Sequence recognition, cooperative interaction, and dimerization of the initiator protein DnaA of Streptomyces.

Using a combined PCR-gel retardation assay, the preferred recognition sequence of the Streptomyces initiator protein DnaA was determined. The protein showed a preference toward DNA containing two Escherichia coli-like DnaA boxes in a head-to-head arrangement (consensus sequence TTATCCACA, whereas the consensus sequence of the DnaA boxes found in the Streptomyces oriC region is TTGTCCACA). In quantitative band shift experiments, the kinetics of the Streptomyces DnaA-DnaA box interaction was characterized. The DnaA protein can form dimers while binding to a single DnaA box; dimer formation is mediated by the domain III of the protein, and the dissociation constant of this process was between 35 and 115 nm. Streptomyces initiator protein DnaA interacts in a cooperative manner with DNA containing multiple binding sites. For the cooperativity effect, which seems to be independent of the distance separating the DnaA boxes, domain I (or I and II) is responsible. The cooperativity constant is moderate and is in the range of 20-110.

The initiator protein DnaA plays an essential role in the initiation of bacterial chromosome replication. The interaction of DnaA protein with its chromosomal origin (oriC) is best understood in Escherichia coli. The E. coli DnaA protein (52 kDa) binds to five nonpalindromic, nonamer sequences, the DnaA boxes. Binding of 10 -20 DnaA monomers promotes a local unwinding of an adjacent AT-rich region. The unwound region provides an entry site for the DnaB/DnaC helicase complex followed by other proteins required to form a replication fork (1)(2)(3)(4).
Apart from its primary function as a replisome organizer, the DnaA protein acts as a transcription factor that represses or activates several genes or terminates transcription, depending on the location and arrangement of DnaA boxes (5).
Both functions of the DnaA protein, replisome organizer and transcription factor, are mediated by its interaction with target DNA. The DNA binding domain of E. coli DnaA has been localized in the 94 C-terminal amino acids. A potential helixloop-helix motif has been reported within this part of the protein (6). However, because x-ray high resolution structure analysis is not yet available for DnaA proteins, the detailed interaction with the DNA is still poorly understood. The consensus sequence of the E. coli DnaA box differs depending on the method used for its evaluation. The most stringent definition for the DnaA box sequence comes from a determination of binding constants: 5Ј-TT(A/T)TNCACA-3Ј (7). T 2 , T 4 , T 7Ј , and T 9Ј were found to be directly involved in DNA-protein interaction (8). E. coli DnaA does not dimerize in solution and interacts with a single DnaA box as a monomer, as measured by gel retardation (7) and by surface plasmon resonance (9). However, binding to DnaA boxes that differ from the stringent consensus sequence by one or two base pairs requires two such boxes and an interaction of DnaA proteins bound to them (9). DnaA binds ATP and ADP with high affinity. Both forms of DnaA protein, ATP-DnaA and ADP-DnaA, recognize DNA in a similar fashion; however, only ATP-DnaA is active in the DNA replication process. Recently, it has been shown that E. coli ATP-DnaA protein recognizes also a hexamer sequence, the ATP-DnaA box 5Ј-AGATCT-3Ј or close match of it (9).
Streptomycetes (Gram-positive soil bacteria) differ from other prokaryotic organisms in their mycelial life cycle and in possessing a large (8-megabase pair), linear and GC-rich (about 72%) chromosome (10,11). Recent discoveries suggest that replication of the linear chromosome of Streptomyces coelicolor A3(2) proceeds bidirectionally from the centrally located oriC region toward the ends of the chromosome (12).
The key elements of initiation of the Streptomyces chromosome replication, oriC region and DnaA protein, show higher complexity than those of E. coli. The Streptomyces oriC region contains numerous DnaA boxes, which are grouped into two clusters (13,14). The Streptomyces DnaA protein consists, like all other DnaA proteins, of four domains. In contrast to the other bacteria, the Streptomyces DnaA protein is larger (70 -73 kDa), since it comprises an additional stretch (ϳ230 amino acids) of predominantly acidic or hydrophobic amino acids within domain II. The residues lower the isoelectric point of the entire Streptomyces DnaA protein (pI ϭ 5.7) (15,16). As it was shown for the E. coli and Bacillus subtilis DnaA proteins, domain III and the C-terminal part (domain IV) of the Streptomyces DnaA protein are responsible for binding of ATP and DNA, respectively (15,17). The consensus sequence of the Streptomyces DnaA box in oriC is 5Ј-TTGTCCACA-3Ј, which differs at the third position (AЈG) from the E. coli DnaA box (13). In contrast to E. coli, the Streptomyces DnaA protein can form a dimer when binding to a single DnaA box. Recently, it has been shown that the domains I and III are independently involved in dimerization of the Streptomyces lividans DnaA protein molecules. The interaction of Streptomyces DnaA protein with two DnaA boxes is cooperative and accompanied by strong DNA bending (16,18). However, we do not know the contribution of the different domains of the Streptomyces DnaA protein to cooperative binding.
In this work, we apply a combined PCR-EMSA 1 technique (19) to elucidate DNA sequence requirements for Streptomyces DnaA protein binding. Using EMSA, we define details of the binding of the Streptomyces wild type DnaA and its truncated forms to DNA. We reveal that the domain I participates in cooperative DnaA protein-DNA interactions. Quantitative analysis of gel mobility shifts allowed us to determine binding constants for dimerization and cooperative DNA protein interactions. In addition, we have varied the spacing between two DnaA boxes and examined the consequences on dimerization and cooperative binding of DnaA protein to these boxes.

EXPERIMENTAL PROCEDURES
Proteins and DNA-The His-tagged wild type DnaA proteins of S. lividans and its truncated mutants DnaA(III-IV) comprising the domain III and the DNA binding domain were purified on a Ni 2ϩ -nitrilotriacetic acid-agarose column (Qiagen) as described earlier (15). The DNA binding domain of the DnaA protein DnaA(BD) was expressed as a C-terminal glutathione S-transferase fusion and purified using glutathione-Sepharose 4B beads (Amersham Pharmacia Biotech) followed by factor X a cleavage as described earlier (16).
Oligonucleotides were chemically synthesized and purified by high performance liquid chromatography. For the kinetic analysis of DnaA protein binding complementary nucleotides were annealed by heating at 95°C for 10 min and gradually cooling to room temperature. The oligonucleotides were end-labeled with [␥-32 P]ATP and T4 polynucleotide kinase and purified from 5% nondenaturing polyacrylamide gels.
In Vitro DNA Binding Site Selection-The recognition motif of the DnaA protein was determined using several cycles of amplification and selection, essentially as described previously (19). The library of 58-bp oligonucleotides for the first selection cycle was prepared by PCR using template oligonucleotides: 5Ј-GGCGGATCCTCGACTAGCGN 20 GCTA-CGAGCTGAGCTCGCG-3Ј and the primer pair 5Ј-GGCGGATCCTCG-ACTAGCG-3Ј and 5Ј-CGCGAGTCTAGCTCGTAGC-3Ј (restriction sites BamHI and SacI are in italic type). The amplification reaction was carried out in 100 l using 1 pmol of template oligonucleotide (1 pmol corresponds to the number of all possible combinations of 20 degenerated bases) and 100 pmol of each primer for 20 cycles, with each cycle consisting of 15 s at 96°C, 30 s at 60°C, and 30 s at 72°C. Doublestranded oligonucleotides were separated on 3% agarose gels, electroeluted for 1 h to TBE buffer, and purified by phenol-chloroform extraction and glycogen-ethanol precipitation. The oligonucleotides were 5Јend-labeled with [␥-32 P]ATP and T4 polynucleotide kinase and then incubated with DnaA protein in concentrations of 5 or 50 nM in binding buffer (20 mM HEPES-KOH, pH 8.0, 5 mM magnesium acetate, 1 mM Na 2 EDTA, 4 mM dithiothreitol, 0.2% Triton X-100, 5 mg/ml bovine serum albumin, and 100 M ATP) The reaction was carried out in 20 l for 30 min at room temperature in the presence of competitor, poly(dI-dC), at a concentration of 2.5 g/l. The complexes were separated on 5% nondenaturing polyacrylamide gels in 0.25ϫ TBE. The bands corresponding to DNA-protein complexes were excised, and the DNA was eluted into gel elution buffer (50 mM Tris-HCl, pH 8.0, 500 mM NaCl, 5 mM MgCl 2 ) for 3 h at 60°C and recovered by phenol-chloroform extraction followed by glycogen-ethanol precipitation. The DNA was then amplified by PCR at conditions described above and subjected to five additional selection cycles of binding and amplification. The subsequent cycles differed only in the amount of competitor added to binding reactions (for cycles 2-6, we used 5, 7.5, 20, 50, and 100 g/l, respectively). Selected oligonucleotides after the fourth, fifth, and sixth round were digested with BamHI and SacI and cloned into a pUC19 vector. The DNA from independent clones was recovered using a Spin Miniprep Kit (Qiagen) and subjected to sequencing using 32 P-5Ј-end-labeled "Ϫ47" primer (New England Biolabs) and the Thermo Sequenase cycle sequencing kit (Amersham Pharmacia Biotech).
The free DNA and complexes were separated by electrophoresis on 5% native polyacrylamide gels, prerun for 1 h (0.25ϫ TBE, 6 V/cm, 20°C). Following electrophoresis, the radioactive gel was dried and analyzed using a PhosphorImager and ImageQuant software (Molecular Dynamics, Inc., Sunnyvale, CA). The variation in experimental data was evaluated by repeating each experiment three or four times. Different protein titration experiments showed variations up to 15%. Equilibrium constants were determined using the modified statistical mechanical model (20) supplemented with the protein dimerization module. For the DNA substrate containing a single binding site A, the concentration of each species in the gel was expressed by the following equations (model 1), where S, SL, and SL 2 represent concentrations of free DNA, monomer, and dimer respectively; L is protein (ligand) concentration; k A is the microscopic equilibrium affinity constant for binding site A, and k Di is the affinity constant describing the dimerization process. For the DNA containing binding sites A and B, the concentrations of the species presented in the gel are described by the following equations (model 2), where k B is the intrinsic affinity constant for the binding site B, and k AB is the cooperativity constant describing interaction of protein molecules occupying both binding sites. Experimental data were simultaneously fitted to the equations using Scientist® for Windows TM software (MicroMath).

Determination of the Optimal DnaA Protein Binding Site-
The consensus sequence of the DnaA box identified within the Streptomyces oriC region is 5Ј-TTGTCCACA-3Ј. The Streptomyces oriC region contains a higher number of the DnaA boxes (19 boxes) than the E. coli oriC region (five boxes). A DNase I footprint of the oriC region with DnaA showed protection of all boxes. However, when analyzed individually by surface plasmon resonance (SPR), only some DnaA boxes were specifically recognized by the DnaA protein (14,15). To characterize in detail the recognition properties of the Streptomyces DnaA protein, we applied a binding site selection technique based on the combined EMSA and polymerase chain reactions ( Fig. 1A; Ref. 19).
The purified DnaA protein was incubated with a substrate pool of double-stranded oligonucleotides, in which a random 20-bp region was flanked by defined 19-bp sequences containing restriction sites to facilitate subsequent cloning. The protein-DNA complexes were separated from the unbound DNA on a 5% polyacrylamide nondenaturing gel. Only one proteinbound fraction was visible. The DNA recovered from the complex was amplified by PCR with the pair of primers complementary to the defined sequences of the oligonucleotides (Fig.  1B) and used as a substrate for the renewed binding assay. In the third selection cycle, we observed a second retarded band with a higher electrophoretic mobility than the previous one. Here, the bands with lower and higher electrophoretic mobility are called a "dimer" (Fig. 1C, ** complex) and a "monomer" (Fig.  1C, * complex), respectively. The DNA from both bands was recovered and used independently in the next selection cycles. The alignments of the selected oligonucleotides after the fourth, fifth, and sixth cycle are shown in Tables II and III. The sequences selected from the "dimer" band after four rounds reveal remarkable features: 63 out of the 67 oligonucleotides analyzed carry two DnaA boxes, one complete 9-bp DnaA box, the "box 1," and one incomplete DnaA box, the "box 2" ("box" as defined by sequence). The incomplete box 2 consists of 4-7 bp (3Ј-part of the DnaA box). Both boxes face each other FIG. 1. In vitro DNA binding sites selection assay. A, the experimental design of the test. B, DNA substrate and the primer pair used in the assay. N, any nucleotide; boldface letters, restriction sites BamHI and SacI. C, subsequent rounds of the DnaA binding site determination. Lanes 1, 2, and 3 in each cycle correspond to 0, 5, and 50 nM concentration of the wild-type DnaA protein, respectively. Frames indicate gel fragments from which DNA used in the next round and/or sequencing was extracted. * and ** signify DNA bands from which selection toward monomer and dimer started.

TABLE II
Selected DNA binding sites for DnaA ("dimer" band) The sequence of cloned binding sites for DnaA, derived after the fourth, fifth, and sixth selection cycles, were aligned for maximum match to the canonical DnaA box definition. The region of 20 originally random nucleotides is shown in uppercase letters, whereas nucleotides from adjacent primer regions are indicated by lowercase letters (only five bases from each side of the random region are shown). Letters in boldface type match the DnaA box consensus sequence. Letters in italic type indicate the lower strand of the DnaA box. For the consensus sequence, letters in boldface represent nucleotides present more than 85%. Uppercase letters indicate nucleotides appearing in the frequency range 50 -85%. Nucleotides that are represented at less than 15% were considered as not significant at the indicated position. Box   The sequence of cloned binding sites for DnaA, derived after the fourth, fifth, and sixth selection cycles, were aligned for maximum match to the canonical DnaA box definition. The region of 20 originally random nucleotides is shown in uppercase letters, whereas nucleotides from adjacent primer regions are indicated by lowercase letters (only five bases from each side of the random region are shown). Letters in boldface type match the DnaA box consensus sequence. Letters in italic type indicate the lower strand of the DnaA box. For the consensus sequence, letters in boldface represent nucleotides present more than 85%. Uppercase letters indicate nucleotides appearing in the frequency range 50 -85%. Nucleotides that are represented at less than 15% were considered as not significant at the indicated position. Box  essential T nucleotides at the 7Ј-and 9Ј-positions. More than half of the oligonucleotides selected from the monomer band after four rounds contained also two boxes. However, the monomer band had only appeared after the third round of the selection (Fig. 1C), and therefore it still contained traces of the oligonucleotides from the dimer band that were subsequently amplified. After the next two selection rounds, only one out of the 44 oligonucleotides contained two boxes. All of the boxes 1 from the monomer band exhibit the E. coli type DnaA box (R1/R4).
The data show that under the conditions of the assay, the Streptomyces DnaA protein interacts only with DnaA boxes; no other consensus sequences (e.g. the newly identified 6-bp ATP-DnaA box (9)) have been found. The results, presented in Tables II and III, show that DnaA protein from Streptomyces possesses a higher affinity toward DnaA boxes from E. coli than toward those from its own oriC region. Binding of the DnaA protein as a dimer apparently requires "head-to-head" orientation of the boxes.
Kinetic Studies with Streptomyces DnaA Protein-Upstream of the promoter region of the Streptomyces dnaA gene are two closely spaced DnaA boxes: a strong one (with the preferred Streptomyces sequence: 5Ј-TTGTCCACA-3Ј) and a weak one (5Ј-TTGTCCCCA-3Ј) in head-to-head arrangement with 3 bp in between (21). Interaction of the DnaA protein with these boxes creates an autoregulatory circuit similar to that known for the E. coli dnaA gene (21). Recently, we have shown that binding of the DnaA protein to both DnaA boxes exhibits a cooperative character (16). Domains I and III independently participate in the dimerization of the DnaA protein molecules (18). To evaluate in detail the kinetics of binding of the DnaA protein to DnaA boxes from the dnaA gene promoter region, we applied gel mobility shift assays to determine the binding constants for cooperativity as well as for dimerization (20). This assay permits the quantitative analysis of the individual protein-DNA complexes. In our gel retardation experiments, we used the wild-type DnaA protein and its truncated forms, the DnaA(III-IV) lacking the two N-terminal domains (I and II) and the DnaA(BD) containing only the DNA binding domain IV. As a prerequisite for the kinetic studies, two DNA substrates containing either a single strong DnaA box, 1s-0 (the weak DnaA box was scrambled), or two DnaA boxes, 2s, derived from the promoter region of the dnaA gene were designed in such a way that the DnaA box(es) is flanked on both sides by sequences corresponding to those within the dnaA promoter region (Table  I). To simplify kinetic calculations, the 2s substrate contains two identical boxes; the weak DnaA box was replaced by the strong one (Table I). For the analysis, we used a fixed input DNA concentration and various protein concentrations, spanning 4 orders of magnitude (see legend to Fig. 2). The concentration of DNA in the reaction mixture was chosen to be at least 5 times lower than the lowest protein concentration used.
According to the statistical mechanical approach (20), the interaction between the DnaA protein and the DNA substrate composed of two binding sites could be described by the following equations, S ϭ 1/Z (Eq. 9) SL ϭ ͑k A ϩ k B ͒L/Z (Eq. 10) where S, SL, and SL 2 are concentrations of free DNA, monomer, and dimer bands, respectively; L is protein (ligand) concentration; k A and k B are the microscopic equilibrium affinity constants for binding sites A and B, respectively; and k AB is the cooperativity constant describing the interaction of protein molecules occupying both binding sites. These equations contain parameters k A , k B , and k AB that occur only in combination and therefore cannot be unambiguously determined. However, if one of the pair k A or k B is known, it is possible to determine the remaining two. One way to achieve that goal is carrying out the gel shift experiment using a substrate containing only one binding site. In our analysis, it is the 1s-0 oligonucleotide. Concentration of free DNA or DNAprotein complexes in this system is given by the Langmuir isotherm.
The interaction of the 1s-0 DNA fragment with increasing amounts of the wild type DnaA and DnaA(III-IV) led to two complexes ( Fig. 2A, a and c), whereas incubation with the DnaA(BD) resulted in a single complex even at the highest protein concentration (Fig. 2A, e). Formation of a dimer band by the first two DnaA derivatives could be explained by an unspecific interaction with the scrambled DnaA box or another sequence motif of the 1s-0 oligonucleotide or a protein dimerization process occurring on the DNA with only one-half of the dimer bound to DNA. The first possibility can be excluded, because domain IV alone does not form slower migrating complexes on that substrate, and all three proteins do not interact with DNA in which both DnaA boxes are scrambled (data not shown). Taking into account the dimerization process, we modified the original equations by adding a constant describing the oligomerization reaction. As a result, we obtained two sets of equations (model 1 and model 2) for the interaction with a single (1s-0) and with two DnaA boxes (2s), respectively (see "Experimental Procedures").
The affinity constant for the single DnaA box, k A , and the affinity constant describing the dimerization process, k Di , were calculated by analyzing the interaction of the wild-type DnaA (or its truncated forms) with the 1s-0 substrate according to model 1 or its simplified version, the Langmuir isotherm for the DnaA(BD). The values obtained from those experiments allowed us to determine k B and k AB , the intrinsic affinity constant for the second DnaA box in substrate 2s, and the cooperativity constant, respectively (Fig. 2B).
The wild type DnaA protein and its truncated forms exhibit similar affinity to the single DnaA box, the 1s-0 substrate (Table IV; K d ϭ 4 -10 nM). The affinity constants calculated for the second DnaA box from the 2s substrate (k B ) were found to be nearly identical to the k A values. Thus, the affinities seem to be independent of flanking sequences of the DnaA box(es).
The calculated cooperativity constants, k AB , showed that only the wild type DnaA is able to interact cooperatively with the two DnaA boxes; the cooperativity effect increases the affinity of the DnaA protein toward the DnaA boxes over 40 times. The values obtained for DnaA(III-IV) and DnaA(BD) are very close to 1; therefore, we consider these interactions as noncooperative. The results demonstrate that domain I or II or both of them are responsible for the cooperativity effect.
The dimerization dissociation constant for the wild-type protein (35 nM) is about 3 times lower than for a protein lacking domains I and II, DnaA(III-IV) (114 nM).
The Distance Separating DnaA Boxes Does Not Influence the Cooperativity of the DnaA-DnaA Box Interaction-The arrangement of the 19 DnaA boxes occurring within the Streptomyces oriC regions is highly conserved. The spacing between DnaA boxes varies from 3 to 20 bp or even more (except for two DnaA boxes that are adjacent). To evaluate the influence of the distance that separates binding sites on the cooperativity, we have designed three double-stranded oligonucleotides with two strong DnaA boxes separated by various spacings. The first one is the previously analyzed 2s substrate that contains 3 bp between the two DnaA boxes. The 2s mimics the spacing between the fifth and sixth DnaA boxes within the Streptomyces oriC region. The other two, 2s ϩ5 and 2s ϩ10 (Table I), contain 5 and 10 additional base pairs between the DnaA boxes, respectively. In the 2s ϩ10 substrate, the increase of the distance separating both boxes by 10 bp does not affect their positioning on the DNA helix, while in the 2s ϩ5 variant, the two DnaA boxes lie on the opposite face of the helix. The calculated cooperativity constants for the three substrates are presented in Table V.
Adding a turn of the helix (2s ϩ10) increased the cooperativity about 2.5 times. Surprisingly, the addition of 5 bp (a half helix turn) caused negligible change of the cooperativity. As it was shown for the "wild-type" substrate, 2s, the interaction of the truncated forms of the DnaA protein, DnaA(III-IV) and DnaA(BD), with the 2s ϩ5 and 2s ϩ10 substrates exhibited a noncooperative character (Table V).
Streptomyces DnaA Protein Prefers DnaA Boxes of the E. coli Type-The binding selection assay suggested that the Streptomyces DnaA protein prefers the DnaA box of E. coli type (5Ј-TTATCCACA-3Ј) over its own DnaA box (5Ј-TTGTCCACA-3Ј). To evaluate in detail the interaction of the DnaA protein with the E. coli type DnaA box, quantitative gel retardation assays were performed. As before, we used a two DnaA box system that enabled us to evaluate cooperativity in addition to intrinsic affinity constants. Thus, two substrates were used for the calculations: Sel 1s-0 and Sel 2s, containing the single box 1 (E. coli DnaA box) or two adjacent boxes 1 in head-to-head orientation, respectively (Table I). The Sel 1s-0 substrate was used to determine the affinity constant for the single binding box and the dimerization affinity constant. These values were then used for the interaction analysis of the DnaA protein with two boxes (substrate Sel 2s). Since the wild-type DnaA protein formed dimers during the interaction with a single DnaA box, the same mathematical models (models 1 and 2; see "Experimental Procedures") were applied. In addition to the substrate with two identical boxes, the Sel 1s-1/2 oligonucleotide selected after the sixth cycle of the binding assay was also analyzed. The results of this analysis and their comparison with the constants obtained for the wild-type DnaA boxes from the promoter region are summarized in Table VI.
The affinity of the wild-type DnaA to the single DnaA box of E. coli is approximately 4 times higher than its affinity to the strong Streptomyces DnaA box. It explains why such DnaA recognition sequences had been found in the selection binding assay. As expected, the dimerization constant does not depend on the DNA recognition sequence; its value is nearly identical for both types of DnaA boxes (Table VI). The cooperativity constant for the Sel 2s substrate is approximately 2 times lower than the cooperativity for the wild type Streptomyces DnaA boxes (2s substrate). This may be due to the fact that the selected sequences did not have the 3-bp spacer. DISCUSSION Despite extensive work on the mechanism of initiation of DNA replication in prokaryotic and eukaryotic systems, several critical aspects of this mechanism and its control still remain obscure. One significant gap is a lack of understanding of the biological and biochemical roles of multiple initiator protein binding sites required by a large group of plasmid (interons) and chromosomal replicons (DnaA boxes). The oriC region of the Streptomyces linear chromosome consists of 19 DnaA boxes that serve as binding sites. Therefore, Streptomyces provides a good model for studying the interaction of initiator protein with multiple cognate binding sites. The initiator protein DnaA of Streptomyces contains two dimerization domains (I and III) that are separated by a long flexible domain II. According to our preliminary results (16,22), interaction of the DnaA protein with two DnaA boxes exhibits cooperativity (e.g. the DnaA protein specifically binds to DNA fragments with two "weak" DnaA boxes, which are not bound as individ-

TABLE VI
Comparison of the binding properties of wild type DnaA protein to the DnaA boxes derived from the promoter region of the dnaA gene and oligonucleotides obtained in the optimal site selection For every constant determination, S.D. is given (expressed as a percentage). *, corresponding dissociation constants (reciprocal values of the affinity constants). **, this substrate comprised only one DnaA box; k B does not exist. ***, the intrinsic affinity constants for the binding sites A and B were found to be not identical. White and black arrows indicate E. coli-and Streptomyces-like DnaA boxes, respectively. The gray arrow indicates box 2 found by the in vitro DNA binding site selection assay. ual boxes). Here, we present the recognition properties of the Streptomyces DnaA protein and the kinetic details of dimerization and cooperative DNA binding of this protein. Finally, we try to answer the questions of why the Streptomyces oriC region contains so many DnaA boxes and how it arose.
The Streptomyces DnaA Protein Prefers the E. coli DnaA Box over Its Own DnaA Box-Analysis of 57 DnaA boxes from three different Streptomyces oriC regions showed that the preferred sequence is 5Ј-TTGTCCACA-3Ј ("strong" DnaA box). However, each of the analyzed oriC regions contains only one strong DnaA box, while the sequences of others, the weak DnaA boxes, differ by one or two bases from the preferred sequence. Our previous results revealed that the Streptomyces initiator protein DnaA interacts specifically only with a few isolated (14) boxes, including the strong one from its own oriC region (Table  VII). To evaluate the recognition properties of the Streptomyces DnaA protein, a binding selection assay based on a combined PCR-EMSA technique was applied. Unexpectedly, we found that the Streptomyces DnaA protein prefers the E. coli DnaA box instead of its own DnaA box. After four cycles of binding, only a fraction of the selected oligonucleotides contained Streptomyces-like DnaA boxes, whereas after subsequent cycles (fifth and sixth), this type of DnaA box did dot appear at all. The results are also supported by the gel retardation experiments (Table VII). The Streptomyces DnaA protein shows about 4 times higher affinity for the E. coli DnaA box (K d ϭ 2.6 nM) than for its own strong DnaA box (K d ϭ 10.4 nM); therefore, the E. coli box was exclusively selected during the PCR-EMSA binding assay. The data obtained by gel retardation are consistent with the dissociation constants determined by surface plasmon resonance; the affinity of the Streptomyces DnaA protein to the E. coli DnaA box and to the strong Streptomyces DnaA box have been calculated to be K d ϭ 3.4 nM and K d ϭ 12 nM, respectively (Table VII) (16,17).
The nucleotides immediately adjacent to the selected binding box exhibit low diversity (5Ј-(A/g)(G/C)-box 1-(G/a)-3Ј); therefore, the possibility cannot be excluded that these sequences influence the binding affinity. The naturally occurring DnaA boxes are usually flanked by G from both sides (5Ј-(a/c/g/t)(G/c)t-DnaA-G-3Ј).
Cooperativity and Dimerization: Features of the Streptomyces DnaA Protein-So far, mainly cooperative DNA binding of transcription factors has been studied extensively (23). Cooperativity is best understood for cI repressor. The recently obtained crystal structure of the cI repressor C-terminal domain provided a comprehensive study of the molecular basis of cooperativity (24). In eukaryotes, activation of genes frequently requires the cooperative assembly of large protein complexes on the DNA. However, cooperative binding may also serve functions other than regulation of gene expression. Protein-protein interactions that result in cooperative DNA binding may lead to bending and/or looping of the intervening DNA, contributing to the formation of higher order structured DNAprotein complexes. According to our previous studies, the formation of the Streptomyces initial nucleoprotein complex involves the sequential binding of the DnaA protein molecules to 19 DnaA boxes. Subsequent protein-protein interaction leads to bending and looping of the Streptomyces oriC region (18).
Our recent experiments indicate that the wild-type Streptomyces DnaA protein binds two or more DnaA boxes in a cooperative manner (16). E. coli DnaA protein also binds cooperatively to two nonamer DnaA boxes and to three hexamer ATP-DnaA boxes located within the dnaA promoter region (9). Here, we have shown that the truncated DnaA protein lacking domains I and II is deficient in cooperative binding to two adjacent sites. Thus, domain I (or domains I and II) is responsible for this effect. However, the domain II of the DnaA proteins is highly variable and does not contain any relevant secondary structure elements. Therefore, it is not a plausible candidate for any functional role.
The wild-type DnaA protein binds to two DnaA boxes separated by various spacings with a cooperativity parameter ranging from 23 to over 100. Surprisingly, cooperativity does not depend severely on the spacer length separating both binding sites (Table V). Adding a turn of the helix (2s ϩ10) increased the cooperativity about 2.5 times, whereas the addition of 5 bp (a half of a helix turn) had a modest effect on the cooperativity. Removing of 3 bp (Sel 2s) resulted in only 50% reduction of the cooperativity parameter. However, the Sel 2s oligonucleotide consists of the E. coli type DnaA boxes and therefore cannot be directly compared with other DNA fragments. The moderate influence of the spacer length on the cooperativity may suggest that the protein domains involved in the intramolecular reactions are very flexible. Therefore, changing the spacing between adjacent binding sites does not affect the ability of the DnaA protein to bind in a cooperative manner. The spacing between DnaA boxes within the oriC region varies from 3 to 20 bp. Thus, in theory, the DnaA protein could be able to interact cooperatively with each pair of the adjacent DnaA boxes. However, in the binding site selection assay, only DnaA boxes that face each other have been selected. Therefore, we speculate that cooperativity occurs only when the binding sites are oriented head-to-head (such a box arrangement has been found in the promoter region of the dnaA gene as well as in the oriC regions, e.g. the fifth and sixth DnaA boxes). Our DNase I footprinting experiments (17) and electron microscopy studies (18) corroborate this hypothesis. DNase I footprinting experiments showed that in the oriC region, at the low protein concentration, DnaA binds first to the fifth and sixth DnaA boxes. According to the electron microscopy studies, the highest incidence of protein binding occurred at the middle of the first cluster of DnaA boxes, which corresponds to the location of DnaA boxes 5 and 6. Therefore, we assume that cooperativity at close distance determines at least the start of DnaA-oriC complex formation. Additional long range interactions may be formed subsequently and may be responsible for loop formation.
We established the kinetic constants for dimerization of the DnaA protein. The Streptomyces DnaA is the first chromosomal initiator for which the kinetics of two dimerization domains, I and III, have been determined. These domains dimerize independently (18). The dimerization of the DnaA protein does not occur in the absence of DNA. The wild-type DnaA protein reveals 3 times higher dimerization capability than the truncated DnaA protein containing only one dimerization domain.
The intermolecular interactions of the DnaA protein are 3-20 times weaker than the interactions between DnaA protein and its DNA target (Table IV). Probably, it facilitates effective interactions of the DnaA molecules with DNA containing multiple recognition sequences (e.g. within the oriC region) and further formation of the nucleoprotein complex. For E. coli DnaA, the N-terminal domain 1 has been shown to promote oligomerization (25). However, as for Streptomyces DnaA, a second interaction face has been postulated in domain 3 or 4 (26).
Why Does the Streptomyces oriC Region Contain so Many DnaA Boxes?-The GC content of 57 Streptomyces DnaA boxes derived from three Streptomyces oriC regions is about 10 and 20% lower than the GC content of the oriC region (63%) and the overall GC content of S. lividans DNA (72%), respectively. However, it is still significantly higher than the GC content of the average E. coli DnaA box (ϳ30%). This difference in GC content explains the difference in the DnaA box consensus sequence between Streptomyces (5Ј-TTGTCCACA-3Ј) and E. coli (5Ј-TTATCCACA-3Ј). It also results in flanking sequences of Streptomyces DnaA boxes that are relatively rich in GC.
As shown in Table VII, the dissociation constant (K d ) for specific binding of individual DnaA boxes derived from the S. lividans oriC region varies between 10 and 78 nM; a few additional DnaA boxes are not recognized by DnaA if they are analyzed outside the context of oriC (K d exceeding 200 nM). Interestingly, the affinity of the S. lividans DnaA protein for the R1/R4 E. coli DnaA box (K d ϭ 3.4 nM) is ϳ3 times higher than its affinity for the strong Streptomyces DnaA box. The apparent dissociation constant for binding of the E. coli DnaA protein to the DnaA box R1/R4 was calculated to be K d ϭ 1.1 nM (7).
Despite the strong difference in GC content between Streptomyces and other microoganisms, including E. coli, the domains I and III and the binding domain of the Streptomyces DnaA are highly conserved. A consequence of the high GC content of Streptomyces structural genes (72-74%) is the very nonrandom codon usage, with an extreme paucity of codons with A or T in the third position (C or G is usually at the third codon position). Like E. coli DnaA protein, the binding domain of Streptomyces DnaA protein contains the same putative DNA binding motive: two amphipathic ␣-helices with the basic loop in between followed by a third long ␣-helix (6,16).
The affinity of the S. lividans protein for DNA fragments containing two or three closely spaced DnaA boxes is 6 -10 times higher than its affinity for the single strong DnaA box (Table VII) and is comparable with the affinity of the E. coli DnaA protein to the R1/R4 box (0.9 -1.2). The data suggest that efficient binding of the Streptomyces DnaA protein to DNA requires the presence of more than one Streptomyces DnaA box. Therefore, due to the high GC pressure exerted during the course of Streptomyces evolution, the structure and the sequence of oriC region has been changed.
The relatively low affinity of Streptomyces DnaA protein for a single Streptomyces DnaA box seems to be compensated by a high number of DnaA boxes that are bound in a cooperative manner. In comparison with the cI repressor, the cooperativity parameter of the wild-type DnaA protein (40 -100) is moderate; cI repressor by binding to its three operator sites shows cooperativity in the range of 250 -950 (20). Such high values in a system with 19 binding sites (Streptomyces oriC) might cause irreversible binding of DnaA protein to the oriC region and consequently would block subsequent replication steps.