|Home | About | Journals | Submit | Contact Us | Français|
The papillomavirus E1 protein is a multifunctional initiator protein responsible for preparing the viral DNA template for initiation of DNA replication. The E1 protein encodes two DNA binding activities that are required for initiation of DNA replication. A well-characterized sequence-specific DNA binding activity resides in the E1 DBD and is used to tether E1 to the papillomavirus ori. A non-sequence-specific DNA binding activity is also required for formation of the E1 double trimer (DT) complex, which is responsible for the local template melting that precedes loading of the E1 helicase. This DNA binding activity is very poorly understood. We use a structure-based mutagenesis approach to identify residues in the E1 helicase domain that are required for the non-sequence-specific DNA binding and DT formation. We found that three groups of residues are involved in nonspecific DNA binding: the E1 β-hairpin structure containing R505, K506, and H507; a hydrophobic loop containing F464; and a charged loop containing K461 together generate the binding surface involved in nonspecific DNA binding. These residues are well conserved in the T antigens from the polyomaviruses, indicating that the polyomaviruses share this nonspecific DNA binding activity.
The first step in initiation of DNA replication is the binding of an initiator protein to the replication origin (ori). Local melting of the DNA sequences in the ori and subsequent loading of the replicative DNA helicase follow this recognition step. Although it is well established that these processes must occur in all organisms with double-stranded DNA (dsDNA) genomes, the mechanistic understanding of these events is limited. For example, it is known that in Escherichia coli the initiator protein DnaA recognizes ori C and generates local melting, but the mechanism by which melting occurs is not known, although various models have been proposed (6, 13). Viral initiator proteins such as T-ag from simian virus 40 (SV40) and E1 protein from papillomaviruses provide good systems for the study of the early events in initiation of DNA replication since these multifunctional proteins can carry out all of the activities required to prepare the template for initiation of DNA replication in vitro.
The papillomavirus initiator E1 is an ~70-kDa AAA+ (ATPases associated with various cellular activities) protein and belongs to the SF3 helicase family (18). E1 proteins provide at least four activities required for replication initiation: specific and nonspecific DNA binding, local DNA melting of ori, and unwinding activity carried out by the E1 DNA helicase in front of the replication fork (11, 12, 14, 16, 20, 24, 26, 28, 30, 32). In recent years the study of viral initiator proteins such as E1 and SV40 T-ag has received a significant boost from a combination of biochemistry and structural biology (2, 8, 10, 15, 17, 19, 21, 27).
The E1 DNA binding domain (DBD) has a high specificity for E1 binding sites (E1 BS) in the papillomavirus ori (28, 29). However, this sequence-specific DNA binding activity is masked in the full-length protein by a nonspecific DNA binding activity present in the E1 helicase domain (21, 28). The combination of these two DNA binding activities results in a modest specificity of E1 for binding to the ori, since the nonspecific DNA binding activity is dominant (28). Due to the low sequence specificity of E1, binding of E1 to the ori in vivo requires the virus-encoded transcription factor E2. E1 and E2 bind cooperatively to the ori in the form of an E12E22 ori complex that has high specificity for ori because the activation domain of E2 interacts physically with the helicase domain of E1 (1, 28). This interaction prevents nonspecific DNA binding by the E1 helicase domain, and the E12E22 complex therefore binds to the ori with much higher sequence specificity than E1 alone (28). After the E12E22 complex has bound to ori, the E2 dimer is displaced and further E1 molecules are added to the complex, eventually resulting in the formation of a complex containing six E1 molecules bound to the ori as a head-to-head E1 double trimer (DT). In vitro, in the absence of competitor DNA, the low specificity of E1 is sufficient to form the DT in the absence of E2 (21). The E1 DT serves as a required precursor for the functional helicase, which is an E1 double hexamer (DH). In the transition from a DT to a DH the template is melted locally, allowing the formation of the DH helicase on single-stranded DNA (ssDNA) (17, 21).
The formation of the DT requires the cooperation of the sequence-specific DNA binding activity present in the E1 DBD and the nonspecific DNA binding activity present in the E1 helicase domain (21). E1 DBD and its binding to the E1 BS have been studied extensively by biochemical, genetic, and structural biology approaches and is well understood (3-5, 7, 9, 29). Much less is known about the nonspecific DNA binding activity and how the E1 helicase domain contacts dsDNA in a sequence-nonspecific manner.
Structural information for the E1 helicase domain is available from several structures: a complex between the E1 helicase domain and the transactivation domain of E2, as well as a fragment containing the oligomerization and helicase domains of E1 as a hexamer assembled around ssDNA (1, 8). In the hexamer structure, the six helicase domains form a channel enclosing the ssDNA. Each helicase domain projects a specific motif, the β-hairpin, into the channel in a right-handed staircase pattern. The oligomerization domains contact each other to form a rigid collar that holds the six subunits together. This structural information does not provide direct information about the interaction between the E1 helicase domain and dsDNA since the structure contains ssDNA. However, based on previous studies, two conserved residues at the tip of the β-hairpin (K506 and H507) have been implicated in non-sequence-specific binding to dsDNA (2, 17, 21). Examination of the E1 structure provides additional candidates for residues that could be involved in dsDNA binding. We performed a mutagenesis screen of these surface residues in the E1 helicase domain together with some residues in the oligomerization domain to determine whether these were involved in non-sequence-specific DNA binding. In this screen we identified five residues in the E1 helicase domain that are required for sequence nonspecific DNA binding. Three of these residues—F464, R505, and K506—have dual function and affect both the helicase activity and the DT formation of E1, while the other two show severe defects only for DT formation.
Wild-type E1 and E1 mutants were expressed in E. coli as N-terminal glutathione S-transferase (GST) fusions and purified by affinity chromatography. After isolation of the fusion proteins, the GST portion was removed by thrombin digestion, and the material was further purified by Mono S ion-exchange chromatography as described previously (25). E1 purified in this manner is monomeric, as determined by glycerol gradient sedimentation and gel filtration. E2 protein was expressed in E. coli without a tag and purified to apparent homogeneity by successive ion-exchange chromatography steps as described previously (25).
All ori probes were generated by PCR using an ori template and primers end labeled with [γ-32P]ATP and T4 polynucleotide kinase. To generate the probe in Fig. Fig.2,2, primers were used that change the DNA sequence of the E1 BS 1 and 2 from AACAAT and AATAAT to AACAAA and AAAAAT, respectively, resulting in the loss of E1 binding. Probes were purified by polyacrylamide gel electrophoresis (PAGE), eluted by diffusion, and precipitated.
DNase footprinting was carried out as described previously (25). Protein and probe (15,000 cpm) were incubated at room temperature in binding buffer [20 mM potassium phosphate (pH 7.4), 0.1 M NaCl, 1 mM dithiothreitol (DTT), 1 mM EDTA, 0.7 mg of bovine serum albumin (BSA)/ml, 0.1% NP-40, 5% glycerol, and 2 ng of poly(dI-dC) in a final volume of 10 μ1]. After 20 min, 10 μ1 of a solution containing 5 mM CaCl2 and 10 mM MgCl2 was added together with DNase I. After 60 s, cleavage was terminated by the addition of 130 μl of STOP solution (0.2 M NaCl, 10 mM EDTA, 1% sodium dodecyl sulfate [SDS]), followed by phenol-chloroform extraction. The aqueous phase was precipitated by the addition of 350 μl of 0.45 M ammonium acetate in ethanol.
Four percent acrylamide gels (39:1, acrylamide-bis) containing 0.5 × TBE lacking EDTA, were used for all electrophoretic mobility shift assay (EMSA) experiments. E1 was added to the probe (~2 fmol) in 10 μl of binding buffer (20 mM HEPES [pH 7.5], 100 mM NaCl, 0.7 mg of BSA/ml, 0.1% NP-40, 5% glycerol, 5 mM DTT) supplemented with 5 mM MgCl2 and 2 mM ADP. After incubation at room temperature for 1 h, the samples were loaded and run for 2 h at 9 V/cm. Competition assays were performed by mixing 2 fmol of radiolabeled probe with 2, 4, or 8 fmol of pUC18 plasmid prior to incubation with E1.
In vitro replication was performed in a 25-μl reaction mixture containing 40 mM HEPES-KOH (pH 7.5); 8 mM MgCl2; 0.5 mM DTT; 3 mM ATP; 0.2 mM (each) GTP, UTP, and CTP; 0.1 mM (each) dATP, dGTP, and dTTP; 10 μM [α-32P]dCTP (2 μCi; 800 Ci/mmol); 40 mM creatine phosphate; 400 ng of creatine kinase; 10 μl of S100 extract; and 2.5 μl of high-salt nuclear extract from H293 cells. Reactions were incubated for 60 min at 37°C. The concentration of template in the in vitro reactions was 2 ng/μl. The reactions were stopped by the addition of SDS to 1% and EDTA to 10 mM and treated with proteinase K, followed by phenol-chloroform extraction. The products were analyzed by electrophoresis on 1% agarose gels in TAE buffer (23, 33). For reactions challenged by competitor DNA 500 ng of poly(dA-dT) was added prior to the addition of E1. For reactions containing E2, 3 ng of E2 purified from E. coli was added prior to the addition of E1 (23).
Transient replication assays were carried out as described (31). Briefly, three plasmids—an ori plasmid (11/12/X), an expression vector for E2 (pCGE2), and an expression vector for wild-type or mutant E1 (pCGE1)—were electroporated into Chinese hamster ovary (CHO) cells. Three days after transfection low-molecular-weight DNA was isolated, digested with DpnI, and analyzed by agarose gel electrophoresis, Southern transfer, and hybridization with a radioactively labeled ori probe.
Two different helicase assays were used. In the first (see Fig. Fig.8A),8A), a radiolabeled helicase substrate was generated by annealing a 41-nucleotide (nt) oligonucleotide to a 59-nt oligonucleotide, generating a molecule with a 15-nt 5′ tail, 27 bp of complementarity, and a 32-nt 3′ tail. E1 was incubated with the substrate (~2 fmol) in a buffer containing 20 mM HEPES (pH 7.5), 5 mM MgCl2, 0.7 mg of BSA/ml, 0.1% NP-40, 5 mM DTT, 2 mM ATP, and 5% glycerol at 37°C for 30 min. The reaction was terminated by the addition of SDS to 0.1%, and the samples were analyzed by PAGE.
For the time course helicase assay (Fig. (Fig.8B),8B), a helicase kit from Perkin-Elmer (TruPoint) was used. Briefly, an europium-labeled 44-mer annealed to a 26-mer containing a fluorescence quencher was incubated with E1 in a buffer containing 50 mM Tris-HCl (pH 7.9), 5 mM MgCl2, 2 mM DTT, 1 mM ATP, and 0.2 mg of BSA/ml and then incubated at 37°C. Time-resolved fluorescence was measured every 2 min by using 1420 Victor software in a Perkin-Elmer fluorometer. The substrate concentration was 4 nM, the capture strand concentration was 15 nM, and the reactions were carried out in a volume of 50 μl. At 30 nM, wild-type E1, maximally 50% of the substrate, is unwound.
It is now well established that a particular structure, the β-hairpin in the E1 helicase domain, is required for formation of the E1 DT on the ori (17, 21). Most likely, the β-hairpin is directly involved in non-sequence-specific binding of the E1 helicase domain to the dsDNA flanking the E1 binding sites in the BPV origin of DNA replication. Structural data demonstrate that the β-hairpin lines the inner surface of the channel formed by the six subunits of the E1 helicase and oligomerization domains (8). We reasoned that if residues other than those in the β-hairpin are involved in binding to dsDNA, we might be able to identify these by modeling a monomer of the E1 oligomerization and helicase domain onto dsDNA (Fig. (Fig.1).1). We extracted a monomer of the E1 oligomerization and helicase domains from the hexamer structure and placed it on dsDNA in a linear arrangement, as suggested by the localization of the E1 DBD immediately adjacent to the oligomerization domain. The oligomerization and helicase domains were oriented such that the residues K506 and H507, which are know to affect interaction with dsDNA, were in contact with DNA. Based on this model, we selected residues on the helicase domain that might interact with dsDNA. These residues (S456, N459, K461, S462, F464, T490, Y491, and D504A) were mutated to alanine. We also expanded the set of mutations in the β-hairpin by generating R505A and K508A, which together with the previously characterized mutants K506A and H507A, accounts for the majority of the residues in the β-hairpin. We also generated mutations in the E1 oligomerization domain (N352A, S353A, K356A, and K359A) based on their proximity to DNA in our model. We changed all of these residues to alanine by site-directed mutagenesis in full-length E1, and expressed and purified the mutant proteins from E. coli. Because the alanine substitution at K508 could not be expressed, a leucine substitution (K508L) was generated instead.
Full-length E1 protein can bind to DNA in multiple ways and depending on the structure of the probe different types of complexes can be generated with E1 (21). Formation of the functional complex, the DT, requires cooperation of the two different DNA binding activities, the sequence specific binding activity located in the E1 DBD and the non-sequence-specific binding activity present in the E1 helicase domain, which depends on the β-hairpin. Formation of the DT also depends on the four paired E1 BS in the ori and requires ADP or ATP and a minimal probe size of 84 bp, as discussed below. The two different DNA binding activities can be studied individually by using different probes. On a very short probe (39 bp) centered on the E1 BS but lacking the flanking sequences E1 can form a head-to-head dimer where E1 DBD is bound as a dimer to paired E1 BS. Formation of this head-to-head dimer depends on the sequence-specific DNA binding activity present in the E1 DBD but does not require the sequence nonspecific DNA binding present in the helicase domain. Mutation of two of the E1 BS in one half of the ori results in a switch from the head-to-head dimer bound to the E1 BS via the E1 DBD to binding of E1 as a trimer. Formation of this trimer does not depend on E1 BS or on the E1 DBD but requires the non-sequence-specific DNA binding activity present in the E1 helicase domain. We used the trimer assay to screen the mutants that we had generated in the helicase and oligomerization domain (Fig. (Fig.22).
Wild-type E1 formed a discrete trimer under these conditions (Fig. (Fig.2A,2A, lanes 1 to 3) and so did S456A, N459A, S462A, T490A, and D504A (Fig. (Fig.2A,2A, lanes 4 to 10, 13 to 15, 20 to 22, and 25 to 27), whereas K461A, F464A, and Y491A failed to from the trimer (Fig. (Fig.2A,2A, lanes 10 to 12, 16 to 18, and 22 to 24). We knew from previous studies that Y491A (data not shown) had structural problems, and this mutant was therefore not analyzed further. As we have described previously, two substitutions at the tip of the β-hairpin, K506A and H507A (Fig. (Fig.2B,2B, lanes 4 to 9), were defective for trimer formation in this assay, and we also found that R505A (lanes 1 to 3) in the β-hairpin was defective for trimer formation, whereas K508L (lanes 10 to 12), the fourth mutant in the β-hairpin, could still form the trimer. Of the oligomerization domain mutants N352A, S353A, K356A, and K359A (Fig. (Fig.2B,2B, lanes 13 to 25), K359A (Fig. (Fig.2B,2B, lanes 22 to 24) showed a very slight defect for trimer formation. K356A (Fig. (Fig.2B,2B, lanes 19 to 21) showed an unusual phenotype in that this mutant formed a predominant dimer instead of the trimer.
A defect for trimer formation could be due to the loss of a number of different activities in the mutant protein, including defects in nucleotide binding since nucleotide binding is required for trimer formation. We therefore tested the trimer defective mutants also for DT formation, since in these assays intermediate phenotypes can be observed. We used the wild-type 84-bp ori probe and performed the EMSA in the presence of ADP (21). We first compared wild-type E1 and the mutants K356A, K359A, K461A, and R505. We also included K506A and H507A, which as we have previously demonstrated are defective for DT formation. All of the mutants (Fig. (Fig.3A,3A, lanes 5 to 22) showed some defects for DT formation. The defects were of different kinds, however. K359A (Fig. (Fig.3A,3A, lanes 8 to 10) formed complexes containing four, five, and six E1 molecules similar to wild-type E1, indicating only a slight defect for DT formation, a finding consistent with the very slight defect observed for trimer formation (Fig. (Fig.2).2). The oligomerization domain mutant K356A (Fig. (Fig.3A,3A, lanes 5 to 7) formed a strong E12 band and a ladder extending up to complexes with four to five E1 molecules. K461A (Fig. (Fig.3A,3A, lanes 11 to 13) was also defective for DT formation forming a predominant E15 complex. The β-hairpin mutants R505A (Fig. (Fig.3A,3A, lanes 14 to 16), K506A (Fig. (Fig.3A,3A, lanes 17 to 19), and H507A (Fig. (Fig.3A3A lanes 20 to 22) all formed strong E12 complexes and ladders up to E15 complexes. The fact that all of these mutants were capable of DNA binding demonstrated that the defect is not caused by a gross overall structural defect of these proteins.
We were particularly interested in the highly conserved residue F464, which based on the hexamer structure and biochemical analysis is required for the helicase activity of the E1 hexamer and is also conserved in SV40 T-ag (2, 8, 27). We therefore performed a more extensive mutagenesis at this position to try to distinguish between the role of this residue in helicase activity and in DT formation. We generated a set of additional substitutions at this position (F464L, M, T, N, K, Y, and H), and in Fig. Fig.3B3B we tested the F464 substitutions in the DT formation assay. Interestingly, the substitutions fell into three different groups. The substitutions F464A (Fig. (Fig.3B,3B, lanes 4 to 6), F464L (Fig. (Fig.3B,3B, lanes 7 to 9), F464M (Fig. (Fig.3B,3B, lanes 10 to 12), F464N (Fig. (Fig.3B,3B, lanes 16 to 18), and F464K (Fig. (Fig.3B,3B, lanes 19 to 21) all had similar effects, and these mutants generated distinct E12, E13, and E14 complexes. F464T (Fig. (Fig.3B,3B, lanes 13 to 15) and F464H (Fig. (Fig.3B,3B, lanes 25 to 27) behaved differently and formed also a strong E15 complex. Finally, F464Y (Fig. (Fig.3B,3B, lanes 22 to 24) formed E15 and E16 complexes. These results demonstrate that the properties of F464 are critical for DT formation. The fact that the tyrosine substitution, which has similar properties as the phenylalanine, is the only mutant that can form the DT demonstrates that it is the hydrophobic ring structure that is critically important for DT formation.
We have previously demonstrated that we can distinguish between the DNA binding by the E1 DBD and the E1 helicase domain by using a DNase footprinting assay (21). E1 DBD binds sequence specifically to the E1 BS and the helicase domain binds to the sequences flanking the E1 BS. Consequently, E1 mutants such as H507A and K506A fail to protect the sequences flanking the E1 BS (21). To test whether the other residues are also involved in binding to the flanking sequences, we tested the alanine substitutions at K356, K359, K461, F464, R505, K506, and H507 in the DNase I footprinting assay (Fig. (Fig.44).
Comparison of the footprints generated by the E1 DBD (Fig. (Fig.4,4, lanes 1 to 2) and full-length E1 (Fig. (Fig.4,4, lanes 3 to 4) illustrate the difference between protections of the E1 BS and the additional protection of the sequences flanking the E1 BS. The protection by the E1 DBD encompasses the E1 BS, ~18 bp. The footprint generated by full-length E1 extends by an additional ~30 bp on both sides of the E1 BS. As described previously, the two β-hairpin substitutions K506A and H507A (Fig. (Fig.4,4, lanes 15 to 18) both give rise to robust protection over the E1 BS but very weak or nonexistent protection over the flanking sequences (Fig. (Fig.4,4, compare lanes 15 to 18 to lane 19). The β-hairpin substitution R505A (Fig. (Fig.4,4, lanes 13 to 14), the oligomerization domain mutant K356A (Fig. (Fig.4,4, lanes 5 to 6), and the helicase domain mutant K461A (Fig. (Fig.4,4, lanes 9 to 10) all behaved similarly, providing robust protection of the E1 BS and weak protection of the flanking sequences. The two substitutions K359A and F464A behaved differently. K359A (Fig. (Fig.4,4, lanes 7 to 8) gave rise to a protection similar to that of wild-type E1, with protection of the sequences on both sides of the E1 BS. F464A (Fig. (Fig.4,4, lanes 11 to 12) generated flanking protection on the upstream side of the E1 BS but not on the downstream side of the E1 BS. These results demonstrate that all of the substitutions, with the exception of K359A, show defects for protection of the sequences flanking the E1 BS. K359A is the substitution that showed the least defect for trimer and DT formation (see Fig. Fig.22 and and3).3). The reason for the asymmetric protection observed with F464A is unclear but may indicate that the sequences on the upstream side of the E1 BS provide better binding sites for the E1 helicase domain.
As discussed above, E1 utilizes two different DNA binding activities to form a functional DT complex. The specific DNA binding activity of the E1 DBD is used to bind to the E1 BS, and the nonspecific DNA binding activity that resides in the E1 helicase and oligomerization domain binds to the DNA sequences flanking the E1 BS. We have previously demonstrated that in DNA binding assays where sequence-specific DNA binding is challenged by the presence of nonspecific competitor DNA, the nonspecific DNA binding activity is dominant (21, 28). Therefore, because full-length E1 contains both DNA binding activities, binding of the full-length E1 to the E1 BS can be readily competed for by nonspecific competitor DNA, whereas binding of the E1 DBD, which lacks the non-sequence-specific DNA binding activity, is completely insensitive to competitor DNA.
A prediction based on these observations is that mutations that affect the nonspecific DNA binding activity of E1 would make E1 less sensitive to nonspecific competitor DNA than the wild-type E1 protein. This would provide us with an assay to measure the level of nonspecific DNA binding activity in the context of the full-length E1 with an intact DBD, a task that would be difficult using any other method. Because the mutants that we wanted to test for nonspecific DNA binding activity are defective for trimer and DT formation, we took advantage of the ability of these mutants to form E1 dimers on short ori probes containing the E1 BS (21). We performed these assays by mixing the probe with nonspecific competitor DNA (pUC19) at three different ratios. Wild-type E1 or E1 mutants were then added to the probe mixes, incubated, and loaded onto EMSA gels. As observed in Fig. Fig.5A,5A, wild-type E1 in the absence of competitor DNA forms a robust dimer band with a trace of E13 (lane 1). In the presence of competitor DNA the level of binding is greatly reduced, reflecting the sensitivity to the competitor DNA (Fig. (Fig.5A,5A, lanes 2 to 4). Complex formation by wild-type E1 was reduced >20-fold at the highest concentration of nonspecific competitor DNA (compare lanes 1 and 4). The mutants showed a different pattern. The mutant that was least sensitive to competitor DNA, K506A, showed a <2-fold reduction in complex formation even at the highest level of competitor DNA, indicating that K506 contributes the most to nonspecific DNA binding (Fig. (Fig.5A,5A, lanes 21 to 24). This low level of sensitivity to competitor DNA was similar to that observed with the E1 DBD (Fig. (Fig.5B,5B, lanes 13 to 16). The remaining mutants, K356A, K359A, K461A, H507A, F464A, and R505A showed 2- to 3-fold reductions in binding at the highest level of competitor DNA. These results demonstrate that mutation of any of the residues that we have identified as affecting DT formation renders E1 a more sequence-specific DNA binder. We interpret these results to mean that all of these residues in some way contribute to the nonspecific DNA binding activity of E1.
A peculiar feature of the formation of the E1 DT on the ori is that this complex is strongly dependent on the absolute length of the probe. Formation of the DT depends on the presence of 4 E1 BS in the ori. However, in addition to the four E1 BS that occupy 18 bp, ~33 bp of flanking sequences on either side of the E1 BS (a total of 84 bp) is required for efficient DT formation. These flanking sequences are bound by the E1 helicase domain but without apparent sequence specificity. We have previously shown that deletion of 17 bp from either the left or the right end of the probe resulted in complexes containing five E1 molecules instead of the six E1 molecules present in the DT (21). We interpreted this result to mean that E1 in the DT is bound to the DNA in a helical arrangement and that deletion of sequences from either end therefore results in the loss of one monomer of E1. To further this observation, we generated a series of probes with different size deletions from the left hand (upstream) end. We tested complex formation of wild-type E1 on the truncated probes in the presence of ADP (Fig. (Fig.6).6). On the 84-bp probe E1 formed a discrete complex containing six E1 molecules corresponding to a DT. Deletion of 7 bp had only a slight effect on complex formation observed as an increase in the E15 band relative to the E16 complex (Fig. (Fig.6,6, lanes 16 to 18). Deletion of 10 bp, however (Fig. (Fig.6,6, lanes 13 to 15), showed a clear effect. With this probe, the E15 complex now becomes predominant over the E16 complex. Further deletion results in the appearance of smaller complexes such that deletion of 22 bp results in a predominant complex containing four E1 molecules (Fig. (Fig.6,6, lanes 1 to 3). Due to the large sizes of these complexes the probe size contributes very little to the mobility of the complex. These results suggested that DT assembly requires the sequences flanking the E1 BS in a length-dependent manner and that, interestingly, removal of the first 10 to 13 bp results in the formation of an E15 complex. Removal of another 9 bp results in an E14 complex. These results are difficult to explain by any other mode of DNA binding than a helical arrangement of the E1 molecules around the dsDNA (Fig. (Fig.6C6C).
The data presented thus far indicated that our screen, which is based on the ability of E1 to bind DNA as a trimer, had yielded a number of mutations defective in nonspecific DNA binding. To determine whether the nonspecific DNA binding activity of E1 was required for viral DNA replication, we tested the substitutions in DNA replication assay in vivo and in vitro (Fig. (Fig.7).7). In Fig. Fig.7A,7A, we compared the ability of wild-type E1 and the 7 E1 mutants to support DNA replication in cell free replication reactions. In these assays, a plasmid containing the ori is incubated with E1 in a H293 cell extract in the presence of radiolabeled nucleotide. The level of DNA synthesis can be quantitated after separation of the replication products by agarose gel electrophoresis. In Fig. Fig.7A7A we compared the ability of the different E1 proteins to support replication under three different conditions. We have previously demonstrated that in vitro DNA replication is independent of the viral E2 protein under standard conditions, although replication in vivo requires the E2 protein (23). However, E2 can be made essential for DNA replication under conditions in which nonspecific competitor DNA is added to the reactions prior to the addition of E1. Replication can then be rescued by the addition of E2, since E2 confers on E1 the ability to bind DNA with high sequence specificity. This phenomenon is illustrated in lanes 1 to 3 with wild-type E1. In lane 1, wild-type E1 gives rise to robust replication. Addition of nonspecific competitor DNA [poly(dA-dT)] suppresses replication by >20-fold (lane 2). The addition of E2 under these conditions restores DNA replication to ~40% of the replication observed in the absence of competitor DNA (lane 3). Only three of the mutants showed detectable replication in the absence of competitor DNA. K356A showed ~40% of wild-type activity in the absence of competitor DNA (lane 4), this activity was completely suppressed in the presence of competitor DNA (lane 5) and was not restored detectably in the presence of E2 (lane 6). K359A had the same activity as wild-type E1 in the absence of competitor DNA (lane 7), was suppressed completely by competitor DNA (lane 8), and was restored by the addition of E2 (lane 9), thus behaving virtually identically to wild-type E1. Finally, K461A showed a trace of replication activity in the absence of competitor DNA (lane 10). The rest of the substitutions (F464A, R505A, K506A, and H507A, lanes 5 to 8) showed <5% of wild-type activity.
In Fig. Fig.7B7B we tested all of the F464 substitutions (F464A, F464L, F464M, F464T, F464N, F464K, F464Y, and F464H, lanes 2 to 9) for in vitro DNA replication. All of these substitutions were defective for in vitro DNA replication showing <5% of wild-type activity.
Finally, to determine what role the substitutions played for replication in vivo, we generated mammalian expression vectors for the substitutions K356A, K359A, K461A, F464A, R505A, K506A, and H507A. These expression vectors were transfected into CHO cells, together with a viral ori plasmid and an expression vector for E2. At 3 days after transfection, low-molecular-weight DNA was isolated, digested with DpnI, and analyzed by Southern blotting and hybridization with an ori probe. Interestingly, only two mutants showed detectable activity. K356A had ~10% of the wild-type activity, while K359A had ~40% of the wild-type activity. These results clearly differ significantly from those obtained in the in vitro DNA replication assays. The fact that the mutants in the oligomerization domain retained significant replication activity while the mutants in the helicase domain did not indicates that there may be a qualitative difference between these two groups of mutants.
Since K506A is required for both nonspecific DNA binding and DNA helicase activity the β-hairpin is clearly involved in both these processes. To determine whether the rest of the mutants defective for nonspecific DNA binding had helicase defects, we tested the mutants for helicase activity in an oligonucleotide displacement assay, as shown in Fig. Fig.8A.8A. Wild-type E1 showed robust helicase activity at the highest concentration of E1. The mutants K356A (Fig. (Fig.8A,8A, lanes 6 to 8) and K359A (lanes 9 to 11) similarly showed robust DNA helicase activity showing 90 and >100% of the wild-type activity, respectively, demonstrating that these residues in the oligomerization domain do not contribute to the helicase activity of E1. The mutants K461A (lanes 12 to 14) and H507A (lanes 21 to 23) also had substantial activity, ~60% of wild-type E1. The remaining mutants R505A (lanes 15 to 17) and K506A (lanes 18 to 20) lacked detectable helicase activity. These results demonstrate that the β-hairpin residues R505 and K506 clearly play essential roles in the E1 helicase activity. In contrast, the two residues in the oligomerization domain play no role in helicase activity.
We wanted to determine whether the highly conserved residue F464, which is important for DT formation, is conserved because of its role in DT formation or because of its role in the helicase activity of E1. For example, although F464Y had significant activity in the DT formation assay (as shown in Fig. Fig.3),3), F464Y lacked activity in the in vitro replication assay. To try to distinguish between a critical role of F464 in DT formation and DNA helicase activity, we tested the F464 substitutions for DNA helicase activity using a real-time fluorescent substrate helicase assay (Fig. (Fig.8B).8B). Interestingly, all of the substitutions at F464 had severe defects for DNA helicase activity. Two mutants F464H and F464M showed a trace of activity, while the remaining mutants were barely above background levels of activity. These results indicate that the absolute conservation of F464 among all papovavirus E1 proteins and T antigens is likely due to the requirement of this particular residue for DNA helicase activity. This conclusion is supported by the fact that this residue is also conserved in viruses with ssDNA genomes, such as adenoassociated viruses, that are unlikely to require template melting.
Viral initiator proteins such as the E1 proteins from the papillomaviruses and T antigens from the polyomaviruses are multifunctional proteins that can carry out all of the activities required to prepare a DNA template for initiation of DNA replication. The multifunctional nature of these proteins has made dissection of functional domains difficult, since single domains frequently have more than one function. However, recent high-quality structural information has made it possible to perform more sophisticated mutational analyses to address how these initiator proteins carry out their varied activities (2, 8, 10, 15, 17, 19, 21, 27).
Two different domains contribute to the DNA binding activity of the E1 protein. The sequence-specific DNA binding activity resides in a distinct DBD, which has been very well characterized by biochemical and structural means, and sequence-specific DNA binding is therefore fairly well understood. However, the E1 proteins also contain a nonspecific DNA binding activity that is essential for template melting. This activity is very poorly characterized. Our previous studies have shown that two residues in the E1 β-hairpin, K506 and H507, are involved in sequence nonspecific DNA binding (21). To identify other residues that take part in sequence-nonspecific DNA binding, we extracted a monomer of the E1 oligomerization and helicase domains from the hexamer structure and modeled this monomer on dsDNA, asking what residues other than K506 and H507 might be capable of interacting with dsDNA. We selected residues at four positions (N352, S353, K356, and K359) in the E1 oligomerization domain and at twelve positions in the E1 helicase domain (S456, N459, K461, S462, F464, T490, Y491, D504, R505, K506, and H507). These mutants were then screened for the ability to form the E1 trimer complex (Fig. (Fig.2).2). Five of the mutants in the helicase domain (K461A, F464A, R505A, K506A, and H507A) failed to form the trimer and lost DNA binding completely. One of the mutants in the oligomerization domain (K356A) was defective for E13 formation but instead formed an E12 complex very efficiently. The same mutants that were defective for trimer formation were, as expected, generally defective also for DT formation (Fig. (Fig.33).
To test whether these mutants were defective for sequence-nonspecific DNA binding, we performed DNase footprinting assays with them (Fig. (Fig.4).4). We have previously demonstrated that the E1 helicase domain binds to the sequences flanking the E1 BS and that E1 protein with mutations in the β-hairpin (K506A and H507A) fails to protect these sequences (21). Consistent with a defect for nonspecific DNA binding, all of the mutants, with one exception, failed to protect the sequences flanking the E1 BS. This exception, K359A, showed robust protection over the flanking sequences (Fig. (Fig.4,4, lanes 7 to 8), demonstrating that this mutant retained nonspecific DNA binding activity. These results indicate that the nonspecific DNA binding activity depends on multiple residues in E1.
We further analyzed the mutants by testing them in the comprehensive in vitro and in vivo DNA replication assays (Fig. (Fig.7).7). The results demonstrated that the majority of the mutants that we have identified as important for non-sequence-specific DNA binding lacked replication activity altogether. Interestingly, however, the two mutants in the oligomerization domain behaved differently. K359A showed no defect for in vitro DNA replication, a finding consistent with the very slight defects in trimer and DT formation and the lack of a defect for sequence-nonspecific DNA binding activity observed by DNase foot printing. K356A showed a slight (~2-fold) reduction for in vitro DNA replication, a surprisingly small effect given the defect of this mutant for trimer formation. However, both K356A and K359A showed more significant defects for DNA replication in vivo where K356A showed barely detectable levels of replication (>10% of the wild type), and K359A had ~40% of the wild-type activity. The results from the in vivo and in vitro replication assays are very interesting. The failure of E2 to rescue K356A for in vitro replication (Fig. (Fig.7A)7A) indicates that this particular mutant may be defective for interaction with E2. The interaction with E2, which is essential for DNA replication in vivo, is dispensable for replication in vitro unless the DNA binding specificity of E1 is challenged by the addition of competitor DNA. This defect in interaction with E2 could therefore explain the discrepancy between in vivo and in vitro replication for this mutant.
These results demonstrate that the two mutants in the oligomerization domain behave differently than the mutants in the helicase domain which all have severe replication defects in vivo and in vitro. K359A had a very mild phenotype in most of the assays, and the lack of a severe replication defect is therefore not surprising. However, K356A only had a very slight defect for in vitro DNA replication, although trimer and DT formation was significantly compromised (Fig. (Fig.22 and and3).3). Consequently, this is the only mutant where we do not observe a correlation between the failure to form a trimer or DT and a failure to replicate in vitro. We believe that this very modest defect for in vitro DNA replication is inconsistent with any significant defect for nonspecific DNA binding. This conclusion is different from that arrived at from similar data by Sanders (19). We believe that the defect of this particular mutant may be in oligomerization rather than in nonspecific DNA binding. In the trimer formation assays in Fig. Fig.2,2, this mutant formed an efficient E1 dimer complex instead of the E1 trimer, indicating that the DNA binding activity was not compromised. This indicates that trimer formation might be affected either by preventing DNA binding or by preventing oligomerization. In the first case, the effects on DNA replication are very severe; in the second case, the effects on DNA replication are modest.
Although structures of various parts of E1 are now available, the structure of the E1 DT-DNA complex is unknown. Our biochemical data provide some information that can be used to propose how the DT is formed. First, our data indicate that the trimer and DT are held together by the interaction with DNA. We have seen no indication that these complexes can form in the absence of DNA. However, the formation of these complexes likely involves some protein-protein interactions, most likely between the oligomerization domains, as indicated by the K356A mutation, which is defective for trimer formation, although it can still bind DNA as a dimer. Our conclusion therefore is that the trimer and DT are generated through a combination of protein-protein and protein-DNA interactions.
The EMSA results using probes shorter than 84 bp illustrate a phenomenon that we have observed previously(21). The formation of the DT is very sensitive to reduction in the length of the probe. Reduction of the size of the probe by as little as 7 to 10 bp results in a very clear reduction in the ability to form a DT. Because these changes in probe length result in very predictable changes in complex formation, we believe that the E1 DT is a symmetrically arranged helical complex that encircles the dsDNA.
The mutational analysis identifies three elements in the BPV1 E1 helicase domain as being involved in nonspecific dsDNA binding and DT assembly. These three elements—the β-hairpin that contains R505, K506, and H507, a hydrophobic loop that contains the residue F464, and a charged loop that contains the residue K461—are located in close proximity to each other (Fig. (Fig.9).9). Of the five residues in the helicase domain that are involved in nonspecific dsDNA binding, F464 and K506 are invariant, whereas H507 is highly conserved in all PV initiator E1 proteins (Fig. (Fig.9).9). The fourth and fifth residues, R505 and K461, show a modest level of conservation, and many other residues are allowed at these positions. However, the specific defects of both of these residues for trimer and DT formation and for in vitro DNA replication indicates that these residues indeed are essential for E1 function.
We have no direct information about how the E1 helicase domain interacts with dsDNA, but we can arrive at a general idea about how the residues that we have identified bind to dsDNA based on the properties of their side chains. Three of the residues (K461, R505, and K506) are positively charged. In one case (K461A) we know that the positive charge is important for function since substitution of this Lys with Arg retains the ability to form a trimer and DT (data not shown), whereas substitution with alanine does not (Fig. (Fig.22 and and3).3). Two other residues, H507 and F464, are ring structures. We have demonstrated that the ring structure of the H507 is important for function since a Tyr substitution at this position is the only substitution that retains the ability to form a DT (17). Similarly, a Tyr substitution at F464 showed less of a defect for DT formation than the other substitutions (Fig. (Fig.3).3). Thus, the nonspecific DNA binding activity is contributed by three positively charged residues and two hydrophobic residues where the ring structure is important. A simple model consistent with these data is that the nonspecific interaction with DNA consists of a charge interaction with the phosphate backbone and a hydrophobic interaction, possibly with the minor groove. A minor groove interaction has been demonstrated to be required for local template melting (22).
Two of the residues that we have identified as being involved in nonspecific DNA binding are also required for other functions. K506 is required for the DNA helicase activity of E1 in addition to its role in trimer and DT formation (17). Similarly, as shown here, R505 and F464 are critical for the DNA helicase activity of E1 in addition to their roles in nonspecific DNA binding and DT formation. The remaining two residues, H507 and K461, appear to have no function in E1 DNA helicase activity, since alanine substitutions at these positions only show modest defects compared to wild-type E1 (Fig. (Fig.8A).8A). This partial overlap between residues required for DT formation and for DNA helicase activity is consistent with the idea that melting of dsDNA and unwinding of dsDNA by the E1 helicase are separate but related processes.
This research was supported by a grant RO1 AI 072345 to A.S.
Published ahead of print on 10 February 2010.