Footprinting Studies of Specific Complexes Formed by RepA, a Replication Initiator of Plasmid pCU1, and Its Binding Site

Abstract

The basic replicon of plasmid pCU1 contains three different replication origins. Replication initiated from the oriB origin requires pCU1-encoded protein RepA. Previously, information analysis of 19 natural RepA binding sequences predicted a 20-bp sequence as a RepA binding site. Guanines contacting RepA in the major groove of DNA have also been determined. In this study, we used the missing-nucleoside method to determine all of the bases relevant to RepA binding. The importance of some thymine bases was also confirmed by a missing-thymine site interference assay. Participation of the 5-methyl groups of two thymines (at positions −6 and 7) in RepA binding was pointed out by a missing-thymine methyl site interference assay. Phosphate groups of the DNA backbone which strongly interfered with RepA binding upon ethylation were also identified. The pattern of contacting positions mapped by hydroxyl radical protection footprinting indicates that RepA binds to one face of B-form DNA. The length of the binding site was found to be 20 bp by dissociation rate measurement of complexes formed between RepA and a variety of binding sequences. The symmetry of the binding site and that of the contacting bases, particularly the reacting 5-methyl groups of two thymines, suggest that pCU1-encoded RepA may contact its site as a homodimer.

The broad-host-range plasmid pCU1 belonging to the IncN group has a 2-kb replicon which contains all of the information required for plasmid replication (4, 7) and for determination of the host range pattern (6). Three distinct origins of replication (oriB, oriS, and oriV) have been identified on this fragment (1). The possibility that these multiple origins may be related to plasmid promiscuity needs to be explored; therefore the detailed structural and functional characterization of each origin is required.

It was shown previously that an intact repA gene is required for the initiation of replication from the oriB origin (5). It was pointed out that RepA is the only plasmid-encoded protein required for oriB usage (11). In vitro studies revealed two separate regions within a 2-kb region which contained multiple binding sites of the RepA protein. Five RepA binding sites are clustered in the oriB region, while 14 others are located in a region (group I iterons) known to be required for controlling the copy number of the plasmid (5, 11). Two kinds of RepA binding sites in the region of the group I iterons, which differed from each other at position −3, were identified. Due to this difference, one type of the RepA binding site overlaps with a putative DnaA binding site and the other type does not. RepA binds with different affinities to the two kinds of sites (11).

Binding sites of RepA were identified by DNase I footprinting experiments. They were not longer than 20 bp, as predicted from the information contents of the 19 natural binding sites. However, the 14-bp palindromic sequences (4-bp inverted repeats separated by a 6-bp spacer) internal to the binding sites failed to bind RepA, indicating that longer sequences are required. Methylation protection and interference footprinting experiments demonstrated that RepA makes base-specific contacts in the major groove of DNA. The guanine bases likely to contact RepA were specifically identified (11).

Here, we have extended our knowledge about the oriB-RepA system by footprinting analyses of RepA binding sites. Missing-nucleoside experiments (3) were performed in order to determine all of the bases relevant to RepA binding. Affinity measurements of RepA binding to differentially altered binding sites were used to specify the length of the binding site. Other footprinting methods were used to identify 5-methyl groups of thymines and phosphate groups of the DNA backbone likely to contact RepA.

MATERIALS AND METHODS

Preparation of labeled DNA fragments.

The 94-bp EcoRI-HindIII fragment from plasmid pPP319 (11) was used for all footprinting experiments as a single-end-labeled fragment. According to previous studies (11) this fragment carried sequences sufficient to constitute a RepA binding site. Labeling was carried out by end filling the 3′ ends of the DNAs, created by either EcoRI or HindIII digestion, using the Klenow fragment of Escherichia coli DNA polymerase I in the presence of [α-³²P]dATP. The samples were extracted with phenol and precipitated in ethanol. In a second digestion, the DNAs labeled at the EcoRI or the HindIII site were cleaved with HindIII or EcoRI, respectively. The labeled fragments were separated by electrophoresis on 5% nondenaturing polyacrylamide gels and eluted from the appropriate gel slices by the crush-and-soak method (9). The 108- or 107-bp-long EcoRI-HindIII fragments from various derivatives of pNEB19 (New England Biolabs) bearing different binding sites were used in dissociation rate measurements. These plasmids were created by cloning either a 26-bp (GATCGTGTGGAAATCCGCCCACCTTG) oligonucleotide or two kinds of 25-bp oligonucleotides into the BamHI restriction site of pNEB193. The sequences of the 25-bp oligonucleotides differed from that of the 26-bp oligonucleotide in such a way that one of the bases indicated with boldface letters was omitted. The plasmids were digested simultaneously with EcoRI and HindIII, and the ends of the fragments were labeled by end filling with Klenow fragments in the presence of [α-³²P]dATP. Labeled DNAs carrying the binding sites were isolated from polyacrylamide gels as described above.

Protein purification.

RepA protein was purified as described in reference 11 except that in the final step glycerol was omitted from the storage buffer. Aliquots of the protein samples were stored at −70°C. The purity of the RepA protein preparation was estimated to be >80%.

RepA-DNA binding reactions.

Binding reactions were performed at room temperature for 15 min. The binding buffer consisted of 20 mM Tris-HCl (pH 8.0), 60 mM NaCl, 40 mM KCl, 0.1 mM EDTA, 1 mM magnesium acetate, 2.5 mM ATP, 1 mM dithiothreitol (DTT), and 5% glycerol for gel mobility shift and interference footprinting experiments. However, glycerol was omitted from the binding reaction mixture for the hydroxyl radical protection footprinting experiment to avoid inhibition of DNA cleavage by the hydroxyl radicals. Each binding reaction mixture contained calf thymus DNA (amounts are indicated in the description of the experiments below) to reduce nonspecific binding of the labeled DNA fragment.

Hydroxyl radical protection footprinting.

Two binding reactions were set up, and each reaction mixture contained 20 ng of end-labeled fragment and 2 μg of calf thymus DNA in a volume of 40 μl of binding buffer. Two micrograms of purified RepA was added to one of the two reaction mixtures. After a 15-min incubation at room temperature the samples were subjected to hydroxyl radical treatment based on the protocol described in reference 3. Ten microliters of cleavage reagent [1 mM Fe(NH₄)₂SO₄, 3% H₂O₂, 20 mM ascorbic acid, 1 mM EDTA] was added to each reaction mixture, and the mixture was incubated for 15 min at room temperature. The reactions were quenched by adding 5 μl of 100 mM thiourea and 5 μl of 500 mM EDTA. Subsequently, 30 μl of 8 M ammonium acetate was added to the mixtures and the DNAs were precipitated in ethanol. The pellets were dissolved in 100 μl of Tris-EDTA buffer, phenol extracted, precipitated again, rinsed, and dried. The pellets were dissolved in formamide-dye mixture, and an equivalent amount of radioactivity from each sample was subjected to electrophoresis in a 10% denaturing polyacrylamide gel. Video images of the autoradiograms were generated and quantified using the system and the analySIS Pro, version 3.00, software package of Soft Imaging System GmbH (Münster, Germany).

Missing-nucleoside footprinting experiments.

The missing-nucleoside method (3) is also referred to as hydroxyl radical interference footprinting. To initiate the hydroxyl radical cleavage of DNA samples, 10 μl of cleavage reagent was added to 40 μl of reaction buffer (20 mM Tris-HCl [pH 7.5], 100 mM NaCl, 10 mM MgCl₂, 1 mM DTT) containing 50 ng of end-labeled fragment and 4 μg of calf thymus DNA. After 15 min of incubation at room temperature the reaction was quenched and the DNA was precipitated in ethanol. Four micrograms of purified RepA was added to the modified DNA (40 ng of labeled DNA plus 3.2 μg of calf thymus DNA) dissolved in binding buffer to allow the formation of complexes between RepA and the DNA fragments. The RepA-bound and unbound fragments were separated by electrophoresis on nondenaturing polyacrylamide gels. The rest of the modified DNAs (10 ng of labeled DNA plus 0.8 μg of calf thymus DNA), which had not been reacted with RepA, were also loaded on the gel to serve as controls. The DNAs were eluted from the appropriate gel slices by the crush-and-soak method (9). The pellets were dissolved in a formamide-dye mixture, and electrophoresis of the samples containing equal amounts of radioactivity was performed on a 10% denaturing polyacrylamide gel. The autoradiograms were quantified as stated above.

Preparation of uracil-containing DNA.

PvuII-digested pPP319 plasmid DNA served as a template to produce uracil-containing DNA carrying a RepA binding site. PCR-mediated DNA amplifications (30 cycles of 94°C for 1 min, 40°C for 1 min, and 72°C for 1 min) were carried out using the −20 sequencing and −21 reverse primer pair (New England Biolabs) either in a standard PCR or under conditions where 5 or 10% of the dTTP was replaced by dUTP. The amplified DNA contained the 94-bp EcoRI-HindIII fragment of plasmid pPP319, and it was labeled as described above.

Missing-thymine site interference footprinting.

One unit of uracil-DNA-glycosylase (Amersham) was added to 0.5 μg of end-labeled uracil-containing DNA in 24 μl of binding buffer supplemented with 0.1 mg of bovine serum albumin/ml, and the reaction mixture was incubated at 37°C for 30 min to allow uracil-DNA-glycosylase to remove all of the incorporated uracil bases. Two micrograms of calf thymus DNA (1 μl) and 5 μg of purified RepA diluted in binding buffer (15 μl) were added, and the binding reaction was allowed to proceed at room temperature for 15 min. The RepA-bound complexes and the unbound fraction of the fragments were separated, and DNAs were eluted as described above. The pellets were dissolved in 1 M piperidine, and the mixture was incubated for 30 min at 90°C. The samples were lyophilized and resuspended in a formamide-dye mixture, and an equivalent amount of radioactivity from each sample was electrophoresed on a 12% denaturing polyacrylamide gel. The uracil-DNA-glycosylase-digested uracil-containing DNA treated with piperidine served as a control (U ladder) to determine the degree of uracil incorporation at each position in the unselected DNA population.

Missing-thymine methyl site interference footprinting.

The binding reaction mixture contained 0.5 μg of end-labeled uracil-containing DNA, 2 μg of calf thymus DNA, and 5 μg of purified RepA in a volume of 40 μl. The binding reaction was allowed to proceed at room temperature for 15 min, after which the unbound and the protein-bound DNA fractions were separated and isolated as described above. The DNA pellets were resuspended in 20 μl of uracil-DNA-glycosylase assay buffer (20 mM Tris-HCl [pH 8.0], 1 mM EDTA, 1 mM DTT, 0.1 mg of bovine serum albumin/ml). Then 1 U of uracil-DNA-glycosylase was added and the reaction mixture was incubated at 37°C for 30 min. The samples were extracted with phenol and precipitated in ethanol. The DNAs were subjected to piperidine treatment and handled further as described above.

Ethylation interference footprinting.

Ethylation interference footprinting experiments were carried out as described in reference 10, except that 50 ng of end-labeled fragment was used for the modification of DNA by ethylnitrosourea (Sigma) and 40 ng of the ethylated labeled DNA was used with 5 μg of purified RepA in the binding reaction.

Dissociation rate measurement of RepA-DNA complexes.

Experiments were done as described in reference 11. Bound fractions of the samples were quantified by the system and software package of Soft Imaging System GmbH.

RESULTS

Missing-nucleoside footprinting experiments to detect bases important for RepA binding.

Guanine bases facing RepA in RepA-DNA sequence-specific complexes were identified in a previous study (11). In addition to guanine we also intended to determine the importance of other bases in the complex formation. By missing-nucleoside footprinting experiments the effect of elimination of any base on RepA binding can be detected (3); thereby, information about the relative importance of all bases can be obtained. Furthermore, it was expected that the boundary of the binding site could also be defined by this method.

Figure 1 shows the autoradiograms and the densitometric evaluation of the results. Thymine bases at the outermost positions (−6 and 7) of the 14-bp palindromic region seem to be very critical for RepA binding. On the basis of these data the RepA binding site could be narrowed down to 19 bp (from positions −7 to 11). A slight interference could be seen on the top strand at positions −7 and 11, while the bottom strand indicated no significant contribution to the binding at these positions. The data are in good agreement with the former result, which indicated that the 14-bp sequence did not constitute a functional RepA binding site. It is interesting to note that the bases playing an important role in the interactions with RepA follow each other uninterruptedly on the top strand while a gapped pattern can be found on the bottom strand. However, the phenomenon has no known significance with regard to specific features of the DNA-protein complex or DNA-protein interactions. It can be noted that the inverted repeats inside the binding site are 5 bp long. This is a feature of each RepA binding site belonging to group I iterons, but the five RepA binding sites in the oriB region contain only 4-bp-long inverted repeats (11).

FIG. 1.

Missing-nucleoside analysis of RepA-DNA complexes. The autoradiograms obtained with the top and bottom strands are indicated. Lane a, G+A Maxam-Gilbert sequencing markers; lane b, chemically modified DNA not subjected to RepA binding; lanes c and d, unbound and bound fractions, respectively, of modified DNA subjected to RepA binding. Brackets, locations of the 20-bp binding sites (positions −8 to 11). The numbering of the positions of the RepA binding site is the same as in reference 11. Some of the positions showing strong interference with RepA binding are indicated within the brackets. The densitometric evaluations of the results aligned to the sequence are shown below the autoradiograms. The heights and signs of the bars were deduced as follows: (band intensity in lane c − band intensity in lane d)/band intensity in lane b. Positive values are shown above the line for the top strand and below the line for the bottom strand. Arrows, 4-bp palindromic regions.

PCR incorporation of uracil residues into RepA binding sites for footprinting experiments.

pPP319 contains a single RepA binding site cloned into the multiple cloning sites of pUC19. We selected a primer pair which can be used to amplify the entire region of the multiple cloning sites. Three parallel reactions were set up, each differing in only one component from the other two. The first PCR mixture contained equivalent amounts of all four nucleotides, as usual in a DNA amplification reaction. The other two reaction mixtures contained a fifth nucleotide (dUTP), and its amount was adjusted in such a way that 5 or 10% of the dTTP was replaced by dUTP. Similar amplifications were detected in all three reactions, which correlated with the observation reported in reference 2 that Taq polymerase did not discriminate strongly between dUTP and dTTP in the course of DNA synthesis. The uracil-DNA-glycosylase-resistant fractions of the amplified DNAs were determined. In the reaction where 10% of the dTTP was replaced with dUTP, 72% of the DNA remained intact after the sample was treated with the enzyme and piperidine (data not shown). The above data indicated that the uracil incorporation was less than one per DNA fragment on average. We used this condition to produce uracil-containing DNA for the footprinting experiments.

Determination of thymine contacts with RepA by missing-thymine site interference footprinting.

The 94-bp-long, end-labeled uracil-containing EcoRI-HindIII fragments from PCR amplification were used in these experiments. In missing-thymine site interference assays, uracil bases were excised by uracil-DNA-glycosylase from the modified DNAs and then subjected to complex formation with RepA. After the separation of RepA-bound and unbound DNA fractions, the DNAs were treated with piperidine and analyzed on a sequencing gel. The results are shown in Fig. 2A and C for the top and bottom strands of the RepA binding site, respectively. Thymines required for RepA binding are indicated by diminished binding activity when the uracil bases were eliminated from positions −8, −6, 0, 10, and 11 in the top strand and −2, −1, and 7 in the bottom strand.

FIG. 2.

Autoradiograms showing the results of missing-thymine (MT) site (A and C) and missing-thymine methyl (MTM) site (B and D) interference assays on the top (A and B) and bottom (C and D) strands of the RepA binding site. Lane a, G+A Maxam-Gilbert sequencing markers; lane b, U ladder (uracil-containing DNA not subjected to RepA binding); lanes c and d, unbound and bound fractions, respectively, of uracil-containing DNA subjected to RepA binding. Five thymine bases on the top strands (positions −8, −6, 0, 10, and 11) and three on the bottom strands (positions −2, −1, and 7) which are essential for RepA-DNA complex formation were identified by MT site assays. Positions of the bases are indicated within the brackets representing the 20-bp RepA binding site. The MTM site assays showed that only one of these thymines on each strand (positions −6 on the top strand and 7 on the bottom strand) uses the 5-methyl group to contact RepA. We have noticed not only that thymine bases were replaced by uracil during the PCR synthesis but also that a minor fraction of cytosine bases were also replaced due to the possibility of G:U base pair formation (lanes b). Elimination of the uracil bases from the binding sites representing cytosine bases may or may not have resulted in interference with RepA binding, but the behavior of these sites does not influence the results that we obtained concerning the role of thymine bases in RepA binding.

Methyl groups of thymines at positions −6 and 7 are involved in RepA binding.

For the missing-thymine methyl site interference assays, the uracil-containing end-labeled DNAs were allowed to form complexes with RepA protein. Both the RepA-bound and unbound DNAs were isolated and digested with uracil-DNA-glycosylase and then treated with piperidine and analyzed in a sequencing gel. Replacement of thymine by uracil at position −6 on the top strand (Fig. 2B) and at position 7 on the bottom strand (Fig. 2D) interfered with RepA binding. Since the only difference between a uracil and a thymine is the methyl substitution at position 5 of the pyrimidine ring of the thymine, our results indicate that the 5-methyl groups of T₋₆ and T₇ are involved in contacting RepA. We presume that RepA makes hydrophobic interactions through these two atomic groups in the major groove of the DNA since the methyl group of the thymine is accessible only in the major groove.

Identification of phosphate contacts by ethylation interference footprinting.

The end-labeled DNAs were modified by ethylnitrosourea treatment, resulting in one ethylated phosphate per fragment on average. The positions of the ethylated phosphates interfering with RepA binding were identified after selection of the RepA-bound and unbound fractions of a binding reaction followed by cleavage of DNAs at the modified positions and analysis on a sequencing gel. The results are shown in Fig. 3.

FIG. 3.

Ethylation interference footprinting of RepA-DNA complexes. Lane a, G+A Maxam-Gilbert sequencing markers; lane b, ethylated DNA not subjected to RepA binding; lanes c and d, unbound and bound fractions, respectively, of ethylated DNA subjected to RepA binding. Partially resolved multiple bands can usually be seen due to alternative ethylations at the two oxygens of each phosphate and occasional ethylation of the bases. The numbers indicate base positions related to interference with RepA binding. Bands marked with arrows indicate that interference at position −5 on the top strand and position 6 on the bottom strand was due to ethylation of the bases. Ethylation of phosphates 3′ to the base at position 2 on the top strand and 3′ to the bases at positions −3, 7, 8, and 9 on the bottom strand interfered with RepA binding. The locations of these phosphates can be seen on B-form DNA in Fig. 5.

It has been established that cleavage of ethylated DNA produces DNA fragments with three different kinds of ends: 3′-OH and 3′-ethylated phosphates derived from DNAs ethylated at phosphate groups and 3′ phosphate that originates from ethylated bases (8, 10, 12). As a result of their different electrophoretic mobilities, up to three bands could be seen in ethylated DNA for each band of the lane representing the products of Maxam-Gilbert sequencing reactions. The fastest-moving species of the triplet migrated similarly to the products of sequencing reactions and, therefore, are assumed to end with a 3′ phosphate. The other two slower-migrating species were assumed to represent the products of ethylated phosphates. Taking these observations into consideration, the strong interference, detected at positions −5 in the top strand and at 6 in the bottom strand, was due to ethylation of the bases. Interference with RepA binding at positions 2 (top strand) and 3, 7, 8, and 9 (bottom strand) indicates phosphate contacts 3′ to the bases. The locations of the identified phosphate contacts along the B-form DNA are shown in Fig. 5.

FIG. 5.

Map of some of the base and backbone contacts of RepA on a schematic B-form DNA. Contacts were deduced from methylation protection and interference footprinting (G), missing-thymine methyl site assays (T), and ethylation interference (P) and hydroxyl radical protection studies (filled circles).

Mapping of contacts between RepA and its site by hydroxyl radical protection footprinting.

The hydroxyl radical protection footprinting method (3) was applied to map the solvent accessibility of the DNA surface in the presence of bound RepA protein (Fig. 4). The protected residues mapped to one face of DNA (Fig. 5). This is the same face with which RepA makes specific contacts in the major groove with the guanine bases (at positions −5 and 6) and the thymine bases (at positions −6 and 7). These results indicate that RepA binds to two consecutive major grooves on one face of the DNA.

FIG. 4.

Hydroxyl radical protection footprinting of RepA binding sites. Lane a, G+A sequencing markers. The end-labeled DNAs were subjected to hydroxyl radical cleavage in the absence (lane b) or presence (lane c) of RepA. The histograms showing the magnitude of protection at each position were drawn as described for Fig. 1, but the heights of the bars were derived from the differences in band intensities in lanes b and c. Arrows, locations of the 4-bp-long palindromic regions. The locations of backbone contacts of RepA were depicted on a schematic B-form DNA in Fig. 5.

Determination of the length of a RepA binding site by dissociation rate measurements.

The results of missing-nucleoside footprinting experiments indicated that the sequence from position −7 to 11 constitutes the RepA binding site. However, the requirement of T at position −8, as shown by the missing-thymine site interference assay, does not fit into this picture. In order to determine the length of the binding site, three oligonucleotides with different binding sites were synthesized. In these experiments we used the RepA binding site which overlaps with a putative DnaA binding site (11). The first oligonucleotide contained the 20-bp RepA binding site from position −8 to 11 (TGTGGAAATCCGCCCACCTT), while the other two carried a different base at either the farthest left (GGTGGAAATCCGCCCACCTT) or farthest right (TGTGGAAATCCGCCCACCTA) positions of the site. The selected bases (boldface) do not occur in homologous positions of the wild-type binding sites. The synthetic binding sites inserted into the BamHI restriction site of pNEB193 and the resulting recombinant plasmids containing the different binding sites were confirmed by sequencing. The dissociation rates of the complexes formed between RepA protein and the different binding sites have been measured (Fig. 6). The affinities of the RepA protein to these binding sites were determined from the half-lives of the complexes, which were calculated from the slopes of straight lines fitted to the data points (Fig. 6). The half-life of the complexes formed by RepA and the wild-type binding sites was 41 ± 2 min, while the half-lives of the complexes formed by RepA and the two other sequences, having a base substitution either at the left or the right end, were 17 ± 2 and 12 ± 1 min, respectively. According to these results we can conclude that bonds of a RepA binding site stretch at least from position −8 to 11.

FIG. 6.

Dissociation of RepA-DNA complexes. (A) Autoradiograms of band shift assays used to monitor the rates of breakdown of RepA complexes with wild type (wt) and mutant binding sites [MP₍₋₈₎ and MP₍₁₁₎, where the numbers in parentheses indicate the positions of the mutations]. The concentration of RepA was kept constant (120 mM), while the concentration of the RepA binding sites varied between 45 and 63 pM. In each case the left lane shows the distribution of free and RepA-bound sites to be at an equilibrium. The other lanes were obtained in the presence of a 1,000-fold excess of unlabeled competitor DNA (supercoiled pPP319 plasmid DNA). This DNA was added at time zero, and the samples were loaded thereafter. Time intervals were 3 min (from 0 to 21 min). (B) The RepA-bound fractions of the samples were determined in each assay, and the data points were fitted to a straight line. The half-lives of the complexes, the times required to dissociate 50% of the complexes, were calculated from the slopes of the straight lines.

DISCUSSION

We have analyzed the interactions between RepA protein and its specific binding sites resident in plasmid pCU1. Binding affinity measurements and different kinds of footprinting analyses were used. From these and other data of previous studies (11) following conclusions can be drawn about RepA binding.

The RepA binding site is 20 bp long.

The measured half-lives of complexes formed between RepA protein and DNAs carrying different sequence variants of RepA binding sites indicated that the sequence from position −8 to 11 contains critical information for binding to the RepA protein. The binding site contains a 14-bp sequence with palindromic flanks (4-bp inverted repeats separated by a 6-bp spacer) which alone was not sufficient to constitute a RepA binding site (11). However, the 14-bp region extended with two additional bases on the left side and with four bases on the right constituted a functional binding site. The experimental data confirmed that the sequence logo (13) predicted correctly the length of the binding site of the pCU1 replication protein (11).

RepA is likely to bind as a homodimer to its site on one face of the DNA.

The symmetry present in the RepA binding site is also reflected in RepA binding shown by similar base contacts within the T₋₆-G₋₅ and G₆-T₇ dinucleotide steps. The similar contacts of RepA in two consecutive major grooves on one face of DNA suggest that RepA recognizes its site as a homodimer. However, asymmetry can be seen in the pattern of contact points between RepA and its site shown by the hydroxyl radical protection footprinting experiments and in the locations of the phosphate contacts. Asymmetry can also be seen in the number of bases required at the two sides of the 14-bp sequence (two bases on the left and four on the right). The asymmetry may provide polarity to the unit complexes (i.e., one RepA dimer bound to one site) when the complexes are arrayed along the oriB region. This polarity might be needed to ensure that RepA molecules bound to the neighboring binding sites could form specific higher-order complexes by protein-protein interactions.