The E1 Initiator Recognizes Multiple Overlapping Sites in the Papillomavirus Origin of DNA Replication

Abstract

A common feature of replicator sequences from a variety of organisms is multiple binding sites for an initiator protein. By binding to the replicator, initiators mark the site and contribute to melting or distortion of the DNA. We have defined the recognition sequence for the papillomavirus E1 initiator and determined the arrangement of binding sites in the viral origin of replication. We show that E1 recognizes a hexanucleotide sequence which is present in overlapping arrays in virtually all papillomavirus replicators. Binding of the initiator to these sites would result in the formation of a closely packed array of E1 molecules that wrap around the double helix.

Initiation of DNA replication requires the recognition of the replicator, or origin of replication (ori), by an initiator protein. Initiator proteins, as the name implies, function by promoting initiation of DNA replication and can perform this function in several ways. In addition to ori recognition, initiators interact with components of the replication machinery, such as DNA polymerases, thereby acting as “landing pads” for the binding of replication factors to the ori. Binding of initiator proteins frequently results in the structural distortion of the ori which may promote local ori melting and subsequent DNA unwinding. In some cases, viral initiators have intrinsic helicase activity and are capable of unwinding DNA without the assistance of other factors.

This latter class of initiator proteins, the T antigens from viruses such as simian virus 40 (SV40) and polyomavirus and the E1 proteins from papillomaviruses can perform all of the functions that are required for unwinding (for reviews about these proteins, see references 10 and 43). These proteins share many features, although the two types of viruses are only distantly related. Both proteins have common domains, such as a DNA-binding domain (DBD) and nuclear localization signal in the N-terminal half of the protein and ATP-binding and helicase domains in the C-terminal half (4, 5, 10, 19, 20, 25, 31, 44). Interestingly, the structures of the DBDs are very similar in spite of a lack of significant sequence homology (9, 22), indicating that the two proteins overall may be more similar than expected from the low degree of sequence similarity. Upon binding, both proteins form oligomeric structures which are capable of locally melting the ori (2, 3, 12, 29, 30, 40). The helicase-active form of both proteins is a hexamer, which unwinds DNA strand specifically (11, 34, 38, 40, 48, 50). Bidirectional unwinding is likely made possible by the formation of two hexamers at the site of unwound DNA, which has been observed for T antigen (26, 39, 48).

How an oligomeric helicase forms from these monomeric ori-binding initiator proteins has resisted elucidation. The relationship between the number of binding sites and the oligomeric forms of the proteins is not obvious. In the case of the well-characterized T antigen, four pentanucleotide repeats with the sequence GAGGC are found in the SV40 ori and are bound with high sequence specificity by T antigen (7, 8, 45). However, only a subset of these sites seem to be required for both local ori distortion and double-hexamer formation (18). Despite the occupation of only a subset of sites by the double hexamer, all four sites are required for extensive unwinding of the DNA (18). Thus, how T antigen utilizes its binding sites to assemble an active oligomeric initiator complex that can distort the ori and subsequently form a helicase has yet to be determined.

The structural and functional homology between SV40 T antigen and papillomavirus E1 proteins suggests that common mechanisms are used by these viruses in the assembly of active initiator complexes. Like T antigen, the E1 protein recognizes and binds to multiple sites within the ori (16, 17, 27, 35, 37, 47, 49). However, in contrast to SV40, a second viral protein, the transcription factor E2, is utilized in combination with E1. Through cooperative interaction with E2, E1 binding to the ori is stimulated, and the specificity with which E1 binds to the ori is also greatly increased (1, 23, 28, 32, 37, 46, 49). E1 therefore requires E2 as a loading factor which confers specificity (24, 30). Several E1-containing ori complexes have been isolated and characterized, lending insight into the steps which occur during the assembly of an active initiator complex capable of DNA unwinding. The cooperative binding of E1 and E2 to the ori results in the formation of the E1E2-ori complex, which is capable of highly sequence-specific DNA binding and ori recognition (32). The E1E2-ori complex is a direct precursor of an E1-ori complex, which contains no E2 and has a greater number of E1 monomers bound to the ori compared to the E1E2-ori complex and which is active for local ori melting (24, 30). The progression from the E1E2-ori complex to the E1-ori complex through the loading of additional E1 monomers onto the ori suggests that the formation of even larger forms of E1 may involve, similarly, the stepwise binding of additional E1 molecules to preexisting ori-bound E1 complexes.

We have previously characterized the DNA-binding properties of the bovine papillomavirus (BPV) E1 DBD (4). We have also demonstrated that the E1 palindrome contains at least two binding sites for E1 and that these binding sites are used in the formation of the E1E2-ori complex. Here, we have used the E1 DBD to further characterize the E1 binding site. We demonstrate that E1 recognizes variants of the hexanucleotide sequence AACAAT which is present in six copies in the E1 binding site. These sites, which are conserved in both sequence and position in other papillomavirus ori's, are arranged in an overlapping array. The occupation of six potential E1 binding sites within the ori suggests a mechanism for the generation of oligomeric E1 complexes, ultimately resulting in E1 hexamers with DNA helicase activity.

MATERIALS AND METHODS

Protein expression and purification.

The expression and purification of the E1 and E2 DBDs have been described previously (4).

Plasmid constructs.

All BPV ori constructs and mutants were generated from a plasmid containing the BPV minimal ori (sequences between nucleotides 7914 and 27 in the BPV genome) cloned between the XbaI and HindIII restriction sites in pUC19 but with a high-affinity E2 binding site, E2 BS9 (21).

Xho probes.

Ori constructs containing an 8-bp XhoI linker flanking half of the E1 palindrome and the BS9 E2 binding site were generated by PCR using a primer containing the XhoI linker for the top strand (GGTCTAGACCTCGAGGAACAAT) and a primer containing the E2 binding site for the bottom strand. The template was an ori containing an XhoI linker inserted into the HpaI restriction site in the center of the E1 palindrome. All point mutations in the E1 binding site in this context were generated by PCR using top-strand primers containing the desired mutation. Xho probes containing inosine substitutions were generated using primers containing the inosine in either the top or the bottom strand depending on the location of the purine to be substituted. The +3 and the B42+3 ori's containing a three-nucleotide insertion between the E1 and E2 binding sites (see Fig. 5) were generated by PCR using a primer containing the high-affinity E2 binding site 9 and a three-nucleotide insertion (CAC) between the E1 palindrome and the beginning of the E2 binding site. The template used was either the wild-type (wt) BPV minimal ori (+3) or an ori containing a single point mutation in nucleotide position 7942 of the BPV sequence (B42+3) (35).

FIG. 5.

Inosine substitutions indicate that E1 recognizes the major groove. Inosine substitutions were generated at the six positions corresponding to the E1 recognition sequence and tested for binding as described above. (A) I/C base pairs were generated at positions 1 (lanes 11 to 15), 2 (lanes 21 to 25), and 3 (lanes 26 to 30) on the top strand. Probes containing G/C base pairs at positions 1 and 2 (lanes 6 to 10 and 16 to 20, respectively) are shown for comparison. Lanes 1 to 3, 6 to 8, 11 to 13, 16 to 18, 21 to 23, and 26 to 28 contained both E1 and E2 DBDs. Lanes 4, 9, 14, 19, 24, and 29 contained E2 DBD alone; lanes 5, 10, 15, 20, 25, and 30 contained probe alone. (B) I/C base pairs were generated at positions 4 (lanes 11 to 15), 5 (lanes 21 to 25), and 6 (lanes 31 to 35) on the top strand. Probes with G/C base pairs at the corresponding positions are shown for comparison (lanes 6 to 10, 16 to 20, and 26 to 30, respectively). Lanes 1 to 3, 6 to 8, 11 to 13, 16 to 18, 21 to 23, 26 to 28, and 31 to 33 contained E1 and E2 DBDs. Lanes 4, 9, 14, 19, 24, 29, and 34 contained E2 DBD alone; lanes 5, 10, 15, 20, 25, 30, and 35 contained probe alone. (C) Summary of the results from inosine substitutions. The wt level of binding is set to 1.0.

DEPC analysis.

Diethyl pyrocarbonate (DEPC) interference experiments were performed as described previously (4, 42).

Electrophoretic mobility shift assay (EMSA).

Probe (5,000 cpm/reaction) was mixed with E1 and/or the E2 DBD together with 20 ng of nonspecific competitor DNA (pUC119) in binding buffer (20 mM potassium phosphate [pH 7.4], 0.1 M NaCl, 1 mM EDTA, 0.1% NP-40, 3 mM dithiothreitol, 0.7 mg of bovine serum albumin/ml, 10% glycerol). Binding reactions were performed in a total volume of 10 μl. After incubation for 30 min at room temperature, the samples were loaded on 6% 40:1 (acrylamide-bisacrylamide) polyacrylamide gels and subjected to electrophoresis in 0.5× Tris-borate-EDTA. Quantitation was performed using a Fuji BAS 1000 imaging system.

RESULTS

The recognition sequence for E1 consists of the hexanucleotide AACAAT.

Mutational analysis of the 18-bp E1 palindrome has indicated that mutations in all positions affect the binding of E1 (17, 35). However, the binding of E1 in multiple different complexes has made it difficult to unambiguously assign the specific sequence that E1 recognizes. To define an E1 recognition sequence, we constructed an ori template which contains one-half of the E1 palindrome, together with the E2 binding site, flanked by an XhoI linker (Xho probe, Fig. 1A). We have previously shown that E2 can stimulate binding of one monomer of E1 to the half of the E1 palindrome proximal to the E2 binding site when either the distal half is mutated or the two halves of the palindrome are separated by an 8-bp linker insertion (4). These results strongly suggested that a recognition sequence for binding of an E1 monomer was contained within one-half of the palindrome. We measured E1 binding in the presence of E2, since the complex containing only the E1 monomer has low stability.

FIG. 1.

The E1 protein binds to the hexanucleotide sequence AACAAT. (A) A probe containing an 8-bp XhoI linker flanking half of the E1 palindrome with the E2 binding site in its natural position was used in EMSA with E1 and E2 DBDs. Under these conditions, a monomer of the E1 DBD binds cooperatively with E2 DBD to the probe. Transversion mutations were generated in the XhoI linker (B), the central 6 bp (C), and the right flank (D) and then tested by EMSA. The results are shown in graphical form in panel E. (B) Two transversion mutations in the XhoI linker (M1, lanes 9 to 16, and M2, lanes 17 to 24) were generated and compared to the wt probe for E1 binding. Three fourfold dilutions of E1 DBD in the absence (lanes 1 to 3, 9 to 11, and 17 to 19) or presence (lanes 4 to 6, 12 to 14, and 20 to 22) of the E2 DBD were used. Lanes 7, 15, and 23 contained E2 DBD alone; lanes 8, 16, and 24 contained probe alone. The position of the combined E1 and E2 DBD complexes and the E2 DBD complex are indicated by arrows. (C) Six transversion mutations were generated in the E1 half-palindrome and tested for E1 binding as described in panel B. Mutations C3 to G (lanes 6 to 10), A4 to T (lanes 11 to 15), and A5 to T (lanes 16 to 20) are shown in comparison to the wt (lanes 1 to 5). Only binding in the presence of E2 is shown. Lanes 1 to 3, 6 to 8, 11 to 13, and 16 to 18 contained both E1 and E2 DBDs. Lanes 4, 9, 14, and 19 contained E2 DBD alone; lanes 5, 10, 15, and 20 contained probe alone. (D) Three transversion mutations were generated in the right flank in positions 7 to 9 and tested for binding of E1 DBD as described above. Mutations A7 to T (lanes 6 to 10), A8 to T (lanes 11 to 15), and T9 to A (lanes 16 to 20) are shown in comparison with the wt (lanes 1 to 5). Lanes 1 to 3, 6 to 8, 11 to 13, and 16 to 18 contained both E1 and E2 DBDs; lanes 4, 9, 15, and 19 contained E2 DBD alone; lanes 5, 10, 15, and 20 contain probe alone. (E) Summary of the results of transversion mutations in E1 binding site. The effects of mutations is shown in a bar graph. The level of E1 binding to the wt sequence is set at 1.0.

Figure 1 shows EMSA results in which mutations were generated in the left flank (the Xho linker, panel B), the central hexanucleotide sequence (panel C), and the right flank (panel D), and the results are summarized in the graph in Fig. 1E. In the presence of E1 alone, little complex formation is observed (Fig. 1B, lanes 1 to 3). In the presence of E2 DBD alone, a single complex is formed (Fig. 1B, lane 7). In the presence of both the E1 and E2 DBDs, two major complexes form (Fig. 1B, lanes 4 to 6). The faster-migrating complex contains only the E2 DBD. The second, slower-migrating complex contains a monomer of the E1 DBD bound cooperatively with the E2 DBD. The identity of this complex was verified by DEPC interference analysis (data not shown). The interference pattern produced by this complex was virtually identical to that observed for a complex that was formed on a probe containing a mutation in the half of the E1 palindrome distal to the E2 binding site which we previously have shown contains a monomer of the E1 DBD together with the E2 DBD (4). The more slowly migrating complex observed at the highest concentration of E1 DBD is of unknown origin but likely corresponds to nonspecific binding of an additional E1 molecule to the XhoI linker sequence to form an E1 dimer since digestion of the probe with XhoI results in loss of the larger complex (data not shown). The two mutations in the XhoI linker, C-to-G transitions (lanes 9 to 16 and 17 to 24) had little or no effect on E1 binding, as expected.

Mutations in the central six nucleotides all had significant effects on E1 binding (see Fig. 1C for specific examples and Fig. 1E for a summary). For example, a change from a C to a G in the third position of the half-palindrome reduced the formation of the complex containing a monomer of E1 bound together with E2 by >10-fold (Fig. 1C, lanes 6 to 8). Similarly, a transversion mutation in the fifth position, A5 to T, also resulted in a severe reduction in the level of E1 binding to the probe together with E2 (lanes 16 to 18). An A-to-T transversion in the fourth position, however, affected E1 binding only slightly (lanes 11 to 13). In contrast, transversion mutations on the right flank (Fig. 1D), namely, A to T at position 7 (lanes 6 to 10), A to T at position 8 (lanes 11 to 15), and T to A at position 9 (lanes 16 to 20), had small effects on E1 binding, indicating that the sequences important for E1 binding are contained within positions 1 to 6. This is consistent with the minor effects of mutations at positions 7 and 8 observed for E1 monomer binding in the absence of E2 (13).

The effects of single mutations on E1 binding are summarized in Fig. 1E. Transversion mutations in any of the first six nucleotides, except the fourth, significantly reduced E1 binding. However, mutations outside of this hexanucleotide sequence, namely, in positions 7 to 9, only slightly impaired E1 binding to the Xho probe in the presence of E2. Mutations within the XhoI linker insertion also did not affect E1 binding. This suggests that the sequence AACAAT constitutes the recognition sequence for binding of a monomer of E1 DBD. This is consistent with previous results which show strong DEPC interference within this sequence produced by the complex containing a single monomer of the E1 DBD together with the E2 DBD (4). The effects of these mutations are also very similar to the effects of transversion mutations on the formation of the full-length cross-linked E1E2-ori complex (35). Importantly, in that study the E2 proximal and distal E1 binding sites (2 and 4) were affected virtually identically by mutations, indicating that the proximity to E2 and the interaction with the E2 DBD does not alter sequence recognition by E1 significantly. Thus, although these binding experiments were performed in the presence of E2 DBD, the results are likely a reflection of the sequence specificity of E1.

We subsequently performed a complete mutagenesis of this hexanucleotide sequence to determine the effects of both transition and transversion mutations. The effects of all of the single-point mutations are shown in the graph in Fig. 2. Transversion mutations within the hexanucleotide recognition sequence, namely, in positions 1, 2, and 3, have significant effects on E1 binding and are more severe than the corresponding transition mutations. Mutations of the remaining three positions, 4, 5, and 6, showed a more complex pattern. In position 4 of the binding site, mutating the A into either a C or a T had little effect on E1 binding, whereas the single transition mutation into a G has a severe effect, suggesting that features specific to a G/C base pair are inhibitory for E1 binding. In position 5, any base pair other than an A/T base pair nearly abolishes E1 binding, suggesting that a specific determinant in the A/T base pair is important for E1 binding. In position 6, T-to-G or T-to-C mutations have modest effects on binding, while a T-to-A transversion has the most severe effect of any single point mutation (>20-fold reduction in binding), indicating that an A/T base pair at this position is particularly deleterious for binding.

FIG. 2.

Summary of the effects of mutations in all positions on E1 binding. The six positions which showed significant effects on E1 binding in Fig. 1 were chosen for a complete mutagenesis, wherein the bases at all positions were changed into the three possible alternatives. The mutants were tested for binding of E1 DBD as shown in Fig. 1, and the results were quantitated and are shown as a bar graph.

The E1 palindrome contains six nested E1 binding sites.

Cooperative binding of E1 and E2 to the wt probe results in binding of E1 to the sequence AACAAT and an identical sequence in the other half of the palindrome and binding of E2 to the E2 binding site, based on interference and mutational analysis (E1 BS 2 and 4, Fig. 3A). Inspection of the E1 palindrome indicated that a second pair of putative E1 binding sites are present (E1 BS 1 and 3). Binding sites 2 and 4 both have the sequence AACAAT. Binding sites 1 and 3 have the sequences AACAAC and AATAAT, respectively. The differences between these sites and binding sites 2 and 4 is a single change in either the sixth or the third position in sites 1 and 3, respectively. These changes are in positions where transition mutations have modest effects on E1 binding (Fig. 2). Further inspection of the E1 palindrome reveals that there are two additional hexanucleotide sequences which are similar to sites 1 to 4 and overlap sites 1 and 4. Binding sites 5 and 6 have the sequences AATCAC and AATTAT, respectively. Both contain a transition at position 3 and a transversion at position 4, each of which should not seriously impair E1 binding (Fig. 2). Binding site 5 also has a C/G base pair instead of an T/A base pair in the sixth position which, according to the mutagenesis results in Fig. 2, is predicted to have only minor effects on E1 binding. Thus, it is conceivable that these other sites can be bound by E1.

FIG. 3.

(A) The information from the mutagenesis experiments was used to identify other potential E1 recognition sequences in the ori based on sequence homology. These new putative sites, 1, 3, 5, and 6, are shown together with the known sites 2 and 4 as shaded boxes. (B) Sites 1, 3, 5, and 6 were inserted into the Xho probe, and binding was measured by EMSA in the presence of E2 DBD as described for Fig. 1. The results are summarized in the bar graph.

To determine whether these flanking sequences could indeed serve as E1 binding sites, we utilized the assay depicted in Fig. 1, which allows us to isolate and measure binding to an individual E1 binding site. We replaced the wt site with sequences comprising the other putative sites and tested the affinity of E1 to binding to these sites in the presence of the E2 DBD. The level of E1 binding observed when site 1, 3, 5, or 6 was substituted for site 2 in the Xho ori construct is shown as a bar graph in Fig. 3B. Binding to all four sites was very similar; sites 3 and 6 showed an approximately threefold reduction compared to site 2, and sites 1 and 5 showed an approximately fivefold reduction compared to site 2.

A second pair of recognition sequences for E1 is present in the E1 binding site.

We previously hypothesized, based on our earlier analysis of E1 complexes formed with the full-length E1 protein, that two different complexes with overlapping binding sites could form on the ori (35). In addition to the paired binding sites 2 and 4, mutagenesis and interference analysis indicated that a second pair of E1 binding sites was present partially overlapping the first pair but shifted by 3 bp. To determine directly whether this second pair of recognition sequences corresponded to sites 1 and 3, we constructed an ori template containing a 3 bp insertion between the E2 binding site and the E1 palindrome (+3 ori). On the wt ori, the E2 DBD can stimulate binding of the E1 DBD specifically to sites 2 and 4. Sites 1 and 3 are shifted by three nucleotides relative to binding sites 2 and 4. Thus, it may be possible to stimulate binding of the E1 DBD to sites 1 and 3 if the E2 binding site is also moved by three nucleotides, which would place the E2 DBD in register with the putative binding sites 1 and 3. We compared complex formation using the wt probe and the +3 ori in EMSA. We also included the probe B42+3. This probe has a point mutation in E1 binding site 4 which prevents the formation of the E1 dimer complex on sites 2 and 4. The results are shown in Fig. 4A.

FIG. 4.

E1 DBD binds to binding sites 1 and 3 when the E2 binding site is moved by 3 bp. (A) EMSA was performed to determine the ability of the E1 DBD to bind to the ori in combination with the E2 DBD by using a probe with the E2 binding site in the wt position (wt, lanes 1 to 8) or when the E2 binding sites was moved 3 bp through an insertion of 3 bp between the E1 and E2 binding sites (+3, lanes 9 to 16) or with a probe containing the +3 insertion and a point mutation in binding site 4 (B42+3, lanes 17 to 24). Lanes 1 to 3, 9 to 11, and 16 to 18 contained three twofold dilutions of E1 DBD alone. Lanes 4 to 6, 12 to 14, and 19 to 21 contained the same dilutions of E1 DBD in the presence of E2 DBD. Lanes 7, 14, and 23 contained E2 DBD alone; lanes 8, 15, and 24 contained probe alone. The arrows indicate the migration of the respective complexes. (B) DEPC interference analysis was performed on both strands of the B42+3 probe. The probes were modified with DEPC, and the complexes formed in the presence of E2 DBD alone (lanes 1 and 4) and both E1 and E2 DBDs (lanes 2 and 5) were isolated and subjected to cleavage with piperidine and analyzed on a sequencing gel. The positions of interference on both strands are indicated by filled circles. (C) Diagram of the positions of interference in the wt and +3 ori templates. The DEPC interference produced by binding of a dimer of E1 DBD in the presence of E2 DBD on the B42+3 probe is shown in comparison with the interference obtained for a complex formed on a wt probe (4). The B42 mutation is shown in boldface.

In the absence of the E2 DBD, the E1 DBD forms a predominant complex with the wt ori, and we know from previous studies that this complex contains two monomers of the E1 DBD bound to sites 2 and 4 (Fig. 4A, lanes 1 to 3). At the highest concentration of E1 DBD, a second, slower-migrating complex can form, and this probably represents the binding of additional molecules of E1 either specifically or nonspecifically to the ori (lane 1). In the presence of the E2 DBD, binding by the E1 DBD to sites 2 and 4 is stimulated through the formation of the E1 DBD-E2 DBD-ori complex (compare lanes 4 to 6 with lanes 1 to 3).

Using the +3 probe, similar levels of binding were observed using the E1 DBD alone, indicating that the three-nucleotide insertion does not affect E1 binding as expected (compare Fig. 4A, lanes 9 to 11, with Fig. 4A, lanes 1 to 3). Since the sequence of the E1 palindrome is unchanged, the E1 DBD-DNA complex most likely contains two E1 monomers bound to sites 2 and 4, which are the preferred sites of E1 binding. This is consistent with the very low level of binding observed with the B42+3 probe, where the E1 binding site 4 has been mutated (lanes 17 to 19). In the presence of both the E1 and the E2 DBD, we observed two complexes that do not form in the presence of the E2 DBD alone (lanes 11 to 13). The lower complex migrates in the same position as that formed in the presence of the E1 DBD alone and, therefore, is likely to contain two E1 DBD monomers bound to the preferred sites 2 and 4. The upper complex migrates slower and likely corresponds to binding of two E1 DBD monomers together with the E2 DBD. Since this complex is unaffected by the B42 mutation (lanes 20 to 22), E1 is unlikely to be bound to sites 2 and 4, and the ability of this complex to form on the B42+3 probe would be consistent with binding to sites 1 and 3.

A peculiarity is that the mobility of this complex is significantly greater than for the complex formed on the wt ori. We have shown that cooperative binding of the E1 and E2 DBDs induces a sharp bend in the DNA (13). The differences in mobility can be explained by the position of the E1 and E2 binding sites relative to the ends of the probes. The +3 ori probes that were used in this experiment contain less sequence flanking the E2 binding site. Consequently, the sharp bend induced by cooperative binding of E1 and E2 is close to the end of the probe which results in a faster-migrating complex compared to when the bend is closer to the center of the probe, as in the wt probe.

To determine whether E1 was bound to sites 1 and 3 in the complex formed on the +3 ori, we performed DEPC interference analysis on the E1 DBD-E2 DBD-ori complex formed on the B42+3 ori. probe. We used the B42+3 ori in order to simplify the isolation of the complex away from the E1 dimer complex that forms on sites 2 and 4 in the +3 probe. An EMSA was performed using the B42+3 ori probe treated with DEPC, which modifies A and G residues. The complex was extracted from the gel, cleaved with piperidine, and analyzed on a sequencing gel. The results for both strands are shown in Fig. 4B. On both the bottom and the top strands we observed strong interferences over both halves of the E1 palindrome (compare lanes 2 and 3 and lanes 5 and 6). E2 DBD alone gave rise to interferences over the E2 binding site as expected (lane 4 for the top strand and data not shown for the bottom strand).

The interference data is summarized in Fig. 4C. The interference pattern produced by binding of E1 DBD, together with E2 DBD on the B42+3 probe, is virtually identical to that observed for the E1E2-ori complex formed on the wt ori (4) except that the interference pattern on the B42+3 probe is shifted downstream by three nucleotides. The strongest interferences are now located primarily over E1 binding sites 1 and 3 instead of E1 binding sites 2 and 4, a result consistent with E1 binding to these sites. Thus, the DEPC interference analysis provides evidence that the second pair of binding sites (E1 binding sites 1 and 3) are functional E1 binding sites.

E1 recognizes determinants in the major groove at some positions in the binding site.

Our previous DEPC interference assays indicated that E1 recognizes the major groove in DNA, since the main site of modification is the N⁷ position of A and G which protrudes into the major groove (4, 35). To determine whether E1 binds to specific base determinants in the major groove, we performed inosine substitutions within the hexanucleotide E1 recognition sequence as described previously (41). Inosine specifically base pairs with C. The pattern of hydrogen bond acceptors and donors present in an I/C base pair is identical to that of a G/C base pair in the major groove and to that of an A/T base pair in the minor groove. The mutational analysis of the E1 binding site allowed us to compare the effects of I/C base-pair substitutions with those of either G/C or A/T base pairs at each position in the E1 binding site. The Xho ori was used in EMSAs to measure the level of E1 binding in the presence of single inosine substitutions at each purine position in site (AACAAT) on both the top and the bottom strands. The results are summarized in Fig. 5C.

If recognition is in the minor groove an I/C base pair can substitute for an A/T base pair but a G/C base pair can not. If recognition is in the major groove, then an I/C base pair can substitute for a G/C base pair, but an A/T base pair cannot. Replacing the A/T base pair in either of the first two positions of the E1 binding site with an I/C base pair will have minor effects on the E1 binding (Fig. 5A, lanes 11 to 15 and 21 to 25). However, a G/C base pair in those positions also does not significantly impair E1 binding (Fig. 5A, lanes 6 to 10 and 16 to 20). This suggests that at these two positions, determinants other than groove recognition are important for E1 binding. In the same way, at the third and sixth positions of the E1 binding site, the effects of inosine substitutions are ambiguous. When an I/C substitution is made at position 3 (Fig. 5A, lanes 26 to 30), the level of E1 binding is severely reduced, which is also observed when a G/C substitution is made at the same position (Fig. 2). However, an A/T substitution at position 3 also inhibits E1 binding and therefore, whether the inosine substitution on the top strand at position three disrupted major or minor groove contacts cannot be discriminated. Making a C/I base-pair substitution at the third position results in approximately twofold reduction in E1 binding (Fig. 5C and data not shown). Compared to a T/A substitution at this position, where E1 binding is reduced approximately threefold, the C/I substitution at this position appears to allow E1 to bind slightly better. It is therefore difficult to determine unequivocally whether the determinants of E1 binding at this position are within the major or minor groove. Indeed, E1 may recognize both major and minor groove determinants. Because of the lack of severity of transition mutations in this position, it is also possible that DNA structure plays a greater role than base-specific contacts.

An I/C base pair at position six of the E1 binding site results in a level of E1 binding greater than that observed with an A/T base pair, which nearly abolishes E1 binding (Fig. 5B, lanes 31 to 35). This suggests that a major groove element may influence E1 binding. However, a C/I base-pair substitution at the sixth position results in only a slight impairment of E1 binding, with a level of binding intermediate between wt (T/A) and a G/C base pair (Fig. 5C and data not shown). The ambiguity of these results may reflect that, at this sixth position of the E1 binding site, there are determinants for E1 binding that are not simply base-specific structures within the major or minor grooves. Alternatively, both major and minor groove-specific determinants are involved in E1 recognition at this position.

However, in positions 4 and 5, the effects of inosine substitutions are more meaningful. Changing the wt A/T base pair to a G/C base pair in position 4 results in a nearly fivefold decrease in the level of E1 binding (Fig. 2). Binding of the E1 DBD to the site when the A/T base pair is replaced by an I/C base pair is inhibited to the same degree as with a G/C mutation, suggesting that a major groove-specific determinant affects E1 binding (Fig. 5B, compare lanes 6 to 10 and 11 to 15 to lanes 1 to 5). In position 5, the effect of replacing the wt A/T base pair with an I/C base pair resulted in a virtually complete loss of E1 binding (Fig. 5B, lanes 21 to 25). This effect is similar to that produced by a G/C base pair at that position (Fig. 5B, lanes 16 to 20), also suggesting a major groove interaction. Together, these results indicate that in both positions 4 and 5, E1 recognizes determinants in the major groove. Whether E1 also recognizes the major groove at the other positions in the E1 binding site cannot be determined by inosine substitutions.

DISCUSSION

The E1 protein is capable of binding to the ori in various forms. E1 binds as a dimer together with a dimer of E2, forming the E1E2-ori complex, which most likely functions for initial recognition and tethering of E1 to the ori. This is clearly one of the functions of the E1 binding site and also the function that traditionally is associated with sequence-specific DNA binding. Using the E1 DBD, we have determined which sequences E1 recognizes. E1 binds to a recognition sequence consisting of the hexanucleotide sequence AACNAT, with relaxed specificity for transition changes in positions 1 to 3 and with N being any base except G. Sequence recognition is at least partly in the major groove since DEPC modifications in the N⁷ position of either A or G produce strong interference. Inosine substitutions at several positions within the E1 binding site are also consistent with major groove binding, although the inosine substitutions yielded unexpectedly ambiguous results, perhaps due to structural determinants which may be important for E1 binding.

Arrangement of the E1 binding sites.

Previous mutagenesis of the E1 palindrome has shown that E1 binding sites 2 and 4 are required for formation of the full-length E1E2-ori complex (35). Furthermore, the same study indicated that additional sequences which overlap sites 2 and 4 by three nucleotides are required for the formation of the larger E1-ori complex. In the present study, we have confirmed the existence of a second pair of sites, which we refer to as E1 binding sites 1 and 3. The binding of two E1 DBD molecules to these sites was stimulated in the presence of the E2 DBD when the E2 binding site was moved in phase with these sites. Using the information regarding the sequence preference of E1, we have also identified two additional putative E1 binding sites, 5 and 6, which have an affinity for E1 similar to that of binding sites 1 and 3. These results indicate that six E1 binding sites may be present in the E1 palindrome of the BPV ori, although we have so far been unable to demonstrate binding of the E1 DBD to the putative sites 5 or 6 in the context of the wt ori.

DNA binding for tethering and for complex assembly.

As discussed above, binding of E1 to ori seems to serve a tethering role when E1 is bound to sites 2 and 4 in the presence of E2. This tethering appears to be completed after the formation of the E1E2-ori complex, and the binding of additional E1 molecules likely represents a function other than ori recognition. Sequence-specific binding to multiple sites could serve to arrange the E1 molecules in a particular configuration that is necessary for biochemical activity. If this is the case, the arrangement of binding sites in the DNA can be viewed as a blueprint for how larger E1 complexes are assembled. Such a mechanism would prevent formation of these complexes in the absence of the correct DNA sequence and thus contribute to the specificity of assembly. By decoding the arrangement of the binding sites we may be able to deduce and understand the architecture of the E1 complexes.

The different affinities of E1 for the E1 binding sites may promote a stepwise assembly of initiator complexes. Sites 2 and 4 with the consensus binding site sequence AACAAT are the preferred sites for the binding of the first two E1 monomers to the ori through the cooperative interaction with E2. The E1E2-ori complex, having no other apparent replication-related activity, therefore represents the first step in the assembly of a replication-competent initiator complex. The initially loaded E1 dimer can subsequently act as a seed for the binding of additional E1 molecules to weaker E1 binding sites. Thus, formation of the E1E2-ori complex facilitates the formation of the E1-ori complex, which involves binding of two additional molecules to the weaker sites 1 and 3 (30). The low affinity of sites 1 and 3 compared to sites 2 and 4 may require that protein-protein interactions between bound E1 molecules play a greater role in stable complex formation than specific DNA binding. Thus, we conclude that binding of E1 to four overlapping sites results in the formation of the E1-ori complex, which has activity for distortion of the ori.

The most distinctive feature of the arrangement of the E1 binding sites is that the sites are overlapping and that 3 bp are shared between any two overlapping sites. In fact, if E1 were to bind to the six E1 recognition sequences this would represent a rather remarkable close packing; the 24 bp that constitute the E1 binding site would bind six molecules of E1, each of which requires a recognition sequence of 6 bp. Although we do not directly demonstrate here that the E1 DBD can bind simultaneously to overlapping sites, the interference and mutational analysis using full-length E1 (30, 35) can now be interpreted in light of the results presented here and clearly indicates that in the cross-linked E1-ori complex, E1 is bound to the overlapping sites 1 to 4. Thus, the DNA binding results obtained with the E1 DBD are consistent with the results obtained for the full-length E1 protein.

It may seem that the overlapping arrangement of the E1 binding sites presents an obstacle to simultaneous occupancy. However, taking the structure of the DNA into account clarifies how binding is accomplished. Our model for binding of E1 is shown in Fig. 6 and is based on hydroxy radical footprinting (33) and ethylation interference (30), as well as DEPC interference and mutational analysis reported here and elsewhere (4, 35). Hydroxy radical footprinting of the E1E2-ori complex where an E1 dimer is bound to sites 2 and 4 places the E1 protections on one face of the helix. Binding of a second pair of E1 molecules to sites 1 and 3 produces a precise duplication of the dimer footprint representing a shift by 3 bp or approximately one-third of a full turn, indicating that if the first dimer is bound on the top of the helix, the second dimer is bound on the side face of the helix consistent with binding of E1 to sites 1 to 4 (Fig. 6C). Ethylation interference (30) results in a similar conclusion; the interference pattern observed with the E1E2-ori complex is precisely duplicated in the E1-ori complex, a result consistent with a transition from binding of one E1 dimer to binding of two E1 dimers.

FIG. 6.

Arrangement of the E1 binding sites in the BPV core ori. (A) Stereo diagram of the BPV ori DNA containing the E1 binding sites. E1 binding sites 1 to 4 are indicated by arrows. The first three bases on the top and bottom strands for each binding site are colored as shown in panel B. Binding sites 2 and 4 are shown in red, binding sites 1 and 3 are shown in blue, and the putative sites 5 and 6 are shown in green. (B) DNA sequence of the E1 binding region. The positions of binding sites 1 to 4 are indicated by solid arrows, and the positions of the putative sites 5 and 6 are indicated by stippled arrows. The three first bases on the top and bottom strands of each E1 binding site are color coded. E1 binding sites 2 and 4 are indicated in red, and E1 binding sites 1 and 3 are indicated in blue. The putative sites 5 and 6 are shown in green. Above the sequence is shown the positions of ethylation interference (30) with a dimer of E1 (E1E2-ori complex, filled circles) and two dimers of E1 (E1-ori complex, open circles). (C) Schematic diagram of the arrangement of the E1 molecules on the ori with the different E1 binding site arrangements. Binding of E1 to the binding sites 1-4 is indicated by filled symbols. Binding to the putative binding sites 5 and 6 is indicated by open symbols.

The overlap between the binding sites can be accommodated based on the repeated sequence of the site. When projected onto a B-DNA helix the three first bases of the site on the top strand and the last three bases on the bottom strand are positioned on opposite sides of the major groove, as indicated in the stereo drawing in Fig. 6A. Thus, if the first three bases of the site are recognized on the top strand and the second three are recognized on the bottom strand, an overlap can be accomplished since an overlapping site could utilize the complement on the other strand for recognition. For site 2, for example, 5′-AAC-3′ would be recognized on the top strand and 5′-ATT-3′ would be recognized on the bottom strand (shown in red in Fig. 6A and B). The overlapping site 1, shown in blue, is positioned on a different face of the helix, with no obvious steric clashes, although part of E1 binding site 1 is base paired with part of binding site 2 (Fig. 6A). This mode of binding is consistent with the positions of ethylation interference on the phosphate backbone, which occur on both sides of the major groove flanking the recognition sequence as indicated in Fig. 6B. This model is also consistent with the positions of DEPC interference.

An interesting question is how E1 complexes that are larger than the two dimers can be assembled. We have previously observed that after glutaraldehyde cross-linking, full-length E1 becomes topologically linked to a plasmid carrying the viral ori but can be released after linearization of the DNA, indicating that E1 encircles the DNA. We also showed that under these conditions a cross-linked oligomer corresponding to a trimer of E1 could be detected by Western blot analysis (33). Using the information that we have regarding binding of the two E1 dimers, there are two obvious ways that trimers could be generated from the two dimers. One is that two additional E1 molecules bind to the putative sites 5 and 6 (see Fig. 6C). This would generate a symmetrical dimer of trimers wherein each trimer encircles the DNA. The putative sites 5 and 6 have similar affinities for E1 compared to sites 1 and 3; however, sites 5 and 6 differ in that they are not paired like the other four sites and therefore cannot benefit from the increase in affinity that dimerization confers on binding of E1 (4). Other arrangements are also possible; for example, a third dimer could be bound on the third, unoccupied face of the DNA helix. Binding of E1 to the putative site 6 would place a dimer partner in the center of the E1 palindrome. The sequence here is GGTAAC which, although it overlaps with both sites 2 and 3, is available for binding (Fig. 6). This sequence, however, is a very poor E1 binding site (data not shown). Thus, with either model, factors other than DNA contacts are likely to be required for higher-order complex formation.

From the scenario discussed above, it can be envisioned how the existence of multiple binding sites for E1, with different affinities, could facilitate the progression from the E1E2-ori complex to the formation of an E1 hexamer. Simultaneous binding to all sites would allow six molecules of E1 to assemble on the ori. How such a complex would be related to an active helicase is unclear. However, an interesting possibility is that each of the two trimers eventually give rise to an E1 hexamer. It has been demonstrated that for T antigen, preformed hexamers are incapable of forming double hexamers on the ori, whereas monomeric ATP-bound T antigen is the preferred substrate for double-hexamer formation (6, 36). This suggests that, like E1, the T-antigen hexamer most likely arises from the sequential binding of monomeric T-antigen molecules. Four of the six binding sites in the E1 palindrome are arranged in a fashion similar to that of the T-antigen binding sites in the SV40 ori, indicating that a similar assembly process is utilized for T antigen. The stepwise assembly of a helicase through the successive binding of monomeric proteins, as inferred from our data, would allow multiple points which could be regulated during initiation of DNA replication.

Comparison to other papillomaviruses.

Although a general sequence similarity is observed between the ori sequences that are believed to bind E1 in different papillomaviruses, it has been difficult to assess exactly how similar these E1 binding regions are. Utilizing the binding site sequence that we have identified and allowing variations as revealed by the mutational analysis (see Fig. 2), it is clear that papillomaviruses in general appear to have the same organization of E1 binding sites. However, the number of recognizable high-affinity sites varies, indicating either that the E1 proteins from different papillomaviruses have differences in sequence recognition or that contribution from other factors such as protein-protein interactions vary in different viral types. The overall high degree of homology in the E1 DBDs and especially in regions involved in DNA recognition argues that differences in sequence recognition are probably subtle (9, 14).

As shown in Fig. 7, certain papillomaviruses such as HPV-1 appear to have a larger number of high-affinity sites than does BPV. Furthermore, the presence of high-affinity sites in positions corresponding to sites 5 and 6 in BPV lends some credence to the functionality of the putative sites 5 and 6 in the BPV ori. It is interesting that the HPV-1 ori is unique in that transient DNA replication in vivo can be achieved with overexpression of E1 in the absence of E2, a finding that may be indicative of higher-affinity E1 binding (15). A further interesting observation is that some HPVs, such as HPV-31 and HPV-32, have three high-affinity sites in one-half of the palindrome, while the sites in the other half are low-affinity sites. This may indicate that the E1 dimer interaction contributes more to complex formation in these viruses than it does for BPV E1 binding.

FIG. 7.

Comparison of the E1 binding site arrangement in select papillomavirus origins of replication. The stippled arrows indicate putative E1 binding sites with significantly lower affinity based on the lack of conservation at positions that are important for binding of BPV E1.