Specific Recognition of DNA by Integration Host Factor

Integration host factor (IHF) is a protein that binds to the H′ site of bacteriophage λ with sequence specificity. Genetic experiments implicated amino acid residue Glu44 of the β-subunit of IHF in discrimination against substitution of A for T at position 44 of the TTR submotif of the binding site (Lee, E. C., Hales, L. M., Gumport, R. I., Gardner, J. F. (1992) EMBO J., 11, 305–313). We have extended this observation by generating all possible single-base substitutions at positions 43, 44, and 45 of the H′ site. IHF failed to bind these H′ site substitution mutants in vivo. The K d (app) value for each H′ site substitution, except for H′45A mutant, was reduced >2000-fold relative to the wild-type site. Substitution of amino acid β-Glu44with alanine prevented IHF from discriminating against the H′44A variant but not the other H′ site substitution mutants. Further analysis with other substitutions at position β44 demonstrated that both oxygens of the wild-type glutamic acid are necessary for discrimination of AT at position 44. Because the β-Glu44residue does not contact the DNA, this residue probably enforces binding specificity indirectly through interaction with amino acids that themselves contact the DNA.

Integration host factor (IHF) 1 was discovered as a protein required for site-specific recombination of bacteriophage (1). It is a small (20 kDa) basic heterodimer and member of the HU-like family of DNA binding proteins. These proteins have two domains, a helix-turn-helix motif involved in protein-protein dimerization interactions and an arm composed of two antiparallel ␤-sheets that bind to and bend DNA (2,3). Unlike other HU-like proteins, IHF binds to DNA with a high degree of sequence-directed specificity (4). How IHF achieves sequence specificity has been a challenging question since its discovery.
The HЈ site of attP from bacteriophage is one of the best characterized IHF binding sites. The HЈ site is defined as the 34 base pairs protected from DNase I digestion by IHF (4). Alignment of the HЈ site and other known IHF binding sites revealed three conserved sequence elements (5). The WATCAR element (where A, C, G, and T are the standard nucleotide bases, W is A or T, and R is A or G) forms the core of the consensus. It is separated from a second element to its 3Ј side, TTR, by four base pairs that are not conserved among IHF binding sites. The third element, a poly(dAT) tract of 4 -6 base pairs, is found in a subset of the known IHF binding sites. It is located 5Ј to the WATCAR element but is separated from it by approximately eight base pairs. The bases between the poly(dAT) and the WATCAR element also appear not to be conserved among IHF binding sites.
The validity of the consensus sequence was confirmed in a genetic study in which base pair substitutions that disrupt IHF binding were isolated within each of these elements (6). One of these mutant HЈ sites, which contains a T to A change at the center position of the TTR element, HЈ44A, was used in a genetic selection to find substitution mutants within IHF that allow it to bind this variant (7). The IHF mutants isolated replaced glutamic acid 44 of the ␤-subunit with a glycine (␤E44G), lysine (␤E44K), or valine (␤E44V). A model based on the crystal structure of HU from Bacillus stearothermophilus predicted that position 44 of the ␤-subunit was in the center of a small ␤-sheet, which is highly conserved among IHF proteins from different species (8). It connects the dimerization domain and the DNA-binding antiparallel ␤-sheets (3). Thus, this ␤-sheet is implicated in specific DNA binding to the TTR element (6). The recent IHF-HЈ site cocrystal structure confirms the conclusions of the studies detailed above and reveals amino acid side chains that contact DNA (2). In this structure, only one amino acid (␤-Arg 46 ) of this ␤-sheet makes a hydrogen bond with a DNA base of the TTR element. The majority of contacts made by IHF with the HЈ site are to the phosphates and riboses of the DNA backbone. Thus, it appears that IHF recognizes the HЈ site sequence largely through indirect sequence-dependent variation of the DNA backbone conformation.
In this report, we examine the importance of the bases of the TTR element for IHF binding by making all of the possible single-base pair substitutions within this element of the HЈ site. These variant HЈ sites were examined for IHF binding in the in vivo challenge phage assay and an in vitro gel mobility shift assay. Interpreted in light of the IHF-HЈ site cocrystal structure, these experiments highlight the importance of alternating pyrimidines and purines within the TTR element and of its flanking sequences in IHF binding. Of particular interest is the role of ␤-Glu 44 in the interaction of IHF with the HЈ site TTR element. The glutamic acid residue was substituted with various amino acids and tested in a series of in vivo assays of IHF function. The results indicate that IHF is exquisitely sensitive to the precise placement of ␤-Glu 44 but that this residue is not required per se for complex formation. Assays with a challenge phage containing the mutant HЈ44 site demonstrate that ␤E44 allows IHF to discriminate between a T:A and an A:T base pair at the central position of the TTR element, although this residue does not itself directly interact with DNA. Both oxygens of the carboxyl group are necessary for specific recognition.

EXPERIMENTAL PROCEDURES
Bacterial Strains- Table I lists the Escherichia coli and Salmonella typhimurium strains used in this study.
Media, Chemicals, and Enzymes-LB broth (10 g/liter tryptone, 5 g/liter yeast extract, 10 g/liter NaCl) was the standard liquid culture medium, with 15 g/liter agar added for solid media. Ampicillin and kanamycin (Sigma) were supplemented at a final concentration of 50 g/ml where required. Isopropyl-␤-D-thiogalactopyranoside (IPTG; Sigma) was supplemented at the concentrations indicated.
SacII endonuclease and Taq DNA polymerase were purchased from Promega. All other DNA modification enzymes were purchased from Life Technologies. Reaction conditions used were according to the suppliers' instructions. Deoxyribonucleoside triphosphates for DNA amplification were purchased from Promega. Oligodeoxyribonucleotides for DNA amplification and cassette mutagenesis were purchased from Genosys.
Challenge Phage Assays-Challenge phages used in this study are derivatives of bacteriophage P22 modified so that IHF binding to the HЈ site regulates expression of the reporter (ant) gene encoding antirepressor from the P ant promoter. Because IHF is the only protein capable of repressing P ant -directed transcription of ant, binding of IHF to the HЈ site controls the lysis-lysogeny decision of infecting phage.
Phage P22-based challenge phage assays were performed as described previously (6). Various strains of JG1151, containing pEKR99a derivatives, were grown overnight in LB plus ampicillin, subcultured 1:50 into LB plus ampicillin, and grown for 3 h at 37°C. The cultures were diluted 1:4 into LB plus ampicillin containing IPTG at concentrations ranging from 1 M to 1 mM to induce IHF synthesis. Challenge phages were added to a multiplicity of infection of 25, and the mixtures were incubated for 1 h at room temperature. Dilutions of each mixture were plated onto LB plates containing ampicillin, kanamycin, and IPTG at the concentrations used during the induction and incubated 24 h at 37°C. Viable cells were determined by plating dilutions of induced, uninfected cells on LB containing ampicillin and IPTG. The percentage of lysogeny was determined as the number of ampicillin-and kanamycin-resistant lysogens multiplied by 100 divided by the number of ampicillin-resistant viable cells. Negative controls were performed with JG1151 containing pTrc99a and a challenge phage lacking the HЈ site. Challenge phages used in this study contain the goh1 mutation that allows efficient P22 lytic growth in the absence of IHF (6).
Gel Mobility Shift Assays-A pair of complementary oligodeoxyribonucleotides, d(GTGACCTGTTCGTTGCAACAAATTGATAAGC-AATGCTTTTTTATAATGCCAACTTA) and d(TAAGTTGGCATTATAA-AAAAGCATGCTTATCAATTTGTTGCAACGAACAGGTCAC) containing the sequence of the HЈ site or mutant derivatives as well as flanking DNA from attP were synthesized. The 5Ј-hydroxyl of each oligonucleotide at a final concentration of 1.6 pmol/l was phosphorylated with [␥-32 P]ATP and T4 polynucleotide kinase (9). The phosphorylated oligonucleotides were heated at 90°C for 3 min and then annealed by reducing the temperature to 20°C to give a final duplex concentration (v/v) glycerol. The reactions were incubated at room temperature for 20 min to allow equilibration. A 5-l aliquot of each reaction was loaded onto a 7% polyacrylamide gel (acrylamide:bisacrylamide ratio of 29:1) in 1ϫ TBE (89 mM Tris borate, pH 8.0, 2 mM EDTA) that had been run at 10 V/cm for 30 min, and the sample was then subjected to electrophoresis for 2.75 h. The gel was dried and exposed to a PhosphorImager cassette (Molecular Dynamics, Inc., Sunnyvale, CA), and the appropriate bands were quantified (10).
Plasmid Construction- Table I lists the plasmids used in this study. The plasmid pEKR7Zf(ϩ) is an IHF-encoding derivative of pGEM7Zf(ϩ) (Promega). It encodes ampicillin resistance and expresses the IHF genes ihfB and ihfA from P lac . DNA containing ihfB and ihfA was amplified with the oligonucleotides d(CGGAATTCGGCCGCCCTT-AATCATT) and d(GGTCTAGAGCGGCCTTTTTAGTTAG) using pHN␤␣ (6) as a template. These oligonucleotides introduce an EcoRI site 53 base pairs 5Ј to the start of ihfB and an XbaI site 10 base pairs 3Ј to ihfA. The amplified fragment and the plasmid pGEM7Zf(ϩ) were individually digested with EcoRI and XbaI, mixed, and joined with DNA ligase.
Plasmid pEKR99a and its mutant derivatives were constructed by ligating the 969-base pair EcoRI-ihfB-ihfA-XbaI fragment from pEKR7Zf(ϩ) into EcoRI-XbaI digested pTrc99a (Amersham Pharmacia Biotech) downstream from P lac . Expression of IHF from pEKR99a is regulated by lacI q .
Oligonucleotide Cassette Mutagenesis-A pair of complementary oligonucleotides, d(GaATTTCAATACGCTCGCCCTGCGCAAGAGTCGA-GGCCA) and d(TATGGCCTCGACTCTTGCGCAGGGCGAGCGTATT-GAAATtCGC), were used for each amino acid substitution introduced into ihfB. Each set comprised codons 32-46, with the underlined bases indicating codon 44. Each of the four duplexes that introduce the four different substitution mutations (alanine, aspartic acid, glutamine, and asparagine) in codon 44 (underlined) changed codon 44 to the appropriate trinucleotide sequence to generate each amino acid substitution. Each duplex also changed the third base of codon 45 from C to T (indicated in the oligonucleotide sequences by lowercase type). This disrupts a natural SacII restriction site without changing the amino acid encoded. After annealing, the oligonucleotides formed complementary duplexes with the top strand providing 5Ј NdeI and 3Ј SacII cohesive ends. The oligonucleotide duplexes were ligated with T4 DNA ligase into pEKR7Zf(ϩ) previously digested with NdeI and SacII and then electroporated into EM424. Ampicillin-resistant isolates were used to prepare miniprep DNA. Mutated plasmids were identified as being resistant to digestion by SacII, and the entire ihfB gene was sequenced to confirm the presence of the base substitution.
Mu Killing Assay-IHF binding to the ihfB site of bacteriophage Mu is required for the stimulation of transcription from the early promoter P e , expression of the lytic genes, and subsequent killing of the host through multiple transpositions and cell lysis (12). Thus, the efficiency  (13), and the assay for this function was performed as described previously (14) with the following modifications. Electrocompetent cells containing pEKRR99a or a mutant derivative were electroporated with a mixture of pSC101 and pET28a(ϩ), a plasmid that does not require IHF for replication or maintenance. After 1 h of recovery incubation, an aliquot of each electroporation mixture was serially diluted and plated onto LB containing ampicillin and spectinomycin or kanamycin. After overnight incubation at 37°C, the effect of IHF mutations was quantified as the number of electroporants maintaining pSC101 (spectinomycin-resistant) divided by the number maintaining pET28a (kanamycin-resistant).
Molecular Modeling-The IHF-HЈ site cocystral structural coordinates reported by Rice et al. (2), 1IHF.pdb, were obtained from the protein data base (available on the World Wide Web). The coordinates were modeled with the program Swiss-pdViewer PPC version 3.5, and amino acid substitutions were generated with the mutate function of the programs (15). Images were exported to POV-Ray Tracer program for rendering and labeled with text in Adobe Photoshop version 5.5.

IHF Binding to Variants Containing Changes in the TTR
Element-Challenge phages containing all nine possible variants that produce single-base substitutions at positions 43-45 in the TTR element of the HЈ site were constructed as described under "Experimental Procedures." With 1 mM IPTG induction, P22-HЈ(II) containing a wild-type HЈ site formed lysogens at a frequency of 20% (Fig. 1). All of the variant phages failed to form lysogens at 1 mM IPTG, suggesting a significant reduction in the affinity of IHF for these sites. The HЈ site DNA (0.25 M) and salmon sperm DNA (75 g/ml) were mixed with a series of 2-fold dilutions of IHF and subjected to gel shift analysis. The K d (app) was defined as the IHF concentration at which 50% of the labeled DNA fragments were shifted into a bound complex when the IHF/DNA concentration ratios were large. IHF had a K d (app) of 1 nM for the wild-type site (Table II), which is comparable with the 1.5 and 1.9 nM values reported previously (10,16). The HЈ45A duplex was bound by IHF with a K d (app) of 10 nM, an approximately 10-fold reduction relative to the wildtype HЈ site. IHF failed to shift 50% of the fragment of each of the remaining HЈ site mutants, and thus these sites have a K d (app) greater than 2.7 M.
Previous methylation interference studies of the HЈ site suggested that the composition of the minor groove at positions 43 and 44 was important for IHF binding (8). Subsequent studies of the HЈ site with the base analogs 2-aminopurine and 2,6diaminopurine, which place an N2 amino group in the center of the minor groove, confirmed the importance of the minor groove of these positions in IHF binding (10). The experiments using base analogs suggested that it is the N2 amino group of guanine in the center of the minor groove that causes the reduction of affinity of IHF for HЈ sites containing C:G or G:C base pairs at these positions. In the IHF-HЈ cocrystal structure, the center of the minor groove of positions 43 and 44 is occupied by ␤-Arg 46 (Fig. 2a). Thus, the N2 amino group of guanine in either a C:G or G:C base pair is predicted to clash with ␤-Arg 46 .
The effect of the HЈ43A and HЈ44A mutations upon IHF binding cannot be explained by the presence of a disruptive N2 amino group in the minor groove. In fact, the hydrogen bond accepting groups in the minor groove of the A:T base pair are formally indistinguishable from those of a T:A base pair because they appear to occupy equivalent positions relative to its pseudodyad axis of symmetry (19). However, the HЈ43A and HЈ44A mutants have an altered sequence of pyrimidine and purine bases within the consensus strand of the TTR element, and this may change the geometry of the bases and backbone of the bases neighboring the changed base. This effect could account for the ability of a protein to distinguish A:T from T:A base pairs because the different nearest neighboring bases alter the precise positioning of hydrogen-bonding groups of the two configurations of the A:T base pair in the minor groove. Upon IHF binding, these alterations in the DNA geometry could disrupt a necessary contact or introduce steric clashes between the TTR element and IHF.
Binding of Wild-type and Mutant IHF Proteins to the HЈ Site-A previous genetic study (7) indicated that glutamic acid 44 might be involved in sequence-specific interactions with the HЈ site because single amino acid substitutions to glycine, lysine, or valine allow the mutant proteins to repress the P22-HЈ44A challenge phage. This phage contains a T to A change at the center position of the TTR element of the HЈ site. However, the mutants also continued to repress the challenge phage containing the wild-type HЈ site. It is possible that ␤-Glu 44 prevents stable IHF-HЈ44A site complex formation because the negative charge of the carboxyl group clashes with some portion of the TTR element and that the ␤-Glu 44 substitution mutants allow complex formation with the mutant site by removing this charge. In order to understand how ␤-Glu 44 is involved in the DNA binding specificity, additional amino acid substitutions at position 44 were tested in a series of assays. The glutamic acid was substituted with alanine (␤E44A), aspartic acid (␤E44D), asparagine (␤E44N), and glutamine (␤E44Q) through site-directed mutagenesis. A specific prediction was that only the ␤E44D protein would prevent repression of P22-HЈ44A(II), because it, like the wild-type glutamic acid, carries a negative charge. The ␤E44N and ␤E44Q mutants were constructed to test directly the effect of the negative charge by converting the carboxylate to the uncharged amide. The ␤E44D and ␤E44N changes shortened the side chains by a methylene group, and the ␤E44A mutation further shortened the side chain.
Most of the mutants had reduced IHF binding activity relative to wild-type IHF with a challenge phage containing the wild-type HЈ site (Fig. 1a). The ␤E44A protein lysogeny curve did not differ significantly from the wild-type IHF lysogeny curve. At an induction concentration of 1 M IPTG, the ␤E44G protein has 50-fold lower lysogeny frequency than the wild type and ␤E44A proteins. The difference in binding between the ␤E44A and ␤E44G proteins may be due to the greater conformational freedom of the peptide backbone in glycine-substituted mutant. At 1 M IPTG, the ␤E44D, ␤E44N, and ␤E44Q proteins promoted only background levels of lysogeny. At an induction concentration of 10 M IPTG, the reduction in percentage lysogeny of these mutants relative to the wild-type protein was less dramatic. With the exception of the ␤E44D protein, all the mutants tested reached a maximum percentage of lysogeny at 100 M IPTG induction. The percentage of lysogeny of ␤E44D was 5000-fold lower than wild-type IHF at 100 M IPTG but eventually reached a maximum at 1 mM IPTG that was similar to the levels achieved by other IHF mutants at 100 M IPTG. The reduction in binding of these mutants (␤E44D, ␤E44G, ␤E44N, and ␤E44Q) relative to the wild-type IHF indicates that IHF is a less efficient repressor of P ant transcription when amino acid ␤44 is not glutamic acid or alanine.
IHF Binding to the Mutant HЈ44A Site-Substitution mutants of glutamic acid ␤44 to glycine, lysine, and valine were originally isolated as mutants capable of repressing the P22-HЈ44A(II) challenge phage mutant DNA binding site (7). These amino acids have such chemically dissimilar side chains that we reasoned that it must be the negative charge of ␤-Glu 44 that prevents IHF from stably binding the HЈ44A site.
Both the ␤E44G (one of the original suppressors) and the ␤E44A proteins repress P22-HЈ44A(II) at levels comparable with those seen with the wild-type P22-HЈ(II) (Fig. 1b). The ␤E44N and ␤E44Q mutants do not repress the P22-HЈ44A(II) phage as effectively as the P22-HЈ(II) phage. Both fail to form lysogens at 1 and 10 M IPTG induction. At higher levels of induction, the percentage of lysogeny increases but remains significantly below the levels seen with P22-HЈ(II). Both the wild-type (␤E44) and ␤E44D fail to repress the P22-HЈ44A(II) phage at the highest induction concentration (1 mM IPTG). This result confirms the initial hypothesis that amino acids carrying a negative charge at ␤44 prevent stable IHF-HЈ44A complex formation.
Binding of IHF Mutants to Variants Containing Changes in the TTR Element-As mentioned, Lee et al. (7) demonstrated that the proteins with glycine, lysine, or valine substitutions at position ␤44 could repress the P22-HЈ(II) phage but could not repress HЈ site challenge phages containing mutations in the poly(dAT) tract, the N8 spacer, or the WATCAR element. This finding suggested that substitution at ␤44 does not compensate for deleterious interactions with distant portions of the HЈ site. Except for the P22-HЈ44A phage, no substitution mutations within this element were available for testing then, so the effects of changes within the TTR element were determined. Here we tested the ␤-Glu 44 substitution mutants to determine if the loss of specificity with the HЈ44A site extended to any of the variants with base substitutions in the TTR element. As was observed for the wild-type protein, none of the mutants formed lysogens with challenge phages containing the altered TTR elements (data not shown). This lack of repression suggests that other amino acids are responsible for the specificity at these positions in the HЈ site and that any deleterious interactions introduced by other TTR substitutions cannot be suppressed by altering ␤-Glu 44 .
Mu Induction and pSC101 Maintenance-Mutants with amino acid substitutions at residues near ␤-Glu 44 such as ␤R42A, ␤R46C, and ␤R46H exhibit a significant reduction in IHF function as measured by the plating efficiency of bacteriophage Mu (17,18). The ␤-Glu 44 mutants constructed in this study were tested for their ability to promote Mu induction and maintenance of the plasmid pSC101 to determine if IHF function was disrupted. Both Mu induction and pSC101 mainte-

Ͼ2750
This study HЈ43C This study HЈ43G This study HЈ44A This study HЈ44G This study HЈ45A This study HЈ45C This study HЈ45T This study nance require IHF binding to unique DNA sites, the ihfB site in Mu and pSC101 ori, respectively (Table III). These two sites have an A at the third position of the TTR element that differs from the G found in the HЈ site and also differ from each other in the sequence of the bases surrounding their TTR elements.
In addition, the pSC101 ori and ihfB sites differ from the HЈ site and each other at the permissively variable bases within the WATCAR element as well as in the composition of the poly(dAT) element and the N8 spacer. We found that the A, D, G, N, and Q substitution mutants functioned as efficiently as wild type for Mu induction and pSC101 maintenance. These results demonstrate that the substitutions we tested do not alter IHF structure enough to result in a mutant phenotype by these assays and suggest that they do not dramatically alter the conformation or stability of the proteins (Table III). They also suggest that the substitutions do not affect the ability of IHF to recognize the TTR element in different IHF binding sites.

DISCUSSION
In this study, we constructed site-directed mutants to examine the role of ␤-Glu 44 in IHF binding to the TTR element in one of its specific DNA sites. The behavior of these mutants indicates that the wild-type ␤-Glu 44 bestows specificity in the IHF-HЈ site complex by buttressing ␤-Arg 42 and ␤-Arg 46 . In light of the IHF-HЈ site cocrystal, we attribute these effects primarily to the hydrogen bonds to ␤-Arg 46 and ␤-Arg 42 made by ␤-Glu 44 . In the absence of either interaction, the ␤1-␤2 turn portion of the protein-clamp collapses and ␤-Arg 46 gains the conformational freedom to adapt to the HЈ44A mutant site but not to other TTR mutant derivatives. The details of our proposal are presented below.
In the IHF-HЈ site cocrystal (2), ␤-Glu 44 is inserted into the minor groove of the TTR element but does not contact the DNA (Fig. 2a). The carboxyl group of ␤-Glu 44 interacts with ␤-Arg 42 and ␤-Arg 46 . One of the oxygens of ␤-Glu 44 interacts with the charged ureido amine groups of ␤-Arg 42 , and the ␤-Arg 42 makes a salt bridge to the phosphate group of Ala 41 (not shown). The other oxygen of ␤-Glu 44 makes hydrogen bonds to both ureido amine groups of ␤-Arg 46 . In turn, ␤-Arg 46 makes hydrogen bonds to the O 2 of Thr 44 through both its terminal ureido amine groups. The aliphatic portion of ␤-Arg 46 also makes van der Waals contacts with the deoxyribose of Gly 45 (not shown). This network of bonds between ␤-Arg 42 , ␤-Glu 44 , ␤-Arg 46 , and the minor groove of the TTR element form the center of a protein clamp that holds the HЈ site in its bent conformation. The periphery of this protein clamp is composed of amide nitrogen contacts from the ␤1-␤2 turn and the N terminus of ␣-helices 1 and 3 to the phosphates of the TTR element. Rice et al. (2) suggested that the ␤E44G, ␤E44V, and ␤E44K suppressors allow the binding of the HЈ44A(II) site by conferring increased conformational freedom to the side chain ␤-Arg 46 so that it does not disrupt the interaction with the TTR element.
The ability of IHF to discriminate against binding to the HЈ44A site TTR element appears to depend upon the bonds ␤-Glu 44 makes with both ␤-Arg 42 and ␤-Arg 46 . Of the mutants  tested, only IHF molecules with glutamic acid or aspartic acid at position ␤44 are unable to repress P22-HЈ44A(II) (Fig. 1b). This finding supports the initial supposition that the removal of the negative charge allows IHF to bind the mutant HЈ44A site. The reduced lysogeny of P22-HЈ(II), when tested with the aspartic acid substitution mutant, indicates that disruptive interactions have been introduced (see below), and the negative charge must be precisely placed to allow optimal IHF function. Computer modeling indicates that the aspartic acid continues to make a hydrogen bond to the secondary amino group of ␤-Arg 46 (Fig. 2b). Binding affinity to wild type is lowered but not abolished, because interactions between ␤-Arg 46 and ␤-Asp 44 force a clash between ␤-Arg 46 and the ribose of Gly 45 or Thr 44 . Additional support for the hypothesis derives from the intermediate levels of lysogeny promoted by the polar amino acid substitution mutations in the ␤E44N and ␤E44Q proteins (Fig. 1b). Neither the ␤E44N nor ␤E44Q residues are predicted to make hydrogen bonds with ␤-Arg 42 , because the oxygen acceptor has been replaced with an amide donor (Fig. 2c). Yet the ␤E44N and ␤E44Q mutants do promote lysogeny with P22-HЈ44A(II) but at levels well beneath that of the ␤E44A mutant with P22-HЈ44A(II). This result suggests that the hydrogen bonds from the ␤44 residue to ␤-Arg 42 control part of the specificity for binding to the TTR element. Indeed, in mutants containing substitutions of ␤-Glu 44 such as alanine, glycine, valine, or lysine, these hydrogen bonds cannot form because these amino acids do not have oxygens in their side chains. Lack of the interactions would confer conformational freedom on both the ␤-Arg 46 and ␤-Arg 42 residues to bind the HЈ44A site without steric clashes. This assertion is supported by the fact that P22-HЈ44A(II) forms lysogens at levels comparable with those seen with the P22-HЈ(II) phage. Based on the behavior of the substitution mutants and the cocrystal structure, we propose the following model. The ␤-Glu 44 residue imposes specificity on the binding at the central base pair of the TTR element through hydrogen bonds from an oxygen that holds ␤-Arg 46 in a position that prevents it from adapting to the disruptive alterations introduced in the HЈ44A site. These hydrogen bonds between the oxygen of ␤-Glu 44 and the ureido amines of ␤-Arg 46 are further stabilized by an additional bond between the oxygen of ␤-Glu 44 and an amino group of ␤-Arg 42 and as a result limit the conformational freedom of ␤-Glu 44 . In turn, ␤-Arg 42 is held in place by an electrostatic bond with the phosphate group of the adenine at position base pair 41 of the HЈ site (not shown). Thus, all three amino acids are necessary for the achievement of full discrimination against the P22-HЈ44A(II) variant.
It remains unclear which portion of the HЈ44A site clashes with ␤-Arg 46 in wild-type IHF. One possibility is that the hydrogen bonds from both of the ureido amine groups of ␤-Arg 46 to the O 2 of the thymine at base pair 44 of the HЈ sites are disrupted by steric clashes with the N3 of adenine in the HЈ44A site. The transversion from T:A to A:T found in the HЈ44A site should formally place the N3 of adenine 44 in the position previously occupied by O 2 of the thymine in the HЈ site. The N3 of adenine at base pair 44 in the HЈ44A site might accept hydrogen bonds from ␤-Arg 46 even if the roll, twist, or propeller twist angles were exaggerated to their possible extremes. It seems likely that any changes in the roll, twist, or propeller twist caused by the substitution within the TTR element would also result in a distortion of the DNA backbone. Such changes could disrupt stable binding by IHF by causing the van der Waals interaction between ␤-Arg 46 and the deoxyribose at guanine at base pair 45 seen in the cocrystal to clash. The removal of the hydrogen bonds bracing ␤-Arg 46 could allow it to move toward the center of the minor groove to alleviate the steric hindrance.
In summary, the wild-type ␤E44 residue confers binding specificity "obliquely" (indirectly) by forming hydrogen bonds to ␤-Arg 42 and ␤-Arg 46 . These results show how interactions not directly at the protein-DNA interface can confer sequence specificity by affecting the positioning of amino acids that themselves contact the nucleic acid. Such oblique but specific effects suggest the difficulties inherent in attempting to engineer sequence specificity changes in proteins that interact with DNA. The genetic methods employed here allow specific predictions concerning the detailed chemical interactions of amino acid side changes critical to site-specific interactions.