ␣ and stimulated pol ␦ and pol ⑀. Isolation of stable pol ␦ and pol ⑀ complexes, formed with DNA in the presence of RF-C, PCNA, and ATP, indicated that pol ␦ and pol ⑀ directly interact with these auxiliary proteins (5)(6)(7)(8). Biochemical studies on RF-C and PCNA suggested that the physiological function of RF-C is to catalyze the loading of PCNA onto a DNA template using the energy of ATP hydrolysis (6,8,9). Using different approaches, it has been shown that the PCNA loaded onto DNA is able to track along the template (10 -14). Preferential binding of RF-C to the template-primer junction, together with the footprinting data, suggested that assembly of the RF-C⅐PCNA complex occurred at the 3Ј-OH end of the primer (15). Analysis of loading of [ 32 P]PCNA onto DNAs of different structure showed that the assembly of the primary RF-C⅐PCNA complex does not require 3Ј-OH ends and might occur directly on dsDNA, whose topological structure and sequence do not present any restriction to loading. Such an initial salt-sensitive RF-C⅐PCNA complex was suggested to slide along DNA to the 3Ј-OH end of a primer, where the final conversion of the intermediate complex to the catalytically competent salt-resistant PCNA clamp takes place (13).
The close cooperation of RF-C and PCNA suggested that these proteins might act together in various DNA transactions. Both RF-C and PCNA were required to reconstitute DNA replication (16) and nucleotide excision repair in vitro (17). Base excision repair has been found to be carried out in two alternative pathways, one of which is PCNA-dependent (18,19). The PCNA dependence of base excision repair implies the requirement for RF-C in this process as well.
RF-C is a multiprotein complex composed of five subunits, one large subunit and four small subunits, named according their apparent molecular masses on SDS-PAGE: p140, p40, p38, p37, and p36. The cDNAs encoding all five subunits of human RF-C (20 -23) and yeast RF-C (24 -27) have been cloned. The five subunits are encoded by different genes that show extensive amino acid sequence homology in the middle part of the polypeptides (22,23,25,27). All five genes encoding the subunits of yeast RF-C have been shown to be essential for viability (24 -27). Each human RF-C subunit corresponds closely to its yeast counterpart (27; reviewed in Ref. 28), implying a conserved function of this protein in eukaryotic organisms.
Some of the human and yeast RF-C subunits have been expressed individually in E. coli to define their role in RF-C function. Interaction of RF-C with PCNA was attributed to p40 (29). p40 and p37 interacted directly with pol ␦ and pol ⑀, respectively, as well as with each other (29). Yeast Rfc3p (counterpart of p36) and Rfc4p (counterpart of p40) yielded a stable complex upon co-expression in E. coli (25). p37 and its yeast homolog Rfc2p showed specific primer end binding (26,29). DNA-dependent ATPase activity was found to reside on yeast Rfc3p (24). Recombinant yeast RF-C has been overexpressed in yeast by using a plasmid encoding all five RF-C subunits (30). Active recombinant human RF-C has been recently reconstituted using an in vitro coupled transcription/translation system (31) and baculovirus expression system (32).
In this paper we explored the recombinant baculoviruses encoding individual RF-C subunits to analyze the pathway of assembly of multisubunit RF-C complex. We report here the expression of functionally active recombinant RF-C and its various stable subcomplexes. A model for assembly of individual subunits into the five-subunit complex is proposed.

Proteins
Calf thymus pol ␦ (1 unit is defined as incorporation of 1 nmol of dTMP into poly(dA)-oligo(dT) (20:1 base ratio) in the presence of 100 ng of PCNA in 60 min at 37°C), calf thymus RF-C (Mono Q fraction), and E. coli SSB have been described (5). Human PCNA was overproduced in E. coli with a plasmid pT7/hPCNA and purified as described (4). The phosphorylatable derivative of PCNA, called ph-PCNA, was overproduced in E. coli, purified, and 32 P-phosphorylated by cAMP-dependent protein kinase (catalytic subunit) using [␥-32 P]ATP (13). Protein gp II was overproduced in E. coli strain K561 bearing pDG117IIA plasmid (33); the latter was kindly provided by K. Horiuchi (National Institute of Genetics, Mishima, Japan). gp II was purified from the soluble fraction according to Ref. 34 (34). The DNA preparations used for PCNA loading assays contained 75% RF II and 25% RF I DNA.

Construction of Baculovirus Transfer Vectors
RFC140 -Plasmid pSK/RFC140 was digested with HaeII, precipitated with ethanol, and treated with mung bean nuclease (20 units of nuclease/4.6 g of digested plasmid, 20-min incubation at 30°C). Digested DNA was phenol-extracted and cut with PstI, resulting in a cDNA fragment containing a blunt-ended 5Ј-terminus upstream of the ORF and a PstI site in the untranslated 3Ј-terminus. In the same way, a plasmid pBacHisC was digested successively by BamHI, mung bean nuclease, and PstI. cDNA was ligated into the pBacHisC and the resulting clones analyzed by sequencing to find clones containing the RFC140 gene in the reading frame of the His-tag leader with a minimal number of amino acids derived from 5Ј-untranslated sequence. Intentional excessive treatment of pSK/RFC140 by mung bean nuclease resulted in partial digestion of the untranslated 5Ј-terminus of cDNA fragment. The clone chosen for protein expression, pBacHis/RFC140, encoded Cys-Gly-Ala-Ala between the His-tag and the initiation Met of the ORF (Fig. 1). The same preparation of RFC140 cDNA fragment (HaeII, mung bean nuclease, and PstI digest) was ligated into pVL1393 digested with SmaI and PstI. The clone chosen for expression contained the ORF of RFC140 starting at position ϩ54 of the mutated polyhedrin gene.
RFC40 -The 5Ј-end of RFC40 cDNA was modified by PCR to gener-ate a BamHI site upstream of the ORF. The reaction was performed with Pwo pol using the forward primer dTCAAGGATCCATATGGAG-GTGGAGGCCG, the backward primer dGCCCCTGTCATTTGAAGC, and pET5a/40 (20) as a template. The PCR product was digested by BamHI and SmaI. The fragment containing the 5Ј-end of RFC40 cDNA (278 nt) was used to replace the BamHI/SmaI fragment in the original pET5a/40 plasmid. Next, the whole RFC40 cDNA was excised from the modified plasmid pET/RFC40 as a BamHI/EcoRI fragment and cloned into a pSK vector. RFC40 cDNA was then recloned as a BamHI/HindIII fragment from pSK/RFC40 into pBacHisC. The BamHI/EcoRI fragment from modified plasmid pET/RFC40 was cloned into plasmid pVL1393. RFC37-RFC37 cDNA in the plasmid pET5b/37 lacked the fragment encoding first 17 amino acids of p37 (21). To complete the cDNA, we performed reverse transcription followed by PCR. Total RNA from 293S cells was kindly provided by Irene Boche (this laboratory). The 5Јterminal 435-nt fragment was cloned by reverse transcription of 1 g of total RNA by Retrotherm reverse transcriptase with primer dACACG-GCTTCCCATCTGAG, followed by PCR using the forward primer dTCGCGGATCCATATGCAAGCATTTCTTAAAGG and the same backward primer used for reverse transcription. The PCR product was digested by BamHI and BstXI. The 272-nt fragment containing the 5Ј-end of RFC37 cDNA was used to replace the BamHI/BstXI fragment in the original pET5b/37. The full-length RFC37 cDNA was transfered from modified plasmid pET/RFC37 as a BamHI/PstI fragment into pBacHisC and as a BamHI/EcoRI fragment into pVL1393.
RFC38 and RFC36 -The cDNAs of both RFC38 and RFC36 were recloned from the corresponding plasmids pET19b/His38 and pET19b/ His36 as BamHI/HindIII fragments into pBacHisA and as BamHI/KpnI fragments into pVL1393.
Growth and maintenance of High Five insect cells (ITC Biotechnology GmbH) in adherent cultures, preparation of recombinant baculoviruses, and infection of the cells were performed according to routine protocols (35). High Five cells were grown in Grace's insect medium (Life Technologies, Inc.) supplemented with 10% fetal calf serum (Hy-Clone). For in vivo recombination, the transfer vectors and BaculoGold DNA (Pharmingen) were co-transfected into insect cells using DOTAP transfection reagent (Boehringer Mannheim). To prepare a control virus, the insect cells were co-transfected with unmodified vector pBacHisC and BaculoGold DNA, thus generating a virus encoding 54 amino acid polypeptide with His-tag at the N terminus. Infected cells were incubated for 48 h prior to harvesting the recombinant proteins.

Purification of Non-fused Recombinant RF-C
Analytical Purification on HAP-3.8 ϫ 10 7 cells infected either with five viruses encoding all RF-C subunits or with control virus were lysed in 1.5 ml of buffer A (20 mM Tris-HCl (pH 7.5), 1 mM DTT, 1 mM EDTA, 0.2% (v/v) Nonidet P-40, 10 mM NaHSO 3 , 1 mM phenylmethylsulfonyl fluoride, 1 g/ml each of aprotinin, leupeptin, and pepstatin) containing 0.1 M NaCl. Cell debris was removed by centrifugation and supernatant loaded onto a 0.2-ml HAP column, equilibrated in buffer A, 0.1 M NaCl. The column was washed with 0.5 ml of buffer A, 0.1 M NaCl, and 1 ml of 0.2 M potassium phosphate in buffer B (potassium phosphate (pH 7.5) at the indicated concentration, 10% (v/v) glycerol, 1 mM DTT, 1 mM phenylmethylsulfonyl fluoride, 1 g/ml each of aprotinin, leupeptin, and pepstatin). Proteins were eluted with 0.4 ml of buffer B, 0.4 M potassium phosphate.
Preparative Purification-2.7 ϫ 10 8 cells infected with five viruses encoding all RF-C subunits were lysed in 10 ml of buffer A, 0.1 M NaCl.

PCNA Loading Assay
Reaction mixtures (final volume of 20 l) contained 40 mM triethanolamine-HCl (pH 7.5), 1 mM DTT, 10 mM MgCl 2 , 1 mM ATP, 25 ng of [ 32 P]ph-PCNA, 100 ng of nicked DNA, and RF-C as indicated. Samples were incubated for 4 min at 37°C, then glutaraldehyde was added to a final concentration of 0.1% (w/v), and the mixtures incubated for another 10 min at 37°C. The samples were loaded and analyzed on a neutral 0.8% agarose gel in the presence of 0.1% SDS as described (13).

Polymerase ␦ Holoenzyme Assay
Reaction mixtures (final volume of 25 l) contained 40 mM Tris-HCl (pH 7.5), 0.2 mg/ml bovine serum albumin, 1 mM DTT, 10 mM MgCl 2 , 1 mM ATP, 50 M each of dATP, dGTP, dCTP, 20 M [␣-32 P]dTTP (500 cpm/pmol), 100 ng of primed ssDNA, 100 ng of PCNA, 1.2 g of E. coli SSB, 0.25 unit of pol ␦, and RF-C as indicated. Samples were incubated for 30 min at 37°C, the reactions were terminated by adding 1 ml of ice-cold 10% (w/v) trichloroacetic acid, and acid-insoluble material was analyzed by scintillation counting. For product length analysis, the reactions were terminated by treating them with proteinase K (60 g/ml) for 30 min at 37°C in the presence of 1% (w/v) SDS and 20 mM EDTA (pH 8.0). The DNA was then precipitated with ethanol and the products analyzed on an alkaline 1.5% agarose gel as described (36).

Other Methods
Protein concentration was determined according to Ref. 37 using Bio-Rad reagents and IgG as a standard.
For Western blot analysis, the proteins separated by SDS-PAGE were blotted to Optitran nitrocellulose membrane (Schleicher & Schuell) and probed with mAb 19 (22). The bands were visualized by using secondary anti-mouse antibody conjugated to alkaline phosphatase (Promega).
Radioactivity in the agarose gels was quantified using a Phospho-rImager (Molecular Dynamics). Ethidium bromide-stained agarose gels and Coomassie-stained polyacrylamide gels were photographed using Image Store 7500 (UVP) and analyzed with NIH Image 1.57 software.

Construction of Recombinant Baculoviruses-
The cDNAs of all five RF-C subunits were cloned into baculovirus vectors as described under "Materials and Methods," to produce each subunit both as a non-fused polypeptide and as an N-terminal His-tagged fusion. The 5Ј-termini of cDNA in all 10 transfer vectors were re-sequenced using a primer complementary to the polyhedrin promoter. The sequences around each start codon are shown on Fig. 1. The recombinant baculoviruses were named according to the encoded protein using the prefix v-(e.g. v140 for the virus encoding p140 protein, etc.).
Expression of Functional Recombinant Human RF-C Using Baculoviruses-Two assays were used to address the question whether the functionally active RF-C can be produced in the insect cells by co-infection with five recombinant baculoviruses encoding individual RF-C subunits. The first step of pol ␦ holoenzyme assembly, the loading of the sliding clamp, was analyzed by assaying the RF-C-dependent formation of the complex between [ 32 P]PCNA and circular dsDNA (13). The RF-C-dependent assembly of pol ␦ holoenzyme was analyzed by measuring DNA synthesis by pol ␦ on singly primed ssDNA template in the presence of RF-C, PCNA, RP-A (or E. coli SSB), and ATP (2,3,38). Extracts from insect cells, infected either with five RF-C viruses or control virus, were chromatographed on HAP (Fig. 2, A and B). Although most of the DNA synthesis activity in the extracts was not dependent on exogenous pol ␦ and was found at about equal level in both extracts, a new pol ␦-dependent activity was evident upon co-infection of the cells with the RF-C viruses. This pol ␦ stimulating activity copurified with PCNA loading activity (Fig. 2B), confirming that these two novel activities correspond to the recombinant RF-C. The PCNA loading assay proved to be a quite specific test for the expression of recombinant human RF-C in insect cells, as host polymerases did not interfere with it and insect RF-C appeared not to interact efficiently with human PCNA. The small scale analytical HAP chromatography with step elution was sufficient to reveal recombinant RF-C activity poorly detectable in crude extracts. PCNA loading activity of recombinant RF-C was about 40-fold higher than the background in the control experiment (Fig. 2C). The recombinant RF-C was further purified on a Mono Q column. The pol ␦ stimulating and PCNA loading activities coeluted at 0.25-0.3 M NaCl and the active fractions were essentially free of host pol activity (data not shown). The chromatographic behavior of the recombinant protein on HAP and Mono Q was similar to that of natural mammalian RF-C (1, 2, 5).
Polymerase ␦ holoenzyme DNA synthesis was linearly dependent on low concentrations of calf thymus RF-C, thus allowing to quantify the amounts of RF-C in terms of activity units (1 unit corresponded to the incorporation of 1 nmol of total dNMP into singly primed ssDNA in the presence of pol ␦, PCNA, and E. coli SSB in 30 min at 37°C, Ref. 5). Additional RF-C in the reaction mixture led to saturation of DNA synthesis (Fig. 3A). These saturation conditions apparently corresponded to the loading of at least one PCNA clamp per available template-primer junction and no additional RF-C was required once pol ␦ holoenzyme DNA synthesis started (36). Under our experimental conditions, pol ␦ holoenzyme DNA synthesis in the presence of saturating amounts of calf thymus RF-C was 40 pmol of dNMP in 30 min. Polymerase ␦ holoenzyme DNA synthesis in the presence of recombinant RF-C also reached saturation, with maximal dNMP incorporation of 14 pmol in 30 min, corresponding to 35% of the value obtained for calf thymus RF-C (Fig. 3B). Analysis of DNA synthesis products on alkaline agarose gel showed that the average size of the fragments synthesized in the presence of recombinant RF-C was about 400 -1000 nt in comparison with products of 600 -5000 nt synthesized by pol ␦ in the presence of calf thymus RF-C (Fig. 3C). The length of the products synthesized by pol ␦ holoenzyme is known to be dependent on incubation time (8,36) and concentration of pol ␦ core enzyme (39), but not on concentration of RF-C (2). To assess whether the slightly decreased activity of recombinant RF-C could be caused by a substoichiometric amount of the large subunit (see Fig. 4 below), we analyzed whether a complex of the four small RF-C subunits could inhibit pol ␦ holoenzyme DNA synthesis. Recombinant four-subunit complex p40-his⅐p38⅐p37⅐p36 (see below in the text and Fig. 6) in amounts up to 1 g, corresponding to about 300-fold molar excess over five-subunit RF-C, did not significantly decrease bovine RF-C-dependent DNA synthesis (data not shown). Therefore the decreased DNA synthesis performed in the presence of recombinant five-subunit RF-C was not due to non-stoichiometric expression of RF-C subunits. Another possible explanation for the apparently lower activity of human recombinant RF-C was that we used bovine pol ␦ for the DNA synthesis assay. Some species specificity in the interaction of human RF-C and bovine pol ␦ cannot be ruled out. Further comparison of natural human and bovine RF-C with recombinant human protein will be required to determine which function or property of RF-C affects the length of DNA products.
Expression of Functional Recombinant RF-C Containing Histagged Subunits-The recombinant baculoviruses were constructed to produce each RF-C subunit both as non-fused and His-tag-fused proteins, the latter designed for purification using affinity chromatography. Ni-NTA resin chromatography of the extracts from cells infected with any recombinant virus encoding non-fused RF-C subunit or with control virus reproducibly contained the same background polypeptides, most of them over 60 kDa. Only four minor host cell polypeptides were detected in the region of 35-50 kDa, and these were independent on the viruses used for infection (see, e.g., Figs. 4C, 5C, and 6A). Upon infection of insect cells with baculoviruses encoding individual His-tagged RF-C subunits, followed by Ni-NTA resin chromatography, the purified p40-his, p38-his, p37-his, and p36-his could be clearly detected on Coomassie-stained denaturing polyacrylamide gels. The amounts of soluble individual RF-C subunits purified using Ni-NTA resin were estimated using the bovine serum albumin standards loaded on the same denaturing gel. 3.8 ϫ 10 7 cells (one 150-cm 2 flask) infected with the corresponding baculovirus yielded 16.5 g of p37-his, 11 g of p40-his, 8.5 g of p36-his, and 1.2 g of p38-his (data not shown).
We investigated whether any of the His-tag fusions can be used to purify the whole five-subunit complex by affinity chromatography on Ni-NTA resin. Cells were infected with five viruses, one of them encoding a His-tagged subunit. The proteins from the cell extracts were purified on Ni-NTA resin as described under "Materials and Methods." When p40, p37, or p36 was His-tag-fused, the purified fractions displayed PCNA loading activity (Fig. 4B), and this activity was reproducibly highest in preparations containing the p36-his subunit. SDS-PAGE analysis followed by Coomassie staining showed that the fractions active in the PCNA loading assay contained polypeptides with molecular masses characteristic for small human RF-C subunits (Fig. 4C; compare the protein bands marked with dots in lanes 4, 6, and 7 with the background in lane 2).
Although the small polypeptides were directly detectable by Coomassie staining, the large RF-C subunit could not be seen, suggesting that this subunit might be underexpressed in comparison with other subunits. To pursue this question, the expression and co-purification of p140 with other subunits was analyzed by Western blot with mAb 19 against RFC140 (22). A 140-kDa polypeptide reacting with mAb was observed in the preparations containing p40-his, p37-his and p36-his (Fig. 4D), confirming that all five subunits copurify with the PCNA loading activity by affinity chromatography.
The results presented in Fig. 4 allowed us to address the question whether all five RF-C subunits are required to reconstitute active human RF-C in insect cells. Affinity-purified fractions were prepared from insect cells co-infected with combinations of viruses as shown at the top of Fig. 5. Omission of any of the five subunits resulted in the loss of PCNA loading activity (Fig. 5B) and the complete absence of p140 in the purified fractions (Fig. 5D), while the subcomplexes of small subunits could be seen (Fig. 5C). This result indicated that all five subunits are required to reconstitute functional RF-C and  figure  (lanes 3-8). Proteins from the cell extracts were purified on Ni-NTA resin (see "Materials and Methods"), and 1 l of fractions analyzed using the PCNA loading assay. After electrophoresis, the agarose gel was first stained with ethidium bromide and photographed (A), then fixed with acid, dried, and autoradiographed (B). Reaction mixture contained the following: 0.25 unit of calf thymus RF-C (lane 1) and fractions from the infected cells (lanes 2-8). C, 20 l of Ni-NTA resinpurified fractions were analyzed by 12.5% SDS-PAGE. The lower half of the gel was stained with Coomassie. The protein bands characteristic of the small RF-C subunits are marked with rectangular brackets. D, the upper half of the gel was blotted to nitrocellulose membrane and analyzed with mAb 19 against RFC140.

FIG. 6. Copurification of small RF-C subunits with p40-his upon co-infection of insect cells with recombinant baculoviruses. Insect cells were infected by various combinations of viruses as
indicated at the top of the figure. The proteins from the cytosolic fraction were purified on Ni-NTA resin as described under "Materials and Methods." The eluted fractions were analyzed by 12.5% SDS-PAGE and stained with Coomassie (A). The protein bands were quantified using the NIH Image 1.57 software (B). Markers at 40 and 50 kDa were from a 10-kDa ladder kit (Life Technologies, Inc.). that p140 binds stably to the small subunits only when all four of them are co-expressed. The conclusion that all of the subunits are required for biochemical activity of RF-C agrees with the findings that all five genes of yeast RF-C are essential for viability (24 -27).
Subunit Interactions in Recombinant RF-C-While the presence of p140 in the purified fractions could be clearly detected with mAb, this subunit appeared to be stoichiometrically underrepresented relative to the small subunits. Thus p140 might bind last to a preformed subcomplex of small subunits. Furthermore, all four small subunits must comprise this subcomplex, since omission of any of the small subunits resulted in the loss of p140 from the purified fractions (Fig. 5). To assess this possibility, we analyzed the co-purification of proteins upon co-infection of the cells with various combinations of the viruses encoding small subunits, one of them as a His-tag fusion.
Co-infection of the cells with v40-his, v38, v37, and v36 resulted in a protein preparation containing the four corresponding subunits: p40-his, p37, and p36 in stoichiometric amounts and p38 in about half the amount of the other subunits (Fig. 6A, lane 3). Upon infection of the cells with corresponding pairs of viruses: v40-his⅐v40, v40-his⅐v38, v40-his⅐v37, and v40-his⅐v36, no protein co-purified with p40-his, indicating that p40 does not form stable complexes with other any single subunit and does not tend to oligomerize (data not shown). Next, various combinations of three viruses were analyzed. Only p40-his was found in the purified fraction upon co-infection of the cells with v40-his, v38, and v37, indicating that these three subunits do not form a stable subcomplex (Fig. 6A,  lane 4). Co-infection of the cells with v40-his, v38, and v36 showed only very weak binding of p36 to p40-his (Fig. 6A, lane  5). In contrast, a stoichiometric complex of p40-his, p37, and p36 could be isolated upon co-infection of the cells with the corresponding viruses (Fig. 6A, lane 6), indicating that these three subunits form a stable intermediate. Since p38 is not required to form p40-his⅐p37⅐p36 complexes, the protein preparation in lane 3 is apparently a mixture of trimer p40-his⅐p37⅐p36 and tetramer p40-his⅐p38⅐p37⅐p36, the amount of the latter being limited by the expression of p38.
Assembly of subcomplexes was further investigated in coinfections using the v37-his virus in place of v40-his. When the cells were co-infected with v37-his, v40, v38, and v36, the four corresponding subunits could be seen in the purified fraction, with p37-his migrating most slowly (Fig. 7A, lane 2). In contrast to the v40-his results, the co-infections with pairs of viruses showed that p38, p37, and p36, but not p40, co-purify to some extent with p37-his (Fig. 7A, lanes 3-6). Upon co-infection of the cells with v37-his, v40, and v38, no proteins copurified with p37-his and even the weak interaction with p38 was inhibited (Fig. 7A, lane 7; see also Fig. 5C, lane 8). This result confirms the above observation that the subunits p40, p38, and p37 do not form a stable subcomplex. Upon co-infection of the cells with v37-his⅐v40⅐v36 and v37-his⅐v38⅐v36, the corresponding tertiary complexes of subunits could be observed, although not with stoichiometric ratios of subunits as for p40-his⅐p37⅐p36 (see Fig. 6A, lane 6).
To characterize these two complexes more carefully, 6-fold larger amounts of the cells were infected with the corresponding viruses, proteins were purified using Ni-NTA resin, and the eluted fractions were chromatographed on a Mono Q column (Fig. 7, B and C). Analysis of the extract from the cells infected with v37-his⅐v38⅐v36 showed that the complex containing p37his and p36 eluted from the Mono Q column first, followed by the complex of p37-his, p38, and p36 (Fig. 7B). We supposed figure. The proteins from the cytosolic fraction were purified on Ni-NTA resin as described under "Materials and Methods." The eluted fractions were analyzed by 12.5% SDS-PAGE and stained with Coomassie. B, 2.7 ϫ 10 8 cells were infected with v37-his, v38, and v36. Cytosolic extract was mixed with 0.36 ml of Ni-NTA resin and incubated for 1 h at 4°C, the resin pelleted by centrifugation, packed in the column, washed with 10 ml of buffer D, and proteins eluted from the resin with 1.5 ml of buffer E. The eluate was diluted 6-fold with buffer C, loaded on a 1-ml Mono Q column, and chromatographed in a 15-ml gradient of NaCl from 50 to 500 mM in buffer C. Proteins collected in 0.5-ml fractions were precipitated with 5% trichloroacetic acid, dissolved in loading buffer, and analyzed by 12.5% SDS-PAGE followed by Coomassie staining (fractions 8 -15). The sample of individually expressed p37-his was loaded onto the far right lane as an additional marker. C, 2.7 ϫ 10 8 cells were infected with v37-his, v40, and v36. Purification and analysis were performed exactly as in panel B. that p38 was underexpressed in comparison with p37-his and p36, resulting in the appearance of the excess of p37-his⅐p36 complex. The ratio of the band intensities for p37-his⅐p36 complex was 1:0.55 (determined in fraction 9), and for p37-his⅐p38⅐p36 complex it was 1:0.62:0.64 (fraction 13). In the case of infection with v37-his, v40, and v36, no p37-his⅐p36 complex could be found, while the complex p37-his⅐p40⅐p36 eluted as a sharp symmetrical peak with a band ratio 1:0.66:0.70 for p37his, p40, and p36, respectively (Fig. 7C).

FIG. 7. Copurification of small RF-C subunits with p37-his upon co-infection of insect cells with recombinant baculoviruses. A, insect cells were infected by various combinations of viruses as indicated at the top of the
To confirm these results, the same series of experiments were carried out using v36-his co-infections. Though p36-his appears to be the most useful for the purification of biologically active RF-C, its subunit interaction analysis was most difficult to interpret, since p36-his overlapped with p40 in SDS-PAGE. Co-infection of the cells with v36-his, v40, v38, and v37 resulted in a poorly resolved complex, and only p37 could be clearly identified (Fig. 8A, lane 2). Co-infections with pairs of viruses showed that p36-his binds to p37 (lane 7), but does not interact detectably with p38 (lane 6); no conclusion could be made for p40 (lane 5). Co-infection of the cells with v36-his, v40, and v38 yielded no polypeptide band co-purifying with p36-his (lane 3). Although p36-his⅐p40⅐p37 complex could not be resolved into three components, it was evident that the amount of p37 bound to p36-his complexes was higher for v36-his⅐v40⅐v37 than for v36-his⅐v37 infections (compare lanes 4 and 7). This cooperative interaction of p40 and p37 further validated the data obtained using p40-his and p37-his (Figs. 6 and 7). The protein complex(es) resulting from co-infection of the cells with v36-his⅐v38⅐v37 were purified using Ni-NTA resin followed by chromatography on the Mono Q column (Fig. 8B). The complex p36-his⅐p37 eluted first, followed by the complex p36-his⅐p38⅐p37 (Fig. 8B). The ratio of the band intensities was 1:0.68 for p36-his⅐p37 (fraction 10) and 1:0.56:0.69 for p36-his⅐p38⅐p37 (fraction 15).
Model of Assembly of Five-subunit RF-C Complex (Fig. 9)-The RF-C intersubunit interactions described above  allow to propose a model for assembly of RF-C subunits into the five-subunit complex. p37 and p36 form a dimeric complex stable enough to copurify on affinity resin and anionexchanger. We found only two stable tertiary subcomplexes, p40⅐p37⅐p36 and p38⅐p37⅐p36, which might be intermediates in the RF-C assembly. Both subcomplexes could be derived from the common precursor p37⅐p36. Therefore, we postulate a bifurcated pathway for the RF-C assembly; either p40 or p38 binds to p37⅐p36 dimer to form tertiary complex. The missing fourth small subunit binds to the tertiary complex, forming a quaternary complex of small subunits. The latter binds the large subunit to form a catalytically competent five-subunit complex (Fig. 9). A more limited model for assembly of RF-C subunits coexpressed using an in vitro coupled transcription/translation system has also been proposed recently (31). Data from that system indicated the following. (i) No stable complex with two subunits was formed, although weak interactions could be observed for several pairs of subunits; (ii) the only stable threesubunit core complex was formed with p40, p37, and p36; (iii) p140 and p38 bound only cooperatively to the core complex p40⅐p37⅐p36 (31). However, co-expression of RF-C subunits using baculoviruses revealed additional intersubunit interactions, resulting in a significantly expanded model (Fig. 9). Thus the model based on the in vitro coupled transcription/translation system appears to represent one of the alternative assembly pathways observed in the baculovirus expression system. The inability to detect some of the subcomplexes in the in vitro expression system (31) that were detected using the baculovirus system could derive from differences between the expression systems, for example in protein modification. Although baculovirus expression does not guarantee the correct processing of mammalian polypeptides, the posttranslational modification of proteins in insect cells might be very close to those occurring in mammalian cells (see Ref. 40, and references therein). Two closely migrating bands of p140 were detected in Western blots of recombinant human RF-C (Fig. 4D). Two closely migrating bands of the large subunit of mouse RF-C were also visible in Western blots of mouse whole cell extracts (41). These results suggest that the baculovirus expression system might allow proper posttranslational modification of RF-C subunits.
The possibility to produce soluble individual RF-C subunits, several subcomplexes, as well as functional five-subunit RF-C using recombinant baculoviruses provides a useful approach to the detailed biochemical analysis of the structure and mechanism of action of this protein, as well as its role in DNA replication and DNA repair machineries as a whole. This will allow the in vitro characterization of functional subcomplexes of the subunits and various mutant polypeptides and will shed light on the biochemical role of each subunit in the overall function of RF-C.