IHF-dependent activation of P1 plasmid origin by DnaA


  • Present addresses: Ambion Inc., 2130 Woodward, Austin, TX 78744, USA; Invitrogen, 1610 Faraday Avenue, Carlsbad, CA 92008, USA.

*E-mail chattoraj@nih.gov; Tel. (+1) 301 496 9194; Fax (+1) 301 480 1493.


In bacteria, many DNA–protein interactions that initiate transcription, replication and recombination require the mediation of DNA architectural proteins such as IHF and HU. For replication initiation, plasmid P1 requires three origin binding proteins: the architectural protein HU, a plasmid-specific initiator, RepA, and the Escherichia coli chromosomal initiator, DnaA. The two initiators bind in the origin of replication to multiple sites, called iterons and DnaA boxes respectively. We show here that all five known DnaA boxes can be deleted from the plasmid origin provided the origin is extended by about 120 bp. The additional DNA provides an IHF site and most likely a weak DnaA binding site, because replacing the putative site with an authentic DnaA box enhanced plasmid replication in an IHF-dependent manner. IHF most likely brings about interactions between distally bound DnaA and RepA by bending the intervening DNA. The role of IHF in activating P1 origin by allowing DnaA binding to a weak site is reminiscent of the role the protein plays in initiating the host chromosomal replication.


Transactions on DNA connected with fundamental life processes such as replication, transcription and recombination involve opening the strands of duplex DNA. This requires formation of nucleoprotein complexes involving multiple DNA–protein and protein–protein interactions. The complex formation apparently puts enough tension on the DNA backbone that strand-opening of nearby DNA follows spontaneously. Our knowledge is not clear on how the complexes organize in space and how they provide the energy for strand-opening. At present, we are limited to knowing the DNA sites and proteins, and mostly their binary interactions. The players generally include at least one DNA binding protein that provides site-specificity to the reaction, and a member from a group of small DNA binding proteins, called DNA architectural proteins. They either bend DNA, best exemplified by IHF, or prefer to bind to distorted DNA and thereby stabilize the structure, as is the case with HU and HNS (Swinger and Rice, 2004).

IHF was discovered as a factor essential for site-specific integration of phage λ to its chromosomal attachment site. The protein was later found to bend DNA strongly and how the bending could help in putting together a higher order nucleoprotein complex has been inferred from DNA footprinting and crystallographic studies (Radman-Livaja et al., 2006). Since its discovery, many diverse reactions showed a requirement for IHF (Nash, 1996). IHF binds DNA in a sequence-specific fashion. However, in some cases, IHF can be substituted with HU protein, which binds DNA without apparent sequence specificity. Strand-opening in the Escherichia coli origin of replication (oriC) is one such case that can be achieved with DnaA and either IHF or HU protein (Skarstad et al., 1990; Hwang and Kornberg, 1992). How a non-specific binding protein, like HU, can substitute for the sequence-specific binding of IHF is not clear in oriC, but studies in transposition of phage Mu and repression of gal operon suggest that HU can also bind site-specifically due to cooperation from other proteins that participate in those reactions (Lavoie and Chaconas, 1993; Aki and Adhya, 1997). The specificity apparently comes not from DNA sequence but from a distorted structure of the HU binding site, imposed by the other proteins of the reaction.

Strand-opening at oriC also requires that DnaA be bound to ATP. More recently, the requirement for the ATP-bound form of DnaA has been made clearer by discovery of three weak (low-affinity) sites that recognize DnaA only in the ATP-bound form (Leonard and Grimwade, 2005). The binding to these weak sites increases in the presence of IHF, hence they were called I-sites. It appears that the rate-limiting step for initiation in oriC is the occupation of I-sites. IHF thus plays a facilitatory role in oriC function. Indeed, initiation in the absence of IHF happens at a higher than the normal amount of DnaA (Von Freiesleben et al., 2000).

The DnaA protein and IHF or HU are also required for replication of several plasmids, particularly the ones belonging to the iteron family, such as pSC101, R6K, P1 and F (Hansen and Yarmolinsky, 1986; Kline et al., 1986; Filutowicz and Appelt, 1988; Manen and Caro, 1991). The defining feature of these plasmids is the presence of iterated initiator binding sites, called iterons, in the origin of replication. Iterons are specific for plasmid-encoded initiators (called RepA in P1) and the initiator–iteron interactions are essential not only for initiation but also for controlling the frequency of initiation. The origin of these plasmids also contain at least one strong DnaA binding site (a DnaA box) but the presence of multiple DnaA boxes is also common (Doran et al., 1999). The three proteins allow opening of the strands of the origin but also contribute to subsequent steps of initiation such as helicase loading (Kawasaki et al., 1996; Konieczny and Helinski, 1997; Park et al., 1998; Zzaman and Bastia, 2005).

The binding sites of initiators to the origin must be precisely ordered to form a functional initiation complex as changing their relative positions often inactivates the origin. This is particularly striking in oriC where even a single bp insertion or deletion is not tolerated (Woelker and Messer, 1993; Hsu et al., 1994). In view of these results we were particularly intrigued by the finding in plasmid P1 that a DnaA box could be moved from one end of the origin to the other without loosing function (Abeles et al., 1990). However, moving the box a few bp at either end did compromise the origin function (Park and Chattoraj, 2001). These results suggested that, as in oriC, the precise context of DnaA binding in the origin is critical, and the role of the boxes is not merely to provide loading sites of the protein to increase its local concentration.

There are two tandem DnaA boxes in the left end of P1 origin and three more on the right end (Fig. 1A). However, one consensus box suffices at either end (Abeles et al., 1990). Somewhat unexpectedly we found that when the right boxes were the only boxes, DNA further downstream was required (Park and Chattoraj, 2001). Here we have defined the extent of that additional DNA. There is an IHF site within the sequence, and the IHF protein becomes essential for replication when DnaA boxes were present only on the right. As the replication also required about 50 more bp downstream of the IHF site, it is likely that the downstream sequences are allowing binding of a protein that participates in the initiation reaction in an IHF-dependent manner. Our results implicate DnaA to be such a protein. From the requirement of IHF for replication and lack of any conspicuous DnaA boxes there, we suggest that there is at least one I-site downstream of the IHF site and, in that case, this study extends the functioning of weak (incognito) DnaA binding sites beyond oriC.

Figure 1.

Requirement of IHF and HU proteins for functioning of P1 origin with DnaA boxes only on the right. miniP1 plasmids are shown diagrammatically on the left and their efficiency to transform hosts is given by number of colonies on transformation plates obtained using ∼100 ng DNA.
A. The wild-type (WT) strain was DH5(λDKC266), where the prophage supplied ∼1X RepA constitutively. infA is BR4543(λDKC266), which is inactivated for the ihfA gene. pinfA+ (= pHNαβ) was used to complement the IHFα deficiency of the infA strain. minipSC101 (= pMLO42) was used here as a control for an IHF-dependent plasmid.
B. The host strains were N99hupA16(λDKC266) and N99hupA16 hupB11(λDKC266), both carrying pKP154, where hupA+ gene was present under arabinose control. miniP1 plasmids pKP151 and pKP152 are same as pDKC412 and pDKC418 except that the drug marker was changed from Cm to Ap. This was necessitated because a Cm cassette was already used to inactivate hupB. pALA96 is used here as a hupAhupB-independent control plasmid. The numbers are means of three different experiments; the individual numbers varied within a twofold range.


Identification of an IHF site on the right side of P1 origin

Previously we showed that when the left two DnaA boxes in the P1 plasmid origin were deleted, the right DnaA boxes were only able to promote replication when additional sequences further downstream were present (Park and Chattoraj, 2001). Analysis of this sequence revealed the presence of a putative IHF binding site (Fig. 1A). A plasmid containing only the right DnaA boxes transformed an IHF mutant host poorly (pDKC418, Fig. 1A). The efficiency of transformation approached that of pSC101, a well-characterized IHF requiring replicon (Manen and Caro, 1991). When the left DnaA boxes were present, replication was not dependent on IHF, as was found earlier (pDKC412 and pDKC416, Fig. 1A; Funnell, 1988).

Whereas oriC can get by with either IHF or HU, replication of plasmids P1 and F depends specifically on HU for efficient origin opening (Kawasaki et al., 1996; Park et al., 1998). HU is normally a heterodimer but in many cases homodimerization of either of the subunits, encoded by genes hupA or hupB, suffices (Nash, 1996). The IHF dependence of the miniP1 plasmid pDKC418 raised the question whether it still required HU. pDKC418 replicated when only hupA was deleted from the chromosome (Fig. 1B). However, deletion of both hupA and hupB was not permissive for pDKC418 replication unless expression of hupA from a complementing plasmid was induced. IHF therefore cannot substitute for HU, and the two proteins most likely perform separate functions in replication of pDKC418.

To show that the IHF requirement was direct through binding to a site on the right side of the P1 origin, electrophoretic mobility shift assay (EMSA) was performed. A 243 bp fragment of the P1 origin was used that contained sequences downstream of the iterons including the putative IHF site. Purified IHF bound to this DNA fragment at nM concentrations, as was found earlier for sequence specific binding of the protein (Hoover et al., 1990; Grimwade et al., 2000; Sze et al., 2001) (Fig. 2). These results indicate that IHF functions by binding to the right side of the P1 origin. As a single retarded band was seen, we can conclude that in the region of interest there is no second binding site for IHF.

Figure 2.

EMSA of IHF binding to DNA that becomes essential for P1ori function in the absence of left DnaA boxes. Concentrations were 2 nM for the pUC19 fragment and 1 nM for the P1ori fragment.

Models of IHF function

We entertained three models that have been proposed by others for the functioning of IHF. First, the binding of IHF to DNA itself facilitates replication. It has been proposed that the binding could alter DNA structure and the structural perturbation could propagate through the DNA to help open-complex formation in transcription (Sheridan et al., 1998; Sze et al., 2001). In the present context the structural perturbation could facilitate DnaA binding to the DnaA boxes or even propagate to the A + T rich region of the origin to help its opening. A prediction of this DNA structural transmission model is that activation should not be sensitive to the phase of the helix to which IHF binds. Second, IHF by bending DNA allows the downstream DNA to lay on top of DnaA and/or RepA of the initiation complex in a manner similar to DNA interactions with the ‘back’ of RNA polymerase (Engelhorn and Geiselmann, 1998). This protein–DNA interaction model predicts that the phasing between the IHF site and DnaA boxes and/or iterons should be critical but the sequences downstream of the IHF site need not be specific. Transcriptional activation studies have implicated, however, that the sequences could also be specific (UP-element; Giladi et al., 1998). Third, in the sequence downstream of the IHF site, a protein binds, which then could interact with distally bound DnaA and/or RepA due to bending of the intervening DNA by IHF (Hoover et al., 1990; Stenzel et al., 1991). This protein–protein interactions model predicts that activation would be sensitive to both phasing and sequence alterations. To choose between the models we first studied the consequences of altering the phasing of the IHF site.

Importance of phasing of IHF site

To test whether simply the bend would be enough to satisfy the requirements for replication as predicted by the first model, we placed from 2 to 12 bp, in 2 bp intervals, between the right DnaA boxes and the IHF site in the P1 origin (Fig. 3). When 4 or 6 bp was inserted plasmid copy number was reduced to background levels (pRF111 and pRF112, Fig. 3). However, when 10 bp was inserted the origin function returned to almost wild-type levels (pRF114, Fig. 3). Addition of another 2 bp again began to impair origin function as the 2 bp insertion did (pRF110 and pRF115, Fig. 3). These results suggest that the presence of the bend is not enough for replication. The fact that introduction of one helical turn of the DNA (approximately 10 bp) retained optimal origin function suggests that the DNA bend must be properly phased for the downstream DNA to be functional, as required by the remaining two models.

Figure 3.

Requirement of proper phasing of the IHF site with respect to DnaA boxes for origin function. The top diagram shows relevant elements of P1 origin. Note that the origin is deleted for the left two DnaA boxes. The position of insertion of 2–12 bp random sequences was between co-ordinates 652–653. The bottom diagram shows copy numbers of plasmids with insertions relative to wild-type pALA657. The number of bp inserted and the resultant plasmid names are given along the x-axis.

Minimal sequence requirement downstream of IHF site

The second and the third models differ in terms of sequence specificity of DNA downstream of the IHF site: random sequences might suffice in the second model but not in the third. To determine if specific sequences were required downstream of IHF, a series of plasmid constructs were made which had DNA downstream of IHF deleted and replaced by vector DNA (Fig. 4). Plasmids containing only 12, 22, 32 or 42 bp of natural sequences downstream of IHF did not replicate (pRF101–pRF104, Fig. 4). However, the presence of 52 bp (co-ordinates 680–731) retained activity almost equal to wild type (pRF105, Fig. 4).

Figure 4.

Determination of the extent of additional DNA downstream of the IHF site required for origin function. Co-ordinates of the right boundary of various deletion derivates of pRF100 and their copy numbers relative to pRF100 are shown. The copy number of RF100 is identical to that of pALA657, the wild-type plasmid of Fig. 3. Grey boxes represent regions whose sequences have been randomized from the ones present in identical regions of pRF100.

To locate sequences important within the 52 bp stretch, sequences from 12 to 22 (co-ordinates 691–701) were randomized and this reduced origin function (pRF106, Fig. 4). However, when the randomization was extended by another 10 bp, the origin was essentially non-functional (pRF107, Fig. 4). As the comparison of pRF103 and pRF105 already suggested the importance of sequences between co-ordinates 711–731, it appears that particular sequences present at least in the last 30 of the 52 bp stretch downstream of the IHF binding site are important. The requirement of specific sequences suggests the presence of protein binding sites in support of the third model, and this was investigated further.

Presence of putative DnaA I-sites downstream of IHF

As the entire stretch of DNA downstream of the iterons could be replaced with a single consensus DnaA box, we expected that the role of the 52 bp stretch downstream of the IHF site is to provide binding sites for DnaA (Abeles et al., 1990; Park and Chattoraj, 2001). Although no recognizable DnaA boxes could be found there, we were encouraged by the recent finding in oriC of additional DnaA binding sites, called I-sites, with little homology to the DnaA boxes (Grimwade et al., 2000). However, the homology among the three known I-sites of oriC is rather weak and the derived consensus sequence (A/T)G(G/C)(A/T)N(G/C)G(A/T)(A/T)(T/C)A is quite degenerate. In any event, we used the sequence to search for matches to the P1 origin by manually moving it 1 bp at a time. Two stretches matched in eight (site I) and seven (site II) positions out of the degenerate 11-mer consensus (Fig.  S1). To test whether DnaA could bind to these sites we attempted DMS footprinting, an approach that led to the discovery of I-sites in oriC.

Our expectation was that to bind DnaA, the I-sites would require assistance not only from IHF but also from DnaA bound to three right boxes. Even iteron-bound RepAs could help DnaA binding to the I-sites if IHF were to bring about interactions between RepA and DnaA. The footprinting patterns in the region 22–52 bp from the IHF site that was expected to have the I-sites changed when any of the factors DnaA, IHF or RepA was absent (Fig. S1). However, the changes were not sharply confined to define a binding site, and the results could not be accepted as confirmation of the presence of I-sites. We therefore proceeded to validate the I-sites by an alternate approach.

A consensus DnaA box can functionally substitute DNA downstream of IHF

To determine if DnaA is the protein that binds to DNA downstream of the IHF site, we randomized sequences downstream of IHF between 691 and 718 before adding a consensus DnaA box (pRF200; Fig. 5). The randomization was done to eliminate the possibility of other protein binding to this region, so that the results could be unambiguously attributed to DnaA binding. In pRF200, a DnaA box was added at 2 bp intervals in both orientations in and around the position of the putative site II that we identified by sequence gazing. Robust origin function could be regained in only two cases (pRF204 and pRF206, Fig. 5), indicating that it is DnaA that binds to this region in a specific orientation in the wild-type origin. Incidentally, the position and orientation of the DnaA box in pRF206 exactly matched those of the putative site II. The facts that pRF204 and pRF206 replicated better than pRF208, and that pRF202 and pRF210 did not function suggest that the site is phasing dependent, as was also the suggestion earlier (Fig. 3). The finding of higher copy number of pRF206 than its wild-type counterpart (pRF105) is also consistent with the expectation that replacing weak I-sites with a stronger DnaA box should stimulate the origin function, if binding of DnaA is the required role of the sequences downstream of IHF. However, replacing the R box consensus sequence of pRF205 and pRF206 with two I-sites of oriC, R5 and I2, in either orientation did not allow detectable P1 origin activity (pTVC142–pTVC145, Fig. 5). As we will discuss, single I-sites are most likely not adequate for origin function.

Figure 5.

Effect of replacing sequences downstream of IHF with a consensus DnaA box. In plasmids pRF200–pTVC145, the sequences covering bp 691–718 have been randomized (grey box). The arrowheads indicate the orientation of the boxes with respect to the natural boxes. The box was placed at different positions (phases) in two orientations at each position illustrated by bidirectional arrowheads. The orientation of the box in odd numbered plasmids is opposite to, and in even numbered plasmids the same as those found naturally in the P1 origin. The box orientation is also identified at the top of copy number columns. Copy numbers are relative to that of the wild-type plasmid, pRF105. For comparison, the consensus box has been replaced with the R5 box of oriC in pTVC142 and pTVC143, and with the I2 site of oriC in pTVC144 and pTVC145.

IHF dependence of the consensus DnaA box function

If in plasmid pRF206 the consensus DnaA box functions similarly to the sequences that are naturally present downstream of the IHF site, we expected the plasmid replication to be dependent on IHF. Such was indeed the case (Fig. 6). To ensure further that the replacement of an I-site with a DnaA box did not change the replication program radically, we determined the requirement for the other three DnaA boxes and the iterons. To our surprise, the construct without the three DnaA boxes, pRF220, was still replication proficient. We then removed the three boxes from a plasmid that was otherwise wild type (pRF224). The origin of pRF224 was also functional without any recognizable DnaA box and remained IHF dependent, however, with a greatly reduced copy number. Removal of the iterons, as in pRF221, abrogated replication. These results suggest that DnaA bound to the I-site/consensus box downstream of the IHF site most likely interacts with RepA bound to the iterons in an IHF-dependent manner. As a control, we also confirmed that when the consensus box was placed next to the iterons, origin function was no longer dependent on IHF (pKP123, Fig. 6).

Figure 6.

Dispensability of right DnaA boxes but not IHF protein when a consensus box is placed downstream of IHF site. Copy numbers are relative to that of pRF105. In pRF220, the three natural DnaA boxes are randomized, illustrated by the absence of arrowed boxes upstream of the IHF site. pKP123 is a control plasmid where a consensus DnaA box replaces the three right DnaA boxes and further downstream DNA of the P1 origin.

It appears that in the normal course both the three internal DnaA boxes and multiple I-sites contribute to maximize the origin function. Removal of the three internal DnaA boxes decreased the copy number (pRF105 versus pRF224, and pRF206 versus pRF220). The involvement of multiple I-sites is suggested by the finding that randomization of the sequences between co-ordinates 691–711 decreased replication in a stepwise fashion (pRF106 and pRF107, Fig. 4), and replacement of the putative I-sites with a single I-site (pTVC142–pTVC145, Fig. 5) proved inadequate for origin function. These results suggest that multiple sites participate in bringing about optimal interactions between RepA and DnaA.

Dependence on DnaA–ATP for replication

In E. coli oriC, DnaA boxes R1 to R5 bind equally well to DnaA whether the protein is bound to ATP or ADP. The I-sites of oriC on the other hand prefer ATP-bound DnaA over ADP-bound DnaA by about sixfold (Grimwade et al., 2000). If the putative I-sites of P1 origin function similarly to the ones in oriC, we expect the plasmid replication to be compromised in hosts where ATP interactions with DnaA is aberrant. Several DnaA mutants have been characterized which either bind ATP poorly due to, for instance, A184V change (Carr and Kaguni, 1996) or bind ATP but still fail to recognize I-sites as is the case with the R285A change (Kawakami et al., 2005). The A184V change makes DnaA temperature sensitive for oriC replication, but increases its DNA binding affinity to oriC (Carr and Kaguni, 1996). With the R285A change, DnaA shows no intrinsic ATPase activity, cannot bind I-sites nor unwind oriC and so is unable to initiate replication from oriC (Kawakami et al., 2005). Unexpectedly, even with the unwinding defect, the mutant protein was capable of loading DnaB at oriC. Two other changes in DnaA, G177D and A184T, make pSC101 replication temperature sensitive (Sutton and Kaguni, 1997). The G177D change is found in the highly conserved P-loop that is thought to be the ATP binding pocket (Saraste et al., 1990).

We tested whether the mutant DnaAs could support a P1 plasmid that depends on IHF and the putative I-sites (pDKC418, Fig. 1). As controls, we used pDKC416 – a P1 plasmid that contains the left DnaA boxes and thus does not depend on I-sites, pST52 – whose replication is known not to require DnaA, and pAL70 – a minioriC plasmid (Leonard and Helmstetter, 1988). We compared efficiencies of these plasmids to transform a ΔdnaA host pre-transformed with plasmids supplying RepA and the mutant DnaAs.

The IHF-independent plasmid, pDKC416, transformed cells with about 10-fold higher efficiency than the IHF-dependent plasmid, pDKC418, in all but one case (Table 2). In cells with the G177D change in DnaA, the number of transformants with pDKC418 was two to three orders of magnitude lower compared with pDKC416. This result thus complies with our expectation that changing the ATP interaction site of DnaA could affect the IHF-dependent and -independent P1 origins differently. However, in the other three mutants that are also affected in interactions with ATP, specific defect in replication of pDKC418 was not apparent. This was particularly evident in the case of DnaA mutant R285A, whose inability to bind I-sites at oriC has been demonstrated clearly (Kawakami et al., 2005). However, in this mutant replication of pDKC418 was at least as robust as it was in the presence of wild-type DnaA. Thus it appears that for the IHF-dependent replication of the P1 origin the requirement for the ATP-bound form of DnaA is not obligatory.

Table 2.  P1 plasmid replication in cells with DnaA altered in the ATP binding domain.
Transforming plasmidsaTemperaturePlasmids in recipient
  • a. 

    miniP1 plasmids, pDKC416 and pDKC418 have been described in Fig. 1. pST52, a DnaA-independent plasmid, and pAL70, an oriC plasmid, are used here as controls. Cells also contained pALA177 to supply RepA constitutively.

  • b. 

    DnaA mutants are identified by their amino acid changes and were provided through plasmid clones in a host deleted for its own dnaA gene (CVC562).

  • c. 

    The efficiency of miniP1 plasmids to transform DnaA mutant hosts is given by the number of colonies on transformation plates.

pDKC416 (miniP1)30°C104c10310210210410
pDKC418 (miniP1)30°C1031010101030
pST52 (RSF1030)30°C104104103104104104
pAL70 (oriC)30°C1031010101010


DnaA, a well-conserved bacterial protein, is essential not only for chromosomal replication but also for replication of many plasmids (Messer, 2002). In both systems the protein helps to open the strands of origin DNA and load DnaB helicase. The activities of the protein are mediated through site-specific DNA binding but the modes of binding can be quite diverse. DnaA strongly binds to sequences 5′-TTATCCACA (consensus DnaA box) and hundreds of similar sequences with a close match in the E. coli chromosome. However, the affinity of binding is not easily predicted because of the influence of sequences flanking the boxes. In fact DnaA can bind with low affinity to sequences far removed from the consensus DnaA box, called the I-sites (so named because of improved binding in the presence of IHF), which seem to play a rate-limiting role in initiating origin opening (Leonard and Grimwade, 2005). The binding efficiency to the low affinity sites can also depend on whether DnaA is complexed with ATP or ADP, or whether there is nearby IHF binding, or both. In terms of these auxiliary factor involvements, the DnaA binding sites in oriC can be classified into three types. Boxes that are identical or nearly so to the consensus box (boxes R1-R5) bind DnaA with varying affinity but it does not change whether the protein is complexed with ATP or ADP. The I-sites on the other hand show sixfold better binding to DnaA–ATP over DnaA–ADP and the binding is further facilitated in the presence of IHF. Finally, the R5 box although binds to DnaA–ATP and DnaA–ADP equally well, the binding is improved in the presence of IHF. Thus R5 can be called an I-site as it was originally defined (Grimwade et al., 2000). Importantly, the results further showed that all IHF-stimulated binding of DnaA does not require that the protein be complexed with ATP. Here we show that P1 plasmid origin has weak DnaA binding sites with features of I-sites. The functioning of the P1 origin was dependent on IHF and on DnaA but not on sequences closely matching the consensus DnaA box. The origin functioned more efficiently when a consensus box substituted one of the purported I-site sequences, supporting the view that I-sites bind DnaA with low affinity. Finally, one DnaA mutant defective in the domain for interactions with ATP (due to G177D change) was compromised in supporting IHF-dependent P1 replication. The P1 origin thus may have sequences that function like I-sites and, in that case, our study extends the functional importance of weak DnaA binding sites beyond oriC.

Three other DnaA mutants altered in the domain for binding ATP (due to A184T, A184V and R285A changes) and defective for oriC replication, supported IHF-dependent P1 replication. Unexpectedly, the DnaA mutant R285A, whose defect in binding I-sites of oriC has been clearly demonstrated (Kawakami et al., 2005), was as good as the wild-type DnaA in supporting P1 replication. The putative I-sites of the P1 origin thus appear to be similar to R5 in the sense that they are low affinity sites but do not demand that the DnaA be bound to ATP. Unlike R5, the P1 sites show little homology to the consensus R boxes. In this respect, the P1 sites are similar to I-sites of oriC, as they all have unrecognizable (incognito) sequence homology. In summary, the studies on oriC and the P1 origin show that the IHF and DnaA–ATP-stimulated binding need not go hand in hand for weak DnaA binding sites.

How is IHF helping? The role is unlikely to be simply to improve DnaA binding as in I-sites of oriC because even when the consensus DnaA box replaced the putative I-sites, IHF was still required (pRF206, Fig. 6). In oriC, in origins of pSC101 and R6K, and in a majority of other systems where IHF participation is important, the role of the protein is believed to bring about interactions between distally bound proteins (Stenzel et al., 1987; Hoover et al., 1990; Filutowicz and Inman, 1991). In oriC, the protein brings about interactions between two distally bound DnaAs, and in the plasmid origins between DnaA and RepA. Our phasing experiments leave little doubt that action at a distance is involved in the P1 origin (Fig. 3). Furthermore, IHF is most likely allowing I-site-bound DnaAs to contact iteron-bound RepAs, as a plasmid devoid of all conspicuous DnaA boxes was still dependent on IHF and iterons (pRF220 and pRF221, Fig. 6).

In some respects, the studies described here are similar to the ones done on oriC deleted for the R4 DnaA box (Bates et al., 1995). The deletion of the box did not inactivate the origin but made it dependent on Fis, a DNA binding protein (IHF was not tested). Similarly, when we deleted the left DnaA boxes in P1 origin, which like R4 sit at one end of the origin, the plasmid became dependent on IHF, another DNA binding protein. We checked whether as in oriC, Fis has become essential in addition to IHF. In a fis::kan mutant (BR5919, Table 1), however, the left DnaA box-deleted plasmid (pDKC418) could be introduced, and it replicated similarly to the undeleted control plasmid (pDKC416) (data not shown). IHF seems to suffice for bringing about the presumptive RepA–DnaA interactions.

Table 1. E. coli strains and plasmids.
 Description, plasmid origin, drug resistanceReference
 BR2845Same as DH5Grant et al. (1990)
 BR3827ΔdnaA zia::pKN500, R1, KnHansen and Yarmolinsky (1986)
 BR4338polAtsD. Bastia
 BR4479BR4338(λDKC311)This study
 BR4543DH5 himA::Tn10This study
 BR4587N99hupA16, KnWada et al. (1988)
 BR4588N99hupA16hupB11, Kn, CmWada et al. (1988)
 BR5919MC1000fis-767Johnson et al. (1988)
 BR8139BR4479himA::Tn10Thus study
 BR8753BR4338himA::Tn10This study
 CVC562A bigger colony former of BR3827This study
 pAL70E. coli minichromosomeLeonard and Helmstetter (1988)
 pALA96A pBR322 derivative deleted for tet, ApChattoraj et al. (1984)
 pALA177Constitutive 3X RepA source, pBR322, ApChattoraj et al. (1985)
 pALA657P1 origin with only right DnaA boxes, pUC19, ApAbeles et al. (1990)
 pDKC412P1 origin with only the left DnaA boxes ligated to the Ω CmR cassettePark and Chattoraj (2001)
 pDKC416Entire P1 origin ligated to Ω Cm cassettePark and Chattoraj (2001)
 pDKC418P1 origin with only right DnaA boxes ligated to Ω Cm cassettePark and Chattoraj (2001)
 pinfA+Same as pHNαβ, pBR322, ApGranston and Nash (1993)
 pKP123P1 origin with only a consensus DnaA box on right, pUC19, ApThis study
 pKP154hupA+ in pBAD24, pBR322, SpThis study
 pMLO42pGB2 with Ω Cm cassette replacing Sp cassette, pSC101M. Lobocka
 pNND3Constitutive 7X RepA source in pST52, RSF1030, CmN. Das
 pRF100pALA657 with EcoRI site beside IHF, pUC19, ApThis study
 pSE418Source of Sp cassette, pUC18, ApS. Elledge
 pST52A dnaA independent replicon, RSF1030, CmChattoraj et al. (1985)
 pTVC101pACYC184 with Sp cassette replacing cat geneThis study

Does IHF play any role in the wild-type P1 origin (when the left DnaA boxes are also present)? In a previous study, a miniP1 plasmid, pDKC416, that included the regions required for both the IHF-independent and IHF-dependent replication, was found to have a higher copy number than pDKC412, which contained the region required for IHF-independent initiation only (see Fig. 1 for the plasmids). We expected that in the absence of IHF the two plasmids would replicate similarly as the IHF-dependent origin would have no effect. However, pDKC416 copy number was about 50% higher compared with pDKC412 in both the hosts (BR2845 and BR4543, Table 1). As we propose that the sequences downstream of the IHF site perhaps cannot function in the absence of IHF, the right three DnaA boxes appear to be responsible for better functioning of pDKC416.

Our results are also intriguing in terms of showing that a plasmid can have two largely overlapping origins. The more efficient IHF-independent origin requires RepA, HU and DnaA but the host initiator need not be in the ATP-bound form (Wickner and Chattoraj, 1987; Skovgaard et al., 1998). The overlapping second origin (which functions without the left DnaA boxes) requires IHF in addition to the above. Plasmid R6K has three overlapping origins, and one of their distinguishing features is the requirement for IHF: oriγ being IHF requiring and the other two, oriα and oriβ, not (Filutowicz and Appelt, 1988; Miron et al., 1992). The multiple origins can be a desirable feature for plasmid maintenance in that the secondary origins can serve as back-up in case the primary origin fails to fire or proves unsuitable in a different growth environment. In fact, P1 also has yet another replicon that is normally used for lytic development but can also maintain P1 prophage as a plasmid (Yarmolinsky et al., 1989).

Experimental procedures

Bacteria and plasmids

Strains are listed in Table 1. Construction of plasmids relevant to this work is described below. The rest of the plasmids are listed in Table 1.

Plasmids of Fig. 3 were constructed using the QuikChangeII Site-Directed Mutagenesis kit (Stratagene, Cedar Creek, TX), where 2, 4, 6, 8, 10 and 12 bases were inserted between the right DnaA boxes and the IHF site using pALA657 as a template, using primers RAF1/2, RAF9/10, RAF11/12, RAF13/14, RAF15/16 and RAF17/18 respectively (supporting Table 1). The resultant plasmids were called pRF110-pRF115 respectively.

Plasmids of Fig. 4 were constructed by site-directed mutagenesis using the same kit. With pALA657 as a template, and RAF19 and RAF20 as primers, an EcoRI site was created by deleting the T at co-ordinate 687 about 10 bp downstream from the IHF site, creating pRF100. The plasmid has another EcoRI site at co-ordinate 1000. Cutting pRF100 with EcoRI and religating removed 316 bp downstream of the IHF site and created pRF101. To create pRF102–pRF107, oligonucleotides RAF25/26, RAF29/30, RAF31/32, RAF33/34, RAF94/95, and RAF96/97 were annealed and ligated to pRF101 cut with EcoRI.

Plasmids of Figs 5 and 6 were constructed first by creating pRF200, which had 25 bp downstream of IHF randomized. To do this, pRF101 (Fig. 4) was cut with EcoRI and NarI and ligated to complementary oligonucleotides RAF163/164. To insert a consensus DnaA box in plasmids pRF201–pRF210, the same method was followed except another oligonucleotide pair was used. This pair allowed cloning in either orientation: RAF165/166 was used for pRF201 and pRF202, RAF167/168 for pRF203 and pRF204, RAF169/170 for pRF205 and pRF206, RAF171/172 for pRF207 and pRF208, and RAF173/174 for pRF209 and pRF210. Plasmids pTVC142 and pTVC143, and pTVC144 and pTVC145 are identical to pRF105 and pRF206 except that the consensus box is replaced with R5 box and I2 site of oriC. The primer pairs were RFR5FW and RFR5REV, and RFI2FW and RFI2REV respectively.

In Fig. 6, pRF220 was created by cutting pRF206 with EcoRI and HindIII and ligating to RAF182/183. pRF221 was a misclone, which deleted 205 bp of pRF220, removing the RepA binding sites (iterons). pRF224 was created by isolating the 214 bp downstream of IHF from pRF105 cut with EcoRI and NdeI, annealing RAF33/34, and ligating these to pRF220 cut with EcoRI and NdeI.

Plasmids of Table 2 were constructed to provide wild-type and mutant DnaA proteins from an isogenic set from the vector, pACYC184. For convenience, the CmR gene of the vector was replaced with a SpR gene. The gene was amplified from pRFG123 (Fekete and Chattoraj, 2005) with primers RAF161/162. Using the λ Red system (Datsenko and Wanner, 2000) the amplified product was inserted into pACYCdnaA+ (Sutton and Kaguni, 1995) using sequences in the primers, which were homologous to the 5′ and 3′ termini of the CmR gene of the vector. This created pRF400, which expressed wild-type DnaA from its own promoter. pRF400 was cut with EcoRI and EcoRV to remove wild-type dnaA and the mutants dnaA genes were inserted from plasmids cut with the same restriction enzymes. The G177D mutant gene was obtained from pET-146a (a plasmid similar to the wild-type dnaA carrying pKC596 except for the mutation causing the G177D change, Carr and Kaguni, 1996), the A184T mutant gene from pET-105c (again derived from pKC596), the A184V mutant gene from pACYCA184V, and the R285A mutant gene from pKA234 (Kawakami et al., 2005). The resultant plasmids were called pRF401–pRF404 respectively.

Electrophoretic mobility shift assays

The P1 origin DNA was a HindIII-PvuII (P1 co-ordinates 607–849) fragment from pALA657 (Abeles et al., 1990) and was labelled by end filling with α-32PdATP using Klenow. A HindIII-PvuII control fragment was obtained from pUC19 (co-ordinates 308–451) and labelled identically. DNA (about 2 ng) and IHF (a gift from S. Goodman) at different concentrations were mixed in a 10 μl reaction volume and incubated for 20 min at 37°C. The reaction buffer was 50 mM HEPES (pH 7.5) containing 150 mM KCl + 1 mM EDTA + 2 mM DTT + 10% glycerol + 0.1 mg ml−1 BSA. The mixture was loaded into a 6% polyacrylamide gel (14 × 0.15 cm) in 1× TBE buffer and electrophoresed at 140 V for 3 h at 4°C. The gel was dried and radioactive bands were recorded using a phosphoimager (Fujix BAS, 2000).

Plasmid copy number analysis

Cells carrying pUC19–P1ori hybrid plasmids were grown in 5 ml of Lurai–Bertani (LB) at 30°C under antibiotic selection for plasmid to an OD600 of ∼0.2. They were diluted to an OD600 of 0.02 in 20 ml LB pre-warmed to 42°C and grown at the same temperature without selection. At an OD600 of ∼0.5, four OD equivalents of cells (e.g. 8 ml of culture at OD600 = 0.5) from different cultures were mixed with 5 μl of an OD600 = 15 culture of a strain containing pNEB193 (New England Biolabs, Beverly, MA) to monitor plasmid recovery in subsequent operations. From the mixture, the cells were pelleted in a 15 ml Falcon Tube (BD, Franklin Lakes, NJ) and plasmids were isolated using a QIAprep Spin Miniprep kit (Qiagen, Valencia, CA) in 50 μl of EB buffer. Ten microlitres of each plasmid preparation was run on 2% agarose gels after digestion with 0.3 μl of DrdI (New England Biolabs, Beverly, MA) for 3 h at 37°C. After staining with ethidium bromide, the intensities of plasmid bands were measured using NIH Image and they were adjusted for fragment size and recovery with respect to the control plasmid pNEB193.


We are grateful to Jon Kaguni and Tsutomu Katayama for dnaA plasmids, Steve Goodman for purified IHF protein, Julia Grimwade and Deepak Bastia for helpful discussions, and Ranajit Ghosh and Michael Yarmolinsky for critical reading of the manuscript.