A Critical DnaA Box Directs the Cooperative Binding of theEscherichia coli DnaA Protein to the Plasmid RK2 Replication Origin*

The requirement of DnaA protein binding for plasmid RK2 replication initiation the Escherichia coli was investigated by constructing mutations in the plasmid replication origin that scrambled or deleted each of the four upstream DnaA boxes. Altered origins were analyzed for replication activity in vivo and in vitro and for binding to the E. coli DnaA protein using a gel mobility shift assay and DNase I footprinting. Most strikingly, a mutation in one of the boxes, box 4, abolished replication activity and eliminated stable DnaA protein binding to all four boxes. Unlike DnaA binding to the E. coli origin, oriC, DnaA binding to two of the boxes (boxes 4 and 3) in the RK2 origin, oriV, is cooperative with box 4 acting as the “organizer” for the formation of the DnaA-oriV nucleoprotein complex. Interestingly, the inversion of box 4 also abolished replication activity, but did not result in a loss of binding to the other boxes. However, DnaA binding to this mutant origin was no longer cooperative. These results demonstrate that the sequence, position, and orientation of box 4 are crucial for cooperative DnaA binding and the formation of a nucleoprotein structure that is functional for the initiation of replication.

Whereas the interactions of DnaA at oriC have been studied in considerable detail (for a review, see Ref. 2), much less is known about the formation of DnaA-nucleoprotein structures at the replication origins of plasmid replicons. Unlike the more disperse arrangement of DnaA boxes at oriC and replication origins of other bacteria, the DnaA boxes at plasmid replication origins in general are closely associated forming one or more clusters. In addition, the DnaA protein must work in concert with a plasmid-encoded initiation protein (for reviews, see Refs. 18 and 19)). Finally, unlike the case with oriC, the ADP form of DnaA protein is sufficient for replication in vitro of plasmids R1 (15) and P1 (20) and for origin opening of F (21), RK2 (11), and R6K (22). These observations likely reflect fundamental differences in the structure of the nucleoprotein complexes formed by the DnaA protein at the replication origin of plasmids and the bacterial chromosome.
The broad host range plasmid RK2 exhibits the remarkable ability to replicate and maintain itself in a wide variety of Gram-negative bacteria (23) and, thus, is a particularly interesting system for studying fundamental questions about the interactions between plasmid-and host-encoded factors at a replication origin. The DnaA protein has been shown to be required for RK2 replication (24,25), but its exact role has been less well studied. Only recently, it was demonstrated that there are two DnaA boxes in addition to two previously described boxes located in the leftward part of the RK2 origin, oriV, and that all four boxes bind the E. coli DnaA protein (11). The four DnaA boxes arranged as two pairs in an indirect orientation can potentially form a cruciform structure (see Fig. 1). Boxes 2, 3, and 4 contain one mismatch from the stringent DnaA box consensus sequence, whereas box 1 contains two mismatches.
Replication initiation of RK2 requires both DnaA and the plasmid-encoded initiation protein, TrfA. The TrfA protein binds to the iteron sequences in oriV (10, 26 -28) and this binding results in the melting of all four 13-mers in the AϩTrich region in the presence of the DnaA and HU proteins (11). The DnaA protein also has the indispensable role of recruiting the DnaB-DnaC complex to the RK2 origin through a proteinprotein (DnaA-DnaB) interaction (29). The loading and activation of the DnaB helicase at the open region requires both the DnaA and TrfA proteins (29) and the correct positioning of the DnaA boxes on the helical axis with respect to the rest of the origin (30). How the DnaA protein and the TrfA protein cooperate to achieve DnaB loading in not known.
Do the requirements for DnaA protein binding and function in plasmid systems differ from that of bacterial chromosomes, specifically oriC? In this paper, we present the molecular and biochemical characterization of the structural requirements for functional binding of DnaA protein during plasmid replication initiation, specifically plasmid RK2. Our results demonstrate that one of the four DnaA boxes, box 4, directs the cooperative binding of DnaA to the other DnaA boxes, essentially acting as the "organizer" in the formation of a functional DnaA-oriV nucleoprotein structure. These features of DnaA protein bind-ing to a plasmid replication origin may be important in understanding the differences between plasmid and chromosomal replication initiation and, in the case of RK2, its broad host range properties.
DNase I Footprinting-DNA probes were prepared as described (11) with [␣-32 P]dATP (6000 Ci/mmol) and the Klenow fragment of E. coli DNA polymerase (NEN Life Science Products). Unincorporated ATP was removed using nucleotide removal spin columns (Qiagen). Binding of DnaA protein to origin DNA containing mutated DnaA box sequences was performed essentially as described previously (11), except that DnaA protein was incubated with DNA at 37°C for 15 min and DNase I (New England Biolabs, Inc.), 0.67 units/reaction, at 37°C for 30 s.
For footprinting of specific nucleoprotein complexes, reactions were assembled as for gel mobility shift assays described below except that the amount of probe used was increased. After incubating the reaction containing labeled DNA template and DnaA protein for 15 min at 37°C, DNase I was added (6.7 units/reaction) for 2 min at the same temperature. Reactions were then run on a 5% nondenaturing polyacrylamide gel. After electrophoresis, specific complexes were identified and isolated from the gel and eluted overnight at 30°C in 1 M LiCl 2 , 20 mM Tris-HCl, pH 7.6, 0.2% SDS, 0.2 mM EDTA. Eluant was collected over a spin column (Bio-Rad), extracted with phenol-chloroform, ethanol-precipitated, and resuspended in loading buffer.

Requirement of DnaA Boxes for oriV Function-
To address the importance of each DnaA box present in the RK2 origin ( Fig. 1), individual box sequences were replaced with the scramble sequence 5Ј-GATATCCTG, which is similar to a sequence shown previously not to bind DnaA protein (35). Thus, mutants oriV A1, A2, A3, and A4 contain the scramble sequence (5Ј-GATATCCTG) in place of the wild-type DnaA box sequence for boxes 1, 2, 3 and 4, respectively (see Fig. 1). Mutated origins were also constructed that contained deletions of two, three, or all four DnaA boxes. Mutagenesis was carried out using polymerase chain reaction (see under "Experimental Procedures") and the resulting origins were subcloned in plasmid pKD19L1 (30). All mutations were confirmed by sequence analysis (data not shown). Plasmid pKD19L1 is a dual origin (RK2-R6K␥) plasmid that replicates by the R6K␥ origin in E. coli strain CC118 (pir) because this strain encodes the R6K initiator protein. In the parental strain, CC118, only the RK2 origin may be utilized. Thus, the ratio of transformation frequencies obtained in CC118 versus CC118(pir) was used to quantitate oriV function in vivo. Ratios obtained for altered origin constructs were normalized to those obtained for pKD19L1 (wild-type oriV) ( Table I). Origins that contained a scrambled sequence in DnaA box 1 (oriV A1) or box 2 (oriV A2) or a deletion of both boxes 1 and 2 (oriV A⌬1-2) did not negatively effect origin function and actually resulted in a somewhat higher level of plasmid transformation frequencies. However, a scrambled sequence in DnaA box 3 (oriV A3) or box 4 (oriV A4) reduced or completely abolished plasmid transformation, respectively. Similarly, an origin containing a deletion of all four DnaA boxes (oriV A⌬1-4) rendered the plasmid inactive. An origin containing only DnaA box 4 (oriV A⌬1-3) was only 74% active compared with wild-type (Table I). The oriV A⌬1-3 mutant is highly variable; in another set of experiments, replication activity was as low as 32% of the activity of a wild-type origin (data not shown). Additionally, the transformants that were obtained with the oriV A⌬1-3 mutant were lost fairly rapidly compared with the wild-type plasmid in the absence of antibiotic selection. 1 Thus, whereas both DnaA boxes 3 and 4 are required for replication activity in vivo, the requirement for DnaA box 4 is the more stringent.
The mutated origins were also analyzed for replication activity in vitro using an in vitro replication assay for RK2 that has been described previously (33). This assay utilizes a 40% (NH 4 ) 2 SO 4 fraction from a soluble E. coli extract to replicate supercoiled DNA containing an RK2 origin in the presence of the initiator protein, TrfA. Fig. 2 shows DNA synthesis for wild-type and altered origin templates. In vitro replication activity with the scrambled DnaA box origin templates was relatively similar to the activity that was observed in vivo ( Fig.  2A). Scramble mutations in DnaA boxes 1 and 2 had no inhibitory effect on DNA synthesis, whereas mutations in DnaA boxes 3 and 4 resulted in a lower level of replication activity, with mutant oriV A4 displaying only 50% of wild-type levels. Replication activity of origin templates with deletion mutations (Fig. 2B) again showed that DnaA boxes 1and 2 can be deleted with out affecting replication levels, whereas deleting all four DnaA boxes abolished DNA synthesis. Replication of an origin template with only DnaA box 4 (oriV A⌬1-3) was slightly less than wild-type except at the highest amount of TrfA protein.
Box 4 Is Essential for DnaA Binding to Other Boxes at the Replication Origin-We analyzed the binding of DnaA protein to oriV fragments containing the DnaA box scramble mutations using DNase I footprinting and a gel mobility shift assay. The DNase I pattern of DnaA protection for wild-type oriV (Fig. 3,  lanes 1-3) was as has been previously shown for both linear and supercoiled oriV templates (11). When a DnaA box contained the scrambled sequence, full DNase I protection was not observed in that box. Thus, the replacement of the wild-type sequence with the scramble sequence had altered DnaA binding to the respective box. Interestingly, the scramble mutation in DnaA box 4 (oriV A4) greatly reduced the amount of protection observed for box 3 (Fig. 3, lanes 13-15). We further analyzed DnaA binding to mutants oriV A3 and oriV A4 using a gel mobility shift assay (Fig. 4). As expected for wild-type oriV, increasing DnaA concentration led to four specific nucleoprotein complexes (Fig. 4, lanes 1-5); however, on the basis of the relative intensity of the various retarded bands, their formation did not proceed sequentially. Even with smaller amounts of DnaA protein, the appearance of complex I never occurred in the absence of complex II, and usually there was much less of complex I (data not shown and Ref. 11). Thus, binding of DnaA protein to linear oriV DNA appears to be more stable when in the form of complex II, suggesting the possibility of cooperative binding by DnaA. DnaA binding to scramble mutants oriV A3 and A4 support this idea. When DnaA box 3 was scrambled, only complex II and, to a lesser extent, complex I was seen (Fig.  4, lanes 6 -10). When DnaA box 4 was scrambled, no nucleoprotein complexes were observed (Fig. 4, lanes 11-15). This is consistent with the lack of replication activity observed for mutant oriV A4 in vivo (Table I) and the substantially reduced activity in vitro ( Fig. 2A), suggesting that the function of DnaA box 4 is to initiate DnaA binding to oriV.
DnaA Binds Initially to Boxes 4 and 3-We further analyzed DnaA binding by carrying out footprint analysis on specific DnaA-oriV nucleoprotein complexes. DnaA was incubated with wild-type oriV probe, treated with DNaseI, and run on a native gel in order to isolate specific complexes. DnaA protein concentrations were adjusted in order to obtain only complexes I and II (see Fig. 4). DnaA-oriV complexes I and II were isolated, gel purified, and run on a denaturing gel in order to visualize  DNaseI protection patterns (Fig. 5). The analysis of nucleoprotein complexes I and II showed that both DnaA boxes 4 and 3 were protected from the DNaseI treatment. These results demonstrate that upon the binding of DnaA to oriV, the first boxes to be filled are boxes 4 and 3.
The fact that the protection pattern for complex I (which may be expected to represent DnaA binding to a single DnaA box) shows that both boxes 4 and 3 are bound by DnaA (Fig. 5, middle lane) suggests that the binding by DnaA to one box is immediately followed by the binding to a second box. We should note that similar results were obtained at even lower DnaA protein concentrations (data not shown). This supports the idea that DnaA binding to either box 4 or 3 alone is unstable compared with the binding of both boxes. Because the treatment with DNase I is done prior to separation by gel electrophoresis, any subsequent disassociation of DnaA protein from either box 4 or 3 would result in the isolation of complex I, which contains a footprint of both boxes 4 and 3. We investigated whether the appearance of complex I was simply a result of the disassociation of DnaA protein from complex II in solution or during gel electrophoresis by looking at the rate of complex I formation; however, neither the incubation time nor the amount of gel running time had any effect on the formation of complex I (data not shown). It is possible that complex I contains DnaA bound to both boxes 4 and 3 but is structurally different than complex II and thus exhibits an altered mobility during gel electrophoresis.
Cooperative DnaA Binding Is Dependent on the Position and Orientation of Box 4 -Because DnaA box 4 appears to be critical in initiating binding to the other DnaA boxes, we investigated whether it is the actual DnaA box sequence and/or the position of box 4 that is crucial. Mutations were constructed as described above and analyzed for in vivo activity (Table II) and ability to bind DnaA protein using the gel mobility shift assay (Fig. 6). When the sequence of box 4 was replaced with the sequence of box 1 (oriV 4-1), in vivo activity was abolished and DnaA binding was greatly reduced (Fig. 6A). The results were the same if box 4 sequence was reintroduced into this mutant construct in the box 3 position (oriV 3-4, see Table II and Fig.  6B). If box 4 was flipped (oriV 4-Flip) so that it faced away from the other boxes, the effect was even more dramatic in that complexes were not formed at all with DnaA protein (Fig. 6C), and the origin was completely inactive in vivo (Table II). Because the sequences surrounding DnaA boxes in oriC have been shown to effect DnaA binding (36), it was possible that DnaA could not bind to box 4 in the oriV 4-Flip mutant because the adjoining sequences were not also placed in the flipped position. Therefore, a mutant was constructed that flipped box 4 and the adjoining sequences (3 base pairs on either side of the box, oriV 4-FlipX). This construct was also inactive in vivo (Table II), but nucleoprotein complexes were observed with DnaA protein (Fig. 6D), suggesting that now DnaA could bind to box 4 as well as the other boxes. However, the binding of DnaA to the boxes in this mutant template proceeded in a stepwise, noncooperative fashion, which is not what is observed for a wild-type oriV template (see Fig. 4, lanes 1-5). Thus, the orientation of DnaA box 4 and its adjoining sequences, relative to box 3, is critical for cooperative binding and it is this binding that is required for replication activity of oriV. DISCUSSION We have investigated the binding of the DnaA protein to the RK2 replication origin and the DnaA box requirement for plasmid RK2 replication. Our results clearly showed that in E. coli, boxes 3 and 4 are essential for optimal replication activity, whereas boxes 1 and 2 are dispensable. The DnaA-oriV interaction is dependent upon one critical box, box 4, as mutations in box 4 abolished stable DnaA binding to all four boxes. DNaseI footprinting of specific DnaA-oriV complexes demonstrated that the first boxes to be filled are boxes 4 and 3. These results, along with the nonlinear progression of DnaA-oriV complex formation observed in the gel mobility shift assay, indicate that DnaA binds cooperatively to the boxes at the RK2 origin. This is in contrast to what has been observed at oriC, where several approaches failed to detect cooperative DnaA binding (37,38). DnaA was shown to bind preferentially to box R4, but the binding preference of DnaA to the other sites was not altered when box R4 was mutated (38).
The DNase I footprinting experiments performed here do not reveal which box in oriV, 4 or 3, actually binds DnaA first. We cannot exclude the possibility that DnaA binds initially to either box 4 or box 3 without a preference. This could account for the observed footprint by DnaA in complex I (Fig. 5, middle lane), which indicates a mixture of both boxes. However, we propose that because a mutation in box 4 abolishes replication activity and the stable binding of DnaA to the other boxes, DnaA probably binds first to box 4 and then cooperatively, possibly via a conformational change in the DNA, to box 3. What is clear is that a fully functional origin requires the formation of a complex II that contains DnaA protein bound to both boxes 4 and 3.
Further analysis of DnaA box 4 argues that there are specific requirements for both its sequence and position for oriV functionality. If box 4 is replaced with the sequence of box 1 (mutant oriV 4-1) the origin is inactive (Table II) and DnaA binding is greatly reduced as judged by gel retardation analysis (Fig.  6A). The sequence of box 1 has less homology to the DnaA box consensus than box 4, and in footprinting experiments more DnaA protein is required to see protection in box 1 than in box 4 ( Fig. 3 and data not shown). Thus, the data suggest that box 4 has a higher affinity for DnaA or forms a more stable complex with DnaA than the other DnaA boxes and, in this capacity, acts to recruit DnaA protein to oriV. However, the critical role of box 4 is also dependent upon its relative position within the origin. In mutant oriV 3-4, the box 4 sequence replaces DnaA FIG. 5. DNase I footprinting of specific nucleoprotein complexes. Linear template (2.5 ng) containing the wild-type oriV region was incubated with 31 ng of DnaA protein followed by DNase I treatment as described under "Experimental Procedures." Complexes I and II were separated by gel electrophoresis, isolated from the acrylamide gel, and analyzed on a 10% denaturing gel. box 3 of the oriV 4-1 mutant (see Table II); this mutant was also inactive and displayed reduced DnaA binding, arguing for the importance of the location of box 4. It should be emphasized that the gel mobility shift assay used in these experiments is a measure of both protein affinity to the individual boxes and the stability of the nucleoprotein complexes. It is, therefore, possible that DnaA protein is still able to bind to the boxes in these mutant templates but requires contact with box 4 in its "correct" position to form stable complexes observable during gel electrophoresis.
It has been shown for oriC that the preference for DnaA box R4 depended on it being in the correct orientation (38). We found for oriV that if box 4 was inverted (mutant oriV 4-Flip), DnaA protein did not form stable complexes with any of the DnaA boxes unless adjoining sequences were also inverted (mutant oriV 4-FlipX) (Fig. 6). Thus, with the oriV 4-FlipX mutant, DnaA did bind to all four boxes, but this binding no longer exhibited cooperativity. Possibly, in this case the binding of DnaA to the high affinity but inverted box 4 increased the local concentration of DnaA protein and, thus, enhanced the probability of other DnaA-DnaA box contacts. Interestingly, the oriV 4-FlipX mutant was not active in vivo. Presumably when DnaA is bound to the inverted box 4, it is now on the opposite face of the DNA helix with respect to box 3, thus disrupting the structure and/or critical protein contacts required for cooperative binding. These results demonstrate that the correct orientation of box 4 is crucial for directing not only the cooperative binding of DnaA but the formation of a nucleoprotein structure that is required for a functional initiation complex. More than likely, this specific structure is critical for the loading of DnaB helicase. We have previously demonstrated that when all four DnaA boxes are on the opposite side of the DNA helix with respect to the rest of the origin, DnaB helicase activity at oriV is reduced even though the origin is fully open (30). Unlike replication initiation at oriC, where origin opening and helicase loading occur as the result of a single initiator protein (DnaA), which is responsible for both steps, initiation at a plasmid origin usually requires two initiator proteins (DnaA and a plasmid-encoded replication initiation protein) and, therefore, may involve a fundamentally different nucleoprotein structure for facilitating each step of replication initiation.
Most plasmid replication origins contain more than one DnaA box or cluster of DnaA boxes. In the case of the narrow host range plasmids P1 (39) and R6K (40), one box is sufficient for a functional origin. Evidence for cooperative DnaA binding to the DnaA boxes at the origin of pSC101 has been demonstrated to occur at a distance (DNA looping) with the combined action of the RepA initiator protein and IHF (41). However, for RK2, the cooperative binding exhibited by the DnaA protein occurs at boxes that are separated by only three to five base pairs and is not dependent on the binding of additional proteins, including the initiator protein, TrfA. Recently it was shown for plasmid R6K that the DnaA protein physically interacts with the R6K plasmid specific initiator protein and that this interaction is required for origin opening (22). There is no evidence for an interaction between DnaA and the RK2 initiator protein, TrfA. Preliminary results using ELISA to detect protein-protein interactions suggest that if an interaction does occur, it is a relatively weak interaction. 2 An interaction between the DnaA and TrfA proteins clearly is not required for binding of the two proteins to their respective sites, DnaA boxes or iteron sequences (11), or for the cooperative DnaA-oriV interaction presented here. It is possible that these differences in molecular interactions at the replication origin observed between the narrow host range plasmids, P1 and R6K, and RK2 at least partially account for their different host range.
It was surprising to us that DnaA boxes 1 and 2 can be deleted without affecting the efficiency of transformation of E. coli (Table I) or in vitro replication activity (Fig. 2), especially 2 K. S. Doran and D. R. Helinski, unpublished observations. because the four boxes potentially can form a cruciform structure (11). However, in view of the broad host range replicating activity of RK2, the experiments performed here with E. coli do not rule out the possibility that a cruciform structure and/or all four DnaA boxes are required in bacterial hosts other than E. coli for optimal replication. Prior to the discovery of all four DnaA boxes, mutagenesis studies of the RK2 origin region demonstrated a differential requirement of DnaA boxes between E. coli and P. aeruginosa (42)(43)(44). Thus, it is of interest to test the DnaA box mutants constructed in this study in other bacterial hosts, as different DnaA and other host replication proteins may act differently at the RK2 origin.
The finding that boxes 4 and 3 are sufficient for oriV activity indicates that the DnaA-oriV complex is a much more compact structure than is found for oriC, where the nucleoprotein complex extends over approximately 200 base pairs. Do the structural differences reflect the differences in DnaA function for the two replicons? In the case of oriC, the DnaA protein does not bind cooperatively to the five boxes, yet it directs the wrapping of DNA around a core of DnaA protein monomers and is subsequently responsible for both the opening of the 13-mer region and the recruitment of the DnaB-DnaC complex. For oriV, the DnaA protein plays a more limited role in that although it recruits the DnaB-DnaC complex to the origin region, it cannot by itself form an open complex. Our results have further demonstrated through mutational analysis and protein binding experiments that one box, box 4, effectively acts as the organizer for DnaA cooperative binding at oriV. This binding then initiates a nucleoprotein structure that is functional for the formation of a prepriming complex. This may represent a novel type of DnaA binding at a prokaryotic origin. It will be of interest to determine whether this type of DnaA interaction is found for other plasmid systems and whether it is a critical factor for the broad host range properties of the RK2 plasmid.